automated methods in chemical risk assessment€¦ · value. chemical properties relevant to hazard...

Automated Methods in Chemical Risk Assessment

by

Trevor N. Brown

A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy

Chemistry University of Toronto

© Copyright by Trevor N. Brown 2011

ii

Automated Methods in Chemical Risk Assessment

Trevor N. Brown

Doctor of Philosophy

Chemistry

University of Toronto

2011

Abstract

In this thesis, topics in chemical hazard and risk assessment are explored through the use of

multimedia mass balance models and high-throughput chemical property prediction methods.

Chemical hazard metrics, as calculated by environmental fate and transport simulations, are

investigated to determine the validity of two common simplifying assumptions in the underlying

models; the use of octanol as a surrogate for organic matter and the use of environmental

parameters that do not vary in time. A major finding is that the use of these common simplifying

assumptions in multimedia mass balance models has little effect on chemical risk assessment,

provided that chemicals are ranked relative to each other rather than a predetermined cutoff

value. Chemical properties relevant to hazard and risk are collected, and applied in a large-scale

chemical hazard assessment to derive a short list of potential Arctic contaminants. Several

further data needs are identified; these are widely applicable and easily calculable metrics for

chemical biodegradation, toxicity and emissions. A new method of predicting chemical

properties is presented to assist in meeting these data needs. The method automatically creates

predictive, quantitative relationships between the structures and properties of chemicals that are

comparable to similar relationships created with expert judgement.

iii

Acknowledgments

First and foremost I would like to thank my supervisor Frank Wania; I am grateful for his

unfailing support and patience. I will always appreciate the excitement and interest he showed in

the projects we worked on together, and how he always made time to discuss ideas about

environmental chemistry, european travels and many other topics. I would also like to thank my

committee members Jamie Donaldson and Derek Muir for their support and interest over the

years, as well as the additional members of my defense committee George Arhonditsis, and my

external examiner Matt MacLeod for being interested and taking the time participate.

I would like to acknowledge the two CEFIC-LRI grants that helped to fund much of my Ph.D.

work, and thank the members of the RLT for their interest and understanding in the projects.

Special thanks especially to the chairmen and regular attendees at the annual project meetings

Andrew Riddle and Dolf van Wijk.

Thanks to my collaborators and coauthors for their insight, helpful comments and hard work.

Michael McLachlan, Knut Breivik, Kai-Uwe Goss, Gerjte Czub and Emma Undeman have my

gratitude for all their help and for making the annual meetings during the IMPS project so

enjoyable, with the addition of Jon Arnot and Xianming Zhang in the Screen-POP project.

I am privileged to have two of the most loving and supportive parents in the world. I will never

be able to thank them enough for their unwavering support through many, many years of school

with all of the related false starts, wrong turns and long distance moves. I am grateful for the

support of both my brothers, my sister inlaw, and far too many other relatives to mention here.

On the topic of family I‘d also like to thank the other mother figures in my life while in Toronto;

Ying Lei and Olga Petrovski for checking to make sure that I was happy, healthy and well-fed.

Over the last five years I have met so many great people I would be forced to add another chapter

here to mention them all by name. Thank you to both the UTSC and downtown chemistry grad

students I‘ve known over the years, especially members of the Wania Group, past, and present.

Thank you to my climbing partners who have been great friends and introduced me to a lifelong

passion, especially the original partners-in-climb Johnny, Erin and Magda.

Finally, thank you to Rachel Chang for greatly enriching my life and for being so incredibly

supportive during the writing of this thesis.

iv

Table of Contents

Acknowledgments .......................................................................................................................... iii

Table of Contents ........................................................................................................................... iv

List of Tables ............................................................................................................................... viii

List of Figures ................................................................................................................................ xi

Chapter 1 Introduction .................................................................................................................... 1

1 Chemical Hazard and Risk Assessment ..................................................................................... 2

2 Environmental, Bioaccumulation and Exposure Modelling ...................................................... 6

3 Automated Chemical Screening and Prioritization .................................................................. 10

4 Objectives and Outline ............................................................................................................. 12

Chapter 2 Screening Chemicals for the Potential to be Persistent Organic Pollutants: A Case

Study of Arctic Contaminants .................................................................................................. 14

1 Introduction .............................................................................................................................. 15

2 Methods .................................................................................................................................... 16

2.1 Outline of the Approach .................................................................................................... 16

2.2 Compilation of Data .......................................................................................................... 17

2.3 Screening for Long Range Transport to the Arctic and Bioaccumulation in Humans ..... 19

2.4 Screening for Persistence in Air ....................................................................................... 22

2.5 Screening for Resemblance with the Structural Profile of Known Arctic Contaminants . 22

3 Results and Discussion ............................................................................................................. 23

3.1 Preliminary Screening ....................................................................................................... 23

3.2 Chemicals Matching the Structural Profile of Known Arctic Contaminants ................... 25

3.3 Potential Arctic Contaminants .......................................................................................... 27

3.4 High Risk Chemicals ........................................................................................................ 28

4 Acknowledgements .................................................................................................................. 34

5 Appendix .................................................................................................................................. 34

5.1 Compiling a List of Known Arctic Contaminants ............................................................ 34

5.2 Profiling Arctic Contaminants .......................................................................................... 40

5.3 Chemical Profiling Criteria ............................................................................................... 41

5.4 Application of Chemical Profiles ...................................................................................... 42

5.5 Interpretation of Figure 9 .................................................................................................. 47

5.6 Definition of the POP score .............................................................................................. 49

5.7 Interpretation of the POP Score ........................................................................................ 50

v

Chapter 3 Development and Exploration of an Organic Contaminant Fate Model Using Poly-

Parameter Linear Free Energy Relationships ........................................................................... 85

1 Introduction .............................................................................................................................. 86

2 Methods .................................................................................................................................... 88

2.1 PP-LFER Equations .......................................................................................................... 89

2.2 Implementation of PP-LFERs in CoZMo-POP2. ............................................................. 90

2.3 Solute Descriptors. ............................................................................................................ 91

2.4 Model Parameters. ............................................................................................................ 93

2.5 Model Comparison. ........................................................................................................... 94

2.6 Linking Solute Descriptors to Chemical Fate. .................................................................. 95

3 Results and Discussion ............................................................................................................. 96

3.1 Comparison of KOC Values. .............................................................................................. 96

3.2 Comparison of KPA Values. .............................................................................................. 97

3.3 Comparison of Model Results. ......................................................................................... 98

3.4 Linking Solute Descriptors to Chemical Fate. ................................................................ 102

4 Acknowledgements ................................................................................................................ 105

5 Appendix ................................................................................................................................ 106

5.1 Derivation of Equations for Enthalpies of Phase Change ............................................... 106

5.1.1 Derivation of PP-LFER Equations for ΔHAW, ΔHOA and ΔHOW ........................ 106

5.1.2 Derivation of PP-LFER Equations for ΔHHA and ΔHHW .................................... 107

5.1.3 Derivation of a PP-LFER Equation for ΔHPA ..................................................... 107

5.2 Calculation of Z-values in CoZMo-POP 2 ..................................................................... 109

5.3 PP-LFER for Sorption to Forest Canopy ........................................................................ 111

5.4 PP-LFER Equation for Hexadecane/Air Partitioning ..................................................... 113

5.5 Dataset Reduction Method .............................................................................................. 114

5.6 Description of Model and Parameters ............................................................................. 124

5.7 Interpolation Method ...................................................................................................... 129

5.8 Validity of Showing Results in the Chemical Space ...................................................... 134

Chapter 4 Investigating the Time Resolution of Water and Organic Carbon Balances and

their Effect on Contaminant Fate Modeling ........................................................................... 147

1 Introduction ............................................................................................................................ 148

2 Methods .................................................................................................................................. 150

3 Results .................................................................................................................................... 156

4 Discussion and Conclusions ................................................................................................... 164

vi


6 Appendix ................................................................................................................................ 169

6.1 Variable Mass Balances of Air, Organic Aerosol, Water and Particulate Organic

Carbon. ............................................................................................................................ 169

6.1.1 General Description. ........................................................................................... 169

6.1.2 Air Balance. ........................................................................................................ 170

6.1.3 Organic aerosol balance. ..................................................................................... 171

6.1.4 Water balance. ..................................................................................................... 172

6.1.5 POC balance. ....................................................................................................... 176

6.2 Meteorological Data and Climatic Scenario Definition. ................................................. 181

6.2.1 Data source and selection. ................................................................................... 181

6.2.2 Heat Transfer Model. .......................................................................................... 181

6.2.3 Insolation Model. ................................................................................................ 182

6.2.4 Precipitation Model. ............................................................................................ 184

6.2.5 Primary Productivity Model. .............................................................................. 185

6.2.6 Atmospheric OH Radical Concentration Model. ................................................ 186

6.3 Derivation of Empirical Relationship for Predicting Temperature Dependence. ........... 191

6.4 Definition of Environmental Half-lives. ......................................................................... 194

Chapter 5 Creating a Fragment-Based QSAR with Iterative Fragment Selection (IFS) ............ 199

1 Introduction ............................................................................................................................ 200

2 Methods .................................................................................................................................. 201

2.1 Data Processing ............................................................................................................... 201

2.2 Overview of the Method ................................................................................................. 202

2.3 Generation of Descriptors ............................................................................................... 202

2.4 Model Selection .............................................................................................................. 204

2.5 Goodness of Fit Metric ................................................................................................... 205

2.6 Fragment Selection Rules ............................................................................................... 206

2.7 Cross Validation .............................................................................................................. 207

2.8 External Validation ......................................................................................................... 210

2.9 Experimental Datasets .................................................................................................... 211

3 Results .................................................................................................................................... 212

3.1 Fitting and External Validation ....................................................................................... 212

3.2 Fragment Coefficients ..................................................................................................... 217

3.3 Fragment Selection ......................................................................................................... 218

vii

4 Discussion and Conclusion .................................................................................................... 219

4.1 Model Averaging and Model Uncertainty ...................................................................... 219

4.2 Comparison to Arnot et al. QSAR .................................................................................. 225

4.3 Fragment Analysis .......................................................................................................... 227


6 Co-Author Contributions ....................................................................................................... 231

7 Appendix ................................................................................................................................ 232

Chapter 6 Conclusions ................................................................................................................ 264

1 Summary of Major Conclusions ............................................................................................ 265

2 Recommendations and Future Directions .............................................................................. 267

References ................................................................................................................................... 269

viii

List of Tables

Table 1 - Summary of highproduction volume chemicals which are predicted to become Arctic

contaminants or which match the structural profile of known Arctic contaminants. ................... 30

Table 2 - Names, CAS numbers, and the overall POP score for a list of known Arctic

contaminants ................................................................................................................................. 36

Table 3 - CAS number, chemical class, name and POP score of chemicals meeting the elevated

AC-BAP, atmospheric oxidation half-life, and POP score criteria. ............................................. 52

Table 4 - CAS number, chemical class, name and POP score of chemicals meeting the elevated

AC-BAP and atmospheric oxidation half-life criteria, but not meeting the POP score criteria. .. 65

Table 5 - Explanation of the Acronyms for Chemical Class Used in Tables S2 and S3 .............. 83

Table 6 - Chemical classifications of the chemicals which meet the elevated AC-BAP and

atmospheric oxidation half-life criteria. ........................................................................................ 84

Table 7 - Phase Descriptors for PP-LFER Equations Used in this Study. .................................... 90

Table 8 - Real Chemicals Dataset ............................................................................................... 115

Table 9 - Hypothetical Chemicals Dataset .................................................................................. 122

Table 10 - Solute Descriptor Inter-Correlations (R2): Real Chemicals ...................................... 124

Table 11 - Solute Descriptor Inter-Correlations (R2): Hypothetical Chemicals ......................... 124

Table 12 - Summary of Model Parameters ................................................................................. 126

Table 13 - Statistical Ratios of Interpolated Solute Descriptors ................................................. 138

Table 14 - Sub-sampling Statistics ............................................................................................. 146

Table 15 - Mass Balance Compartments and Fluxes: their Interdependence and Fitting. ......... 179

Table 16 - Relative Precipitation Intensities. .............................................................................. 188

Table 17 - Average, Minimum and Maximum Characteristic Travel Distances of Persistent and

Labile Chemicals at Hourly, Daily, Weekly, Monthly and Seasonal Environmental Parameter

Resolutions, in Kilometers. ......................................................................................................... 197

Table 18 - Summary of Statistics for Fitted and External Validation Datasets. ......................... 213

ix

Table 19 - 2D Fragment Based IFS-HLN QSAR Averaged from Three Cross-Validations. ...... 221

Table 20 - Chemicals with log HLN Residuals Greater than 1 in IFS-HLN and EPI HLN .......... 226

Table 21 - Qualitative Prediction Accuracy for Fish log HLN Fit on Dataset Splitting from this

Work Using Average Model from Three Different Cross-Validations. ..................................... 227

Table 22 - Comparison of Fragment Coefficients from IFS-HLN and EPI-HLN. ....................... 229

Table 23 - Training and Validation Datasets Summary. ............................................................. 232

Table 24 - 2D Fragment Based QSAR from IFS-HLN with a 1:2 Internal Training to Internal

Validation Ratio During Cross-Validation. ................................................................................ 245





Table 27 - 2D Fragment Based QSAR from IFS-EPI with a 1:2 Internal Training to Internal






Table 30 - 2D Fragment Based QSAR from IFS-yscr with a 1:2 Internal Training to Internal






Table 33 - 2D Fragment Based QSAR from IFS-KOW with a 1:2 Internal Training to Internal




x


Validation Ratio during Cross-Validation. ................................................................................. 262

xi

List of Figures

Figure 1. Schematic Diagram of the Factors Considered in Chemical Risk Assessment. .............. 3

Figure 2 – Chemical Space plot of elevated Arctic contaminant potential. .................................. 20

Figure 3 - Schematic diagram of the screening process. The 105,584 chemicals in the EPI Suite

database were screened for their potential to accumulate in the Arctic food chain. ..................... 24

Figure 4 - Locations of all chemicals in the EPISuite database which match the structural profile

of known Arctic contaminants. Plotted using the "HLC data set". ............................................... 26

Figure 5 - Plot of all 105,581 chemicals in the EPI Suite database. ............................................. 39

Figure 6 - Frquency distribution of the number of structural atoms (SA). distributions are shown

for five different chemical "populations"...................................................................................... 44

Figure 7 - Frequency distribution of the degree of halogenation (CX). distributions are shown for

five different chemical "populations" ........................................................................................... 45

Figure 8 - Frequency distribution of the degree of internal connectivity (XS). distributions are

shown for five different chemical "populations" .......................................................................... 46

Figure 9 - Mean and standard deviation of the total number of structural atoms (SA), degree of

halogenation (CX), and degree of internal connectivity (XS) for five different chemical

"populations". The deviation from zero is shown in orange ......................................................... 47

Figure 10 - Locations of the 1460 chemicals for which solute descriptors were obtained in the log

KOA/log KAW chemical space. Values for log KOA and log KAW are calculated using equation (9)

and the phase descriptors from Table 7. The reasonable domain of model applicability is shown

in red. ............................................................................................................................................ 92

Figure 11 - (A) Correlation between the log KOC values obtained by the SP-LFER equation and

values obtained from the PP-LFER obtained for humic acid. (B) Correlation between the logKPA

values obtained by the SP-LFER and values obtained from the PP-LFER obtained for whole

aerosol. .......................................................................................................................................... 97

xii

Figure 12 - Chemical space plots of phase distribution of the PP-LFER based model and the sum

of absolute differences in the percent phase distribution between the SP-LFER and PP-LFER

based models for each of the three emission scenarios. .............................................................. 100

Figure 13 - Vector plots of movement in the log KAW/log KOA chemical space that can be related

to the contributions of each phase descriptor. Vector magnitudes are the value of the upper

quartile minus the value of the lower quartile of the distribution of solute descriptors multiplied

by the corresponding phase descriptor. ....................................................................................... 103

Figure 14 - Linear Regressions of Plant Sorption with log KOA ................................................. 113

Figure 15 - Solute Descriptor Plots ............................................................................................. 137

Figure 16 - KOC and KPA Correlations of Hypothetical Chemicals ............................................. 138

Figure 17 - Götz et al. PP-LFER KPA versus SP-LFER KPA ....................................................... 139

Figure 18 - Aspvreten PP-LFER KPA versus SP-LFER KPA ....................................................... 139

Figure 19 - Chemical Space Plots for Hypothetical Chemicals .................................................. 140

Figure 20 - Relative Difference in Model Outputs for Air Emissions ........................................ 141

Figure 21 - Relative Difference in Model Outputs for Soil Emissions ....................................... 142

Figure 22 - Relative Difference in Model Outputs for Water Emissions ................................... 143

Figure 23 - Chemical Space Plots for Degrading Real Chemicals ............................................. 144

Figure 24 - Characteristic Travel Distance of Degrading Real Chemicals ................................. 145

Figure 25 - Schematic of the Major Pools and Fluxes of the Four Environmental Mass Balances

in CoZMoMAN-EV and their Interconnectivity. ....................................................................... 151

Figure 26 - Air temperature (A) and primary biological production rate in fresh water (B) at

hourly, daily, weekly, monthly, and seasonal resolutions. ......................................................... 154

Figure 27 - Percentage of Hourly Resolved Variability Captured with Daily, Weekly, Monthly

and Seasonal Resolved Data for Labile and Persistent Chemicals in all Model Compartments.

Each Cell is a Chemical Space Plot Showing the Fraction of Intra-Annual Variability Captured at

Each Point; Labile and Persistent Environmental Distributions are Provided as a Guide. ......... 160

xiii

Figure 28 - Hourly Characteristic Travel Distances (CTDs) in Kilometers for Labile (A-E) and

Persistent (F-J) Chemicals using Hourly (A&F), Daily (B&G), Weekly (C&H), Monthly (D&I),

and Seasonal (E&J) Resolved Environmental Parameters. ........................................................ 163

Figure 29 - Moles of Chemical in Air Simulated using Hourly Resolved Environmental

Parameters for Labile (A) and Persistent (B) Chemicals, Ranging from a log KOA of 4 in Red to

a log KOA of 10 in Blue, and a log KAW of -5 to 1. ................................................................. 164

Figure 30 - Hourly Resolved Air, Water and Soil Temperatures in Kelvin. .............................. 187

Figure 31 - Hourly Resolved Insolation and Corrected Insolation as Fraction of Maximum. ... 187

Figure 32 - Hourly Precipitation Amounts as a Fraction of Annual Precipitation. .................... 188

Figure 33 - Primary Biological Production Rate as a Fraction of Annual Production. .............. 189

Figure 34 - Time Averaging of Environmental Parameters Soil Temperature (A), Marine Water

Temperature (B), Marine Primary Biological Production Rate (C), Atmospheric OH Radical

Concentration (D), Rainfall (E) and Wind Speed (F). ................................................................ 190

Figure 35 - Correlation of Enthalpies of Phase Change in kJ/mole with the Logarithm of their

Corresponding Partition Coefficients. ........................................................................................ 193

Figure 36 - Annual Pattern of Variation in the Molar Amount of Chemical in Air, Cultivated

Soil, and Fresh Water for Hypothetical Labile Chemicals with the original CoZMoMAN Model

(Red Lines) and the New Model (Blue Lines) with Monthly Resolved Environmental Parameters.

..................................................................................................................................................... 195

Figure 37 - Annual Pattern of Variation in the Molar Amount of Chemical in Air, Cultivated

Soil, and Fresh Water for Hypothetical Persistent Chemicals with the original CoZMoMAN

Model (Red Lines) and the New Model (Blue Lines) with Monthly Resolved Environmental

Parameters. .................................................................................................................................. 196

Figure 38 - Extrapolation to Steady State Characteristic Travel Distance for Labile Chemicals

(A) and Persistent Chemicals (B). .............................................................................................. 198

Figure 39 - Model fit and predictions from IFS-HLN for (A,B) 1:2, (C,D) 1:1, and (E,F) 2:1 ratios

between internal training and internal validation datasets during cross-validation. ................... 214

Figure 40 - Model fit and predictions for the averaged IFS HLN QSAR for training (A) and

external validation (B) dataset. ................................................................................................... 224

xiv

Figure 41 - Model fit and predictions for fish HLN using IFS-EPI for (A,B) 1:2, (C,D) 1:1, and

(E,F) 2:1 ratios between internal training and internal validation datasets during cross-validation.

..................................................................................................................................................... 242

Figure 42 - Model fit and predictions for fish HLN using IFS-yscr for (A,B) 1:2, (C,D) 1:1, and


..................................................................................................................................................... 243

Figure 43 - Model fit and predictions for log KOW using IFS-KOW for (A,B) 1:2, (C,D) 1:1, and


..................................................................................................................................................... 244

1

Chapter 1

Introduction

2

1 Chemical Hazard and Risk Assessment

A large and ever-increasing number of chemicals are used in commerce; the chemical

abstracts service (CAS) reports more than 282 thousand inventoried or regulated substances,1

and the OECD reported that member countries produced more than 4600 chemicals in

quantities of 1000 tonnes or more in 2007.2 It is impossible to determine the exact number of

chemicals with significant commercial use by searching publicly available data, because

many countries do not gather data on chemical production, or report production to the

OECD. Some of these chemicals are likely to pose a human health risk, but chemical risk

assessment is a lengthy, data intensive process. Many factors must be considered when

determining if a chemical is a human health risk; human exposure to a chemical is a function

of the amount produced, the usage pattern, the route of exposure, and the properties of the

chemical.3 If the potential for human exposure exists this still only constitutes a hazard. In

order to translate into risk the chemical must additionally be toxic and have properties that

allow it to resist metabolism and accumulate to dangerous levels in human tissues.4

Furthermore, if a chemical does not pose a human health risk it may still be an environmental

risk, which poses the daunting challenge of assessing the risk for a myriad of different

organisms in different environments. At the same time there has never been more public

pressure on governments to protect their citizens and the environment from the adverse

effects of chemical pollution, and regulators have responded at the national and international

level with new chemical regulations and risk assessment programs.5-7

To ensure the success

of these programs regulators have called on academia and industry to assist them with new

sources of data and new tools to allow for rapid screening and prioritizing of chemicals that

may be a human health or environmental risk.

Figure 1 shows a simplified schematic of the factors that must be considered when

performing a chemical risk assessment for a single anthropogenic parent chemical acting on a

single target organism. The schematic approximately follows the course of a chemical as it

moves through the environment and into an ecosystem, and covers most of the fields that

make up environmental organic chemistry. In order to constitute a risk a chemical must make

its way from Production through the schematic to Adverse Effects. Assessing a chemical for

only a subsection of steps between Production and Adverse Effects is a hazard based

assessment rather than a risk based assessment. There are multiple pathways for a chemical

3

to follow on its way from Production to Adverse Effects, some short and others more

circuituous, but all of them move through Exposure, and end in Adverse Effects. These two

factors make up the accepted definition of risk and form the basis of various risk metrics,

where risk is a multiplicative function of Exposure and Adverse Effects measured or

predicted in comparable units.3

Figure 1. Schematic Diagram of the Factors Considered in Chemical Risk Assessment.

4

Chemicals which take the shortest path in Figure 1 from Production to Adverse Effects;

flowing through Usage and Emission directly to Exposure, and then directly from Exposure

to Adverse Effects, are of course the most regulated chemicals, i.e. acutely toxic chemicals

with the potential for direct consumer exposure.8 In the past any chemical with either one of

these attributes, a high potential for direct consumer or occupational exposure (―near-field‖

exposure) or acute toxicity, were more closely regulated than those chemicals where

exposure occurs indirectly through the environment (―far-field‖ exposure) or those chemicals

which bioaccumulate to a threshold level before adverse effects are observed.8 There have

been relatively recent changes in this regard though as international agreements such as the

Stockholm Convention,5 and national programs such as REACH in the European Union

6

have been brought into force seeking to regulate chemicals taking a more circuituous path

from Production to Adverse Effects. These regulatory efforts fall into two related categories

that seek to regulate POPs (Persistent Organic Pollutants) or PBTs (Persistent,

Bioaccumulative and Toxic chemicals). Both paradigms seek to regulate chemicals that are

persistent in the environment, bioaccumulate in target organisms and have adverse effects,

with the POPs oriented regulatory efforts further seeking to regulate chemicals susceptible to

long range transport.9,10

This combination of properties creates the potential for chemicals to

follow the more circuituous path in Figure 1 through Environmental Fate and Transport to

Exposure, and then through Bioaccumulation to Adverse Effects.

If a chemical is determined to be a risk then the aim of regulators is to cut the flow of

chemical off somewhere in the schematic of Figure 1 before it reaches Adverse Effects.

Regulatory action is, however, restricted to acting within the first three steps, Production,

Usage and Emission, and Environmental Fate and Transport by banning production, banning

or restricting further use and by requiring remediation of environmental contamination,

respectively.8 If the only route of exposure is direct exposure from usage then a usage ban is

effective in mitigating the risk a chemical poses. For chemicals with significant

environmental emissions however shutting the process down at the production or usage steps

may not eliminate the risk. If a chemical is persistent in the environment then there will be a

remaining pool of chemical causing continued exposure. Additionally, if a chemical is

bioaccumulative then target organisms will have a pool of the chemical stored in their bodies

5

that may continue to cause adverse effects even if regulation is effective in halting exposure.

Some classic examples of these complications are the legacy POPs targeted by the

Stockholm Convention such as PCBs (polychlorinated biphenyls).11

At every one of the stages included in Figure 1 the properties of the chemical that is being

assessed for risk play a role. Chemicals are produced and used because they have properties

that are useful. Some examples are PCBs, produced and used because of their high resistance

to heat and stress,12

perfluorinated surfactants and polymers, produced for their non-stick and

stain-repellent properties,13

and chlorinated pesticides, of course, were produced and used

specifically because they were toxic to certain organisms, and persistence was desirable

because it meant that they were effective for longer. As the previous examples demonstrate,

it is sometimes the very properties that make a chemical useful that also make the chemical a

risk. Chemical Emissions, and Environmental Fate and Transport are dependent on the

interaction between chemical properties, the usage of the chemical, and the properties of the

environment. How a chemical distributes itself among environmentally relevant phases and

resists degradation by physical processes and native microbes are important factors in

determining how the chemical enters the environment, where it goes and how long it stays

there.14,15

Exposure, Bioaccumulation and Adverse Effects are also functions of chemical

properties, and of the characteristics of the target organism. Chemical properties and the

presence of specific functional groups play a role in determining the absorption efficiency of

chemicals through different exposure routes, resistance to metabolism and toxicity.4

The direct pathways in Figure 1; direct exposure to chemical from usage and acute toxicity,

are only loosely linked to chemical properties or are very difficult to predict from known

chemical properties. Direct exposure can instead be estimated from industry and usage data,

assuming that these data exist and are accessible,16

and while it is difficult to predict which

organisms will experience acute toxicity from which chemicals the effect is often easily

observable and can be attributed to a specific chemical if the route of exposure is

determined.4 However, the pathway followed by POPs and PBT substances through Figure 1

is intrinsically linked to their properties; namely resistance to environmental degradation and

6

metabolism and the correct combination of partitioning properties that allow for

environmental mobility, high exposure potential and bioaccumulation.17

If a chemical has

properties consistent with those of other POPs and PBTs it may be considered a candidate for

POP or PBT related regulation.18

Risk assessment is, however, rarely an easy task because Figure 1, while a useful

demonstrative tool, is a gross oversimplification of all the factors involved in chemical risk

assessment. Many of the difficulties in chemical risk assessment are due to large possible

variations in chemical usage, emissions, environmental effects,19

and the behaviour and

physiology of the target organism.20

Additionally, many factors are not so clearly delineated

as Figure 1 suggests; for example in many cases environmental degradation or metabolism in

target organisms can transform a chemical into secondary products that may be more or less

harmful than the parent chemical, breaking the chain between Production and Adverse

Effects.21,22

Bioaccumulation comprises far more than chemical uptake in a single organism;

to accurately determine the chemical concentration in a single organism its food chain and

other effects such as maternal transfer must be considered, so bioaccumulation involves

really a series of organisms being exposed to a chemical.23

By far the largest difficulty in

chemical risk assessment though is simply gathering enough information to make an

informed decision. Experimental measurement of chemicals in the environment and in target

organisms has become routine and sensitive, but gathering experimental data at every point

in Figure 1 is a lengthy expensive process, especially if large lists of chemicals need to be

assessed for risk. For this reason many regulatory efforts have turned to the use of

environmental, bioaccumulation and exposure models to assist in predicting if chemicals

constitute a risk.3,24-26

2 Environmental, Bioaccumulation and Exposure Modelling

Environmental chemical fate and transport models are tools used to understand, explain and

predict the movement of chemicals through the environment. Foodweb bioaccumulation

models are similarly applied to study the accumulation of chemicals in biota, and together

these computer models are used to assess the potential exposure of humans, and other

7

organisms, to chemical contaminants.3,24

Environmental models explicitly consider the major

storage reservoirs of chemicals, also referred to as sinks or pools, and movements, referred to

as transport, or dispersion. Long term storage reserviors of organic chemicals may be soils,

sediments or deep waters, and typical movements are by diffusion, or in moving air and

water which act as carriers.14

Bioaccumulation models also consider storage reservoirs and

movements of chemical within an organism; with fatty tissues generally acting as storage

reservoirs and blood circulation, respiration, ingestion and egestion being typical movements

of chemical within an organism.23

Models are attractive tools in chemical risk assessement

because they help to fill in knowledge gaps about the environmental transport, fate and

bioaccumulation of chemicals using information about the chemical‘s properties and

parameters describing the relevant storage reservoirs and movements in the environment and

in organisms of interest, without the need for direct measurements of chemical

concentrations in the environment and biota.

Some of the most useful tools in assessing environmental exposure potential are multimedia

mass balance models which simulate the transport and fate of chemicals in the physical

(abiotic) environment.24,27

Multimedia models divide the environment into compartments,

with typical compartments including the atmosphere, aquatic environments, sediments, soil

and possibly plant matter. Each compartment is composed of one or more subcompartments

made up of pure phases such as air, water, mineral matter and organic matter. A compartment

is essentially a unit of the world which is assumed to be homogenous in both its physical

properties and the distribution of chemical within it, meaning that chemical equilibrium is

assumed within compartments. Compartments vary in size depending on the spatial

resolution of the environment being modelled. Although equilibrium is assumed within a

compartment this may not be the case in different parts of the environment depending on the

complexity of the model. Using the nomenclature of Mackay,15

environmental models may

be Level I, II, III, or IV models, with Level I models assuming equilibrium in a closed

system with no variance in time; and at the other end of the spectrum are Level IV models

that employ a non-equilibrium, non-steady state, time-variant description of the chemical

mass balance. Chemicals move between compartments either by diffusion or in a carrier

phase, and these movements, along with emissions and degradation are called fluxes. The

8

mass balance of chemical in a multimedia model essentially keeps track of the amount of

chemical in each compartment of the model and adjusts the amounts by summing the fluxes

which add or remove chemical in each of them.

Many mass balance models use fugacity as an equilibrium criterion to descibe how chemicals

partition between different parts of the environment and are often referred to as Mackay

models, after the author of the first manuscript that described their definition and use.28

Fugacity is somtimes referred to as the ―escaping tendency‖, because chemicals tend to have

a net movement from areas of high fugacity to areas of low fugacity, and is linearly related to

the concentration as shown in equation (1), where C is the molar concentration, f is the

fugacity in Pa, and Z is the fugacity capacity in mol m-3

Pa-1

.

C = f · Z (1)

When a chemical is at equilibrium between two locations its fugacity will be equivalent, so

the assumption of equilibrium within compartments in mass balance models is an assumption

of equi-fugacity, and fugacities between compartments will be different if the compartments

are not at equilibrium. Using fugacity as an equilibrium criterion simplifies the formulation

of the mass balance equations by removing the need for a partitioning coefficient between

each pair of compartments. The fugacity capacity, or Z value, essentially represents one half

of a partition coefficient, so the mass balance for each compartment can be calculated

separately with its fugacity and Z value instead of needing to simultaneously calculate the

equilibrium condition between every pair of comparments.15

Chemical fluxes are calculated

as shown in equation (2), where N is the flux in mol h-1

, and D is in units of mol Pa-1

h-1

.

N = f · D (2)

The calculation of the D values varies depending on the type of flux, diffusive exchange,

degradation or advection, but are all directly additive in the mass balance equation if they

occur in parallel and apply to phases having equivalent fugacities. In each case the relevant

fugacity is used to calculate the flux in moles per hour; the fugacity of the chemical in the

compartment in the case of degradation and the fugacity of the chemical in the source

9

compartment in the case of transport fluxes.15

Mass balances are constructed for each

compartment using the fluxes, concentrations, and compartment volumes.

Mass balance models of bioaccumulation and food webs work in much the same way as they

do for the abiotic environment.29

Each organism may be treated as a compartment, with sub-

compartments composed of water, lipids and important tissues in equilibrium. A more

sophisticated model may treat each tissue as a compartment on its own and calculate fluxes

between the tissues via blood flow. An important distinction however is that instead of

chemical emissions, as in the environmental mass balance models, biota experience chemical

exposure. Exposure is through three main routes; respiration (either air or water), ingestion of

water and food (other biota in the case of a food web), and by dermal contact.30

Fluxes of

chemical losses from biota are through respiration, egestion and metabolic biotransformation.

The concentration of chemical in biota may also be reduced by growth dilution, where the

amount of chemical remains constant but the organism grows to a larger size.29

Human

chemical exposure and bioaccumulation modelling is especially of interest for the purposes

of chemical risk assessment, and a number of models exist to predict the extent of chemical

exposure from the abiotic environment and the human food chain.23,31-33

Recent research

efforts have worked towards linking abiotic and human exposure mass balance models.

These linked models allow a chemical to be traced from the point of emission, through the

environment to exposure and simulate the bioaccumulation in the food chain and

humans,25,31,34

essentially covering most of Figure 1 in a single model.

There is one primary difficulty with using mass balance models for chemical risk assessment;

not surprisingly this difficulty is a lack of information. If a chemical is being assessed for risk

it is likely to be a relatively new chemical, although this is not always the case, and therefore

the chemical property data required to assess the chemical for every step in Figure 1 is

unlikely to be available or will be uncertain. Parameters describing the influence of the

physical environment and bioaccumulation may also be unavailable or uncertain. Many

simplifying assumptions have been made in both the abiotic and bioaccumulation models

with regards to the sorption capacity of various phases and the variability, both spatially and

10

temporally, in many parameters.14,15

These assumptions simplify the data requirements and

avoid unnecessary model complexity, but the effect of these simplifications on their

effectiveness as risk assessment tools is not always known.35

3 Automated Chemical Screening and Prioritization

Environmental, exposure and bioaccumulation models are used to calculate a number of

chemical hazard metrics, which along with data on adverse effects contibute to chemical risk

assessments. Hazard metrics calculated from the abiotic environmental models fall under two

main categories; long range transport metrics,36,37

and environmental persistence metrics. 37,38

Long range transport metrics give a measure of the exposure potential in locations that are

remote from the point of emission, which is important for the international regulation of

chemicals such as that called for by the Stockholm Convention.5 Environmental persistence

metrics measure how long a chemical will remain in the environment; a long environmental

residence time means that exposure will continue long after a chemical‘s use has been

discontinued due to storage in environmental reservoirs.11

Bioaccumulation and food web

models can be used to calculate a chemical‘s bioaccumulation potential from different routes

of exposure and with time variable dosing regimes. Knowing the peak and average internal

chemical concentrations (doses) is a useful piece of information when judging the possibility

of adverse effects.39

Linked environmental and bioaccumulation models can be used to

calculate overall exposure metrics such as intake fractions and human body burdens that

directly relate the amount of chemical in an organism to the amount of emissions released

into the environment.26,31,34,40

Chemical risk assessment is a difficult, data-intensive process that is expensive both in terms

of time and in terms of the tests which must be performed for proper assessment. Because of

the large number of chemical that are used in commerce it is virtually impossible to perform

a detailed risk assessment on all of them. This has led to the development of chemical hazard

screening and prioritization methods which are used to reduce the length of chemical lists by

automatically filtering out chemicals which are likely to be a low risk based on their hazard

metrics. A number of tools exist which can perform automated hazard metric screening on

11

chemicals given a set of chemical properties or a chemical structure.41,42

A common practice

is to set cut-off values for specific chemical properties or hazard metrics and focus only on

chemicals that meet these criteria; a good example is the logarithm of the octanol-water

partition coefficient (log KOW) being screened for values greater than five as a surrogate for

bioaccumulation potential.43

Chemicals may also be ranked by their properties or hazard

metrics, with the top-priority chemicals being those with the hazard metrics most likely to

make them the highest risk to human or environmental health.3,26

These methods are not

without problems however, because if data inputs for chemical properties are unavailable it

may not be possible to prioritize some chemicals, and if the properties are available but

uncertain there may be false negatives and false positives for chemical screening methods.

There are many chemical property prediction methods available to deal with the problem of

missing properties when performing a chemical risk assessment. Paritioning of chemicals

between environmentally relevant phases can be predicted by a large number of methods

with a wide range of sophistication. Sorption of chemicals to organic matter in the

environment and biota can be predicted with single-parameter regressions with partitioning

into n-octanol,44

with multi-parameter empirical equations,45

or with quantum chemical

calculations.46

Paritioning of a chemical between its pure form, octanol, water and air also

have a variety of prediction methods available, some also quite sophisticated such as

qunatum chemical predictions,47

and others quite simplistic such as quantitative structure

property relationships (QSPRs) that predict chemical properties based only on the 2D

structure of a chemical.41

Property prediction methods also exist for degradation in the

environment,48

biotransformation in fish,49

and adverse effects,50

but these methods are

generally more simplistic and more uncertain due to the chemical-specific nature of these

properties and the difficulty in measuring them experimentally. The prediction methods

mentioned here have varying degrees of uncertain. Unfortunately, the usual trade-off for

chemical property prediction methods is that the faster and easier to use methods are by far

the least accurate. Two very important parameters related to the chemical cannot be predicted

from chemical structure at all, these are the production volume and the usage profile of the

chemical. These data must be gathered from the few publically available databases, solicited

12

from industry, or estimated; if none of these options are possible then the exact level of

exposure cannot be known, and only a hazard-based assessment is possible.

4 Objectives and Outline

The overall objective of this thesis is to explore the feasibility of automated calculation of

hazard metrics for chemicals that have little existing data available, and to explore the

uncertainties involved in the models used to calculate these hazard metrics. Chapter 2 of this

thesis is a chemical screening exercise which involves a preliminary chemical hazard

assessment of a large list of chemicals, evaluating them for the potential to be contaminants

that fit into the POP paradigm and that may be found in the Arctic environment. Important

issues related to chemical risk that are discussed in Chapter 2 are the lack of publicly

availabile chemical production values, the uncertainties associated with using simplistic

chemical property predictions, chemical hazard screening, and the hazardous attributes of

POPs.

Chapter 3 and Chapter 4 explore the uncertainties associated with the application of

multimedia mass balance models due to the description of environmental phase distribution,

and the description of environmental temporal variability, respectively. Essentially, there

exist more sophisticated methods for describing environmental distribution and

environmental variability than are commonly used in multimedia models. Incorporating these

more sophisticated methods may improve the accuracy of the model outputs, but there is a

corresponding increase in the complexity of the model and in the data required to run the

model. Chapter 3 and Chapter 4 examine if this trade-off is worth making in the context of

automated chemical risk and hazard assessment.

As discussed throughout the Introduction, one of the greatest challenges when performing a

chemical risk or hazard assessment is often the lack of information regarding a chemical, its

uses, how it behaves in the environment and in biota. Chapter 5 of this thesis is a first step

towards providing another chemical property prediction tool intended to provide information

13

required for high-throughput chemical hazard and risk assessments. This tool is first applied

to the problem of predicting rates of metabolic transformation of chemicals, because this has

been shown to have a large potential impact on the human health risk of some chemicals.

14

Chapter 2

Screening Chemicals for the Potential to be Persistent Organic

Pollutants: A Case Study of Arctic Contaminants

Trevor N. Brown, Frank Wania

Environmental Science & Technology 2008 42 (14), 5202-5209

Reproduced with permission from Environmental Science and Technology

Copyright 2008 American Chemical Society

15

1 Introduction Considering the large number of chemicals in commerce, the complication of varied and

unknown chemical degradation pathways in the environment, and the highly selective nature

of most analytical detection systems, it is conceivable, if not highly likely, that the majority

of contaminants present in the environment, in wildlife and in humans, remains unidentified.

Even though there are more than 82,000 substances covered by the US Toxic Substances

Control Act (TSCA)51

, the US Centers for Disease Control monitor less than 120 substances

in human blood52

. Contaminants whose environmental occurrence has not previously been

documented are neither detected, studied, monitored, nor evaluated, let alone regulated. For

example, even though synthesized and used for more than half a century, the ubiquitous

presence of perfluorinated alkyl-compounds in people, wildlife and the physical environment

was not recognized until a decade ago. Now these substances are of the highest priority to

chemical regulators around the world 53

. Discovery of previously undetected environmental

contaminants is rarely the result of a directed search, but often owes much to serendipity.

Clearly, there is a need for a more rational approach with a greater promise to succeed in

identifying undetected contaminants.

Based on a quantitative understanding of the relationship between chemical properties and

environmental behavior, it is now becoming possible to identify the properties of substances

that qualify as contaminants of concern. For example, we now can predict what chemical

characteristics make an organic substance susceptible to accumulation in the Arctic physical

environment54

. Similar approaches seek to constrain the properties of substances

bioaccumulating in human food chains23

. It should thus be possible to screen the multitude of

chemicals of commerce for those that do possess such properties, and which therefore are

most likely to be found in environmental and human tissue samples.

Prioritizing chemicals with respect to environmental concern has been an ongoing effort for

several decades. Muir and Howard55

provide an in-depth review of screening methods and

projects aimed specifically at identifying persistent organic pollutants (POPs). The

motivation for such screening has intensified in response to the United Nations Environment

16

Program‘s Stockholm Convention on POPs5, which calls for the identification and global

regulation of chemicals that fulfill the four criteria of persistence, bioaccumulation, long

range transport and toxicity. Nearly all screening initiatives include parameters that address

these criteria when searching for new potential POPs55

.

The task of developing an effective screening method is difficult and complex. A method

must be highly discriminatory to eliminate the many chemicals that are of little concern, and

must simultaneously identify those chemicals which constitute a large hazard. In order to be

considered effective a method must demonstrate a strong selectivity for those chemicals

which cause the highest concern, capturing as many high risk chemicals and as few low risk

chemicals as possible. We present here an approach which is specifically designed to be

highly selective for chemicals which may become Arctic contaminants. The rationale is that

traditional dietary habits place Northern indigenous residents among the human

subpopulations most vulnerable to elevated exposure to POPs40

, and many of the traditional

POPs are of especially high concern to this specific population.

2 Methods

2.1 Outline of the Approach

The approach consists of two parallel screening methodologies; one methodology screens

chemicals based on substance properties and the other screens chemicals based on a

structural profile of known Arctic contaminants. Since substance properties are determined

by molecular structure, both methodologies might be expected to flag the same chemicals as

being of high concern. However, comparing the sets of substances identified by either

methodology can provide valuable insight into their relative strengths and limitations.

The pathway a chemical must follow to become a contaminant in the upper trophic levels of

the Arctic food chain is circuitous; emission, long range transport, deposition, exposure,

bioconcentration and biomagnification all play a role. The primary screening methodology

applied here, and in most screening exercises, examines each step along this pathway and

17

defines a set of criteria which chemicals must meet. Here we rely on three such criteria; the

chemical must have distribution properties that allow it to reach the Arctic marine food

chain, it must have distribution properties that allow it to bioaccumulate in humans, and it

must be sufficiently persistent in the atmosphere to reach the Arctic without degrading. This

methodology rests on three main assumptions; the first is that the potential to accumulate in

the Arctic marine food chain depends on chemical distribution, meaning chemicals with

certain combinations of partitioning properties will have a higher potential to accumulate,

and the second assumption is that long range transport happens mainly via the atmosphere;

and therefore atmospheric oxidation is the primary factor limiting long range transport to the

Arctic. Although a chemical must also be resistant to metabolism to achieve high levels in

top predators of the Arctic food chain, we do not include a metabolic criterion because of the

difficulty of predicting susceptibility to metabolism.

The second screening methodology rests upon the observation that chemicals which have

already been identified as contaminants in the higher trophic levels of the Arctic food chain

must have the correct combination of substance properties to traverse the entire pathway

from emission to human. A chemical profiling method is employed which compares a

chemical's structure to the structures of known Arctic contaminants, and those chemicals

which match the structural profile are assumed to have a higher likelihood of being Arctic

contaminants. Cases where these two paradigms overlap and where they differ are used to

identify possible false positives and negatives, and to assist in classification. These screening

methodologies test a chemical for the potential to be an Arctic contaminant, but to be a risk

the chemical must also be emitted. Ideally usage profiles and emissions data should be

consulted but this is impractical for a large number of chemicals and the data is largely

unavailable, so it is assumed that chemicals with a higher production volume will have a

greater likelihood of significant emissions.

2.2 Compilation of Data

Identifying information (CAS number, SMILES string), physical-chemical properties (molar

mass, octanol-water partition coefficient KOW, Henry's law constant, vapor pressure, aqueous

18

solubility), and the atmospheric oxidation half-lives for 105,584 individual chemicals were

compiled and calculated using the extensive CAS/SMILES string database of the EPISuite

software package41

. These data served in the first screening method. In all cases experimental

values were used where available. It should be noted that EPISuite assumes that all

molecules are in their neutral form, so it is not possible to assess the effects of any

dissociation equilibria. Two separate sets of partitioning properties were calculated from

these data, with the same log KOW used in both cases. In the first set (―HLC data set‖) a value

for the air-water partition coefficient KAW was derived from the Henry's Law constant (KAW =

HLC/(RT)) and a value for the octanol-air partition coefficient KOA was obtained by applying

a thermodynamic cycle with the KOW (KOA = KOW/KAW). If the Henry's Law constant was

unavailable then it was estimated using the ratio of vapor pressure and aqueous solubility. In

the second set of partitioning properties (―VP data set‖) the vapor pressure was used to

estimate a value for the KOA based on an empirical equation from Xiao et al.56

, then a value

for KAW was calculated from KOW (KAW = KOW/KOA).

Ideally experimental values should be used for all partitioning properties and OH oxidation

rates, because the applicability of the EPISuite prediction methods to a very wide range of

organic compounds is untested and doubtful57-59

. However, the list of organic substances with

reliable experimental data is much smaller than the number of chemicals with SMILES

strings in the EPISuite database. Experimental values are also much less likely to be

available for chemicals which have not already been identified as harmful, and screening will

therefore not identify any new potential POPs. Properties predicted from chemical structure

alone, such as those available in the EPISuite database, are of limited and unknown accuracy

but at this time they are the most extensive available and can help identify high concern

chemicals which have not yet been closely examined.

For the second methodology based on structural resemblance, a list of 86 known Arctic

contaminants was compiled. This list (Table 2), references for concentrations found in Arctic

biota, the procedure and the criteria used in compiling the list, as well as a short description

of the six major structural classes of known Arctic contaminants is given in the Appendix.

19

SMILES strings were found in the EPISuite database or composed and used to quantify the

86 chemicals‘ structural properties.

Five different lists were used to determine if a chemical has a high production volume

(HPV): the Canadian Domestic Substances List which flags substances having an annual

production volume greater than 1,000 t (metric tonnes)7, the US EPA's HPV Challenge

Program which includes chemicals that are produced or imported into the US in quantities of

454 t (1 million pounds) or more per year60

, the HPV list compiled by the European

Chemical Bureau‘s European Chemical Substances Information System which includes

chemicals produced or imported to the European Union in quantities of 1000 t or more61

, the

OECD's list of HPV chemicals produced in quantities of 1,000 t or greater in the EU or

another member country62

, and any chemical listed under the US Toxic Substances Control

Act (TSCA) produced in a quantity of 454 t (1 million pounds) or more in any year for which

data is available63

. Chemicals were further checked against three databases of registered

pesticides; the Canadian Domestic Substances List7, the US EPA's list of registered

pesticides64

, and the World Health Organization's list of current use pesticides65

.

2.3 Screening for Long Range Transport to the Arctic and Bioaccumulation in Humans

By linking global transport calculations66

with a food chain bioaccumulation model23

, Czub

et al.40

have delineated the combination of partitioning properties KAW, KOW, and KOA, which

result in a high potential for organic chemicals to be transported to the Arctic and to

accumulate in the Arctic human food chain. They illustrated these combinations by plotting

an Arctic Contamination and Bioaccumulation Potential (AC-BAP) after seventy years for a

multitude of hypothetical perfectly persistent chemicals as a function of the chemical

partitioning space defined by KOA and KAW (Figure 2). The area of elevated AC-BAP is

defined here as exceeding at least 10% of the maximum AC-BAP70 value (red outline in

Figure 2). Over half of the 86 known Arctic contaminants fall within the defined area of

elevated AC-BAP. Some, such as heavily chlorinated PCDD/Fs and PCBs, fall outside of the

red outline because the global transport calculations predict a low potential to reach the

20

Arctic for particle-bound substances with log KOA values above 1140,66

. This subject is

explored further below. The area of elevated AC-BAP comprises the following thresholds:

log KOW ≥ 3.5

log KOA ≥ 6

0.5 ≥ log KAW ≥ -7

log KAW ≤ -1.78 · log KOA + 14.56

(3)

Figure 2 – Chemical Space plot of elevated Arctic contaminant potential.

The red outline delineates chemical partitioning property combinations that result in a

high (i.e. in excess of 10 % of the maximum) potential for accumulation in the Arctic

physical environment and the Arctic human food chain as described by Czub et al.40

.

21

Values for log KOA and log KAW of 86 known Arctic contaminants (experimental values

referenced in the Appendix, and predicted with SPARC67

if not available) are also

shown.

The selection of the 10% line is arbitrary and not very conservative, but was deemed to be a

compromise between selectivity and the possibility of excluding chemicals which should be

flagged as of concern. Using the 1% line instead would greatly improve the agreement

between the model results and the monitoring data, as all but two of the chemicals fall within

the line indicating 1% of the maximum AC-BAP70. However, doing so would essentially

neglect the effects of long range transport because the 1% line is nearly equivalent to the area

of elevated Arctic bioaccumulation40

. The contours of the plot of AC-BAP suggests that the

elevated area extends below log KAW = -5. There are computational reasons for not extending

the plot below this line but no mechanistic reasons, so we have extrapolated two log units

into this area to ensure we capture as many potential Arctic contaminants as possible.

By checking if the partitioning properties of an organic chemical fall within the red outline in

Figure 2, the potential for environmental transport to the Arctic and bioaccumulation in the

Arctic food chain are being screened for simultaneously. Whereas the screening process is

simple, the definition of the area of elevated AC-BAP is based on detailed mechanistic

models and represents a sophisticated and effective exclusion method. Using an enclosed

chemical space has intuitively a higher power of discrimination than using cut-off values for

only one of the three partitioning properties, such as the common use of a log KOW threshold

of 523

, which in itself may not even be valid (Equation (3),ref 23). The partitioning properties

of the entire chemical data set were screened using the thresholds of Equation (3). Both the

HLC data set and the VP data sets were checked and chemicals were considered to meet the

criterion of elevated AC-BAP, if either data set placed the chemical within the red

boundaries.

22

2.4 Screening for Persistence in Air

In the screening step involving the partitioning space, chemical persistence is not being

considered. To filter out chemicals which are quickly degraded in the atmosphere and

therefore have low long range transport potential, an atmospheric oxidation half-life cut-off

is used. Hydroxyl radicals are the primary atmospheric oxidant for most substances68

; so a

hydroxyl radical atmospheric oxidation half-life (OH thalf) was used as the screening

parameter. Some experimental OH thalf values are available in the EPISuite database and the

rest were predicted using EPISuite41

. There will be contributions from other oxidants to the

total atmospheric oxidation rate but the reaction rates with OH are the most reliable69,70

.

Reaction with other oxidants will only increase the overall rate of oxidation, so OH thalf

values may be viewed as the maximum half life for a chemical in air. A cut-off value of two

days was used as suggested by the Stockholm Convention5 and Muir and Howard

55. We

should caution that this criterion could eliminate chemicals that may undergo long range

transport to the Arctic in the oceans, or have a longer oxidation half-life when sorbed to

atmospheric particles.

2.5 Screening for Resemblance with the Structural Profile of Known Arctic Contaminants

There is a distinct set of combinations of partitioning and persistence properties which meet

all the criteria of the first screening methodology, but these properties are themselves

governed by chemical structure. This implies that there may be a limited number of possible

combinations of structural features that give rise to the properties of an Arctic contaminant.

A structure property relationship (SPR) was developed to predict if a chemical has the

potential to be an Arctic contaminant based on chemical structure. The vast majority of SPRs

rely on multiple linear regressions but that approach is impractical in this case as no

continuous scale of the potential to become an Arctic contaminant exists that can be used to

parameterize a regression equation. Instead the list of 86 known Arctic contaminants is

compared to the full list of chemicals and deviations from a "typical" molecule are quantified

for 3 structural parameters; halogenation, internal connectivity and molecular size. A detailed

23

description of the derivation and application of the structural profile of Arctic contaminants

is provided in the Appendix (Section 5.1).

3 Results and Discussion

3.1 Preliminary Screening

Figure 3 shows a schematic representation of how the two screening methodologies classify

the screened chemicals. From the list of 105,584 chemicals 16,888 chemicals fall within the

boundaries of elevated AC-BAP using either the HLC or VP data sets (see Figure 5 of the

Appendix). A total of 13,905 chemicals are identified by the HLC data set, 14,957 are

identified by the VP data set and 11,974 chemicals are identified by both data sets. Neither

data set shows a significant bias towards over- or underestimating the log KAW or log KOA

values so it can not be immediately determined which data set is more accurate. A total of

16,151 chemicals have experimental or estimated OH thalf values greater than two days. There

are 2,025 chemicals which both fall within the chemical space and have an OH thalf value

greater than two days. There are 3,088 chemicals which match the structural profile of known

Arctic contaminants.

More than two thirds of the chemicals screened (74,016 chemicals or 70.1%) did not meet

any of the three screening criteria and are unlikely to reach the Arctic or bioaccumulate in

humans in this region. A further 12,669 (12.0%) chemicals do not match the structural profile

of known Arctic contaminants and fall outside of the area of elevated AC-BAP, but are

persistent in air. These chemicals may possibly reach the Arctic environment but because the

area of elevated AC-BAP is strongly influenced by the potential to bioaccumulate they are

unlikely to lead to significant human exposure. An additional 14,608 (13.8%) chemicals fall

within the area of elevated AC-BAP but are not persistent in air and do not match the

structural profile of known Arctic contaminants. It is unlikely that a significant fraction of

these chemicals will reach the Arctic and become contaminants. Eliminating these three

groups of chemicals from further consideration excludes 101,293 chemicals (95.9%).

24

Is the chemical

persistent in air?

no

yes

Do the chemicals have

partitioning

properties that allow

for accumulation in

the physical

environment and in the

human food chain of

the Arctic?

no

yes

Does the chemical match

the structural profile of known

Arctic contaminants?

yesno

yesno

yesno

yesno

Very unlikely to be an

Arctic contaminant.

Possible false positive by profiling criteria

or both unusual exposure route and

bioaccumulation mechanism

Chemical is unlikely to reach the Arctic

human food chain in significant amounts

or bioaccumulate there

Possible false negative by

LRT/B/P methodology due to unusual

bioaccumulation or transport mechanism

Chemical will likely degrade en route to

the Arctic human food chain

Possible false negative by LRT/B/P

methodology due to direct emissions or

non-gas phase persistence

Possible false positive by

LRT/B/P methodology or possible

non-POP like Arctic contaminant

Very likely to have the properties that

cause a chemical to become an

Arctic contaminant

1,203 (1.1%, 32 HPVC) 822 (0.8%, 27 HPVC)

255 (0.2%, 6 HPVC)14,608 (13.8%, 535 HPVC)

1,457 (1.4%, 49 HPVC)12,669 (12.0%, 1,042 HPVC)

554 (0.5%, 6 HPVC)74,016 (70.1%, 2,626 HPVC)

Figure 3 - Schematic diagram of the screening process. The 105,584 chemicals in the EPI Suite database were screened for

their potential to accumulate in the Arctic food chain.

25

3.2 Chemicals Matching the Structural Profile of Known Arctic Contaminants

There are 2,266 chemicals (2.1%) which despite matching the structural profile of known

Arctic contaminants fail to meet one or both of the other two screening criteria. Of these

there are 554 (0.5%) that fail to meet both the persistence in air and chemical space criterion

and 255 (0.2%) that fall within the area of elevated AC-BAP but are not persistent in air.

Due to low air persistence these chemicals are unlikely to reach the Arctic, however it is

possible that their degradation products will be of concern. The oxidation products could be

more persistent and their partitioning properties may be different, but predicting the major

oxidation products of the 809 chemicals identified is beyond the scope of this study.

An additional 1,457 (1.4%) chemicals match the structural profile of known Arctic

contaminants and are persistent in air but fall outside of the area of elevated AC-BAP. Figure

4 shows the location in the chemical space of all 3,088 chemicals which match the structural

profile of known Arctic contaminants. There are thus large numbers of substances that are

structurally similar to known Arctic contaminants, but nevertheless do not possess the right

combination of partitioning properties to become Arctic contaminants based on the model

predictions and assumptions used in this study. A closer inspection of Figure 4 reveals three

types of chemicals outside of the red boundaries. It is enlightening to explore what prevents

these chemicals from being classified as potential Arctic contaminants, because it could

possibly reveal the limitations of both the screening and the chemical profiling technique.

Type 1: Volatile Chemicals. Many chemicals are located to the upper left of the high AC-

BAP area, indicative of high log KAW (> 0.5) and low log KOA (< 8) values. These chemicals

are similar to known Arctic contaminants in size and degree of halogenation but are

considerably more volatile, and thus will not bioaccumulate in air-breathing organisms40

.

The explanation is that these are mostly highly fluorinated chemicals, which experience

weak van der Waals interactions in comparison to their highly chlorinated analogs, and so

partition relatively weakly to condensed phases71

. Perfluorinated alkanes are examples of

such chemicals.

26

Figure 4 - Locations of all chemicals in the EPISuite database which match the

structural profile of known Arctic contaminants. Plotted using the "HLC data set".

Type 2: Highly Sorptive Chemicals. A significant number of chemicals which match the

structural profile of known Arctic contaminants lie to the right of the high AC-BAP area,

having a predicted log KOA above 8 and a log KOW above 5. These are low-volatility and

sparingly water-soluble chemicals that sorb predominately to particles in both the

atmosphere and aqueous media. Particle settling processes are predicted to prevent efficient

long range transport in both atmosphere and oceans66

. Examples are octachlorodibenzo-p-

dioxin and decabromodiphenyl ether. It is noteworthy to point out that these chemicals are

predicted to have considerable potential for bioaccumulation23

and that some known Arctic

contaminants indeed have such partitioning properties (Figure 2, Table 2). These chemicals

27

have the potential to become Arctic contaminants if they are emitted close to the Arctic. It is

also possible that particle-bound atmospheric transport is more efficient in delivering

contaminants to the Arctic than is currently described in global transport calculations; one

reason may be because the effect of intermittent rain is ignored72

. Another possible long

range transport mechanism for these chemicals is biotransport, as migrating organisms may

take up these contaminants in their southern wintering grounds73

.

Type 3: Small, Polar Chemicals. The majority of the chemicals which match the structural

profile of known Arctic contaminants and fall outside of the area of elevated AC-BAP do so

because their log KOW values are too low. Many of these chemicals are small and polar due

to the placement of polar functional groups on the "back-bone" of the molecule, making

them too hydrophilic to bioaccumulate.

3.3 Potential Arctic Contaminants

There are 2,025 (1.9%) chemicals which both fall within the area of elevated AC-BAP and

are persistent in air; 822 (0.8%) of these also match the structural profile of known Arctic

contaminants. All 822 chemicals that have partitioning properties corresponding to high AC-

BAP, are persistent in air, and match the structural profile of known Arctic contaminants are

listed in Table 3 of the Appendix. The majority of these chemicals have already been

detected in Arctic biota; the general chemical classes and the number of chemicals flagged

are shown in Table 6 of the Appendix. Major chemical groups of environmental concern,

such as PCBs, PCNs, PCDDs, PCDFs and PBDEs are identified, along with a number of

homologues which are halogenated with fluorine, bromine, iodine or combinations thereof

with chlorine atoms. There are also a large number of homologues of various chemical

classes which are substituted with halogenated alkyl side chains. One of the larger chemical

classes is polychlorinated diphenylethers (PCDEs) which have been previously quantified in

environmental samples74

. Chlorinated benzenes, phenols, nitrophenols and nitrobenzenes are

named along with other chlorinated benzene derivatives such as toluene, aniline and others.

An almost equally large number of these are flagged which contained halogens other than

chlorine. A structurally diverse group of halogenated or halo-alkyl triazine derivatives are

28

also among the potential POPs. Organochlorine pesticides designated by the Stockholm

Convention (mirex, chlordanes, toxaphene, HCH, HCB) are also identified along with a

number of other obsolete and current-use pesticides.

Remaining are 1,203 (1.1%) chemicals which both fall within the area of elevated AC-BAP

and are persistent in air but do not match the structural profile of known Arctic contaminants.

This group thus includes chemicals that we would not necessarily suspect to become Arctic

contaminants based on structural similarity, but are predicted to have the respective

properties. It would thus contain any truly novel type of Arctic contaminant. Out of the 1,203

chemicals 793 have some degree of halogenation, and 483 of these fall into one of the

chemical classifications that comprise the majority of the list of 822 POP-like chemicals. For

example there are 20 mono- and dichloro- PCBs which are excluded from the list of POP-

like chemicals because of their low degree of chlorination. Almost all of the 1,203 non POP-

like chemicals have at least some portion which is aromatic. A large number of the chemicals

that are not halogenated are nitro-substituted aromatics or contain a silane or siloxane

substituent. There are also 43 current-use or obsolete pesticides which fall into this category.

3.4 High Risk Chemicals

There are 4,291 chemicals which are identified as potential Arctic contaminants by one or

both screening methodologies. These chemicals were checked against the five HPV chemical

lists and 120 were found to be HPV chemicals; the identified HPV chemicals are

summarized in Table 1. A number of chemicals which were included in the set of 86 known

Arctic contaminants (PBDEs, HBCD, endosulfan, DDT, -HCH and HCB) appear in the list

of 120 HPV chemicals. Some of these have been banned or severely restricted in North

America but appear on the European HPV list, which was most recently updated in 2000.

There are three chemicals structurally related to Endosulfan (CAS 115275, 115286,

2157199) among the non-persistent molecules identified, including two of the chemicals

which also fall outside of the area of elevated AC-BAP. Representatives of the three types of

29

chemicals that fall outside of the area of elevated AC-BAP (Figure 4) are present.

Perfluoroundecane, which is intended to represent perfluoroalkanes with chain length

between 5 and 18 carbons in the EPISuite SMILES database (CAS 86508421), falls under

the volatile category (type 1), PBDEs are particle-bound substances (type 2) and the three

pesticides chlorothalonil, picloram and nitrapyrin are small, polar chemicals (type 3).

Twenty-seven chemicals are identified by both screening methods, three of which are in the

training set. Exact usage profiles for some of these chemicals are difficult to obtain but most

appear to be used as intermediates in the synthesis of halogenated pesticides or polymers.

Some may not be a major environmental concern, for example the two

trichloromethylpyridines (CAS 1817136, 69045836) and pentachloropyridine (CAS

2176627) are reportedly used only in closed systems75

. The brominated species (CAS

632791) is used as a flame retardant in polymers76

. Perhaps of more concern is the presence

of a current use pesticide, pentachloronitrobenzene (CAS 82688). Also of concern are the

chlorinated cyclopentanes and cyclopentenes which are highly chlorinated but have virtually

no information available regarding their usage profiles.

30

Table 1 - Summary of highproduction volume chemicals which are predicted to become

Arctic contaminants or which match the structural profile of known Arctic

contaminants.

CAS AC-

BAP(a)

Persistent

(b) Profile

(c) Pesticide

(d) Name

115286 no no yes no 1,4,5,6,7,7-hexachloro-5-norbornene-2,3-dicarboxylic acid

1691992 no no yes no N-ethyl-1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-N-(2-hydroxyethyl)-1-octanesulfonamide

2157199 no no yes no Endosulfan alcohol

25637994 no no yes no hexabromocyclododecane (1,3,5,7,9,11-hexabromocyclododecane)

(e)

27905459 no no yes no 3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,10-heptadecafluorodecyl, 2-propenoic acid ester

61262531 no no yes no 1,1'-[1,2-ethanediylbis(oxy)]bis-2,3,4,5,6-pentabromobenzene

77474 no yes yes yes 1,2,3,4,5,5-hexachloro-1,3-cyclopentadiene

87616 no yes yes no 1,2,3-trichlorobenzene

87683 no yes yes no 1,1,2,3,4,4-hexachloro-1,3-butadiene

98157 no yes yes no 1-chloro-3-(trifluoromethyl)-benzene

98464 no yes yes no 1-nitro-3-(trifluoromethyl)-benzene

98566 no yes yes no 1-chloro-4-(trifluoromethyl)-benzene

108770 no yes yes no 2,4,6-trichloro-1,3,5-triazine

115253 no yes yes no Octafluorocyclobutane

120821 no yes yes yes 1,2,4-trichlorobenzene

307357 no yes yes no 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-1-octanesulfonyl fluoride

311897 no yes yes no 1,1,2,2,3,3,4,4,4-nonafluoro-N,N-bis(nonafluorobutyl)-1-butanamine

328847 no yes yes no 1,2-dichloro-4-(trifluoromethyl)-benzene

329011 no yes yes no 1-isocyanato-3-(trifluoromethyl)-benzene

335422 no yes yes no heptafluoro-butanoyl fluoride

338841 no yes yes no 1,1,2,2,3,3,4,4,5,5,5-undecafluoro-N,N-bis(undecafluoropentyl)-pentanamine

423507 no yes yes no 1,1,2,2,3,3,4,4,5,5,6,6,6-tridecafluoro-1-hexanesulfonyl fluoride

428591 no yes yes no trifluoro(trifluoromethyl)-oxirane

647427 no yes yes no 3,3,4,4,5,5,6,6,7,7,8,8,8-tridecafluoro-1-octanol

678397 no yes yes no 3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,10-heptadecafluoro-1-decanol

719324 no yes yes no 2,3,5,6-tetrachloro-1,4-benzenedicarbonyl dichloride

865861 no yes yes no 3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11,12,12,12-heneicosafluoro-1-dodecanol

1163195 no yes yes no decabromodiphenylether (BDE-209)

1737935 no yes yes no 3,5-dichloro-2,4,6-trifluoropyridine

1897456 no yes yes yes Chlorothalonil

1918021 no yes yes yes Picloram

31

1929824 no yes yes yes Nitrapyrin

2043530 no yes yes no 1,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8-heptadecafluoro-10-iododecane

2043541 no yes yes no 1,1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10-heneicosafluoro-12-iodododecane

2043574 no yes yes no 1,1,1,2,2,3,3,4,4,5,5,6,6-tridecafluoro-8-iodooctane

2402791 no yes yes no 2,3,5,6-tetrachloropyridine

3194556 no yes yes no 1,2,5,6,9,10-hexabromocyclododecane

3825261 no yes yes no Ammonium perfluorooctanoate

5848931 no yes yes no 5-chloro-3-(trichloromethyl)-1,2,4-thiadiazole

10469097 no yes yes no 3,4,5,6-tetrachloro-2-pyridinecarboxylic acid

14143603 no yes yes no 4-amino-3,5,6-trichloro-2-pyridinecarbonitrile

17824838 no yes yes no 3,4,5,6-tetrachloro-2-pyridinecarbonitrile

32534819 no yes yes no pentabromodiphenylether (BDE-99)(e)

32536520 no yes yes no octabromodiphenylether (BDE-203)(e)

36483600 no yes yes no hexabromodiphenylether (BDE-167)(e)

40088479 no yes yes no tetrabromodiphenylether (BDE-55)(e)

52314677 no yes yes no 3-(2,2-dichloroethenyl)-2,2-dimethyl-cyclopropanecarbonyl chloride

59808785 no yes yes no tetrachlorocyclopentane (1,2,3,4-tetrachlorocyclopentane)

(e)

60825265 no yes yes no (3,5,6-trichloro-2-pyridinyl)oxy-acetic acid, methyl ester

63936561 no yes yes no nonabromodiphenylether (BDE-206)(e)

68928803 no yes yes no heptabromodiphenylether (BDE-173)(e)

69045789 no yes yes no 2-chloro-5-trichloromethylpyridine

86508421 no yes yes no perfluoro compounds C5-18 (perfluoroundecane)

(e)

138495428

no yes yes no 1,1,1,2,2,3,4,5,5,5-decafluoropentane

163702076

no yes yes no 1,1,1,2,2,3,3,4,4-nonafluoro-4-methoxy-butane

101053 yes no yes yes Anilazine

115275 yes no yes no 1,4,5,6,7,7-hexachloro-5-norbornene-2,3-dicarboxylic anhydride

115297 yes no yes yes Endosulfan

3278895 yes no yes no 1,3,5-tribromo-2-(2-propenyloxy)-benzene

3734483 yes no yes no Chlordene

24448097 yes no yes no 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-N-(2-hydroxyethyl)-N-methyl-1-octanesulfonamide

78637 yes yes no no 1,1'-(1,1,4,4-tetramethyl-1,4-butanediyl)bis[2-(1,1-dimethylethyl) peroxide

80079 yes yes no no 1,1'-sulfonylbis[4-chlorobenzene]

80104 yes yes no no Dichlorodiphenylsilane

81141 yes yes no no 1-[4-(1,1-dimethylethyl)-2,6-dimethyl-3,5-dinitrophenyl]-ethanone

81152 yes yes no no 1-(1,1-dimethylethyl)-3,5-dimethyl-2,4,6-trinitrobenzene

32

84515 yes yes no no 2-ethyl-9,10-anthracenedione

88062 yes yes no yes 2,4,6-trichlorophenol

88857 yes yes no yes Dinoseb

98737 yes yes no no 4-(1,1-dimethylethyl)-benzoic acid

101633 yes yes no no 1,1'-oxybis 4-nitrobenzene

102363 yes yes no no 1,2-dichloro-4-isocyanatobenzene

115322 yes yes no yes Dicofol

118796 yes yes no no 2,4,6-tribromophenol

136607 yes yes no no benzoic acid, butyl ester

144796 yes yes no no chloromethyldiphenylsilane

147820 yes yes no no 2,4,6-tribromobenzenamine

320729 yes yes no no 3,5-dichloro-2-hydroxybenzoic acid

540976 yes yes no no 2,2,4,4,6,6,8,8,10,10,12,12-dodecamethylcyclohexasiloxane

1185097 yes yes no no 1,1,2,2-tetrachloro-ethanesulfenyl chloride

1539044 yes yes no no 1,4-benzenedicarboxylic acid, 1,4-diphenyl ester

1836755 yes yes no yes Nitrofen

2116849 yes yes no no 1,1,1,5,5,5-hexamethyl-3-phenyl-3-[(trimethylsilyl)oxy]-trisiloxane

2392485 yes yes no no 4-chloro-1-(2,4-dichlorophenoxy)-2-nitrobenzene

3457612 yes yes no no 1,1-dimethylethyl 1-methyl-1-phenylethyl peroxide

13472087 yes yes no no 2,2'-azobis-2-methylbutanenitrile

34893920 yes yes no no 1,3-dichloro-5-isocyanatobenzene

35578473 yes yes no no bis(4-bromophenyl)-ethanedione

50594440 yes yes no no 5-[2-chloro-4-(trifluoromethyl)phenoxy]-2-nitrophenol, 1-acetate

50594779 yes yes no no 3-[2-chloro-4-(trifluoromethyl)phenoxy]-phenol, 1-acetate

52270447 yes yes no no Neodecanoic acid, cobalt(2+) salt

63734623 yes yes no no 3-(2-Chloro-4-trifluoromethylphenoxy)-benzoic acid

64667330 yes yes no no 4,6,6,6-Tetrachloro-3,3-dimethylhexanoic acid methyl ester

50293 yes yes yes yes p,p'-DDT

58899 yes yes yes yes -HCH

68360 yes yes yes no 1,4-bis(trichloromethyl)-benzene

82688 yes yes yes yes pentachloronitrobenzene

87843 yes yes yes no 1,2,3,4,5-pentabromo-6-chlorocyclohexane

95943 yes yes yes no 1,2,4,5-tetrachlorobenzene

117088 yes yes yes no 4,5,6,7-tetrachloro-1,3-isobenzofurandione

118741 yes yes yes yes HCB

133493 yes yes yes no pentachlorobenzenethiol

626391 yes yes yes no 1,3,5-tribromobenzene

632791 yes yes yes no 4,5,6,7-tetrabromo-1,3-isobenzofurandione

634662 yes yes yes yes 1,2,3,4-tetrachlorobenzene

1134049 yes yes yes no 2,3,4,5-tetrachloro-6-(trichloromethyl)-pyridine

33

1203867 yes yes yes no 2,2-dichloro-1-(2,4,5-trichlorophenyl)-ethanone

1817136 yes yes yes no 3,6-dichloro-2-(trichloromethyl)-pyridine

2176627 yes yes yes no pentachloropyridine

5216251 yes yes yes no 1-chloro-4-(trichloromethyl)-benzene

17700093 yes yes yes no 2,3,4-trichloronitrobenzene

29091096 yes yes yes no 2,4-dichloro-1,3-dinitro-5-(trifluoromethyl)-benzene

30554724 yes yes yes no tetrabromodichlorocyclohexane (1,2,3,4-tetrabromo-5,6-dichlorocyclohexane)

(e)

62111471 yes yes yes no heptachlorocyclopentene (1,2,3,3,4,4,5-heptachloro-1-cyclopentene)

(e)

68258902 yes yes yes no heptachlorocyclopentane (1,2,3,3,4,5,5-heptachlorocyclopentane)

(e)

68258913 yes yes yes no hexachlorocyclopentane (1,2,2,3,4,4-hexachlorocyclopentane)

(e)

69045836 yes yes yes no 2,3-dichloro-5-trichloromethylpyridine

72030263 yes yes yes no hexachlorocyclopentene (1,2,3,4,5,5-hexachloro-1-cyclopentene)

(e)

73588428 yes yes yes no 1,3-dichloro-5-(1,3,3,3-tetrachloro-1-methylpropyl)-benzene

75147205 yes yes yes no 2,2,3,4,4-pentachloro-3-butenoic acid, butyl ester

(a) Indicates if the chemical falls within the area of elevated AC-BAP.

(b) Indicates if the chemical has an OH thalf greater than 2 days.

(c) Indicates if the chemical matches the structural profile of known Arctic contaminants.

(d) Indicates if the chemical appears on one of the three pesticide lists.

(e) The CAS is for an unspecified isomer or analogue, the specific chemical noted in brackets corresponds to

the SMILES string found in the EPISuite database.

The majority of the HPV chemicals which are persistent in air and fall within the area of

elevated AC-BAP but do not match the structural profile of known Arctic contaminants are

halogenated to some degree. Three chlorinated pesticides are included in this group; 2,4,6-

trichlorophenol, dicofol and nitrofen. An interesting result is the presence of a non-

halogenated pesticide Dinoseb, and dodecamethylcyclohexasiloxane (D6) which is an

ingredient found in cosmetic and personal hygiene products77

.

The selectivity of the screening method presented here could be varied by changing the

criteria thresholds. For example, the net could be cast wider by defining the thresholds for

the partitioning properties (Equation (3)) based on 1 % of maximum AC-BAP70 or it could

be made more selective by raising the threshold for the atmospheric oxidation half life. The

thresholds for partitioning and atmospheric oxidation could be made less stringent, i.e. the

danger of false-negatives could be reduced, if reliable screening steps for biotransformation

34

and environmental persistence could be devised and added, which would limit the number of

false positives. Another research need is the development of methods to predict the

partitioning, environmental fate and food chain accumulation of charged chemical species.

We suggest the next steps with respect to the identified potential Arctic contaminants include

the verification of the property data used in this study through laboratory measurements or

more sophisticated estimation techniques as well as laboratory tests of bioaccumulation to

assess the potential for biotransformation. If a substance should indeed be shown to be

resistant to metabolism, Arctic monitoring programs should search for it.

4 Acknowledgements We acknowledge funding from the Long-range Research Initiative of the European Chemical

Industry Association (CEFIC), and Michael McLachlan for insightful comments.

5 Appendix

5.1 Compiling a List of Known Arctic Contaminants

To minimize possible "sampling error" in constructing the list of known Arctic contaminants

four criteria were applied: (1) Chemicals must have been positively identified and quantified

in the tissues of Arctic biota, preferably in biota at the higher trophic levels of the food chain.

(2) Congeners with consistently higher concentrations in Arctic biota are selected over

congeners with lower concentrations, (3) Chemical structure groupings should be balanced to

contribute approximately equal numbers of chemicals to the list, and (4) in cases where some

chemicals must be excluded to balance the list, the remaining chemicals should be

representative of the structural diversity of the chemical group. Note that all of the

contaminants selected are traditional POP-like contaminants, meaning references to "known

Arctic contaminants" here and elsewhere in the Appendix refer specifically to "POP like

Arctic contaminants".

Three recent review papers identified major POPs present in marine78

, freshwater79

and

terrestrial biota80

from the Canadian Arctic. The anthropogenic contaminants in higher

35

trophic organisms (fish, birds and mammals) identified in these reviews included PCBs,

HCHs, DDTs, dieldrin, endosulfan, chlordanes, chlorobenzenes, toxaphenes, PCDDs,

PCDFs, PBDEs, PCNs, perfluorinated alkyl acids (PFAs) and chlorinated paraffins (CPs).

All of the identified chemicals are halogenated and have "backbone" structures composed

primarily of carbon. Sample chemicals containing all three of the above halogens (F, Br, Cl)

are included, but the chemical structure grouping is based on the "backbone" of the

molecule. This is not an exhaustive list of known Arctic contaminants and there are many

that have been detected but are not listed here.

The first chemical group is biphenyls and naphthalenes. Coplanar biphenyls and

naphthalenes are structurally similar and have similar chemical and toxicological

properties81,82

. Ortho-substituted biphenyls are also included in this group to represent the

structural diversity of group. The second chemical group is bridged diphenyls which includes

PBDEs and DDTs. DDT has different steric factors which affect the orientation of the rings

than those that affect PBDEs83,84

, but they are structurally similar enough for the purposes

here because SMILES string contain no information regarding steric factors. The third

chemical group is three-ring heterocycles which includes PCDDs and PCDFs. The fourth

chemical group is monocyclics which includes HCBs, HCHs and polychlorinated styrenes

(PCSs). The fifth chemical group is straight-chain alkyls which include PFAs and CPs. The

sixth and final chemical group is chemicals containing a norbornene or norbornane structure

which includes dieldrin, endosulfan, chlordanes and toxaphenes. Another chemical, mirex, is

also included in this group. Though mirex does not contain a norbornene or norbornane

structure it is more similar to this chemical group than the other groups.

36

Table 2 - Names, CAS numbers, and the overall POP score for a list of known Arctic

contaminants

Also given is the contribution of the three chemical profiling criteria SA (structural

atoms), CX (degree of halogenation), XS (degree of internal connectivity) to the POP

score. The references provide concentrations for the contaminants in Arctic biota and

physical-chemical properties. See page S2 for description of how this list was compiled.

biphenyls and naphthalenes

Arctic Reference

Property Ref.

CAS SA

Cont. CX

Cont. XS

Cont. SCORE

(tetra) PCN-28 85 86 53555638 -0.591 2.749 1.875 4.033

(tetra) PCN-29 85 86

149864788 -0.591 2.749 1.875 4.033


67922218 -0.591 2.749 1.875 4.033


149864802 -0.591 2.749 1.875 4.033

(penta) PCN-52 85,87 86

53555650 -0.591 3.208 1.875 4.492

(penta) PCN-54 85,87 86

150224161 -0.591 3.208 1.875 4.492

(penta) PCN-61 85 86

150224229 -0.591 3.208 1.875 4.492

(hexa) PCN-66 85,78 86

103426966 -0.591 3.609 1.875 4.893

(hexa) PCN-69 85 86

103426944 -0.591 3.609 1.875 4.893

(hexa) PCN-71 85 86

90948280 -0.591 3.609 1.875 4.893

(tetra-o) PCB-52 87 81

35693993 -0.235 2.406 1.563 3.733


32598100 -0.235 2.406 1.563 3.733


32690930 -0.235 2.406 1.563 3.733

(penta-o) PCB-99 90 89

38380017 -0.235 2.830 1.563 4.157

(penta-no) PCB-118 88 81

31508006 -0.235 2.830 1.563 4.157

(penta-no) PCB-126 87,91 89

57465288 -0.235 2.830 1.563 4.157

(hexa-o) PCB-138 88,90 81

35065282 -0.235 3.208 1.563 4.535

(hexa-o) PCB-153 88,90 81

35065271 -0.235 3.208 1.563 4.535

(hexa-no) PCB-169 87,91 89

32774166 -0.235 3.208 1.563 4.535

(hepta-o) PCB-170 88,90 89

35065306 -0.235 3.545 1.563 4.872

(hepta-o) PCB-180 88,90 81

35065293 -0.235 3.545 1.563 4.872

bridged diphenyls

p,p'-DDD 91 92

72548 -1.062 2.138 1.339 2.416

p,p'-DDE 91,93 92

72559 -1.062 2.138 1.339 2.416

p,p'-DDT 91,93 92

50293 -1.062 2.532 1.339 2.810

BDE-28 94 95,96

41318756 -0.649 1.804 1.442 2.598

BDE-47 94,97,98,99 95,96

5436431 -0.649 2.264 1.442 3.058

BDE-99 94,98,99 95,96

60348609 -0.649 2.673 1.442 3.467

BDE-100 94,98,99 95,96

189084648 -0.649 2.673 1.442 3.467

BDE-153 94,98,99 95,96

68631492 -0.649 3.039 1.442 3.833

BDE-154 94,98,99 95,96

207122154 -0.649 3.039 1.442 3.833

three-ring heterocycles

2,3,7,8-tetraCDD 87 100

1746016 -1.062 2.138 2.009 3.086

1,2,3,7,8-pentaCDD 87 100

40321764 -1.062 2.532 2.009 3.479

1,2,3,4,7,8-hexaCDD 87 100

39227286 -1.062 2.887 2.009 3.834

37

1,2,3,6,7,8-hexaCDD 87 100

57653857 -1.062 2.887 2.009 3.834

1,2,3,7,8,9-hexaCDD 87 100

19408743 -1.062 2.887 2.009 3.834

1,2,3,4,6,7,8-heptaCDD

87 100 35822469 -1.062 3.208 2.009 4.155

OCDD 87 100

3268879 -1.062 3.499 2.009 4.446

2,3,7,8-tetraCDF 87 100

51207319 -0.649 2.264 2.164 3.779

1,2,3,7,8-pentaCDF 87 100

57117416 -0.649 2.673 2.164 4.188

2,3,4,7,8-pentaCDF 87 100

57117314 -0.649 2.673 2.164 4.188

1,2,3,4,7,8-hexaCDF 87 100

70648269 -0.649 3.039 2.164 4.554

1,2,3,6,7,8-hexaCDF 87 100

57117449 -0.649 3.039 2.164 4.554

2,3,4,6,7,8-hexaCDF 87 100

60851345 -0.649 3.039 2.164 4.554

1,2,3,7,8,9-hexaCDF 87 100

72918219 -0.649 3.039 2.164 4.554

1,2,3,4,6,7,8-heptaCDF

87 100 67562394 -0.649 3.368 2.164 4.883

1,2,3,4,7,8,9-heptaCDF

87 100 55673897 -0.649 3.368 2.164 4.883

OCDF 87 100

39001020 -0.649 3.666 2.164 5.181

monocyclics

PeCB 93,101 92

608935 -2.244 4.374 1.563 3.693

HCB 93,98,101,102 92

118741 -2.244 4.811 1.563 4.130

-HCH 90,93 103

319846 -2.244 4.811 1.563 4.130

-HCH 90,91,93 103

319857 -2.244 4.811 1.563 4.130

-HCH 93 103

58899 -2.244 4.811 1.563 4.130

HBCD 94,99 67

25637994 -0.235 3.208 0.781 3.753

PtCS (pentachlorostyrene)

104 67 14992815 -1.418 3.701 1.172 3.456

HxCS (hexachlorostyrene-

beta-trans) 104 67

90301921 -1.418 4.124 1.172 3.878

HpCS (heptachlorostyrene-

beta,beta) 102(a),104 67

29082755 -1.418 4.491 1.172 4.245

OCS (octachlorostyrene)

102,104,105 67 29082744 -1.418 4.811 1.172 4.566

alkyls

perfluorooctane sulfonate (PFOS)

90,106 58 45298906 -0.235 5.641 0.000 5.406

perfluorooctane sulfonamide (PFOSA)

91 58 754916 -0.235 5.641 0.000 5.406

perfluorooctanoic acid (PFCA-8)

106 58 335671 -0.591 5.774 0.000 5.183

perfluorononanoic acid (PFCA-9)

90,106 58 375951 -0.178 5.842 0.000 5.665

perfluorodecanoic acid (PFCA-10)

90,106 58 335762 -0.235 5.898 0.000 5.662

perfluoroundecanoic acid (PFCA-11)

90,106 58 2058948 -0.649 5.944 0.000 5.295

chlorinated paraffin (C10,Cl7)

b

107 67 32534784 -0.591 3.962 0.000 3.371


b

107 67 214327669 -0.591 4.277 0.000 3.686

chlorinated paraffin 107 67

219697082 -0.591 4.558 0.000 3.967

38

(C10,Cl9)b


b

107 67 none

assigned -0.178 3.742 0.000 3.564

chlorinated paraffin C11,Cl8)

b

107 67 36312819 -0.178 4.052 0.000 3.874


b

107 67 none

assigned -0.178 4.330 0.000 4.152


b

107 67 214327681 -0.235 3.208 0.000 2.972


b

107 67 214327692 -0.235 3.545 0.000 3.310


b

107 67 214327716 -0.235 4.124 0.000 3.889

norbornenes-norbornanes

oxychlordane 90,93,98 100

27304138 -0.178 4.052 3.409 7.283

Heptachlor-exo-epoxide

91,98 92 1024573 -0.178 3.742 3.409 6.974

trans-nonachlor 91,98 100

39765805 -0.591 4.558 2.813 6.780

MC 6 98 67

98318979 -0.591 4.558 2.813 6.780

Aldrin 101(a) 92

309002 -0.235 3.208 3.125 6.097

Dieldrin 91,93,102,101(a) 92

60571 -0.649 3.039 3.606 5.996

Endrin 91 92

72208 -0.649 3.039 3.606 5.996

a-endosulfan 93(a),101(a) 92

959988 -0.649 3.039 2.164 4.554

b-endosulfan 93(a) 92

33213659 -0.649 3.039 2.164 4.554

endosulfan-sulfate 93 67

1031078 -1.062 2.887 2.009 3.834

Toxaphene 26 91 67

142534712 -0.591 4.277 1.875 5.561

Toxaphene 50 91 67,108

66860808 -0.591 4.811 1.875 6.095

Toxaphene 62 91 67,108

154159065 -0.591 4.558 1.875 5.842

Mirex 91,101 67,108

2385855 -0.591 5.249 4.688 9.346

a The authors present concentrations for chemical groups which include this congener, but do not

explicitly list the concentrations for individual congeners.

b There is no chemical structure available for an "average" chlorinated paraffin, chemical structures

were made by taking a carbon chain of the correct length and adding the correct number of chlorine

atoms evenly spaced along the chain.

39

Figure 5 - Plot of all 105,581 chemicals in the EPI Suite database.

Experimental values are used for the plot where available, otherwise predicted values

from the "HLC data set" are used. Chemicals with an OH oxidation half-life less than

two days are highlighted in green, and the area of elevated AC-BAP is delineated in

red.

40

5.2 Profiling Arctic Contaminants

All of the data needed to create a profile of known Arctic contaminants is theoretically

contained in the SMILES strings of the EPISuite database. Parameters such as the number

and types of atoms and the first-order bond connectivity are easily extracted by simple

parsing algorithms. A more difficult task is selecting the parameters to use in the chemical

profile. Multiple linear regression is not possible because we do not have a continuous and

universal scale against which to correlate a multiple linear regression; we can only say with

certainty whether a chemical accumulates in Arctic biota or not, which is discreet data. The

basic approach taken to profiling Arctic contaminants is as follows; first a hypothetical

"typical" molecule is characterized using the list of all available SMILES strings, then this

molecule is compared with a ―typical‖ Arctic contaminant, derived from the list of 86

contaminants mentioned above, to determine how the structural attributes of known Arctic

contaminants deviate from the hypothetical "typical" molecule. The observed deviations are

finally combined to create a "POP score" which indicates how closely a chemical matches

the profile of known Arctic contaminants. Several criteria must be met in order to ensure that

the POP score is robust and has predictive power. The most important factor is the selection

of known Arctic contaminants which has already been discussed. There should be as few

chemical parameters as possible, each should have a sound mechanistic reason for being

included, and the parameters should not be inter-correlated.

Examining the list of 86 known contaminants some general observations can be made about

the structural characteristics they share and which therefore may be important in rendering a

chemical a potential Arctic contaminant. A widely accepted characteristic used to help

identify POPs by structure is chlorination109

and many known Arctic contaminants have a

relatively high degree of chlorination. More recently perfluorinated organic acids and

brominated flame retardants have become a concern and halogenation in general can be

viewed as a possible indicator that a chemical may be a POP99,110

. All of the 86 known Arctic

contaminants fall within a fairly narrow range of molecular size, suggesting that there is an

optimum. If a chemical is too small it will be too volatile to be deposited in the Arctic and it

will be exhaled by organisms; if it is too large it will not be subject to long range

atmospheric transport and may not bioaccumulate. A third common attribute of the known

Arctic contaminants is a tendency towards cyclic structures. Many contain aromatic rings

and some of the most toxic, persistent and bioaccumulative chemicals such as those

identified by the Stockholm Convention have highly cyclized structures, e.g. mirex and

41

chlordane. Perfluorinated acids are obviously the exception to this rule, but they also have a

higher degree of halogenation. Based on this examination, molecular size, halogenation and

internal connectivity were selected as the criteria for profiling POP-like structures.

5.3 Chemical Profiling Criteria

The criterion of molecular size is quantified as the total number of structural atoms (SA),

where structural atoms are defined as carbon, oxygen, nitrogen, silicon, sulfur and

phosphorous atoms.

SA = Σ C, O, N, Si, S and P atoms (4)

SA does not include hydrogen atoms. SMILES strings do not contain hydrogen atoms in

most cases and determining the valence of each atom and adding hydrogen atoms needlessly

increases the complexity of the parsing algorithms. Furthermore, hydrogen atoms contribute

relatively little mass or size to molecules compared to other atoms, so can be safely

neglected. Halogen atoms are also not included in SA, because otherwise the parameter

indicating the degree of halogenation would be correlated with the molecular size parameter,

which should be avoided. SA only refers to the structural "back-bone" of a molecule, it will

not be strongly correlated with the degree of halogenation, and it will have the same value

for groups of homologous molecules (PCBs for example) which is desirable.

The criterion quantifying the degree of halogenation of a molecule is annotated CX. The

number of halogen-carbon bonds is used to define CX instead of the number of halogen

atoms, because there are a number of salts in the EPISuite database and only halogenated

organics are of interest. To prevent any weak correlation with the size of a molecule (larger

molecules have more possible halogenation sites) CX is normalized to the total number of

non-hydrogen atoms in the molecule (TA):

number of halogen-carbon bonds = CX

TA (5)

42

This also restricts the possible values of CX to between zero and approximately one which

makes the degree of halogenation of large and small molecules more comparable.

The third criterion of internal connectivity could be defined as the number of rings, but

identifying and counting rings again needlessly complicates the SMILES string parsing

algorithms. A more readily obtained measure of internal connectivity is the number of bonds;

for example n-hexane has five carbon-carbon bonds and cyclohexane has six carbon-carbon

bonds. This is a characteristic of the structural back-bone of a molecule so only structural

atoms as defined by Equation (4) and bonds between structural atoms as defined by Equation

(6) will be used.

SB = Σ bonds between C, O, N, Si, S and P atoms (6)

The degree of internal connectivity is defined as the ratio of "excess structural bonds" (the

number of structural bonds minus the number of bonds that would be present in an n-alkane

with the same number of atoms) to the number of structural atoms (XS) as shown in Equation

(7).

[ - ( -1)] =

SB SAXS

SA (7)

As with CX, using a ratio restricts the possible values of XS to between zero and

approximately one.

5.4 Application of Chemical Profiles

All three of the chemical profile criteria SA, CX, and XS were calculated for the 105,584

molecules and statistically analyzed. distribution graphs which show the relative frequency

of values were constructed for five different "populations" of chemicals; the full list of

chemicals, only those chemicals which fall within the area of elevated AC-BAP (only

accumulative), only those chemicals which have an OH thalf value greater than two days

(only persistent), the chemicals which are both accumulative and persistent, and finally the

43

86 known Arctic contaminants. The distribution graphs are Figure 6 to Figure 8. The SA has

a normal distribution only slightly skewed to larger values and so can be adequately

described by the mean and standard deviation of the values. For CX and XS however the

distributions are dominated by an enormous number of chemicals with a CX or XS value of

zero which has a large influence on the mean and standard deviations of the values. For the

full, only accumulative and only persistent populations a more statistically sound method of

analyzing the variation in the population is to use the mode (the most common value, in this

case zero) instead of the mean and calculate the deviation from the mode. This is not valid

for the populations of both accumulative and persistent chemicals or for the known Arctic

contaminants, but in these cases the deviation from zero becomes very similar to the mean of

the population as the distribution of CX and XS values migrate further from zero. The mean

and standard deviation of SA, CX and XS as well as the deviation from zero of CX and XS are

shown in Figure 9. Comparisons of the CX and XS values of the different populations are

primarily based on the mean and standard deviation to make the comparisons easier, but

subsequent calculations are based on the deviation from zero.

44

Figure 6 - Frquency distribution of the number of structural atoms (SA). distributions

are shown for five different chemical "populations"

The full list of chemicals in the EPI Suite database (All), chemicals which fall within the

defined chemical space (Accumulative), chemicals with an OH oxidation half-life of

greater than 2 days (Persistent), chemicals which are both Accumulative and Persistent

(Both), and the list of 86 known Arctic contaminants (Known).

0 20 40 60 80 100 120 140 160

02468

10121416182022

SA

Kn

ow

n

050

100150200250300350

Bo

th

0200400600800

100012001400

Pe

rsis

ten

t

0

500

1000

1500

2000

Accu

mu

lative

01000200030004000500060007000

All

45

Figure 7 - Frequency distribution of the degree of halogenation (CX). distributions are

shown for five different chemical "populations"





0.0 0.2 0.4 0.6 0.8

02468

10

CX

Kn

ow

n

0

100

200

300

400

500

Bo

th

0

2000

4000

6000

8000

Pe

rsis

ten

t

02000400060008000

10000120001400016000

Accu

mu

lative

0

20000

40000

60000

80000

100000

All

46

Figure 8 - Frequency distribution of the degree of internal connectivity (XS).

distributions are shown for five different chemical "populations"





0.0 0.2 0.4 0.6 0.8

02468

1012141618

XS

Kn

ow

n

050

100150200250300

Bo

th

0100020003000400050006000

Pe

rsis

ten

t

0

1000

2000

3000

4000

Accu

mu

lative

05000

1000015000200002500030000

All

47

Figure 9 - Mean and standard deviation of the total number of structural atoms (SA),

degree of halogenation (CX), and degree of internal connectivity (XS) for five different

chemical "populations". The deviation from zero is shown in orange

5.5 Interpretation of Figure 9

The mean number of structural atoms in accumulative molecules is virtually the same as for

the whole population, but the mean value for persistent molecules is significantly lower

(Figure 9). For the molecules which are both accumulative and persistent the mean value of

SA is much closer to the "only persistent" population, and the mean value for known Arctic

contaminants is encouragingly close to this SA value as well. Also notable is the very tight

clustering of the SA values for known Arctic contaminants; the standard deviation is only

2.42. This seems to confirm the conjecture that there is an optimum size for Arctic

contaminants.

0 5 10

15 20 25 30

SA

0.0 0.1 0.2 0.3 0.4 0.5

CX

0.0

0.1

0.2

0.3

Population

XS

Full Population

Only Accumulative

Only Persistent

Accumulative & Persistent

Known Contaminants

48

Figure 9 shows that the pattern of the mean values is closely approximated by the deviations

from zero. Figure 7 shows that while most chemicals have a CX value of zero, there is a large

amount of variation in the degree of halogenation for those chemicals with CX greater than

zero. For the full population as well as for the accumulative and the persistent sub-

populations a pattern similar to SA is observed; the accumulative population has a mean CX

value similar to the entire population, but the value for persistent molecules is significantly

different, much higher in this case. For molecules which are both accumulative and

persistent CX is much higher again, and for the known Arctic contaminants the mean CX is a

full order of magnitude higher than the mean of the full population. The deviation from zero

increases by a factor of three from the full population to the known Arctic contaminants.

Much higher CX values for the known Arctic contaminants suggest that halogenation

contributes significantly to both persistence and bioaccumulation of individual molecules,

however regarding the population of molecules as a whole, halogenation contributes more to

persistence. Another notable piece of information is that the magnitude of the standard

deviation of the known contaminants is comparable to the standard deviation of the whole

population; this implies that while there is not an optimum degree of halogenation,

increasing the amount of halogenation increases the persistence and accumulation of a

molecule.

Variations in the mean values of XS are more complex than SA and CX. The distribution of

XS is dominated by a large number of chemicals with no internal connectivity (Figure 8), but

Figure 9 shows that the deviation from zero is a good approximation of the mean for all

populations. Looking only at accumulative molecules slightly increases the mean XS value,

but for the persistent molecules the mean XS value slightly decreases. However for

molecules which are both persistent and accumulative the mean XS value is much higher,

and is higher again for known Arctic contaminants. There are most likely a number of factors

affecting the results, for example methyl groups are more resistant to OH oxidation than

secondary or tertiary carbons69

which may account for the decrease in XS for the persistent

sub-population. There are many molecules containing aromatic rings in the sub-population

which is both accumulative and persistent, so it appears that increasing the degree of

aromaticity may increase the persistence of a molecule and this in turn increases the mean

value of XS in this specific sub-population of accumulative molecules. However, there are

many molecules in the list of known Arctic contaminants which contain non-aromatic rings

so it could also be concluded that increasing the degree of internal connectivity of any kind

contributes to the potential to become an Arctic contaminant. The magnitude of the standard

49

deviation of XS is similar for all populations and slightly increases for the list of known

Arctic contaminants so it does not appear that there is an optimal value. Clearly there is a

finite limit to how many internal rings there can be in a stable molecule; for a group of

molecules with the same number of atoms there will be an optimal number of internal

connections which maximizes stability. However it must be remembered that the molecules

in EPI Suite database are chemicals in commercial use and a tendency to more stable

molecules must exist. Given these assumptions we conclude that the optimal degree of

internal connectivity is not generally exceeded in our population of molecules, so increasing

the XS value tends to increase the bioaccumulation of a molecule, and persistence is also

often increased due to the correlation of XS with the number of aromatic rings.

5.6 Definition of the POP score

The most important conclusions of the previous analysis are that SA has an optimal value,

indicated by the small standard deviation of SA for the population of known Arctic

contaminants (Figure 9), and that increases in CX and XS make a molecule more persistent

and bioaccumulative. The latter is obvious from greatly elevated mean CX and XS values for

known Arctic contaminants (Figure 9). One difficulty remaining is how to weight the

contribution of each of these three criteria to the POP score. As stated before this cannot be

done using multiple linear regression so we instead look at the deviations from a hypothetical

"typical" molecule. The first restrictions we make are that the hypothetical "typical"

molecule must have a lower POP score than a molecule fitting the profile of known Arctic

contaminants. We define the contribution of CX and XS to the POP score as the number of

standard deviations above the mode of the full population. The CX or XS value of a molecule

is divided by the standard deviation of the full population. This means that a "typical"

molecule with the most common CX and XS values of zero would have a POP score of zero

if only these two criteria were used. The contribution of SA to the POP score is defined as the

negative of the absolute number of standard deviations from the optimal value, meaning that

deviations from the optimal value of SA decrease the POP score. The mean value of SA for

known Arctic contaminants is assumed to be the optimal value, and is subtracted from the SA

of a molecule and then the absolute value is divided by the standard deviation of SA for

known Arctic contaminants. The complete calculation of the POP score is shown in Equation

(8).

50

| - 11.43|POP score = + -

0.1039 0.1067 2.420

CX XS SA (8)

5.7 Interpretation of the POP Score

When interpreting the POP scores some discretion must be exercised. As was stated

previously there is no universal scale of susceptibility to becoming an Arctic contaminant,

and the POP score is not nearly sophisticated enough to be interpreted as such. When

comparing molecules the one with a higher POP score is not necessarily more likely to

accumulate in Arctic biota. Qualitative data was used to create the POP score, therefore only

qualitative data can be provided by it. Interpretation of the score should keep in mind the

molecules that were used to parameterize it. All 86 known Arctic contaminants are

halogenated and the lowest contribution that CX makes to the POP score is 1.809, so any

chemical with a CX contribution to the POP score of less than 1.8 should be excluded. For all

but two of the 86 known Arctic contaminants the CX makes a larger contribution to the POP

score than XS (Table 2), and these two chemicals (dieldrin and endrin) have an unusually

high level of cyclization, a likely explanation for this is that halogenation stabilizes the

highly cyclized structures. Based on these observations there does not seem to be a need to

restrict the interpretation of the POP score based on the XS value. Finally, the number of

structural atoms for the list of known Arctic contaminants ranges between six and fourteen

so any molecule too far outside these bounds should be regarded with skepticism; a molecule

one standard deviation below 6 (SA <= 3) or one standard deviation above 14 (SA >= 17)

should not be considered a potential POP.

The mean POP score calculated for the 86 known Arctic contaminants was 4.447 ± 1.122.

The lower limit of the 95% prediction interval of the POP score is 2.57 and the lowest POP

score for a known Arctic contaminant is 2.416 (DDE). To ensure that all possible Arctic

contaminants are captured by the POP score all chemicals with a score of 2.4 or greater,

while also meeting the SA and CX requirements previously mentioned, are defined as

matching the profile of known Arctic contaminants.

Although the POP score should not be interpreted quantitatively, it is worthwhile to look at

which chemicals have particularly high POP scores. Any molecule with a POP score greater

than the mean score of known Arctic contaminants (4.447) was defined as a chemical which

51

has an above average likelihood of being an Arctic contaminant. There are 79 chemicals out

the 822 chemicals which are both persistent and fall within the area of elevated AC-BAP

which exceed this cut-off (identified in Table 3). Among these are known Arctic

contaminants such as mirex, toxaphenes, chlordanes, hexachlorobiphenyls, and a number of

perfluorinated organic acids, as well as variations of known Arctic contaminants, such as

some highly fluorinated biphenyls. Included are also a series of chlorinated cyclopentanes,

pentenes and pentadienes, a number of highly halogenated benzimidazoles, several triazines

substituted with trichloromethyl or trichloroethyl groups, two chlorinated PAHs and a

number of different halogenated bridged diphenyls.

52

Table 3 - CAS number, chemical class, name and POP score of chemicals meeting the

elevated AC-BAP, atmospheric oxidation half-life, and POP score criteria.

A “yes” in the column entitled “Hi POP” implies that the chemical has a POP score

greater than 4.447.

CAS CLASS NAME HI-POP SCORE

82688 CUP benzene, pentachloronitro- 3.475 87865 CUP phenol, pentachloro- 3.519 88824 CUP benzoic acid, 2,3,5-triiodo- 2.443

117180 CUP 2,3,5,6-tetrachloronitrobenzene 2.999 634662 CUP benzene, 1,2,3,4-tetrachloro- 3.168

2678219 CUP OLPISAN 2.470 6379460 CUP Vancide PB 2.470 7159344 CUP Pyroxychlor 2.999 8003461 CUP trichlorodinitrobenzene 2.470 25167833 CUP phenol, tetrachloro- 3.008 58138082 CUP Oxirane, 2-(3,5-dichlorophenyl)-2-(2,2,2-trichloroethyl)- yes 4.534 297789 OP Isobenzan yes 6.498

1344327 OP benzene, trichloro(chloromethyl)- 3.008 1715408 OP bromociclen yes 4.776 2550756 OP HERCULES426 yes 5.608 3495429 OP chlorquinox 4.033 3513937 OP Hercules 426D yes 5.738 41318756 PBDE 2,4,4'-tribromodiphenyl ether 2.598 49690940 PBDE tribromodiphenyl oxide 2.598 1336363 PCB PCBs 3.733 2437798 PCB 2,2',4,4'-PCB 3.733 2542292 PCB pentachlorobiphenyl 4.157 7012375 PCB 2,4,4'-PCB 3.251 11097691 PCB AROCLOR 1254 4.157 11126424 PCB AROCLOR 5460 3.733 12672296 PCB AROCLOR 1248 3.733 12674112 PCB Aroclor 1016 3.251 15862074 PCB 2,4,5-PCB 3.251 15968055 PCB 1,1'-biphenyl, 2,2',6,6'-tetrachloro- 3.733 16606023 PCB 2,4',5-PCB 3.251 18259057 PCB 2,3,4,5,6-PCB 4.157 25323686 PCB trichlorobiphenyl 3.251 25429292 PCB pentachlorobiphenyl 4.157 26914330 PCB tetrachlorobiphenyl 3.733 31508006 PCB 2',3,4,4',5'-PCB 4.157 32598100 PCB 2,4,3',4'-PCB 3.733 32598111 PCB 2,3',4',5-PCB 3.733 32598122 PCB 1,1'-biphenyl, 2,4,4',6-tetrachloro- 3.733 32598133 PCB 3,3',4,4'-PCB 3.733 32598144 PCB 1,1'-biphenyl, 2,3,3',4,4'-pentachloro- 4.157 32690930 PCB 2,4,4',5-PCB 3.733 33025411 PCB 2,3,4,4'-PCB 3.733 33284525 PCB 3,3',5,5'-PCB 3.733 33284536 PCB 2,3,4,5-PCB 3.733 33284547 PCB 2,3,5,6-PCB 3.733 33979032 PCB 2,2',4,4',6,6'-PCB yes 4.535 35693926 PCB 2,4,6-PCB 3.251 35693993 PCB 2,2',5,5'-PCB 3.733 36559225 PCB 1,1'-biphenyl, 2,2',3,4'-tetrachloro- 3.733 37680652 PCB 2,2',5-PCB 3.251 37680663 PCB 2,2',4-PCB 3.251 37680685 PCB 1,1'-biphenyl, 2,3',5'-trichloro- 3.251 37680696 PCB 3,3',4-trichlorobiphenyl 3.251 37680732 PCB 2,4,5,2',5'-PCB 4.157 38379996 PCB 2,2',3,5',6-PCB 4.157 38380028 PCB 2,3,4,2',5'-PCB 4.157

53

38380039 PCB 2,3,3',4',6-PCB 4.157 38380040 PCB 2,2',3,4',5',6-PCB yes 4.535 38411222 PCB 2,3,6,2',3',6'-PCB yes 4.535 38444734 PCB 2,2',6-PCB 3.251 38444767 PCB 2,3',6-trichlorobiphenyl 3.251 38444778 PCB 2,4',6-PCB 3.251 38444789 PCB 2,2',3'-PCB 3.251 38444814 PCB 2,3',5-PCB 3.251 38444847 PCB 2,3,3'-PCB 3.251 38444858 PCB 2,3,4'-PCB 3.251 38444869 PCB 2',3,4-PCB 3.251 38444870 PCB 3,3',5-PCB 3.251 38444881 PCB 3,4',5-trichlorobiphenyl 3.251 38444905 PCB 3,4,4'-PCB 3.251 38444938 PCB 2,2',3,3'-PCB 3.733 39485831 PCB 1,1'-biphenyl, 2,2',4,4',6-pentachloro- 4.157 39635331 PCB 3,3',4,5,5'-pentachlorobiphenyl 4.157 41464395 PCB 2,2',3,5'-PCB 3.733 41464408 PCB 2,2',4,5'-PCB 3.733 41464419 PCB 2,2',5,6'-tetrachlorobiphenyl 3.733 41464420 PCB 1,1'-biphenyl, 2,3',5,5'-tetrachloro- 3.733 41464431 PCB 1,1'-biphenyl, 2,3,3',4'-tetrachloro- 3.733 41464464 PCB 1,1'-biphenyl, 2,3',4',6-tetrachloro- 3.733 41464475 PCB 1,1'-biphenyl, 2,2',3,6'-tetrachloro- 3.733 41464486 PCB 1,1'-biphenyl, 3,3',4,5'-tetrachloro- 3.733 41464497 PCB 1,1'-biphenyl, 2,3,3',5'-tetrachloro- 3.733 41464511 PCB 2,2',3',4,5-PCB 4.157 52663588 PCB 1,1'-biphenyl, 2,3,4',6-tetrachloro- 3.733 52663599 PCB 2,2',3,4-PCB 3.733 52663602 PCB 2,2',3,3',6-PCB 4.157 52663613 PCB 2,2',3,5,5'-PCB 4.157 52663624 PCB 2,2',3,3',4-pentachlorobiphenyl 4.157 53469219 PCB Aroclor 1242 3.733 53555661 PCB 3,4,5-trichlorobiphenyl 3.251 54230227 PCB 2,3,4,6-tetrachlorobiphenyl 3.733 55215173 PCB 2,2',3,4,6-pentachlorobiphenyl 4.157 55312691 PCB 2,2',3,4,5-pentachlorobiphenyl 4.157 55702459 PCB 2,3,6-PCB 3.251 55702460 PCB 1,1'-biphenyl, 2,3,4-trichloro- 3.251 55712373 PCB 1,1'-biphenyl, 2,3',4-trichloro- 3.251 55720440 PCB 2,3,5-trichlorobiphenyl 3.251 56558168 PCB 2,2',4,6,6'-pentachlorobiphenyl 4.157 56558179 PCB 1,1'-biphenyl, 2,3',4,4',6-pentachloro- 4.157 56558180 PCB 1,1'-biphenyl, 2,3',4,5',6-pentachloro- 4.157 57465288 PCB 3,3',4,4',5-pentachlorobiphenyl 4.157 60145202 PCB 2,2',3,3',5-pentachlorobiphenyl 4.157 60145213 PCB 1,1'-biphenyl, 2,2',4,5',6-pentachloro- 4.157 60233241 PCB 2,3',4,6-tetrachlorobiphenyl 3.733 60233252 PCB 1,1'-biphenyl, 2,2',3,4',6'-pentachloro- 4.157 62796650 PCB 1,1'-biphenyl, 2,2',4,6-tetrachloro- 3.733 65510443 PCB 2,3,4,4',5-pentachlorobiphenyl 4.157 65510454 PCB 2,2',3,4,4'-PCB 4.157 68194047 PCB 1,1'-biphenyl, 2,2',4,6'-tetrachloro- 3.733 68194058 PCB 2,3,6,2',4'-PCB 4.157 68194069 PCB 2,2',4,5,6'-pentachlorobiphenyl 4.157 68194070 PCB 1,1'-biphenyl, 2,2',3,4',5-pentachloro- 4.157 68194105 PCB 1,1'-biphenyl, 2,3,3',5',6-pentachloro- 4.157 68194116 PCB 1,1'-biphenyl, 2,3,4',5,6-pentachloro- 4.157 68194127 PCB 2,4,5,3',5'-PCB 4.157 70362413 PCB 2,3,3',4,5'-pentachlorobiphenyl 4.157 70362457 PCB 1,1'-biphenyl, 2,2',3,6-tetrachloro- 3.733 70362468 PCB 1,1'-biphenyl, 2,2',3,5-tetrachloro- 3.733 70362479 PCB 2,2',4,5-tetrachlorobiphenyl 3.733 70362480 PCB 2,3',4',5-tetrachlorobiphenyl 3.733 70362491 PCB 3,3',4,5-tetrachlorobiphenyl 3.733 70362504 PCB 3,4,4',5-tetrachlorobiphenyl 3.733 70424678 PCB 2,3,3',5-tetrachlorobiphenyl 3.733 70424689 PCB 2,3,3',4',5-pentachlorobiphenyl 4.157 70424690 PCB 2,3,3',4,5-pentachlorobiphenyl 4.157 70424703 PCB 2',3,4,5,5'-pentachlorobiphenyl 4.157

54

73575527 PCB 2,3',4,5'-tetrachlorobiphenyl 3.733 73575538 PCB 1,1'-biphenyl, 2,3',4,5-tetrachloro- 3.733 73575549 PCB 2,2',3,6,6'-pentachlorobiphenyl 4.157 73575550 PCB 2,2',3,5,6'-pentachlorobiphenyl 4.157 73575561 PCB 1,1'-biphenyl, 2,2',3,5,6-pentachloro- 4.157 73575572 PCB 1,1'-biphenyl, 2,2',3,4,6'-pentachloro- 4.157 74338231 PCB 2,3',5',6-tetrachlorobiphenyl 3.733 74338242 PCB 1,1'-biphenyl, 2,3,3',4-tetrachloro- 3.733 74472336 PCB 2,3,3',6-tetrachlorobiphenyl 3.733 74472347 PCB 1,1'-biphenyl, 2,3,4',5-tetrachloro- 3.733 74472358 PCB 2,3,3',4,6-pentachlorobiphenyl 4.157 74472369 PCB 1,1'-biphenyl, 2,3,3',5,6-pentachloro- 4.157 74472370 PCB 2,3,4,4',5-pentachlorobiphenyl 4.157 74472381 PCB 1,1'-biphenyl, 2,3,4,4',6-pentachloro- 4.157 74472392 PCB 2,3',4',5',6-pentachlorobiphenyl 4.157 76842074 PCB 2',3,3',4,5-pentachlorobiphenyl 4.157

50748 PCBD 2,3,4,5-tetrachlorobenzoic acid 2.999 51398 PCBD benzoic acid, 3,4,5-trichloro- 2.443 58902 PCBD phenol, 2,3,4,6-tetrachloro- 3.008 82622 PCBD 3,4,6-trichloro-2-nitrophenol 2.567 87876 PCBD 1,4-benzenediol, 2,3,5,6-tetrachloro- 2.962 95943 PCBD benzene, 1,2,4,5-tetrachloro- 3.168

117975 PCBD benzenethiol, pentachloro-, zinc salt 3.210 133493 PCBD benzenethiol, pentachloro- 3.519 484673 PCBD phenol, 2,3,5,6-tetrachloro-4-methoxy- 2.999 527208 PCBD pentachloroaniline 3.519 606075 PCBD pentachloroethylbenzene 3.456 608935 PCBD benzene, pentachloro- 3.693 634833 PCBD 2,3,4,5-tetrachloroaniline 3.008 634902 PCBD benzene, 1,2,3,5-tetrachloro- 3.168 781157 PCBD benzene, 1,2,3,4-tetrachloro-5,6-dinitro- (6CI,7CI,8CI,9CI) 2.952 875401 PCBD benzene, 1,2,3,5-tetrachloro-4-methyl- (9CI) 3.008 877098 PCBD benzene, 1,2,3,5-tetrachloro-4,6-dimethyl- (9CI) 2.962 877101 PCBD 1,2,4,5-tetrachloro-3.6-dimethylbenzene 2.962 877112 PCBD pentachlorotoluene 3.519 879390 PCBD 1,2,3,4-tetrachloro-5-nitrobenzene 2.999 935955 PCBD phenol, 2,3,6-trichloro- 3.008 938227 PCBD benzene, 1,2,3,5-tetrachloro-4-methoxy- 2.962 938863 PCBD benzene, 1,2,3,4-tetrachloro-5-methoxy- 2.962 944616 PCBD benzene, 1,2,3,4-tetrachloro-5,6-dimethoxy- 3.096 944785 PCBD 1,4-dimethoxytetrachlorobenzene 3.096 994785 PCBD tetrachloro-1,4-dimethoxybenzene 3.096

1006311 PCBD 2,3,5,6-tetrachlorotoluene 3.008 1006322 PCBD 2,3,4,5-tetrachlorotoluene 3.008 1012846 PCBD benzoic acid, pentachloro- 3.475 1198556 PCBD 1,2-benzenediol, 3,4,5,6-tetrachloro- 2.962 1441027 PCBD pentachlorophenyl acetate 3.554 1825190 PCBD pentachloro(methylthio)benzene 3.456 1825214 PCBD pentachloroanisole 3.456 1897412 PCBD 1,4-benzenedicarbonitrile, 2,3,5,6-tetrachloro- 3.096 1953997 PCBD 1,2-benzenedicarbonitrile, 3,4,5,6-tetrachloro- 3.096 2438882 PCBD tetrachloronitroanisole 3.241 2539175 PCBD 2-methoxytetrachlorophenol 2.999 2668248 PCBD 4,5,6-trichloroguaiacol 2.443 2877147 PCBD pentachlorophenoxyecetic acid 3.682 3215654 PCBD (2,3,6-trichlorophenyl)-acetonitrile 2.443 3481207 PCBD 2,3,5,6-tetrachloroaniline 3.008 3714623 PCBD 2,3,4,6-tetrachloronitrobenzene 2.999 4824720 PCBD 4-nitro-2,3,5,6-tetrachlorophenol 3.096 4901513 PCBD 2,3,4,5-tetrachlorophenol 3.008 5324685 PCBD benzene, 1,3,5-trichloro-2,4,6-trimethyl- 2.443 6284839 PCBD benzene, 1,3,5-trichloro-2,4-dinitro- 2.470 6936409 PCBD benzene, 1,2,4,5-tetrachloro-3-methoxy- 2.962 7476826 PCBD Butanoic acid, 2,3,4,6-tetrachlorophenyl ester (9CI) 2.952 10460330 PCBD phenol, 2,3,4,6-tetrachloro-5-methyl- 2.962 12408105 PCBD tetrachlorobenzeneS 3.168 13061284 PCBD ethanone, 1-(2,4,5-trichlorophenyl)- (9CI) 2.443 13608872 PCBD 2,3,4-trichlor-acetophenon 2.443 13801508 PCBD 1,2-benzenedithiol, 3,4,5,6-tetrachloro- 2.962 16022698 PCBD benzenemethanol, 2,3,4,5,6-pentachloro- 3.456

55

16766293 PCBD benzene, 1,2,3-trichloro-4,5-dimethoxy- 2.567 17700093 PCBD 2,3,4-trichloronitrobenzene 2.443 18708708 PCBD benzene, 1,3,5-trichloro-2-nitro- 2.443 19303889 PCBD phenol, 2,3,4,5-tetrachloro-6-methyl- (9CI) 2.962 20098388 PCBD benzene, 1,2,4,5-tetrachloro-3,6-dinitro- 2.952 20098480 PCBD benzene, 1,2,3-trichloro-5-nitro- 2.443 20404028 PCBD 2,3,6-trichloro-4-nitrophenol 2.567 20925853 PCBD benzonitrile, pentachloro- 3.456 23399908 PCBD 2,4,6-trichlorophenylacetate 2.567 27735644 PCBD phenol, pentachloro-, sodium salt, monohydrate (8CI,9CI) 3.519 27864137 PCBD benzene, 1,2,4-trichloro-3-nitro- (6CI,8CI,9CI) 2.443 29450633 PCBD benzaldehyde, pentachloro-, oxime (8CI,9CI) 3.475 29733708 PCBD tetrachloromethylbenzene 3.008 33715627 PCBD 3,4,5-trichloroacetanilide 2.567 42138727 PCBD benzenamine, 2,3,4,5-tetrachloro-6-methoxy- 2.999 53014418 PCBD benzene, tetrachloro(methylthio) 2.962 53452816 PCBD tetrachloroanisoles 2.962 54965707 PCBD benzene, 1,3,5-trichloro-2-(1-methylethyl)- 2.443 55538733 PCBD phenol, 3,4,6-trichloro-2-propyl- 2.567 56680650 PCBD phenol, 2,3,5,6-tetrachloro-4-methyl- (9CI) 2.962 57057837 PCBD phenol, 3,4,5-trichloro-2-methoxy- 2.443 61465790 PCBD 2,3,6-trichlorocumene 2.443 61465814 PCBD trichlorodimethoxybenzene 2.567 68671909 PCBD benzene, 1,2,4,5-tetrachloro-3-(methylthio)- 2.962 69576803 PCBD benzene, 1,2,3,4-tetrachloro-5-methoxy-6-nitro- 3.241 69911611 PCBD pentachlorotoluene 3.519 70439962 PCBD benzenamine, 2,3,5,6-tetrachloro-4-methoxy- 2.999 70833635 PCBD ar,ar,ar-trichlorobenzeneacetonitrile 2.443 74313001 PCBD benzene, 1-methoxy-4-methyl-, trichloro DERIV. 2.443 81686422 PCBD 2,3,6-trichloro-p-cymene 2.567 81686433 PCBD 2,3,5,6-tetrachloro-p-cymene 3.096 85298073 PCBD 1,2,5-trichloro-3,4-dimethoxybenzene 2.567

107409529 PCBD 1,2,4,5-tetrachloro-3-(methylsulfinyl) 3.096 1746016 PCDD 2,3,7,8-TCDD 3.085 30746588 PCDD 1,2,3,4-tetrachlorodibenzo-p-dioxin 3.085 33423926 PCDD 1,3,6,8-tetrachlorodibenzo-p-dioxin 3.085 36088229 PCDD pentachlorodibenzo-p-dioxin 3.479 40581940 PCDD 1,4,7,8-tetrachlorodibenzo-p-dioxin 3.085 62470535 PCDD 1,3,7,9-tetrachlorodibenzo-p-dioxin 3.085 67028197 PCDD 1,2,3,4,6-pentachlorodibenzo-p-dioxin 3.479 71669255 PCDD 1,2,3,6-tetrachlorodibenzo-p-dioxin 3.085 71669266 PCDD 1,2,3,9-tetrachlorodibenzo-p-dioxin 3.085 22274426 PCDE 2,3,4,5,6-pentachlorodiphenylether 3.467 24910734 PCDE 3,4',5-trichlorodiphenyl ether 2.598 28076735 PCDE 2,2'4,4'-tetrachlorodiphenyloxide 3.058 31242930 PCDE hexachlorodiphenyl oxide 3.832 31242941 PCDE tetrachlorodiphenyl ether 3.058 42279298 PCDE pentachlorophenyl ether 3.467 52322802 PCDE benzene, 1,2,4-trichloro-5-phenoxy- 2.598 56348722 PCDE 3,3',4,4'-tetrachlorobiphenyl ether 3.058 57321638 PCDE trichlorophenyl ether 3.832 59039213 PCDE 2,4,4'-trichlorodiphenyl ether 2.598 60123640 PCDE 2,2',4,4',5-pentachlor-diphenylether 3.467 60123651 PCDE 2,3',4,4',5-pentachlorodiphenyl ether 3.467 61328447 PCDE 1,2-dichloro-4-(2-chlorophenoxy)benzene 2.598 61328458 PCDE 2,4,4',5-tetrachlorodiphenyl ether 3.058 63553300 PCDE 2,4,4',6-tetrachlorodiphenyl ether 3.058 63646515 PCDE 3,4,4'-trichlorodiphenyl ether 2.598 63646559 PCDE 2,3,4',5,6-pentachlorodiphenylether 3.467 65075005 PCDE 2,4',5-trichlorodiphenyl ether 2.598 65075016 PCDE 2,3,4,4'-tetrachlorodiphenyl ether 3.058 66794603 PCDE 3,3',4-trichlorodiphenyl ether 2.598 68914976 PCDE 2,2',4-trichlorodiphenyl ether 2.598 71585370 PCDE 2,2',3,4,4'-pentachlorodiphenylether 3.467 71585381 PCDE 2,2',3,4,4',5-hexachlorodiphenylether 3.832 71585392 PCDE 2,2',3,3',4,4'-hexachlorodiphenylether 3.832 71859308 PCDE 2,2',4,4',5,5'-hexachlorodiphenylether 3.832 85918316 PCDE 2,3,3',4,4'-pentachlorodiphenylether 3.467 85918338 PCDE 2,3,4,6-tetrachlorodiphenyl ether 3.058 85918350 PCDE 2,2',3,4,6'-pentachlorodiphenylether 3.467

56

85918372 PCDE 2,2',3,4',5',6-hexachlorodiphenylether 3.832 94339590 PCDE 3,3',4,4',5-pentachlorodiphenylether 3.467

104294168 PCDE 2,2',4,4',6-pentachlorodiphenylether 3.467 106220819 PCDE 2,2',4,4',5,6'-hexachlorodiphenylether 3.832 113464178 PCDE 2,3,4,4',5-pentachlorodiphenylether 3.467 116995201 PCDE 2,2',3,4',6-pentachlorodiphenylether 3.467 124076660 PCDE 2,2',3,3',4,6'-hexachlorodiphenylether 3.832 130892669 PCDE 2,3',4',6-tetrachlorodiphenyl ether 3.058 130892670 PCDE 2,2',4,5,6'-pentachlorodiphenylether 3.467 131138211 PCDE 2,2',4,5,5'-pentachlorodiphenylether 3.467 147102634 PCDE 2,2',3,4'-tetrachlorodiphenyl ether 3.058 147102645 PCDE 2,3',4,5'-tetrachlorodiphenyl ether 3.058 155999921 PCDE 2,2',4,5'-tetrachlorodiphenyl ether 3.058 155999932 PCDE 2,3',4-trichlorodiphenyl ether 2.598 155999976 PCDE 2,3,3',4',5,6-hexachlorodiphenylether 3.832 157683711 PCDE 2,3,4'-trichlorodiphenyl ether 2.598 157683722 PCDE 2,4',6-trichlorodiphenyl ether 2.598 157683733 PCDE 2,2',3,4',5-pentachlorodiphenylether 3.467 157683744 PCDE 2,3',4,4',6-pentachlorodiphenylether 3.467 159553681 PCDE 2,3,3',4',5-pentachlorodiphenylether 3.467 159553692 PCDE 2,3,3',4',6-pentachlorodiphenylether 3.467 160282042 PCDE 2,3',4,5,5'-pentachlorodiphenylether 3.467 160282053 PCDE 2,3,4,4',6-pentachlorodiphenylether 3.467 160282075 PCDE 2,3,3',4,5'-pentachlorodiphenylether 3.467 160282086 PCDE 2,2',3',4,5-pentachlorodiphenylether 3.467 160282097 PCDE 2,2',3,4,5'-pentachlorodiphenylether 3.467 160282100 PCDE 2,2',3,3',4-pentachlorodiphenylether 3.467 162853249 PCDE 2,3,5-trichlorodiphenyl ether 2.598 162853250 PCDE 2,3,6-trichlorodiphenyl ether 2.598 162853261 PCDE 2,2',4,5-tetrachlorodiphenyl ether 3.058 162853272 PCDE 2,3,3',4'-tetrachlorodiphenyl ether 3.058 162853283 PCDE 2,2',3,4',5,5'-hexachlorodiphenylether 3.832 24478726 PCDF dibenzofuran, 1,2,3,4-tetrachloro- 3.779 24478737 PCDF 1,2,4-trichlorodibenzofuran 3.319 30402143 PCDF 1,2,3,4-tetrachlorodibenzofuran 3.779 30402154 PCDF pentachlorodibenzofuran 4.188 43048006 PCDF trichlorodibenzofuran 3.319 51207319 PCDF 2,3,7,8-tetrachlorodibenzofuran 3.779 54589718 PCDF 2,4,8-trichlorodibenzofuran 3.319 55722275 PCDF tetrachlorodibenzofuran 3.779 57117325 PCDF 2,3,8-trichlorodibenzofuran 3.319 57117336 PCDF 2,3,6-trichlorodibenzofuran 3.319 57117347 PCDF 2,3,4-trichlorodibenzofuran 3.319 57117358 PCDF 1,3,7,8-tetrachlorodibenzofuran 3.779 57117369 PCDF 1,3,6,7-tetrachlorodibenzofuran 3.779 57117370 PCDF dibenzofuran, 2,3,6,8-tetrachloro- 3.779 57117381 PCDF dibenzofuran, 2,4,6,7-tetrachloro- 3.779 57117392 PCDF dibenzofuran, 2,3,6,7-tetrachloro- 3.779 57117405 PCDF 3,4,6,7-tetrachlorodibenzofuran 3.779 57117427 PCDF dibenzofuran, 1,2,3,6,7-pentachloro- 4.188 57117438 PCDF 2,3,4,6,7-pentachlorodibenzofuran 4.188 58802145 PCDF 2,4,6-trichlorodibenzofuran 3.319 58802156 PCDF 1,2,4,7,8-pentachlorodibenzofuran 4.188 58802167 PCDF 1,3,4,7,8-pentachlorodibenzofuran 4.188 58802178 PCDF 2,3,7-trichlorodibenzofuran 3.319 58802189 PCDF 1,7,8-trichlorodibenzofuran 3.319 58802190 PCDF 2,4,6,8-tetrachlorodibenzofuran 3.779 58802203 PCDF dibenzofuran, 1,2,7,8-tetrachloro- 3.779 62615081 PCDF 1,2,3,8-tetrachlorodibenzofuran 3.779 64126870 PCDF 1,2,4,8-tetrachlorodibenzofuran 3.779 64560141 PCDF 1,4,8-trichlorodibenzofuran 3.319 64560152 PCDF 1,2,6-trichlorodibenzofuran 3.319 64560163 PCDF 1,3,7-trichlorodibenzofuran 3.319 64560174 PCDF dibenzofuran, 1,3,7,9-tetrachloro- 3.779 66794590 PCDF 1,4,6,7-tetrachlorodibenzofuran 3.779 67481225 PCDF dibenzofuran, 2,3,4,6,8-pentachloro- 4.188 69433007 PCDF 1,2,6,7,8-pentachlorodibenzofuran 4.188 69698573 PCDF 1,2,4,6,8-pentachlorodibenzofuran 4.188 70648134 PCDF 1,4,9-trichlorodibenzofuran 3.319 70648156 PCDF 1,3,4,6,9-pentachlorodibenzofuran 4.188

57

70648167 PCDF dibenzofuran, 1,3,4,7-tetrachloro- 3.779 70648189 PCDF 1,2,6,9-tetrachlorodibenzofuran 3.779 70648190 PCDF 1,4,6,9-tetrachlorodibenzofuran 3.779 70648203 PCDF 1,3,4,7,9-pentachlorodibenzofuran 4.188 70648214 PCDF 1,3,6,7,8-pentachlorodibenzofuran 4.188 70648225 PCDF 1,2,8,9-tetrachlorodibenzofuran 3.779 70648236 PCDF 1,2,4,8,9-pentachlorodibenzofuran 4.188 70648247 PCDF dibenzofuran, 1,2,4,6,9-pentachloro- 4.188 70872821 PCDF dibenzofuran, 1,2,6,7,9-pentachloro- 4.188 71998726 PCDF dibenzofuran, 1,3,6,8-tetrachloro- 3.779 71998737 PCDF 1,2,4,6-tetrachlorodibenzofuran 3.779 71998748 PCDF 1,2,4,7,9-pentachlorodibenzofuran 4.188 76621120 PCDF 1,3,8-trichlorodibenzofuran 3.319 82911588 PCDF 1,4,6,8-tetrachlorodibenzofuran 3.779 82911599 PCDF 1,6,8-trichlorodibenzofuran 3.319 82911602 PCDF 1,4,6-trichlorodibenzofuran 3.319 82911613 PCDF 1,3,4-trichlorodibenzofuran 3.319 83636479 PCDF 1,2,3-trichlorodibenzofuran 3.319 83690986 PCDF 1,3,6,9-tetrachlorodibenzofuran 3.779 83704216 PCDF 1,2,3,6-tetrachlorodibenzofuran 3.779 83704227 PCDF 1,2,3,7-tetrachlorodibenzofuran 3.779 83704238 PCDF 1,2,3,9-tetrachlorodibenzofuran 3.779 83704249 PCDF 1,2,4,9-tetrachlorodibenzofuran 3.779 83704250 PCDF 1,2,6,7-tetrachlorodibenzofuran 3.779 83704261 PCDF 1,2,7,9-tetrachlorodibenzofuran 3.779 83704272 PCDF dibenzofuran, 1,3,4,6-tetrachloro- 3.779 83704283 PCDF 1,3,4,9-tetrachlorodibenzofuran 3.779 83704294 PCDF 1,4,7-8-tetrachlorodibenzofuran 3.779 83704307 PCDF dibenzofuran, 2,3,4,6-tetrachloro- 3.779 83704318 PCDF 2,3,4,7-tetrachlorodibenzofuran 3.779 83704329 PCDF 2,3,4,8-tetrachlorodibenzofuran 3.779 83704330 PCDF 1,6,7,8-tetrachlorodibenzofuran 3.779 83704341 PCDF 1,2,8-trichlorodibenzofuran 3.319 83704363 PCDF 1,3,4,6,7-pentachlorodibenzofuran 4.188 83704374 PCDF 1,2,7-trichlorodibenzofuran 3.319 83704385 PCDF 1,2,9-trichlorodibenzofuran 3.319 83704396 PCDF 1,3,6-trichlorodibenzofuran 3.319 83704409 PCDF 1,3,9-trichlorodibenzofuran 3.319 83704410 PCDF 1,4,7-trichlorodibenzofuran 3.319 83704421 PCDF 2,4,7-trichlorodibenzofuran 3.319 83704432 PCDF 3,4,6-trichlorodibenzofuran 3.319 83704443 PCDF 3,4,7-trichlorodibenzofuran 3.319 83704454 PCDF 2,6,7-trichlorodibenzofuran 3.319 83704465 PCDF 1,6,7-trichlorodibenzofuran 3.319 83704501 PCDF 1,2,4,6,7-pentachlorodibenzofuran 4.188 83704512 PCDF 1,2,3,6,8-pentachlorodibenzofuran 4.188 83704534 PCDF 1,2,3,7,9-pentachlorodibenzofuran 4.188 83704545 PCDF 1,2,3,8,9-pentachlorodibenzofuran 4.188 83704556 PCDF 1,3,4,6,8-pentachlorodibenzofuran 4.188 83710070 PCDF 1,2,6,8-tetrachlorodibenzofuran 3.779 83719408 PCDF 1,2,4,7-tetrachlorodibenzofuran 3.779 92341043 PCDF 1,3,4,8-tetrachlorodibenzofuran 3.779 1321648 PCN naphthalene, pentachloro- yes 4.492 1321659 PCN naphthalene, trichloro- 3.505 1335871 PCN naphthalene, hexachloro- yes 4.893 1335882 PCN naphthalene, tetrachloro- 4.033 20020024 PCN 1,2,3,4-tetrachloronaphthalene 4.033 31604281 PCN 1,3,5,8-tetrachloronaphthalene 4.033 50402512 PCN 1,2,4-trichloronaphthalene 3.505 50402523 PCN 1,2,3-trichloronaphthalene 3.505 53555638 PCN 1,2,3,5-tetrachloronaphthalene 4.033 53555649 PCN 1,3,5,7-tetrachloronaphthalene 4.033 53555650 PCN 1,2,3,5,7-pentachloronaphthalene yes 4.492 55720371 PCN 1,3,7-trichloronaphthalene 3.505 55720406 PCN 2,3,6-trichloronaphthalene 3.505 55720439 PCN 1,4,6,7-tetrachloronaphthalene 4.033 67922218 PCN 1,2,4,7-tetrachloronaphthalene 4.033 67922263 PCN 1,2,3,4,6-pentachloronaphthalene yes 4.492 90948280 PCN 1,2,4,5,6,8-hexachloronaphthalene yes 4.893

103426977 PCN 1,2,3,5,6,7-hexachloronaphthalene yes 4.893

58

150224241 PCN 1,2,3,5,8-pentachloronaphthalene yes 4.492 29086382 PCS Trans-Heptachlorostyrene 4.246 29086393 PCS Cis-Heptachlorostyrene 4.246 61255810 PCS Heptachlorostyrene 4.246

355464 PFA 1-hexanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,6-tridecafluoro- yes 4.849

375928 PFA 1-Heptanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,7-pentadecafluoro- yes 5.375

423541 PFA Octanamide, 2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-pentadecafluoro- (7CI,8CI,9CI) yes 5.184

1763231 PFA 1-Octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro- yes 5.407

2795393 PFA 1-Octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-, potassium salt yes 5.218

41006411 PFA 1-Butanesulfinyl chloride, 1,1,2,2,3,3,4,4,4-Nonafluoro- 3.170

45298906 PFA 1-Octanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-, ion(1-) yes 5.407

51735843 PFA 1-Butanesulfinamide, 1,1,2,2,3,3,4,4,4-Nonafluoro- 3.583

68259074 PFA 1-Heptanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,7-pentadecafluoro-, ammonium salt yes 5.112

68259085 PFA 1-hexanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,6-tridecafluoro-, ammonium salt yes 5.036

68259096 PFA 1-pentanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,5-undecafluoro-,ammonium salt yes 4.451

68259121 PFA 1-Nonanesulfonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,9-nonadecafluoro- yes 5.066

68555679 PFA 1-Octanesulfinic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8,8-heptadecafluoro-, sodium salt yes 5.464

50293 POP p,p'-DDT 2.810 53190 POP o,p'-DDD 2.416 57749 POP chlordane yes 6.498 58899 POP gamma-HCH 4.131 72548 POP p,p'-DDD 2.416

118741 POP HCB 4.131 143500 POP chlordecone yes 8.666 319846 POP alpha-HCH 4.131 319857 POP beta-HCH 4.131 319868 POP delta-HCH 4.131 608731 POP hexachlorocyclohexane 4.131 789026 POP o,p'-DDT 2.810

1024573 POP Heptachlor epoxide yes 6.973 2385855 POP Mirex yes 9.345 3224826 POP o,p'-DDE 2.416

4329071 POP (m,p'-DDT) benzene, 1-chloro-3-[2,2,2-trichloro-1-(4-chlorophenyl)ethyl]- (9CI) 2.810

5103719 POP cis-chlordane (alpha-) yes 6.498 5103731 POP Cis-Nonachlor yes 6.780 5103742 POP trans-chlordane yes 6.498

6108107 POP Cyclohexane, 1,2,3,4,5,6-hexachloro-, (1.alpha.,2.alpha.,3.alpha.,4.beta.,5.beta.,6.beta.)- 4.131

6108118 POP ZETA-hexachlorocyclohexane 4.131 6108129 POP ETA-hexachlorocyclohexane 4.131 8001352 POP Toxaphene yes 5.561 8017343 POP p,p'-DDT 2.810 12789036 POP chlordane yes 5.964 21161580 POP 1,2,3,6,9,10,10-Heptachloropentacyclodecane yes 8.058 27154445 POP hexachlorocyclohexane 4.131 27304138 POP Oxychlordane yes 7.283 33442830 POP Photoheptachlor yes 8.058 39765805 POP Trans-nonachlor yes 6.780

39801144 POP 1,3,4-metheno-1H-cyclobuta[cd]pentalene, 1,1a,2,2,3,3a,4,5,5,5a,5b-undecachlorooctahydro- (9CI) yes 9.137

55570848 POP 1,3,4-metheno-1H-cyclobuta[cd]pentalene, 1a,2,2,3,3a,4,5,5,5a,6-decachlorooctahydro- (9CI) yes 8.907

115384946 POP 1,2,3,4,5,6,7,8,8,-Nonachlorooctahydro-4,7-METH* yes 6.780 87821 PXBD benzene, hexabromo- 4.131 87832 PXBD benzene, pentabromomethyl- 3.519

319879 PXBD benzene, pentachlorofluoro- 4.131 392574 PXBD benzene, 1,2,4,5-tetrafluoro-3,6-diiodo- 4.131 576556 PXBD phenol, 2,3,4,5-tetrabromo-6-methyl- 2.962

59

607954 PXBD phenol, 2,4,6-tribromo-, acetate (8CI,9CI) 2.567 608219 PXBD benzene, 1,2,3-tribromo- 2.526 608297 PXBD benzene, 1,2,3-triiodo- 2.526 608902 PXBD pentabromobenzene 3.693 615543 PXBD benzene, 1,2,4-tribromo- 2.526 626391 PXBD benzene, 1,3,5-tribromo- 2.526 626448 PXBD 1,3,5-triiodoenzene 2.526 634684 PXBD 1,2,3,4-tetraiodobenzene 3.168 634899 PXBD 1,2,3,5-tetrabromobenzene 3.168 634924 PXBD 1,2,3,5-tetraiodobenzene 3.168 636282 PXBD benzene, 1,2,4,5-tetrabromo- 3.168 636317 PXBD 1,2,4,5-tetraiodobenzene 3.168 832531 PXBD benzenesulfonyl chloride, pentafluoro- (7CI,8CI,9CI) 3.245

2338207 PXBD benzoic acid, 3,4,5-triiodo- (7CI,8CI,9CI) 2.443 2401210 PXBD benzene, 1,2-dichloro-3-iodo- (7CI,8CI,9CI) 2.526 2708976 PXBD benzene, 1,2,3,4-tetrafluoro-5,6-diiodo- (9CI) 4.131 2810697 PXBD benzene, 1,2,3,4-tetrabromo-5,6-dimethyl- (9CI) 2.962 3032813 PXBD benzene, 1,3-dichloro-5-iodo- 2.526 5075616 PXBD phosphine, dimethyl(pentafluorophenyl)- 3.475 13075019 PXBD benzene, 1,2,4-tribromo-3,5,6-trichloro- (6CI,7CI,8CI,9CI) 4.131 13311723 PXBD phenol, 4-bromo-2,3,6-trichloro- (7CI,8CI,9CI) 3.008 14400943 PXBD 2,3,4,6-tetrabromophenol 3.008 14862523 PXBD benzene, 1,3-dibromo-5-chloro- 2.526 15396368 PXBD benzoic acid, 2-chloro-3,5-diiodo- (8CI,9CI) 2.443 17274123 PXBD benzoic acid, 2,3,5-triiodo-, sodium salt (8CI,9CI) 2.443 19393921 PXBD 1-bromo-2,6-dichlorobenzene 2.526 19752557 PXBD benzene, 1-bromo-3,5-dichloro- 2.526 20020193 PXBD benzenamine, 2,4,6-tribromo-N-nitro- 2.567 20555913 PXBD benzene, 1,2-dichloro-4-iodo- 2.526 22311257 PXBD 1,2,3,4-tetrabromobenzene 3.168 23488382 PXBD benzene, 1,2,4,5-tetrabromo-3,6-dimethyl- 2.962 25376389 PXBD phenol, tribromo- 2.526 28779080 PXBD tribromobenzene 2.526 29682415 PXBD benzene, 1,4-dichloro-2-iodo- (8CI,9CI) 2.526 29682448 PXBD benzene, 1-bromo-2,4,5-trichloro- (6CI,8CI,9CI) 3.168 29898326 PXBD benzene, 2,4-dichloro-1-iodo- (8CI,9CI) 2.526 35488176 PXBD 2,3,4,5-tetrabromo-6-methoxyphenol 2.999 36059219 PXBD tetrabrom-O-xylol 2.962 38926851 PXBD 2,3,4-tribromo-6-methoxyphenol 2.443 39569216 PXBD tetrabromo-2-chlorotoluene 3.519 41424366 PXBD benzene, 1,3,5-tribromo-2-methoxy-4-methyl- 2.443 46495663 PXBD benzene, 1-chloro-2,3,5-trifluoro-4,6-diisocyanato- (9CI) 2.952 56759569 PXBD 1,2-benzenediol, 3,4,6-tribromo-5-methyl- (9CI) 2.443 56961774 PXBD 1-bromo-2,3-dichlorobenzene 2.526 58169996 PXBD phenol, 2,3,4,6-tetrabromo-5-methyl- 2.962 58683708 PXBD benzene, 1,2,4-tribromo-5-(1-methylethyl)- (9CI) 2.443 61878555 PXBD triiodobenzene 2.526 62778192 PXBD benzene, 1,3-dichloro-2-iodo-5-nitro- 2.443 63884435 PXBD 1,4-benzenediol, 2,3,5-tribromo-6-methyl- (9CI) 2.443 68084297 PXBD phenol, pentabromo-, aluminum salt 3.210 90077784 PXBD bromotetrachlorobenzene 3.693

115172162 PXBD dibromodichlorobenzene 3.168 434902 PXBP Decafluorobiphenyl yes 5.701

3883861 PXBP 1,1'-biphenyl, 2,2',3,3',5,5',6,6'-octafluoro- yes 5.176

10386842 PXBP 1,1'-biphenyl, 4,4'-dibromo-2,2',3,3',5,5',6,6'-octafluoro- (9CI) yes 5.701

59080330 PXBP 2,4,6-tribromobiphenyl 3.251 50585427 PXDD 2,3-dichloro-7,8-difluorodibenzo-p-dioxin 3.085 84761819 PXDF 1,2,8-tribromodibenzofuran 3.319 84761820 PXDF 2,3,8-tribromodibenzofuran 3.319

107227565 PXDF bromotrichlorodibenzofuran 3.779 18134582 silane silane, trichloro(1,4,5,6,7,7-hexachloro-5-norbornen-2-yl)- 4.323

68360 XAB benzene, 1,4-bis(trichloromethyl)- 3.879 81196 XAB benzene, 1,3-dichloro-2-(dichloromethyl)- 3.008

134258 XAB benzene, 2,4-dichloro-1-(dichloromethyl)- 3.008 401774 XAB benzene, 1-fluoro-3-(trichloromethyl)- (9CI) 3.008 402426 XAB benzene, 1-fluoro-4-(trichloromethyl)- (9CI) 3.008 488982 XAB benzene, 1-fluoro-2-(trichloromethyl)- 3.008 881992 XAB alpha,alpha'-hexachloro-m-xylene 3.879

1079170 XAB benzene, 1,2,4,5-tetrachloro-3,6-bis(chloromethyl)- (9CI) 3.879

60

1133579 XAB benzene, 1,2,3,5-tetrachloro-4,6-bis(chloromethyl)- (9CI) 3.879 1424799 XAB benzene, 1,2,4-trichloro-3-(chloromethyl)- 3.008 1592310 XAB benzene, 1,4-bis(dibromomethyl)- 2.962 2136814 XAB benzene, 1-chloro-3-(trichloromethyl)- 3.008 2136892 XAB benzene, 1-chloro-2-(trichloromethyl)- 3.008 2136950 XAB benzene, pentachloro(dichloromethyl)- (9CI) 4.321 2741573 XAB benzene, 1-(dichloromethyl)-2-(trichloromethyl)- (9CI) 3.456 3955268 XAB benzene, 1,2,4-trichloro-5-(chloromethyl)- 3.008 4960489 XAB benzene, 2-(bromomethyl)-1,3,4-trichloro- (9CI) 3.008 5216251 XAB benzene, 1-chloro-4-(trichloromethyl)- 3.008 7398825 XAB benzene, 1,4-bis(dichloromethyl)- (9CI) 2.962 10541716 XAB benzene, 1,4-dichloro-2-(trichloromethyl)- (9CI) 3.519 13014181 XAB benzene, 2,4-dichloro-1-(trichloromethyl)- 3.519 13014249 XAB benzene, 1,2-dichloro-4-(trichloromethyl)- 3.519 13209159 XAB benzene, 1,2-bis(dibromomethyl)- 2.962 13911029 XAB benzene, 1,2,3-trichloro-4-(chloromethyl)- (9CI) 3.008 17293037 XAB benzene, 1,3,5-trichloro-2-(chloromethyl)- (9CI) 3.008 17299977 XAB 1,3,5-tri(alpha-chloromethyl)benzene 2.443 23691272 XAB benzene, (1,3,3,3-tetrachloropropyl)- 2.999 24653858 XAB benzene, 1,2,4-trichloro-3-(dichloromethyl)- (9CI) 3.519 25393980 XAB benzene, 1,4-bis(1,2-dibromoethyl)- (9CI) 3.096 25641990 XAB benzene, 1,2-bis(dichloromethyl)- 2.962 25850491 XAB benzene, 1,3-bis(1,2-dibromoethyl)- (9CI) 3.096 28744458 XAB benzene, 1,3,5-trichloro-2-(dichloromethyl)- (9CI) 3.519 30359536 XAB benzene, 1-(2,2,2-trichloroethyl)-3-(trifluoroM* 3.887 31904184 XAB Toluene, .alpha.,.alpha.,.alpha.,ar,ar-pentachloro- 3.519 33429708 XAB benzene, 1,2,4-trichloro-5-(dichloromethyl)- (9CI) 3.519 36323281 XAB benzene, 1,3-bis(dibromomethyl)- 2.962 41999842 XAB benzene, 1,4-dichloro-2,5-bis(dichloromethyl)- (9CI) 3.879 54965401 XAB benzene, 1-chloro-4-(2,2-dichloro-1-methylethyl)- 2.443 56682872 XAB 1,3,5-triS(dichloromethyl)benzene 3.887 56762233 XAB benzene, (1,2,3-tribromopropyl)- (9CI) 2.443 56961821 XAB benzene, 1,2,3-trichloro-4-(dichloromethyl)- (9CI) 3.519 56961832 XAB benzene, 1,4-dichloro-2-(dichloromethyl)- (9CI) 3.008 56961843 XAB benzene, 1,2-dichloro-4-(dichloromethyl)- 3.008 56961854 XAB benzene, 1,3-dichloro-5-(dichloromethyl)- (9CI) 3.008 68238937 XAB benzene, dichloro(dichloromethyl)- 3.008

73588428 XAB benzene, 1,3-dichloro-5-(1,3,3,3-tetrachloro-1-methylpropyl)- 3.956

93962664 XAB benzene, 1-(3-bromopropyl)-2,4-dichloro- (9CI) 2.443 94022987 XAB benzene, 1-(2-iodoethyl)-2-(trifluoromethyl)- (9CI) 2.999 95998644 XABP chloro-AR,AR'-bis(trifluoromethyl)-1,1'-biphenyl 3.485 95998666 XABP trichloro-AR,AR'-bis(trifluoromethyl)-1,1'-biphenyl* 4.043 99686529 XABP dichlorobis(trifluoromethyl)biphenyl 3.777 346554 XAH Quinoline, 4-chloro-7-(trifluoromethyl)- 4.093 613536 XAH Quinoline, 2-(tribromomethyl)- 3.589 949428 XAH 1,3,5-triazine, 2-methyl-4,6-bis(trichloromethyl)- 3.887 956387 XAH 2-(2-bromoethyl)-4,6-bis(trichloromethyl)-S-TR* 4.309

1128161 XAH Pyridine, 3,5-dichloro-2-(trichloromethyl)- 3.519 1129197 XAH Pyridine, 2,4-dichloro-6-(trichloromethyl)- 3.519 1134049 XAH Pyridine, 2,3,4,5-tetrachloro-6-(trichloromethyl)- 4.321 1145444 XAH 2-(2-chloroethyl)-4,6-bis(trichloromethyl)-S-T* 4.309 1201305 XAH Pyridine, 3,4,5-trichloro-2-(trichloromethyl)- 3.950 1817136 XAH Pyridine, 3,6-dichloro-2-(trichloromethyl)- 3.519 2338309 XAH 4567-tetraBr-2-CF3 benzimidazole yes 5.247 3393597 XAH 4,5,7-triCL-2-CF3-benzimidazole yes 4.893 3599711 XAH 2-ethyl-4,6-bis(trichloromethyl)-S-triazine 3.956 3599744 XAH 2,4-bis(trichloromethyl)-S-triazine 3.879 3599766 XAH 2-methylthio-4,6-bis(trichloromethyl)-s-triazine 3.956 3671601 XAH 2-CF3-5-bromobenzimidazole 4.033 5311251 XAH 2,4-dimethoxy-6-trichloromethyl-S-triazine 2.737 5516483 XAH 2-methylthio-4,6-bis(1,1,2-trichloroethyl)-S-T* 3.754 6542672 XAH 2,4,6-triS(trichloromethyl)-S-triazine yes 4.850 6587219 XAH 2-CF3-5,6-dibromo-benzimidazole yes 4.492 7041249 XAH Pyridine, 2,3,5-trichloro-6-(trichloromethyl)- 3.950 7682328 XAH benzimidazole-4,5,6-triBr-2-CF3 yes 4.893 10243831 XAH N,N'-(6-chloro-S-triazine-2,4-diYL)bis(2,2,2-*) yes 4.670 13704900 XAH 2-chloro-4-phenyl-6-trichloromethyl-S-triazine 3.058 14946180 XAH 2,4-bis(methylthio)-6-trichloromethyl-S-triazine 2.737 21384338 XAH 2-(1-aziridinyl)-4,6-bis(trichloromethyl)-S-triazine* yes 4.923

61

22652148 XAH Pyridine, 2,6-dichloro-4-(trichloromethyl)- 3.519 23779977 XAH Quinoline, 4-chloro-8-(trifluoromethyl)- (8CI,9CI) 4.093 24481332 XAH 2-(1,1,2-trichloroethyl)-4,6-bis(trichloroMET*) yes 4.905 24481376 XAH 2-PenTYL-4,6-bis(trichloromethyl)-S-triazine 3.112 24481401 XAH 2-isobutyl-4,6-bis(trichloromethyl)-S-triazine 3.754 24481412 XAH 2-butyl-4,6-bis(trichloromethyl)-S-triazine 3.754 24481423 XAH 2-isopropyl-4,6-bis(trichloromethyl)-S-triazine 4.071 24481434 XAH 2-propyl-4,6-bis(trichloromethyl)-S-triazine 4.071 24481558 XAH 2-(O-chlorophenyl)-4,6-bis(trichloromethyl)-S-* 3.485 24504221 XAH 2-phenyl-4,6-bis(trichloromethyl)-S-triazine 3.164 24802855 XAH 2-(tert-butylamino)-4,6-bis(trichloromethyl)-S* 3.112 24803052 XAH 2-dimethylamino-4,6-bis(trichloromethyl)-S-triazine* 4.071 24803245 XAH 2-methylamino-4-propyl-6-trichloromethyl-S-triazine* 2.470 24803267 XAH 2-isopropyl-4-methylamino-6-trichloromethyl-S-triazine* 2.470 24803643 XAH 2-methylamino-4,6-bis(trichloromethyl)-S-triazine* 3.956 30339505 XAH 2-amino-4-(1,1,2-trichloroethyl)-6-trichloroMET* 3.956 30339572 XAH 2-amino-4,6-bis(1,1-dichloroethyl)-S-triazine 3.241 30339583 XAH 2,4-bis(1,1-dichloroethyl)-6-methylamino-S-triazine* 2.952 30339607 XAH 2-amino-4,6-bis(1,1,2-trichloroethyl)-S-triAZI* 4.071 30339709 XAH 2-amino-4-butyl-6-trichloromethyl-S-triazine 2.470 30339732 XAH 2-amino-4-(SEC-butyl)-6-trichloromethyl-S-triazine* 2.470 30339798 XAH 2-amino-4-isobutyl-6-trichloromethyl-S-triazine 2.470 30361978 XAH 2-(chloromethyl)-4,6-bis(trichloromethyl)-S-TR* 4.248 30362313 XAH 2-(dichloromethyl)-4,6-bis(trichloromethyl)-S-* yes 4.566 30362448 XAH 4,6-bis(trichloromethyl)-S-triazine-2-Methanol* 3.112 30362459 XAH alpha-methyl-4,6-bis(trichloromethyl)-S-trazine 2.495 30362620 XAH 2,4-diethyl-6-(trichloromethyl)-S-triazine 2.737 30362744 XAH 2,4,6-triS(1,1-dichloroethyl)-S-triazine 3.754 30362788 XAH 2,4,6-triS(1-bromoethyl)-S-triazine 2.470 30369154 XAH 2-amino-4-(2,4-dichlorophenyl)-6-(trichloroME*) 2.810 30377243 XAH 2-amino-4-(P-chlorophenyl)-6-trichloromethyl-S-triazine 2.416 30863004 XAH 2,4-bis(1,1-dichloroethyl)-6-(methylthio)-S-triazine 2.952 30863220 XAH ethyl 4,6-bis(trichloromethyl)-S-triazine-2-carboxylic acid* 3.112 30863253 XAH 4,6-bis(trichloromethyl)-S-triazin-2-YL isothiozole 4.071 30863515 XAH 2-methoxy-4,6-bis(trichloromethyl)-S-triazine 3.956 30894894 XAH 2-chloro-4,6-bis(trichloromethyl)-S-triazine 4.246 30894907 XAH 2-chloro-4,6-bis(1,1-dichloroethyl)-S-triazine 3.554 30894918 XAH 2-chloro-4,6-bis(1,1,2-trichloroethyl)-S-triazine 4.309 30894985 XAH 2,4-dibromo-6-(1,1,2-trichloroethyl)-S-triazine 3.456 30894996 XAH 2-bromo-4,6-bis(1,1,2-trichloroethyl)-S-triazine 4.309 31120237 XAH 2,4-dimethoxy-6-(1,1,2-trichloroethyl)-S-triazine 2.470 51492014 XAH Pyridine, 2,3-dichloro-6-(trichloromethyl)- 3.519 55366308 XAH 2,6-dichloro-3-(trichloromethyl)pyridine 3.519 69045836 XAH 2,3-dichloro-5-trichloromethylpyridine 3.519 70788544 XAH Pyridine, 2,3,4-trichloro-6-(trichloromethyl)- 3.950

108030779 XAH Pyridine, 3,5-dichloro-2,4-dimethoxy-6-(trichloromethyl)- 3.682 1573575 XAlka 1,2,2,3,3,4-hexachlorobutane 2.705 2431552 XAlka 1,1,2,2,3,4-hexachlorobutane 2.705 6820742 XAlka Perchloroisobutane 3.805 18585381 XAlka 1,2,3,4,5,6-hexachlorohexane 2.569 18791190 XAlka 1,1,1,2,3,4,4,4-Octachlorobutane 3.346 20338265 XAlka 1,1,2,2,3,3,4,4-Octachlorobutane 3.346 21483625 XAlka 1,1,1,2,2,3,3,4,4-Nonachlorobutane 3.593 26523637 XAlka Butane, hexachloro- (6CI,7CI,8CI,9CI) 2.705 32694761 XAlka 1,1,1,2,3,3,4,4-Octachlorobutane 3.346 34973416 XAlka 1,1,2,2,3,4,4-Heptachlorobutane 3.054 68512163 XAlka Decane, brominated chlorinated 2.617 79458541 XAlka 1,1,1,4,4,4-hexachlorobutane 2.705 83682284 XAlka 2,2,3,4,5,5-hexachlorohexane 2.569 83682295 XAlka 1,2,2,5,5,6-hexachlorohexane 2.569 83682693 XAlka 1,1,2,2,3,3-hexachlorobutane 2.705 83682706 XAlka 1,1,1,2,2,3,3-Heptachlorobutane 3.054

87843 XAlkaC Cyclohexane, 1,2,3,4,5-pentabromo-6-chloro- 4.131 707551 XAlkaC Cyclohexane, 1,1,2,3,4,5,6-heptachloro- (6CI,7CI,8CI,9CI) yes 4.501

1837918 XAlkaC Cyclohexane, 1,2,3,4,5,6-hexabromo- 4.131 3194578 XAlkaC 1,2,5,6-tetrabromocyclooctane 2.962 3322938 XAlkaC Cyclohexane, 1,2-dibromo-4-(1,2-dibromoethyl)- 2.962 22138392 XAlkaC 1-pentachlorocyclohexane 3.693 30554724 XAlkaC Cyclohexane, tetrabromodichloro- 4.131 30554735 XAlkaC Cyclohexane, tribromotrichloro- 4.131

62

31454485 XAlkaC tetrabromocyclooctane 2.962 33489279 XAlkaC 3-I-pentachlorocyclohexane 4.131 33489280 XAlkaC 1-I-pentachlorocyclohexane 4.131 36635026 XAlkaC 2-Br-pentachlorocyclohexane 4.131 36635037 XAlkaC 1-Br-pentachlorocyclohexane 4.131 51962631 XAlkaC 4-(1,2-dichloroethyl)-1,2-dichlorocyclohexane 2.962 55265695 XAlkaC 2,4-diBr-tetrachlorocyclohexane 4.131 55298458 XAlkaC 1,2-diBr-tetrachlorocyclohexane 4.131 55332893 XAlkaC 2,3-diBr-tetrachlorocyclohexane 4.131 56421444 XAlkaC 1-methylpentachlorocyclohexane 3.519 56421455 XAlkaC 1,4-dimethyltetrachlorocyclohexane 2.962 68258902 XAlkaC Cyclopentane, heptachloro- yes 4.832 68258913 XAlkaC Cyclopentane, hexachloro- yes 4.467 83682626 XAlkaC 1,2-bis(1,2-dichloroethyl)cyclobutane 2.962 83682648 XAlkaC 1,2,4-trichloro-4-(1,1,2-trichloroethyl)cyclohexaNE 3.879 376794 XAlke 1-Hexene, 1,3,4,5,6,6-hexachloro-1,2,3,4,5,6-hexafluoro- 4.173

1725742 XAlke 1,2,3,4,5,6-hexachloro-3-hexene 2.569 2482680 XAlke 2-Perchlorobutene 3.346 3050428 XAlke Perchlorobutene 3.346 36678452 XAlke 2-Butene, 1,1,2,3,4,4-hexabromo- 2.705 83682319 XAlke 2,2,3,4,5,5-hexachloro-3-hexene 2.569 83682400 XAlke 1,1-dichloro-2,3-bis(trichloromethyl)propene 3.266 706785 XAlkeC Octachlorocyclopentene yes 5.140

6317255 XAlkeC 1,3-cyclopentadiene, 1,2,3,4-tetrachloro-5-(dichloromethylene)- 4.131

6928570 XAlkeC 1,3-Cyclopentadiene, 1,2,3,4,5-pentachloro-5-(trichloromethyl)- (8CI,9CI) yes 4.818

55044467 XAlkeC cyclobutane, 1,2-dichloro 3,4-bis(dichloromethylene)- 4.131 57722153 XAlkeC hexachlorocyclohexene 4.131 57722164 XAlkeC hexachlorocyclohexene 4.131 57722175 XAlkeC hexachlorocyclohexene 4.131 62111471 XAlkeC Cyclopentene, heptachloro- yes 4.832 72030263 XAlkeC Cyclopentene, hexachloro- yes 4.467 737202 XAN Heptafluoro-2-(trichloromethyl)naphthalene yes 6.110 345904 XBDP benzene, 1,1'-(bromomethylene)bis[4-fluoro- (9CI) 2.598 361079 XBDP benzene, 1,1'-(2,2,2-trifluoroethylIdene)bis[4-chloro- 2.810 782081 XBDP chlorobis(p-chlorophenyl)methane 2.598

5216535 XBDP benzene, 1,1'-(2,2-dichloroethylidene)bis[4-bromo- (9CI) 2.416

5736492 XBDP benzene, 1,1'-(bromomethylene)bis[2,3,4,5,6-pentafluoro- (9CI) yes 5.204

87581 XH 2,3,4,5-tetraiodofuran 3.495 2176627 XH Pyridine, pentachloro- 3.693 2338105 XH 4,5,6,7-tetrachlorobenzotriazole 4.040 5599246 XH 2,4-dichloro-6-pentachlorophenoxy-S-triazine 4.162 7682340 XH benzimidazole-2,4,5,6,7-pentaCL yes 4.516 10202467 XH 2,4-dichloro-6-(P-chlorophenyl)-S-triazine 3.251 10351061 XH 4-Pyridinethiol, 2,3,5,6-tetrachloro- (7CI,8CI,9CI) 3.008 17717167 XH pyridine, 2,3,4,5-tetrachloro-6-fluoro- 3.693 18247773 XH 2,4-dichloro-6-(2,4-dichlorophenoxy)-S-triazine 3.058 22963628 XH Pyridine, 2,3,5,6-tetrachloro-4-(methylthio)- 2.962 30886261 XH 2,4-dichloro-6-(P-chlorophenoxy)-S-triazine 2.598 30886272 XH 2,4-dichloro-6-(2,4,5-trichlorophenoxy)-S-triazine 3.467 30894634 XH 2,4-dichloro-6-((P-chlorophenyl)thio)-S-triazine 2.598 22039389 XNOR 1,2,3,4,7,7-hexachloro-2-norbornene yes 5.289 7090417 XPAH biphenylene, 2,3,6,7-tetrachloro- yes 4.514 90077795 XPAH tetrachloroacenaphylene yes 4.514

90175 XUC benzenemethanol, .alpha.-(trichloromethyl)-, acetate 2.470 117088 XUC 1,3-Isobenzofurandione, 4,5,6,7-tetrachloro- 4.093 127902 XUC 1,1'-Oxybis[2,3,3,3-tetrachloropropane] 3.303

307711 XUC

Undecanoic acid, 2,2,3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,11,11-eicosafluoro-, potassium salt yes 5.013

358236 XUC methanesulfonic acid, trifluoro-, anhydride 2.846 376501 XUC hexanedioic acid, octafluoro-, diethyl ester 2.438 398992 XUC benzothiazole, 2-methyl-5-(trifluoromethyl)- 3.589 402380 XUC 3,3-diMe-1-(3-CF3 Ph)triazene 2.470 605367 XUC naphthalene, 1,2,3,4-tetrachloro-1,2,3,4-tetrahydro- 4.033 632791 XUC 1,3-Isobenzofurandione, 4,5,6,7-tetrabromo- 4.093

711626 XUC carbonic acid, CYCLIC 3,4,5,6-tetrachloro-O-phenylene ester 4.033

63

725893 XUC benzoic acid, 3,5-bis(trifluoromethyl)- (7CI,8CI,9CI) 4.071 791504 XUC 1,3-diEHIETANE, 2,2,4,4-tetraKIS(trifluoromethyl yes 5.529 853394 XUC methanone, bis(pentafluorophenyl)- (9CI) 4.287

1043501 XUC benzene, 1,2,3,4,5-pentafluoro-6-[(2,3,4,5,6-pentafluorophenyl)thio]- yes 4.978

1203867 XUC ethanone, 2,2-dichloro-1-(2,4,5-trichlorophenyl)- 3.475 1536238 XUC methanone, (2,3,4,5,6-pentafluorophenyl)phenyl- 2.810 1717506 XUC 1,3-dithietane, 2,2,4,4-tetrafluoro- (8CI,9CI) 4.085 1825236 XUC benzoyl chloride, pentachloro- (7CI,8CI,9CI) 3.879 1928376 XUC acetic acid, (2,4,5-trichlorophenoxy)-, methyl ester 2.470 2200706 XUC 4,4'-dihydroxyoctafluorobiphenyl 3.777 2207274 XUC 1,2,3,4-tetrachloro-5,5-dimethoxy-1,3-cyclopentadiene 2.999 2338274 XUC 1H-benzimidazole, 4,5,6-trichloro-2-(trifluoromethyl)- yes 4.893 2338296 XUC 1H-benzimidazole, 4,5,6,7-tetrachloro-2-(trifluoromethyl)- yes 5.247 2341868 XUC benzenamine, N-[(pentafluorophenyl)methylenE]- 2.810 2666708 XUC Carbonimidic dichloride, (2,4-dichlorophenyl)- (9CI) 2.962 2694066 XUC benzoic acid, 2,3,6-trichloro-, methyl ester 2.567 2948201 XUC chloroacetic acid, pentachlorophenyl ester 4.071 3141251 XUC thiophene, 2,3,4-tribromo- (6CI,7CI,8CI,9CI) 2.827 3584665 XUC 1H-benzimidazole, 5-chloro-2-(trichloromethyl)- (9CI) 4.033 3760665 XUC 1-Butanone, 2-bromo-4-chloro-1-(4-chlorophenyl)- (9CI) 2.737 3958030 XUC thiophene, tetrabromo- 3.495

4089581 XUC

propanoyl fluoride, 2,3,3,3-tetrafluoro-2- 1,1,2,3,3,3-hexafluoro-2- 1,1,2,2-tetrafluoro-2-(fluorosulfonyl)ethoxy prop 3.750

4400060 XUC [1,1'-biphenyl]-4-OL, 3,4',5-trichloro- 2.598 5358065 XUC benzoic acid, 2,3,4,5-tetrachloro-6-cyano-, methyl ester 2.952 5634377 XUC ethanol, 2,2,2-trichloro-, carbonate (2:1) (9CI) 2.707

5902692 XUC Acetic acid, chloro-, 2,4,5-trichlorophenyl ester (6CI,8CI,9CI) 3.096

6022339 XUC Carbamimidic chloride, N'-(3,4-dichlorophenyl)-N,N-dimethyl- (9CI) 2.737

13252147 XUC propanoic acid, 2,3,3,3-tetrafluoro-2-(1,1,2,3,* yes 4.805 14047097 XUC 3,3',4,4'-tetrachloroazobenzene 2.416 14815873 XUC benzoic acid, 4-(trichloromethyl)-, methyl ester (9CI) 2.737

16090145 XUC ethanesulfonyl fluoride, 2- 1- difluoro (trifluoroethenyl)oxy methyl -1,2,2,2-tetrafluoroethoxy -1,1,2,2-tetrafluoro- yes 4.577

16143843 XUC 3,5-diCF3 acetanilide 3.754 17025477 XUC benzene, [(tribromomethyl)sulfonyl]- 2.567 17587223 XUC 3,5-Octanedione, 6,6,7,7,8,8,8-heptafluoro-2,2-dimethyl- 3.310 18237940 XUC 1H-benzimidazole, 4,5,6,7-tetrachloro-2-methyl- 4.033 19259111 XUC thiophene, tetraiodo- (6CI,7CI,8CI,9CI) 3.495 20405305 XUC ethanol, 2,2,2-trichloro-, phosphate (3:1) (8CI,9CI) 4.153 21132309 XUC Urea, N,N'-bis(pentafluorophenyl)- (9CI) 2.985 21757824 XUC Acetic acid, 1-(3,4-dichlorophenyl)-2,2,2-trichloroethyl ester 3.376 23383962 XUC 2,4-diCL CF3-methanesulfonanilide 3.682

24169322 XUC 1,3-dioxolane, 2-(bromomethyl)-2-(2,3,4-trichlorophenyl)- (8CI,9CI) 3.733

25108694 XUC benzene, 1-chloro-4-(2,2-dichloro-1-methylcyclopropyl)- (8CI,9CI) 3.505

25711777 XUC propanoyl fluoride, 2,3,3,3-tetrafluoro-2- 1,1,2,3,3,3-hexafluoro-2-(pentafluorophenoxy)propoxy - 4.117

26654977 XUC ethanesulfonyl fluoride,2-[1-[difluoro[(triFLUO* 4.051

27619972 XUC 1-Octanesulfonic acid, 3,3,4,4,5,5,6,6,7,7,8,8,8-tridecafluoro- yes 4.769

29091096 XUC benzene, 2,4-dichloro-1,3-dinitro-5-(trifluoromethyl)- 2.746 29684262 XUC benzenediazonium, 3,5-bis(trifluoromethyl)- 3.956 32919043 XUC Acetic acid, tribromo-, phenylmethyl ester (9CI) 2.737

39108344 XUC 1-Decanesulfonic acid, 3,3,4,4,5,5,6,6,7,7,8,8,9,9,10,10,10-heptadecafluoro- (9CI) 4.216

39234861 XUC benzenesulfonyl chloride, 3,5-bis(trifluoromethyl)- 3.883 42016911 XUC benzoyl chloride, 2-hydroxy-3,5-diiodo- (9CI) 2.443 50274834 XUC benzoyl chloride, 3,4,5-trichloro-2-hydroxy- (9CI) 2.999 51366257 XUC 2,2,2-trichloro-1-(3,4-dichlorophenyl)ethyl acetate 3.376 51735901 XUC 1-butanesulfinyl isoCYANATE, 1,1,2,2,3,3,4,4,4-Nonafluoro- 3.808 55759881 XUC benzene, 2,4-dichloro-1-[(4-chlorophenyl)thio]- 2.598 55773905 XUC ethylpentachlorophenoxyacetate 2.746 56046083 XUC 3-methoxypentachlorocyclohexane 3.456 56400114 XUC 3(CF3-MeO)pentachlorocyclohexane yes 4.566 56421320 XUC 2-ethoxypentachlorocyclohexane 3.475

64

56421353 XUC 3-ethoxypentachlorocyclohexane 3.475 56421364 XUC 1(CF3-MeO)pentachlorocyclohexane yes 4.566 56961912 XUC 1,4-naphthalenedione, 5-bromo-2,3-dichloro- (9CI) 3.251 57846067 XUC thiophene, 2,5-dibromo-3-(dibromomethyl)- (9CI) 3.168

59819522 XUC phosphorochloridic acid, 2-chlorophenyl 2,2,2-trichloroethyl ester (9CI) 2.810

59866727 XUC propanoic acid, 2,3-dibromo-, 2,2,2-tribromo-1-chloroethyl ester (9CI) 2.612

60066844 XUC Cyclopropanecarboxylic acid, 2,2-dimethyl-3-(2,2,2-trichloroethyl)-, ethyl ester 2.470

60207309 XUC 1,3-dioxolane, 2-(bromomethyl)-2-(2,4-dichlorophenyl)- (9CI) 3.251

60254140 XUC Cyclopropanecarboxylic acid, 2,2-dimethyl-3-(2,2,2-trichloroethyl)-, ethyl ester, trans- 2.470

62601176 XUC benzene, pentachloro[(chloromethyl)thio]- (9CI) 3.879 63980121 XUC benzoic acid, 4-(trichloromethyl)-, 2-chloroethyl ester (9CI) 2.952 64630644 XUC 2',3,4,4',5-pentachloro-2-aminodiphenyl ether 2.810 64694855 XUC Acetamide, N-(3,4-dichlorophenyl)-2,2,2-trifluoro- 3.554

65072482 XUC benzenemethanol, 3,4-dichloro-.alpha.,.alpha.-bis(trifluoromethyl)- yes 4.624

65601403 XUC 2,2,4,4-tetrachloro-2,2,4,4,6,6-hexahydro-6,6-B* 2.569 68413718 XUC benzene, 1,3,5-tribromo-2-(2-bromoethoxy)- 2.999 68516541 XUC benzenesulfonyl chloride, 4-(1,1,2,2-tetrafluoroethoxy)- 2.810

69087463 XUC propanoic acid, 3- 1- difluoro (trifluoroethenyl)oxy methyl -1,2,2,2-tetrafluoroethoxy -2,2,3,3-tetrafluoro- yes 4.769

69116730 XUC

propanoic acid, 3- 2- 1,2-difluoro-2-oxo-1-(trifluoromethyl)ethoxy -1,2,2-trifluoro-1-(trifluoromethyl)ethoxy -2,2,3,3 3.337

74298694 XUC naphthalene, 2,4,5,8-tetrachloro-1-methoxy- 3.733 74664305 XUC benzenediazonium, 3,5-bis(trifluoromethyl)- 3.956 75147205 XUC 3-Butenoic acid, 2,2,3,4,4-pentachloro-, butyl ester 2.617 76253606 XUC tetrachlorobenzyltoluene 2.416

95998702 XUC alpha,alpha-difluorodichlorotrifluoromethyl diphenylmethane* 3.485

99308244 XUC trichloro-1-(dichloropropoxy)hexane 2.617 99308255 XUC tetrachloro-1-(dichloropropoxy)hexane 3.018

121107435 XUC 2,2'-4,4'-tetrachloro-3-methyl diphenylmethane 2.416 121107479 XUC 2,2'-4,6'-tetrachloro-3-methyl diphenylmethane 2.416 121107480 XUC 2,2'-4,6'-tetrachloro-5-methyl diphenylmethane 2.416 121107833 XUC 2',3-4,6'-tetrachloro-6-methyl diphenylmethane 2.416

65

Table 4 - CAS number, chemical class, name and POP score of chemicals meeting the

elevated AC-BAP and atmospheric oxidation half-life criteria, but not meeting the POP

score criteria.

CAS Class NAME SCORE

80068 CUP chlorfenethol 0.91 84548 CUP 9,10-Anthracenedione, 2-methyl- -0.65 84628 CUP 1,2-benzenedicarboxylic acid, diphenyl ester -4.02 88062 CUP phenol, 2,4,6-trichloro- 2.40 88857 CUP phenol, 2-(1-methylpropyl)-4,6-dinitro- -1.75 92933 CUP 1,1'-biphenyl, 4-nitro- -0.23 95954 CUP phenol, 2,4,5-trichloro- 2.40 97165 CUP genite 0.35

103333 CUP diazene, diphenyl- 0.28 115322 CUP dicofol 2.18 116290 CUP tetradifon 1.80 117260 CUP bulan -1.22 117271 CUP 2-nitro-1,1-bis(p-chlorophenyl)propane -0.71 510156 CUP chlorobenzilate -1.22 513928 CUP ethene, tetraiodo- 2.52

1420071 CUP dinoterb -1.75 1836755 CUP benzene, 2,4-dichloro-1-(4-nitrophenoxy)- 0.35 1836777 CUP chlornitrofen 0.80 1861321 CUP DCPA 1.75 2439012 CUP oxythioquinox 0.40 2487016 CUP medinoterb acetate -3.51 2759719 CUP cypromide 2.70 3765579 CUP glenbar 1.75 3861414 CUP bromoxynil butyrate 0.81 3996596 CUP medinoterb -2.19 5335240 CUP [1,1'-biphenyl]-2-ol, 3,5-dichloro- 2.08 8027007 CUP dilan -0.71 13738631 CUP 1,5-dichloro-3-fluoro-2-(4-nitrophenoxy)benzene 0.80 14255880 CUP Fenazaflor 0.36 15457053 CUP benzene, 2-nitro-1-(4-nitrophenoxy)-4-(trifluoromethyl)- -1.35 27253298 CUP Neodecanoic acid, zinc salt -5.19 28159980 CUP IRGAROL 1051 -1.20 42576023 CUP bifenox -1.73 42874011 CUP nitrofluorfen 0.63 50594677 CUP acifluorfen-Me -1.52 60168889 CUP fenarimol -1.26 61444620 CUP N-[2-amino-3-nitro-5-(trifluoromethyl)phenyl]-2* 1.63 79622596 CUP fluazinam -0.41

83164334 CUP 3-pyridinecarboxamide, N-(2,4-difluorophenyl)-2-[3-(trifluoromethyl)phenoxy]- -1.84

88283414 CUP ethanol, 1-(2,4-dichlorophenyl)-2-(3-pyridinyl)-, O-methyloxime -0.19 106917526 CUP flusulfamide -0.68 131086425 CUP HC-252 Herbicide -2.75 3740929 CUP fenchlorim 2.70 101553 PBDE benzene, 1-bromo-4-phenoxy- 1.48

2050477 PBDE benzene, 1,1'-oxybis 4-bromo- 2.08 36563470 PBDE bromophenoxybenzene 1.48 2050671 PCB biphenyl, 3,3'-dichloro- 2.70 2050682 PCB 4,4'-dichlorobiphenyl 2.70 2051607 PCB 2-chlorobiphenyl 2.07 2051618 PCB 3-chlorobiphenyl 2.07 2051629 PCB 4-chlorobiphenyl 2.07 2974905 PCB biphenyl, 3,4'-dichloro- 2.70 2974927 PCB biphenyl, 3,4-dichloro- 2.70 11104282 PCB Aroclor 1221 2.07 11141165 PCB Aroclor 1232 2.07 13029088 PCB 2,2'-dichlorophenyl 2.70 16605917 PCB 2,3-dichlorobiphenyl 2.70 25569806 PCB 2,3'-PCB 2.70 27323188 PCB chlorobiphenyl 2.07 33039815 PCB dichlorobiphenyl 2.70

66

33146451 PCB 2,6-dichlorobiphenyl 2.70 33284503 PCB 2,4-dichlorobiphenyl 2.70 34883391 PCB 2,5-dichlorobiphenyl 2.70 34883415 PCB biphenyl, 3,5-dichloro- 2.70 34883437 PCB 2,4'-PCB 2.70

132567227 PCB 4,4'-dichlorobiphenyl 2.70 87401 PCBD benzene, 1,3,5-trichloro-2-methoxy- 2.38 88493 PCBD benzenesulfonyl chloride, 4-chloro-2,5-dimethyl- 1.41 98317 PCBD benzenesulfonyl chloride, 3,4-dichloro- 1.64 98602 PCBD benzenesulfonyl chloride, 4-chloro- 0.91

102363 PCBD benzene, 1,2-dichloro-4-isocyanato- 1.79 320729 PCBD benzoic acid, 3,5-dichloro-2-hydroxy- 1.95 551768 PCBD 3-methyl-2,4,6-trichlorophenol 2.38 585477 PCBD 1,3-benzenedisulfonyl dichloride 0.55 609198 PCBD 3,4,5-trichlorophenol 2.40 609609 PCBD benzenesulfonyl chloride, 2,4-dimethyl- 0.67 634935 PCBD benzenamine, 2,4,6-trichloro- 2.40 882144 PCBD propanamide, N-(3,4-dichlorophenyl)- 1.92 933755 PCBD phenol, 2,3,6-trichloro- 2.40 933788 PCBD 2,3,5-trichlorophenol 2.40

1196130 PCBD benzene, 1,3,5-trichloro-2-nitroso- 2.38 1570678 PCBD 2,6-dichlor-3,4-dimethylphenol 1.79 1965259 PCBD benzene, 1-azido-2,4-dichloro- 1.79 2057649 PCBD 2,4-dichlor-3,6-dimethylphenol 1.79 2131557 PCBD benzene, 1-chloro-4-isothiocyanato- 1.00 2213812 PCBD benzene, 2-chloro-5-(1,1-dimethylethyl)-1,3-dinitro- -0.74 2612579 PCBD benzene, 2,4-dichloro-1-isocyanato- 1.79 2888064 PCBD M-chlorobenzenesulphonyl chloride 0.91 2891170 PCBD 1,3-benzenedisulfonyl dichloride, 4-chloro- 1.19 2905239 PCBD 2-chlorobenzenesulfonyl chloride 0.91 3296057 PCBD benzene, 1-azido-4-chloro- 1.00 3426899 PCBD phosphorodichloridous acid, phenyl ester -0.25 3773146 PCBD 2,4,5-trichlorobenzenethiol 2.40 3780856 PCBD carbonylcyanide26dichlorophenylhydrazone 1.36 4163784 PCBD benzene, 1,2,4-trichloro-5-(methylthio)- 2.38 5392825 PCBD benzene, 1,4-dichloro-2-isocyanato- 1.79 5402733 PCBD benzenesulfonyl chloride, 2,5-dichloro- 1.64 6130752 PCBD benzene, 1,2,4-trichloro-5-methoxy- 2.38 6590949 PCBD benzene, 1,2-dichloro-4-isothiocyanato- 1.79 6590961 PCBD benzene, 2,4-dichloro-1-isothiocyanato- 1.79 6639301 PCBD benzene, 1,2,4-trichloro-5-methyl- 2.40 6752381 PCBD benzenesulfonyl chloride, 4-isocyanato- 0.55 7160227 PCBD propanamide, N-(3,4-dichlorophenyl)-2,2-dimethyl- 1.36 7359720 PCBD 2,3,4-trichlorotoluene 2.40 7606873 PCBD benzoic acid, 3,5-dichloro-2-hydroxy-, methyl ester 2.16 7669547 PCBD benzenesulfenyl chloride, 2-nitro- 0.35 13334731 PCBD benzene, 1,5-dichloro-2-methoxy-3-methyl- 1.79 13790142 PCBD benzene, 1,3-dichloro-4-methyl-2-nitro- 1.95 14593283 PCBD 4-(tert-butyl)-2-chloro-6-nitrophenol 0.25 14611702 PCBD thiocyanic acid, 3,5-dichloro-4-hydroxyphenyl ester 1.95 15196835 PCBD benzoic acid, 5-chloro-2-hydroxy-, ethyl ester 1.29 15945070 PCBD benzenesulfonyl chloride, 2,4,5-trichloro- 2.26 15950660 PCBD 2,3,4-trichlorophenol 2.40 17199212 PCBD 2-chloro-6-nitro-4-(1,1,3,3-tetramethylbutyl)-P* -1.69 17302828 PCBD benzoic acid, 3,5-dichloro-4-hydroxy-, ethyl ester 1.92 19040621 PCBD benzenesulfonyl chloride, 2,5-dimethyl- 0.67 21472866 PCBD benzene, 1,2,3-trichloro-5-methyl- 2.40 23165539 PCBD benzene, 4-chloro-1-isothiocyanato-2-methyl- 1.22 23378883 PCBD benzenesulfonyl chloride, 3,5-dichloro-2-hydroxy- 1.83 25097443 PCBD phosphinic chloride, (1,1-dimethylethyl)[4-(1,1-dimethylethyl)phenyl]- (9CI) -1.30 25167822 PCBD phenol, trichloro- 2.40 25167844 PCBD trichloro-1,2-benzenediol 2.38 27165226 PCBD benzenediazonium, 4-chloro-2-nitro- 1.48 28225889 PCBD benzenethiol, 2,4-dichloro-5-methyl- 1.68 28286864 PCBD benzenesulfonyl chloride, 2,4-dichloro-5-methyl- 1.83

28343284 PCBD pentanenitrile, 4,4-dimethyl-3-oxo-2-[(2,4,5-trichlorophenyl)hydrazono]- (9CI) -0.31

29682460 PCBD benzene, 1,3-dichloro-2-methyl-4-nitro- 1.95 30928246 PCBD benzenediazonium, 2,6-dichloro-4-nitro- 2.16 32139723 PCBD 1,2-benzenediol, 3,4,6-trichloro- 2.38

67

34123507 PCBD benzene, 2-chloro-4-isocyanato-1-(1-methylethyl)- 1.29 34732097 PCBD 2,3,4-trichlorobenzenesulphonyl chloride 2.26 34893920 PCBD benzene, 1,3-dichloro-5-isocyanato- 1.79 34981389 PCBD benzenesulfonyl chloride, 5-chloro-2-methyl- 1.15 35282838 PCBD benzene, 1,2,4-trichloro-3,6-dimethoxy-5-nitro- 1.88 36663759 PCBD benzene, 2-chloro-5-(1-methylethyl)-1,3-dinitro- -0.25 37693155 PCBD phenol, 2,6-dichloro-3-methyl-4-nitro- 2.16 38815393 PCBD phosphorodichloridic acid, (1,1-dimethylethyl)phenyl ester 0.07 39549274 PCBD phenol, 2,4-dichloro-3-methyl-6-nitro- 2.16 39920371 PCBD benzene, 1,3-dichloro-2-isocyanato- 1.79 41195908 PCBD benzene, 1,2-dichloro-3-isocyanato- 1.79 54135807 PCBD benzene, 1,2,3-trichloro-4-methoxy- 2.38 54135818 PCBD benzene, 1,2,5-trichloro-3-methoxy- 2.38 54135829 PCBD benzene, 1,2,3-trichloro-5-methoxy- 2.38 54932842 PCBD benzene, 1,4-dichloro-2,5-diethyl- 1.95 55191167 PCBD benzene, 1,3-dichloro-5-(1,1-dimethylethyl)-2-methoxy- 1.92 56157927 PCBD 4-chloro-2-methylbenzenesulphonyl chloride 1.15 56961207 PCBD 1,2-benzenediol, 3,4,5-trichloro- 2.38 56961321 PCBD benzoic acid, 3-chloro-2-hydroxy-, ethyl ester 1.29 56961456 PCBD phenol, 2,4,5-trichloro- 2.40 56962095 PCBD 1-propanone, 1-(4-chlorophenyl)-, oxime 1.48 57135689 PCBD benzene, 1-chloro-2-isothiocyanato-4-nitro- 1.29 60741517 PCBD phenol, 2,4-dichloro-6-methyl-3-(1-methylethyl)- 2.16 64070811 PCBD arsinic chloride, 1,2-phenylenebis- (9CI) -1.42 65724121 PCBD dichloro-p-cymene 1.95 68631049 PCBD benzenesulfonyl chloride, 2,3-dimethyl-5-nitro- -0.39 69158265 PCBD 2,4,5,6-tetrachloro-1,3-benzenediol diacetate 1.75 71463597 PCBD pentanoic acid, 2,6-dichlorophenyl ester 1.36 74098296 PCBD phenol, 6-chloro-3,4-dimethyl-2-nitro- 1.29 74220158 PCBD phenol, 2,4,6-trichloro-, calcium salt (9CI) 2.40 74664338 PCBD benzenediazonium, 4-chloro-2-nitro-, trichlorozincate(1-) (9CI) 1.48 81686411 PCBD 2,5-dichloro-p-cymene 1.95 81686444 PCBD 2,3-dichloro-p-cymene 1.95 81686466 PCBD 2,6-dichloro-p-cymene 1.95 92366353 PCBD dichloro-p-cymen-8-ol 2.16 93942611 PCBD ethanone, 1-(2,3-dichlorophenyl)-, oxime 1.95 94213226 PCBD benzeneacetaldehyde, 2,3-dichloro-a-oxo-, aldoxime (9CI) 2.16 99817364 PCBD 2-nitro-4,6-dichloro-5-ethylphenol 1.92 29446159 PCDD 2,3-dichlorodibenzo-p-dioxin 2.15 33857260 PCDD 2,7-dichlorodibenzo-p-dioxin 2.15 33857282 PCDD 2,3,7-trichlorodibenzo-p-dioxin 2.64 38178380 PCDD 1,6-dichlorodibenzo-p-dioxin 2.15 38964226 PCDD 2,8-dichlorodibenzo-p-dioxin 2.15 39227582 PCDD 1,2,4-trichlorodibenzo-p-dioxin 2.64 50585392 PCDD 1,3-dichlorodibenzo-p-dioxin 2.15 54536173 PCDD 1,2,3-trichlorodibenzo-p-dioxin 2.64 54536184 PCDD 1,2-dichlorodibenzodioxin 2.15 54536195 PCDD 1,4-dichlorodibenzodioxin 2.15 64501004 PCDD dichlorodibenzodioxin 2.15 64560130 PCDD 4,6-dichlorodibenzodioxin 2.15 67028175 PCDD 1,3,7-trichlorodibenzodioxin 2.64 69760969 PCDD trichlorodibenzodioxin 2.64 82291336 PCDD 1,3,6-trichlorodibenzo-p-dioxin 2.64 82306614 PCDD 1,3,8-trichlorodibenzo-p-dioxin 2.64 82306625 PCDD 1,3,9-trichlorodibenzo-p-dioxin 2.64 2444895 PCDE benzene, 1,1'-oxybis[4-chloro- 2.08 2689078 PCDE benzene, 1-chloro-2-phenoxy- 1.48 6842622 PCDE benzene, 1-chloro-3-(4-chlorophenoxy)- 2.08 6903657 PCDE 2,4'-dichlorodiphenyl ether 2.08 7005723 PCDE benzene, 1-chloro-4-phenoxy- 1.48 24910687 PCDE 3,5-dichlorodiphenyl ether 2.08 24910698 PCDE benzene, 1,4-dichloro-2-phenoxy- 2.08 28419694 PCDE 2,6-dichlorodiphenyl ether 2.08 28675083 PCDE dichlorophenyl ether 2.08 51892263 PCDE 2,4-dichloro-1-phenoxybenzene 2.08 55398862 PCDE monochloro diphenyl oxide 1.48 55538697 PCDE benzene, 1,2-dichloro-4-phenoxy- 2.08 5409836 PCDF 2,8-dichlorodibenzofuran 2.80 24478748 PCDF 2,4-dichlorodibenzofuran 2.80 25074673 PCDF 3-chlorodibenzofuran 2.20

68

43047990 PCDF dichlorodibenzofuran 2.80 51230490 PCDF 2-chlorodibenzofuran 2.20 58802214 PCDF 3,7-dichlorodibenzofuran 2.80 60390274 PCDF 2,6-dichlorodibenzofuran 2.80 64126858 PCDF 1,2-dichlorodibenzofuran 2.80 64126869 PCDF 2,3-dichlorodibenzofuran 2.80 70648145 PCDF 1,9-dichlorodibenzofuran 2.80 74918404 PCDF 3,6-dichlorodibenzofuran 2.80 74992964 PCDF 4-chlorodibenzofuran 2.20 74992975 PCDF 1,6-dichlorodibenzofuran 2.80 74992986 PCDF 2,7-dichlorodibenzofuran 2.80 81638371 PCDF 1,8-dichlorodibenzofuran 2.80 84761864 PCDF 1-chlorodibenzofuran 2.20 94538008 PCDF 1,3-dichlorodibenzofuran 2.80 94538019 PCDF 1,4-dichlorodibenzofuran 2.80 94538020 PCDF 1,7-dichlorodibenzofuran 2.80 94570839 PCDF 3,4-dichlorodibenzofuran 2.80 1825305 PCN naphthalene, 1,5-dichloro- 2.89 1825316 PCN naphthalene, 1,4-dichloro- 2.89 2050693 PCN naphthalene, 1,2-dichloro- 2.89 2050728 PCN naphthalene, 1,6-dichloro- 2.89 2050739 PCN naphthalene, 1,7-dichloro- 2.89 2050740 PCN naphthalene, 1,8-dichloro- 2.89 2050751 PCN naphthalene, 2,3-dichloro- 2.89 2065705 PCN naphthalene, 2,6-dichloro- 2.89 2198756 PCN naphthalene, 1,3-dichloro- 2.89 2198778 PCN naphthalene, 2,7-dichloro- 2.89 28699889 PCN naphthalene, dichloro- 2.89

98588 PXBD benzenesulfonyl chloride, 4-bromo- 0.91 98613 PXBD benzenesulfonyl chloride, 4-iodo- 0.91 99285 PXBD phenol, 2,6-dibromo-4-nitro- 1.95

106376 PXBD benzene, 1,4-dibromo- 1.72 106536 PXBD benzenethiol, 4-bromo- 0.71 118796 PXBD phenol, 2,4,6-tribromo- 2.40 147820 PXBD benzenamine, 2,4,6-tribromo- 2.40 305851 PXBD phenol, 2,6-diiodo-4-nitro- 1.95 455209 PXBD benzenesulfonyl fluoride, 4-ethyl- 0.67 583551 PXBD benzene, 1-bromo-2-iodo- 1.72 589877 PXBD benzene, 1-bromo-4-iodo- 1.72 591184 PXBD benzene, 1-bromo-3-iodo- 1.72 601843 PXBD benzoic acid, 2,6-dibromo- 1.79 607998 PXBD benzene, 1,3,5-tribromo-2-methoxy- 2.38 609223 PXBD phenol, 2,4-dibromo-6-methyl- 1.68 609234 PXBD phenol, 2,4,6-triiodo- 2.40 615429 PXBD benzene, 1,2-diiodo- 1.72 624384 PXBD benzene, 1,4-diiodo- 1.72 626006 PXBD benzene, 1,3-diiodo- 1.72 827236 PXBD 2,4-dibrom-6-nitro-anilin 1.95

1074244 PXBD benzene, 1,4-dibromo-2,5-dimethyl- 1.68 1139522 PXBD phenol, 4-bromo-2,6-bis(1,1-dimethylethyl)- -0.25 1646544 PXBD benzene, 1,4-dibromo-2,3,5,6-tetramethyl- 1.95 1940427 PXBD phenol, 4-bromo-2,5-dichloro- 2.40 1985122 PXBD benzene, 1-bromo-4-isothiocyanato- 1.00 2059769 PXBD benzene, 1-iodo-4-isothiocyanato- 1.00 2101884 PXBD benzene, 1-azido-4-bromo- 1.00 2101895 PXBD benzene, 1-azido-3-bromo- 1.00 2109128 PXBD phenol, 2,4,6-triiodo-3-methyl- 2.38 2131591 PXBD benzene, 1-bromo-3-isothiocyanato- 1.00 2432146 PXBD phenol, 2,6-dibromo-4-methyl- 1.68 2437492 PXBD 1,3-benzenediol, 2,4,6-tribromo- 2.38 2489523 PXBD benzenesulfonyl fluoride, 3-(chlorosulfonyl)- 0.55 2747173 PXBD 3,4,5-tribromocatechol 2.38 2905240 PXBD M-bromobenzenesulphonyl chloride 0.91 3217150 PXBD 4-bromo-2,6-dichlorophenol 2.40 3302394 PXBD benzene, 1-azido-2-bromo- 1.00 3336263 PXBD benzonitrile, 3-bromo-4-hydroxy-5-iodo- 1.79 3460182 PXBD benzene, 1,4-dibromo-2-nitro- 1.79 4068751 PXBD benzoic acid, 2-hydroxy-5-iodo-, methyl ester 1.48 4186521 PXBD phenol, 2,4-diiodo-6-methyl- 1.68 4524770 PXBD 6-bromo-2,4-dichlorophenol 2.40

69

4526561 PXBD 4,6-dibromo-2-chlorophenol 2.40 4619743 PXBD phenol, 2,4,6-tribromo-3-methyl- 2.38 5324130 PXBD 2,6-dibromo-4-chlorophenol 2.40 5326385 PXBD benzene, 1-iodo-2-methyl-4-nitro- 1.22 5398276 PXBD benzenamine, 2,6-diiodo-4-nitro- 1.95 5469192 PXBD benzene, 1-bromo-2,4,5-trimethyl- 1.00 6320407 PXBD benzene, 1,3,5-tribromo-2-methyl- 2.40 6380343 PXBD phenyl arsenic diiodide -0.68 6942990 PXBD benzene, 2,4-dibromo-1,3,5-trimethyl- 1.79 7191460 PXBD ethanone, 1-(2-hydroxy-3,5-diiodophenyl)- 1.95 7530270 PXBD phenol, 4-bromo-2-chloro-6-methyl- 1.68 7587157 PXBD phenol, 2,5-dichloro-4-iodo- 2.40 7745928 PXBD benzene, 2-iodo-1-methyl-4-nitro- 1.22 15046841 PXBD 2,4-dibromoanisole 1.68 17199223 PXBD 2-bromo-6-nitro-4-(1,1,3,3-tetramethylbutyl)-PH* -1.69 17199234 PXBD 2-bromo-4-(tert-butyl)-6-nitrophenol 0.25 19241384 PXBD benzene, 4-bromo-1-isothiocyanato-2-methyl- 1.22 21702841 PXBD benzene, 2,4-dibromo-1-methoxy- 1.68 22621438 PXBD salicylic acid, 5-iodo-3-nitro- 0.76 22802400 PXBD phenol, 2,6-dibromo-3,4-dimethyl- 1.79 23456931 PXBD 3,3-dimethyl-1-(4-jodphenyl)-triazen 1.48 26249127 PXBD benzene, dibromo- 1.72 27476228 PXBD tribromotoluene 2.40 28166065 PXBD benzene, 4-azido-1-fluoro-2-nitro- 1.29 28805905 PXBD dibromodimethylbenzene 1.68 29052071 PXBD butyric acid, M-iodophenyl ester 1.29 29052082 PXBD butyric acid, P-iodophenyl ester 1.29 30672709 PXBD benzenesulfonyl fluoride, 2-(chlorosulfonyl)- (9CI) 0.55 30672721 PXBD benzenesulfonyl fluoride, 4-(chlorosulfonyl)- 0.55 30812874 PXBD benzene, dibromoethyl- 1.68 34586497 PXBD benzenesulfonyl fluoride, 2-ethyl- 0.67 34586500 PXBD benzenesulfonyl fluoride, 3-ethyl- 0.67 40371640 PXBD benzene, 1-bromo-5-chloro-2-methyl-4-nitro- 1.95 41252964 PXBD benzene, 2-chloro-1-iodo-4-nitro- 1.79 41252975 PXBD benzene, 4-iodo-1-methyl-2-nitro- 1.22 41727473 PXBD benzoic acid, 3,5-dibromo-4-hydroxy-, methyl ester 2.16 42228660 PXBD benzenediazonium, 2-bromo-4,6-dinitro- 0.25 50638476 PXBD benzene, 4-bromo-2-chloro-1-methoxy- 1.68 50702380 PXBD M-iodobenzenesulphonyl chloride 0.91 51699899 PXBD benzene, 1,3-dibromo-2-methoxy-5-methyl- 1.79 52628372 PXBD tribromoaniline 2.40 54852685 PXBD phenol, 2-bromo-4-chloro-6-methyl- 1.68 55771818 PXBD benzoic acid, 3,5-dibromo-4-hydroxy-, ethyl ester 1.92 56701305 PXBD benzene, 2-iodo-1,3,5-trimethyl-4-nitro- 1.29 57018129 PXBD phenol, 2,6-dibromo-4-ethyl- 1.79 62265990 PXBD benzene, 1,3-dibromo-2-methoxy-4-methyl-5-nitro- 1.92 62778181 PXBD phenol, 2,6-diiodo-3,4-dimethyl- 1.79 63059290 PXBD benzenesulfonyl chloride, 2-iodo- 0.91 64046602 PXBD 1,3-benzenediol, 2,4,6-tribromo-5-methyl-, diacetate (9CI) 0.75 65436875 PXBD tribromomethylphenol 2.38 66974581 PXBD 1-(o-I Ph)-3,3-dimethyltriazene 1.48 68084300 PXBD phenol, 2,4,6-tribromo-, aluminum salt 2.13 71412254 PXBD Butanoic acid, 4-cyano-2,6-diiodophenyl ester 0.81 71672883 PXBD benzene, 1-bromo-4-isothiocyanato-2-methyl- 1.22 86006442 PXBD dibromochloromethylphenol 2.38

92660 PXBP 1,1'-biphenyl, 4-bromo- 2.07 92864 PXBP 1,1'-biphenyl, 4,4'-dibromo- 2.70

321608 PXBP 1,1'-biphenyl, 2-fluoro- 2.07 324743 PXBP 1,1'-biphenyl, 4-fluoro- 2.07 398232 PXBP 4,4'-difluorobiphenyl 2.70

1591317 PXBP 1,1'-biphenyl, 4-iodo- 2.07 2052075 PXBP 1,1'-biphenyl, 2-bromo- 2.07 2113511 PXBP 1,1'-biphenyl, 2-iodo- 2.07 2113577 PXBP 1,1'-biphenyl, 3-bromo- 2.07 3001158 PXBP 1,1'-biphenyl, 4,4'-diiodo- 2.70 13029099 PXBP 2,2'-dibromo-1,1'-biphenyl 2.70 20442799 PXBP 1,1'-biphenyl, 3-iodo- 2.07 27479658 PXBP dibromo-1,1'-biphenyl 2.70 37847522 PXBP 1,1'-biphenyl, 2,4-difluoro- 2.70 41604197 PXBP 1,1'-biphenyl, 4-bromo-2-fluoro- 2.70

70

53591983 PXBP 1,1'-biphenyl, 2-bromo-4-fluoro- 2.70 39073079 PXDD 2,7-dibromodibenzo-p-dioxin 2.15 50585370 PXDD 2,3-dibromodibenzo-p-dioxin 2.15 50585381 PXDD 2,3-difluorodibenzo-p-dioxin 2.15 91371141 PXDD 1,6-dibromodibenzo-p-dioxin 2.15

105836962 PXDD 2,8-dibromodibenzo-p-dioxin 2.15 86760 PXDF 2-bromodibenzofuran 2.20

10016521 PXDF 2,8-dibromodibenzofuran 2.80 65489807 PXDF 2,7-dibromodibenzofuran 2.80

83534 PXN naphthalene, 1,4-dibromo- 2.89 36316888 PXN 2,6-diiodonaphthalene 2.89

80104 silane silane, dichlorodiphenyl- 0.79 144796 silane silane, chloromethyldiphenyl- 0.28 312403 silane silane, difluorodiphenyl- 0.79 775122 silane silane, diphenyl- 0.79 776761 silane silane, methyldiphenyl- 0.28 825945 silane silane, trichloro(4-chlorophenyl)- 0.38 830466 silane benzene, 1,4-bis(trichlorosilyl)- -0.25

1000700 silane silamine, N,N'-methanetetraYLbis[1,1,1-trimethyl- -0.18 1078962 silane butanenitrile, 4-(dichlorophenylsilyl)- 0.55 1078973 silane silane, 1,4-phenylenebis[chlorodimethyl- 0.55 1631830 silane chloro-diphenyl silane 0.79 2602553 silane silane, dichlorobis(trichloromethyl)- 1.77 3401261 silane silane, dichloro(3-chloropropyl)phenyl- 1.09 3449261 silane silanamine, N-(dimethylphenylsilyl)-1,1-dimethyl-1-phenyl- -2.14 4774758 silane diphenyl methyl silicon azide -1.20 5599326 silane triethyl silicon azide -0.59 13465844 silane tetraiodosilane -4.31 13688909 silane silane, trichloro[4-(chloromethyl)phenyl]- 0.56 17887411 silane silane, dichloro(1,1-dimethylethyl)phenyl- 0.67 18081124 silane silane, chloro(4-chlorophenyl)methylphenyl- 0.88 27137855 silane silane, trichloro(dichlorophenyl)- 1.11 30540342 silane silane, diazidomethylphenyl- -0.39 33317656 silane silane, dichlorobis(3-chloropropyl)- -0.08 33434638 silane silane, trichloro(tetrachlorophenyl)- 2.26 58479611 silane silane, chloro(1,1-dimethylethyl)diphenyl- -1.20 64426364 silane silanediamine, N-(chlorodimethylsilyl)-N'-(chlorodiphenylsilyl)-1,1-dimethyl- -3.06 79793003 silane benzenesulfonyl chloride, 4-[2-(trichlorosilyl)ethyl]- 0.67

56337 Silox disiloxane, 1,1,3,3-tetramethyl-1,3-diphenyl- -2.14 540976 Silox cyclohexasiloxane, dodecamethyl- -4.80 556683 Silox hexadecamethylcyclooctasiloxane -8.21

1014660 Silox silane, trimethyl(4-nitrophenoxy)- -0.39 2116849 Silox trisiloxane, 1,1,1,5,5,5-hexamethyl-3-phenyl-3-[(trimethylsilyl)oxy]- -3.94 2351135 Silox disiloxane, 1,3-bis(bromomethyl)-1,1,3,3-tetramethyl- 0.75 2565073 Silox silanediol, diphenyl-, diacetate -3.06 2943706 Silox disiloxane, 1,3-bis(dichloromethyl)-1,1,3,3-tetramethyl- 1.96 3582727 Silox disiloxane, 1,3-dichloro-1,3-dimethyl-1,3-diphenyl- -1.20 4353779 Silox silanol, 1,1,1-trimethyl-, 1-sulfuryl chloridoyl chloride -1.42 5360043 Silox methanol, (1,1,3,3-tetramethyl-1,3-disiloxanediyl)bis-, diacetate -2.30 7288280 Silox pyrimidine, 5-methyl-2,4-bis[(trimethylsilyl)oxy]- -1.75 10448096 Silox heptamethylphenylcyclotetrasiloxane -3.06 10497059 Silox silanol, trimethyl-, phosphate (3:1) -2.30 13176697 Silox 1-tetrasiloxanol, 1,1,3,3,5,5,7,7,7-nonamethyl- (9CI) -2.30 14920924 Silox pentamethylphenyldisiloxane -0.39 17875557 Silox trisiloxane, 1,1,5,5-tetramethyl-3,3-diphenyl- -3.06 17882063 Silox silanol, trimethyl-, benzenesulfonate (7CI,9CI) -0.39 17906091 Silox 1,1,1,3,3,5,7,7,7-Nonamethyl-5-phenyltetrasilo* -3.94 17962344 Silox trisiloxane, 1,1,3,5,5-pentamethyl-3-phenyl- -1.30 18027457 Silox trisiloxane, 3-[(dimethylsilyl)oxy]-1,1,5,5-tetramethyl-3-phenyl- -2.63 18105312 Silox hexanedioic acid, bis(trimethylsilyl) ester -2.71 18132724 Silox disiloxane, 1,3-bis(3-chloropropyl)-1,1,3,3-tetramethyl- 0.63 18306291 Silox silanol, trimethyl-, 1,1'-sulfate -0.65 18407164 Silox 1,1,1,3,3,5,5-Heptamethyl-5-phenyltrisiloxane -2.19 37843268 Silox 2-Butanone, O,O'-(dimethylsilylene)dioxime -1.48 60587102 Silox 1,1,1,3,3,5,7,7,9,9,9-UNDecamethyl-5-phenyipen* -5.66

78637 UC peroxide, (1,1,4,4-tetramethyl-1,4-butanediyl)bis (1,1-dimethylethyl) -3.54 81141 UC ethanone, 1- 4-(1,1-dimethylethyl)-2,6-dimethyl-3,5-dinitrophenyl - -3.51 81152 UC benzene, 1-(1,1-dimethylethyl)-3,5-dimethyl-2,4,6-trinitro- -3.51 83669 UC benzene, 1-(1,1-dimethylethyl)-2-methoxy-4-methyl-3,5-dinitro- -2.63 84479 UC 9,10-anthracenedione, 2-(1,1-dimethylethyl)- -2.14

71

84515 UC 9,10-anthracenedione, 2-ethyl- -1.15 86000 UC 2-nitrobiphenyl -0.23 92922 UC 1,1'-biphenyl -4-carboxylic acid -0.23 93992 UC benzoic acid, phenyl ester -0.23 98737 UC benzoic acid, 4-(1,1-dimethylethyl)- 0.07

101633 UC benzene, 1,1'-oxybis 4-nitro- -2.14 116665 UC 1H-indene, 2,3-dihydro-1,1,3,3,5-pentamethyl-4,6-dinitro- -2.60 118274 UC benzoic acid, 2-[[[2-(ethoxycarbonyl)phenoxy]carbonyl]oxy]-, ethyl ester -5.30 119517 UC 1,2-propanedione, 1-phenyl-, 2-oxime 0.55 136607 UC benzoic acid, butyl ester 0.07 145391 UC benzene, 1-(1,1-dimethylethyl)-3,4,5-trimethyl-2,6-dinitro- -2.63 247165 UC thieno[2,3-B][1]benzothiophene 2.11 247529 UC thieno[3,2-B][1]benzothiophene 2.11 256917 UC 11H-dibenzo[C,F][1,2]diazepine 0.40 303219 UC Phenol, 3-methyl-6-(1-methylethyl)-2,4-dinitro- -1.75 311568 UC 5,6,11,12-tetraoxadispiro[3.2.3.2]dodecane (7CI,8CI,9CI) 2.11 464880 UC cyclopentanecarboxylic acid, 1,2,2,3-tetramethyl- 0.55 501600 UC bis(4-methylphenyl)diazene -0.72 527311 UC 1,2,3,4,5,6-cyclohexanehexone 0.55 535466 UC benzenestibonic acid -1.00 584907 UC diazene, bis(2-methylphenyl)- -0.72 585488 UC 2,6-di-t-butylpyridine -0.39 588045 UC bis(3-methylphenyl)diazene -0.72 599666 UC benzene, 1,1'-sulfonylbis 4-methyl- -1.20 602608 UC anthracene, 9-nitro- -0.65 607012 UC phosphine, ethyldiphenyl- -0.23 620553 UC benzene, 1-nitro-3-phenoxy- -0.72 620882 UC 4-nitro diphenyl ether -0.72 621307 UC benzene, 1-isothiocyanato-3-methyl- 0.35 643652 UC methanone, (3-methylphenyl)phenyl- -0.23 643947 UC 1,2-benzenediol, dibenzoate (9CI) -4.02 691247 UC 2-propanamine, N,N'-methanetetraylbis[2-methyl- -0.18 716767 UC [1,1'-biphenyl]-3-carboxylic acid -0.23 720752 UC 1,1'-biphenyl -4-carboxylic acid, methyl ester -0.72 728405 UC 2,6-ditert.butyl-4-nitro-phenol -2.19 736301 UC benzene, 1,1'-(1,2-ethanediyl)bis 4-nitro- -2.60 829834 UC diphenylarsine 0.55 880933 UC 1-methyl-4-nitronaphthalene 0.28 881038 UC naphthalene, 2-methyl-1-nitro- 0.28 892217 UC 3-nitro-fluoranthen -1.15 943157 UC benzene, 1-methyl-4-(1-methylethyl)-2-nitro- 0.07 943271 UC ethanone, 1- 4-(1,1-dimethylethyl)phenyl - 0.07 945482 UC arsine, methyldiphenyl- (6CI,7CI,8CI,9CI) -0.65 949871 UC diazene, (4-methylphenyl)phenyl- -0.23 954074 UC 9,10-anthracenedione, 1-methyl- -0.65 954461 UC 9-nitrophenanthrene -0.65

1131471 UC 3H-1,5-benzodiazepine, 2,4-dimethyl- 0.79 1195706 UC 1,1-cyclopropanedicarbonitrile, 2,2,3,3-tetramethyl- 0.67 1202364 UC benzene, 1,1'-tellurobis- 0.55 1205067 UC benzoic acid, 4-isothiocyanato-, ethyl ester -0.39 1271541 UC chromium, bis(h6-benzene)- (9CI) 0.55 1320167 UC benzoic acid, (1,1-dimethylethyl)- 0.07 1344305 UC phenol, 2-(1-methylpropyl)dinitro- -1.75 1421494 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)-4-hydroxy- -2.19 1486288 UC phosphine, methyldiphenyl- 0.28 1519364 UC 9,10-Anthracenedione, 1,4-dimethyl- -1.15 1539044 UC 1,4-benzenedicarboxylic acid, diphenyl ester -4.02 1562932 UC benzoic acid, 4-(phenylazo)- -1.20 1605534 UC diethylphenylphosphine 0.67 1620640 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)-4-hydroxy-, ethyl ester -3.07 1636824 UC 2-pentanone, (2,4-dinitrophenyl)hydrazone -2.63 1636835 UC 3-pentanone, (2,4-dinitrophenyl)hydrazone -2.63 1640397 UC 3H-Indole, 2,3,3-trimethyl- 1.33 1719217 UC phenol, 2,4,6-trimethyl-3-nitro- 0.07 1816962 UC benzene, 1,4-dimethyl-2- (4-methylphenyl)sulfonyl - -1.67 1817749 UC benzene, 1,1'-methylenebis 4-nitro- -2.14 1886573 UC benzene, 1-(1,1-dimethylethyl)-2-nitro- 0.07 1942616 UC diazene, (4-isocyanatophenyl)phenyl- -1.20 2033241 UC 1,3-dioxane-4,6-dione, 2,2-dimethyl- 0.35 2113588 UC 3-nitrobiphenyl -0.23

72

2131615 UC benzene, 1-isothiocyanato-4-nitro- 0.55 2131626 UC benzoic acid, 4-isothiocyanoato- 0.55 2155717 UC 1,2-benzenedicarboperoxoic acid, bis(1,1-dimethylethyl) ester -3.94 2167239 UC peroxide, (1-methylpropylidene)bis (1,1-dimethylethyl) -1.89 2273510 UC stannane, oxodiphenyl- -0.65 2348176 UC 2-hexanone, (2,4-dinitrophenyl)hydrazone -3.07 2359093 UC Isophthalic acid, 5-tert-butyl- -1.30 2444293 UC p-nitrophenyl o-tolyl ether -1.20 2491523 UC 4-nitroazobenzene -1.20 2510556 UC 9-phenanthrenecarbonitrile -0.13 2511220 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)-4-hydroxy-, methyl ester -2.63 2523480 UC 9H-fluorene-2-carbonitrile 0.40 2622891 UC borinic acid, diphenyl- 0.28 2852688 UC methanone, bis(3-methylphenyl)- -0.72 3056642 UC 4-tert-butylphenol, Acetate -0.39 3077972 UC 2-butanone, 3-methyl-, (2,4-dinitrophenyl)hydrazone -2.63 3282562 UC benzene, 1-(1,1-dimethylethyl)-4-nitro- 0.07 3285981 UC 9,10-anthracenedione, 1,2-dimethyl- -1.15 3382567 UC 4-t-butylnitrobenzene 0.07 3402742 UC benzene, 1-methyl-4-(4-nitrophenoxy)- -1.20 3457612 UC peroxide, 1,1-dimethylethyl 1-methyl-1-phenylethyl -0.85 3529826 UC benzene, 1-isothiocyanato-3-nitro- 0.55 3581702 UC 1-propanol, 2,2-dimethyl-, benzoate -0.39 3581735 UC 1-propanol, 2,2-dimethyl-, 4-nitrobenzoate -1.75 3646579 UC diazene, bis(4-nitrophenyl)- -2.60 3682562 UC benzoic acid, 2-(phenylazo)- -1.20 3709180 UC 1,3-dioxane-4,6-dione, 2,2,5-trimethyl- 0.67 4044659 UC benzene, 1,4-diisothiocyanato- 0.55 4067145 UC benzoic acid, 4-(1,1-dimethylethyl)-, aluminum salt 0.07 4097498 UC 4-tert-butyl-2,6-dinitrophenol -1.75 4171339 UC diazene, bis(2-nitrophenyl)- -2.60 4196898 UC 1,3-propanediol, 2,2-dimethyl-, dibenzoate -3.97 4237405 UC benzene, 1-(1-methylpropyl)-4-nitro- 0.07 4262617 UC peroxide, 1,1'-(1-methylethylidene)bis[2-(1,1-dimethylethyl) -1.48 4507986 UC 1,2-dioxolane, 3,3'-dioxybis[3,5,5-trimethyl- (8CI,9CI) -1.67 4519328 UC diarsene, diphenyl- (9CI) 0.55 4731366 UC tri(sec-butyl) aluminum -0.24 5097121 UC benzene, 1,1'-sulfonylbis[2-methyl- -1.20 5180596 UC 2,6-bis(1,1-dimethylethyl)-4-nitrobenzenamine -2.19 5393497 UC benzenamine, N,N'-(1,2-dimethyl-1,2-ethanediylidene)bis- -1.67 5406575 UC benzoic acid, 4-(1,1-dimethylethyl)-, ethyl ester -0.85 5410979 UC 3-nitrodibenzofuran -0.13 5465134 UC benzene, 1,2,4,5-tetramethyl-3,6-dinitro- -1.30 5617709 UC 5,7-dioxaspiro[2.5]octane-4,8-dione, 6,6-dimethyl- 1.33 5689190 UC benzenemethanol, a,a-dicyclopropyl- 0.95 5737133 UC 4H-cyclopenta(DEF)phenanthren-4-one 0.45 5892091 UC bismuthine, (benzoyloxy)oxo- -0.59 5950834 UC benzene, 1-nitro-2-(4-nitrophenoxy)- -2.14 6002342 UC phosphine, (1,1-dimethylethyl)diphenyl- -1.20 6099805 UC benzene, 1-(1,1-dimethylethyl)-2-methoxy-3,5-dinitro- -2.19 6338096 UC 9,10-anthracenedione, 1-mercapto- -0.65 6531357 UC 9,10-anthracenedione, 2,3-dimethyl- -1.15 6575140 UC Benzonitrile, 2,6-diethyl- 0.55 6676900 UC (2-methylphenyl)phenyldiazene -0.23 6720281 UC diazene, (4-methylphenyl)phenyl-, (Z)- -0.23 6720394 UC diazene, (4-methylphenyl)phenyl-, (E)- -0.23 7030184 UC 2-methyl-4'-nitroazobenzene -1.67 7137555 UC benzene, 1-butyl-2-nitro- 0.07 7373264 UC benzene, 1,3,5-triisocyanato-2-methyl- -1.30 7470146 UC 3-phenanthrenecarboxylic acid -0.65 7507484 UC 1,4-benzenediol, 2-(1,1-dimethylethyl)-, diacetate (9CI) -2.19 7508681 UC benzoic acid, 4-(phenylazo)-, ethyl ester -2.14 7599237 UC 1,1'-biphenyl, 2-azido- -0.23 7612966 UC 4-Isothiocyanoazobenzene -1.20 7621161 UC phosphine, (1,1-dimethylethyl)methylphenyl- 0.55 10342606 UC benzene, 1-(2-methylpropyl)-4-nitro- 0.07 10605240 UC 1,3-propanediol, 2-methyl-2-propyl-, dinitrate (7CI,8CI,9CI) -1.48 13009911 UC 1,3,5-triazine, 2,4,6-tris(2-methyl-1-aziridinyl)- -0.63

13102324 UC propaneperoxoic acid, 2,2-dimethyl-, 1,1,4,4-tetramethyl-1,4-butanediyl ester (9CI) -5.19

73

13177281 UC nitrofluoranthene -1.15 13246327 UC arsine, As,As'-1,2-phenylenebis[As,As-dimethyl- -0.59 13471697 UC benzene, 4-isocyanato-1-methyl-2-nitro- 0.07 13472087 UC Butanenitrile, 2,2'-azobis[2-methyl- -1.06 13509336 UC carbonothioIic acid, O,S-diphenyl ester -0.72 13509369 UC carbonodithioIic acid, S,S-diphenyl ester -0.72 13615388 UC 2-methyl-8-nitronaphthalene 0.28 13653628 UC Peroxide, (1-methylpropylidene)bis (1,1-dimethylpropyl) -2.71 14213015 UC 1,1'-biphenyl, 3-azido- -0.23 14284936 UC ruthenium, tris(2,4-pentanedionato-o,o )-, (oc-6-11)- -3.95 14387305 UC benzoic acid, 4,4'-oxybis-, 1,1'-dimethyl ester -3.06 15085208 UC 4-Indancarbonitrile, 7-methyl- 1.33 15128504 UC dibenz[B,E]oxepin-6,11-dione -0.65 15231784 UC 1,3-dioxane-4,6-dione, 2,2-dimethyl-5-phenyl- -0.72 15815391 UC 9,10-Anthracenedione, 1,5-dimethyl- -1.15 15986922 UC pyrazine, 2,3-dihydro-5,6-dimethyl- -0.25 15986933 UC pyrazine, 2-ethyl-5,6-dihydro-3-methyl- 0.04 16225266 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)- -1.75 16268987 UC 1,3,5-triazine-2,4,6-triamine, N,N',N''-tris(1,1-dimethylethyl)- (9CI) -3.51 16424661 UC 4-HeptaNone, 2,2,3,3,5,5,6,6-Octamethyl- -1.89 16714184 UC benzene, 2-azido-1-methyl-3-nitro- 0.07 16738888 UC 1,1-cyclopropanedicarbonitrile, 2-methyl-2-propyl- 0.67 16738899 UC 1,1-cyclopropanedicarbonitrile, 2-butyl-2-methyl- 0.55 16888758 UC 1,2-ethanediamine, N,N'-bis(1-methylethylidene)- -0.59 16968197 UC benzene, 1,1'-(1,2-ethanediyl)bis[2-nitro- -2.60 17024189 UC 2-nitrophenanthrene -0.65 17082121 UC diazene, diphenyl-, (E)- 0.28 17484285 UC benzoic acid, 3-(1,1-dimethylethyl)-5-methyl- -0.39 18213735 UC 2,4-dimethoxy-6-phenyl-S-triazine -0.72 18264954 UC 2-phenanthrylamine, 9,10-dihydro-1,3-dinitro- -2.62 18277913 UC benzoic acid, 2-(phenylazo)-, ethyl ester -2.14 18709518 UC pyrazine, 2,5-bis(1,1-dimethylethyl)- -0.39 18794393 UC pyridine, 4-(1,1-dimethylethyl)-3-[(1,1-dimethylethyl)thio]- -0.85 18794462 UC 3-picoline, 6-(tert-butylthio)- 0.55 18794484 UC pyridine, 2-[(1,1-dimethylethyl)thio]-3,5-dimethyl- 0.07 18833879 UC 3- picoline, 2-(tert-butylthio)- 0.55 19370344 UC benzene, 1-(1-methylpropyl)-2-nitro- 0.07 19377958 UC BIcyclo[3.1.0]hexaN-2-one, 1,5-bis(1,1-dimethylethyl)-3,3-dimethyl- -1.20 19694123 UC diazene, bis(1,1-dimethylpropyl)- -0.24 19715196 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)-2-hydroxy- -2.19 19881186 UC benzene, 1-isothiocyanato-4-(4-nitrophenoxy)- -2.14 20332109 UC stannane, diphenylthioxo- (8CI,9CI) -0.65 20615752 UC benzene, 1-azido-4-methyl-2-nitro- 0.07 20651756 UC benzene, 1-butyl-4-nitro- 0.07 20651767 UC benzene, 1-butyl-3-nitro- 0.07 20927951 UC 2-nitrodibenzofuran -0.13 21578949 UC 1,4-benzenedicarboperoxoic acid, bis(1,1-dimethylethyl) ester (9CI) -3.94 21850408 UC hexanediperoxoic acid, 2,4,4-trimethyl-, bis(1,1-dimethylethyl) ester (9CI) -4.78 22397337 UC 1,2,4,5-Tetroxonane, 3,3,6,6,9,9-hexamethyl- -0.85 23132527 UC benzene, 1-(1,1-dimethylethyl)-3-nitro- 0.07 23383597 UC propaneperoxoic acid, 2,2-dimethyl-, 1-methyl-1-phenylethyl ester -1.75 24582545 UC methyl diphenylarsinite -1.06 24624291 UC 9,10-anthracenedione, 1-ethyl- -1.15 25005952 UC pyrazine, 2,3-dihydro-5,6-di-2-pyridyl- -1.15 25149466 UC propanenitrile, 2- (1,1-dimethylethyl)azo -2-methyl- -0.18 25722661 UC cyanic acid, (1-methylethylIdene)di-4,1-phenylE* -2.14 26328596 UC 4-Isothiocyanobenzophenone -1.20 26444202 UC AR,AR'-azotoluene -0.72 27253334 UC neodecanoic acid, calcium salt -5.19 27936341 UC methylanthracene-9,10-dione -0.65 28122841 UC benzenecarbothioIC acid, 4-nitro-, S-(4-methylphenyl) ester -2.14 28361435 UC benzene, 1,1'-sulfonylbis 3,4-dimethyl- -2.14 28842059 UC diazene, (2-methylphenyl)(3-methylphenyl)- (9CI) -0.72 28984852 UC nitrobiphenyl -0.23 29418222 UC o,p'-azotoluene -0.72 29418573 UC 2-methyl-4-nitroazobenzene -1.67 29418584 UC (4-methylphenyl)(4-nitrophenyl)diazene -1.67 30026927 UC hydroperoxide, 1-[(1,1-dimethylethyl)phenyl]-1-methylethyl (9CI) -0.85 30448432 UC 1,2-benzenedicarboxylic acid, 1,2-bis(1,1-dimethylethyl) ester -3.07 30863651 UC 2-(diazomethyl)-4,6-bis(methylthio)-S-triazine 0.07

74

31656914 UC 1,1'-biphenyl, 4-azido- -0.23 32294603 UC ditelluride, diphenyl 0.55 32366260 UC benzene, (1-azido-1-methylethyl)- 0.55 32588548 UC 9,10-anthracenedione, 2-(1,1-dimethylpropyl)- -2.62 32673259 UC phosphine, bis(1,1-dimethylethyl)phenyl- -0.85 32703790 UC 1,3-isobenzofurandione, 5-(1,1-dimethylethyl)- -0.23 33720740 UC azete, 2,3-dihydro-4-phenyl- 1.28

34296554 UC spiro[cyclopropane-1,9'-[9H]fluorene]-2,3-dicarboxylic acid, dimethyl ester, cis- (9CI) -3.15

34372093 UC pentanenitrile, 2,4-dimethyl-2-(phenylazo)- -1.30 34874878 UC benzene, 2-methoxy-1,3,5-trimethyl-4-nitro- -0.39 35147456 UC 1,2-dioxolane, 3,5-bis[(1,1-dimethylethyl)dioxy]-3,5-dimethyl- -2.63 36050927 UC phosphine, methyl(1-methylethyl)phenyl- 0.67 36210716 UC methanone, (4-azidophenyl)phenyl- -1.20 36760437 UC carbonothioic acid, O-(1,1-dimethylethyl) S-phenyl ester -0.39 37467655 UC ethanone, 1-(3,5-diethyl-2,6-dihydroxyphenyl)- -0.85 38675795 UC methanone, cyclopropyl[4-(1,1-dimethylethyl)phenyl]- -0.23 39049053 UC neodecanoic acid, titanium salt -0.24 39950931 UC 3,4-hexanedione, 2,2,5,5-tetramethyl-, monooxime -0.65 40515182 UC benzene, 4-azido-1-methyl-2-nitro- 0.07 40515193 UC benzene, 2-azido-1-methyl-4-nitro- 0.07 40888979 UC 2-propanol, 2,2'-azobis-, diacetate (ester) (9CI) -1.89 41037132 UC 1-methyl-3-nitronaphthalene 0.28 41825289 UC plumbane, bis(nitrooxy)diphenyl- (9CI) -1.89 41902425 UC 3-pentanol, 3-(1,1-dimethylethyl)-2,2,4,4-tetramethyl- -1.06 41981708 UC thiazole, 2,4-dimethyl-5-propyl- 0.35 41981719 UC thiazole, 2,5-diethyl-4-methyl- 0.35 41981720 UC thiazole, 4,5-dimethyl-2-propyl- 0.35 41981753 UC thiazole, 4-ethyl-2-methyl-5-propyl- -0.18 43052397 UC pentanedioic acid, bis(1,1-dimethylethyl) ester (9CI) -2.30 43126836 UC tert-dodecanethiol, silver(1+) salt -0.65 43133944 UC ethanone, 1-[(1,1-dimethylethyl)phenyl]- 0.07 49540854 UC 4-isothiocyanophenyl benzoate -1.67 49556163 UC neodecanoic acid, tin(2+) salt -0.24 50670503 UC 1,1'-biphenyl -4-carbonitrile, 4'-methyl- -0.23 51677419 UC butane, 2-azido-2,3,3-trimethyl- -0.59 51818554 UC neodecanoic acid, iron salt -0.24 51818565 UC neodecanoic acid, nickel salt -5.19 52235208 UC butanenitrile, 2- (1,1-dimethylethyl)azo -2-methyl- -0.24 52270447 UC neodecanoic acid, cobalt(2+) salt -0.24 52414898 UC thiazole, 2,4-diethyl-5-methyl- 0.35 52625259 UC benzoic acid, 3,5-bis(1,1-dimethylethyl)-4-hydroxy-, nickel(2+) salt (2:1) -2.19 52909883 UC benzenecarbothioIC acid, 2-nitro-, S-(4-methylphenyl) ester -2.14 52909894 UC benzenecarbothioIC acid, 3-nitro-, S-(4-methylphenyl) ester -2.14 53044487 UC benzene, 1-(2-ethylphenoxy)-2,4-dinitro- -3.06 53274194 UC benzeneacetic acid, 4-nitro-, 4-methylphenyl ester -2.60 53498309 UC 2-isopropyl-4,5-dimethylthiazole -0.59 53498310 UC thiazole, 5-ethyl-4-methyl-2-(1-methylethyl)- 0.67 54106400 UC benzene, 1-methyl-2-(2-nitrophenoxy)- -1.20 54187198 UC methanone, bis(2-methylcyclopropyl)-, oxime 1.53 54357083 UC 2-methyl-6-nitronaphthalene 0.28 54518091 UC benzo[B]thiophen-2(3H)-one, 3-diazo- 1.33 54644416 UC propanoic acid, 2,2-dimethyl-, 2-(1,1-dimethylethyl)phenyl ester -1.75 54738931 UC nitroanthracene -0.65 54755203 UC 2-methyl-5-nitronaphthalene 0.28 54755214 UC 2-methyl-8-nitronaphthalene 0.28 55044525 UC benzoic acid, 4-methyl-, 4-(methoxycarbonyl)phenyl methyl ester -3.06 55044694 UC 2-butanol, 2-methyl-3-(nitroimino)-, benzoate (ester) -2.19 55076363 UC benzoic acid, 2-hydroxy-6-nitro-, phenyl ester -2.14 55172980 UC neodecanoic acid, barium salt -5.19 55203593 UC 3H-Indole, 2,3,3-trimethyl-5-(phenylsulfonyl)- -2.62 55591112 UC 1H-Indene-4-carboxylic acid, 2,3-dihydro-1,1-dimethyl-, methyl ester -0.23 55712384 UC 1H-Indene-4-carboxylic acid, 2,3-dihydro-1,1-dimethyl- 0.28 55794202 UC butanoic acid, 3,3-bis[(1,1-dimethylethyl)dioxy]-, ethyl ester -3.54 55912180 UC pentanenitrile, 2- (1,1-dimethylethyl)azo -2,4-dimethyl- -1.06 56623337 UC 2-propanol, 2,2'-azobis-, dipropanoate (ester) (9CI) -2.71 57908482 UC 2-butanol, 2,2'-azobis-, diacetate (ester) (9CI) -2.71 57910799 UC 2-butanol, 2- (1,1-dimethylethyl)azo - -0.18 58743752 UC 1,1'-biphenyl -4-carbonitrile, 4'-ethyl- -0.72 60263447 UC pyridine, 2-(3,3-dimethylbutyl)- 0.55

75

60451072 UC neodecanoic acid, vanadium salt -0.24 60633218 UC phosphine, bis(1,1-dimethylethyl)(2,2-dimethylpropyl)- -1.06 61142516 UC phosphine, (2,2-dimethylcyclopropyl)diphenyl- -1.15 61951960 UC neodecanoic acid, cadmium salt -5.19 62245470 UC 2-amino-7-nitrophenanthrene -1.15 62337911 UC phosphine, bis(2,2-dimethylcyclopropyl)phenyl- -0.65 62637983 UC 1-naphthalenol, 2-(phenylazo)-, acetate (ester) -3.09 63017878 UC 1-methylnitronaphthalene 0.28 63042080 UC carbonic acid, methyl 2-methyl-4-[(2-methylphenyl)azo]phenyl ester (9CI) -3.06 63084139 UC 1H-Indene-5-carbonitrile, 2,3-dihydro-1,1,2,3,3,6-hexamethyl- -1.20 63906558 UC 3,6-diethyl-3,6-dimethyl-1,2,4,5-tetroxane 0.55

64047274 UC Iron, [[3,3'-(1,2-ethanediyldinitrilo)bis[2-butanone] dioximato](2-)-N,N',N'',N''']- (9CI) -1.89

67801530 UC 1,4-benzenedicarboxylic acid, 2-(4-methylphenyl)-, dimethyl ester -3.06 67801563 UC 1,1'-biphenyl -2,4'-dicarboxylic acid, 5-methyl-, dimethyl ester -3.06 68132956 UC borate(1-), dimethyldiphenyl-, lithium, (T-4)- -0.23 68425592 UC benzene, (1-methylpropyl)nitro- 0.07 68455925 UC nitrophenanthrene -0.65 68459814 UC benzene, 1-(1-methylpropyl)-3-nitro- 0.07 68683181 UC neodecanoic acid, silver(1+) salt -0.24 68958521 UC neodecanoic acid, iron(3+) salt -0.24 71735198 UC phenol, 2-(1-methylpropyl)-4,6-dinitro- -1.75 71850776 UC benzene, 1-(1,1-dimethylethyl)-3,5-dimethyl-2,4-dinitro- -2.19 74630005 UC benzene, 2-(1,1-dimethylethyl)-4-methyl-1-(1-methylethyl)-3,5-dinitro- -3.07

74810392 UC tricyclo[5.1.0.02,4]octane-5-carboxylic acid, 3,3,8,8-tetramethyl-, methyl ester (9CI) -0.13

75790828 UC benzene, 1-methyl-2-nitro-4-[(4-nitrophenyl)methyl]- -2.60 77468361 UC nitrofluoranthene -1.15 80182270 UC dimethylnitrophenanthrene -1.65 80191442 UC methylnitrophenanthrene -1.15 85895814 UC pyridine, 2-(1-ethyl-1-methylpropyl)- 0.55 91137278 UC 1-methyl-5-nitronaphthalene 0.28 91137289 UC 2-methyl-7-nitronaphthalene 0.28 94158062 UC [1,1'-biphenyl]-2-ol, 2'-azido-, acetate (ester) (9CI) -2.14

105752678 UC 1-methyl-6-nitronaphthalene 0.28 116211951 UC naphthalenetetrone, methyl -0.23 116211984 UC hydroxynitrophenanthrene -1.15 116530075 UC 1-methyl-7-nitronaphthalene 0.28

91134 XAB benzene, 1,2-bis(bromomethyl)- 1.68 93527 XAB benzene, (1,2-dibromoethyl)- 1.68 94995 XAB benzene, 2,4-dichloro-1-(chloromethyl)- 2.40

102476 XAB benzene, 1,2-dichloro-4-(chloromethyl)- 2.40 589151 XAB benzene, 1-bromo-4-(bromomethyl)- 1.65 589173 XAB benzene, 1-bromo-4-(chloromethyl)- 1.65 612124 XAB benzene, 1,2-bis(chloromethyl)- 1.68 623245 XAB benzene, 1,4-bis(bromomethyl)- 1.68 623256 XAB benzene, 1,4-bis(chloromethyl)- 1.68 626153 XAB benzene, 1,3-bis(bromomethyl)- 1.68 823789 XAB benzene, 1-bromo-3-(bromomethyl)- 1.65

2014837 XAB benzene, 1,3-dichloro-2-(chloromethyl)- 2.40 2387180 XAB benzene, 1,4-bis(chloromethyl)-2-methyl- 1.79 2735059 XAB benzene, 2,4-bis(chloromethyl)-1-methyl- 1.79 2745495 XAB benzene, 1,4-dichloro-2-(chloromethyl)- 2.40 3290015 XAB benzene, 1,2-dichloro-3-(chloromethyl)- 2.40 3290060 XAB benzene, 1,3-dichloro-5-(chloromethyl)- 2.40 3335339 XAB benzene, 1-methyl-2-(trichloromethyl)- 2.38 3335340 XAB M-xylene, .alpha.,.alpha.,.alpha.-trichloro- 2.38 3335351 XAB benzene, 1-methyl-4-(trichloromethyl)- 2.38 3433805 XAB benzene, 1-bromo-2-(bromomethyl)- 1.65 3883134 XAB benzene, (2,2,2-trichloroethyl)- 2.38 6298722 XAB benzene, 1,4-bis(chloromethyl)-2,5-dimethyl- 1.95 6529539 XAB benzene, 1-(2-bromoethyl)-4-chloro- 1.68 7398449 XAB benzene, 1-(chloromethyl)-4-(dichloromethyl)- 2.38 16799056 XAB benzene, 1-(2-bromoethyl)-3-chloro- 1.68 18880007 XAB benzene, 1-(bromomethyl)-4-(1,1-dimethylethyl)- 1.48 20443985 XAB benzene, 2-(bromomethyl)-1,3-dichloro- 2.40 28347139 XAB benzene, bis(chloromethyl)- 1.68 31195172 XAB benzene, tribromoethyl- 2.38 38721710 XAB benzene, dichloro(chloromethyl)- 2.40 53459408 XAB benzene, 1-(2-chloroethyl)-4-(chloromethyl)- 1.79

76

54789309 XAB benzene, 1-(1-chloroethyl)-4-(chloromethyl)- 1.79 54789310 XAB benzene, 1-(1,2-dichloroethyl)-4-methyl- 1.79 54965003 XAB benzene, 1,4-dichloro-2-(1-chloroethyl)- 2.38 54965014 XAB benzene, 1,2-dichloro-4-(1-chloroethyl)- 2.38 54965025 XAB benzene, 1,4-dichloro-2-(2-chloroethyl)- 2.38 59216170 XAB benzene, 2,4-dibromo-1-(2-bromoethyl)- 2.38 59473459 XAB benzene, 1-(chloromethyl)-2-iodo- 1.65 74298945 XAB benzene, 1-chloro-4-(1,2-dichloroethyl)- 2.38 1667103 XABP 1,1'-biphenyl, 4,4'-bis(chloromethyl)- 1.48 1940574 XABP 9H-fluorene, 9-bromo- 2.20 24802991 XAH 2-methylamino-4-phenyl-6-trichloromethyl-S-triA* 1.38 24848484 XAH 3-(sec-butyl(4-methyl-6-(trichloromethyl)-S-*)) 0.22 24848519 XAH 3-(methyl(4-propyl-6-(trichloromethyl)-S-tri*)) 0.22 30339743 XAH 1,3,5-triazin-2-amine, 4-pentyl-6-(trichloromethyl)- 1.88 30357392 XAH 2-(tert-butylamino)-4-methyl-6-trichloromethyl* 1.88 30357427 XAH 2-methyl-4-((1,1,3,3-tetramethylbutyl)amino)-6-* -0.31 30362335 XAH 2-methyl-4-phenyl-6-(trichloromethyl)-S-triAZI* 1.98 30362471 XAH alpha,alpha-dimethyl-4,6-bis(trichloromethyl)-*- 1.90 30369165 XAH 2-amino-4-(P-tolyl)-6-(trichloromethyl)-S-triAZ* 1.38 30369574 XAH 2-amino-4-((P-chlorophenyl)thio)-6-(trichloro*) 1.80 30369585 XAH 2-((P-chlorophenyl)thio)-4-(methylamino)-6-(T*) 1.21 30576319 XAH 2-amino-4-(2,4-dichlorophenyl)thio-6-trichloro* 2.18 30863606 XAH 2-methoxy-4-phenyl-6-trichloromethyl-S-triazine 1.38 30863628 XAH 2-(P-chlorophenyl)-4-methoxy-6-trichloromethyl* 1.80

103112352 XAH ethyl-1-(2,4-dichlorophenyl)-5-trichloromethyl-* 0.99 75956 XAlka pentabromoethane 2.98

354472 XAlka ethane, pentabromofluoro- 3.32 422264 XAlka 1,1,1,2,2,3-hexachloro-3-fluoropropane 3.25 422286 XAlka 1,1,2,2,3,3-hexachloro-1-fluoropropane 3.25 422402 XAlka 1,1,1,2,3,3-hexachloro-2-fluoropropane 3.25 422786 XAlka 1,1,1,2,2,3,3-heptachloro-3-fluoropropane 3.52 422811 XAlka 1,1,1,2,3,3,3-heptachloro-2-fluoropropane 3.52 431798 XAlka 1,1,1,2,3,3-hexachloro-3-fluoropropane 3.25 507255 XAlka methane, tetraiodo- 3.39 558167 XAlka bromotriiodomethane 3.39 594218 XAlka 1,1,1-triiodoethane 1.88 594672 XAlka pentaiodoethane 2.98 594730 XAlka hexabromoethane 3.32 594898 XAlka 1,1,1,2,2,3,3-heptachloropropane 3.25 594901 XAlka octachloropropane 3.52 629094 XAlka hexane, 1,6-diiodo- 0.16 630251 XAlka ethane, 1,2-dibromo-1,1,2,2-tetrachloro- 3.32

1522889 XAlka 2,2-bis(jodmethyl)-1,3-dijodpropan 1.62 1529686 XAlka Butane, 1,2,3,4-tetrabromo- 1.74 2657672 XAlka 1,2,3,4-tetrabromobutane 1.74 3228997 XAlka propane, 1,3-dichloro-2,2-bis(chloromethyl)- 1.62 3229003 XAlka propane, 1,3-dibromo-2,2-bis(bromomethyl)- 1.62 3607781 XAlka 1,1,1,3,3,3-hexachloropropane 2.93 3849330 XAlka propane, 1,1,1,2,3,3,3-heptachloro- 3.25 14059906 XAlka dibromodiiodomethane 3.39 14349827 XAlka chlorotriiodomethane 3.39 21981339 XAlka Butane, 1,1,1,3,3-pentachloro- 2.28 24425976 XAlka 1,1,1,2,2,3-hexachloropropane 2.93 54268029 XAlka 1,2,2,3-tetrabromopropane 2.02 58443860 XAlka hexane, 1,2,5,6-tetrabromo- 1.61 79504022 XAlka bromopentachloroethane 3.32 83293827 XAlka 1,2,2,3,3-pentachlorobutane 2.28

90993 XBDP benzene, 1,1'-(chloromethylene)bis- 1.48 101768 XBDP benzene, 1,1'-methylenebis 4-chloro- 2.08 134838 XBDP benzene, 1-chloro-4-(chlorophenylmethyl)- 2.08 776749 XBDP benzene, 1,1'-(bromomethylene)bis- 1.48

2642800 XBDP 1,1-bis(p-chlorophenyl)-2-chloroethane 1.98 5789300 XBDP benzene, 1,1'-(1,2-dibromo-1,2-ethanediyl)bis- 1.48 5963495 XBDP 1,2-dichloro-1,2-diphenylethane 1.48 12110393 XBDP methylium, chlorodiphenyl-, (OC-6-11)-hexachloroantimonate(1-) 1.48 13391394 XBDP benzene, 1-chloro-3-(chlorophenylmethyl)- 2.08 24161146 XBDP methylium, chlorodiphenyl- (8CI,9CI) 1.48 25249392 XBDP bis(chlorophenyl)methane 2.08 56961478 XBDP benzene, 1-chloro-2-(chlorophenylmethyl)- 2.08 59485346 XBDP benzene, 1,1'-(1,2-ethanediyl)bis[2-bromo- 1.48

77

86986 XH quinoline, 4,7-dichloro- 2.89 3141240 XH thiophene, 2,3,5-tribromo- 0.95 3842555 XH 2-chloro-4,6-diphenyl-S-triazine -0.65 4203183 XH quinoline, 4,6-dichloro- 2.89 4350418 XH pyridine, 2-[(4-chlorophenyl)methyl]- 1.48 4409114 XH pyridine, 4-[(4-chlorophenyl)methyl]- 1.48 6484259 XH quinazoline, 4-chloro-2-phenyl- 0.43 13838341 XH 2,4-dichloro-6-(P-tolyloxy)-S-triazine 1.48 20815547 XH pyridine, 4-[(P-bromophenyl)azo]- 0.92 21039781 XH cinnoline, 6-chloro-4-phenyl- 0.43 25117504 XH pyridine, 2-[(P-bromophenyl)azo]- 0.92 25117515 XH quinoline, 2-[(P-bromophenyl)azo]- -0.65 25117526 XH quinoline, 7-[(P-bromophenyl)azo]- -0.65 25187193 XH quinoxaline, 6-chloro-2-phenyl- 0.43 30358168 XH 2-amino-4,6-bis(P-chlorophenoxy)-S-triazine -1.78 30369212 XH 2-amino-4,6-bis(P-chlorophenyl)-S-triazine -0.73 30894770 XH 2-chloro-4,6-bis(O-chlorophenoxy)-S-triazine -0.88 30894929 XH 2-chloro-4-(diazomethyl)-6-phenyl-S-triazine 0.38 30894930 XH 2-chloro-4-(P-chlorophenyl)-6-phenyl-S-triazine -0.19 30894963 XH 2,4-dibromo-6-phenoxy-S-triazine 2.08 31721923 XH 5H-benzo[4,5]cyclohepta[1,2-b]pyridin-5-one, 7-chloro-10,11-dihydro- 0.43 59666169 XH quinoline, 2,7-dichloro-4-methyl- 3.01 63905964 XH 4-pyridinethiol, 3,5-diiodo- 1.65 74129150 XH pyridine, 2-[(4-bromophenyl)methyl]- 1.48 464415 XNOR bicyclo[2.2.1]Heptane, 2-chloro-1,7,7-trimethyl-, endo- 2.16

22819701 XNOR tricyclo[5.1.0.03,5]octane, 8,8-dichloro-1,4,4-trimethyl-, (1a,3b,5b,7a)- (9CI) 3.86

30462534 XNOR bicyclo[2.2.1]heptane, 2-chloro-1,7,7-trimethyl-, (1R,2S,4R)- 2.16 259790 XPAH biphenylene 2.11

17219942 XPAH 9,10-dichlorophenanthrene 2.15 59116880 XPAH dichlorophenanthrene, isomer 2.15 86329604 XPAH dichlorofluoranthene, isomer 1.52

80079 XUC benzene, 1,1'-sulfonylbis 4-chloro- 0.91 81867 XUC 1H,3H-naphtho 1,8-cd pyran-1,3-dione, 6-bromo- 1.00 82439 XUC 9,10-anthracenedione, 1,8-dichloro- 0.94 82440 XUC 9,10-anthracenedione, 1-chloro- 0.43 82462 XUC 9,10-anthracenedione, 1,5-dichloro- 0.94 83056 XUC 2,2-bis(p-chloro-phenyl)acetic acid 0.35 84457 XUC 9,10-anthracenedione, 2,3-dichloro- 0.94 85290 XUC methanone, (2-chlorophenyl)(4-chlorophenyl)- 1.48 85461 XUC 1-naphthalenesulfonyl chloride 0.79 90904 XUC methanone, (4-bromophenyl)phenyl- 0.92 90982 XUC methanone, bis(4-chlorophenyl)- 1.48 93118 XUC 2-naphthalenesulfonyl chloride 0.79 94177 XUC peroxide, bis(4-chlorobenzoyl) -0.71

104245 XUC benzoyl chloride, 4-(phenylazo)- -0.15 117237 XUC 3H-pyrazol-3-one, 2-(2-chloro-6-methylphenyl)-2,4-dihydro-5-methyl- 0.92 129351 XUC 9,10-anthracenedione, 1-chloro-2-methyl- -0.11 133142 XUC peroxide, bis(2,4-dichlorobenzoyl) 0.08 134850 XUC methanone, (4-chlorophenyl)phenyl- 0.92 135126 XUC benzene, 4-chloro-1-(4-chlorophenoxy)-2-nitro- 0.35 322758 XUC benzene, 1-(4-chlorophenoxy)-2-nitro-4-(trifluoromethyl)- 0.63 330938 XUC di-(P-fluorphenyl)-ether 2.08 340578 XUC mecloqualone -0.65 342234 XUC methanone, bis(2-fluorophenyl)- 1.48 342256 XUC methanone, (2-fluorophenyl)(4-fluorophenyl)- 1.48 345551 XUC ethanone, 1-(2'-fluoro[1,1'-biphenyl]-4-YL)- 0.38 345926 XUC 4,4'-difluorobenzophenone 1.48 447314 XUC ethanone, 2-chloro-1,2-diphenyl- 0.38 569062 XUC 9,10-anthracenedione, 1-fluoro- 0.43 602211 XUC 1,2-benzenedicarboxylic acid, 3,4,5,6-tetrachloro-, monoethyl ester (9CI) 1.75 602255 XUC 9,10-anthracenedione, 1,4-dichloro- 0.94 602379 XUC naphthalene, 1-chloro-8-nitro- 1.48 602733 XUC 9,10-anthracenedione, 1,3-dichloro- 0.94 605403 XUC 9,10-anthracenedione, 2,6-dichloro- 0.94 605436 XUC 9,10-anthracenedione, 2,7-dichloro- 0.94 605618 XUC naphthalene, 1-chloro-4-nitro- 1.48 605630 XUC naphthalene, 1-chloro-5-nitro- 1.48 606008 XUC benzoic acid, 2-amino-3,5-dibromo-, methyl ester 2.16 607227 XUC naphthalene, 1-chloro-2-nitro- 1.48

78

607374 XUC naphthalene, 7-chloro-1-nitro- 1.48 625887 XUC thiophene, 2,5-diiodo- 1.97 631936 XUC benzenesulfonamide, 4-chloro-N-(4-chlorophenyl)-N-methyl- -0.19 632837 XUC 1-bromoantrhraquinone 0.43 677678 XUC acetyl fluoride, difluoro(fluorosulfonyl)- 0.64 712481 XUC diphenylchloroarsine 0.55 720741 XUC ethanone, 1-(4'-fluoro[1,1'-biphenyl]-4-YL)- 0.38 727515 XUC 9,10-anthracenedione, 1,6-dichloro- 0.94 727991 XUC methanone, phenyl[2-(trifluoromethyl)phenyl]- 1.38 728814 XUC methanone, phenyl[3-(trifluoromethyl)phenyl]- 1.38 728869 XUC methanone, phenyl[4-(trifluoromethyl)phenyl]- 1.38 734598 XUC phosphine, (4-bromophenyl)diphenyl- -1.17 790410 XUC benzoic acid, 4-chloro-, anhydride with 4-chlorobenzoic acid -0.19 958009 XUC 11H-dibenzo[C,F][1,2]diazepin-11-one, 3,8-dichloro- 0.94

1016779 XUC methanone, (3-bromophenyl)phenyl- 0.92 1016780 XUC methanone, (3-chlorophenyl)phenyl- 0.92 1079669 XUC phosphinous chloride, diphenyl- 0.79 1084986 XUC 11H-dibenzo[C,F][1,2]diazepine, 3,8-dichloro- 1.53

1107002 XUC 1,3-Isobenzofurandione, 5,5'- 2,2,2-trifluoro-1-(trifluoromethyl)ethylidene bis- -2.25

1111917 XUC germane, triiodomethyl- (7CI,8CI,9CI) -4.31 1137593 XUC [1,1'-biphenyl]-4-ol, 3,5-dichloro- 2.08 1185097 XUC ethanesulfenyl chloride, 1,1,2,2-tetrachloro- 1.33 1483723 XUC iodonium, diphenyl-, chloride 1.92 1483734 XUC iodonium, diphenyl-, bromide 1.92 1484500 XUC ethanone, 2-bromo-1,2-diphenyl- 0.38 1488422 XUC iodonium, (2-carboxyphenyl)phenyl-, hydroxide, inner salt 0.35 1545831 XUC diazene, (4-fluorophenyl)phenyl- 0.92 1594690 XUC 9,10-Anthracenedione, 1,7-dichloro- 0.94 1596367 XUC benzenesulfonamide, N-(4,5-dichloro-2-nitrophenyl)- -1.22 1602002 XUC P,P'-dichloroazobenzene 1.48 1613667 XUC germane, dichlorodiphenyl- 0.55 1623934 XUC [1,1'-biphenyl]-4-sulfonyl chloride -0.23 1714507 XUC 4,4'-dichlorobenzophenone, oxime 0.91 1729993 XUC 1-naphthalenecarboxylic acid, 8-bromo- 1.48 1806231 XUC methanone, (2-chlorophenyl)(4-fluorophenyl)- 1.48 1836733 XUC benzene, 1-chloro-3-methyl-2-(4-nitrophenoxy)- -0.66 1836744 XUC benzene, 1-chloro-4-(4-nitrophenoxy)- -0.15 1928398 XUC Acetic acid, (2,4,5-trichlorophenoxy)-, ethyl ester (6CI,7CI,8CI,9CI) 1.88 2005085 XUC benzoic acid, 4-chlorophenyl ester -0.15 2069412 XUC methanone, (4-bromophenyl)(4-fluorophenyl)- 1.48 2091614 XUC benzene, 1-chloro-2-(4-nitrophenoxy)- -0.15 2117693 XUC Plumbane, dichlorodiphenyl- 0.55 2181422 XUC trimethylsulfonium iodide -3.07 2303233 XUC benzene, 1-chloro-3-(4-nitrophenoxy)- -0.15 2392485 XUC benzene, 4-chloro-1-(2,4-dichlorophenoxy)-2-nitro- 0.80 2547617 XUC methanesulfonyl chloride, trichloro- 0.54 2629472 XUC stibine, chlorodiphenyl- 0.55 2712836 XUC butyranilide, 2,2,3,3,4,4,4-heptafluoro-2'-hydroxy-4'-nitro- 1.63 2729115 XUC phosphine oxide, triS(pentafluorophenyl)- 1.99 2750814 XUC naphthalene, 1,4-dichloro-5-nitro- 2.08 2790161 XUC cyclopropanecarboxamide, N-(3,4-dichlorophenyl)-1-methyl- 2.08 2892617 XUC 3-butenoic acid, 2,3,4,4-tetrachloro-, butyl ester 2.16 2899027 XUC urea, N,N'-dichloro-N,N'-bis(2,4,6-trichlorophenyl)- 1.69

2915017 XUC 4(3H)-Quinazolinone, 2-methyl-3-(a,a,a-trifluoro-o-tolyl)-, monohydrochloride (8CI) -0.34

2965777 XUC 4(3H)-Quinazolinone, 2-methyl-3-(a,a,a-trifluoro-m-tolyl)-, monohydrochloride (8CI) -0.34

3033736 XUC Peroxide, bis(2-chlorobenzoyl) -0.71 3096477 XUC 2-chloro-9H-fluoren-9-one 1.59 3236973 XUC stannane, triiodomethyl- (8CI,9CI) 1.46 3457463 XUC ethanedione, bis(4-chlorophenyl)- 0.35 3622400 XUC benzothiazole, 2-bromo-4-chloro- 2.83 3808897 XUC ethanone, 1-(2'-chloro[1,1'-biphenyl]-4-YL)- 0.38 3905962 XUC 4,4'-dinitro-octafluorobiphenyl 1.29 3988032 XUC methanone, bis(4-bromophenyl)- 1.48 4053081 XUC 1H,3H-naphtho[1,8-cd]pyran-1,3-dione, 6-chloro- 1.00 4185631 XUC naphthalene, 2-chloro-1-nitro- 1.48 4189100 XUC 2-ethylbuttersaeure-2,2,2-trichlorethylester 1.63 4340776 XUC diazene, (4-chlorophenyl)phenyl- 0.92

79

4418842 XUC diazene, (4-bromophenyl)phenyl- 0.92 4537002 XUC 1-naphthalenecarboxylic acid, 8-chloro- 1.48 4559960 XUC 1-butanone, 1-(4-bromophenyl)-4-chloro- 2.16 4693009 XUC 2H-3,1-benzoxazine-2,4(1H)-dione, 6,8-dichloro- 2.70 4841207 XUC propanoic acid, 2-(2,4,5-trichlorophenoxy)-, methyl ester 1.88 4889707 XUC benzoic acid, 2-[(4-chlorophenyl)methyl]- -0.15 5074715 XUC phosphine, bis(2,3,4,5,6-pentafluorophenyl)phenyl- 1.67 5162038 XUC methanone, (2-chlorophenyl)phenyl- 0.92 5317668 XUC benzeneacetic acid, a,a-dichloro-, ethyl ester 1.92 5323671 XUC benzenemethanol, 2,4-dichloro-, carbonate (2:1) (9CI) 0.08 5659416 XUC 2,3,4,5,5-pentachloro-2,4-pentadienoic acid 2.18 5731011 XUC ethanone, 1-(4'-bromo[1,1'-biphenyl]-4-yl)- 0.38 5774061 XUC 1-naphthalenecarboxylic acid, 3-chloro- 1.48 5858195 XUC benzo[b]thiophen-3(2H)-one, 5,7-dichloro-4-methyl- 3.01 6141953 XUC diazene, (4-chlorophenyl)phenyl-, (E)- 0.92 6240557 XUC naphthalene, 1,2-dichloro-3-nitro- 2.08 6284793 XUC methanone, (3,4-dichlorophenyl)phenyl- 1.48 6297116 XUC 9H-fluoren-9-one, 2,7-dichloro- 2.15 6530978 XUC diazene, (4-chlorophenyl)phenyl-, (Z)- 0.92 6639276 XUC diazene, (4-iodophenyl)phenyl- 0.92 6657052 XUC 2-chloro-4-phenylazophenol 0.38 6723406 XUC 1H-indene-1,3(2H)-dione, 2-[4-(trifluoromethyl)phenyl]- 0.22 6740869 XUC methanone, (1-bromocyclopentyl)(2-chlorophenyl)- 2.08 6915577 XUC 1,3,2-benzodioxabismole, 4,5,6,7-tetrabromo-2-hydroxy- 1.75 7012160 XUC 9H-fluorene, 2,7-dichloro- 2.80 7334330 XUC diazene, bis(2-chlorophenyl)- 1.48 7335833 XUC benzoic acid, 2-benzoyl-3,4,5,6-tetrachloro-, methyl ester -0.30 7791277 XUC pyrosulfuryl chloride -1.83 8072206 XUC [(4-chlorophenyl)thio](2,4,5-trichlorophenyl)diazene (9CI) 0.91 10182840 XUC iodonium, diphenyl- 2.03 13036754 XUC disulfuryl fluoride -1.83 13102346 XUC 3H-Pyrazol-3-one, 2-(2,5-dichlorophenyl)-2,4-dihydro-3-methyl- 2.08 13124179 XUC 3H-Pyrazol-3-one, 2-(3,4-dichlorophenyl)-2,4-dihydro-5-methyl- 2.08 13191361 XUC thiophene, 2,5-dibromo-3-methyl- 1.72 13304240 XUC benzoic acid, 2-[(4-chlorophenyl)azo]- (9CI) -0.66 13455000 XUC hypodiphosphorous tetraiodide -3.90 13779925 XUC niobium pentaiodide -4.72 14529125 XUC 2H-3,1-benzoxazine-2,4(1H)-dione, 6-chloro-1-methyl- 1.48 15426149 XUC diazene, 1,2-bis(3-chlorophenyl)- 1.48

15608374 XUC 1,3,5,2,4,6-triazatriphosphorine, 2,2,4,4,6-pentabromo-6-chloro-2,2,4,4,6,6-hexahydro- (8CI) -0.68

15945285 XUC benzoic acid, 3,5-dichloro-4-methoxy-, ethyl ester 1.36

15965001 XUC 1,3,5,2,4,6-triazatriphosphorine, 2,2,4,6-tetrabromo-4,6-dichloro-2,2,4,4,6,6-hexahydro- (8CI) -0.68

16032523 XUC 1,3,5,2,4,6-triazatriphosphorine, 2,4,6-tribromo-2,4,6-trichloro-2,2,4,4,6,6-hexahydro- (8CI) -0.68

16187943 XUC isothiazole, 4-bromo-3-phenyl- 2.33 16442107 XUC benzoic acid, 2,6-dichloro-, anhydride (8CI,9CI) 0.63 16574520 XUC methanone, (4-fluorophenyl)[4-(trifluoromethyl)phenyl]- 1.80 16650525 XUC 1-naphthalenecarboxylic acid, 5-chloro- 1.48 16965085 XUC carbonic acid, 1,1-dimethylethyl 2,4,5-trichlorophenyl ester 1.31 17175132 XUC carbonic acid, P-bromophenyl phenyl ester -0.15 17318706 XUC benzenesulfonyl chloride, 4-benzoyl- -1.20 17640083 XUC propionic acid, 2,2-dichloro-, pentyl ester 1.01 17969221 XUC thiazole, 4-(chloromethyl)-2-(4-chlorophenyl)- 2.70 18018092 XUC 9,10-anthracenedione, 1,3-dichloro-2-methyl- 0.37 18018332 XUC benzimidazole, 1-ethyl-2-methyl-6-nitro-5-(trifluoromethyl)- 0.80 18189190 XUC 1,2-ethanedione, 1-[4-(bromomethyl)phenyl]-2-phenyl- -0.66 18264841 XUC phenanthrene, 2-bromo-9,10-dihydro-3-nitro- -0.11

18812812 XUC 2-propanamine, N,N'-[4-(chloromethyl)-4-methyl-2,3-oxetanediylidene]bis[2-methyl- (9CI) -0.74

19055700 XUC 1H-indene-1,3(2H)-dione, 2-[3-(trifluoromethyl)phenyl]- 0.22 19762799 XUC 4-isothiazolecarbonitrile, 3-(P-bromophenyl)-5-methyl- 0.92 19811053 XUC methanone, (2,4-dichlorophenyl)phenyl- 1.48

19982877 XUC hydrazinecarboximidamide, 2-[(4-fluorophenyl)[4-(trifluoromethyl)phenyl]methylene]-, monohydrochloride (9CI) -0.47

20357793 XUC 1H-indene, 1,2-dibromo-2,3-dihydro- 2.83 20717797 XUC 2-naphthalenecarboxylic acid, 1-bromo- 1.48 21221932 XUC methanone, [3,5-bis(trifluoromethyl)phenyl]phenyl- 1.91 21854955 XUC ethanedione, bis(2-chlorophenyl)- 0.35

80

22071245 XUC methanone, [3-(bromomethyl)phenyl]phenyl- 0.38 22532689 XUC benzene, 1,2,4-trichloro-5-(4-nitrophenoxy)- 0.80 22544021 XUC benzene, 4-chloro-1-(2-chlorophenoxy)-2-nitro- 0.35 22544076 XUC benzene, 2-chloro-1-(4-chlorophenoxy)-4-nitro- 0.35 22711235 XUC 1,2-ethanedione, 1-(4-chlorophenyl)-2-phenyl- -0.15 23038360 XUC benzene, 1-bromo-4-(phenylsulfonyl)- 0.38 25004959 XUC methanesulfinyl chloride, 1,1,1-trichloro- 0.64 25006320 XUC leptophos-O-analog 0.24 25100918 XUC thiazole, 2-(4-chlorophenyl)-4-methyl- 2.07 25187068 XUC methanone, (2-chlorophenyl)(2,5-dichlorophenyl)- 1.98 25512429 XUC dichloro-1,1'-biphenyl 2.08 25981833 XUC 3H-indole, 5-chloro-2,3,3-trimethyl- 2.07 26203816 XUC acetic acid, (2,4,6-tribromophenoxy)-, ethyl ester (8CI) 1.88 27087463 XUC carbonic acid, P-chlorophenyl phenyl ester -0.15 27701662 XUC 1,1'-biphenyl, 4-bromo-3-nitro- 0.38 28202304 XUC benzene, 2-methyl-4-nitro-1-(1,1,2,2-tetrafluoroethoxy)- 2.34

28313522 XUC acetic acid, [[2-chloro-5-(trifluoromethyl)phenyl]hydrazono]cyano-, ethyl ester (9CI) 0.08

28313577 XUC acetic acid, [[4-chloro-3-(trifluoromethyl)phenyl]hydrazono]cyano-, ethyl ester (9CI) 0.08

28313691 XUC acetic acid, cyano[[2,6-dichloro-4-(trifluoromethyl)phenyl]hydrazono]-, methyl ester (9CI) 0.99

28313771 XUC acetic acid, [[4-chloro-2-(trifluoromethyl)phenyl]hydrazono]cyano-, methyl ester (9CI) 0.62

28317568 XUC butanenitrile, 2-[[2-chloro-5-(trifluoromethyl)phenyl]hydrazono]-3-oxo- (9CI) 1.18

28317944 XUC pentanenitrile, 2-[[4-chloro-2-(trifluoromethyl)phenyl]hydrazono]-4,4-dimethyl-3-oxo- (9CI) -0.44

28322783 XUC acetic acid, cyano[(2,4,5-trichlorophenyl)hydrazono]-, ethyl ester (9CI) 0.22

28343251 XUC pentanenitrile, 2-[[2-chloro-5-(trifluoromethyl)phenyl]hydrazono]-4,4-dimethyl-3-oxo- (9CI) -0.44

28343273 XUC pentanenitrile, 2-[[3,5-bis(trifluoromethyl)phenyl]hydrazono]-4,4-dimethyl-3-oxo- (9CI) -0.32

28801914 XUC 9H-fluorene, bromo- 2.20 29170089 XUC 1,1'-biphenyl, 4-iodo-4'-nitro- 0.38 29328992 XUC 1H-benzotriazoLE, 5-chloro-1-(4-chlorophenyl)- 1.53 29421736 XUC thiophene, 3,5-dibromo-2-methyl- 1.72 29558881 XUC benzoic acid, 4-chloro-, 4-fluorophenyl ester 0.91 30070634 XUC benzoic acid, 3-chloro-, 1,1'-anhydride -0.19 31039722 XUC 1H-isoindole-1,3(2H)-dione, 4,5,6,7-tetrachloro-2-(4-chlorophenyl)- 1.54 31431160 XUC methanone, (4-chloro-3-nitrophenyl)(4-fluorophenyl)- -0.19 31643148 XUC benzenemethanol,a-(trichloromethyl)-, propanoate 1.88 32752548 XUC methanone, [4-(bromomethyl)phenyl]phenyl- 0.38 33146575 XUC methanone, (4-chlorophenyl)(2,4-dichlorophenyl)- 1.98 33184520 XUC methanone, (5-fluoro-2-methylphenyl)phenyl- 0.38 33184553 XUC (5-chloro-2-methylphenyl)phenylmethanone 0.38 33713132 XUC benzamide, N-[2-chloro-5-(trifluoromethyl)phenyl]-2-hydroxy-3-nitro- -1.00 34052374 XUC methanone, (2-chloro-5-nitrophenyl)phenyl- -0.66 34113467 XUC 2-(2-chlorophenyl)-2-(4-chlorophenyl)acetic acid 0.35 34183067 XUC methanone, (3,5-dichloro-4-hydroxyphenyl)phenyl- 0.91 34223820 XUC 9H-fluoren-9-one, 3,6-dichloro- 2.15 34593754 XUC 4-(tert-butyl)-2,6-dichlorophenol 2.16 34645846 XUC Fenclofenac -0.19 35578473 XUC ethanedione, bis(4-bromophenyl)- 0.35

36865540 XUC acetic acid, [[3,5-bis(trifluoromethyl)phenyl]hydrazono]cyano-, methyl ester, (Z)- (9CI) 0.76

36865722 XUC acetic acid, [[2-chloro-4-(trifluoromethyl)phenyl]hydrazono]cyano-, methyl ester (9CI) 0.62

36905007 XUC acetic acid, [[3,5-bis(trifluoromethyl)phenyl]hydrazono]cyano-, methyl ester, (E)- (9CI) 0.76

36905041 XUC acetic acid, cyano[(2,4,5-trichlorophenyl)hydrazono]-, methyl ester, (Z)- (9CI) 0.75

37572444 XUC isothiazole, 4-bromo-3,5-bis(methylthio)- 1.00 38396213 XUC naphthalene, 2-chloro-3-nitro- 1.48 38396291 XUC naphthalene, 6-chloro-1-nitro- 1.48 38461299 XUC benzene, 2,4-dichloro-1-(2-nitrophenoxy)- 0.35 39145476 XUC benzene, 1-(4-chlorophenoxy)-2-nitro- -0.15 39145487 XUC benzene, 1,4-dichloro-2-(4-nitrophenoxy)- 0.35 40877096 XUC 1-butanone, 4-chloro-1-(4-chlorophenyl)- 2.16 41182858 XUC benzenecarboximidoyl bromide, N-methyl- 1.00 41893703 XUC benzene, [bromo(1-bromocyclopropyl)methyl]- 2.89

81

42728621 XUC methanone, [4-(chloromethyl)phenyl]phenyl- (9CI) 0.38 42771799 XUC ethanone, 1-(2-fluoro[1,1'-biphenyl]-4-yl)- 0.38 42959999 XUC methanone, (2-bromo-3-nitrophenyl)phenyl- -0.66 42989162 XUC benzoic acid, 2,4,5-trichloro-6-cyano-3-phenoxy-, methyl ester -0.83 49619434 XUC benzoic acid, 2-chloro-, 1,1'-anhydride -0.19 50274641 XUC methanone, (2-chloro-4-nitrophenyl)(4-chlorophenyl)- -0.19 50274969 XUC benzene, 2-chloro-1-[(2,4-dichlorophenyl)methyl]-4-nitro- 0.80 50594440 XUC phenol, 5- 2-chloro-4-(trifluoromethyl)phenoxy -2-nitro-, acetate (ester) -1.52 50594779 XUC phenol, 3- 2-chloro-4-(trifluoromethyl)phenoxy -, acetate 0.08 50618923 XUC benzene, 1,1'-(1,2-ethanediyl)bis[4-fluoro-2-nitro- -1.73 50807175 XUC ethanone, 2,2-dichloro-1-[4-(1,1-dimethylethyl)phenyl]- 1.36 50817726 XUC 1H,3H-naphtho[1,8-cd]pyran-1,3-dione, 4-chloro- 1.00 51282690 XUC methyl 5-(2-chloro-4-fluorophenoxy)-2-nitrobenz* -1.73 51326375 XUC ethanone, 2,2,2-trichloro-1- 4-(1,1-dimethylethyl)phenyl - 1.88 51735876 XUC phophorimidic trichloride, [(trifluoromethyl)sulfinyl]- -0.03 53218114 XUC benzeneacetic acid, 4-nitro-, 4-chlorophenyl ester -1.66 53317058 XUC 2-thiophenecarboxylic acid, 3,4,5-tribromo- 2.38 53526297 XUC naphthalene, 1-chloro-3-nitro- 1.48 53591790 XUC ethanone, 1-(2',4'-difluoro[1,1'-biphenyl]-4-yl)- 0.91 53911685 XUC 2H-Pyran-2,6(3H)-dione, 4-(4-chlorophenyl)dihydro- 0.92 54683919 XUC benzoic acid, 2-chlorophenyl ester 0.38 54934495 XUC propanoic acid, 2,2-dichloro-, 1,2-ethanediyl ester 2.17

55521147 XUC Carbonic acid, 1,1,2,2,3,3,4,4,5,5,6,6,7,7,8,8-hexadecafluoro-1,8-octanediyl dimethyl ester (9CI) 1.81

55683244 XUC benzeneacetic acid, 3,4-bis(2,2,3,3,3-pentafluoro-1-oxopropoxy)-, methyl ester (9CI) -0.40

55702426 XUC [1,1'-biphenyl]-2-carboxylic acid, 4,4'-dichloro-, methyl ester 0.35 55702448 XUC 1H-benzimidazole, 2-ethyl-7-nitro-1-propyl-5-(trifluoromethyl)- -0.30 55780411 XUC cycloPROP[A]INdene, 6-bromo-1,1A,6,6A-tetrahydro- 3.10 56041759 XUC benzenemethanol, 4-bromo-a-cyclopropyl-a-methyl- 2.07 56107029 XUC methanone, (4-chloro-3-nitrophenyl)phenyl- -0.66 56454150 XUC benzenebutanoic acid, G-bromo-, ethyl ester 0.25 56882521 XUC ethyl-2,4-dichlorobenzoate 2.16 56960953 XUC benzoic acid, 4-chloro-3-nitro-, (4-chlorophenyl)methyl ester -1.22 56960964 XUC benzoic acid, 2-chloro-5-nitro-, (4-chlorophenyl)methyl ester -1.22 56961003 XUC benzenesulfonic acid, 4-chloro-, (4-chlorophenyl)methyl ester -0.19 56961365 XUC naphthalene, 1-chloro-6-nitro- 1.48 56961376 XUC naphthalene, 2-chloro-6-nitro- 1.48 56961387 XUC naphthalene, 2-chloro-7-nitro- 1.48 56961398 XUC naphthalene, 3-chloro-1-nitro- 1.48 56961898 XUC 2-naphthalenecarboxylic acid, 5-chloro- 1.48 56961989 XUC 9,10-Phenanthrenedione, 2-chloro- 0.43 56961990 XUC 9,10-Phenanthrenedione, 3-chloro- 0.43 56966699 XUC benzene, 2-chloro-4-nitro-1-phenoxy- -0.15 56978508 XUC 1,1'-biphenyl, 4,4'-dichloro-2,2'-dinitro- -0.71 57025760 XUC MC-15608 [2,4'-dicl-4-CF3-3'-CO2Me-diPh ether] 0.42 57147054 XUC 1,4-benzenedimethanol, 2,3,5,6-tetrabromo-, diacetate (9CI) 0.62 57340213 XUC diazene, 1-[4-(bromomethyl)phenyl]-2-phenyl- 0.38 57396879 XUC [1,1'-biphenyl]-4-OL, 4'-chloro-, acetate -0.15 57479604 XUC ethanone, 2-(2-chlorophenyl)-1-phenyl- 0.38 58359538 XUC benzenesulfonyl chloride, 4-(phenylazo)- -1.20 58748166 XUC bis(2,4-dichlorobenzoyl) peroxide (9CI) -0.71 59089677 XUC [1,1'-biphenyl]-4-ol, 2',4'-difluoro-, acetate (9CI) 0.35 59189514 XUC methanone, (2,6-difluorophenyl)phenyl- 1.48 59396508 XUC methanone, (4-fluoro-3-methylphenyl)(4-fluorophenyl)- 0.91 59396519 XUC methanone, (2-fluoro-5-methylphenyl)phenyl- 0.38 59584224 XUC benzoic acid, 4-[[(4-chlorophenyl)imino]methyl]-, methyl ester -1.17 59700571 XUC pentachloro(2,2,3,3-tetrafluoropropoxy)cyclotri* 2.37 60066537 XUC hexanoic acid, 4,6,6,6-tetrachloro-3,3-dimethyl-, ethyl ester 2.17 61405489 XUC RH-1460 diPh thioether -1.52 62336247 XUC phosphine, (2-bromophenyl)diphenyl- -1.17 62476577 XUC benzene, 2-chloro-1-(3-nitrophenoxy)-4-(trifluoromethyl)- 0.63 62586963 XUC benzenemethanol, a,a-dicyclopropyl-4-fluoro- 1.59 62625245 XUC ethanone, 2-bromo-1- 4-(phenylazo)phenyl - -0.66 63028273 XUC 2,5-dichloro-4'-isopropylbiphenyl 0.91 63549337 XUC 1-propanone, 3-chloro-1-(5-chloro-2-methylphenyl)- 2.16 63549348 XUC 1-propanone, 3-chloro-1-(2-chloro-5-methylphenyl)- 2.16 63734623 XUC benzoic acid, 3- 2-chloro-4-(trifluoromethyl)phenoxy - 0.63 63867072 XUC Butanoic acid, 2-ethyl-, trichloro-1-methylethyl ester (9CI) 1.88 63867083 XUC Butanoic acid, 2-bromo-2-ethyl-, 2,2,2-tribromoethyl ester 2.16

82

63867107 XUC Butanoic acid, 2-chloro-2-ethyl-, 2,2,2-trichloroethyl ester 2.16 63867174 XUC 2-ethylbuttersaeure-2,2,2-tribromethylester 1.63 63979442 XUC pentanoic acid, 2,2-dibromo-, ethyl ester 0.75 64047478 XUC butanoic acid, 2-bromo-3-methyl-, 2,2,2-tribromoethyl ester 1.96 64047489 XUC butanoic acid, 2-bromo-3-methyl-, 2,2,2-trichloroethyl ester 1.96 64048904 XUC arsonous dichloride, [3-(trifluoromethyl)phenyl]- (9CI) 1.73 64057767 XUC benzenemethanol, 3,4-dichloro-, carbonate (2:1) (9CI) 0.08 64059376 XUC butanoic acid, 2-bromo-3-methyl-, trichloro-1-methylethyl ester (9CI) 2.16 64123644 XUC benzene, 2-(chloromethyl)-1-(1-methylethyl)-4-nitro- 0.76 64399264 XUC cyclopentanecarbonitrile, 1-(4-chlorophenyl)- 1.48 64667330 XUC hexanoic acid, 4,6,6,6-tetrachloro-3,3-dimethyl-, methyl ester 2.39 65237174 XUC 1,3-Isobenzofurandione, 5,6-dibromo- 3.01 65601414 XUC 2,2,4-trichloro-2,2,4,4,6,6-hexahydro-4,6,6-tri* 1.31 68015952 XUC benzene, 2-(chloromethyl)-1-(1-methylpropyl)-4-nitro- 0.25 69943471 XUC methanone, (2-chloro-4-fluorophenyl)phenyl- 1.48 70788566 XUC methanone, 4-(dichloromethyl)phenyl (2,5-dichlorophenyl)- 1.80 71463495 XUC benzoic acid, 2,6-dichloro-, 2,6-dichlorophenyl ester 1.80 71463553 XUC cyclopropanecarbonitrile, 1-(2,4-dichlorophenyl)- 3.01 71501514 XUC [1,1'-biphenyl]-2-carboxylic acid, 3'-chloro-4'-methyl- -0.15 71849984 XUC benzoic acid, 2,6-dichloro-, phenyl ester 0.91 73806710 XUC 2,6-dimethyl-2,3,5,6-tetrabromo-4-heptanone 2.16 74129105 XUC ethanone, 1-[2-bromo-4-(phenylazo)phenyl]- (9CI) -0.66 74204049 XUC 9,10-anthracenedione, 1-bromo-4-methyl- -0.11 74298912 XUC 1,1'-biphenyl, 3,4',5-trichloro-4-methoxy- 1.98 74298923 XUC 1,1'-biphenyl, 2,3',4-trichloro-4'-methoxy- 1.98 74610507 XUC benzoic acid, 2-(acetyloxy)-3,5-dibromo-, 4-bromophenyl ester -0.83 74630925 XUC 1-propanol, 1,1-dichloro-, phosphate (3:1) (9CI) 1.83

83721464 XUC methanesulfonamide, 1-chloro-N-[4,5-dichloro-2-(2,4-dichlorophenoxy)phenyl]-, Na salt 0.33

83721475 XUC methanesulfonamide, 1-chloro-N-[2,3,4-trichloro-6-(2,4-dichlorophenoxy)phenyl]-, Na salt 0.64

84392176 XUC 4'-(trifluoromethyl)-2-biphenylcarboxlic acid 0.80 84532729 XUC 1-naphthalenecarboxylic acid, 6-methoxy-5-(trifluoromethyl)- 0.80 85262948 XUC benzenemethanol, 2,4-dichloro-, propanoate 1.92 85721080 XUC methanone, [2-chloro-5-(trifluoromethyl)phenyl](4-fluorophenyl)- 2.18 85721091 XUC methanone, [2-bromo-4-(trifluoromethyl)phenyl](4-fluorophenyl)- 2.18 85897296 XUC chlorofluorenone 1.59 86914729 XUC methanone, (5-chloro-2-hydroxy-3-methylphenyl)(4-chlorophenyl)- 0.35 87750503 XUC benzene, 1-(phenylsulfonyl)-4-(trifluoromethoxy)- 0.24 87750592 XUC methanone, (2-fluoro-5-methylphenyl)[2-(trifluoromethyl)phenyl]- 1.21 87750616 XUC methanone, (2-chloro-4-fluorophenyl)(4-fluorophenyl)- 1.98 88185222 XUC 3-[2-chloro-4-(trifluoromethyl)phenoxy]B -3.04 88927428 XUC ethyl (2,3,4,5-tetrachlorophenoxy)acetate 2.34 90077740 XUC dichloro-9H-fluoren-9-one 2.15 90077751 XUC trichloro-9H-fluoren-9-one 2.64 94070856 XUC naphthalene, 1,3-dibromo-1,2,3,4-tetrahydro- 2.89

94248267 XUC methanesulfonamide, 1-chloro-N-(2-phenoxyphenyl)-, pentachloro deriv., sodium salt 0.64

95998699 XUC dichloro(trifluoromethyl)benzophenoNE 2.18

97659353 XUC 1,3,4-oxadiazol-2(3H)-one, 3-(2-chloro-4-isocyanatophenyl)-5-(1,1-dimethylethyl)- -1.66

104668679 XUC propanamide, 3,3,3-trifluoro-2-methyl-N-[4-nitro-3-(trifluoromethyl)phenyl]-, (±)- 1.32

108548727 XUC (dichlorophenyl)methyl chlorobenzoate 0.80 121325448 XUC benzoic acid, 5-[2-chloro-5-(trifluoromethyl)phenoxy]-2-nitro-, methyl ester -1.52

83

Table 5 - Explanation of the Acronyms for Chemical Class Used in Tables S2 and S3

CUP current use pesticide (or safener)

OP Obsolete pesticide

PBDE polybrominated diphenyl ether

PCB polychlorinated biphenyl

PCBD polychlorinated benzene derivative (chlorination is on benzene ring, functional

groups are simple)

PCDD polychlorinated dibenzo-p-dioxin

PCDE polychlorinated diphenyl ether

PCDF polychlorinated dibenzofuran

PCN polychlorinated naphthalene

PCS polychlorinated styrene

PFA perfluorinated organic acid

POP persistent organic pollutants designated by the Stockholm Convention

PXBD polyhalogenated benzene derivative (same as PCBD but including at least one

fluorine, bromine or iodine atom)

PXBP poly halogenated biphenyl (same as PCB but including at least one fluorine,

bromine or iodine atom)

PXDD polyhaloginated dibenzo-p-dioxin

PXDF polyhaloginated dibenzofuran

PXN polyhaloginated naphthalene

silane contains one or more silicon atoms

Silox contains one or more silicon atoms bonded to an oxygen atom

UC non-halogenated unclassified chemical, does not fall into any of the other

classifications

XAB halogenated alkyl benzene (benzene (possibly directly halogenated) substituted

with a halogenated alkyl)

XABP halogenated alkyl biphenyl (same as XAB but biphenyl instead of benzene)

XAH halogenated alkyl heterocycle (same as XAB but a heterocycle instead of

benzene)

Xalka halogenated alkane

XalkaC halogenated cycloalkane (excluding halogenated cyclohexanes)

Xalke halogenated alkene

XalkeC halogenated cycloalkene

XAN halogenated alkyl naphthalene (same as XAB but a naphthalene instead of

benzene)

XBDP alkyl-bridged halogenated diphenyl

XH halogenated heterocycle

XNOR halogenated norbornene or norbornane

XPAH halogenated polycyclic aromatic hydrocarbon

XUC halogenated unclassified chemical, does not fall into any of the other

classifications

84

Table 6 - Chemical classifications of the chemicals which meet the elevated AC-BAP

and atmospheric oxidation half-life criteria.

Elevated AC-BAP Yes Yes

Long atmospheric oxidation half-life Yes Yes

Fulfills POPs score criterion Yes No

Chemical Class Number Identified Number Identified

PCBs 123 20

Other halogenated biphenyls 4 16

Halo-alkyl biphenyls 3 2

PCNs 19 11

Other halogenated naphthalenes 0 2

Halo-alkyl naphthalenes 1 0

PCDDs 9 17

Other halogenated dibenzo-p-dioxins 1 5

PCDFs 90 20

Other halogenated dibenzofurans 3 3

PBDEs 2 3

Polychlorinated diphenylethers (PCDEs) 60 12

PCSs 3 0

chlorinated benzene derivatives 87 101

Other halogenated benzene derivatives 57 97

Halo-alkyl benzenes 48 40

Perfluorinated organic acids 13 0

Other Stockholm POPs and derivatives 33 0

Pesticides (obsolete) 6 0

Pesticides (current use) 11 43

Halogenated PAHs 2 4

Halogenated alkanes 16 32

Halogenated cycloalkanes 23 0

Halogenated alkenes 7 0

Halogenated cycloalkenes 9 0

Halogenated diphenylmethanes 5 13

Halogenated heterocycles 13 24

Halo-alkyl heterocycles 72 15

Halogenated norbornenes/norbornanes 1 3

silane 1 27

Siloxane 0 27

Unclassified: Non-Halogenated 0 319

Unclassified: Halogenated 100 347

Total 822 1203

85

Chapter 3

Development and Exploration of an Organic Contaminant Fate

Model Using Poly-Parameter Linear Free Energy Relationships


Environmental Science & Technology 2009 43 (17), 6676-6683

Reproduced with permission from Environmental Science and Technology

Copyright 2009 American Chemical Society

86

1 Introduction Several years ago Breivik and Wania proposed that the domain of applicability of

multimedia fate models could be expanded by using poly-parameter linear free energy

relationships (PP-LFERs) to quantify phase partitioning111

. The argument for implementing

models of this type was that single-parameter linear free energy relationships (SP-LFERs),

primarily those based on regressions with KOW, make poor predictions of phase partitioning

when applied to chemicals outside of the datasets used in their parameterization112

.

Subsequent work has shown that this approach is viable for chemicals, such as

pharmaceuticals, which may be considered to be outside the range of applicability of SP-

LFER based models113

. More recently Götz et al. implemented PP-LFER based partitioning

into atmospheric transport models and demonstrated that the model results can be

significantly different from SP-LFER based models, especially for polar chemicals114

.

The PP-LFER equations applied in these studies were the linear solvation energy

relationships developed by Abraham et al.115

, or extensions thereof116,117

. These equations

include terms that describe the interactions of the solute (solute descriptors) and of the

phases involved (phase descriptors). Specific intermolecular interactions, such as dipole

interactions and hydrogen bonding, and non-specific intermolecular interactions, such as

cavitation energy and dispersive van der Waals interactions, are accounted for separately.

Three major factors limited the implementation of a PP-LFER-based multimedia fate model

at the time of Breivik and Wania‘s publication; PP-LFER equations were missing for several

environmentally relevant phases, there was no PP-LFER based method for calculating the

temperature dependence of phase partitioning, and the availability of solute descriptors for

environmentally relevant chemicals was limited. There has since been progress on all three

of these issues, and sufficient data is now available for the full implementation of a PP-

LFER-based multimedia fate model.

Much work has been done recently to address the lack of PP-LFER equations for some

environmentally relevant phases. Goss118

has proposed a modification of Abraham‘s PP-

87

LFERs which uses a single form to describe all partitioning processes, instead of separate

forms for partitioning between two condensed phases and between a gas phase and a

condensed phase, as in the traditional Abraham equations115

. This modification has a

statistical disadvantage in that two of the parameters are highly correlated, but as Goss notes

the equations are more convenient for environmental modelling because it allows for the use

of thermodynamic cycles to indirectly derive equations for partitioning systems that have not

been experimentally measured. Goss et al. have subsequently characterized the partitioning

of a wide range of chemicals to environmentally relevant phases, including natural organic

matter57,119

and atmospheric aerosol120,121

, and provided modified PP-LFERs to describe

partitioning to these phases.

A solution to the problem of describing the effect of temperature on phase partitioning has

also been presented. Mintz et al. derived Abraham PP-LFER equations for the prediction of

enthalpies of solvation for water/air and octanol/air partitioning122

, building upon previous

work deriving equations for the enthalpy of phase change for systems of two condensed

phases123,124

. Using the enthalpies compiled by Mintz et al.122

, and provided by Niederer et

al.119

and Arp et al.121

the necessary data is now available to derive modified PP-LFERs for

the enthalpies of phase change required for multimedia fate modelling.

Constructing a PP-LFER-based multimedia fate model is only a worthwhile endeavour if

there are solute descriptors available for environmentally relevant chemicals. Such

descriptors have been published for a large number of chemicals, but due to experimental

limitations many of these chemicals are small and have a single functional group45,125

. This

limitation in the datasets used to parameterize the PP-LFERs however does not appear to

limit their applicability to environmentally relevant phases126

. Some recent additions to the

available solute descriptors are more environmentally relevant; these include a series of

commonly used pesticides and pharmaceuticals127

, and polychlorinated biphenyls (PCBs)89

.

88

We outline here the implementation of PP-LFER equations into CoZMo-POP2, a non-

equilibrium, non-steady state, fugacity based multimedia fate model128

In working towards

the goal of expanding the range of multimedia fate models to more polar chemicals, a non-

steady state model is preferred. Polar chemicals will intuitively be more sensitive to seasonal

variations in the water balance, and possibly other environmental parameters, so a non-

steady state model will be required to properly describe their environmental fate.

Additionally, if significant differences are observed with a a non-steady state simulation on a

relatively short time scale then the potential differences between SP-LFER and a PP-LFER-

based models may be environmentally relevant, because we expect larger differences in

predicted environmental fate to be caused by differences in the description of environmental

partitioning as chemicals near equilibrium.

2 Methods The details concerning the selection of PP-LFER equations from the literature, and the

derivation of additional equations required are provided here and in the Appendix. A far

more difficult task than the implementation of a PP-LFER based model though is devising a

thorough and systematic comparison of the results of the two models. Where data is

available individual chemicals can be run in both the original SP-LFER based and the new

PP-LFER based CoZMo-POP2, and the results can be compared directly. This was the

strategy of Götz et al. and the differences in the model outputs for the specific chemicals

modelled are clearly demonstrated117

. We prefer to evaluate the differences for the entire

range of chemicals that the model can be applied to, and to offer specific guidance on what

results can be expected for chemicals with diverse environmental fates. Previous studies of

this type have made use of chemical space maps which show the variation in model outputs

as a function of partitioning coefficients, usually KAW, KOA and KOW129

. This approach uses

hypothetical chemicals with various combinations of the partitioning coefficients to

systematically map the chemical space. Applying this method to a PP-LFER based model,

however, is difficult, because for each combination of partitioning coefficients there are

multiple valid combinations of solute descriptors. It is, in theory, also possible that chemicals

in close proximity to each other within the chemical space may have widely different model

89

results, but as is discussed below and in the Appendix this is generally found not to be the

case; which allows the chemical space map method of testing and displaying model

sensitivity to be modified to evaluate the PP-LFER model.

2.1 PP-LFER Equations

Equation (9) is the modified PP-LFER equation presented by Goss to describe the

partitioning coefficient K of solute i between two phases j and k118

.

log Kijk = sjkSi + ajkAi + bjkBi + ljkLi + vjkVi + cjk (9)

Lowercase letters denote phase descriptors and uppercase solute descriptors. Each term

represents how different types of interactions between the solute and two phases contribute

to the overall partitioning. Specific interactions of the solute are described by Si

(polarity/dipolarity), Ai (hydrogen bond acidity) and Bi (hydrogen bond basicity) with the

corresponding phase descriptors describing the relative affinity of the two phases for those

kinds of interactions. Non-specific interactions are described by Li (log of the hexadecane air

partition coefficient) and Vi (McGowan volume) and cjk is a system constant. An analogous

equation is used to calculate the enthalpy of phase change.

ΔHijk = sjkSi + ajkAi + bjkBi + ljkLi + vjkVi + cjk (10)

Strictly speaking this is a misapplication of the PP-LFER equation, because the solute

descriptors are intended to describe free energy changes, not enthalpy changes alone.

However previous studies have used this type of equation with good results122-124

: the reason

for this success is very likely that in many cases the free energy and the enthalpy of a

partitioning process are linearly related, as has been noted by previous studies112,124,130

, and

recently validated mathematically by Macleod et al. for the enthalpy of vaporization131

. For

the purpose of modelling chemical fate at environmental temperatures and dilute

concentrations this assumption of linearity appears to be a fair one.

90

Table 7 - Phase Descriptors for PP-LFER Equations Used in this Study.

sjk ajk bjk ljk vjk cjk reference

log KAW -2.07 -3.67 -4.87 -0.48 2.55 0.59 118

log KOA 0.66 3.49 1.42 0.91 -0.14 -0.25 (a)

log KOW -1.41 -0.18 -3.45 0.43 2.41 0.34 118

log KHA 1.01 3.18 1.86 0.75 -0.17 -0.24 (b) 57

log KHW -1.07 -0.49 -3.01 0.27 2.38 0.34 (b) 57

log KQA 1.38 3.21 0.42 0.63 0.98 -7.24 (c) 121

ΔHAW (kJ/mol) -0.86 33.63 43.79 1.52 16.63 8.59 (d)

ΔHOA (kJ/mol) 6.04 -53.66 -9.19 -9.66 1.57 -6.67 (d)

ΔHOW (kJ/mol) 5.18 -20.03 34.60 -8.14 18.19 1.92 (e)

ΔHHA (kJ/mol) -4.59 -31.87 -17.81 -8.15 -2.82 -5.90 (f)

ΔHHW (kJ/mol) -5.45 1.76 25.98 -6.63 13.81 2.69 (g)

ΔHPA (kJ/mol) -

14.03 -20.73 -0.03 -3.37 -20.22 -2.89 (h)

(a) Derived by thermodynamic cycle from log KAW and log KOW of reference 118. (b) Equations from

reference 57 for 25 °C, normalized to OC content. (c) Equation for Berlin Winter aerosol, as

recommended for generic terrestrial aerosol, from reference 121. (d) Derived by for this study from the

data in the supporting information of reference 122. (e) Derived by thermodynamic cycle from ΔHWA and

ΔHOA. (f) Derived for this study from the data in the supporting information of reference 119. (g) Derived

by thermodynamic cycle from ΔHWA and ΔHHA. (h) Arp et al. suggest using the enthalpy of vaporization

as a substitute for ΔHQA, the equation shown is for the enthalpy of vaporization and was derived for this

study from the data in the supporting information of reference 130.

Table 7 summarizes the phase descriptors for Equations (9) and (10) that are used in this

study. Phases are symbolized by the subscripted letters A (air), O (octanol), W (water), H

(Leonardite humic acid), Q (water-insoluble aerosol fraction), and P (whole aerosol). Phase

descriptors provided in Table 7 for the enthalpies of phase change could not be drawn

directly from the literature, but have instead been derived for this study using data from the

relevant papers. Details concerning the derivation of these phase descriptors are provided in

the Appendix (Section 5.1).

2.2 Implementation of PP-LFERs in CoZMo-POP2.

As a fugacity based model partitioning in CoZMo-POP2 is described by Z-values, which are

half of a partition coefficient and describe the capacity of a phase to hold the chemical

91

modelled. The exact equations used to calculate the Z-values in CoZMo-POP2 are provided

in ref.128

, and the alterations to these equations required to implement PP-LFERs are

provided in the Appendix (Section 5.2). Two important SP-LFER equations are replaced

with PP-LFER equations; these are the equation for partitioning between organic carbon and

water, and the equation for partitioning between atmospheric particles and air. All other Z

values are calculated as described in ref.128

, the only difference is that all partitioning

coefficients are calculated with equation (9) and the phase descriptors from Table 7.

Temperature correction of the partitioning coefficients also remains unaltered, but the

enthalpies are now calculated with equation (10) using the phase descriptors from Table 7,

and then converted to internal energies. One SP-LFER remains in the model; partitioning to

the forest canopy, described by a regression with KOA132

. We prefer this equation over the

available PP-LFER for partitioning to tomato cuticle133

; the reasons for this are discussed in

the Appendix (Section 5.3).

2.3 Solute Descriptors.

Experimentally measured solute descriptors have been compiled from a number of literature

sources45,125,127,134-137

. In several of these sources Li values are missing for some45

, or all of

the chemical,127,135,137

and the missing values have been filled in from other sources where

available138-140

. Values for Li are not available for some chemicals; to fill in these data gaps a

regression has been parameterized to predict the Li value from the other solute descriptors,

the details of this regression are provided in the Appendix (Section 5.4). Two additional

sources of solute descriptors are used which calculated the descriptors from literature values

for various partitioning coefficients; these are PCBs89

, and polychlorinated naphthalenes86

. In

total, solute descriptors have been obtained for 1460 individual chemicals.

Values for log KAW and log KOA have been calculated for all 1460 chemicals using equation

(9) and the phase descriptors from Table 7, and are plotted in the log KAW/log KOA chemical

space (Figure 10). The domain of reasonable model applicability is outlined in red. This

domain is defined primarily by computational costs; outside this range the time step required

to ensure stability of the model‘s numerical calculation is very short resulting in long

computation times. In addition, many of the chemicals outside of this range partition so

92

strongly to a single phase that multimedia fate modelling is a misguided exercise. Chemicals

falling outside of the range of reasonable model applicability were removed from the list

leaving 932 individual chemicals.

Figure 10 - Locations of the 1460 chemicals for which solute descriptors were obtained

in the log KOA/log KAW chemical space. Values for log KOA and log KAW are calculated

using equation (9) and the phase descriptors from Table 7. The reasonable domain of

model applicability is shown in red.

Many of the remaining 932 chemicals have very similar solute descriptors, meaning the

model results would also be very similar for these chemicals. To save computational costs

the list has been further reduced by removing chemicals with very similar solute descriptors.

93

The reduction method is pseudo-random and designed to preserve all of the variability in the

combinations of solute descriptors in the full data set. A sample set of 235 chemicals has

been selected and used to perform calculations. The list of these chemicals and the details of

the reduction procedure are provided in the Appendix (Section 5.5, Table 8).

A statistical analysis of the sample set of 235 chemicals shows that there are a number of

strong inter-correlations in the data, most notably between Li and Vi, and Li and Si (Table

10). This is because solute descriptors are not random but vary systematically within

chemical classes. This presents a problem for the evaluation of the model results. Because of

the systemic variation in solute descriptors any patterns observed in the model results may be

due to the nature of the data set, and not directly related to the properties of the PP-LFER

model. To help quantify the extent of this effect a second data set of hypothetical solute

descriptors has been created which contains no inter-correlations (Table 11) or systemic

variation and the results for this data set are also discussed. Hypothetical chemicals have

been created by taking a random number from within the range of each solute descriptor

contained in the data set of real chemicals. This second data set contains 246 hypothetical

chemicals. The solute descriptor combinations are provided in the Appendix (Table 9).

2.4 Model Parameters.

The parameterization used in this study is for an environment representing the Baltic Sea

drainage basin141

. This parameterization includes both the aquatic environment, composed of

two water and four sediment compartments, and the terrestrial environment, composed of

three soil compartments and two forest canopy compartments. A single air compartment is in

contact with all surface compartments. All environmental parameters are the same for both

the SP-LFER and PP-LFER models. The model details are provided in the Appendix

(Section 5.6, Table 12).

Aside from partitioning coefficients, the model requires the input of a chemical‘s

degradation half-lives. Such values cannot be realistically defined for the data set of

94

hypothetical chemicals, and so to ensure comparability between the data sets generic values

are used for all calculations. All chemicals are assumed to be perfectly persistent in all

compartments. This is not only required because persistence cannot be calculated for the

hypothetical chemicals but because it would have a confounding effect on the results;

chemicals with very similar solute descriptors might have widely different fates if their

susceptibility to degradation is different.

Simulated chemical fate can be widely different depending on which model compartment

receives the emissions142

. Three different emission scenarios are tested; continuous

emissions to air, agricultural soil and fresh water. Each simulation is run for 10 years and the

model outputs for the final year of the simulation are used to evaluate the results.

2.5 Model Comparison.

All partitioning coefficients required for the SP-LFER model are calculated using equation

(9) and the phase descriptors from Table 7. The advantage of this method is that there is no

need to consider experimental error in the partitioning coefficients when comparing the

results, because the only source of error will be from the solute descriptors and this will

equally affect the results of both the SP-LFER and PP-LFER models. This means that the

only difference between the model inputs is the replacement of octanol based SP-LFERs

with PP-LFERs, and any observed differences will be attributable to this alteration.

However, the disadvantage of this approach is that the full variability in the model inputs is

not captured for either model because the input data is derived from linear fits of

experimental data. This shortcoming should be considered when interpreting the results of

the model comparison.

The output parameter used to compare the models is the amount in various compartments in

moles. Using the amount of chemical as the primary model output is intuitive because it is

linearly related to environmental concentrations and inter-compartment fluxes. The amount

of chemical is summed together for like compartments to obtain five output amounts for air,

soil, forest canopy, water and sediments. Amounts are averaged for the entire final year of

the simulation. To display the model outputs as chemical space plots we first calculate the

95

log KOA and log KAW values of each chemical using equation (9) and the phase descriptors

from Table 7. As can be observed in Figure 10 the chemicals are unevenly scattered in the

chemical space, so the next step is to interpolate the model outputs being plotted in the

chemical space to a grid of evenly spaced log KOA and log KAW values. Finally, extrapolation

is used to fill in the gaps where interpolation is not possible. A custom interpolation and

extrapolation method has been devised to take into consideration the fact that moving in

different directions in the chemical space has different chemical meanings. The details of this

method are provided in the Appendix (Section 5.7). The validity of this method is primarily

demonstrated by the fact that the resultant plots are sensible and relatively smooth, but

further discussion of the validity can be also be found in the Appendix (Section 5.8).

2.6 Linking Solute Descriptors to Chemical Fate.

A PP-LFER model allows for the elucidation of the relationship between individual solute

descriptors and environmental fate. This is primarily investigated using sub-sampling

statistical analysis. The basic principle is this; a random sub-sample drawn from a data set of

chemicals should have the same mean value and standard deviation for any solute descriptor.

For example, the mean value and standard deviation of Bi in the hypothetical data set are

0.574 and 0.367 respectively. Now we take a sub-sample of this data set, all hypothetical

chemicals with Ai above 0.394, and the mean value and standard deviation of this sub-

sample are 0.550 and 0.367 respectively. The statistics are almost identical, indicating that

this was a random sub-sampling with respect to Bi. If the mean value of the sub-sample

deviates from the mean value of the entire data set then this is a biased sub-sample and if the

standard deviation is significantly reduced compared to the full data set this is a selective

sub-sample. Sub-sampling statistical analysis of chemicals based on PP-LFER model outputs

allows us to detect if environmental phases have biases or selectivity for chemicals with

certain solute descriptor values, linking molecular interactions to environmental fate.

96

3 Results and Discussion

3.1 Comparison of KOC Values.

Organic carbon/water partitioning coefficients (KOC) of all 932 chemicals within the model‘s

domain of reasonable applicability, as calculated in the two models are compared in Figure

11A. In the SP-LFER model, KOW is calculated using equation (9), and then KOC is

calculated with a regression derived by Seth et al.44

. In the PP-LFER model, the KOC is the

KHW taken directly as calculated with equation (9) and the phase descriptors in Table 7.

There is surprisingly little scatter in the plot and a very strong correlation between the SP-

LFER and PP-LFER values. This is due partly to a strong similarity in the phase descriptors

for KOW and KHW (Table 7), but is also due to systematic variations in the solute descriptors

of real chemicals which reduce the variability; the same plot for the hypothetical chemicals

(Figure 16A in the Appendix) shows far more scatter. From the regression shown in Figure

11A we expect that chemicals with low KOC values will partition more strongly to organic

carbon in the PP-LFER model, and chemicals with high KOC values will partition more

weakly to organic carbon. The SP-LFER used in CoZMo-POP2 to calculate KOC is: KOC =

0.35∙KOW44

. If the SP-LFER equation is instead changed to: KOC = 0.85 KOW then there is a

nearly perfect match with the PP-LFER results. This alternate equation is similar to an

equation initially presented by Seth et al. (equation 10 of ref.44

), which they discarded

because some of the data in the regression were suspected to be inaccurate.

97

Figure 11 - (A) Correlation between the log KOC values obtained by the SP-LFER

equation and values obtained from the PP-LFER obtained for humic acid. (B)

Correlation between the logKPA values obtained by the SP-LFER and values obtained

from the PP-LFER obtained for whole aerosol.

3.2 Comparison of KPA Values.

Figure 11B compares values for atmospheric particle/air partitioning coefficients (KPA)

calculated using SP-LFERs and PP-LFERs. The SP-LFER for sorption to atmospheric

particles in CoZMo-POP2 assumes that the organic matter in the aerosol has the same

sorption properties as octanol143

. The PP-LFER values for KPA are calculated using the dual

phase model (consisting of water-insoluble and aqueous fractions) by Arp et al.120

, with an

assumed relative humidity of 80% to calculate the aqueous volume fraction of the aerosol.

While most chemicals in Figure 11B are close to the 1:1 line, there is significantly more

scatter than in the corresponding plot for KOC. On the left side of Figure 11B (log KPA < 6)

the chemicals with the largest deviations have much higher KPA values predicted by the PP-

LFER equation. These chemicals are biased towards low Si, Li and Vi, high Ai and Bi values,

and also have low calculated KAW values (log KAW < -3.5). We conclude that absorption into

the aqueous fraction of the aerosol increases the overall KPA for these chemicals. It should be

noted that this is unlikely to affect the model results because sorption of chemicals with log

KPA < 6 to atmospheric particles will be negligible. On the right side of Figure 11B (log KPA

98

> 6) the chemicals with the largest deviations have lower KPA values predicted by the PP-

LFER equation. These chemicals are biased towards large molecules with hydrogen bond

accepting groups (high Bi, Li and Vi values) and also have low KAW values (log KAW < -3.5).

Despite partitioning strongly to water these chemicals have a significantly lower predicted

KPA using the dual phase PP-LFER equation. An inspection of the phase descriptors (Table

7) reveals the reason; KOA increases more rapidly with increasing Bi and Li than KQA, so we

conclude that partitioning into the aqueous phase of the aerosol fails to offset the decreased

sorption capacity of the water-insoluble fraction of the aerosol for these specific chemicals.

A plot similar to that shown in Figure 11B was created using the PP-LFER for partitioning to

atmospheric aerosol by Götz et al.117

(Figure 17 in the Appendix). The KPA from the dual-

phase absorption mechanism by Arp et al.121

is comparable to the KPA derived from KOA,

lying close to the 1:1 line, as opposed to the KPA from the model of Götz et al. which is

consistently higher by an order of magnitude. This is likely because the latter assumes

perfect additivity of various absorptive and adsorptive fractions, which is probably an

overestimate of the sorptive capacity of atmospheric particles. Arp et al. also have derived a

PP-LFER for other aerosol collected in southern Sweden121

, which is arguably more

appropriate for the current model parameterization. However, the mean absolute deviation in

calculated KPA values from the results of the equation used here is only 0.34 log units and the

fit is very similar, as can be seen in Figure 18, so we prefer to use the equation recommended

by Arp et al.

3.3 Comparison of Model Results.

Figure 12 shows the primary model outputs as six chemical space plots, all with log KOA on

the x-axis and log KAW on the y-axis. As there are no chemicals in the dataset with a log KOW

above approximately 11 (Figure 10), this area is blacked out in Figure 12. Anomalous

contour shapes observed in some of the plots are due to irregularly spaced chemical data.

The corresponding plots for hypothetical chemicals are provided in the Appendix (Figure

19). Figure 12A to C show the phase distribution of real chemicals in the PP-LFER model

expressed as a fraction of the total mass of chemical; the coloured areas are where more than

99

50% of a chemical is found in a single phase. No chemical is ever found to be more than

50% in the forest canopy. Figure 12D to F are plots of the absolute difference between the

SP-LFER and PP-LFER model results, which is calculated by summing the absolute

differences in model results for all five phases and then dividing by two. This represents the

fraction of chemical that shifts from one compartment to another when comparing the model

results.

In all three emission scenarios partitioning to air and water is very similar, the largest

absolute differences in the phase distribution occur at high log KOA values where chemicals

primarily partition to the three solid phases; forest canopy, soils and sediment. Two

mechanisms are responsible for most of these differences; deposition to the forest canopy,

and soil water run-off. In the region defined by log KOA > 4.5 and log KOW > 4.5 (marked as

region 1) some chemicals show enhanced partitioning to soil at the expense of partitioning to

sediments in the PP-LFER model (soil/sediment enhanced by factor of up to 2.2). This is

primarily due to decreased sorption to atmospheric particles (lower KPA in PP-LFER model)

which in turn results in a faster uptake in the canopy (gaseous deposition to the canopy is

faster than particle-bound deposition) and a greater flux of chemical to the soil with falling

foliage. In the region defined by log KAW < -1.5 and 1 < log KOW < 4.5 (marked as region 2)

chemicals show enhanced partitioning to either soil or sediments, which is due to either

enhanced or depressed sorption to particulate organic carbon in run-off water (differences in

KOC) depending on the solute descriptors of the chemical (soil/sediment enhanced or

depressed by up to a factor of 1.7).

100

Figure 12 - Chemical space plots of phase distribution of the PP-LFER based model

and the sum of absolute differences in the percent phase distribution between the SP-

LFER and PP-LFER based models for each of the three emission scenarios.

The primary effect of these mechanisms is that when emissions are to air or water, most of

the differences between the SP-LFER and PP-LFER models occur in the three solid phases

(Figure 12D and F). However, when emissions are to soil most of the largest absolute

differences between the models occur in the transition zones where chemicals are distributed

between the air or water and the solid phases (Figure 12E). One of the main effects of the

mode of emission is that the amount of chemical in the receiving compartment is elevated.

This can be seen in Figure 12A, B and C: the compartment receiving emissions makes up the

largest portion of the phase distribution in each case. Considering this effect, a general rule

can be stated as follows: Whereas the primary environmental fate of chemicals is similarly

101

predicted by both models, the largest relative differences in results are found mostly in the

environmental phases which contain only a small fraction of emitted chemical (less than

1%). This is because even small absolute differences in the phase distribution will have a

significant relative effect (up to a factor of; air: 10.9, canopy: 14.5, water: 5.29, soil: 11.8,

sediment: 1.73). This is demonstrated by plotting the fraction of chemicals in any single

phase and then overlaying this with a plot of the relative differences in model results for that

phase; the largest relative differences in model results are always located outside of the area

with the highest fraction of chemicals (Figure 20, Figure 21, and Figure 22 in the Appendix).

A good example is the environmental distribution of PCB-194 resulting from emissions to

water. An uncertainty analysis was performed for PCB-194 by calculating the 95%

confidence intervals of the SP-LFER parameters and performing additional simulations with

these values, however the same could not be done for the PP-LFER model as standard errors

were not available for the solute descriptors. In the SP-LFER model 98.03% (94.75%-

99.10%) of PCB-194 is in sediments and 0.008% (0.003%-0.025%) is in air, in the PP-LFER

model 94.44% is in sediments and 0.033% is in air. The primary environmental fate of PCB-

194 in both models is sorption to sediments; the PP-LFER model results are depressed by a

negligible factor of 1.04 (1.00-1.05) relative to the SP-LFER model results. However, if we

are specifically interested in air concentrations of PCB-194 the PP-LFER model results are

elevated by a significant factor of 4.24 (1.33-11.7) relative to the SP-LFER model results.

An additional simulation has been performed to test the possible effect of degradation. Real

chemicals are assigned uniform degradation half-lives of two days in air, two months in

water and six months in soils, sediments and forest canopy. Additionally, the Characteristic

Travel Distance (CTD) is calculated for each chemical as described in ref.72

. Chemical space

plots corresponding to Figure 12A and D are provided in the Appendix (Figure 23) along

with a plot of how CTD varies with location in the chemical space for both the SP-LFER and

PP-LFER models (Figure 24). The phase distribution shifts with the incorporation of

degradation because a smaller fraction of emissions makes their way from air to the

condensed phases. However, the absolute differences in model results follow the same

102

general pattern. Interestingly, the absolute difference in percent distribution of the SP-LFER

and PP-LFER models is smaller when degradation is included (up to a factor of 23.3).

Oxidative degradation in air is assumed to only affect chemical in the gas phase; particle

bound chemical is protected. This has the effect of making the relative amount of chemical in

the gas phase smaller and therefore chemicals are less susceptible to gas phase deposition to

forests, which has been identified above as one of the mechanisms causing differences in

model results. Another observation is that the CTD for particle-bound chemicals is enhanced

in the PP-LFER model versus the SP-LFER model (up to a factor of 2.61), due to a smaller

net loss to the surface media relative to the amount of particle-bound chemicals.

A major conclusion resulting from this work is that differences in the output from the two

models are relatively small; larger differences in model outputs have been demonstrated by

varying the environmental input parameters of a multimedia fate model129

. When using a

multimedia fate model as an evaluative or screening tool we recommend that the choice of

using a SP-LFER or PP-LFER model should be based on the quality of the available

chemical input values. The more accurate mechanistic description of partitioning in a PP-

LFER model should allow for better predictive power, but this will only become realized if

the quality of the solute descriptors is as good as, or better than, the quality of the octanol-

based partition coefficients.

3.4 Linking Solute Descriptors to Chemical Fate.

A simple graphical method of comparing how each solute descriptor affects the location of a

chemical in the log KAW/log KOA chemical space is to plot the vectors of the movement.

Because the solute descriptors are of different magnitudes plotting the vector of the phase

descriptors would be misleading, so the vectors must be normalized to the magnitudes of the

phase descriptors. Figure 13 shows the vectors in the log KAW/log KOA chemical space

corresponding to the range defined by the upper quartile and lower quartile of the possible

values of each solute descriptor. This can be interpreted as the range of movement in the

chemical space induced by varying each solute descriptor across the majority of the range of

103

its values. The plot shows that the Li descriptor has the largest effect on location in the

chemical space, which in turn controls chemical fate.

A more detailed investigation of the link between chemical fate and solute descriptors can be

achieved with sub-sampling statistical analysis. Both real and hypothetical chemicals were

divided into four sub-samples based on their simulated environmental fate; chemicals which

are found primarily in air in all three emission scenarios, chemicals found primarily in water,

chemicals found primarily in one or more of the solid phases (soil, sediment, forest canopy),

and multimedia chemicals which have different environmental fates depending on the mode

of emission. Mean values and standard deviations for the solute descriptors of the full data

sets and for all sub-samples are provided in Table 13 of the Appendix. A phase is considered

to be biased towards a solute descriptor if the mean value of the sub-sample deviates from

the value of the full data set by more than 10%, and the phase is considered to be selective if

the standard deviation of the sub-sample is more than 20% lower than the value for the full

data set.

Figure 13 - Vector plots of movement in the log KAW/log KOA chemical space that can be

related to the contributions of each phase descriptor. Vector magnitudes are the value

of the upper quartile minus the value of the lower quartile of the distribution of solute

descriptors multiplied by the corresponding phase descriptor.

S A B L V

0.45

-1.42

0.79

-0.83

0.59

-2.02

3.57

-1.88

1.89

-0.10

log KOA

log K

AW

104

Hypothetical chemicals favouring the air compartment are biased and selective towards low

values of Si and Li but show no bias or selectivity to the other solute descriptors.

Hypothetical chemicals partitioning primarily into water are biased towards high values of Si

and Bi and low values of Li and Vi, with selectivity for low Li values. Hypothetical chemicals

which are primarily sorbed to the three solid phases are biased towards low Bi values and

both biased and selective towards high Li and Vi values. Multimedia hypothetical chemicals

have no biases and are selective only to intermediate Li values. The biases and selectivity of

each phase are intuitive, for example chemicals found in the air tend to be those that only

weakly experience van der Waals dispersive interactions (Li) and dipole interactions (Si).

Chemicals sorbed to solid phases tend to be large (Vi) and experience strong van der Waals

dispersive interactions (Li). Chemicals which experience strong dipole interactions (Si) and

strong specific interactions (Bi) tend to be found in the water. Two features of this analysis

are of further interest; first there is no observable bias or selectivity in any of the sub-samples

for chemicals which are hydrogen bond donors (Ai). Additionally, the multimedia chemicals

are very close to a random sub-sampling of the full data set of hypothetical chemicals.

Statistical analysis of the sub-samples of the dataset of real chemicals reveals far more biases

and selectivity than the hypothetical data set; this is undoubtedly because inter-correlations

in solute descriptors of real chemicals cause additional biases in the phase distribution. The

first notable difference is the Ai solute descriptor; real chemicals in the water phase are

biased towards high Ai values, real chemicals in the air and solid phases are biased and

selective for low Ai values, and real multiphase chemicals are biased towards low Ai values.

To understand this, the data set of real chemicals was divided into two sub-samples;

chemicals that have Ai = 0 (n = 138) and chemicals that have Ai > 0 (n = 97). The mean and

standard deviation for the solute descriptors of these sub-samples are provided in Table 14 in

the Appendix. Chemicals with Ai > 0 are biased and selective for low values of Si and high

values of Bi, and biased towards low values of Li and Vi. This means that hydrogen bond

donors in the dataset are small chemicals and also hydrogen bond acceptors, in short there

are too many alcohols. This observation explains most of the differences between the real

and hypothetical datasets. The water phase is biased towards chemicals with high Bi, and low

Li and Vi, many of the chemicals with these properties also have high Ai values causing the

water phase to be biased towards real chemicals with high Ai values. A similar argument

105

explains the differences between the two data sets for chemicals sorbed to the solid phases.

The net effect of this is that chemicals with high Ai and Bi end up in the water phase, and

chemicals with high Si, Li and Vi end up sorbed to the solid phases. This causes real

chemicals in air to be biased and selective for chemicals with low values of all descriptors.

A general conclusion is that the PP-LFER model has no selectivity for multimedia

chemicals; they are simply those chemicals that are not strongly selected for by any single

phase. Real multimedia chemicals are biased and selective for large Li and Vi, and are biased

towards high Bi and low Ai. A closer inspection of the biases, however, shows that they are

smaller in magnitude than the biases of the air, water or solid phases, so we conclude that

these biases are the result of the data set being skewed by the selection process, as discussed

in detail in the Appendix. Multimedia chemicals tend to have intermediate solute descriptor

values because chemicals with relatively high or low values of any solute descriptor

generally experience a bias towards one or more environmental phases, as demonstrated by

the selectivity of the model for multimedia chemicals with intermediate Li values. The fact

that Li descriptor has such a strong influence on chemical fate may provide an explanation

for the strong similarity between the SP-LFER and PP-LFER model results: KOA-based SP-

LFERs likely capture the majority of the van der Waals interactions represented by the Li

descriptor and so on a large scale do a reasonable job of predicting environmental fate.


Industry Association (CEFIC). The study benefitted greatly from discussions with Kai-Uwe

Goss, Michael S. McLachlan and Knut Breivik.

106

5 Appendix

5.1 Derivation of Equations for Enthalpies of Phase Change

5.1.1 Derivation of PP-LFER Equations for ΔHAW, ΔHOA and ΔHOW

Solute descriptors and experimental ΔHAW values have been obtained from the Supporting

Information of Mintz et al.122

. As noted by Mintz et al. the L descriptor for one chemical,

erythritol, is unavailable, but a value has been calculated for this study using the regression

described elsewhere in the Appendix. Multiple linear regression is used to parameterize a

PP-LFER equation following the form proposed by Goss118

. Equation (11) is the PP-LFER

for ΔHAW: where S, A, B, L and V are the solute descriptors and the values in brackets are

the standard error associated with the phase descriptor.

ΔHAW = -0.86(±1.28)∙S + 33.63(±1.48)∙A + 43.79(±0.95)∙B

+ 1.52(±0.58)∙L + 16.63(±2.47)∙V + 8.59(±1.01)

(11)

(n = 369, R2 = 0.943, RMSE = 4.76 kJ/mol)

The standard error for s, the phase descriptor for polar/induced dipole interactions, is larger

than the phase descriptor which means that this phase descriptor has a negligible effect on

the value of ΔHAW. Despite this, the s descriptor is included in the equation to allow for the

calculation of ΔHOW by thermodynamic cycle.

Solute descriptors and experimental ΔHOA values have also been obtained from the

Supporting Information of Mintz et al. and a PP-LFER is parameterized by multiple linear

regression. The result is shown in equation (12).

ΔHOA = 6.04(±1.10)∙S – 53.66(±2.37)∙A – 9.19(±1.12)∙B

- 9.66(±0.34)∙L + 1.57(±1.40)∙V – 6.67(±0.74)

(12)

(n = 138, R2 = 0.989, RMSE = 2.58 kJ/mol)

Summing the solute descriptors from equations (11) and (12) gives ΔHOW by thermodynamic

cycle. This PP-LFER was tested by calculating ΔHOW for individual chemicals found in both

data sets by the same thermodynamic cycle. Equation (13) shows the PP-LFER and the

statistics for the comparison.

107

ΔHOW = 5.18∙S – 20.03∙A – 34.60∙B - 8.14∙L + 18.19∙V + 1.92 (13)

(n = 96, R2 = 0.784, RMSE = 4.54 kJ/mol)

All of the ΔHOW values are relatively close to zero so another test of the predictive power of

the regression is the correct prediction of the sign of ΔHOW. The correct sign is predicted for

84 of 96 chemicals which have available ΔHOW values, and in the remaining 12 cases the

values are very close to zero, within the range of -5 to 5 kJ/mol.

5.1.2 Derivation of PP-LFER Equations for ΔHHA and ΔHHW

Solute descriptors and ΔHHA values were obtained from the Supporting Information of

Niederer et al.119

. Equation (14) shows the PP-LFER equation parameterized for ΔHHA.

ΔHHA = -4.59 (±5.98)∙S – 31.87(±4.62)∙A – 17.81(±5.25)∙B

- 8.15(±2.28)∙L – 2.82(±9.42)∙V – 5.90(±5.57)

(14)

(n = 138, R2 = 0.595, RMSE = 10.99 kJ/mol)

PP-LFER equation (15) for ΔHHW has been derived by thermodynamic cycle from the

equations for ΔHHA and ΔHAW. Values of ΔHHW for have also been derived for individual

chemicals found in both data sets and these are used to test the regression.

ΔHHW = -5.45∙S + 1.76∙A + 25.98∙B - 6.63∙L + 13.81∙V + 2.69 (15)

(n = 60, R2 = 0.464, RMSE = 8.29 kJ/mol)

Values for ΔHHW are close to zero, as are the ΔHOW values, so the ability of the regression to

properly predict the sign was tested again. The correct sign is predicted for 50 of the 60

chemicals, and of the remaining chemicals an additional 7 chemicals have ΔHHW values

close to zero, within the range of -5 to 5 kJ/mol.

5.1.3 Derivation of a PP-LFER Equation for ΔHPA

Arp et al. recommend using the enthalpy of vaporization (ΔvapH) to predict the temperature

dependence of sorption to atmospheric particles121

. Values for ΔvapH were obtained from the

supporting information of Goss and Schwarzenbach130

and solute descriptors were obtained

108

from the database of solute descriptors described in the main text of this work. A total of 161

chemicals were found which had both ΔvapH values and solute descriptors.

Goss and Schwarzenbach note a number of outliers when ΔvapH is regressed versus the

natural log of liquid vapour pressure (ln VP) and attribute this to deviations from Trouton‘s

rule because of inter-molecular interactions (such as dimerization) in the gas phase or the

liquid phase130

. Some of these same chemicals are also in the dataset of Arp et al. (e.g. acetic

acid) and show the strongest deviations between ΔvapH and ΔHPA121

. We suggest that this is

because the experimental concentrations of Arp et al. were too low to allow for these kinds

of inter-molecular interactions, causing these chemicals to essentially obey Trouton‘s Rule

and deviate from the experimental ΔvapH values. In light of this we attempt to create a PP-

LFER for ΔvapH which intentionally neglects any deviations from Trouton‘s Rule.

The predicted ΔvapH values for a number of alcohols and nitroanilines had strong negative

deviations from the experimental values (ethanol, propanol, butanol, pentanol, hexanol,

octanol, cyclohexanol, 2-nitroaniline, 3-nitroaniline, 4-nitroaniline), and because they are all

relatively strong hydrogen bond donors and acceptors they very likely do not obey Trouton‘s

rule. In this case it would be because of intermolecular interactions in the liquid phase. The

ΔvapH of two carboxylic acids and two chlorophenols had strong positive deviations from the

experimental values and are also both hydrogen bond donors and acceptors indicating

possible intermolecular interactions in the gas phase (acetic acid, butanoic acid, 3-

chlorophenol, 4-chlorophenol). All of these chemicals were removed from the dataset. Four

other chemicals were identified as outliers, most likely due to erroneous ΔvapH values, and

removed from the correlation (hexachlorobenzene, 2,3-benzofuran, fluoranthene, 1-

propylamine). The most obvious of these was hexachlorobenzene, for which the

experimental value was given as 98.61 kJ/mol. The ln VP of hexachlorobenzene provided in

Goss and Schwarzenbach was -5.93130

, but in Shen and Wania the value presented is -2.3692

.

This value when used in the correlation of Goss and Schwarzenbach gives a ΔvapH value of

73.27 kJ/mol, which is close to the prediction of the PP-LFER equation. The PP-LFER

equation derived from the remaining data points is presented in equation (16):

ΔvapHPPLFER = 14.04 (±1.32)∙S + 20.74 (±1.44)∙A + 0.02 (±1.09)∙B (16)

109

+ 3.36 (±0.52)∙L + 20.23 (±1.94)∙V + 2.89

(n = 143, R2 = 0.995, RMSE = 1.82 kJ/mol)

Comparing the experimental values of ΔvapH to the ΔHPA values of Arp et al. we find the

relation shown in equation (17):

ΔHPA = (-0.995 ± 0.135) ΔvapH (17)

(n = 38, R2 = 0.297, RMSE = 8.46 kJ/mol)

Equation (18) shows the same relationship using the ΔvapH values predicted from equation

(16):

ΔHPA = (-0.988 ± 0.100) ΔvapHPPLFER (18)

(n = 38, R2 = 0.476, RMSE = 6.42 kJ/mol)

Most of the improvement in the correlation statistics is due to a dramatically improved

prediction of the ΔHPA values for chemicals which are known to deviate from Trouton‘s rule.

5.2 Calculation of Z-values in CoZMo-POP 2

As a fugacity based model, environmental phase partitioning in CoZMo-POP 2 is not

described directly by partitioning coefficients but instead by fugacity capacities (Z, mol m-3

Pa-1

). A detailed overview of the usage and nomenclature of fugacity based environmental

models is given by Mackay15

.

Seth et al.44

derived the following KOW based SP-LFER, equation (19), for partitioning

between organic carbon and water (KOC).

KOC = 0.35•KOW (19)

Equation (20) shows how KOC is used to calculate the fugacity capacity of particulate organic

carbon (ZPOC), which is in turn used to calculate partitioning to soil and sediment organic

carbon, as well as to organic carbon suspended in the water column.

110

ZPOC = ZWATER • KOC • δOC / 1•106 kg/L (20)

In these equations KOC is in units of L/kg, δOC is the density of organic carbon in g m-3

and

1•106 kg/L is a unit correction. To replace this SP-LFER equation with a PP-LFER equation

KOC is simply replaced with KHW yielding equation (21).

ZPOC = ZWATER • KHW • δOC / 1•106 kg/L (21)

Sorption to atmospheric particles in CoZMo-POP 2 follows the SP-LFER of Finizio et al.143

,

which essentially assumes that the organic matter in atmospheric aerosol has the same

sorption properties as octanol; equation (22) shows the form of the equation.

ZP = ZAIR • KOA • VFOQ • δOC / δoctanol (22)

In this equation VFOQ is the volume fraction of organic carbon in aerosol and δoctanol is the

density of octanol in g m-3

. A modified version of the two phase PP-LFER model of aerosol

presented by Arp et al. is used to describe sorption to aerosol120

, but assuming that there is no

ionization or ―salting out‖ effect. The volume of the water-soluble phase relative to the

water-insoluble phase is calculated with the growth factor (GF), which is a function of the

relative humidity (RH). The growth factor is calculated using an exponential fit of the data

provided by Arp et al.120

, shown in equation (23).

GF = 1 + 0.0245 • e (4.033 • RH)

(23)

An equation for partitioning to whole aerosol (KPA, m3 g

-1) is obtained by substituting the

growth factor into equation (23) of reference 120, producing equation (24).

KPA = KQA + (GF – 1) / KAW • δwater (24)

In this equation KQA is the partitioning coefficient between air and the water-insoluble

fraction of aerosol in units of m3 g

-1, and δwater is the density of water in g m

-3. It should be

noted that KQA was parameterized at 15 ºC, so KAW is temperature corrected to 15 ºC as well

when applying equation (24). Equation (25) shows the PP-LFER modified calculation of ZP

where δaerosol is the density of whole aerosol in units of g m-3

.

111

ZP = ZAIR • KPA • δaerosol (25)

All other Z values are calculated as described by Wania et al128

, the only difference is that all

partitioning coefficients are calculated with equation (9) and the phase descriptors from

Table 7 of the main publication.

5.3 PP-LFER for Sorption to Forest Canopy

Two equations of type (26) were derived by Horstmann and McLachlan to estimate

partitioning between deciduous (KDEC/AIR) and coniferous (KCON/AIR) forest canopies and

air132

.

KCANOPY/AIR = M • KOAN (26)

This equation can be converted to the log-linear form shown in equation (27);

log KCANOPY/AIR = N ∙ log KOA + m (27)

Where m is equal to log M and N is the same value from equation (27). The values of the

regression coefficients from Horstmann and McLachlan were m = 1.58 and N = 0.69 for

coniferous canopy and m = 1.15 and N = 0.76 for deciduous canopy. A similar study by Su

et al. for a deciduous canopy calculated a regression with coefficients of m = 2.04 and N =

0.67144

. A study by Welke et al., from which Platts et al. derived the data for their PP-LFER,

also calculates a log linear regression with coefficients m = 0.820 and N = 0.668145

. Figure

14 shows all of these regressions, where it can clearly be seen that the regression of Welke et

al. is dissimilar from the forest canopy regressions. Tomato cuticle clearly has different

partitioning properties from forest canopy, and for this reason we prefer to use the regression

of Horstmann and McLachlan.

Two additional regressions were derived using PP-LFER equations to test the effect the

training set of chemicals has on the resulting regression with KOA. The dataset Horstmann

and McLachlan used to derive their equations included a number of PAHs, HCB, PCBs and

PCDD/PCDFs. Solute descriptors were found in the database outlined in the main text of this

work for the PAHs, HCB, and PCBs, but none could be located for PCDD/PCDFs, so PCB-

112

194 and PCB-209 were added to the dataset to extend it to less volatile chemicals. The PP-

LFER of Platts et al. for partitioning between tomato cuticle matrix and air (KMXa/AIR) was

used to calculate values for KMXa/AIR for this dataset133

, and the KOA values were calculated

from the PP-LFER of Goss118

. The regression coefficients are m = -0.41 and N = 1.19.

The second regression was performed on the dataset of Welke et al.; solute descriptors were

found for 54 of the chemicals used by Welke et al. in the database described in the main text.

These solute descriptors were then used to calculate KOA values as described above, and a

regression of the form shown in equation (27) was calculated using the experimental KMXa

values. The regression coefficients are m = -0.38 and N = 1.02. The regression of log KMXa

with log KOA is similar for both the non-polar organic dataset of Horstmann and Mclachlan

and the more diverse dataset of Welke et al., so the regression for the forest canopies should

similarly be applicable to a more diverse dataset

113

Figure 14 - Linear Regressions of Plant Sorption with log KOA

Plots of the log-linear regressions of log KCANOPY/AIR with log KOA of Horstmann and

McLachlan132

and Su et al.144

, and the regression of log KMXa/AIR with log KOA of Welke

et al.145

.

5.4 PP-LFER Equation for Hexadecane/Air Partitioning

A large number of chemicals are missing values for the L descriptor. This descriptor is

unique in that it is the only Abraham descriptor that is itself a partitioning coefficient,

implying that it is possible to derive a PP-LFER to describe the interactions involved in the

partitioning process. To our knowledge this has not been done directly, but an equation can

be drawn indirectly from the literature. Abraham et al. provide an equation for partitioning

between hexadecane and water (K16/W) using the phase descriptors sjk, ajk, bjk, ejk and vjk (ref

2

4

6

8

10

12

14

2 4 6 8 10 12 14

log KOA

H&M Dec.

H&M Con.

Su et al.

Welke et al.

1:1

log K

Pla

nt/

AIR

114

45, equation 8). A regression for partitioning between water and air (KWA) using the same

phase descriptors has also been presented125

. Assuming that these processes form a complete

thermodynamic cycle, equation (28) for L (K16/A) can be derived by adding the phase

descriptors of the two equations:

log K16/A = 0.932∙S + 0.226∙A – 0.028∙B + 1.244∙E + 3.564∙V – 0.907 (28)

The two phase descriptors describing hydrogen bonding nearly cancel out; the magnitudes of

ajk and bjk are approximately -3.6 and -4.9 respectively in the PP-LFER for K16/W and 3.8 and

4.8 respectively in the PP-LFER for KWA. This is expected, as hydrogen bonding should play

a negligible role in the partitioning between hexadecane and air. To completely eliminate

hydrogen bonding a new PP-LFER was parameterized using most of the chemicals with

available experimental L values from the database outlined in the main text of this work.

Many of the solute descriptors for PCBs and PCNs were left out of the regression because

they would have skewed the dataset; the chemicals retained were those with the most

experimental data used in the original derivation of the solute descriptors. The regression

was performed only versus the S, E and V descriptors; this equation is shown in equation

(29). E is the excess molar refraction, a measure of polarizability.

log K16/A (L) = 0.951(±0.056)∙S + 1.205(±0.041)∙E

+ 3.492(±0.045)∙V – 0.816(±0.044)

(29)

(n = 621, R2 = 0.979, RMSE = 0.329)

The phase descriptors in (29) almost exactly match those of the literature derived equation

(28). The fact that it is possible to create a regression between the solute descriptors that has

such excellent statistics reveals strong inter-correlations among the descriptors, which is not

desirable for multiple linear regressions. Using this regression to fill in the missing L values

will only increase the degree of inter-correlation in the dataset, but it does include chemical

information not found in the other equations used in this work, namely the E descriptor.

5.5 Dataset Reduction Method

Two criteria were used to determine if two chemicals were similar enough to eliminate one

of the chemicals from the dataset; the first is the distance between the chemicals in the log

115

KOA / log KAW chemical space, and the second is the differences in their solute descriptors.

The full list of 932 chemicals considered was randomly ordered using Microsoft Excel by

assigning each chemical a random number and then sorting the list by the random number.

The mean and standard deviation of each solute descriptor was calculated for the full set of

932 chemicals and then the standard deviation was used to normalize all of the solute

descriptors to allow the deviations of different descriptors to be directly compared.

Each chemical is compared sequentially to every other chemical in the list that has not

already been identified as redundant and the following determined:

• If the distance between the chemicals in the log KOA / log KAW

chemical space is below a given threshold the criteria for distance is

considered met.

• The sum of the deviations between the normalized solute descriptors

of the chemicals is calculated.

• The maximum deviation between the normalized solute descriptors of

the chemicals is calculated.

• If both the sum and the maximum deviations are below the thresholds

the criteria for similar solute descriptors is considered met.

If both the distance and the similar solute descriptor criteria are met for any other chemical in

the list the current chemical is identified as redundant and the next chemical in the list is

considered. The thresholds used are distance = 0.5 units, sum of differences = 1.2 standard

deviations, maximum difference = 0.6 standard deviations. These thresholds were set to

reduce the list to a manageable size while not eliminating too much of the variability in the

dataset. In the area defined by log KOA > 6 and log KAW > -3.5 (i.e. the area containing the

PCBs and PCNs) the chemicals are much more numerous and closely spaced, so in this area

the thresholds were divided by two in order to retain more of the chemicals.

Table 8 - Real Chemicals Dataset

CAS NAME S A B L V Ref. Ref. L

54115 Nicotine 0.88 0.00 1.09 5.85 1.37 137

a

55389 Fenthion 1.75 0.00 0.65 9.43 1.99 127

a

116

57103 Hexadecanoic acid 0.60 0.60 0.45 8.31 2.44 45

a

57114 Octadecanoic acid 0.60 0.60 0.45 9.27 2.72 45

a

59507 4-Chloro-3-methylphenol 1.02 0.65 0.23 5.29 1.04 125

125

62533 Aniline 0.96 0.26 0.41 3.93 0.82 45

138

64186 Formic acid 0.60 0.75 0.38 1.25 0.32 45

a

64197 Acetic acid 0.65 0.61 0.45 1.75 0.46 125

125

67561 Methanol 0.44 0.43 0.47 0.97 0.31 125

125

68122 N,N-Dimethylformamide 1.31 0.00 0.73 3.17 0.65 125

125

69727 Salicylic acid 0.84 0.71 0.38 4.51 0.99 137

a

72208 Endrin 1.35 0.00 0.61 9.66 2.01 127

a

74953 Dibromomethane 0.67 0.10 0.10 2.89 0.60 125

125

75116 Diiodomethane 0.69 0.05 0.23 3.85 0.77 45

139

75898 2,2,2-TrifluoroethanoI 0.60 0.57 0.25 1.22 0.50 125

125

76017 Pentachloroethane 0.66 0.17 0.06 4.27 1.00 125

125

77101 phencyclidine 0.97 0.00 0.82 9.09 2.15 137

a

78400 Triethyl phosphate 1.00 0.00 1.06 4.75 1.39 125

125

78819 Isobutylamine 0.32 0.16 0.63 2.42 0.77 45

a

78875 1,2-Dichloropropane 0.60 0.10 0.11 2.86 0.78 125

125

79118 Chloroacetic acid 1.08 0.74 0.36 2.71 0.59 45

a

79243 Nitroethane 0.95 0.02 0.33 2.41 0.56 125

125

79345 1,1,2,2-Tetrachloroethane 0.76 0.16 0.12 3.80 0.88 125

125

80466 4-tert-Pentylphenol 0.89 0.56 0.41 6.17 1.48 45

a

83329 Acenaphthene 1.04 0.00 0.20 6.47 1.26 125

125

84662 Diethyl phthalate 1.40 0.00 0.88 7.37 1.71 45

a

86737 Fluorene 1.03 0.00 0.20 6.92 1.36 125

125

88095 2-Ethylbutanoic acid 0.57 0.60 0.50 3.53 1.03 45

a

88744 2-Nitroaniline 1.37 0.30 0.36 5.63 0.99 125

125

88857 Dinoseb 1.75 0.17 0.46 7.82 1.69 127

a

89714 Methyl 2-methylbenzoate 0.87 0.00 0.43 5.05 1.21 45

138

91203 Naphthalene 0.92 0.00 0.20 5.16 1.09 125

125

91236 2-Nitroanisole 1.47 0.22 0.30 5.44 1.09 127

a

92240 Naphthacene 1.70 0.00 0.33 10.75 1.82 45

138

92671 4-Aminobiphenyl 1.48 0.26 0.48 7.45 1.42 45

a

92944 p-Terphenyl 1.48 0.00 0.30 9.69 1.93 45

138

93550 Ethyl phenyl ketone 0.95 0.00 0.51 5.09 1.16 45

a

93992 Phenylbenzoate 1.42 0.00 0.47 7.51 1.54 45

a

94257 n-Butyl 4-aminobenzoate 1.47 0.32 0.59 7.38 1.60 45

a

94257 butyl 4-aminobenzoate 1.20 0.29 0.75 7.11 1.60 137

a

95578 2-Chlorophenol 0.88 0.32 0.31 4.18 0.90 125

125

95932 1,2,4,5-Tetramethylbenzene 0.60 0.00 0.19 5.11 1.28 45

a

96480 gamma-Butyrolactone 1.74 0.00 0.45 3.51 0.64 45

a

98566 4-Chlorobenzotrifluoride 0.58 0.00 0.01 3.98 1.03 137

a

99081 3-Nitrotoluene 1.10 0.00 0.28 5.10 1.03 125

125

99661 Valproic acid 0.57 0.60 0.50 4.47 1.31 137

a

100174 4-Nitroanisole 1.32 0.20 0.24 5.29 1.09 127

a

100469 Benzylamine 0.88 0.10 0.72 4.32 0.96 45

138

117

100481 4-Cyanopyridine 1.21 0.00 0.59 4.03 0.83 125

125

100845 3-Methylanisole 0.78 0.00 0.30 4.39 1.06 45

138

101815 Diphenylmethane 1.04 0.00 0.28 6.31 1.47 45

138

101848 Diphenyl ether 1.08 0.00 0.20 6.29 1.38 45

138

101973 Ethyl phenylacetate 1.01 0.00 0.57 5.67 1.36 45

a

103297 1,2-Diphenylethane 1.03 0.00 0.28 6.76 1.61 45

138

103833 N,N-Dimethylbenzylamine 0.80 0.00 0.69 5.08 1.24 45

a

104723 n-Decylbenzene 0.47 0.00 0.15 7.75 2.13 45

a

104949 4-Methoxyaniline 1.10 0.23 0.65 4.95 1.02 45

138

106445 p-Cresol 0.87 0.57 0.31 4.20 0.92 125

125

106478 4-Chloroaniline 1.13 0.30 0.32 4.89 0.94 125

125

106934 1,2-Dibromoethane 0.76 0.10 0.17 3.39 0.74 125

a

107186 Prop-2-en-1-ol 0.46 0.38 0.48 1.94 0.55 45

a

107197 propargyl alcohol 0.57 0.38 0.59 1.97 0.50 137

a

108861 Bromobenzene 0.73 0.00 0.09 4.04 0.89 125

125

108894 4-Methylpyridine 0.82 0.00 0.55 3.64 0.82 125

125

108918 Cyclohexylamine 0.56 0.16 0.58 3.80 0.95 125

125

108930 Cyclohexanol 0.54 0.32 0.57 3.76 0.90 125

125

109024 N-Methylmorpholine 0.74 0.00 0.90 3.30 0.86 125

a

109977 Pyrrole 0.73 0.41 0.29 2.63 0.58 45

a

110565 1,4-Dichlorobutane 0.95 0.00 0.17 3.79 0.92 125

a

110587 n-Pentylamine 0.35 0.16 0.61 3.14 0.91 125

125

110623 Pentanal 0.65 0.00 0.45 2.85 0.83 125

125

110805 2-Ethoxyethanol 0.50 0.30 0.83 2.82 0.79 45

138

110894 Piperidine 0.46 0.10 0.69 3.30 0.80 45

138

110918 Morpholine 0.79 0.06 0.91 2.98 0.72 125

a

111717 Heptanal 0.65 0.00 0.45 3.87 1.11 125

125

111762 2-butoxyethanol 0.50 0.30 0.83 3.81 1.07 125

125

111842 n-Nonane 0.00 0.00 0.00 4.18 1.38 125

125

111875 Octan-1-ol 0.42 0.37 0.48 4.62 1.30 45

139

111922 Di-n-butylamine 0.30 0.08 0.68 4.35 1.34 125

125

112129 2-Undecanone 0.68 0.00 0.51 5.73 1.67 125

125

112298 1-Bromodecane 0.40 0.00 0.12 5.87 1.69 45

a

112301 Decan-1-ol 0.42 0.37 0.48 5.63 1.58 45

139

112403 n-Dodecane 0.00 0.00 0.00 5.70 1.80 45

139

112425 Undecan-1-ol 0.42 0.37 0.48 5.80 1.72 45

a

112538 Dodecan-1-ol 0.42 0.37 0.48 6.28 1.86 45

a

112709 Tridecan-1-ol 0.42 0.37 0.48 6.77 2.00 45

a

112721 Tetradecan-1-ol 0.42 0.37 0.48 7.25 2.14 45

a

118741 Hexachlorobenzene 0.99 0.00 0.00 7.62 1.45 45

138

119335 4-methyl-2-nitrophenol 1.04 0.06 0.40 5.22 1.09 135

a

120127 Anthracene 1.34 0.00 0.26 7.57 1.45 125

125

120729 Indole 1.12 0.44 0.22 5.51 0.95 45

138

121142 2,4-Dinitrotoluene 1.27 0.07 0.51 6.01 1.21 127

a

121448 Triethylamine 0.15 0.00 0.79 3.04 1.05 125

125

123320 2,5-Dimethylpyrazine 0.90 0.00 0.69 3.99 0.92 45

a

118

123864 n-Butyl acetate 0.60 0.00 0.45 3.35 1.03 125

125

126738 Tri-n-butyl phosphate 0.90 0.00 1.21 7.74 2.24 45

a

129000 Pyrene 1.71 0.00 0.29 8.83 1.58 125

125

131113 Dimethyl phthalate 1.41 0.00 0.88 6.05 1.43 45

138

132650 Dibenzothiophene 1.31 0.00 0.18 7.58 1.38 45

138

134327 1-Naphthylamine 1.26 0.20 0.57 6.49 1.19 125

125

135193 2-Naphthol 1.08 0.61 0.40 6.20 1.14 125

125

135988 1-methylpropyl-benzene 0.48 0.00 0.16 4.51 1.28 125

125

136607 n-Butyl benzoate 0.80 0.00 0.46 5.97 1.50 45

a

139662 Diphenyl sulfide 1.20 0.00 0.30 7.48 1.49 45

a

140294 Phenylacetonitrile 1.15 0.00 0.45 4.70 1.01 45

138

142289 1,3-DichIoropropane 0.74 0.00 0.17 3.09 0.78 45

a

142847 Di-n-propylamine 0.30 0.08 0.69 3.35 1.05 45

138

142961 Di-n-butyl ether 0.25 0.00 0.45 3.92 1.29 125

125

143088 Nonan-1-ol 0.42 0.37 0.48 5.12 1.44 45

139

150196 3-Methoxyphenol 1.17 0.59 0.38 4.80 0.97 125

125

198550 Perylene 1.76 0.00 0.40 12.05 1.95 45

138

206440 Fluoranthene 1.55 0.00 0.20 8.83 1.59 45

138

243196 Benzo[b]fluorene 1.57 0.00 0.20 9.52 1.73 45

138

253667 Cinnoline 1.00 0.00 0.78 5.18 1.00 45

a

253827 Quinazoline 1.15 0.00 0.65 4.90 1.00 45

a

260946 Acridine 1.33 0.00 0.58 7.64 1.41 45

138

271896 Benzofuran 0.83 0.00 0.15 4.36 0.91 45

138

275514 Azulene 1.17 0.15 0.16 5.71 1.09 45

138

288131 Pyrazole 1.00 0.54 0.45 3.15 0.54 45

138

288471 Thiazole 0.80 0.00 0.45 3.00 0.60 45

a

289805 Pyridazine 0.85 0.00 0.81 3.43 0.63 45

138

290379 Pyrazine 0.95 0.00 0.62 2.92 0.63 45

138

292648 Cyclooctane 0.10 0.00 0.00 4.33 1.13 45

138

305033 tert-butyl chlorambucil 1.49 0.03 0.89 11.90 2.83 137

a

319846 alpha-HCH 1.20 0.00 0.47 7.34 1.58 134

134

319857 beta-HCH 1.18 0.12 0.58 7.63 1.58 134

134

333415 Diazinon 1.01 0.00 1.19 9.45 2.31 127

a

334485 Decanoic acid 0.60 0.60 0.45 5.46 1.59 45

a

367124 2-Fluorophenol 0.69 0.61 0.26 3.45 0.79 125

125

371415 4-Fluorophenol 0.97 0.63 0.23 3.84 0.79 125

125

402675 3-Fluoronitrobenzene 1.09 0.00 0.35 4.10 0.89 45

a

502410 Cycloheptanol 0.54 0.32 0.58 4.41 1.05 125

125

506309 Eicosanoic acid 0.60 0.60 0.45 10.24 3.00 45

a

512561 Trimethyl phosphate 1.10 0.00 1.00 3.76 0.97 45

a

513086 Tri-n-propyl phosphate 1.00 0.00 1.15 6.42 1.82 45

a

527548 3,4,5-Trimethylphenol 0.88 0.55 0.44 5.57 1.20 45

138

533584 2-Iodophenol 1.00 0.40 0.35 4.96 1.03 125

125

536743 Phenylethyne 0.58 0.12 0.24 3.74 0.91 45

a

537462 methamphetamine 0.79 0.09 0.70 5.83 1.38 137

a

542552 Isobutyl formate 0.60 0.00 0.40 2.79 0.89 125

125

119

544252 Cyclohepta-1,3,5-triene 0.46 0.00 0.20 3.53 0.86 45

a

544638 Tetradecanoic acid 0.60 0.60 0.45 7.35 2.16 45

a

555033 3-Nitroanisole 1.20 0.14 0.25 5.18 1.09 127

a

575371 1,7-Dimethylnaphthalene 0.89 0.00 0.20 6.45 1.37 45

a

583539 1,2-Dibromobenzene 0.96 0.00 0.04 5.46 1.07 45

138

584021 Pentan-3-ol 0.36 0.33 0.56 2.83 0.87 45

a

589184 4-Methylbenzyl alcohol 0.88 0.33 0.60 4.69 1.06 45

a

591355 3,5-Dichlorophenol 1.17 0.77 0.00 5.65 1.02 45

138

601898 2-Nitroresorcinol 1.35 0.01 0.48 5.44 1.01 135

a

608935 Pentachlorobenzene 0.96 0.00 0.00 6.72 1.33 45

138

616477 N-Methylimidazole 0.95 0.00 0.83 3.81 0.68 45

138

623370 3-Hexanol 0.36 0.33 0.56 3.34 1.01 125

125

623427 Methyl butanoate 0.60 0.00 0.45 2.89 0.89 125

125

625809 Diisopropyl sulfide 0.32 0.00 0.37 3.60 1.12 125

125

626028 3-Iodophenol 1.20 0.70 0.18 5.53 1.03 45

138

628057 1-Nitropentane 0.95 0.00 0.29 3.94 0.99 125

125

628717 Hept-1-yne 0.23 0.13 0.10 3.00 1.01 125

125

628819 Ethyl n-butyl ether 0.25 0.00 0.45 2.97 1.01 45

a

629594 n-Tetradecane 0.00 0.00 0.00 6.71 2.08 45

139

630206 1,1,1,2-Tetrachloroethane 0.63 0.10 0.08 3.64 0.88 125

125

634662 1,2,3,4-TetrachIorobenzene 0.92 0.00 0.00 6.17 1.21 125

125

644086 4-Methylbiphenyl 0.98 0.00 0.23 6.61 1.47 45

138

700129 Pentamethylbenzene 0.66 0.00 0.20 5.80 1.42 45

a

758963 N,N-Dimethylpropanamide 1.30 0.00 0.78 4.07 0.93 45

a

768956 Adamantan-1-ol 1.20 0.32 0.56 5.71 1.25 45

a

779022 9-Methylanthracene 1.30 0.00 0.26 8.44 1.60 45

138

781431 9,10-Dimethylanthracene 1.30 0.00 0.26 9.28 1.74 45

138

829265 2,3,6-Trimethylnaphthalene 0.86 0.00 0.21 6.99 1.51 45

a

830137 Cyclododecanone 0.86 0.00 0.56 6.67 1.71 45

a

994058 tert-Amyl methyl ether 0.21 0.00 0.60 2.98 1.01 137

a

1122549 4-Acetylpyridine 1.13 0.00 0.84 4.66 0.97 125

125

1582098 Trifluraline 0.40 0.00 1.35 8.54 2.20 127

a

1730376 1-Methylfluorene 1.06 0.00 0.20 7.44 1.50 45

138

1731846 Methyl nonanoate 0.60 0.00 0.45 5.32 1.59 45

24

1825305 PCN-6 1.06 0.00 0.09 6.78 1.33 86

86

1836755 Nitrofen 0.82 0.00 0.85 8.40 1.80 127

a

2050682 PCB-15 1.18 0.00 0.16 7.58 1.57 89

89

2050740 PCN-9 1.12 0.00 0.11 7.15 1.33 86

86

2051243 PCB-209 2.26 0.00 0.02 11.70 2.55 89

89

2051607 PCB-1 1.07 0.00 0.20 6.34 1.45 89

89

2051629 PCB-3 1.05 0.00 0.18 6.72 1.45 89

89

2189608 n-Octyl benzene 0.48 0.00 0.15 6.78 1.84 45

a

2234131 PCN-75 1.54 0.00 0.00 12.88 2.06 86

86

2303175 Triallate 0.92 0.20 0.75 8.77 2.12 127

a

2416946 2,3,6-Trimethylphenol 0.78 0.36 0.41 5.14 1.20 135

a

2819865 Pentamethylphenol 0.90 0.35 0.43 6.29 1.48 135

a

120

2921882 Chlorpyrifos 0.34 0.00 1.13 8.75 2.15 127

a

2986198 S-methylisothiourea 0.62 0.26 0.84 3.29 0.71 137

a

3376247 Phenyl-N-tert-butylnitrone 1.03 0.01 0.49 6.41 1.54 137

a

3452093 Non-1-yne 0.23 0.12 0.10 4.09 1.29 45

a

4658280 Aziprotryn 1.12 0.11 0.87 7.90 1.63 136

136

4920778 3-methyl-2-nitrophenol 1.01 0.14 0.48 5.19 1.09 135

a

7732185 Water 0.45 0.82 0.35 0.26 0.17 125

125

13429248 hexafluoropropene 0.55 0.77 0.10 1.39 0.70 125

125

15687271 Ibuprofen 0.59 0.59 0.81 6.83 1.78 137

a

26530201 Octhilinone 1.69 0.13 0.62 8.10 1.79 127

a

32598111 PCB-70 1.46 0.00 0.13 8.69 1.81 89

89

32774166 PCB-169 1.70 0.00 0.07 10.61 2.06 89

89

33091177 PCB-197 2.00 0.00 0.06 10.10 2.30 89

89

33146451 PCB-10 1.22 0.00 0.20 6.77 1.57 89

89

34622587 Orbencarb 1.17 0.00 0.86 8.75 1.96 127

a

35256850 Tebutam 1.11 0.02 1.12 8.54 2.10 127

a

35694087 PCB-194 2.00 0.00 0.06 11.19 2.30 89

89

38379996 PCB-95 1.61 0.00 0.13 8.63 1.94 89

89

38444778 PCB-32 1.35 0.00 0.17 7.67 1.69 89

89

38444905 PCB-37 1.31 0.00 0.13 8.39 1.69 89

89

39635319 PCB-189 1.85 0.00 0.07 10.89 2.18 89

89

39635353 PCB-159 1.72 0.00 0.09 9.96 2.06 89

89

40186729 PCB-206 2.13 0.00 0.04 11.45 2.43 89

89

41411625 PCB-160 1.74 0.00 0.11 9.59 2.06 89

89

41411647 PCB-190 1.87 0.00 0.09 10.45 2.18 89

89

44604902 S-2-fluoroethylisothiourea 0.60 0.27 0.81 3.71 0.87 137

a

51570446 PCN-16 1.18 0.00 0.00 7.70 1.45 86

86

51570457 PCN-33 1.24 0.00 0.00 8.27 1.58 86

86

52663782 PCB-195 2.00 0.00 0.06 10.83 2.30 89

89

52663793 PCB-207 2.13 0.00 0.04 10.90 2.43 89

89

52744135 PCB-135 1.74 0.00 0.11 9.29 2.06 89

89

53555638 PCN-28 1.24 0.00 0.00 8.54 1.58 86

86

54910893 Fluoxetine 1.30 0.10 0.93 9.45 2.24 137

a

55720382 PCN-22 1.12 0.00 0.00 8.04 1.45 86

86

58863153 PCN-74 1.48 0.00 0.00 11.41 1.94 86

86

58877886 PCN-63 1.42 0.00 0.00 10.54 1.82 86

86

67922241 PCN-40 1.24 0.00 0.00 8.86 1.58 86

86

67922263 PCN-50 1.36 0.00 0.00 9.20 1.70 86

86

68194081 PCB-150 1.74 0.00 0.11 8.92 2.06 89

89

68194161 PCB-173 1.87 0.00 0.09 10.09 2.18 89

89

69782907 PCB-157 1.72 0.00 0.09 10.27 2.06 89

89

70362504 PCB-81 1.44 0.00 0.11 9.08 1.81 89

89

74472370 PCB-114 1.59 0.00 0.11 9.39 1.94 89

89

74472494 PCB-186 1.87 0.00 0.09 9.73 2.18 89

89

103426977 PCN-67 1.42 0.00 0.00 10.04 1.82 86

86

149864813 PCN-31 1.24 0.00 0.00 9.10 1.58 86

86

121

150224207 PCN-57 1.36 0.00 0.00 9.41 1.70 86

86

150224252 PCN-59 1.36 0.00 0.00 9.59 1.70 86

86

239103572 1-(8-fluorooctyl)-2-nitroimidazole 1.62 0.00 0.76 8.11 1.86 137

a

a. The L solute descriptor was calculated with equation (29).

122

Table 9 - Hypothetical Chemicals Dataset

ID S A B L V

1 0.62 0.48 1.04 5.76 1.95 2 1.55 0.03 0.47 2.98 0.82

3 0.50 0.19 0.22 2.45 0.51

4 0.17 0.36 0.27 8.43 0.32 5 0.34 0.65 1.33 2.69 2.95

6 1.49 0.71 0.21 2.28 0.69

7 1.73 0.81 0.86 0.47 1.97 8 0.71 0.74 0.52 0.28 2.27

9 1.93 0.48 0.25 2.39 1.74

10 0.74 0.24 1.23 2.60 2.37 11 0.35 0.23 0.61 4.19 2.18

12 2.20 0.01 0.80 9.56 2.85

13 1.04 0.32 0.92 5.19 1.83 14 0.43 0.30 0.25 10.13 2.76

15 1.57 0.46 0.46 3.65 1.69

16 0.39 0.43 0.51 2.54 2.44 17 0.78 0.45 0.44 8.41 1.96

18 0.97 0.41 1.26 3.58 2.77

19 0.20 0.26 1.31 1.35 1.06 20 0.26 0.52 0.66 1.93 1.89

21 0.32 0.27 1.21 3.97 2.30

22 1.72 0.47 0.77 0.63 2.46 23 0.32 0.03 1.18 6.97 1.39

24 0.23 0.18 0.17 8.39 2.50

25 0.80 0.27 0.39 10.12 2.74 26 0.37 0.36 0.41 11.54 1.36

27 0.07 0.48 0.08 9.92 0.39

28 0.63 0.18 0.98 5.25 2.58 29 0.46 0.05 0.40 8.16 2.46

30 1.89 0.71 0.75 7.02 2.76

31 0.18 0.54 0.89 2.33 1.81 32 1.01 0.48 0.42 2.48 0.78

33 0.01 0.01 0.50 8.54 0.95

34 1.38 0.76 0.73 0.55 2.54 35 0.80 0.13 0.45 3.79 2.12

36 0.88 0.49 0.21 5.57 1.24

37 0.75 0.67 0.44 10.37 1.93 38 1.76 0.08 0.03 6.15 0.44

39 1.62 0.02 0.16 9.09 2.44

40 1.28 0.75 0.68 2.67 1.87 41 0.26 0.41 0.25 9.21 1.76

42 0.80 0.31 0.21 9.94 0.90

43 1.00 0.60 0.27 8.17 2.63 44 0.74 0.27 0.91 2.19 2.50

45 1.63 0.02 0.44 10.52 2.06

46 0.38 0.65 0.82 5.40 1.35 47 0.06 0.44 0.07 9.43 2.31

48 0.01 0.66 0.91 10.47 2.61 49 1.50 0.44 0.30 9.93 2.91

50 0.89 0.30 0.52 3.40 2.32

51 1.81 0.44 0.79 2.54 2.59 52 1.00 0.39 0.20 7.74 1.92

53 0.33 0.54 0.77 6.48 1.70

54 0.82 0.23 0.24 5.74 1.72

55 1.75 0.77 0.22 3.23 2.01

56 0.77 0.33 0.44 4.17 2.76

57 1.00 0.51 1.00 1.69 2.67 58 0.85 0.69 0.08 5.51 2.59

59 2.21 0.78 0.17 3.35 2.23

60 0.62 0.81 1.14 1.61 3.00 61 0.63 0.61 0.60 4.11 2.22

62 0.00 0.31 1.24 3.89 1.50

63 1.82 0.12 0.88 1.12 2.33 64 0.27 0.11 0.94 4.85 2.01

65 1.66 0.64 0.66 6.74 2.80

66 0.54 0.52 1.27 9.65 2.96

ID S A B L V

67 1.28 0.42 1.10 8.84 2.91

68 1.95 0.22 0.98 3.56 2.69

69 2.17 0.50 0.40 6.67 2.39 70 0.01 0.33 0.30 9.65 1.83

71 1.36 0.37 1.22 5.64 2.67

72 1.85 0.05 0.13 4.51 0.59 73 0.41 0.45 0.11 6.45 0.39

74 0.41 0.69 0.52 4.65 1.35

75 1.19 0.72 0.03 9.98 1.76 76 1.14 0.18 0.45 3.85 2.89

77 0.62 0.82 0.60 7.42 2.78

78 0.52 0.04 0.41 4.83 1.75 79 1.47 0.10 0.49 11.88 2.03

80 0.62 0.43 1.30 1.13 1.92

81 1.60 0.06 0.40 5.05 2.15 82 0.46 0.32 0.57 4.21 1.25

83 0.68 0.42 1.09 1.80 1.06

84 0.92 0.17 0.14 10.35 1.80 85 1.47 0.48 0.16 11.66 2.22

86 0.15 0.47 0.45 3.23 2.67

87 1.19 0.39 1.14 1.82 2.13 88 0.12 0.79 0.67 0.56 1.11

89 1.77 0.64 0.56 1.02 1.20

90 0.61 0.19 0.33 9.81 1.70 91 1.30 0.18 0.63 5.28 1.59

92 2.06 0.68 0.43 2.17 1.79

93 0.11 0.65 0.01 8.04 1.43 94 0.75 0.78 0.21 4.38 1.69

95 0.54 0.12 1.26 1.44 2.57

96 2.02 0.37 0.34 1.84 1.21 97 1.30 0.11 1.32 6.88 2.46

98 0.88 0.28 0.80 3.89 1.64

99 0.09 0.30 0.09 10.20 0.54 100 0.52 0.57 0.53 7.66 2.46

101 0.06 0.06 1.16 2.35 2.46

102 0.31 0.34 1.16 0.70 1.07

103 1.51 0.57 0.65 1.88 2.01

104 1.01 0.42 0.03 7.12 1.64

105 1.37 0.45 0.24 2.31 0.78 106 1.92 0.10 0.20 2.77 0.74

107 1.38 0.19 0.47 10.02 2.32

108 2.14 0.16 0.40 6.39 1.71 109 0.25 0.53 1.08 10.76 2.57

110 0.33 0.07 0.32 12.84 2.79

111 1.30 0.17 0.47 2.47 2.40 112 0.77 0.61 1.09 5.84 2.52

113 0.31 0.20 1.10 3.18 1.01

114 0.25 0.35 1.26 4.27 2.04 115 1.04 0.74 0.02 5.83 1.69

116 0.68 0.26 0.93 10.04 2.31 117 0.39 0.03 0.45 11.24 2.56

118 0.77 0.57 0.11 2.46 1.40

119 0.38 0.60 0.56 10.66 1.99 120 1.01 0.41 0.10 3.80 2.37

121 0.78 0.17 0.07 5.34 1.83

122 0.62 0.50 0.12 5.49 1.68 123 0.17 0.78 0.81 0.72 2.93

124 1.53 0.06 0.60 5.15 0.94

125 0.09 0.37 0.56 6.63 0.87 126 0.98 0.36 0.83 3.27 1.06

127 0.49 0.46 0.25 7.77 2.38

128 1.71 0.27 0.63 5.50 2.34 129 0.78 0.20 0.03 9.99 2.46

130 1.32 0.25 0.05 4.78 2.51

131 1.86 0.56 0.55 5.52 2.72 132 1.29 0.29 0.68 0.98 0.66

123

ID S A B L V

133 1.11 0.24 0.88 8.67 2.62

134 0.11 0.22 0.33 3.00 0.18 135 2.17 0.08 0.09 1.78 1.28

136 0.64 0.25 0.28 4.35 2.65

137 1.37 0.59 0.59 8.85 2.62 138 1.32 0.71 0.21 4.59 2.41

139 1.95 0.73 0.01 9.96 2.37

140 0.74 0.27 0.88 2.67 1.77 141 0.26 0.48 0.42 1.96 1.96

142 0.73 0.24 1.02 1.44 1.41

143 1.03 0.63 0.04 8.90 2.40 144 0.43 0.73 0.09 11.46 1.30

145 1.53 0.18 1.22 1.12 2.28

146 0.04 0.70 0.74 2.47 2.91 147 1.90 0.55 0.80 5.04 2.84

150 1.04 0.56 0.45 6.29 1.27

151 1.80 0.09 0.21 11.42 2.34 152 0.78 0.25 0.16 4.21 2.63

153 0.28 0.72 0.37 4.37 2.54

154 1.24 0.14 0.51 10.90 2.32 155 1.15 0.65 0.26 2.15 2.90

156 1.07 0.49 1.02 1.62 1.67

157 1.86 0.42 0.02 8.48 2.95 158 0.18 0.18 0.66 4.14 2.91

159 0.54 0.81 0.44 5.19 2.38 160 1.34 0.49 0.62 5.22 2.05

161 1.15 0.60 0.13 7.59 1.00

162 1.21 0.53 0.13 2.64 1.63 163 0.28 0.63 0.25 9.38 2.59

164 1.48 0.38 0.72 5.55 2.36

165 0.21 0.63 1.25 7.37 2.50 166 0.11 0.18 0.92 5.55 0.87

167 0.03 0.06 0.48 5.99 2.66

168 1.49 0.24 0.64 7.87 2.85 169 0.67 0.14 0.63 7.32 2.94

170 0.05 0.32 0.56 4.43 1.41

171 1.15 0.54 0.74 1.79 1.75 172 0.33 0.35 0.46 10.80 1.18

173 0.23 0.71 0.72 0.63 1.62

174 0.83 0.05 0.90 7.31 2.73 175 0.30 0.43 0.68 8.46 2.14

176 0.34 0.74 0.20 3.05 2.47

177 0.55 0.73 0.77 8.24 2.85 178 1.55 0.04 0.74 9.54 2.64

179 0.28 0.36 0.94 1.09 1.78

180 1.01 0.21 0.97 1.42 2.42 181 1.36 0.36 0.15 1.29 0.52

182 0.06 0.76 0.58 4.76 2.52

183 1.12 0.03 0.37 2.86 0.24 184 0.20 0.34 0.80 6.06 1.51

185 0.48 0.07 0.84 4.83 0.72

186 0.58 0.13 0.05 12.51 1.77 187 1.14 0.43 0.41 7.07 2.78

188 0.17 0.68 0.96 9.20 2.18

189 0.71 0.77 0.20 0.63 2.22 190 0.99 0.57 1.03 4.92 2.67

191 0.37 0.04 0.19 6.34 0.25

192 1.45 0.14 0.72 1.40 0.94 193 0.16 0.67 1.05 5.85 1.89

194 2.08 0.28 0.21 1.64 0.57

195 0.49 0.34 0.87 4.54 2.54

ID S A B L V

196 0.29 0.27 0.93 6.48 2.30

197 0.78 0.14 0.23 5.79 0.86 198 0.09 0.10 0.30 6.73 1.49

199 0.85 0.45 0.88 2.69 1.53

200 1.18 0.40 0.72 10.47 2.60 201 0.94 0.27 0.13 7.39 1.15

202 0.27 0.56 0.35 0.99 1.37

203 2.18 0.54 0.44 7.25 2.53 204 0.56 0.44 0.09 9.52 1.59

205 1.89 0.40 0.80 10.08 2.96

206 2.24 0.64 0.19 8.67 2.20 207 2.21 0.14 0.64 0.74 1.92

208 1.12 0.37 0.07 2.97 1.74

209 1.79 0.44 0.21 10.95 2.06 210 2.17 0.23 0.35 0.88 0.72

211 1.30 0.39 0.89 2.60 2.60

212 1.69 0.25 0.12 6.39 1.36 213 0.16 0.61 1.18 9.72 2.78

214 0.43 0.15 0.86 3.28 2.21

215 0.03 0.27 0.57 10.37 2.32 216 1.97 0.11 0.33 3.74 2.08

217 0.86 0.68 0.50 8.22 2.08

218 0.00 0.62 0.43 5.63 1.41 219 0.82 0.43 0.56 5.02 2.52

220 0.26 0.68 0.32 2.57 2.02 221 0.88 0.46 1.01 3.83 1.96

222 0.81 0.69 0.87 1.83 2.86

223 0.32 0.01 0.66 10.34 2.69 224 0.25 0.55 1.26 5.30 2.79

225 0.19 0.50 0.91 6.25 2.11

226 0.32 0.27 0.76 7.65 2.07 227 1.52 0.71 0.64 2.27 2.96

228 0.10 0.57 0.44 3.04 0.64

229 0.16 0.37 0.96 8.77 1.59 230 1.00 0.07 0.88 11.34 2.78

231 1.70 0.09 1.20 5.67 2.29

232 2.16 0.14 0.56 2.69 1.53 233 1.88 0.03 1.16 6.12 2.95

234 2.14 0.73 0.47 4.49 2.89

235 0.18 0.23 0.34 11.59 2.47 236 1.42 0.33 1.16 4.85 2.91

237 0.43 0.50 1.33 2.27 2.42

238 1.51 0.57 0.57 0.32 0.96 239 0.74 0.71 0.78 4.86 1.55

240 1.61 0.39 0.01 5.73 2.56

241 0.29 0.24 1.04 1.79 2.99 242 1.83 0.20 0.12 8.27 2.52

243 2.00 0.41 0.02 8.35 2.88

244 0.65 0.27 0.63 2.32 2.93 245 0.70 0.77 1.13 6.33 2.60

246 1.34 0.19 0.04 7.10 2.53

124

Table 10 - Solute Descriptor Inter-Correlations (R2): Real Chemicals

R2 S A B L V

S - 0.08 0.05 0.48 0.19

A 0.08 - 0.00 0.12 0.07

B 0.05 0.00 - 0.04 0.00

L 0.48 0.12 0.04 - 0.82

V 0.19 0.07 0.00 0.82 -

Table 11 - Solute Descriptor Inter-Correlations (R2): Hypothetical Chemicals

R2 S A B L V

S - 0.01 0.04 0.01 0.01

A 0.01 - 0.00 0.03 0.02

B 0.04 0.00 - 0.06 0.07

L 0.01 0.03 0.06 - 0.03

V 0.01 0.02 0.07 0.03 -

5.6 Description of Model and Parameters

Model details and compartments are as described in Wania et al.128

, except where alterations

have been made for the PP-LFER model as noted in the Section 5.2. An additional alteration

is the addition of a dynamic time step to ensure model stability. The nomenclature in

Appendix A of Wania et al. has been used to summarize the model parameters listed in Table

12.

The finite difference approximation remains unaltered in the model but the length of the time

step is no longer fixed and may differ for each time step. At each time step the maximum

125

relative change in fugacity of any compartment is calculated; if the maximum change is

greater than 50% the solution does not proceed, instead the length of the time step is

decreased to one third of the current value and the calculations for the current time step are

repeated. This procedure is repeated until the maximum relative change in the fugacity of

any compartment is 50% or less, and then the program proceeds to the next time step. The

time step is not permitted to fall below a value of 0.001 hours (3.6 seconds). If the maximum

relative change in the fugacity in any compartment is less than 10% the length of the time

step is increased to four thirds of the current value and the solution proceeds to the next time

step. The time step is not permitted to rise above 24 hours. This dynamic time step ensures

stability in all compartments for chemicals with extreme properties and speeds up the

calculation time for many other chemicals.

An emission rate of 1 tonne per year is used for all three emission scenarios, and a generic

molecular weight of 200 g/mol is used for all real and hypothetical chemicals to ensure that

the same number of moles of chemical is emitted for all chemicals and scenarios. Chemical

degradation half-lives and activation energies are set to arbitrarily high values to prevent

degradation from occurring. The density of atmospheric particles is calculated from the

densities of organic matter and mineral matter and the fraction of organic matter in particles,

assuming that the remainder of the particles is composed of mineral matter. The same

density is used in the PP-LFER model to ensure that the amount of particles in the air is the

same in both models.

126

Table 12 - Summary of Model Parameters

Chemical and Emission

MW 200 Molecular weight (g/mol)

EX 1 Emission rate to the compartment receiving emissions (tonnes/year)

RfA 1 Ratio relating fugacity in inflowing air to fugacity of the outflowing air

RfC 1 Ratio relating fugacity in inflowing water to fugacity of the outflowing water

Miscellaneous Model Parameters

DNOM 1000000 Density of organic matter (g/m3)

DNMM 2400000 Density of mineral matter (g/m3)

Air

HA 6000 Height of atmospheric compartment (m)

AS 1.8E+12 Surface area of drainage basin (m2)

VFSA 4 Mass of aerosol in air (μg/m3)

VFSAIN 4 Mass of aerosol in inflowing air (μg/m3)

VFOQ 0.1 Volume fraction of organic carbon on aerosol (-)

tA 42 Atmospheric residence time (in hours)

Q 6.8E+04 Particle scavenging ratio (-)

facStability 3 Stability of the winter atmosphere relative to summer conditions (-)

U3S 57 Rain rate of drainage basin and estuary (cm/year)

RHA 80 Relative Humidity of air (%)

TA1 268.8 Air temperature in January (Kelvin)

TA2 266.8 Air temperature in February (Kelvin)

TA3 271.1 Air temperature in March (Kelvin)

TA4 278 Air temperature in April (Kelvin)

TA5 282.8 Air temperature in May (Kelvin)

TA6 286.4 Air temperature in June (Kelvin)

TA7 288.2 Air temperature in July (Kelvin)

TA8 290.5 Air temperature in August (Kelvin)

TA9 282.8 Air temperature in September (Kelvin)

TA10 280.2 Air temperature in October (Kelvin)

TA11 277.7 Air temperature in November (Kelvin)

TA12 269.9 Air temperature in December (Kelvin)

WSS1 5.1 Wind speed in January, terrestrial environment (m/sec)

WSS2 5.8 Wind speed in February, terrestrial environment (m/sec)

WSS3 4.9 Wind speed in March, terrestrial environment (m/sec)

WSS4 4.6 Wind speed in April, terrestrial environment (m/sec)

WSS5 5 Wind speed in May, terrestrial environment (m/sec)

WSS6 4.8 Wind speed in June, terrestrial environment (m/sec)

WSS7 4.9 Wind speed in July, terrestrial environment (m/sec)

WSS8 4.6 Wind speed in August, terrestrial environment (m/sec)

WSS9 5.1 Wind speed in September, terrestrial environment (m/sec)

WSS10 5.6 Wind speed in October, terrestrial environment (m/sec)

WSS11 6.4 Wind speed in November, terrestrial environment (m/sec)

WSS12 5.7 Wind speed in December, terrestrial environment (m/sec)

Fresh Water

HW 5 Average (fresh) water depth (m)

frAW 0.06 Fraction of drainage basin covered by rivers and lakes (-)

frUW 0.14 Evaporation from fresh water as fraction of input (-)

vWD-P 1.03 Dry particle deposition to fresh water (m/h)

BPW 70 Primary productivity in fresh water ( g C/m2 yr )

facOWmiw 0.9 POC mineralisation fraction in fresh water (-)

MW 0.35 Fresh water partitioning into POC: Koc = M · Kow (Factor M in L/Kg)

TS1 268 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in January (Kelvin)

TS2 266.8 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in February (Kelvin)

TS3 270.9 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in March (Kelvin)

TS4 277.9 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in April (Kelvin)

TS5 282.4 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in May (Kelvin)

TS6 287.4 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in June (Kelvin)

TS7 288 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in July (Kelvin)

TS8 289.6 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in August (Kelvin)

TS9 282.1 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in September (Kelvin)

TS10 279.8 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in October (Kelvin)

TS11 275.6 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in November (Kelvin)

TS12 269.3 Terrestrial (all canopies, soils and fresh water/sed compartments) temperature in December (Kelvin)

Fresh Water Eroding Sediments (WS1)

127

HWS1 0.05 Surficial fresh water sediment depth, compartment 1 (m)

VFSWS1 0.3 Volume fraction of solids in fresh water sediment, compartment 1 (-)

frAWS1 0.6 Sediment area, compartment 1, as fraction of fresh water area (-)

VsedWS1 0.05 POC sedimentation velocity, fresh water sediment 1 (m/h)

VburWS1 5E-09 POC burial velocity, fresh water sediment 1 (m/h)

VlatWS12 1E-11 Lateral POC transport velocity within fresh water basin (m/h)

VlatWS1TS1 1E-11 Lateral POC transport between fresh water and estuary basins (m/h)

facOWS1res 0.95 POC resuspension intensity (fraction of deposition) (-)

facOWS1mis 0.9 POC mineralisation in sediment (fraction of net deposition) (-)

BWbio1poc 1E-07 Bioturbation diffusivity for sediment solids (m2/h)

U8bblW1 0.05 Mass transfer coefficient for benthic boundary layer (m/h)

Fresh Water Accumulating Sediments (WS2)

HWS2 0.05 Surficial fresh water sediment depth, compartment 2 (m)

VFSWS2 0.3 Volume fraction of solids in fresh water sediment, compartment 2 (-)

VsedWS2 0.05 POC sedimentation velocity, fresh water sediment 2 (m/h)

VburWS2 5E-08 POC burial velocity, fresh water sediment 2 (m/h)

VlatWS21 1E-11 Lateral POC transport velocity within fresh water basin (m/h)

facOWS2res 0.5 POC resuspension intensity (fraction of deposition) (-)

facOWS2mis 0.9 POC mineralisation in sediment (fraction of net deposition) (-)


U8bblW2 0.01 Mass transfer coefficient for benthic boundary layer (m/h)

Estuary Water

HT 52 Average water depth of estuary (m)

AT 4.77E+11 Surface area of estuary (m2)

frUT 0.19 Evaporation from estuary as fraction of input (-)

facFTC 25 Factor by which the net flux between the estuary and coastal water compartments is increased by mixing

vTD-P 1.03 Dry particle deposition to estuary water (m/h)

BPT 134 Primary productivity in estuarine water ( g C/m2 yr )

facOTmiw 0.9 POC mineralisation in estuary water (% of PP)

MT 0.35 Estuarine water partitioning into POC: Koc = M · Kow (Factor M in L/Kg)

CpocOut 0.32 POC Concentration of outside marine water (mg/L)

TT1 269.9 Estuary temperature in January (Kelvin)

TT2 267.6 Estuary temperature in February (Kelvin)

TT3 270.9 Estuary temperature in March (Kelvin)

TT4 276.5 Estuary temperature in April (Kelvin)

TT5 280 Estuary temperature in May (Kelvin)

TT6 284.7 Estuary temperature in June (Kelvin)

TT7 286.7 Estuary temperature in July (Kelvin)

TT8 289.6 Estuary temperature in August (Kelvin)

TT9 284.2 Estuary temperature in September (Kelvin)

TT10 281.6 Estuary temperature in October (Kelvin)

TT11 277.8 Estuary temperature in November (Kelvin)

TT12 272.7 Estuary temperature in December (Kelvin)

WST1 5.8 Wind speed in January, Estuary (m/sec)

WST2 6.8 Wind speed in February, Estuary (m/sec)

WST3 5.5 Wind speed in March, Estuary (m/sec)

WST4 5.3 Wind speed in April, Estuary (m/sec)

WST5 5.9 Wind speed in May, Estuary (m/sec)

WST6 5.3 Wind speed in June, Estuary (m/sec)

WST7 5.8 Wind speed in July, Estuary (m/sec)

WST8 5.3 Wind speed in August, Estuary (m/sec)

WST9 6 Wind speed in September, Estuary (m/sec)

WST10 6.7 Wind speed in October, Estuary (m/sec)

WST11 7.7 Wind speed in November, Estuary (m/sec)

WST12 6.8 Wind speed in December, Estuary (m/sec)

Estuary Eroding Sediments (TS1)

HTS1 0.01 Surficial estuary sediment depth, compartment 1 (m)

VFSTS1 0.2 Volume fraction of solids in estuary sediment, compartment 1 (-)

frATS1 0.7 Sediment area, compartment 1, as fraction of estuary area (-)

VsedTS1 1.75E-02 POC sedimentation velocity, estuary sediment 1 (m/h)

VburTS1 1E-10 POC burial velocity, estuary sediment 1 (m/h)

VlatTS12 1E-07 Lateral POC transport velocity within estuary basin (m/h)

VlatTS1CS1 1E-11 Lateral POC transport between estuary and coastal water basins (m/h)

facOTS1res 0.75 POC resuspension intensity (fraction of deposition) (-)

facOTS1mis 0.05 POC mineralisation in sediment (fraction of net deposition) (-)

BTbio1poc 1E-07 Bioturbation diffusivity for sediment solids (m2/h)

128

U8bblT1 0.05 Mass transfer coefficient for benthic boundary layer (m/h)

Estuary Accumulating Sediments (TS2)

HTS2 0.1 Surficial estuary sediment depth, compartment 2 (m)

VFSTS2 0.2 Volume fraction of solids in estuary sediment, compartment 2 (-)

VsedTS2 1.75E-02 POC sedimentation velocity, estuary sediment 2 (m/h)

VburTS2 3E-08 POC burial velocity, estuary sediment 2 (m/h)

VlatTS21 1E-12 Lateral POC transport velocity within estuary basin (m/h)

facOTS2res 0.1 POC resuspension intensity (fraction of deposition) (-)

facOTS2mis 0.85 POC mineralisation in sediment (fraction of net deposition) (-)


U8bblT2 0.01 Mass transfer coefficient for benthic boundary layer (m/h)

Agricultural Soil (E)

HE 0.18 Soil depth, agricultural soil (m)

VFAE 0.25 Volume fraction of air in agricultural soil (-)

VFWE 0.25 Volume fraction of water in agricultural soil (-)

VFSE 5E-05 Volume fraction of solids in run-off water from agricultural soil (-)

frUE 0.43 Evaporation from other soil compartments (agr.+unc.) as fraction of input (-)

OCE 0.02 Mass fraction of organic carbon solids in agricultural soil (-)

U7Emax 2.08 Mass transfer coefficient through air boundary over agricultural soil layer (m/h)

BsolidE 2.28E-08 Solid phase diffusivity in agricultural soil (m2/h)

vED-P 1.03 Maximum dry deposition velocity to agricultural soil (m/h)

ME 0.35 Partitioning relationship for POC in agricultural soil (Koc = M·Kow), Factor M (L/Kg)

Forest soil (B)

HB 0.1 Soil depth, forest soil (m)

frAB 0.479 Fraction of drainage basin covered by forest (-)

VFAB 0.25 Volume fraction of air in forest soil (-)

VFWB 0.25 Volume fraction of water in forest soil (-)

VFSB 1E-05 Volume fraction of solids in run-off water from forest soil (-)

frUB 0.26 Evaporation from forest soil as fraction of input (-)

OCB 0.02 Mass fraction of organic carbon solids in forest soil (-)

U7Bmax 0.416 Mass transfer coefficient through air boundary over forest soil layer (m/h)

BsolidB 2.28E-08 Solid phase diffusivity in forest soil (m2/h)

vBD-P 0.206 Maximum dry deposition velocity to forest soil (m/h)

MB 0.35 Partitioning relationship for POC in agricultural soil (Koc = M·Kow), Factor M (L/Kg)

Uncultivated Soil (U)

HU 0.1 Soil depth, uncultivated soil (m)

frAU 0.06 Fraction of drainage basin covered by uncultivated soil (-)

VFAU 0.25 Volume fraction of air in uncultivated soil (-)

VFWU 0.25 Volume fraction of water in uncultivated soil (-)

VFSU 3E-05 Volume fraction of solids in run-off water from uncultivated soil (-)

OCU 0.02 Mass fraction of organic carbon solids in uncultivated soil (-)

U7Umax 2.08 Mass transfer coefficient through air boundary over uncultivated soil layer (m/h)

BsolidU 2.28E-08 Solid phase diffusivity in uncultivated soil (m2/h)

vUD-P 1.03 Maximum dry deposition velocity to uncultivated soil (m/h)

MU 0.35 Partitioning relationship for POC in agricultural soil (Koc = M·Kow), Factor M (L/Kg)

Coniferous Canopy

frACon 0.73 Fraction of forest area covered by coniferous trees (-)

VFCon 1.5E-03 Coniferous canopy volume per ground area (m3/m2)

Mcon 38 Paritioning relationship for coniferous canopies, factor M (Kcan/air = M·KowN)

Ncon 0.69 Paritioning relationship for coniferous canopies, exponent N (Kcan/air = M·KowN)

vFCD-G 42.1 Dry gaseous deposition velocity to coniferous canopy, summer average (m/h)

vFCD-P 3.4 Dry particle deposition velocity to coniferous canopy, summer average (m/h)

tFC 5 Average residence time of needles (years)

frUFC 0.4 Evaporation from coniferous canopy as fraction of input (-)

Deciduous Canopy

VFDec 1.1E-03 Deciduous canopy volume per ground area (m3/m2)

Mdec 14 Paritioning relationship for deciduous canopies, factor M (Kcan/air = M·KowN)

Ndec 0.76 Paritioning relationship for deciduous canopies, exponent N (Kcan/air = M·KowN)

vFDD-G 129.6 Dry gaseous deposition velocity to deciduous canopy, summer average (m/h)

vFDD-P 27 Dry particle deposition velocity to deciduous canopy, summer average (m/h)

facVLeaf 0.1 Fraction of leaves which stays on trees during winter

tStartGrow 113 Start of deciduous canopy development (day)

tStopGrow 143 End of deciduous canopy development (day)

tStartFall 302 Start of falling leaves (day)

tStopFall 332 End of falling leaves (day)

frUFD 0.4 Evaporation from deciduous canopy as fraction of input (-)

129

5.7 Interpolation Method

In the log KAW / log KOA chemical space moving in different directions has different

chemical meanings. Moving along the x axis corresponds to changes in octanol/air

partitioning, moving along the y axis corresponds to changes in air/water partitioning,

moving in the diagonal up to the right corresponds to changes in octanol/water partitioning

and moving in the diagonal down to the right corresponds to changes in both octanol/air and

air/water partitioning while octanol/water partitioning remains fixed. A simple distance

weighted interpolation method might possibly interpolate points from data in only one

direction effectively ignoring the other types of partitioning.

The goal of the interpolation method outlined here is to ensure that points are interpolated

from data in at least three of the directions outlined above, guaranteeing that data is included

from at least two directions corresponding to the three partitioning coefficients. The

complete procedure is as follows:

• All points with input data in four directions are interpolated.

• All points with input data in three directions are interpolated.

• All interpolated points are added to the input data.

• All points with input or interpolated data in four directions are

interpolated; these newly interpolated points are also added to the

list of input data as they are interpolated.

• All points with input or interpolated data in three directions are

interpolated; these newly interpolated points are also added to the

list of input data as they are interpolated.

The last two steps are referred to as extrapolation because they are partly, or possibly even

entirely, calculated based on previously interpolated points. Given enough data points to start

from the procedure will always run to completion and fill in the entire chemical space,

extrapolating where necessary. Interpolation for each point is performed as follows:

• The angle of all points within 2.83 log units is calculated.

130

• Points are assigned to the four directions based on their angles; any

points on the dividing line between directions are counted for both

directions.

• Any direction with three or more points available for interpolation is

flagged as having sufficient data for interpolation.

• If there is data in a sufficient number of directions the mean for each

direction is calculated, and then the mean of the directional values is

calculated and used as the interpolated value.

A maximum distance of 2.83 was selected to ensure sufficient points are available for

interpolation in all directions when interpolating with a resolution of 1 log unit or less. Using

the mean value instead of a distance weighted method has a smoothing effect on the results,

making the surfaces less prone to spikes caused by outlying values. Python 2.5.2 code for the

entire interpolation and extrapolation routine is provided below.

131

#chemical space interpolation routine #written by Trevor Brown #in python 2.5.2 from math import acos from math import asin def get_angle(x1, y1, x2, y2): #return the angle that point x2,y2 is found from point x1,y1 in degrees distance = pow(pow(x2 - x1,2) + pow(y2 - y1,2),0.5) if distance == 0: return 0 else: if y2 >= y1 and x2 > x1: return_value = asin((x2 - x1) / distance) * 180 / 3.14159265358979 elif y2 < y1 and x2 >= x1: return_value = 90 + acos((x2 - x1) / distance) * 180 / 3.14159265358979 elif y2 <= y1 and x2 < x1: return_value = 180 + asin((x1 - x2) / distance) * 180 / 3.14159265358979 elif y2 > y1 and x2 <= x1: return_value = 270 + acos((x1 - x2) / distance) * 180 / 3.14159265358979 if return_value >= 360: return_value = return_value - 360 if return_value < 0: return_value = return_value + 360 return return_value def get_mean(data_points): #calculate the mean value of data_points if len(data_points) > 0: return_value = 0 for index in range(len(data_points)): return_value = return_value + data_points[index][1] return_value = return_value / (index+1) return return_value else: return 0 #read in chemical IDs, Koa, Kaw and value to be interpolated #from space delimited input file input_file = open("pyinput.txt","r") store_var = [] count = 0 for index in input_file: #id position = index.find(" ") store_var.append({"id":index[:position]}) index = index[position+1:] #x position = index.find(" ") store_var[count]["koa"] = eval(index[:position]) index = index[position+1:] #y position = index.find(" ") store_var[count]["kaw"] = eval(index[:position]) index = index[position+1:] #z position = index.find(" ") store_var[count]["z"] = eval(index[:position]) index = index[position+1:] count = count + 1 input_file.close() #define chemical space and resolution to parse over #in this case 3 <= Koa <= 13, -6 <= Kaw <= 4 #and a resolution of 0.5 log units koa_vals = [] for index in range(21): koa_vals.append(3 + float(index) * 0.5) kaw_vals = [] for index in range(21): kaw_vals.append(-6 + float(index) * 0.5) #create variables to track output

132

points = len(koa_vals) * len(kaw_vals) point_list = [] out_put = dict() #n_req is the required number of directions that interpolation #points are available from n_req = 4 #determines if the interpolation/extrapolation loop is in #interpolation or extrapolation mode extrapolate = 0 #start interpolation/extrapolation loop while points > 0: #update keeps n_req from decreasing until no more points #can be added when in extrapolation mode update = 0 for koa in koa_vals: for kaw in kaw_vals: #if this point has not already been interpolated calculate #the number of points in range and their directions if not out_put.has_key((koa,kaw)): arc_n_s = [] arc_ne_sw = [] arc_e_w = [] arc_se_nw = [] for index in store_var: angle = get_angle(koa, kaw, index["koa"], index["kaw"]) distance = pow(pow(index["koa"] - koa,2) + pow(index["kaw"] - kaw,2),0.5) if distance <= 2.83: if angle > 337 or angle < 23: arc_n_s.append([distance,index["z"]]) if angle > 22 and angle < 68: arc_ne_sw.append([distance,index["z"]]) if angle > 67 and angle < 113: arc_e_w.append([distance,index["z"]]) if angle > 112 and angle < 158: arc_se_nw.append([distance,index["z"]]) if angle > 157 and angle < 203: arc_n_s.append([-distance,index["z"]]) if angle > 202 and angle < 248: arc_ne_sw.append([-distance,index["z"]]) if angle > 247 and angle < 293: arc_e_w.append([-distance,index["z"]]) if angle > 292 and angle < 338: arc_se_nw.append([-distance,index["z"]]) #only directions with 3 or more points contribute to interpolation n = 0 if len(arc_n_s) >= 3: n = n + 1 if len(arc_ne_sw) >= 3: n = n + 1 if len(arc_e_w) >= 3: n = n + 1 if len(arc_se_nw) >= 3: n = n + 1 #if there is data in the required number of directions interpolate this point if n >= n_req: sum_weight = 0 if len(arc_n_s) >= 3: mean = get_mean(arc_n_s) sum_weight = sum_weight + mean if len(arc_ne_sw) >= 3: mean = get_mean(arc_ne_sw) sum_weight = sum_weight + mean if len(arc_e_w) >= 3: mean = get_mean(arc_e_w) sum_weight = sum_weight + mean if len(arc_se_nw) >= 3: mean = get_mean(arc_se_nw) sum_weight = sum_weight + mean z = sum_weight / n #add this point to the output out_put[(koa,kaw)] = z points = points - 1 #if in extrapolation mode add this point to the extrapolation data

133

#and set the flag "update" not to decrease n_req if extrapolate == 1: store_var.append({"koa":koa,"kaw":kaw,"z":z}) update = 1 #decrease n_req unless "update" flag is set and in extrapolation mode if not(extrapolate == 1 and update == 1): n_req = n_req – 1 #once all possible points have been interpolated start extrapolation if n_req == 2: #append all points interpolated to the list of data for extrapolation for key,value in out_put.items(): store_var.append({"koa":key[0],"kaw":key[1],"z":value}) n_req = 4 extrapolate = 1 #create list of points and interpolated values to output print_list = [] for key,value in out_put.items(): print_list.append([key[1],key[0],value]) print_list.sort() #output list to space delimited file in a matrix with row and column headers output_file = open("pyoutput.txt","w") this_row = print_list[0][0] out_string = "0 " for index in print_list: if index[0] == this_row: out_string = out_string + str(index[1]) + " " else: break out_string = out_string + "\n" output_file.write(out_string) out_string = str(print_list[0][0]) + " " for index in print_list: if index[0] == this_row: out_string = out_string + str(index[2]) + " " else: this_row = index[0] out_string = out_string + "\n" out_string = out_string + str(index[0]) + " " out_string = out_string + str(index[2]) + " " output_file.write(out_string) output_file.close() #note: to get Excel to properly plot the matrix as a surface you must delete #the zero at position 1,1 of the matrix after pasting into Excel

134

5.8 Validity of Showing Results in the Chemical Space

We expect to find a strong link between environmental fate and the solute descriptors of real

chemicals. This expectation comes intuitively from the following observation; if we plot a

series of chemicals, for example n-alcohols, in the log KAW / log KOA chemical space they are

all located close to each other in the plot. If we now overlay the plot with the chemical fate

predicted by an environmental model we find that the n-alcohols are primarily predicted to

be in the aqueous phase. Therefore we can safely assume that chemicals with solute

descriptors similar to n-alcohols will also be found in the aqueous phase of an environmental

model. However, it is in theory possible that chemicals with combinations of solute

descriptors completely different from n-alcohols will also be located in the same region of

the chemical space. If this is true then plotting the results of the PP-LFER model in the log

KAW / log KOA chemical space would not be sensible because the environmental fate, as

plotted in the chemical space, would be largely unrelated to specific combinations of solute

descriptors.

To investigate the relationship between location in the log KAW / log KOA chemical space and

solute descriptors, chemical space plots of each solute descriptor were constructed for both

real and hypothetical chemicals using the interpolation method outlined in Section 5.7. These

plots are shown in Figure 15, the real chemicals are labelled (R) and the hypothetical

chemicals are labelled (H). In addition to these plots are plots of the standard deviation of

each interpolated point for real and hypothetical chemicals; that is the variation in the data

used to interpolate each point, these plots are labelled (RD) and (HD).

For real chemicals all five solute descriptors show clear trends in the log KAW / log KOA

chemical space and for the descriptors S, B, L and V the standard deviation is relatively

small. These plots indicate that, for the four solute descriptors mentioned, areas in the

chemical space correspond to a very limited number of combinations of solute descriptors.

The A solute descriptor shows a clear trend in the chemical space, but the standard deviation

is relatively large and follows the same trend. We interpret this to mean that in the areas

where A = 0 all chemicals in the area generally also have A = 0. However, the areas in the

135

chemical space which contain values of A greater than zero have much more variability in

the possible values of A. For hypothetical chemicals the trends shown in the log KAW / log

KOA chemical space are less defined, and the standard deviations are larger relative to the

interpolated values. Only the L descriptor shows a strong trend and a strong similarity to the

corresponding plot of real chemicals. We interpret this to mean the relationship between

areas in the chemical space and solute descriptor combinations is in general much weaker for

the hypothetical chemicals; there are more possible combinations of solute descriptors that

may give each pair of KAW/KOA values.

Another method of investigating the relationship between the log KAW / log KOA chemical

space and solute descriptors is to consider the interpolated points as a second data set, and

compare the statistics of this data set with the original. The mean values and standard

deviations of both datasets have been calculated. The statistics for the interpolated points

were then divided by the statistics for the original datasets. This procedure was also

performed for a set of randomly generated numbers to check the effect of the distribution of

real and hypothetical chemicals in the chemical space; all of the calculated ratios are shown

in Table 13.

If a solute descriptor has no relation to chemical location in the chemical space each

interpolated point will have a random sampling of the possible phase descriptor values and

the interpolated value will be near the mean of the original dataset. The result will be a much

smaller standard deviation for the interpolated dataset, as much of the variation is ―averaged

out‖ by the interpolation method. However, if the relationship between a solute descriptor

and location in the chemicals space is quite strong then like values will be localized in the

same regions of the chemical space and much of the variation in the original dataset will be

preserved in the interpolation, meaning the standard deviation of the interpolated dataset will

be comparable to the standard deviation of the original dataset. The ratio of the means will

show if the interpolated dataset is skewed towards higher of lower values.

136

All solute descriptors for both the real and the hypothetical datasets show evidence of

localization, except for the A descriptor in the hypothetical dataset which shows no more

localization than the random data. Interpolated values for hypothetical chemicals are not

skewed to any great extent and, with the exception of the L descriptor, have weaker

localization than the dataset of real chemicals. For real chemicals the S, B, L and V

descriptors have strong localization, with interpolated dataset standard deviations

approximately 70-80% of the magnitude of the original dataset. The A descriptor for real

chemicals shows a relatively weak localization, but along with the B descriptor the mean

ratios show the interpolated dataset is skewed towards lower values. The statistics for A,

along with the plots in Figure 15 suggest the following explanation; chemicals with A equal

to zero are distributed randomly throughout the chemical space, but chemicals with A greater

than zero are localized. Interpolated data for A is skewed towards lower values because

chemicals with A equal to zero are collocated with chemicals that have high A values,

causing the higher values to be partially averaged out.

From the above discussion we conclude that the use of the log KAW / log KOA chemical space

to display the results of a PP-LFER model run using real chemicals is valid. The strong

localization of solute descriptors in the chemical space indicates that areas within the

chemical space correspond not just to specific combinations of log KAW and log KOA, but also

to a small number of similar combinations of the solute descriptors S, A, B, L and V.

Plotting hypothetical chemicals in the log KAW / log KOA chemical space is less valid because

the localization is weaker for all descriptors, and non-existent for the A descriptor. The plots

are useful for comparison to the plots created with real chemicals, but should not be used for

quantitative purposes such as the chemical screening exercise of Brown and Wania146

.

137

Figure 15 - Solute Descriptor Plots

Plots of the surface interpolated from the solute descriptors (S, A, B, L, V) of real

chemicals (R) and hypothetical chemicals (H) along with plots of the surface

interpolated from the RMSE of each point from the data used in the interpolation (RD

and HD).

138

Table 13 - Statistical Ratios of Interpolated Solute Descriptors

Real Chemicals

Mean Ratio

Real Chemicals

Standard Deviation

Ratio

Hypothetical

Chemicals Mean

Ratio

Hypothetical

Chemicals Standard

Deviation Ratio

S 0.949 0.792 0.916 0.287

A 0.705 0.307 0.971 0.152

B 0.727 0.663 0.880 0.378

L 1.079 0.813 1.024 0.806

V 1.103 0.737 1.077 0.484

Random 0.972 0.135 0.992 0.130

Figure 16 - KOC and KPA Correlations of Hypothetical Chemicals

(A) Correlation between the log KOC values obtained by the SP-LFER equation and

values obtained from the PP-LFER obtained for humic acid for hypothetical chemicals.

(B) Correlation between the log KPA values obtained by the SP-LFER and values

obtained from the PP-LFER obtained for whole aerosol for hypothetical chemicals.

139

Figure 17 - Götz et al. PP-LFER KPA versus SP-LFER KPA

Correlation between the log KPA values obtained by the SP-LFER and values obtained

from the PP-LFER of Götz et al. for urban aerosol117

.

Figure 18 - Aspvreten PP-LFER KPA versus SP-LFER KPA

Correlation between the log KPA values obtained by the SP-LFER and values obtained

from the PP-LFER of Arp et al. for rural aerosol from Aspvreten, Sweden121

.

y = 0.970x + 0.970

R² = 0.9762

4

6

8

10

12

14

2 4 6 8 10 12 14

log

KP

A(G

otz

)

log KPA (spLFER)

140

Figure 19 - Chemical Space Plots for Hypothetical Chemicals

Chemical space plots of phase distribution of the PP-LFER based model for

hypothetical chemicals and the sum of absolute differences in the percent phase

distribution between the SP-LFER and PP-LFER based models for each the three

emission scenarios.

141

Figure 20 - Relative Difference in Model Outputs for Air Emissions

Chemical space plots of phase distribution of the PP-LFER based model for real

chemicals, where the coloured areas correspond to more than 50% of chemicals being

partitioned to the phase noted, and the relative absolute differences in model outputs

for each compartment between the SP-LFER and PP-LFER based models for emissions

to air.

142

Figure 21 - Relative Difference in Model Outputs for Soil Emissions





to soil.

143

Figure 22 - Relative Difference in Model Outputs for Water Emissions





to water.

144

Figure 23 - Chemical Space Plots for Degrading Real Chemicals

Chemical space plots of phase distribution of the PP-LFER based model for degrading

real chemicals and the sum of absolute differences in the percent phase distribution

between the SP-LFER and PP-LFER based models for emissions to air.

145

Figure 24 - Characteristic Travel Distance of Degrading Real Chemicals

Chemical space plots of the characteristic travel distance in kilometres for the SP-

LFER and PP-LFER models.

146

Table 14 - Sub-sampling Statistics

n = 246 S A B L V

Hypothetical Chemicals Mean 0.916 0.394 0.572 5.500 1.999 All Data points St.Dev 0.630 0.226 0.368 3.210 0.710

n = 235 S A B L V

Real Chemicals Mean 0.985 0.136 0.392 6.068 1.367 All Data points St.Dev 0.453 0.214 0.293 2.570 0.525

n = 24 S A B L V

Hypothetical Chemicals Mean 0.615 0.387 0.517 2.209 2.193 Primarily in Air St.Dev 0.480 0.244 0.349 1.090 0.605

n = 30 S A B L V

Real Chemicals Mean 0.518 0.038 0.212 3.764 1.018 Primarily in Air Stdev 0.254 0.058 0.166 0.692 0.237

n = 78 S A B L V

Hypothetical Chemicals Mean 1.041 0.376 0.763 3.063 1.512 Primarily in Water St.Dev 0.645 0.212 0.357 1.854 0.710

n = 87 S A B L V

Real Chemicals Mean 0.888 0.246 0.558 4.183 0.954 Primarily in Water St.Dev 0.323 0.247 0.230 1.470 0.304

n = 88 S A B L V

Hypothetical Chemicals Mean 0.835 0.389 0.417 8.813 2.256 Primarily Sorbed to Solids St.Dev 0.597 0.232 0.315 1.970 0.533

n = 68 S A B L V

Real Chemicals Mean 1.328 0.060 0.194 8.982 1.895 Primarily Sorbed to Solids Stdev 0.455 0.163 0.207 1.606 0.371

n = 56 S A B L V

Hypothetical Chemicals Mean 0.999 0.430 0.575 5.101 2.188 Multimedia Chemicals St.Dev 0.669 0.228 0.355 1.889 0.666

n = 50 S A B L V

Real Chemicals Mean 0.969 0.105 0.480 6.768 1.579 Multimedia Chemicals St.Dev 0.404 0.195 0.335 1.448 0.350

n = 117 S A B L V

Hypothetical Chemicals Mean 0.899 0.587 0.540 5.319 2.109 Chemicals with High A Stdev 0.623 0.125 0.360 3.195 0.630

n = 97 S A B L V

Real Chemicals Mean 0.817 0.329 0.462 4.898 1.175 Chemicals with High A St.Dev 0.354 0.219 0.223 2.134 0.544

n = 118 S A B L V

Hypothetical Chemicals Mean 0.913 0.201 0.599 5.751 1.847 Chemicals with Low A St.Dev 0.642 0.113 0.362 3.266 0.757

n = 138 S A B L V

Real Chemicals Mean 1.104 0.000 0.343 6.891 1.503 Chemicals with Low A St.Dev 0.479 0.000 0.325 2.537 0.468

147

Chapter 4

Investigating the Time Resolution of Water and Organic Carbon

Balances and their Effect on Contaminant Fate Modeling


Manuscript in Preparation.

148

1 Introduction Multimedia mass-balance models describing the fate and transport of contaminants in the

environment are integral tools in both the prediction of environmental hazard metrics for

hazard indicators such as long-range transport (LRT) and environmental persistence, and in

the interpretation of observed environmental distributions.24,147

Comparison of the LRT

metric Characteristic Travel Distance (CTD) and the environmental persistence metric

Overall Persistence (POV) among different models has shown that model scale, structure and

assumptions have a stronger influence on the prediction of CTD than POV.37

The observed

differences are likely an underestimation of the variability possible in CTD predictions

because all of the models assessed made the assumption of continuous rainfall, a factor that

is well known to have a strong influence on the long range transport of many

contaminants.148

Comparing empirically derived LRT metrics with model predictions for

specific chemicals has highlighted the importance of other variable atmospheric parameters

such a variable atmospheric oxidation rates and wind conditions.149-151

Interactions of the

atmosphere with the surface are also important for contaminant fate and transport; the

cycling of carbon through the terrestrial and marine systems and other fluctuations of

environmental parameters on global and local scales can exert a strong influence on the fate

of many contaminants.19,152

Atmospheric transport and deposition of contaminants is very complex, with many variable

and interacting environmental parameters to consider. Modeling chemical transport in a well-

defined, consistent climatic scenario with explicit consideration of all the major parameters

is possible with detailed parameterization,151

but many of the tools used to screen chemicals

for LRT potential use generic, time averaged environmental parameters. There is nothing

necessarily wrong with using steady state, time averaged models to screen chemicals for

environmental hazard metrics; the data inputs are minimal which facilitates high-throughput

screening of large chemical lists. Furthermore, many chemical parameters such as

partitioning properties, degradation rates, emission amounts and modes of emission are often

very uncertain and are assumed to introduce errors into the prediction of environmental risk

that are comparable to the error introduced by assuming steady state. Steady state

149

approximations and time averaging of important environmental parameters have been

explored to some degree as a compromise between model simplicity and the accuracy of the

resulting predictions;148,153

the errors and uncertainties associated with these assumptions are

explored in greater detail in this work.

In this study the human exposure model CoZMoMAN34

is altered to explore the effects of

environmental variability and time averaging of environmental parameters. CoZMoMAN is

composed of a food chain bioaccumulation model linked with a regional scale multimedia

mass balance model.128

The model has previously been parameterized for the western Baltic

Sea drainage basin, which is left largely unaltered in the current work. In addition to a

chemical mass balance the model solves mass balance equations for water and particulate

organic carbon (POC) balances. Water and POC mass balances are important because the

pools and fluxes of these two phases act as reservoirs and carrier phases for contaminants in

the environment, controlling their fate and transport. Pools and fluxes of organic carbon are

the most important for many semi-volatile, hydrophobic chemicals such as PCBs and

chlorinated pesticides, but many new chemicals of concern are less hydrophobic and the

water balance will be equally or more important for these chemicals. The CoZMoMAN

model ensures that the water and POC balances are internally consistent by optimizing them

to environmental parameters defined by the user at the beginning of each simulation,meaning

that the magnitudes and directions of the fluxes, and the relative sizes of the pools, cannot

violate the law of mass conservation.

The aim of the current work is to introduce variability into the environmental mass balances

of CoZMoMAN and then to apply the model to a range of chemicals and determine the

effect of this variability on chemical fate and transport at different time resolutions. A

previous version of the environmental portion of CoZMoMAN included a variable water

balance,154

but had many empirical fitting parameters that needed to be manually fit and

environmental input parameters were only at a monthly resolution. There are a number of

reasons to be interested in the short term variability of contaminants; long term averages of

environmental contaminant concentrations may be within safe levels, but at short time

150

resolutions spikes related to sudden strong fluxes in the water and POC balances may exceed

safe levels. Interpretation and design of environmental sampling and monitoring also

requires the consideration of variability, to determine how representative samples collected

over a short period of time are of the average concentrations. Long range transport of

chemicals through the atmosphere is also affected by short term variability due to the

interacting effects of relevant environmental parameters. To investigate these effects the time

resolution of environmental input parameters is varied on a scale ranging from hours to

months and the effects on model outputs and the LRT metric CTD are presented.

2 Methods The environmental portion of the CoZMoMAN model has five complete mass balances;

128

these are the mass balance of the contaminant, which has remained unaltered in the new

model, and the mass balances of air, organic aerosol, water and organic carbon which for this

work have all been altered to allow for more flexibility in defining intra-annual variability.

Figure 25 shows the four environmental mass balances in the version of CoZMoMAN

developed for this work, referred to as CoZMoMAN-EV. Some details of the mass balances

shown in Figure 25 are different from CoZMoMAN, these are described in the Appendix

(Section 6.1). The environmental mass balances are distributed among four subsystems of

the model; air and organic aerosol are restricted to the atmospheric subsystem, the water

balance spans the terrestrial and the aquatic subsystems and the organic carbon balance spans

the terrestrial, aquatic and sediment subsystems. Water and air balances are independently

defined, but the organic carbon balance is strongly dependant on the water balance and the

organic aerosol balance is dependent on the air balance. The mass balance of chemical is

dependent on all four environmental mass balances and spans all subsystems of the model.

Note that in Figure 25 some fluxes (arrows) cross subsystem boundaries; these fluxes are

used to introduce variability into the mass balances and the mass balances then respond to

the variability. Production and degradation fluxes also essentially cross the system

boundaries by introducing and eliminating organic aerosol or organic carbon from the mass

balance and are also used to introduce variability to the mass balances. Variability is

introduced into the model by reading in a deterministic pattern of variability for each of four

151

major environmental parameters that act on the environmental mass balances at the system

boundaries; these are temperature, wind speed, rainfall and biological production of POC in

the aquatic subsystem. A fifth environmental parameter, atmospheric OH radical

concentration, is also used to introduce variability into the model but affects only the mass

balance of contaminant.

Figure 25 - Schematic of the Major Pools and Fluxes of the Four Environmental Mass

Balances in CoZMoMAN-EV and their Interconnectivity.

152

Most of the fluxes spanning the subsystem boundaries in Figure 25 have their pattern of

intra-annual variation determined by relating them to one of the four variable environmental

parameters. Air and organic aerosol inflows and outflows are a function of wind speed. Wet

deposition of organic aerosol is a function of rainfall, and rainfall is a direct function of the

intra-annual pattern of rainfall rate. Evaporation is a function of wind speed and temperature.

Production of POC in the aquatic system is naturally a function of the primary biological

production rate, and of temperature. Degradation of POC is also a function of temperature.

Resuspension of sediments from accumulating sediments is a function of wind speed, and

resespension of sediments from eroding sediments is an indirect function of rainfall through

water flow rates. Other fluxes vary in time, but their pattern of variation is directly related to

the size of their originating pools by the user defined parameterization, rather than being

determined by external forces. All of the environmental mass balances are complete at all

points during the year, but in addition to ensuring closure there is a desire to set the relative

magnitudes of the fluxes and the sizes of the pools to specific values.

Magnitudes and sizes are variable over the course of each year, but many parameters have a

user-defined annual average to which the fluxes and pools are fitted. This is achieved with

fitting parameters which, in the case of fluxes, scale the magnitude of each flux to match the

user defined annual average magnitude without directly affecting its pattern of intra-annual

variation. A more refined approach is required to ensure the annual average sizes of the pools

are fitted to the user defined annual average. In each pool of the water, organic aerosol and

POC balances that has a user-defined annual average size one flux in the mass balance

equation is selected as the fitting parameter (the pool of air is fixed in size). This is generally

the flux in the mass balance that is the most uncertain, or difficult to determine, and therefore

adjusting the magnitude introduces little net uncertainty into the model. This procedure of

fitting the parameters of the fluxes and pools to user defined values is called optimization.

Mass balances for air and atmospheric particles are optimized simultaneously, then the mass

balances for water and POC are optimized sequentially. The goal of the mass balance fitting

algorithm is to have a continuous pattern of intra-annual variability, where the state of each

153

mass balance matches at the beginning and the end of the year, so that the same pattern can

be used during every year of the simulation when modeling contaminant fate. The fitting

algorithm solves the mass balance equations at each hour of the year using the finite

difference approximation. When the end of the year is reached the state of the mass balances

are carried over to the next year and the pattern of variable environmental parameters is

repeated starting from the beginning. An initial guess at the fitting parameters is made using

average values defined by the user and the steady state mass balances of CoZMoMAN.

Before any of the fitting parameters are altered the algorithm runs until there are no inter-

annual variations in the mass balances. Fitting parameters are adjusted at the end of each

year; if the alteration of a fitting parameter causes inter-annual variability in the associated

mass balance then fitting parameters associated with that mass balance are not adjusted again

until the inter-annual variation subsides.

Hourly resolved meteorological data consistent with the temperature and wind speed patterns

of the western Baltic Sea drainage basin were obtained from the online Canadian Climate

Norms Database155

including air temperatures, wind speed, visibility, air pressure and

weather observations. Precipitation data were obtained from the same source at a daily

resolution. Using simplistic heat transfer, solar insolation and phytoplankton growth models

an interconnected set of hourly resolved patterns was created from the initial data for air,

terrestrial and water temperatures, fresh water and marine primary production rates, rainfall,

wind speed and atmospheric OH radical concentrations; these models are described in the

Appendix (Section 6.2). Hourly resolved data were then averaged for longer time periods

evenly divisible into 8760 hours per year; 24h (daily resolution), 146h (~weekly resolution),

730h (~monthly resolution), and 2190h (seasonal resolution). Time averaged values were

then interpolated to an hourly resolution. An example of the resulting data is shown in Figure

26, for air temperature and biological production. Similar figures for other parameters,

details regarding the selection of the dataset, and the models used to process the data are

discussed in the Appendix (Figure 34 and Section 6.2).

154

Figure 26 - Air temperature (A) and primary biological production rate in fresh water

(B) at hourly, daily, weekly, monthly, and seasonal resolutions.

155

Because the environmental model used in these calculations has a low spatial resolution the

validity of the simulations with a high time resolution is uncertain. Possible differences in

the results presented here that would arise from increasing the spatial resolution of the

environmental model are difficult to predict. Depending on the implementation details the

CTDs predicted at higher time resolutions may decrease to values closer to the steady-state

predictions, but it is also possible that they may increase and the spread of possible outcomes

would certainly increase, so the exact effects are still an open question.

Simulations were run for hypothetical contaminants with specific combinations of chemical

partitioning properties. Three linked pure-phase partitioning properties are used to

approximate contaminant fate in the environment; air-water partitioning, octanol-air

partitioning and octanol-water partitioning. These properties are calculated as the logarithm

of their respective partition coefficients and are locked in a thermodynamic cycle defined by

equation (30).

log KOW = log KAW + log KOA (30)

In combination with empirical relationships relating octanol sorption to sorption in various

organic carbon pools these chemical properties predict contaminant distribution between air,

water, soil and sediment compartments. A chemical partitioning space is a grid of points

where each point corresponds to a specific combination of the three partitioning properties,

with a corresponding set of model outputs predicting the chemical fate of a hypothetical

chemical with those properties. The chemical partitioning space used to explore the model

outputs of CoZMoMAN-EV is contained within the bounds -5 ≤ log KAW≤ 1, 4 ≤ log KOA≤

12, and simulations were run for five years of continuous emissions. Temperature

dependences for each combination of log KAW and log KOA were calculated with an empirical

equation relating the partition coefficients to the enthalpies of phase change, the derivation of

which is described in the Appendix (Section 6.3). Two simulations were run for all

hypothetical chemicals; one ―labile‖ simulation and one ―persistent‖ simulation.

Environmental half-lives roughly corresponding to the persistence screening criteria of the

Stockholm Convention on POPs5 were calculated for air, terrestrial, water and sediment

compartments at their respective annual average temperatures. These values were then

156

shifted to lower and higher values to create the labile and persistent half-life values. The

temperature dependence of the degradation rates is assumed to be uniform for air and for

surface compartments at all points in the chemical space.

The mode of emission, i.e. the fractional distribution of emissions to the compartments air,

cultivated soil and fresh water, is assumed to be different for different hypothetical chemicals

and dependent on their location in the chemical partitioning space. The rationale for this is

that the mode of emission is not independent of chemical properties. A very volatile

chemical is unlikely to be emitted directly to soil or water, as fugitive emissions from

commercial or industrial use are the most likely route of release. Low-volatility chemicals

are likely to remain in place wherever they are emitted, or wash out with flowing water if

they are water soluble. Emissions of low-volatility chemicals used in commerce may be from

direct applications to soil (i.e. pesticides), amendment to agricultural soil in sewage

treatment plant biosolids, accidental spills, or leaching from landfills. Equilibrium

distribution between fresh water, cultivated soil, and a volume of air corresponding to the

surface area of these two compartments is calculated for each hypothetical chemical at the

annual average temperatures. Emissions are then distributed among these three

compartments according to the equilibrium distribution for each hypothetical chemical. It

was further assumed that the inflowing air contained no contaminants, which causes the air

concentrations to reach pseudo-steady state very rapidly, and therefore the amount of

contaminant in air is largely determined by advection into and out of the system. This means

that the simulations essentially represent a ―source region‖ for the contaminants.

3 Results A number of test calculations were run to confirm that the CoZMoMAN-EV produced model

outputs within the range of similar calculations done in CoZMoMAN; because although the

pattern of intra-annual variability is different in each model the annual average magnitudes

of the contaminant amounts should be fairly similar. This was done using monthly resolved

environmental parameters in both models. In each case the same chemical properties and

emission fractions, derived as described above, were used for both the new model and the

157

old model. Combinations of log KOA and log KAW representative of volatile, water soluble

and lipophilic chemicals for both labile and persistent chemicals were selected for the

comparison. Factor differences were calculated for the amounts of chemical in all

compartments at each hour of the last year of the simulation and then averaged. For

persistent water soluble and lipophilic chemicals the new model is within a factor of 1.5 of

the original model in all compartments, and within a factor of 2 for persistent volatile

chemicals in all compartments except for sediments which are within a factor of 7.

Differences between the models for labile chemicals are much more pronounced in some

compartments; labile water soluble and lipophilic chemicals are within a factor of 1.5 for

aquatic compartments but only within a factor of 3 for air and terrestrial compartments, and

labile volatile chemicals are within a factor of 3 in aquatic compartments and a factor of 3.5

for air and terrestrial compartments. For both labile and persistent water soluble and

lipophilic chemicals the annual pattern of variation predicted by the new and old models in

water and soils respectively are well matched in both the magnitude of the variations and in

the timing of the maximum and minimum values. However, the pattern of variation for labile

and persistent volatile chemicals is more complex in the new model than in the old model,

showing more sensitivity to air temperature, winds speed and atmospheric OH radical

concentrations. Figures comparing the patterns of variation are provided in the Appendix

(Figure 36 and Figure 37).

Simulation results from the five different time resolutions were inspected to ensure that the

total chemical amounts in the model were stable from year to year, so that differences in

other metrics could be reliably attributed to differences in time resolution. Year to year

percent differences in total chemical mass were calculated between the fourth and fifth years

of the simulation at each hour of the year and averaged. For labile chemicals at all points in

the chemical space and at all time resolutions the average year to year differences were <<

1% indicating that a pseudo-steady state had been reached by the fifth year. Persistent

chemicals at all time resolutions had year to year variations in the range of 2.7% (hourly

resolution) to 2.8% (seasonal resolution) when averaged over the entire chemical space;

strongly lipophilic persistent chemicals were furthest from pseudo-steady state peaking at

5.1% (hourly resolution) to 5.2% (seasonal resolution) year to year variation. In the fifth year

158

of the simulation the total chemical mass present in the model environment as a fraction of

the total emissions (retention fraction) was calculated for all points in the chemical space of

each time resolution. Retention fractions for labile chemicals ranged from 0.04% to 3.38%

(average 1.83%) at hourly resolutions and ranged from 0.03% to 3.37% (average 1.82%) at

seasonal resolutions. Retention fractions for persistent chemicals ranged from 0.08% to

31.1% (average 15.7%) at hourly resolutions and ranged from 0.06% to 31.3% (average

15.8%) at seasonal resolutions. In addition to very similar year to year variations and

retention fractions, the chemical distribution at each time resolution was virtually identical.

Simulations using hourly resolved environmental parameters had the largest intra-annual

variation in all compartments of the model, with simulations using lower resolutions

capturing only a fraction of the variation. The percentage of the annual variation captured for

each compartment by each time resolution was calculated using equation (31) where M is the

mass of chemical in an arbitrary compartment at times i and resolutions j:

𝐶𝑎𝑝𝑡𝑢𝑟𝑒𝑑 𝑉𝑎𝑟𝑖𝑎𝑡𝑖𝑜𝑛 = 100% × 1 − 𝑀𝑖 ,𝑗 −𝑀𝑚𝑒𝑎𝑛 ,𝑗 − 𝑀𝑖 ,ℎ𝑜𝑢𝑟 −𝑀𝑚𝑒𝑎𝑛 ,ℎ𝑜𝑢𝑟

𝑀𝑖 ,ℎ𝑜𝑢𝑟 −𝑀𝑚𝑒𝑎𝑛 ,ℎ𝑜𝑢𝑟 (31)

Figure 27 summarizes the percentage of the intra-annual variation that is captured for each

compartment and time resolution for persistent and labile chemicals. As Figure 27 shows,

daily resolved environmental parameters capture most of the variability in surface

compartments for most points in the chemical space. In a number of aquatic compartments

the simulations with averaged environmental parameters capture ~0% of the intra-annual

variation for chemicals with high log KOW values because the pattern of variation in those

compartments is out of phase with the results of the simulation run with hourly resolved

parameters, but the amount of high log KOW chemical present in these compartments is small

so this result is largely unimportant. In both of the soil compartments the pattern of intra-

annual variability is poorly captured for chemicals that are primarily in air or water, and for

multimedia chemicals. The most notable result is the poor performance of the time averaged

environmental parameters for capturing the intra-annual variation of the amount of chemical

in air; the variations in air are much larger because the amount of chemical in air responds

159

faster to changes in environmental parameters than in any other compartment and only

hourly resolution is sufficient to capture this variability.

160

Figure 27 - Percentage of Hourly Resolved Variability Captured with Daily, Weekly,

Monthly and Seasonal Resolved Data for Labile and Persistent Chemicals in all Model

Compartments. Each Cell is a Chemical Space Plot Showing the Fraction of Intra-

Annual Variability Captured at Each Point; Labile and Persistent Environmental

Distributions are Provided as a Guide.

161

CTDs were calculated using two different methods. In method one, referred to as annual

CTDs, an annual averaged atmospheric residence time was calculated by taking the average

mass of chemical in air for the last year of the simulation and dividing by the sum of the

annual average net depositional flux and the annual average degradative flux and then the

CTD was calculated with the annual average wind speed. In method two, referred to as

hourly CTDs, the CTDs were calculated using the mass, net depositional flux, degradative

flux and wind speed at each hour of the final year of the simulation, and then these hourly

CTDs were averaged. However, a problem arises when using hourly resolved environmental

parameters because the atmospheric OH radical concentrations fall to zero at night and in

most cases there is no net deposition (because the system is near steady state and emissions

are mostly to surface media), creating times when the hourly atmospheric residence time

approaches infinity. Conceptually this makes sense; it simply means that for the duration of

the periods of no net losses the chemical will travel as far as the wind will carry it, but when

averaging the hourly CTDs the result is heavily skewed to unrealistically large values by the

near-infinity CTDs. To estimate more realistic CTDs when using hourly resolved

environmental data the hourly CTDs were calculated only for daytime hours, and nighttime

hours in which the net deposition exceeded 1% of average daytime degradative losses.

Nighttime travel was approximated by assuming no net loss of chemical at night, then the

number of days it would take to deplete the chemical from the air with day time loss only

was calculated, and the amount of nighttime travel that would occur during this period was

estimated with the annual average nighttime wind speed. Using this method CTDs for

persistent chemicals were in the range of 20000 to 70000 kilometers, values corresponding to

global distribution, and CTDs for labile chemicals were in the range of 2000 to 8000

kilometers. These CTDs are not necessarily unrealistic but they are very different from those

calculated for simulations using lower resolution environmental parameters and are likely

inaccurate. Therefore, for simulations that were run using hourly resolved environmental

parameters the wind speed, mass of chemical in air, depositional flux and degradative flux

were all calculated as 24 hour running averages instead of using hourly values; which

generates CTD values much more in line with the other resolutions. Figure 28 summarizes

the CTDs calculated by the hourly methods, a numerical comparison of the values calculated

by the annual and hourly methods is provided in Table 17 in the Appendix. In a steady state

162

chemical mass balance model (a level III model for example) the two methods outlined here

would produce identical CTD values, the differences between them essentially shows the

effect of periods when high air concentrations coincide with high winds, low degradation and

low deposition to the surface (Table 17 in the Appendix).

Figure 29 shows the amounts of chemical in air for each of the hypothetical labile and

persistent chemicals calculated with hourly resolved environmental parameters. The annual

average amounts in air are determined by chemical partitioning properties (log KOA,

essentially volatility) but the pattern of variability is mostly determined by environmental

parameters and degradation. During warm, sunny periods the pattern of variation for labile

chemicals is dominated by diurnal variations in temperature and atmospheric OH radical

concentrations, whereas for the persistent chemicals wind speed and temperature are

primarily responsible for the pattern of variation. Model predictions and observations of

PCBs (semi-volatile persistent chemicals) drew similar conclusions on the causes of diurnal

variation and short term variability.156

For both labile and persistent chemicals during winter

the pattern of variation is primarily determined by the effects of weather. This is partly due

to the long length of nights in the winter versus the length of nights during the summer; long

nights and low daytime atmospheric OH radical concentrations mean there is less absolute

difference between daytime and nighttime degradation during the winter. For example the

prominent peak in late January corresponds to a period of cold temperatures, low

precipitation, low winds with high cloud cover and low visibility which reduces the predicted

concentration of atmospheric OH radicals. During this period the concentration of

atmospheric particles also peaks, which increases the capacity of bulk air to hold particle

bound chemicals. These effects combine to allow chemicals to accumulate in the atmosphere

because of an elevated storage capacity with few of the normal removal processes active. To

quantify the relative LRT effectiveness during summer and winter CTDs were calculated for

April to September and October to March of the final year of the simulation using the same

method described above to calculate annual average CTDs with the results from simulations

using hourly resolved environmental parameters. For both labile and persistent chemicals

winter CTDs were on average a factor of 6 times higher than during summer.

163

Figure 28 - Hourly Characteristic Travel Distances (CTDs) in Kilometers for Labile (A-

E) and Persistent (F-J) Chemicals using Hourly (A&F), Daily (B&G), Weekly (C&H),

Monthly (D&I), and Seasonal (E&J) Resolved Environmental Parameters.

164

Figure 29 - Moles of Chemical in Air Simulated using Hourly Resolved Environmental

Parameters for Labile (A) and Persistent (B) Chemicals, Ranging from a log KOA of 4

in Red to a log KOA of 10 in Blue, and a log KAW of -5 to 1.

4 Discussion and Conclusions From the data presented in Figure 27 we suggest that weekly to monthly resolved

environmental data is sufficient for modeling organic contaminant fate in the aquatic

environment. Simulations with monthly resolved environmental parameters capture more

than 75% of the intra-annual variability in water and sediment compartments for chemicals

which are primarily found in the aquatic environment. Chemicals found primarily in the air

also have their pattern of variability in the aquatic environment well captured by monthly

165

resolved values, but chemicals primarily found in the soil may require higher resolutions to

capture their pattern of intra-annual variation in the aquatic environment, because their high

log KOW values makes them quite sensitive to changes in POC concentrations and time

averaging tends to shift the peak POC concentration to later in the year. A similar statement

about the soil environment cannot be made with certainty because many chemicals found

primarily in soil are not at pseudo-steady state, and so although greater than 90% of the intra-

annual variability in cultivated soil is captured by monthly resolved environmental

parameters this is likely because it is a simple linearly increasing trend due to continuing

emissions. Forest canopy, and therefore forest soil as well, responds quite rapidly to changes

in air concentrations, and therefore daily resolved environmental parameters are advised

when the goal is to capture the pattern of intra-annual variability. Although the water and

POC balances play an important role in the transport and fate of chemicals in the terrestrial

and aquatic environments, environmental parameters on a resolution greater than daily, and

in some cases greater than monthly, do not appear to be required to model the patterns of

seasonal variation, at least on a regional scale, because of the relatively slow response of

chemicals in these compartments to changing environmental conditions.

If the goal is to capture the pattern of intra-annual variability in air, however, higher than

daily resolution is required; especially for labile chemicals which respond rapidly to the daily

cycle of high and low temperatures and atmospheric OH radical concentrations. Parameters

affecting the amount of chemical in air should be highly resolved; hourly resolution may not

be required; it should be sufficient to capture the diel cycles of temperature and atmospheric

OH radical concentrations. Rainfall and wind speed data should be of sufficient resolution to

capture the patterns of dry and wet spells, and calm and windy spells, which in many cases

may be on the order of daily resolution. Amounts of labile and persistent chemicals in air

simulated with hourly resolved environmental parameters were correlated against air

temperature, atmospheric OH radical concentration, wind speed, rainfall and primary

biological production to determine the effects of each parameter. For labile chemicals the

order of decreasing correlation was air temperature > atmospheric OH radical concentration

> wind speed > primary biological production > rainfall. For persistent chemicals the order

of decreasing correlation was wind speed > air temperature > atmospheric OH radical

166

concentration > primary biological production> rainfall. In all cases the correlations were

negative, which may seem counter intuitive for air temperature, but this is because peak air

temperatures correspond approximately to peak atmospheric OH radical concentrations and,

due to the temperature dependence of degradation, periods of high temperature also

correspond to periods of rapid oxidation in the atmosphere. Primary biological production is

correlated with both air temperature and atmospheric OH radical concentration so it is not

clear how much of the correlation of this parameter with air concentration is due to draw

down by increased POC concentration and how much is due to coincidental correlation.

Chemical amounts in the model respond to changing environmental parameters, but because

of kinetic limitations there is a lag between the change in an environmental parameter and

the response in chemical amounts. A separate response time will be associated with each

environmental parameter, and these response times correspond approximately to the time

resolutions of the environmental parameters required to capture the variability observed in

the chemical concentration, because any fluctuations on a smaller time scale are too rapid to

produce a response in the chemical concentration. Response times of the amount of chemical

in air to environmental parameters were estimated by comparing the environmental input

parameters to the model outputs and slowly shifting the annual patterns of each

environmental parameter to later times by one hour increments. As the environmental

parameters are shifted the correlation with the amount of chemical in air increases as each

parameter is shifted closer to being in phase with the corresponding response. The shift in

time that produces the peak correlation of each environmental parameter with the amount of

chemical in air is assumed to correspond approximately to the response time to that

parameter. Using this method the response time of labile chemicals to air temperature,

atmospheric OH radical concentration, wind speed, rain fall and primary biological

production were found to be 2 hours, 4 hours, 5 hours, 12 hours and 4 hours, respectively.

The response time of persistent chemicals to the same parameters was found to be 5 hours, 6

hours, 8 hours, 15 hours and 6 hours, respectively. Primary biological production has

response times identical to those of atmospheric OH radical concentrations so the observed

effect on air concentration is likely coincidental and due to their shared dependence on solar

radiance in the underlying models used to generate their values.

167

As can be seen from Figure 29 there are large fluctuations in air concentration possible from

day to day. To quantify the maximum possible day to day variations in air concentration the

respective average amounts of chemical in air of consecutive 24-hour periods were compared

for all points in the final year of the simulation using hourly resolved environmental

parameters. For labile chemicals the maximum day to day variation in the amount of

chemical in the air was a factor of 3.8 to 4.1 difference. For persistent chemicals a maximum

factor of 4.4 to 4.8 difference in day to day amounts of chemical in air was possible. The

annual average day to day variability for labile chemicals was in the range of 14% to 21%

and for persistent chemicals was in the range of 11% to 29%. Because the model is assumed

to be well mixed on a regional scale this is likely a relatively conservative estimate of the

day to day variability in air concentration that can be attributed to changing meteorological

conditions. This is a factor to consider when analyzing data from 24h hi-volume air samples

and attempting to attribute observed day to day variability to changes in meteorological

conditions or changes in emission strength.

Figure 28 and Table 17 in the Appendix show that as the environmental parameters are

averaged to lower resolutions the hourly CTD values approach the annual CTD values. If

this trend was continued to a hypothetical annual resolution it can be seen that these two

values would converge to a single steady state CTD of about 2500 km on average for the

persistent chemicals and about 300 km on average for the labile chemicals (Figure 38). This

means that using a steady state annual averaged resolution for environmental parameters

underestimates the CTD for persistent chemicals by a factor of 3.4 on average and for labile

chemicals by a factor of 2.8 on average. It should be noted however that the range of

differences across the chemical space is relatively small; ranging from 2.9 to 3.5 for

persistent chemicals and from 2.2 to 3.0 for labile chemicals, so although the

underestimation of CTDs by a factor of approximately three seems quite significant the

relative ranking of chemicals for LRT potential is likely affected very little. This means that

when using CTD as an LRT screening metric for the prioritization of chemicals only relative

prioritization should be performed if the CTDs are calculated by a steady state mass balance

168

model, because using a cut-off value for LRT screening will result in false negatives as the

LRT potential is underestimated.


Industry Association (CEFIC LRI-ECO3A2-USTO-0607).

169

6 Appendix

6.1 Variable Mass Balances of Air, Organic Aerosol, Water and Particulate Organic Carbon.

6.1.1 General Description.

All compartments and sub-compartments in the CoZMoMAN have fixed volumes that were

set by the user. In CoZMoMAN-EV some of the compartments have been made variable in

time; these compartments are the volume fraction of organic aerosol in air, the volume

fraction of water in soils, the volume of water in water compartments, and the volume

fraction of particulate organic carbon (POC) in water. The user no longer sets a fixed volume

for these compartments but instead sets the desired value for the annual average volume of

these compartments. The water balance is driven by rainfall, which in the original model was

assumed to be a constant value defined by the user. The organic carbon balance, which is

only solved for the aquatic compartments, is driven by primary biological production, also

having a constant value defined by the user in CoZMoMAN. In CoZMoMAN-EV both of

these driving forces have been made variant in time and the user now defines a pattern of

variation, and the desired annual averages. Many inter-compartment fluxes have also been

altered to be directly dependent on these driving forces, linked to two other environmental

parameters, temperature and wind speed, or are now dependent on the newly variable

compartment volumes. The user now also defines these inter-compartment fluxes as annual

averages instead of fixed values. Although each of the mass balances for air, organic aerosol,

water and organic carbon are allowed to vary in time their mass balance is complete at each

hour of the year, and when the compartment volumes and fluxes are averaged over the whole

year their averages are equal to the annual average values defined by the user.

For inter-compartment fluxes this is achieved simply by fitting a general empirical scaling

factor for each variable flux. All variable compartment volumes have one of the terms in

their mass balances selected as the ―fitting term‖ which is altered to fit the annual average

volume of the compartment to that defined by the user. As an initial guess to the values of

the compartment volumes and the empirical scaling factors the steady state solution to each

of the mass balances is solved with the annual averages set by the user, and then the

170

compartment volumes are set to the annual averages and the empirical scaling factors are

calculated from the steady state solution. Starting from these initial estimates the mass

balances are calculated for an entire year with a one hour time step, then the annual averages

of the compartment volumes are calculated and compared to the user defined annual

averages. The empirical scaling factors are then altered if required and the next year is

calculated starting from the compartment volumes at the end of the previous year. This

continues until a stable, optimized solution is found. Discussion of each mass balance and

equations describing the calculation of fluxes follow; nomenclature and equations are

derived from Wania et al 2006.128

6.1.2 Air Balance.

Air movement into and out of the system is determined by a user defined residence time. The

atmospheric compartment in CoZMoMAN-EV is not spatially resolved, so wind direction

and the exact shape of the air compartment are undefined. In order to link the movement of

air to the wind speed it is assumed that air moving through the system has a path length

roughly determined by the residence time and the wind speed; conceptually this means that

the air follows a fixed path determined by the topography and prevailing weather of the

modeled environment, not necessarily in a straight line. The effective wind cross section is

then defined by the atmospheric height and the surface area of the system divided by the path

length. Air fluxes are the product of wind speed and wind cross section, and the annual

magnitude of air fluxes are fitted to the user defined average residence time by altering the

path length.

The flux of air into and out of the air compartment (aG) in m3/h is solved using equation

(32), where XSX is the wind cross-section in meters of segment X (terrestrial, estuary,

coastal, open), WSX is the wind speed in m/s of segment X (terrestrial, estuary, coastal or

open), and HTA is the height of the atmosphere in meters.

171

aG = ∑ {XSX· HTA· WSX· 3600 s/h} (32)

Total wind cross section for atmospheric compartment (XST) is determined by equation (33),

where PL is the path-length of air travelling through the atmospheric compartment in meters

and ARA is the area of the atmospheric compartment in m2; and the cross section of each

segment X is solved with equation (34).

XST = ARA / PL (33)

XSX = XST· ARX / ARA (34)

Path-length is initially estimated with equation (35), where tA is the user defined atmospheric

residence time and WSS is the wind speed of the terrestrial segment, then equation (36) is

used to calculate the annual average fitted atmospheric residence time (tA(fit)) which is fit to

match tA by altering the path-length.

PL(guess) = tA· WSS· 3600 s/h (35)

tA(fit) = HTA· ARA / aG (36)

6.1.3 Organic aerosol balance.

A mass balance for atmospheric particles was constructed with inputs from inflowing air and

a term to describe particle formation, and outputs in the outflowing air, particle settling and

precipitation scavenging of particles. The annual average volume fraction of particles in air

is fitted to the user defined volume fraction by altering the value for particle formation.

Equation (37) shows the finite difference approximation of the mass balance of organic

aerosol, where VFSA is the volume fraction of organic aerosol in air, VOA is the volume of

the air compartment in m3, qGpro is surface production of aerosol, qGin is the volume of

organic aerosol flowing into the system with inflowing air, qGout is the volume of organic

aerosol flowing out of the system with outflowing air, qGset is the volume of organic aerosol

settling out of the atmosphere and qGscv is the volume of organic aerosol scavenged by

precipitation, all in units of m3/h.

172

VFSA (t+1) = {VFSA (t) · VOA + qGpro + qGin - qGout - qGset - qGscv} / VOA (37)

Values for qGin, qGout, qGset, qGscv are calculated with equations (38) through (41) where

VFSAin is the user defined volume fraction of organic aerosol in inflowing air, DDVX is the

dry deposition velocity to surface compartment x (forest, cultivated or uncultivated soil,

coniferous or deciduous canopy, and fresh, estuary, coastal and open waters), Q is the

particle scavenging ratio and wGAx is the flux of water in rainfall to the surface compartment

x.

qGin = aG·VFSAin (38)

qGout = aG· VFSA (39)

qGset = VFSA·∑ARx·DDVx (40)

qGscv = VFSA· Q · ∑wGAx (41)

As an initial estimate of the volume fraction of organic aerosol in air the VFSA(fit) is set to

the user defined annual average, VFSA(AA), and an initial estimate of the production of

organic aerosol, qGpro(guess), is calculated by solving a steady state version of equation (37)

(dVFSA·VOA/dt = 0) for qGpro and using annual average values for aG, VFSA, DDVx and

wGAx to calculate the other parameters.

6.1.4 Water balance.

Rainfall remains the only net input of water to the aquatic system but evaporation, instead of

being defined as a fraction of rainfall, is now defined as a function of temperature and wind

speed. However, the annual magnitude of evaporation from each surface compartment is

fitted to the fractional evaporation rates defined by the user with a generic fitting parameter.

Evaporation from soil is additionally a function of the volume fraction of water in soil, the

field capacity (a newly introduced parameter) and an empirical evaporation soil stress

function. The volume fraction of water in soil is allowed to vary with time and is fitted to the

user defined annual average volume fraction by altering the field capacity of the soil. Runoff

from each soil compartment to the fresh water compartment is determined by the volume

fraction of water in soil, the field capacity and an empirical runoff soil stress function. If the

173

volume fraction of water in soil reaches the porosity of the soil then all additional net water

inputs flow directly to the fresh water compartment as surface flow. The magnitude of the

annual average runoff is fitted to the user defined runoff fraction with a generic fitting

parameter. Forest canopy now intercepts only a fraction of the rainfall on the forest soil, for

deciduous canopy this fraction is additionally dependent on the volume of the canopy. Net

rainfall interception by the forest canopy is assumed to enter the forest soil by stem flow. All

bulk water compartments are now permitted to vary in depth, and therefore volume, and are

fitted to the user defined annual average depths by altering empirical outflow fractions that

determine the downstream flow of water based on the current volume of water.

Rainfall to any surface compartment x is calculated with equation (42), where wGAx is the

water flux of rain in m3/h and U3X is the rainfall to any segment X in m/h (fresh water,

estuary, soils and forest canopy comprise the terrestrial ―S‖ segment for rainfall and so all

have the same rainfall rate), and ARx is the area of the surface compartment x.

wGAx = U3X·ARx (42)

This equation is modified for the forest soil and canopy compartments and equations (43) to

(45) are used instead, where wGAFD, wGAFC, wGAB are rainfall water fluxes to deciduous

forest canopy, coniferous forest canopy and forest soil, respectively, frUFDinter and frUFCinter

are the fraction of rainfall intercepted by the deciduous and forest canopies (set to values of

0.13 and 0.22 according to Dunne and Leopold157

) and frCovD is the seasonal forest cover

fraction of the deciduous canopy.

wGAFD = U3S·ARFD·frUFDinter·frCovD (43)

wGAFC = U3S·ARFC·frUFCinter (44)

wGAB = U3S· ARB - wGAFD - wGAFC (45)

Evaporation from any water or forest canopy compartment is calculated with equation (46),

where wGxA is the evaporation flux of water from compartment x, θevapxA is a generic

empirical scaling factor, and VPH2O is the vapor pressure of water in Pa at temperature Tx in

Kelvin, estimated with equation (47), derived for this work from the experimental values.

174

wGxA = θevapxA·WSX· VPH2O (46)

VPH2O = 2.875×10-42

· Tx18.2

(47)

For soil compartments equation (46) is modified to equation (48), where SSevap is the

evaporation soil-stress function. The evaporation soil-stress function is estimated with

equation (49), where VFWx is the volume fraction of water in soil compartment x, VFWxcap

is the field capacity of soil compartment x expressed as a volume fraction, and the values of

20 and 3 are empirical parameters. The soil-stress evaporation function was derived for this

work to capture the correct behavior without the use of a detailed hydrological model. The

soil-stress evaporation function quickly falls to essentially zero when the volume fraction of

water in soil is below the field capacity, and increases rapidly at first and then starts to levels

off above the field capacity.

wGxA = θevapxA· WSX· VPH2O·SSevap (48)

SSevap = {arctan(20 · {VFWx - VFWxcap}) + arctan(20 ·VFWxcap)}3 (49)

Empirical scaling factors, θevapxA, are initially estimated by substituting equation (50) for the

term wGxA in equations (46) and (48) and then solving for θevapA(guess), where frUx is the user

defined fraction of rainfall that evaporates for surface compartment x, and other parameters

are set to their annual average values.

wGxA = frUx·wGAx (50)

The annual average fraction of rainfall that evaporates, frUx(fit), is then optimized by altering

the value for θevapxA.

Runoff from any soil compartment x to fresh water W is calculated with equation (51),

where θrunoffxW is a generic empirical scaling factor and SSrunoff is the runoff soil-stress

function. A value for the SSrunoffis calculated with equation (52), where 10 and 2 are

empirical parameters. The soil-stress runoff function has the same general shape as the soil-

stress evaporation fraction, but falls off more slowly below the field capacity and levels off

faster above the field capacity.

175

wGxW = θrunoffxW· SSrunoff (51)

SSrunoff = {arctan(10 · {VFWx - VFWxcap}) + arctan(10 ·VFWxcap)}2 (52)

An initial estimate of the value for θrunoffxWis obtained by substituting the net water

deposition (wGAx - wGxA) for wGxW in equation (51) and solving for θrunoffxWusing annual

average values for rainfall and evaporation. The annual average value for wGxW is then

optimized to fit the user defined runoff fraction for the soil (1 - frUx) by altering the value for

θrunoffxW.

Equation (53) shows the finite difference approximation for the mass balance of water in any

soil compartment x, where VOB is the volume of the soil compartment. For forest soil stem

flows from forest canopy (wGFCB and wGFDB), defined as total net deposition to forest

canopy, are also counted as inputs in the mass balance equation.

VFWx (t+1) = {VFWx (t) · VOB + wGAx - wGxA - wGxW} / VOB (53)

As an initial estimate the volume fraction of water in soil is set to the user defined annual

average volume fraction VFWX(AA), and the field capacity is set to an initial estimate of 60%

of VFWX(AA). The field capacity is then altered to optimize the annual average volume

fraction of water in soil to match user defined annual average.

Net down-stream flow fwGxy from water compartment x to water compartment y is

calculated with equation (54), where θoutflowxy is a generic empirical scaling factor, HTx is the

depth of water compartment x and ARx is the area.

fwGxy = θoutflowxy·HTx·ARx (54)

Equation (55) is the finite difference approximation for the mass balance of water in

compartment x, where fwGzx is the net flow of water from the upstream water compartment

z, which is replaced with the sum of water runoff from soils in the case of the fresh water

compartment.

176

HTx (t+1) = {HTx (t) ·ARx + wGAx + fwGzx - wGxA - fwGxy} / ARx (55)

As an initial estimate the water depth is set to the annual average defined by the user, and an

initial estimate of θoutflowxy is made by assuming steady state conditions (dHTx/dt = 0),

substituting equation (54) into equation (55) and the solving the equation for θoutflowxy using

annual averaged values for all parameters. The water depth is then optimized to the annual

average by altering the value for θoutflowxy.

6.1.5 POC balance.

Primary biological productivity is still the main source of POC in the model, with an

additional contribution from soil runoff in the fresh water compartment as described in

Wania et al.128

Organic carbon fluxes and volume fractions in the terrestrial environment are

not variable in time other than those previously described for the forest canopy, which

remain unaltered. This is a reasonable assumption because the magnitude of intra-annual

variability is quite small in comparison to the amount of organic carbon present. Two major

alterations were made to the POC balance; first, mineralization rates are now a function of

water temperature, with the annual magnitude of mineralization fitted to the user defined

mineralization fraction with a generic fitting parameter. Second, re-suspension intensities are

now a function of water inflow from upstream or runoff, in the case of eroding sediments,

and wind speed, in the case of accumulating sediments. Annual magnitudes of re-suspension

are fitted to user defined re-suspension fractions with a generic fitting parameter.

An initial estimate of all POC fluxes, oG, concentrations of POC in water compartments x,

CPOCx in units of g POC/m3, and the volume fractions of POC in sediment compartments y,

and VFSy, are made using the steady state solution described by Wania et al.128

. Starting

from this initial condition the mass balance of POC is much simpler. For each water

compartment the finite difference approximation for the concentration of POC is shown in

equation (56), where inputs and outputs are as described in Wania et al.,128

but calculated at

every hour of the year instead of only at the beginning of the simulation and δOC is the

density of organic carbon in g/m3.

177

CPOCx (t+1) = {CPOCx (t) · HTx (t) ·ARx + δOC · (inputs – outputs)} / HTx (t+1) · ARx (56)

It was found that the volume fraction of POC in sediments was virtually constant over the

year and therefore solving the mass balance at each hour of the year was unnecessary.

Instead, the volume fraction of POC in sediments was solved using the annual average

values for all optimized parameters and the equations presented by Wania et al.128

at the end

of each year during the optimization process.

Mineralization of POC in water compartment x is described by equation (57), where oGxmiw

is the flux of organic carbon mineralized in units of m3/h, θmiwx is a generic empirical scaling

factor, and Tx and Txavg are the current and average temperatures of compartment x.

oGxmiw = θmiwx· 1.047(Tx-Txavg)

·CPOCx·VOx / δOC (57)

An initial estimate of oGxmiw is made by calculating a value for the annual average,

oGxmiw(AA), using equation (58), where oGxpro(AA) is the annual average primary production

defined by the user, and facOxmiw is the user defined fraction of the primary production that

is mineralized. An initial estimate for θmiwx is made by substituting equation (58) into

equation (57) and solving for θmiwx, using annual average values to calculate the initial

estimate.

oGxmiw(AA) = oGxpro(AA)·facOxmiw (58)

The mineralization in water, oGxmiw, is then optimized to the annual average value,

oGxpro(AA), by altering the value for θmiwx. A similar solution for mineralization in any

sediment compartment xS1 or xS2is followed using equation (59) instead of equation (57),

where VFOxS1/2is the volume fraction of organic carbon in sediment compartment xS1 or

xS2 and using equation (60) instead of equation (58), where oGxS1/2sed(AA) is the annual

average sedimentation flux to sediment compartment xS1 or xS2, and facOxS1/2mis is the user

defined fraction of the sedimentation flux that is mineralized in sediments.

178

oGxS1/2mis = θmisxS1/2· 1.047(TxS-TxSavg)

·VFOxS1/2 · VOxS1/2 (59)

oGxS1/2mis(AA) = oGxS1/2sed(AA)· facOxS1/2mis (60)

Resuspension of sediments is calculated differently for the eroding sediments, xS1, and

accumulating sediments, xS2, of water compartment x. The resuspension flux of organic

carbon from eroding sediments, oGxS1res, is calculated with equation (61), and the

resuspension flux for organic carbon from accumulating sediments, oGxS2res, is calculated

from equation (62), where θresxS1 and θresxS2are generic empirical scaling factors, wGxy is the

downstream flow from water compartment x, and WSX is the wind speed in the overlying

segment X.

oGxS1res = θresxS1· wGxy2 (61)

oGxS2res = θresxS2· WSX2 (62)

Initial estimates for oGxS1res and oGxS2res are made with equations (63) and (64), where

oGxS1sed(AA) and oGxS2sed(AA) are annual average sedimentation fluxes, and facOxS1res and

facOxS2res are the user defined fractions of the sedimentation fluxes that are resuspended.

oGxS1res = oGxS1sed(AA)· facOxS1res (63)

oGxS2res = oGxS2sed(AA)· facOxS2res (64)

The resuspension fluxes are then optimized to the annual average values by altering the

empirical scaling factors θresxS1 and θresxS2.

179

Table 15 - Mass Balance Compartments and Fluxes: their Interdependence and Fitting.

Compartment Volumes

Compartment Fitted to Fitted with

volume of air not fitted, assumed fixed in time

volume fraction of organic aerosol

in air

annual average volume fraction of

aerosol in air

surface production of organic

aerosol

volume fraction of water in soil annual average volume fraction of

water in soil field capacity of soil

depth of water annual average depth of water net downstream flow of water

volume fraction of water in

sediments not fitted, assumed fixed in time

volume fraction of organic carbon

in soils not fitted, assumed fixed in time

concentration of particulate organic

carbon in water

not fitted, allowed to reach a stable

solution consistent with inputs and

outputs

volume fraction of organic carbon

in sediments

not fitted, assumed fixed in time,

but set to a value consistent with

the annual average of inputs and

outputs

Inter-compartment Fluxes

Flux Fitted to Fitted with

inflowing and outflowing air residence time of air path-length

inflowing organic aerosol not fitted, dependant on the flux of

inflowing air

outflowing organic aerosol

not fitted, dependant on the flux of

outflowing air and the volume

fraction of organic aerosol in air

dry particle settling of organic

aerosol

not fitted, dependant on the


in air

scavenging of organic aerosol by

precipitation

not fitted, dependant on the


in air and the rainfall flux of water

to the surface

rainfall flux of water to the surface not fitted, user defined pattern of

variation and annual magnitude

evaporation of water from surface

user defined fraction of rainfall

flux to the surface, dependant on

volume fraction of water in soil

empirical scaling factor

stemflow of water from forest

canopy to forest soil

not fitted, dependant on rainfall

interception and evaporation from

forest canopy

runoff of water from soil to fresh

water

user defined fraction of rainfall

flux to the surface, dependant on

volume fraction of water in soil


production flux of organic carbon

to water

not fitted, user defined pattern of

variation and magnitude

mineralization flux of organic

carbon in water

user defined fraction of production,

dependant on concentration of

organic carbon in water


sedimentation flux of organic

carbon to sediments

not fitted, dependant on

concentration of organic carbon in

water

180

resuspension of organic carbon

from eroding sediments

user defined fraction of

sedimentation, dependant on net

downstream flow of water


resuspension of organic carbon

from accumulating sediments


sedimentation empirical scaling factor

mineralization flux of organic

carbon in sediments


sedimentation empirical scaling factor

erosion of organic carbon from soil

to fresh water

not fitted, dependant on runoff of

water from soil to fresh water

downstream transport of sediments

not fitted, dependant on volume

fraction of organic carbon in

sediments

lateral intra-basin transport of

sediments



sediments

burial rate of sediments



sediments

181

6.2 Meteorological Data and Climatic Scenario Definition.

Objectives. The objectives of the climatic scenario definition were to create a data set of all

environmental parameters required at a resolution of one hour, to ensure that the parameters

were internally consistent, and that they represent realistic weather patterns compatible with

the parameterization of the model. A perfect match to the parameterized environment was

not required however, because the overall objective of the current work is to investigate the

time resolution of the environmental parameters, not necessarily to model the conditions of a

specific environment. Additionally, the aim of the current work is to investigate the effects of

environmental variability on contaminant fate, not to construct a detailed climatic model, so

the models used to process the environmental data are simplistic.

6.2.1 Data source and selection.

Meteorological data were sourced from the Canadian Climate Norms Database.155

The data

is available online for free from Environment Canada, has been quality checked and is

available at hourly resolution. It was found that the climate norms of the city of Prince

George, British Columbia, Canada compared favorably to the default meteorological

parameterization in CoZMoMAN.34

Monthly resolved meteorological data were obtained for

1943 to 2009 and temperature and precipitation were compared to monthly climate norms. It

was determined that data from the year 2000 had the best fit to the Prince George climate

norms and hourly resolved meteorological data were obtained for that year. The year 2000

was a leap year, but the CoZMoMAN model uses a 365 day year so the data for February 29

was discarded. Hourly resolved data were obtained for air temperature, wind speed,

visibility, air pressure, relative humidity and records of qualitative visual weather

observations. Precipitation amounts for the same year were obtained from the same source at

daily resolution.

6.2.2 Heat Transfer Model.

In addition to air temperatures the model requires the input of terrestrial and marine

temperatures. The temperatures in these compartments generally change more slowly than in

air, so the pattern of air temperature variation should not be used directly. A simple model

182

was used to estimate heat transfer from the air to a unit of soil and water instead. For heat

transfer to soil the approximate mass of a 1×1×1 meter cubed block of a dry soil with a

porosity of 0.5 was calculated to be 1.45×106 g, which when combined with the approximate

heat capacity (~0.8J/gK) and thermal conductivity (~1.5W/mK) of dry soil gives the

approximate model in equation (65), where ΔTsoil is the change in soil temperature per hour,

Tsoil is the current temperature of soil and Tair is the temperature of air.

ΔTsoil = 4.66×10-3

·(Tair - Tsoil) (65)

The model was applied to the collected air temperatures with a time step of one hour, at each

step the average air temperature over that time step was used to calculate the change in soil

temperature expected for the next step. An initial guess was made of the soil temperature by

setting it to the initial air temperature, and then the model was run until the final soil

temperature of each year matched the first soil temperature of the same year to within two

decimal places. An identical procedure was used to generate water temperatures with

equation (66), derived from mass, heat capacity and thermal conductivity values of 1.0×106

g, 4.18J/gK and 0.6W/mK respectively. See Figure 30 for the pattern of air, soil and water

temperatures derived this way.

ΔTwater = 5.17×10-4

· (Tair - Twater) (66)

6.2.3 Insolation Model.

A simple model of solar zenith angle was use to calculate the maximum insolation, which

was subsequently corrected for cloud cover, gas scatter and reduced visibility related to

precipitation using the meteorological data collected, following accepted methods.158

Equations (67) through (71) are used to calculate the approximate uncorrected insolation at

the surface, where JH is the Julian hour of the year starting from zero, JD is the Julian day of

the year starting from zero, fy is the year fraction in radians, D is the declination in radians,

HA is the hour angle in radians, ZA is the zenith angle in radians, LT is the latitude in

decimal degrees, and IN is the pattern of insolation at the surface as a fraction of maximum

insolation in the range 0-1.A latitude of 57.52 decimal degrees was used, corresponding

approximately to the middle latitude of the Baltic Sea drainage basin.

183

fy = 2 · π · JH / 8760 (67)

D = 0.006918-0.399912 · cos(fy)+0.070257 · sin(fy)-0.006758 · cos(2·fy)

+0.000907 · sin(2·fy)-0.002697 · cos(3·fy) + 0.00148 · sin(3·fy) (68)

HA = 1/12· π · {JH - JD · 24 - 12} (69)

ZA = 0.5·π-arcsin{sin(LT·π/180) · sin(D)+cos(LT·π/180) · cos(D) · cos(HA)} (70)

IN = cos(ZA) if > 0 else 0 (71)

Gas scattering of sunlight as it passes through the atmosphere is calculated with equations

(72) and (73), where PL is the path length of light through the atmosphere, AP is the

atmospheric pressure in kPa, and GC is the gas scattering correction to insolation, a value

between -1 and 1. All values of GC less than 0 coincide with times where the insolation IN is

set to zero.

PL = {1.002432 · cos(ZA) · cos(ZA) + 0.148386 · cos(ZA) + 0.0096467}

/ {cos(ZA) · cos(ZA) · cos(ZA) + 0.149864 · cos(ZA) · cos(ZA)

+ 0.0102963 · cos(ZA) + 0.000303978} (72)

GC = 1.021-0.084· {PL · 949· AP · 0.000001+0.051}0.5

(73)

Scattering of sunlight from atmospheric water vapor was corrected for using equations (74)

through (76), where AH is the absolute humidity in Pascals, AHavg is the annual average

absolute humidity, TA is the air temperature, RH is the relative humidity in percent, WP is

the water path, and WC is the water scattering correction to insolation.

AH = (2.875×10-42

· TA18.2

) · RH / 100 (74)

WP = 1.42· {1 + (AH - AHavg) / AHavg} (75)

WC = 1 - 0.077 · (PL · WP)0.3

(76)

Aerosol absorbance and scattering of sunlight was corrected for using equation (77), where

AC is the aerosol correction to insolation.

184

AC = 0.935PL

(77)

Weather observations from the Canadian Climate Norms database are composed of either a

cloud cover observation, or - if there is precipitation -an observation of the precipitation type

and amount. A complete cloud cover dataset was created by filling in precipitation

observations with the last cloud cover observation. Cloud cover observations are Clear,

Mainly Clear, Mostly Cloudy and Cloudy, which correspond to percent approximate

coverage fractions of 0, 0.33, 0.67 and 1. It was assumed that complete cloud cover reduced

insolation at the surface by 40%, which was used to scale the percent cloud cover, as shown

in equation (78), where CC is the cloud cover correction to insolation and FC is the fraction

of cloud cover.

CC = 1 - 0.4· FC (78)

Reduced insolation due to effect of precipitation obscuring sunlight was calculated using the

weather observations and the visibility. Whenever the weather observation recorded any

precipitation, or any other weather effect that was not cloud cover such as fog, equation (79)

was used to calculate the effect, where PC is the precipitation obscurity correction to

insolation, and VI is the visibility in kilometers, in the range of 0.1km to 80.5km with a

mean value of 38.25km.

PC = 1 / exp(1 / VI) (79)

Effective insolation is finally calculated with equation (80), and is a value between 0 and 1

representing the fraction of maximum theoretical insolation at the surface.

INeff = IN · GC · WC· AC· CC · PC (80)

Figure 31 shows the annual pattern of variation for effective insolation.

6.2.4 Precipitation Model.

Precipitation rates at hourly resolutions were obtained from the daily resolution data by using

the weather precipitation observations to assign fractions of the daily total precipitation to

specific hours of each rainfall event. Precipitation observations were consolidated into

different intensities of rain and snow, with rare observations such as ―Ice Pellets‖ and

185

―Freezing Rain‖ being binned into comparable rain or snow intensities. The relative

intensities of each type and precipitation observation were estimated by counting the number

of hours each observation had in each day and multiplying by the rainfall amount for that day

and summing these amounts for the whole year. These relative intensities were then all

divided by the lowest intensity precipitation type (Snow Showers) to get the relative

precipitation intensities (RPI) shown in Table 16. Using the weather observations the length

of each precipitation event was estimated. It was possible, likely even, that the precipitation

events included hours from two or more days; to account for this the total precipitation fall

during each event was calculated using equation (81), where Pevent is the precipitation

occurring in a single event, Pi is the total precipitation occurring during day i and ti is the

number of hours of the precipitation event that fell on day i.

Pevent = ∑Pi·ti/ 24 (81)

The precipitation that occurred during each hour j of the event Pj was then calculated with

equation (82), where RPIj is the relative precipitation intensity of each hour j during the

precipitation event.

Pj = ∑ {Pevent·RPIj} / ∑RPIj (82)

All of the daily precipitation amounts were converted into hourly amounts and then

converted into fractions of the annual precipitation that fell during each hour; these are

shown in Figure 32.

6.2.5 Primary Productivity Model.

Hourly resolved primary biological productivity was estimated using hourly resolved water

temperatures (TW) and effective insolation (INeff) as input parameters for a simple biomass

growth model. INeff was normalized using equation (83).

INeff(norm) = INeff / ∑INeff (83)

Equation (84) calculates the biomass growth rate (primary biological productivity) BP where

BM is current biomass, NU is the current available nutrients and TWavg is the annual average

water temperature. In this equation the first term represents to maximum possible biomass

186

growth rate, the second term represents the uptake efficiency of the nutrients and the third

term represents the temperature dependence of the biomass growth rate.

BP = {INeff(norm)·BM ·125} · {NU / (NU + 75)} · {1.066(TW-TWavg)

} (84)

Equation (85) shows the death rate of biomass DR, where the first term represents the annual

average death rate of biomass (31.18 is a fitted value to give the model the desired pattern

and range), and the second term represent the temperature dependence of the death rate.

DR = {BM · 31.18 / 8760} · {1.047(TWavg - TW)

} (85)

The biomass model is run at an hourly time step using the finite difference approximation in

a closed system, meaning there is no net loss or gain of either nutrients or biomass other than

conversion of one to the other. Equations (86) and (87) show the mass balance calculations

for biomass and nutrients at each time step.

BM(t+1) = BM(t) + BP - DR (86)

NU(t+1) = NU(t) - BP + DR (87)

As an initial condition the biomass was set to a value of 1 and the nutrients set to a value of

99, which constrains the system to a total biomass plus nutrient sum of 100. The simulation

was run for consecutive years until the final biomass amount at the end of the year matched

the initial biomass amount of the same year within 0.01 units. The pattern of primary

biological production was normalized by dividing by the sum of the total annual production;

this pattern is shown in Figure 33.

6.2.6 Atmospheric OH Radical Concentration Model.

An hourly resolved pattern of variation in the atmospheric OH radical concentration was

derived by assuming that it follows the same pattern as effective insolation, INeff.

187

Figure 30 - Hourly Resolved Air, Water and Soil Temperatures in Kelvin.

Figure 31 - Hourly Resolved Insolation and Corrected Insolation as Fraction of

Maximum.

188

Table 16 - Relative Precipitation Intensities.

Precipitation Type Relative Precipitation Intensity (RPI)

Moderate Rain 19.05

Moderate Rain Showers 13.02

Rain 8.47

Rain Showers 5.49

Drizzle 3.71

Moderate Snow 12.87

Snow 3.09

Moderate Snow Showers 3.07

Snow Showers 1

Figure 32 - Hourly Precipitation Amounts as a Fraction of Annual Precipitation.

189

Figure 33 - Primary Biological Production Rate as a Fraction of Annual Production.

190

Figure 34 - Time Averaging of Environmental Parameters Soil Temperature (A),

Marine Water Temperature (B), Marine Primary Biological Production Rate (C),

Atmospheric OH Radical Concentration (D), Rainfall (E) and Wind Speed (F).

191

6.3 Derivation of Empirical Relationship for Predicting Temperature Dependence.

Simulations to generate chemical space plots have previously assumed fixed temperature

dependence across the range of log KAW and log KOA, which is a reasonable approximation

but unsatisfying. To make an empirical prediction useful for application to chemical space

calculations two requirements should be met; first the enthalpies of phase change should be

predicted using only the data available for hypothetical chemicals, namely the partition

coefficients, and second the enthalpies of phase change should be thermodynamically

consistent with each other. MacLeod et al. investigated the validity of predicting the enthalpy

of vaporization from the vapor pressure of a pure substance,159

a similar method is employed

here to create predictions for other phase changes. Enthalpies of phase change for air-water

and octanol-air partitioning were taken from Mintz et al.160

and experimental values for the

partitioning coefficients were found for as many of the same chemicals as possible from the

EPI Suite experimental database and Abraham et al. 41,161

Enthalpies of phase change for

octanol-water partitioning were calculated by applying a thermodynamic cycle for chemicals

where the two datasets of Mintz et al. overlapped, and missing partition coefficients were

similarly calculated when the other two values were available.

The enthalpies of phase change were then plotted versus the logarithms of the partition

coefficients, as shown in Figure 35. Note that the slopes of these three plots are similar.

Regression statistics are quite good for octanol-air partitioning and moderate for air-water

partitioning but are poor for octanol-water partitioning, in addition to the fact that the

enthalpies of phase change for octanol-water partitioning have a narrow range of values in

comparison to the other two regressions. To make the best use of all of the data, and to meet

the requirement for thermodynamic consistency, the regressions were locked into a

thermodynamic cycle by using only one slope to fit all three datasets, and fitting only the

intercepts of the octanol-air and air-water partitioning, then calculating the intercept of

octanol-water partitioning by summing the other two intercepts. The three remaining

parameters were fitted by minimizing the error of the regressions weighted by the spread of

the enthalpy values, shown in equation (88), where RMSEXY is the root mean squared error

192

of the regression for partitioning between phase X and Y, and SDXY is the standard deviation

of the enthalpies of phase change for partitioning between phases X and Y.

weighted error = RMSEOA· SDOA + RMSEAW· SDAW +RMSEOW· SDOW (88)

Equations (89) through (91) show the regression equations and statistics used to calculate the

enthalpies of phase change in kJ/mol at each point in the chemical space of log KOA and log

KAW.

ΔHOA = -7.87 · log KOA- 11.5 (89)

(n=120, R2=0.93, RMSE = 5.92)

ΔHAW = -7.87 · log KAW + 31.0 (90)

(n=323, R2=0.66, RMSE = 10.0)

ΔHOW = -7.87 · log KOW + 19.5 (91)

(n=90, R2=0.46, RMSE = 7.93)

193

Figure 35 - Correlation of Enthalpies of Phase Change in kJ/mole with the Logarithm

of their Corresponding Partition Coefficients.

194

6.4 Definition of Environmental Half-lives.

The Stockholm Convention on POPs recommends environmental half-life cut-off values for

environmental persistence of 2 days in air, 2 months in water, and 6 months in soil and

sediment.5 Because anaerobic degradation in sediments is generally considered to be slower

than degradation in soils, for most chemicals, a ratio of 1:3:9 was preferred for

water:soil:sediment environmental half-lives instead of the 1:3:3 ratio of the Stockholm

Convention.162

Environmental half-lives were calculated similarly to the methods presented

by Arnot et al.162

Equation (92) was used to calculate the half-life in hours for air, water, soil

and sediment, where P is the phase correction with values of -2.5, -1, -0.5 and 0.0 for air,

water, soil and sediments respectively, and S is -0.5 for labile chemicals and +0.5 for

persistent chemicals.

HL = 8760 · 10(P+S)

(92)

The CoZMoMAN model assumes environmental half-lives are input at 25°C, but the half-

lives calculated with equation (92) are the desired half-lives at average environmental

temperatures. To account for this the half-lives of surface media were temperature corrected

using the assumed temperature dependence of 30kJ/mol from the average annual temperature

of each compartment to 25°C. A rate constant for reaction with atmospheric OH radicals is

required for the degradation in air; this was calculated from the half-life using the annual

average atmospheric OH radical concentration and temperature correcting the value similarly

to the HL in surface media using an assumed temperature dependence of 10kJ/mol for

degradation in air and the annual average air temperature.

195

Figure 36 - Annual Pattern of Variation in the Molar Amount of Chemical in Air,

Cultivated Soil, and Fresh Water for Hypothetical Labile Chemicals with the original

CoZMoMAN Model (Red Lines) and the New Model (Blue Lines) with Monthly

Resolved Environmental Parameters.

196

Figure 37 - Annual Pattern of Variation in the Molar Amount of Chemical in Air,

Cultivated Soil, and Fresh Water for Hypothetical Persistent Chemicals with the

original CoZMoMAN Model (Red Lines) and the New Model (Blue Lines) with

Monthly Resolved Environmental Parameters.

197

Table 17 - Average, Minimum and Maximum Characteristic Travel Distances of

Persistent and Labile Chemicals at Hourly, Daily, Weekly, Monthly and Seasonal

Environmental Parameter Resolutions, in Kilometers.

Annual CTD (km)

a Hourly CTD (km)

b

Resolution Persistent Labile Persistent Labile

Hourly 2845 (2429-3141) 492 (454-512) 8531 (7147-8875)c 845 (673-909)

c

Daily 2912 (2488-3215) 475 (435-496) 8114 (7087-8330) 794 (661-833)

Weekly 2738 (2385-3000) 437 (402-455) 7972 (7135-8135) 788 (687-814)

Monthly 2605 (2316-2823) 397 (370-411) 7184 (6059-7279) 709 (633-728)

Seasonal 2605 (2420-2746) 331 (318-338) 4052 (3882-4077) 405 (395-408)

aCTD calculated from annual averaged wind speeds, masses, depositional and

degradative fluxes. bCTD calculated at each hour and then averaged over the year.

cCTD calculated with a 24 hour running average instead of hourly values.

198

Figure 38 - Extrapolation to Steady State Characteristic Travel Distance for Labile

Chemicals (A) and Persistent Chemicals (B).

199

Chapter 5

Creating a Fragment-Based QSAR with

Iterative Fragment Selection (IFS)

Trevor N. Brown, Jon Arnot, Frank Wania

Manuscript in Preparation.

200

1 Introduction National and international regulatory programs and a global treaty mandate chemical hazard

and risk assessments to protect human and environmental health. Chemical assessments

require monitoring, exposure, toxicity and chemical property information such as the

octanol-water partition coefficient (KOW) and reaction half-lives (HLs). Compared to the

number of chemicals legislated for assessment there is a paucity of available measured data,

making it necessary to use theoretical approaches to estimate chemical properties, exposure

and toxicity.26,146,163

Chemical distribution properties and HLs can be predicted from 2D

structure using quantitative structure-property relationships (QSPRs) or quantitative

structure-activity relationships (QSARs) such as those implemented in EPI SuiteTM41

and

elsewhere. Existing fragment-based QSA(P)Rs generally require some initial expert

knowledge to identify molecular fragments that influence the chemical property of interest.

More sophisticated methods for predicting chemical properties have also been developed.

For some physical chemical properties, such as KOW, multiple prediction tools are available;

however, the variability and uncertainty between predictions can be substantial.164

For other

chemical parameters, such as HLs, the availability of high throughput screening-level tools is

very limited.41

There is thus a need to develop, evaluate and compare more QSARs to

address data gaps, identify possible sources of uncertainty in training and testing datasets and

in QSAR methodologies, and quantify and reduce uncertainty in applying these predictive

methods for chemical assessments.

Metabolic biotransformation HLs are important chemical parameters for calculating

ecological and human exposure to chemicals.31

In general, the influence of metabolic

biotransformation on human exposures is much greater than the influence of chemical

partitioning properties.25

Currently, there is only one publicly available QSAR for whole

body biotransformation HLs in vertebrates (fish).41,49

Experimental data and expert

knowledge for whole body biotransformation HLs are limited and, due to the complex and

variable nature of this process, also highly uncertain.165

Expert knowledge and ―rules of

thumb‖ of environmental (microbial) degradation mechanisms and recalcitrant chemical

groups have been built into chemical class specific models, (see review by Pavan and

Worth166

) and more general models such as the BIOWIN set of models in EPI SuiteTM.48

Mechanistic knowledge of environmental microbial degradation was initially applied in the

201

development and evaluation of a fish biotransformation HL-QSAR and the results suggest

some overlap in the importance of selected chemical descriptors (fragments) and

biotransformation mechanisms in microbes and fish.41,165

The success in extrapolating this

knowledge is encouraging; however, it is possible that other mechanisms and molecular

fragments contribute to the degree of chemical biotransformation, for which there is no

current ―expert recognition‖. A tool that automatically generates hypotheses for mechanisms

and molecular fragments related to a chemical property of interest could provide new and

deeper insights into QSA(P)Rs and expand expert knowledge.

The primary objective of this research is to develop a new automated tool for the prediction

of chemical properties. The overall method of fragment generation, dataset splitting, and

model selection is referred to as Iterative Fragment Selection (IFS). This new QSA(P)R

method does not require pre-existing expert knowledge (or ―rules-of-thumb‖) on structural

information related to a specific chemical property. 2D fragments are generated

automatically from the dataset and the method iteratively selects fragments and QSARs

through an automated process of fragment selection and model testing. The secondary

objective of this research is to apply the new method to the development and evaluation of a

new set of QSARs for predicting biotransformation HLs in fish from chemical structure. Two

new sets of biotransformation HL-QSARs are developed and evaluated and then compared

with an existing fragment-based HL-QSAR.41,49

The first new HL-QSAR (referred to as IFS-

HLN) uses the dataset splitting methods developed herein and the second new HL-QSAR

(referred to as IFS-EPI) relies on nearly identical datasets used in the development and

evaluation of the HL-QSAR included in the BCFBAF model in EPI SuiteTM.41,49

Similarities and differences between the QSARs are discussed with a focus on model

performance and fragment selection.

2 Methods

2.1 Data Processing

Data processing for the generation of descriptors, model selection, cross validation and

dataset splitting was primarily performed in python 2.6, with the mathematic and scientific

202

computing modules numpy and scipy employed for multiple linear regression (MLR) and

other mathematical functions.

2.2 Overview of the Method

The first step of the method, and one of the most important, is the generation of fragments;

because all subsequent steps rely on the fragment counts for either model selection or to

calculate of chemical similarity scores used to split the dataset into training, external

validation, and cross-validation datasets. All possible fragments contained in the dataset are

generated automatically to be used as the pool of descriptor during model selection. No pre-

selection or sorting of the pool of fragments is done except for those steps required to ensure

that the final model is stable, with no collinear fragments and minimal over-fitting. A simple

set of selection rules is defined for this purpose based strictly on fragment counts rather than

chemical structure. The model selection algorithm is essentially composed of nested

iterations of fragment selection and removal, and the overall method of fragment generation,

dataset splitting, and model selection is referred to as Iterative Fragment Selection (IFS).

2.3 Generation of Descriptors

With the exception of the molecular weight (MW), the pool of descriptors drawn from for

model selection is composed entirely of fragments created from the training dataset. All of

the chemicals in the training dataset are fragmented by breaking bonds according to the

fragmentation rule: all single and aromatic bonds are broken, with the exception of those

bonds involving hydrogen. The resulting single unit fragments lack chemical context which

in some cases would make their valence and number of attached hydrogen atoms ambiguous

if converted to standard SMILES notation. Therefore, fragments were converted to a custom

variation of SMILES notation which includes explicit hydrogen atoms and differentiates

between fused aromatic atoms bonded only to other aromatic atoms and aromatic atoms with

a single bond to hydrogen or another group. To accommodate string matching the custom

SMILES notations are made unique according to the original method defined by Weininger

et al.,168

but with atom priority based on the number and type of connections instead of

counts of hydrogen and non-hydrogen attachments.

203

Larger, more complex fragments are created by recursively combining the single unit

fragments with their neighbours. The maximum number of single unit fragments contained

in a complex fragment (the recursion depth) has been set to eight, to allow for fragments that

consist of a six membered ring with the positions of two substituents defined. Larger

recursion depths suffer from diminishing gains with regards to the total calculation time, as

the number of unimportant fragments that must be sorted through increases more rapidly

than the number of useful fragments. As a result of the recursive combination algorithm the

complex fragments are non-exclusive. For example; if a fragment containing two

neighbouring aromatic carbons with attached hydrogen atoms (HccH) is searched for in

benzene the count will be six, rather than three. This means that some fragments are not

additive, causing the fragment count to sometimes be non-linearly related to the number of

atoms that are covered by the fragment, an undesirable property for a fragment-based QSAR.

Equation (93) defines a metric measuring the degree of additivity, where A is additivity, C is

the number of atoms that are contained in exactly one copy of the fragment summed for all

chemicals in the training dataset, T is the number of atoms in a single fragment and Nsum is

the sum of fragment counts for all chemicals in the training dataset.

A = C / {T · Nsum} (93)

A adopts values between 0 and 1, where a value of 0 refers to a completely non-additive

fragment and 1 a completely additive fragment. Single unit fragments always have an A of 1.

For model selection MW is treated as a single unit fragment.

There are many fragments in the pool of descriptors that are perfectly cocorrelated;

depending on the dataset, at a recursion depth of eight approximately 90% of the complex

fragments added to the fragment pool have either identical counts or are perfectly collinear

with other fragments. All fragments that have identical counts or are perfectly collinear are

grouped together into aggregate fragments and one of the fragments is selected to represent

the entire aggregate. Selection is based on the following prioritized criteria: fragments with

fewer total counts in the training dataset are preferred, fragments with perfect additivity (i.e.

A = 1) are preferred; fragments containing fewer single units are preferred; fragments with

204

higher additivity are preferred; fragments containing fewer atoms are preferred. Fragments

are filtered by each criterion sequentially, only moving to the next criterion in the case of a

tie, until only a single fragment remains.

Fragments are divided into two classes which are treated differently in the model selection.

Fragments contained in three or more chemicals are referred to as frequent fragments.

Fragments contained in only one or two chemicals are referred to as rare fragments and are

subject to special restrictions during the model selection.

2.4 Model Selection

Model selection proceeds according to an Iterative Fragment Selection (IFS) algorithm.

Fragments are added iteratively to the MLR starting from a model containing only the

intercept (the average value of the observations). Forward selection and backward removal

of fragments are based primarily on improvements to the goodness of fit (GoF) to the

training dataset, with the fragments that produce the largest or smallest improvement in GoF

being selected or removed first. Forward selection considers all fragments in the fragment

pool that are not already in the current model. A selected fragment may replace one or more

fragments already included in the model if selection rules prohibit the fragments from being

included together. Scheme 1 shows the general algorithm for model selection.

Scheme 1: Model Selection Algorithm

1. one forward selection from the pool of single unit frequent fragments

2. iterative backwards removal of rare fragments

3. repeat steps 1-2 until no further improvement in GoF is possible

4. one forward selection from the pool of single unit rare fragments




205

8. one forward selection with replacement from the pool of all frequent fragments


10. iterative backwards removal of frequent fragments


12. one forward selection with replacement from the pool of all rare fragments


14. iterative backwards removal of frequent fragments



Steps 1 through 7 usually proceed as a simple stepwise MLR with addition of rare fragments

occurring only rarely. The resulting model from steps 1 through 7 is referred to as the base

model. In subsequent backwards removal steps (steps 9 ,10 , 13, and 14) fragments in the

base model may be removed like any other fragment. Steps 8 through 16 add and remove

fragments from the base model; no fitting is done on residuals.

2.5 Goodness of Fit Metric

The GoF metric used for model selection is the Akaike Information Criterion, corrected for

dataset size (AICC).169

AICC penalizes model complexity and was found to limit the number

of fragments selected to a favorable fraction of the number of chemicals in the training

dataset (less than a 1:5 ratio), as suggested by Dearden et al.167

The assumptions on which

the AICC is based are well suited to fitting measured, and therefore uncertain, data with

fragments derived from two-dimensional structures; namely that a model which perfectly fits

the data does not exist and instead the goal is to find the simplest model which neglects the

least information.170

206

2.6 Fragment Selection Rules

During model selection the MLR was found to be prone to instability and collinearity

between multiple fragments, likely due to the nature of the descriptors; a sparse matrix of

integers. Fragment selection rules were created to prevent combinations of fragments that

lead to instability and collinearity from being included together during model selection. In

general fragment selection rules are intended to exclude instances of fragment coincidence;

meaning fragments which are not perfectly collinear but which are present in the same

chemicals in the training dataset. Coincident fragments create the possibility of additively

combining fragment counts to create perfect collinearity between three or more fragments,

and over-fitting and instability are also possible. Single unit fragments are of course

coincident with many complex fragments and therefore applying selection rules to single unit

fragments would prevent most other fragments from being included in the model selection,

so fragment selection rules are relaxed for single unit fragments.

Four different types of fragment coincidence are prevented by fragment selection rules.

Selection rules preventing perfect coincidence, asymmetrical coincidence and imperfect

coincidence are applied to all complex fragments. General coincidence is prevented when

one of a pair of fragments is a complex fragment and the other is a single unit fragment. No

fragment selection rules are applied if a pair of frequent, single unit fragments is considered.

Perfect coincidence occurs when two fragments are present in all of the same chemicals in

the training dataset. Asymmetrical coincidence is when one fragment of a pair of fragments

occurs only in chemicals that contain the other fragment, but not in all of the chemicals.

Imperfect coincidence is when two fragments are perfectly coincident, except that each

fragment is also present in one additional chemical. General coincidence is when there are

fewer than three chemicals which contain one of the fragments but not the other. As a

general rule then, for any pair of fragments in a fragment-based model there should be three

or more chemicals which contain one of the fragments and not the other, preferably with

both fragments each having some independent occurrences.

207

Further restrictions are placed on adding rare fragments during model selection. This is

because having many rare fragments present was found to increase the instances of model -

fitting and instability, and because the regression coefficients are very sensitive to any errors

in the observed values of the one or two chemicals in which the fragment is present. First the

GoF must be significantly reduced by including the rare fragment; this has been defined as

reducing the AICC by a value of one or more. Second the standard error of the regression

coefficient for the rare fragment must be smaller than the value of the regression coefficient.

Finally the absolute residual for the chemical containing the rare fragment, or for at least one

of the chemicals if the rare fragment is contained in two chemicals, must be greater than

0.674 times the standard error of the MLR (the student‘s t-statistic corresponding to the

upper 50% of the distribution of residuals) if the rare fragment is removed. These restrictions

must be met at all times, not just when the rare fragment is selected, and rare fragments are

removed in steps 2, 5, 9 and 13 of scheme 1 if they no longer meet any of these restrictions,

even if this causes the GoF to be poorer.

2.7 Cross Validation

During model selection variations of k-fold cross validation are applied. Internal training

datasets are all combinations of k minus m folds (k is the total number of clustered sub-

groups of chemicals and m is the number of groups used for internal cross validation) with

the internal validation datasets composed of the remaining m folds in each case. Various

values of k and m are used, but in all cases a separate model is created by MLR on each

internal training dataset using the same set of fragments and the final model presented is the

average of the models developed on the individual internal training datasets. Increasing k

means there are more fold combinations possible and more separate models are created

during the model selection. Increasing m means that each model is fitted on a smaller internal

training dataset and applied to a larger internal validation dataset. Three different value

combinations are used; k = 6 and m = 2,3,4 so the same folds are used in all three cases but

distributed so that one third, one half and two thirds of the chemicals respectively are in the

internal validation datasets.

208

As multiple models are created simultaneously an overall AICC value is calculated to test the

GoF. Each model is applied to the full training dataset and the residual sum of squares

(RSOS) is calculated. These RSOS values contain the residuals for the training dataset points

used to fit the model as well as the data points from the corresponding internal validation

dataset in some ratio defined by k and m. Increasing m has the effect of weighting the GoF

metric towards measuring the predictive power of the regressions and away from measuring

the fit on the internal training datasets. The RSOS value are all summed and then divided by

the number of models to obtain an average RSOS which is then used in the standard AICC

formula.170

Division of the full training dataset into the k folds was done by clustering similar chemicals

together into the same fold. This was found to promote the selection of fragments that are

more widely applicable across different chemical classes. Chemical similarity was

determined by the similarity of the observed values and by the similarity of fragment counts,

which were quantified separately using Tanimoto coefficients to describe the degree of

similarity,171

and the total similarity taken as the product. Fragment counts and observed

values were normalized (as detailed below) and then Tanimoto coefficients were calculated

for each pair of chemicals.171

The set of fragment counts compared for each pair of

chemicals is the set of all fragments contained in one or both of the chemicals.

Fragment counts were normalized according to equation (94) where Fij is the normalized

fragment count, Nij is the fragment count for fragment j in chemical i and Sj is the number of

single unit fragments contained in fragment j.

Fij = {Nij / Sj}0.5

+ 1 (94)

The normalization gives greater weight to smaller fragments (those containing fewer single

unit fragments), and for large fragment counts the differences between similar counts are

smaller. If a fragment is present in one of a pair of chemicals but not in the other then the

value of Fij for the missing fragment is set to Fij = 0 when calculating the Tanimoto

coefficient.

209

Observed values were normalized according to equation (95), where Vi is the normalized

observation for chemical i, Oi is the observed value for chemical i, and Omean is the mean of

all observations.

Vi = {1/3} · {Oi - Omean} / max( |Oi - Omean| ) + 0.5 (95)

This equation normalizes the observed values to between 1/6 and 5/6 with the mean value of

the dataset centered at 1/2. The Tanimoto coefficient is calculated twice, once for Vi and

once for 1-Vi, and the results are averaged so that approximately the same similarity is

calculated for the same deviation between observed values regardless of whether the values

are at the top or the bottom of the range of observed values.

Normalized fragment counts and observed values are used to calculate two separate

Tanimoto similarity coefficients and the total similarity coefficient between any pair of

chemicals is taken as the product of these. Chemicals are then divided into k folds according

to scheme 2. The aim of Scheme 2 is to seed each fold with chemicals which are the most

similar to others in the dataset and then to fill the fold with chemicals that the most similar to

the seeded chemicals.

Scheme 2: Division of Training Dataset into k Folds

1. Chemicals not yet assigned to a fold are sorted by the sum of their similarity

coefficients with each other chemical that also remain unassigned.

2. The chemical that is the most similar to other chemicals that are not yet assigned to a

fold (maximum value from step 1) is added to an empty fold.

3. Chemicals not yet assigned to a fold are sorted by their similarity to the chemicals

that have already been assigned to the fold, with priority given to chemicals with the

fewest low similarity scores.

4. The chemical with the maximum value from step 3 is added to the fold.

5. Steps 3-4 are repeated until the correct size for the fold is reached.

6. Steps 1-5 are repeated until k folds have been assigned chemicals.

210

2.8 External Validation

Division into new training and testing datasets was done by sorting all chemicals into a large

number of small folds, each containing a few similar chemicals, and then selecting some

chemicals from each fold to be placed into the training dataset. Dividing the dataset into

folds was done exactly as described in the cross validation section, with the fold size set to a

value of 6 and the number of folds set to the number of chemicals divided by 6. Small folds

are best for this procedure because it ensures the chemicals in each fold are strongly similar

and the chemical classes in the dataset will therefore be well represented in both the training

and testing datasets. A value of 6 was chosen for fold size because it is the smallest number

that satisfies three criteria; first that three or more chemicals from each fold must remain in

the training dataset to ensure any unique fragments in a chemical class are frequent

fragments in the training dataset, second that at least one chemical from each fold is added to

the training dataset, and finally because it is evenly divisible by three, so that a 2:1 training

to testing dataset size ratio is obtained. When selecting chemicals to be placed in the training

dataset the first chemical selected is that with the maximum summed similarity with other

chemicals in the same fold. The second chemical selected is that with the maximum summed

similarity with other chemicals remaining in the same fold divided by the similarity with the

first chemical selected. This method ensures that from each fold of six chemicals the most

representative chemical is chosen for the testing dataset, and the second chemical, while also

representative, is as dissimilar from the first selected chemical as possible, to ensure

diversity in the testing dataset. The distribution of chemicals into the various training and

validation datasets is summarized in Table 23.

Aggregate fragments have special restrictions applied when generating fragment counts for

external validation datasets. Individual fragments that make up an aggregate fragment are

likely collinear because they are part of a larger structure, and when an aggregate fragment is

selected by the model selection algorithm this larger structure is the information that

contributes to the GoF. Aggregate fragments are only counted in chemicals in the external

testing dataset if the larger structure is also present. This is done by applying two restrictions;

211

first all of the individual fragments in the aggregate fragment must be present in a chemical

and in the same count ratios as in the training dataset. Additionally, any fragments which are

present in all of the same chemicals in the training dataset as the aggregate fragment must

also be present, though there is no restriction on the exact counts. If either of these two

restrictions is not met for a chemical in an external testing dataset then the aggregate

fragment count is set to zero for that chemical.

2.9 Experimental Datasets

The database of fish biotransformation rates compiled by Arnot et al. was used to test the IFS

model selection algorithm.165

The biotransformation rates were averaged and converted to

the base-ten logarithm of the biotransformation half-lives (log HLN according to the

nomenclature used by Arnot et al.) as described in a subsequent publication by Arnot et al.49

In addition, the model selection algorithm was applied to log KOW values, as compiled in the

supporting information of Arnot et al. using the same division into training and testing

datasets.49

These two dataset splittings and their respective 2D fragment-based QSARs are

referred to as the IFS-HLN and the IFS-KOW datasets. Two-dimensional chemical structures

(SMILES notation172

) were retrieved from the EPI SuiteTM

software package database and

manually checked for correctness. A number of chemicals in the database differed only by

their stereochemistry, which is not captured by 2D structures; for each unique 2D structure

the log HLN and log KOW values were averaged and treated as single data points. An

additional dataset that retained the training and testing dataset division of Arnot et al.49

also

had the IFS model selection algorithm applied it, and along with the resulting QSAR is

referred to as the IFS-EPI dataset. In addition, a y-scramble was performed on the full

dataset of fish biotransformation half-lives (the log HLN values were randomly reassigned to

different chemical in the training dataset), another division into training and testing datasets

was made based on this, and another model for fish biotransformation half-life was fitted on

the y-scrambled data. This dataset and the resulting QSAR are referred to as the IFS-yscr

dataset.

212

3 Results

3.1 Fitting and External Validation

A total of six predictive models were fit for fish log HLN, three variations of IFS-EPI and

three variations of IFS-HLN. An additional three variations of the IFS-KOW dataset were also

created. Three variations of non-predictive models were also fit using the IFS-yscr dataset.

For all four datasets three 2D QSAR models were generated by applying three different

internal cross-validations, one with a 1:2 training to testing dataset split, one with 1:1 and

one with 2:1. In all cases the number of folds, k was equal to 6 and the values of m were 4, 3

and 2 respectively. In each case an averaged model was generated as the linear combination

of the three separately cross validated models. Statistics are summarized for all models in

Table 18, and fits are shown in Figure 39 and Figure 41 to Figure 43 in the Appendix.

213

Table 18 - Summary of Statistics for Fitted and External Validation Datasets.

Dataseta Fragments n ME

b MAE

c R

2 AICC

d

IFS-HLN; training dataset; 1:2 c.v. ratio 36 413 -0.024 0.419 0.794 -369.8

IFS-HLN; testing dataset; 1:2 c.v. ratio 206 -0.040 0.471 0.719

IFS-HLN; training dataset; 1:1 c.v. ratio 51 413 0.001 0.352 0.855 -458.6


IFS-HLN; training dataset; 2:1 c.v. ratio 67 413 -0.008 0.300 0.885 -496.8


IFS-HLN; training dataset; average 107 416 -0.010 0.334 0.870 -319.9

IFS-HLN; testing dataset; average 206 -0.036 0.451 0.745

IFS-EPI; training dataset; 1:2 c.v. ratio 31 417 -0.009 0.453 0.764 -340.3

IFS-EPI; testing dataset; 1:2 c.v. ratio 211 -0.009 0.497 0.691





IFS-EPI; training dataset; average 94 417 -0.008 0.365 0.845 -314.9

IFS-EPI; testing dataset; average 211 -0.016 0.457 0.743

IFS-yscr; training dataset; 1:2 c.v. ratio 11 413 -0.014 0.887 0.147 86.9

IFS-yscr; testing dataset; 1:2 c.v. ratio 206 -0.039 0.967 0.009

IFS-yscr; training dataset; 1:1 c.v. ratio 25 413 -0.020 0.796 0.288 54.0


IFS-yscr; training dataset; 2:1 c.v. ratio 59 413 -0.034 0.607 0.554 -28.4


IFS-yscr; training dataset; average 70 413 -0.023 0.736 0.456 104.7

IFS-yscr; testing dataset; average 206 -0.089 0.974 0.008

IFS-KOW; training dataset; 1:2 c.v. ratio 38 413 0.008 0.335 0.948 -504.1

IFS-KOW; testing dataset; 1:2 c.v. ratio 206 0.020 0.407 0.898

IFS-KOW; training dataset; 1:1 c.v. ratio 62 413 0.001 0.247 0.971 -657.4

IFS-KOW; testing dataset; 1:1 c.v. ratio 206 -0.001 0.354 0.923

IFS-KOW; training dataset; 2:1 c.v. ratio 69 413 -0.010 0.218 0.978 -727.8

IFS-KOW; testing dataset; 2:1 c.v. ratio 206 -0.016 0.361 0.911

IFS-KOW; training dataset; average 109 413 0.000 0.241 0.973 -526.0

IFS-KOW; testing dataset; average 206 0.001 0.346 0.925

aInternal k-fold cross-validation (c.v.) ratios are noted according to relative sizes;

internal training dataset : internal testing dataset. bMean Error.

cMean Absolute Error,

dAkaike Information Criterion, corrected for dataset size

214

.

Figure 39 - Model fit and predictions from IFS-HLN for (A,B) 1:2, (C,D) 1:1, and (E,F)

2:1 ratios between internal training and internal validation datasets during cross-

validation.

The training dataset of Arnot et al.49

contained 421 chemicals, but this has been reduced to

417 for the IFS-EPI dataset by averaging the observed values for chemicals with identical 2D

structures, the number of chemicals in the testing dataset remains unaltered (211). The

training and testing dataset splits generated in this work for the IFS-HLN, dataset contain

fewer total chemicals (413 and 207 respectively) because averaging of identical 2D

structures was done before splitting the dataset into training and testing datasets instead of

afterwards.

215

In all four datasets as the cross validation ratio of data points in the internal training and

testing datasets progresses from 1:2 to 2:1 the fit on the training dataset increases, as

demonstrated by the falling mean absolute error (MAE) and the rising correlation coefficient

(R2). The number of fragments selected in all cases increases as well, but the increase in fit

offsets the increasing complexity as shown by the falling AICC values. This is an expected

result, because at a ratio of 1:2 only a third of the training dataset is in the internal training

dataset fit by MLR at any time and so the residuals are larger overall. Each fragment added

has a direct effect on the residuals of fewer chemicals, so the point at which model

complexity begins to outweigh increases in the GoF comes sooner and fewer fragments are

added to the model. At the other extreme, a 2:1 ratio means two thirds of the training dataset

chemicals are present in each internal training dataset fit by MLR so the fit to the entire

training dataset is better. Each fragment added directly decreases the residuals of more of the

chemicals in the dataset and so more fragments are added before model complexity begins to

outweigh the GoF. The training dataset for IFS-HLN has a better GoF than IFS-EPI, but also

has more fragments in each of the three cross validation ratios. The GoF on the training

dataset for IFS-KOW is considerably better than for either IFS-HLN or IFS-EPI, but this

QSAR should not be used for predictive purposes because the full range of log KOW values is

not captured by the dataset and some of the values are likely predictions from EPI Suite,

which would lend the values to prediction by a fragment based model. The fit on the training

dataset of IFS-yscr demonstrates that the model selection algorithm will fit some amount of

random noise, but the GoF is much lower than the predictive models (Figure 42 in the

Appendix).

As the fit to the IFS-yscr training dataset increases, the prediction of the corresponding

external testing dataset becomes increasingly poor, as shown by the increasing MAE and the

wider spread of the predictions, see Figure 42. The R2 values are all essentially 0 for the IFS-

yscr external testing dataset, indicating the fragment generation, model selection, cross

validation and dataset splitting described in this work are not creating an artificially high

predictive power for the resulting 2D QSAR models.

216

For the IFS-HLN and IFS-EPI and IFS-KOW datasets the prediction of the external testing

dataset follows a different order than the y-scrambled dataset and is the same in all three

cases. The best prediction of the external training dataset is for a 1:1 ratio of internal training

to internal testing dataset chemicals during the cross validation, followed by 2:1 (more

chemicals in the internal training dataset) and 1:2 (more chemicals in the internal testing

dataset). The GoF for the IFS-HLN 2D QSAR models are marginally better than the IFS-EPI

models.49

Prediction of the external testing dataset for IFS-KOW values is better than for

either the IFS-HLN of IFS-EPI datasets, so the general pattern is that models with a better fit

for their training dataset have a better fit for the external validation dataset. However,

different models fit on the same dataset do not follow this pattern, and tend to follow the

opposite pattern as demonstrated by the results for IFS-yscr. This is due to the influence of

fitting versus over-fitting; the training dataset must be fit well in order for the model to select

all of the fragments with predictive power, but if the training dataset is over-fitted then

predictive power degrades as the noise in the data is fitted as well with extraneous fragments.

Average models were created for each dataset by taking the linear combination of the three

different cross validated models and then applying them to both the training and external

testing datasets. In the case of both the training and testing dataset for the IFS-yscr dataset

and the training datasets for the predictive models the GoF of the averaged model fell within

the range of the GoF of the individual models. For the external testing datasets for the three

predictive models however the GoF was higher for the averaged model than for any of the

individual models.

Three different log HLN values can be calculated for each chemical using the models fitted

with different cross validation ratios. This was done for the IFS-HLN and IFS-EPI datasets

and the range of predictions for each chemical in the training and external training datasets

were calculated. The average range of predictions for chemicals in the training and external

testing datasets were 0.41 and 0.48 for the IFS-EPI, and 0.35 and 0.36 for the IFS-HLN.

217

3.2 Fragment Coefficients

Summaries of the fragments in each model, along with their respective coefficients, standard

errors, and counts are given in Table 24 to Table 35. For most of the fragments the standard

error of the coefficients (SE) is less than the absolute value of the coefficient. In all three

IFS-KOW models the median relative SE (SE/|coefficient|) of the fragments in each model is

less than 0.5; corresponding to a median p value of less than 0.05. The median relative SE is

higher for the IFS-HLN, IFS-EPI and IFS-yscr datasets, in the range of approximately 0.7 for

the 1:2 cross validation ratio models decreasing to approximately 0.5 for the 2:1 cross

validation ratio models. This means that for IFS-KOW the uncertainty associated with each

fragment is smaller than for the fragments in IFS-HLN or IFS-EPI, which both have an

uncertainty comparable to the fragments IFS-yscr. Fragments in the IFS-HLN and IFS-EPI

models clearly have some predictive power, as opposed to the fragments in IFS-yscr, but

noise in the log HLN data makes the predictions more uncertain than the log KOW predictions.

Uncertainty in the coefficients of the fragments in the models can lead to large prediction

errors; as noted by Arnot et al.49

chemicals may have a number of fragments with both

negative coefficients (labile groups) and positive coefficients (recalcitrant groups) and the

MLR may over-compensate for the presence of the recalcitrant groups by making the

coefficients of the labile groups even more negative. The result is fragments with inflated

coefficients and greater uncertainty, which if applied to chemicals composed of primarily

labile or primarily recalcitrant groups, will lead to large over- or under-predictions. Similar

errors are possible if the models are applied to chemicals much larger or smaller than the

chemicals in the training datasets; the fragment counts are more likely to be outside of the

range of counts in the training datasets and the uncertainty in the coefficients is compounded

by extrapolating to larger or smaller fragment counts. A 2D QSAR should be more robust

and predictive if the absolute magnitudes of the fragment coefficients are small and

chemicals with a wide range of counts are included in the training dataset.

Median absolute coefficient values for the IFS-yscr models range from 0.59 to 0.87,

considerably higher than the median absolute coefficients for the IFS-EPI models. (0.31-

218

0.41) and the IFS-HLN models (0.28-0.37). Median absolute coefficient values for IFS-KOW

models range from 0.29 to 0.35, but the log KOW values span a much larger range than the

log HLN values, so in relative terms the coefficients tend to be smaller in magnitude. This

follows the same pattern as the model GoF for the training datasets, indicating that a good fit

to the training dataset results in fragment coefficients of smaller magnitudes.

In all of the predictive models the absolute fragment coefficients of rare fragments and

aggregate fragments tend to be higher than the absolute coefficients of the other fragments;

this pattern is not apparent in the models fit to the y-scrambled dataset. This supports the

application of the restrictions on fitting and counting the rare and aggregate fragments

outlined in this work; because these fragments seem to be more prone to over-fitting, as

indicated by their generally larger coefficients, and therefore should be used with caution. To

test the effect of aggregate fragment restrictions three different scenarios were applied to two

of the IFS-HLN models (the 1:1 and 2:1 cross validation ratios). Aggregate fragments

restrictions were applied to none of the fragments, all of the fragments and only to aggregate

fragments. In both cases the best prediction of the external testing dataset was made by

applying the restrictions only to aggregate fragments (R2 = 0.724 and 0.712 as shown in

Table 18). The worst predictions were obtained when applying no restrictions (R2 = 0.674

and 0.630) and slightly poorer predictions were made when applying restrictions to all

fragments (R2 = 0.685 and 0.664).

3.3 Fragment Selection

An inspection of the fragments selected in the models fit on the y-scrambled models should

provide an idea of which types of fragments are prone to over-fitting and therefore should be

viewed with skepticism when they are present in the predictive models. By far the most

common fragments in these models are fragments which essentially capture substitution

patterns of chlorines or bromines on aromatic rings and diphenylethers, which is not

surprising given the number of PCBs, PBDEs, dioxins and furans in the dataset. Two other

relatively common types of fragments are long alkyl chains, some branching and others not,

and large fragments containing phosphorus.

219

The IFS-HLN and IFS-EPI models were compared with the IFS-yscr models to check for

fragment overlap. None of the fragments in the IFS-HLN models were present in the IFS-yscr

models, however one to three fragments in each of the IFS-EPI models were present in the

IFS-yscr.

Overlap between models developed on the same dataset but with different cross validation

ratios was also checked. At the extremes were IFS-yscr models with only 5 fragments in

common, and IFS-KOW models with 20 fragments plus MW in common. IFS-HLN models

had 14 fragments plus MW in common, and IFS-EPI models had 8 fragments in common.

This follows the same pattern as the model GoF on the training datasets, so well fit models

consistently select the same fragments when applying different cross validation ratios.

As discussed by Arnot et al.49

log KOW and MW have an important influence on

biotransformation in fish by affecting the bioavailability and equilibrium within the fish.

Residuals for model predictions of log HLN for the chemicals in the external testing datasets

were regressed against the log KOW and the MW. In all cases no relationship between the

residuals and either the log KOW or the MW was observed, even for the two IFS-EPI models

in which the MW was not included by the model selection algorithm. This result implies that

the information about fish biotransformation half-lives represented by the log KOW and the

MW are captured by the IFS algorithm.

4 Discussion and Conclusion

4.1 Model Averaging and Model Uncertainty

One of the most interesting results of the work presented here is the increased predictive

power for the log HLN external testing datasets when the three different cross validated

models were averaged. The reasons for this can be inferred from the observed results for the

four different datasets. In each 2D QSAR model there will be some fragments which explain

the observed values and have predictive power, and some extraneous fragments which are

instead fitted to noise in the dataset. Fragments which consistently have predictive power are

220

more likely to be present in two or more of the three different cross-validation models and so

will retain their full effect after averaging the models. Extraneous and over-fitted fragments

in the models however will likely be different in each case due to different splitting for the

internal training and testing datasets, and so when the models are averaged the absolute

values of their coefficients will be diminished. The median absolute coefficient value for the

average of the IFS-HLN models is 0.11, significantly smaller than the range of median values

for the individual models (0.29-0.37), so the averaged model will be less prone to over- and

under-prediction errors when applying the model to chemicals with fragment counts that are

outside of the range of counts in the training dataset. Averaging of the models must be done

after applying the IFS algorithm in order to gain the increase in predictive power; when

multiple sizes of internal training and testing datasets were included during the IFS cross

validation step the GoF was no better than the models presented in this work (data not

shown).

For predicting values for log HLN the IFS-HLN models are recommended over the IFS-EPI

models. There are a number of reasons to recommend the IFS-HLN models, the first of which

is better GoF for the training and external testing datasets. Other reasons are more consistent

prediction of log HLN between the different models, more consistent selection of fragments,

smaller absolute values for the coefficients and the absence of any fragments which appeared

in the IFS-yscr models. The average IFS-HLN model is show in Table 19 and the fit of the

training dataset and prediction of the external testing dataset are shown in Figure 40

221

Table 19 - 2D Fragment Based IFS-HLN QSAR Averaged from Three Cross-

Validations.

Range of counts and

number of chemicals containing the fragment

Fragmentd Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.

C=C 0.4685 0.1610 1-2 7 1-1 4

C(H)=CH 0.4412 0.2041 1-3 8 1-1 4

c(Cl)c-ccH 0.3153 0.1030 2-4 39 2-4 22

O=CC(H)(H)CH 0.2990 0.1393 1-2 3 1-2 3

O=CC(C(H)C(H)H)H 0.2883 0.1635 1-2 4 1-1 2

O=COH a 0.2327 0.0883 2-2 2 0-0 0

HCC(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)H a 0.2150 0.0870 1-1 2 0-0 0

c(c{c}c(H)cH)(H)cH 0.1980 0.0710 1-2 6 1-1 1

O=CC(C(H)H)H a 0.1959 0.1427 1-2 2 1-1 1

Hccc(cN(=O)=O)H 0.1870 0.0889 1-2 8 1-1 5

HC(H)H 0.1824 0.0281 1-14 205 1-12 97

HccccN(=O)=O 0.1470 0.0845 1-2 3 1-1 1

Hcc(cN(=O)=O)H 0.1276 0.0747 1-2 23 1-2 10

c(cH)Cl 0.1233 0.0309 1-7 125 1-6 70

c(H)(cH)cNH c 0.1089 0.0530 1-4 10 1-5 5

HC=C(H)H 0.1068 0.0550 1-2 9 1-1 7

c(cH)Br 0.1064 0.0434 1-8 19 1-8 8

HccOP(OC(H)(H)H)(OC(H)(H)H)=S 0.1013 0.0850 1-2 10 1-2 3

ccc(c(cC(H)H)H)H 0.0918 0.0739 1-2 3 1-2 2

c(ccc)-cccc 0.0908 0.0954 1-4 6 1-2 2

HC(C(C(C(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.0820 0.0826 1-2 17 1-2 10

c(H)c(H)c(OP(O)=S)c(H)cH 0.0719 0.0347 1-2 4 2-2 1

C(H)(CH)C(H)C 0.0684 0.0526 1-4 4 2-2 2

c(H)(cH)c-cc(H)c 0.0671 0.0267 1-4 21 1-4 10

HCC(C(C(C(H)H)(H)H)(H)H)(H)H 0.0657 0.0389 1-4 21 1-4 11

c(H)(cNH)c(H)cH 0.0621 0.0775 1-4 5 1-4 5

c(c(Cl)c(c)Cl)(c)Cl 0.0565 0.0449 1-6 18 1-6 9

c(H)(cc)c(H)cH c 0.0543 0.0168 1-6 5 4-4 1

c(H)(cH)ccc 0.0504 0.0450 2-4 35 2-4 15

CC(H)C(H)H 0.0504 0.0474 1-4 5 1-2 3

HCC(C(C(H)H)H)(H)H 0.0486 0.0189 1-12 15 2-6 6

c(H)(ccH)ccO 0.0460 0.0439 1-3 15 1-2 7

c(H)(cCcc(H)cH)c(H)c 0.0453 0.0157 2-8 4 8-8 3

Cl 0.0420 0.0164 1-12 165 1-10 90

c(H)(cH)cc-cc 0.0402 0.0198 1-8 13 1-2 6

cH 0.0367 0.0099 1-15 303 1-15 151

C(H)H 0.0367 0.0169 1-28 153 1-20 74

c(cH)c-cc 0.0323 0.0189 1-8 23 1-8 10

c(H)(c)c-ccH 0.0282 0.0465 1-6 27 1-4 11

c(H)cC(H)H 0.0241 0.0149 1-7 32 1-4 16

c(Cl)ccH c 0.0205 0.0156 1-5 97 1-6 49

c 0.0167 0.0161 1-12 303 1-12 152

{c}(cH){c}{c} -0.0129 0.0275 1-20 20 2-18 10

c(c)(cc)Br c -0.0262 0.0140 2-12 5 2-20 3

O(H)c -0.0425 0.0755 1-2 29 1-1 11

HccC(C(H)H)H -0.0429 0.0332 1-4 9 2-4 4

O=COC(C(H)H)(H)H -0.0449 0.0960 1-2 8 2-2 2

c(H)(c(H)cH)c(H)cC=O -0.0497 0.0335 1-4 16 2-2 3

BrCH c -0.0519 0.0412 3-6 3 1-1 1

HC(C(C(H)(H)H)C(H)(H)H)H c -0.0596 0.0469 1-6 12 1-8 7

c(H)(cc){c}{c}cH -0.0605 0.0597 1-4 8 1-2 3

HC(C(C(H)(H)H)(C(H)(H)H)C(H)(H)H)H -0.0635 0.1023 1-2 6 1-2 4

Hcc(c(cC(H)(H)H)H)H c -0.0662 0.0709 1-2 7 1-3 5

Hccc(c(cC(H)(H)H)H)H -0.0690 0.0524 1-2 7 1-2 5

HCC(H)(H)O -0.0733 0.0657 1-3 7 1-1 4

HC(C(C(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)(H)H)(H)H)H -0.0748 0.0539 1-2 4 2-2 2

Br c -0.0749 0.0312 1-7 30 1-10 12

222

c(H)({c}{c})c(H)c -0.0783 0.0737 1-2 13 1-1 6

{n}{c} -0.0789 0.0349 1-2 5 1-1 1

c(H)c-c(c)c -0.0829 0.0957 1-2 18 1-2 9

c(CH)(cH)c(H)c(H)cH -0.0853 0.0729 1-4 8 1-2 3

O(CH)c -0.1120 0.1138 1-2 6 0-0 0

{c}({c}c){c}{c} -0.1373 0.0797 1-1 6 1-1 4

ccc(c(C(H)(H)H)cH)H -0.1453 0.0657 1-4 6 1-2 8

O=CN -0.1517 0.0948 1-2 10 1-1 1

OC(C(H)(H)O)H -0.2006 0.1474 1-1 3 0-0 0

HC(NC(H)(H)H)(H)H -0.2339 0.0668 1-2 5 1-2 4

HCC(C(C(C(H)(H)H)(C(H)H)H)(H)H)(H)H -0.2427 0.0722 1-2 4 0-0 0

ccC(C(H)(H)H)(H)H -0.2554 0.1060 1-2 3 0-0 0

Occ(cC(H)(H)H)H -0.2608 0.0942 1-1 5 1-1 3

O=COC(H)(H)H -0.2622 0.1478 1-1 3 0-0 0

O -0.2852 0.0394 1-8 104 1-8 46

Hcc(c(C(H)(H)H)cH)C(H)(H)H -0.2863 0.1616 1-2 6 1-1 3

C(H)(CH)=CH -0.2863 0.1240 1-2 3 1-2 3

ClC(C(H)(H)Cl)H -0.3204 0.1384 1-2 3 2-2 1

s -0.3539 0.1382 1-2 5 1-1 2

N -0.4526 0.1165 1-2 18 1-2 6

N(=O)=O -0.5125 0.1518 1-2 29 1-1 17

{n}(c)cH -0.5354 0.2119 1-1 3 0-0 0

N(H)H -0.5726 0.1141 1-2 23 1-2 14

NH -0.5889 0.1368 1-2 18 1-2 7

c(c)(c(H)c)Occc -0.6073 0.2560 1-1 3 1-1 1

O=C -0.6227 0.0742 1-4 48 1-2 14

OH -0.7307 0.1022 1-3 48 1-3 19

P=O -1.1274 0.1708 1-1 9 1-1 7

ClcccO

L.C.F. : c(Cl)(cCl)c(cO)Cl b 0.9384 0.2439 1-2 5 1-1 1

HCC(H)(H)C

L.C.F. : HC(CC(C(C(H)(H)H)H)(H)H)(H)H b 0.6835 0.2080 1-2 3 2-2 1

Nc(c)cH L.C.F. : c(c)(N)c(H)c

b 0.4049 0.1691 1-1 3 0-0 0

{c}(c)({c}c)c(H)cH

L.C.F. : {c}(c)(c(H)cH){c}(c)cH b 0.2500 0.1290 1-2 4 0-0 0

{c}(cH)({c}cH)cc(H)cH L.C.F. : {c}1(cc(H)cH){c}c(H)c(H)c(H)c1H

b 0.2190 0.0672 1-2 7 0-0 0

c(H)({c}{c}o)c(H)cH

L.C.F. : {c}1(c(H)cH){c}(cH)o{c}{c}1c(H)c(H)cH b 0.1511 0.1309 1-2 3 1-1 2

cC(H)=C(H)H

L.C.F. : Hcc(c(C(=C(H)H)H)cH)H a,b 0.1407 0.0982 1-2 2 0-0 0

Hc{c}{c}c(cC(H)(H)H)H L.C.F. : Hc{c}{c}(c(cC(H)(H)H)H)cH

b 0.0905 0.0995 1-2 3 0-0 0

c(c-ccc)cc

L.C.F. : c(c-cc(Cl)cCl)(Cl)c(c)Cl b 0.0780 0.0334 1-8 6 2-6 2

c(cOc(cH)cH)(H)c L.C.F. : c(c(H)c)(Oc(cH)c(H)cH)c(H)cH

b 0.0444 0.0536 1-8 11 1-4 4

c(H)(cC)c(H)cH

L.C.F. : c(H)(c(H)cH)c(H)cC b -0.0523 0.0280 1-6 5 4-4 1

Occ(c(cCC(H)(H)H)H)H

L.C.F. : FC(C(Oc1c(c(c(c(c1H)H)CC(H)(H)H)H)H)(F)F)(H)F b -0.0584 0.0326 4-4 3 4-4 1

HC(C(H)(H)Br)H Unique Chemical : 4101682

a,b -0.0761 0.0656 2-2 1 0-0 0

HC(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)H

L.C.F. : HC(C(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)(H)H)H b -0.0771 0.0427 1-2 5 2-2 2

cc(ccN(=O)=O)H L.C.F. : cc(cc(N(=O)=O)cH)H

b -0.0939 0.1507 1-1 6 1-1 2

{c}(cH)({n}){c}cH

L.C.F. : {c}({n})({c}cH)c(H)cH b -0.1171 0.0438 1-2 5 1-1 1

c(H)(ccH)cO

L.C.F. : c(cH)(O)c(H)ccH b -0.1319 0.1000 1-2 5 1-1 2

O=COC(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H L.C.F. : O=COC(C(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H)(H)H

b -0.1393 0.0564 2-2 3 2-2 2

O=COC(C(H)(H)CH)(H)H

Unique Chemical : 85507795 a,b -0.1632 0.1205 1-1 1 0-0 0

223

c1(H){c}({c}(cH)c(H)cc1)cH

L.C.F. : c1(H){c}({c}(cH)c(H)cc1)cH b -0.1997 0.0809 1-2 5 1-1 1

C(H)(Br)C(H)(Cl)C(H)Br


c(H)(cH){n}

L.C.F. : c(H)({n})c(H)cH . c(H)c(H)c(H)cH a,b -0.2596 0.1370 1-1 2 0-0 0

MW c 0.00459 0.00043 68.1-748 413 68.1-

959.1 206

intercept -1.0727 0.0925

aRare fragment.

bAggregate Fragment. L.C.F. is the largest common fragment(s)

among the chemicals containing the aggregate fragment. Unique chemical is the CAS of

the single chemical containing the aggregate fragment. Testing range and frequency

exclude counts forbidden by aggregate fragment selection rules. cTesting range exceeds

training range. dMolecular weight (MW) or fragment in pseudo SMILES format. {}

indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms by

aromatic bonds. All hydrogens are explicit. eRegression coefficient.

fStandard Error of

the regression coefficient.

224

Figure 40 - Model fit and predictions for the averaged IFS HLN QSAR for training (A)

and external validation (B) dataset.

225

4.2 Comparison to Arnot et al. QSAR

Statistics for the averaged IFS-HLN model are comparable to those reported for the Arnot et

al. model,49

hereafter referred to as the EPI-HLN model, for both the training dataset

(R2=0.82, MAE=0.38 versus R

2=0.870, MAE=0.334 for IFS-HLN) and the external testing

dataset (R2=0.73, MAE=0.45 versus R

2=0.745, MAE=0.451 for IFS-HLN). The number of

chemicals with absolute residual values greater than one log unit is 9 in the training dataset

and 19 in the external testing dataset, as opposed to the 21 and 22 reported by Arnot et al.

Table 23 shows the predicted values and residuals for chemicals in both the training and

external testing datasets. There are 11 chemicals with residuals over one log unit for both the

averaged IFS-HLN model and the EPI-HLN model, these chemicals are shown in Table 20.

Most of the errors appear to be due to the inherent shortcomings of 2D fragment based

QSARs. Tetrachloromethane, 1,1,1-trichloroethane and tetrachloroethylene are all predicted

to be more recalcitrant than they actually are, likely because they are all small, highly

chlorinated molecules and the resulting steric effects are not captured by either of the 2D

fragment-based QSARs. Dimethylphosphoric acid 3-methyl-4-nitrophenyl is an example of a

chemical which is erroneously under-predicted because of the presence of too many

fragments with large negative coefficients (P=O and N(=O)=O) without any corresponding

fragments with large positive coefficients to balance them. Other chemicals‘ log HLN are

probably erroneously predicted by both QSARs because they have unique features which are

poorly captured by 2D fragment descriptors, for example hexachlorobenzene, or they

contain groups which are poorly represented in the dataset, for example

decamethylcyclopentasiloxane.

226

Table 20 - Chemicals with log HLN Residuals Greater than 1 in IFS-HLN and EPI HLN

This Work Arnot et al

CAS Chemical Name Obs. Dataset Pred. Resid. Dataset Pred. Resid.

56235 Tetrachloromethane -1.19 train -0.20 1.00 test 0.08 1.27

71556 1,1,1-Trichloroethane -1.39 train -0.15 1.24 train -0.08 1.31

132649 Dibenzofuran 1.35 train 0.22 -1.12 train -0.08 -1.43

541026 Decamethylcyclopentasiloxane 2.74 train 1.03 -1.71 test 1.72 -1.02

6639301 2,4,5-Trichlorotoluene 2.01 train 0.71 -1.31 train 0.69 -1.32

103446 2-Ethylhexylvinylether 0.95 test -0.06 -1.01 train -0.05 -1.00

118741 Hexachlorobenzene 2.44 test 0.93 -1.51 train 1.33 -1.11

127184 Tetrachloroethylene -0.88 test 0.33 1.20 test 0.53 1.41

947728 9-Chlorophenanthrene 1.80 test 0.21 -1.58 test 0.49 -1.31

1839630 1,3,5-Trimethylcyclohexane 1.71 test 0.46 -1.25 test 0.56 -1.15

2255176 Dimethylphosphoric acid 3-

methyl-4-nitrophenyl 0.01 test -1.89 -1.90 train -1.36 -1.37

Table 21 shows the accuracy of the predicted qualitative biotransformation rates of the

external training dataset made by the averaged IFS-HLN model. Qualitative rankings are the

same as defined by Arnot et al.49

There is a centrist trend in the predictions, slowly

biotransformed chemicals generally have their rates over-predicted and rapidly

biotransformed chemicals generally have their rates under-predicted. The reasons for this are

clear from Figure 40B; the slope of the regression is less than one, so the full range of log

HLN values is not completely captured. If the qualitative predictions are categorized as very

slow or slow, moderate and very fast or fast the correctly predicted fractions are 73%, 77%

and 73%, in contrast to the values 82%, 67% and 75% reported by Arnot et al.

227

Table 21 - Qualitative Prediction Accuracy for Fish log HLN Fit on Dataset Splitting

from this Work Using Average Model from Three Different Cross-Validations.

Observed Qualitative Category

Very Slow Slow Moderate Fast Very Fast Sum

Pre

dic

ted

Qu

alit

ativ

e

Cat

ego

ry

Very Slow 69% (20/29) 5% (2/40) 0% (0/62) 0% (0/64) 0% (0/11) -

Slow 28% (8/29) 50% (20/40) 2% (1/62) 0% (0/64) 0% (0/11) 0% (0/206)

Moderate 3% (1/29) 45% (18/40) 77% (48/62) 31% (20/64) 0% (0/11) 0% (0/206)

Fast 0% (0/29) 0% (0/40) 19% (12/62) 59% (38/64) 100% (11/11) 0% (0/206)

Very Fast 0% (0/29) 0% (0/40) 2% (1/62) 9% (6/64) 0% (0/11) 17% (34/206)

Sum - 0% (0/206) 0% (0/206) 1% (2/206) 21% (44/206) 61% (126/206)

4.3 Fragment Analysis

A number of the fragments in the averaged IFS-HLN model shown in Table 19 have exactly

or partially corresponding fragments in the model presented by Arnot et al., hereafter

referred to as the EPI-HLN model. However, direct comparison of coefficients is problematic

because the EPI-HLN model includes the log KOW and MW, whereas the IFS-HLN model

only includes the MW. The inclusion of these parameters means that the fragment

coefficients are essentially relative to the effect of log KOW and MW. In the IFS-HLN model

MW has a coefficient of +0.00459, in the EPI-HLN model it is approximately -0.00256.

However the EPI-HLN model has a positive coefficient for log KOW which is partially

correlated with MW. In many cases the log KOW in the EPI-HLN model will itself be

calculated with a QSAR that has hundreds of coefficients,173

many of which are also present

in the log HLN model. Taking the coefficients from the log KOW model of Meylan et al.,173

the MW of the fragments and the coefficients from Arnot et al., the total effect of each

fragment can be estimated. The fragments of both Meylan et al. and Arnot et al. are ―atom

centered fragments‖, essentially equivalent to the single unit fragments of this work, but

generally with aliphatic or aromatic attachment differentiated. Larger fragments from this

work could be compared with some of the correction factors in the works of Meylan et al.

and Arnot et al. but the relationship is even less clear.

228

Single unit fragments that appeared in all three of the individually cross-validated models

used to create the averaged IFS-HLN model were selected for comparison with the EPI-HLN

model, along with chlorine and bromine fragments; the results are shown in Table 22. The

relationship between the fragments of the averaged IFS-HLN model and the log KOW model

of Meylan et al. along with the EPI-HLN model is not always clear, because the single unit

fragments may be included in larger fragments or in correction factors. In most cases shown

in Table 22 the signs of the total effect of each fragment are in agreement, and the total effect

predicted by the averaged model from this paper falls within the same range of effects

estimated for the model of Arnot et al. This is quite good agreement given the different

methodologies, training datasets, and fragment definitions.

229

Table 22 - Comparison of Fragment Coefficients from IFS-HLN and EPI-HLN.

IFS-HLN EPI-HLN./Meylan et al.

Fragment Coef. MW eff.a sum Coef. log KOW eff.

b MW eff.

a sum

C(H)=CHc 0.441 26 0.120 0.561 2 × =CH- 0.198 0.767 0.236 26 -0.067 0.367

HC(H)H 0.182 15 0.069 0.251 CH3-aliphatic 0.245 0.547 0.168 15 -0.038 0.375

CH3-aromatic -0.087 0.547 0.168 15 -0.038 0.043

O -0.285 16 0.073 -0.212

O-aliphatic -0.023 -1.257 -0.386 16 -0.041 -0.450

O-aromatic -0.069 0.262 0.081 16 -0.041 -0.029

aliph.-O-arom. 0.000 -0.466 -0.143 16 -0.041 -0.184

N(=O)=O -0.513 46 0.211 -0.301 NO2-aliphatic 0.000 -0.813 -0.250 46 -0.118 -0.367

NO2-aromatic -0.022 -0.182 -0.056 46 -0.118 -0.196

NH -0.589 15 0.069 -0.520 NH-aliphatic 0.407 -1.496 -0.459 15 -0.038 -0.091

NH-aromatic -0.289 -0.917 -0.282 15 -0.038 -0.609

N(H)H -0.573 16 0.074 -0.499 NH2-aliphatic 0.407 1.415 0.434 16 -0.041 0.800

NH2-aromatic -0.289 -0.917 -0.282 16 -0.041 -0.612

OHd -0.731 17 0.078 -0.653

OH-aliphatic -0.062 -1.409 -0.432 17 -0.044 -0.538

OH-aromatic -0.473 -0.480 -0.147 17 -0.044 -0.664

P=O -1.127 47 0.216 -0.912 P=O -0.603 -2.424 -0.744 47 -0.120 -1.467

O=C -0.623 28 0.129 -0.494 O=C-aliphatic 0.000 -1.559 -0.478 28 -0.072 -0.550

aliph.-C(=O)-arom. 0.000 -0.867 -0.266 28 -0.072 -0.338

Cl 0.042 35 0.163 0.205

Cl-aliphatic 0.361 0.310 0.095 35 -0.091 0.365

Cl-olefinic 0.000 0.492 0.151 35 -0.091 0.060

Cl-aromatic 0.378 0.645 0.198 35 -0.091 0.485

Br -0.075 80 0.367 0.292

Br-aliphatic 0.273 0.400 0.123 80 -0.205 0.191

Br-olefinic 0.000 0.393 0.121 80 -0.205 -0.084

Br-aromatic 0.396 0.890 0.273 80 -0.205 0.465

aEffect of MW, the MW multiplied by the appropriate coefficient.

bEffect of log KOW, the corresponding

coeffiecient for log KOW from Meylan et al. multiplied by the coefficient of Arnot et al. cEach of the Arnot et al.

and Meylan et al. coefficients are counted twice. Other olefinic hydrogens appear in the averaged model, so it is

not clear if these values are directly comparable. dThe fragment O(H)c from the averaged model essential creates

a specific effect for all OH groups with an aromatic attachment, the summed effect is -0.695.

230

Several other single unit groups not in Table 22 appear in both the averaged IFS-HLN model

and the EPI-HLN model; ‗N‘ (tertiary amines), ‗C(H)H‘ (CH2 groups), and ‗cH‘ (aromatic

hydrogens). Some larger fragments also have directly comparable fragments in the EPI-HLN

model such as ‗O=COH‘ (alcohol group) and ‗O=COC(H)(H)H‘ (ester group, specifically an

acetyl group). Other fragments have a noticeable but less clear relationship to fragments in

EPI-HLN model such as the fragment ‗s‘ (thiofuran and thiazole corrections) and ‗O=CN‘

(urea correction). This is encouraging because it suggests that the IFS method can capture

the same information as fragments selected based on expert judgment and knowledge of

biodegradation pathways. The initial set of fragments used by Arnot et al. was the same as

that used in a previously developed model for aerobic biodegradation,174

and then modified

by inspecting the residuals of the regression and manually selecting fragments that helped to

correct these residuals.49

Indeed some of the manually selected fragments have

corresponding fragments in the averaged IFS-HLN model, such as the aforementioned

thiofuran and thiazole corrections, and the fragment ―o-Chloro/mono-aromatic ether‖ which

roughly corresponds to the aggregate fragment ‗ClcccO‘ from in the IFS-HLN model. Other

than the fragmentation methods described no pre-selection of fragments was performed in

this work, so this agreement illustrates the effectiveness of the model selection algorithm.

An inspection of the remaining fragments in the averaged model should reveal additional

fragments with no analogous fragments in the EPI-HLN model which are important for

predicting fish log HLN. Based on the previous discussion the aggregate fragments are

ignored because they are too uncertain. Fragments which are similar to those found in the

IFS-yscr models are also ignored, which excludes several large fragments containing

phosphorus, several long aliphatic carbon chains, and fragments which seem to capture

specific substitution patterns on aromatic rings. Fragments with a high relative standard error

for their coefficients are also considered too uncertain. Four interesting fragments meet these

requirements, two of which appear in all three of the individual IFS-HLN cross validation

models; these are the fragments ‘c(Cl)c-ccH‘ and ‗c(cH)Cl‘. The other two fragments are

‗HC(NC(H)(H)H)(H)H‘ and ‗c(cH)Br‘. The fragments ‗c(cH)Cl‘ and ‗c(cH)Br‘ increase the

predicted fish log HLN if there is a chlorine or bromine on an aromatic ring which has no

231

vicinal group; the fragment will be counted twice if there is no vicinal group on either side of

the halogen. These fragments have a roughly inverse relationship with the fragments ‗Clcc‘

and ‗Brcc‘ (not included in the model) which capture the effect of having a vicinal group

next to an aromatically substituted chlorine or bromine. All chlorine and bromine atoms

increase the HLN; the information captured by these fragments is essentially that chlorines

and bromines with vicinal groups increase the HLN values to a smaller degree. The fragment

‘c(Cl)c-ccH‘ captures an ortho-effect for PCBs (only PCBs in the dataset have the fragment

‗c-c‘ which represents two aromatic rings joined by a single bond). Biphenyl and a PCB with

four ortho-chlorines will not have this fragment, but PCBs with one to three ortho-chlorines

will have their HLN values elevated. Finally, the fragment ‗HC(NC(H)(H)H)(H)H‘ is a

tertiary amine substituted with two methyl groups, the presence of which makes a chemical

more labile, this is additive with the fragments ‗N‘ and ‗HC(H)H‘ essentially cancelling out

the positive coefficient of the methyl groups if they are on a tertiary amine.

The IFS methods illustrated here for biotransformation can be applied to other datasets for

key chemical parameters. Future work is likely to include a new QSPRs for log KOW and

vapor pressure, in addition to a QSAR for various toxic modes of action and other

parameters relevent to chemical hazard and risk assessment.


Industry Association (CEFIC LRI-ECO13-USTO-081212).

6 Co-Author Contributions Jon Arnot assissted in the comparison of the results of the final model with the previous

model of fish biotransformation half-lives, and contributed text for the chapter introduction.

232

7 Appendix

Table 23 - Training and Validation Datasets Summary.

CAS no. or ID + Name

Average

IFS-HLN

prediction Residual

IFS-HLN

Splitting

EPI-HLN

Splitting

IFS-yscr

Splitting

(1)

6-n-Butyl-2,3-dimethylnaphthalene 0.47 -0.23 train test train

(2)

XDE-536 methyl ester -1.10 -0.22 train test train

(4) 175 Factor J

0.96 -0.62 train test test

(20)

Octaethylene glycol monotridecyl ether -0.41 -0.20 train train test

(40)

NL-93 2.27 0.34 train train train

(60) NL-133

1.73 0.71 train train train

(70)

NL-63a 0.64 -0.28 train train test

(80)

NL-63b 0.91 -0.51 train test train

(201) XDE-537 n-butyl ester

-0.82 0.04 train train train

(205)

Spinosad Factor D -0.39 0.84 train train test

(692)

2-isopropyl decalin 0.61 0.30 train train train

(694) 1 isobutyl 2,5 dimethyl cyclohexane

0.48 0.32 train train train

(696)

1-octylpyrene 1.20 0.06 train train test

(697)a

Methylnaphthalenes 0.11 0.84 train train train

91576a - - train train train (698)b

Di-Me & Et naphthalenes 0.34 0.26 train train test

575417b - - train train test (2355)

Methylisocyanothion 0.31 0.34 train train train

(6011) 3,3',4,4'-Tetrabromodiphenyl ether (BDE 77)

1.62 0.63 train test train

(6017)

2,2',4,4',5,5'-Hexabromodiphenyl ether (BDE 153) 2.07 -0.06 train train train

53190 1.42 0.23 train test train

53703 0.56 -0.35 train train test

56235 -0.20 -1.00 train test test 59507 -0.66 -0.62 train test test

60297 -0.58 -0.12 train train test

60515 -1.22 0.21 train train train 60571c 1.42 0.22 train train test

72208c - - train test test 62442 -0.95 0.40 train train train

63252 -0.97 -0.21 train test test

67663 -0.40 -0.36 train test train 67721 0.27 0.49 train train train

71556 -0.15 -1.24 train train train

72559 1.88 -0.17 train train train 74975 -0.47 -0.27 train test test


75354 -0.54 -0.18 train train train 76448 1.34 0.10 train test train


78400 -1.56 0.47 train train train 78422 0.37 -0.66 train test test

78637 0.90 -0.13 train train train

78795 -0.47 -0.14 train train train 78875 -0.57 -0.49 train train test

233




80057 -0.48 0.75 train train train 81038 0.72 -0.03 train train train

82053 -0.15 -0.54 train test train

82440 -0.48 0.25 train test train 83794 -0.52 -0.07 train train train

84151 0.67 0.19 train test test

84742 -0.93 0.52 train train train 85018 0.04 0.66 train test train

85223 0.40 -0.38 train test train

85687 -0.73 -0.30 train train test 86737 0.50 0.20 train train train

86748 -0.06 0.12 train train train

87616 0.54 0.38 train train train 87683 1.31 0.76 train train train





88197 -0.45 -0.44 train train test 88733 -0.26 -0.32 train test test


88982 -1.05 0.48 train train train 89690 0.06 -0.70 train train test



92513 0.32 0.06 train train train 92693 -0.68 -0.13 train train test


92864 1.00 0.57 train train train 94520 -0.82 -0.33 train train train


95487 -0.94 0.42 train train train 95501 0.26 0.24 train train test


95578 -0.78 0.40 train train train 95647 -0.92 -0.07 train train train


95772 -0.57 -0.35 train test train 95829 -0.27 -0.86 train train test


96297 -1.00 -0.33 train test train 96764 0.36 -0.73 train train test

96968 -1.14 0.27 train test train

97234 -0.40 -0.11 train test train 98088 -0.31 -0.17 train train train


98511 0.38 -0.07 train test train 98840 -0.88 -0.02 train train test


99081 -0.34 -0.45 train test train 99718 -0.66 -0.18 train train train


100016 -1.09 0.58 train train test



101611 -0.05 0.02 train test test 101848 -0.18 0.09 train train test


104405 -0.05 -0.18 train test train 104723 0.77 -0.14 train train test


105055 0.26 -0.27 train test test 105066 0.10 0.40 train train test



234


107506 1.87 -0.39 train test test 108429 -0.59 -0.12 train train test

108430 -0.79 0.10 train test test

108576 0.10 0.49 train train test 108678 0.19 0.21 train train train


109091 -1.02 0.17 train train train 110021 -0.89 0.07 train test train


111444 -0.47 -0.34 train train train 111842 0.30 0.67 train test test


112414 0.40 0.43 train train test 115322 1.32 0.66 train train train

117180 0.48 0.01 train train train

117793 -1.52 0.46 train test train 117817 0.04 0.40 train train train





119562 -0.30 -0.59 train train train 119619 -0.66 -0.52 train test train


120785 -0.32 -0.59 train train train 120832 -0.46 -0.48 train train train

121142 -0.67 0.45 train test test 122145 -0.48 0.52 train train test


123488 0.77 -0.07 train train test 124118 0.02 0.50 train train train


132649 0.22 1.12 train train train 134327 -0.30 0.05 train test test


140669 -0.07 0.03 train train train 142961 -0.17 -0.08 train train train


156434 -0.90 -0.34 train train train 198550 0.27 -0.33 train test train


208968 0.32 0.20 train train train 214175 0.59 -0.01 train train train


226368 0.28 -0.42 train train train 243174 0.33 -0.72 train train train


298000 -0.09 0.64 train train test 298044 0.35 -0.32 train train train


314409 -1.49 0.16 train train train 320605 0.55 1.26 train train train


330552 -1.02 0.05 train train test 333415 0.28 -0.22 train train train





535773 0.12 -0.05 train test train 541026 1.03 1.71 train test train


542187 -0.17 0.58 train train test 544014 0.25 -0.69 train test test

554007 -0.27 -0.39 train test test

554847 -1.22 0.74 train test test 555033 -0.62 -0.42 train train train

579102 -1.05 -0.07 train test test


235


591355 -0.36 -0.55 train train train 603112 -2.21 0.64 train train train


606280 -1.29 -0.01 train train train 608935 0.75 1.12 train test test


611063 -0.03 0.87 train train test 611212 -0.35 -0.39 train train train


616444 -0.67 -0.61 train train train 618622 0.41 0.16 train train test


629594 0.81 0.68 train test train 634673 -0.21 -0.06 train train train


636282 1.00 0.34 train train train 760236 -0.59 -0.69 train train train


789026 1.62 0.09 train test test


877101 0.68 0.44 train train test

877112 0.80 0.60 train train train 883205 0.35 -0.24 train train test


933120 0.84 -0.12 train train train 935955 -0.20 0.11 train test train

950378 -1.18 0.03 train test train 961115 -0.27 -0.75 train train train


1460022 1.86 0.84 train test train 1490046 -0.62 -0.57 train test train


1634044 -0.22 -0.66 train train train 1634782 -1.81 0.77 train test train




1795159 0.73 0.37 train test train 1825214 2.47 0.20 train train test


1836777 0.88 -0.13 train test test 1889674 0.94 -0.07 train train train


1912249 -0.58 -0.47 train train train 2051243 2.90 -0.03 train train train


2173571 0.09 0.15 train test train 2212671 -1.00 -0.08 train train train


2234131 1.25 0.37 train train train 2243621 -0.54 0.02 train test train


2385855 1.94 -0.14 train train train 2437561 0.50 0.23 train test test





2631405 -0.85 -0.41 train train test 2636262 0.08 0.15 train test train




3229003 0.56 0.20 train test train 3296900 -1.33 -0.06 train train test

3674757 0.40 -0.11 train test test


236

3761420 0.33 -0.36 train test test

3766812 -0.79 0.27 train train train 3811492 -0.40 -0.75 train test train




4130421 0.73 0.12 train test train 4175546 0.30 0.42 train train train


4821196 0.17 0.03 train train train 4883721 -1.28 0.38 train train train


4920950 0.48 -0.49 train train test 5323568 0.83 -0.24 train train train


5428546 -1.25 0.01 train train train 5707448 0.41 -0.35 train train train





7045718 0.70 0.14 train train test 10394577 0.47 -0.17 train train train

13116535 0.00 -0.28 train test train

13151343 0.51 0.14 train test train 13358117 0.12 0.18 train test train

13674845 -0.77 -0.54 train train test 13674878 -0.60 0.09 train train train


15087248 0.28 -0.49 train test train 15254258 0.43 0.16 train test train


16219753 -0.35 -0.22 train test test 16606023 2.03 -0.12 train test train

16958922 0.34 -0.77 train train test

17109498 -0.44 -0.59 train train train 17700093 0.11 -0.30 train test train




19408743 1.44 -0.13 train test train 19666309 0.43 0.34 train train train





25311711 -0.37 -0.15 train test train

26444495 -0.69 0.42 train train train 26898179 1.26 0.30 train train test

28106301 0.18 -0.02 train test train



29761215 -0.33 0.19 train train test


32598100 2.16 -0.09 train test train

32598133 1.58 -0.23 train train test 32598144 2.22 -0.16 train train train


32690930 2.26 0.30 train train test 33460025 0.46 -0.01 train test train

33576920 -0.01 0.77 train test train

33857260 0.76 -0.53 train train train 34883391 1.61 0.77 train test test



237


35822469 1.47 -0.26 train train test 35860378 1.67 0.07 train train train




38380028 2.39 -0.08 train test train 38380039 2.69 -0.05 train train test


38411255 2.55 -0.09 train train test 38444734 2.12 0.37 train train test


38444938 2.13 -0.14 train train train 39001020 1.30 -0.08 train train test


39635319 2.55 -0.23 train test test 40186718 2.19 0.28 train train train



41122707 0.91 -0.06 train train test





51207319 1.12 -0.56 train train train 51655653 0.64 0.24 train train train

51908168 2.68 -0.28 train train test

52663588 2.47 -0.12 train test train 52663680 2.63 -0.18 train train test


52663715 2.59 -0.07 train test train 52663726 2.58 -0.15 train test train

52663748 2.66 -0.15 train test train

52663759 2.62 -0.15 train train test 52663782 2.82 -0.28 train test train

52663793 2.52 0.14 train test test


52886358 0.67 0.30 train test test

52918635 0.58 -0.07 train test test 54135807 1.27 0.13 train train train

57018049 0.27 -0.38 train test train

57117314 1.17 0.19 train test test 57117438 0.80 0.10 train train train



59365605 -1.27 0.14 train test train

60123640 1.59 0.43 train train train 60168889 0.83 -0.26 train test train


60348609 1.85 0.11 train train train 60782583 0.67 -0.63 train test train


62338094 1.02 -0.05 train train test



66332965 -0.43 -0.18 train train train 67562394 1.25 -0.35 train test test

68515479 0.13 -0.74 train test train

68515491 -0.12 -0.35 train test train 68526852 -0.09 0.68 train train test

68526863 -0.22 0.30 train test train

69806402 -0.85 -0.22 train train train 70362457 2.14 0.24 train train train

70362468 2.22 0.58 train test test

70362504 1.67 0.02 train test test

238

71585369 1.93 -0.18 train train test

71888896 0.03 0.10 train test test 74472347 2.24 0.19 train test train


74472449 2.79 -0.18 train test train 74472530 3.03 -0.38 train test train


79060609 0.91 -0.48 train train train 81406373 -0.33 -0.28 train test train

83992692 2.07 -0.10 train test train


106220819 1.92 0.09 train train test

106220831 2.55 0.21 train train test 107534963 0.86 -0.09 train train test

120068362 0.96 0.09 train train test

126690662 0.52 0.40 train test train 156052685 -0.23 0.39 train test train

182346210 1.68 -0.34 train train test


189084682 1.65 -0.07 train test train


(5) 2-hexyl tetralin

0.78 -0.15 test test train

(6)

Spinosad Factor A -0.76 1.17 test test train

(9)

Tri-Me naphthalenes 0.27 0.30 test test test

(10)

C-12-2-LAS 0.63 -0.29 test train test

(30) C-12-5-LAS

0.47 -0.38 test train train

(50)

NL-123 1.68 -0.02 test test test

(90)

NL-83 0.62 -0.32 test train train

(203) 175 Factor L

1.08 -0.59 test train train

(693)

1,1,1 trimethyl butyl benzene 0.40 -0.39 test train test

(6014)

2,3',4,4'-Tetrabromodiphenyl ether (BDE 66) 1.47 -0.04 test train train

50293 1.62 0.31 test train train 50328 0.29 -0.25 test train test

55389 0.18 -0.54 test train train

56382 -0.09 -0.32 test train test 56553 0.34 -0.32 test test test

57749d 1.84 0.42 test test test

5103742d - - test train test 5103719d - - test test test

12789036d - - test train test

58899e 0.51 0.98 test train test 319846e - - test test test

319857e - - test test test

319868e - - test train test 62533 -1.02 0.24 test train train

72548 1.42 1.02 test train test

75092 -0.56 0.05 test train test

75650 -0.92 -0.21 test test test

78308f -0.15 0.67 test train test

1330785f - - test test test 78591 -0.56 -0.73 test train train

79005 -0.94 -0.08 test test test

80433 0.52 -0.53 test train test 82451 -1.31 0.71 test train test


86306 -0.21 -0.70 test train test 87832 0.81 -0.94 test train train

88722 -0.42 -0.06 test train train

89612 0.12 0.44 test train train

239

89634 -0.73 -0.23 test test train

90415 -0.38 -0.68 test train train 90948 -0.75 -0.02 test train train


91203 -0.19 0.60 test train train 91941 -0.12 -0.02 test train train

95498 0.10 -0.82 test train test

95567 -0.73 0.49 test train train 95636 -0.31 0.00 test test train

95783 -0.78 -0.22 test test test

95943 0.80 0.99 test train test 96128 -0.07 -1.07 test test train

98157 0.23 0.42 test test train

98544 -0.43 -0.45 test test test 98839 -0.15 -0.56 test test test

99092 -1.03 0.26 test test train

99547 -0.04 -0.23 test test train 99627 0.41 0.14 test test test


100174 -0.55 -0.40 test train test

100403 0.08 -0.50 test test test

100618 -0.44 -0.43 test train test

101144 0.02 -0.17 test train train 103446 -0.06 1.01 test train train

103695 -0.34 -0.31 test test train

106434 0.02 -0.86 test train train 106467 0.36 0.74 test train test

106489 -0.79 0.37 test test train 106934 -0.29 -0.54 test train train


108361 0.47 0.69 test train test 108601 -0.05 -1.02 test test train


108872 -0.12 0.70 test test test 108907 -0.07 -0.95 test train test

109693 -0.31 -0.14 test test train

110009 -0.61 -0.23 test train test 115866 -0.96 0.20 test train train

115968 -1.40 0.74 test train train

118741 0.93 1.51 test train test 119642 0.02 0.73 test train test

120718 -0.87 -0.12 test train test

120821 0.58 0.62 test test test 121697 -0.64 -0.33 test train train

121755 -0.61 0.90 test test test

122349 -0.79 -0.15 test train test 126738 -0.96 0.20 test train test

127184 0.33 -1.20 test test test

128370 0.29 0.03 test train test 129000 0.02 0.30 test test train

131099 -0.57 0.20 test test test

132650 -0.34 1.17 test test train 135886 0.19 -0.60 test train test


206440 0.40 -0.41 test test test 217594 0.26 -0.34 test train test




479276 -1.24 0.50 test test train

493016g 0.12 1.34 test train test 493027g - - test test test

510156 0.30 -0.39 test train test

527606 -0.82 -0.48 test train train 540976 1.45 0.89 test train train


613127 0.42 -0.44 test train test 623267 -0.30 -0.51 test test train



240


634662 0.62 0.08 test train test 634913 -0.19 0.47 test train test


696446 -0.46 -0.33 test train train 732263 1.14 0.67 test train train


821954 0.30 -0.21 test train train 832699 0.07 0.56 test test train


879390 0.12 -0.81 test train train 947728 0.21 1.58 test test train

1163195 1.97 -0.81 test train test

1241947 -0.52 0.37 test train train 1544190 1.41 1.38 test train test

1570645 -0.56 -0.27 test test test

1839630 0.46 1.25 test test train 2027170 0.27 0.73 test train train


2051301 0.46 0.15 test train test

2104963 0.76 0.21 test train test


2221956 0.93 0.59 test train train 2255176 -1.89 1.90 test train train


2497065 0.49 -1.16 test train test 2980714 0.16 -0.22 test test train

3209221 0.19 0.35 test train test 3268879 1.51 -0.35 test train train


4316658 0.14 0.93 test train test 4390049 2.48 -0.67 test test train


5436431 1.62 -0.03 test train test 5510996 0.19 -0.41 test test test

5566347 1.79 0.44 test test test

5617414 0.55 -0.38 test test test 6842155 0.94 -0.43 test train train


7116963 0.72 -0.68 test test train 13150817 0.66 -0.34 test test train


13540506 0.45 0.00 test test test 14816183 0.08 0.76 test test train

15258738 -0.25 -1.12 test train train

15716082 1.01 0.46 test train train 15862074 1.83 0.01 test train train

16435497 0.54 0.10 test train test

16605917 1.40 0.75 test train train 17088221 0.51 -0.27 test test train

17301234 0.83 0.19 test test test

17312446 0.73 -0.08 test train train 18708708 0.21 0.50 test train train

21564170 -0.33 -0.68 test test test

25321099 0.69 -0.12 test train train 26087478 -0.43 -0.77 test test train

26952205 -0.24 0.70 test test train

28076735 1.42 0.86 test test train

28249776 -0.12 0.78 test test test

29446159 0.71 -0.38 test test test

30746588 0.79 0.23 test test train 31508006 2.44 -0.25 test train test

31710302 1.97 -0.16 test train test

32598111 2.22 0.43 test train train 34883415 1.27 -0.17 test train train

35065271 2.73 -0.19 test train train

36335678 -0.63 0.43 test train test 37680652 2.09 -0.22 test train train

38380051 2.41 -0.18 test train train

38380073 2.46 0.18 test train test

241

38380084 2.41 0.32 test test train

38444858 1.83 0.33 test test train 38640629 0.81 -0.08 test test test

39227582 0.87 -0.44 test train train


51630581 0.81 -0.12 test test train


57057837 0.42 -0.31 test train test

57117449 1.20 -0.20 test train train 57653857 1.44 -0.22 test train train


59261084 2.75 -0.09 test test test 60145202 2.45 0.00 test test train

60145235 2.62 -0.12 test train train

60851345 1.08 0.04 test test train 61328458 1.22 0.71 test train test



68194058 2.38 0.23 test train test

68194149 2.50 -0.27 test test train

68515480 -0.33 -0.13 test train test 71859308 1.77 0.21 test test test

74472336 2.47 0.00 test test train


84852153 0.05 -0.01 test train test 90411511 -0.47 -0.14 test train train

112281773 0.97 0.26 test test test

120068373 0.88 -1.10 test train train 125116236 0.81 -0.58 test test train

169107215 -0.11 0.56 test test train


207122154 2.27 -0.23 test test test

242

Figure 41 - Model fit and predictions for fish HLN using IFS-EPI for (A,B) 1:2, (C,D)

1:1, and (E,F) 2:1 ratios between internal training and internal validation datasets

during cross-validation.

243

Figure 42 - Model fit and predictions for fish HLN using IFS-yscr for (A,B) 1:2, (C,D)



244

Figure 43 - Model fit and predictions for log KOW using IFS-KOW for (A,B) 1:2, (C,D)



245

Table 24 - 2D Fragment Based QSAR from IFS-HLN with a 1:2 Internal Training to

Internal Validation Ratio During Cross-Validation.

Range of counts and


Fragmentd

Coefficiente S.E.f Tr.Rg. Tr.Fr. Te.Rg. Te.Fr.

O=CC(C(H)H)H a 0.588 0.428 1-2 2 1-1 1

C(H)=CH 0.297 0.387 1-3 8 1-1 4

c(Cl)c-ccH 0.282 0.216 2-4 39 2-4 22 O=CC(C(H)C(H)H)H 0.261 0.395 1-2 4 1-1 2

C(H)(CH)C(H)C 0.205 0.158 1-4 4 2-2 2

c(c(Cl)c(c)Cl)(c)Cl 0.170 0.135 1-6 18 1-6 9 c(H)(cH)ccc 0.151 0.135 2-4 35 2-4 15

c(cH)Cl 0.146 0.069 1-7 125 1-6 70

HC(H)H 0.143 0.071 1-14 205 1-12 97 c(H)(c)c-ccH 0.085 0.140 1-6 27 1-4 11

c 0.050 0.048 1-12 303 1-12 152

C(H)H 0.046 0.042 1-28 153 1-20 74

cH 0.035 0.023 1-15 303 1-15 151

c(c)(cc)Br c -0.079 0.042 2-12 5 2-20 3

c(CH)(cH)c(H)c(H)cH -0.101 0.159 1-4 8 1-2 3 O(H)c -0.128 0.227 1-2 29 1-1 11

O -0.225 0.095 1-8 104 1-8 46

{n}{c} -0.237 0.105 1-2 5 1-1 1 Hcc(c(C(H)(H)H)cH)C(H)(H)H -0.291 0.348 1-2 6 1-1 3

N(=O)=O -0.298 0.305 1-2 29 1-1 17 NH -0.331 0.302 1-2 18 1-2 7

N -0.525 0.312 1-2 18 1-2 6

c(c)(c(H)c)Occc -0.528 0.478 1-1 3 1-1 1 OC(C(H)(H)O)H -0.602 0.442 1-1 3 0-0 0

N(H)H -0.621 0.273 1-2 23 1-2 14

OH -0.676 0.257 1-3 48 1-3 19 O=C -0.709 0.153 1-4 48 1-2 14

{n}(c)cH -0.822 0.505 1-1 3 0-0 0

P=O -1.067 0.350 1-1 9 1-1 7 ClcccO


HCC(H)(H)C L.C.F. : HC(CC(C(C(H)(H)H)H)(H)H)(H)H

b 0.550 0.335 1-2 3 2-2 1

{c}(c)({c}c)c(H)cH

L.C.F. : {c}(c)(c(H)cH){c}(c)cH b 0.385 0.327 1-2 4 0-0 0



cc(ccN(=O)=O)H L.C.F. : cc(cc(N(=O)=O)cH)H

b -0.282 0.452 1-1 6 1-1 2

O=COC(C(H)(H)CH)(H)H


MW c 0.00429 0.00101 68.1-

748 413

68.1-

959.1 206

intercept

-1.028 0.206

aNon-significant fragment.

bAggregate Fragment. L.C.F. is the largest common fragment(s) among the

chemicals containing the aggregate fragment. Unique chemical is the CAS of the single chemical

containing the aggregate fragment. Testing range and frequency exclude counts forbidden by aggregate

fragment selection rules. cTesting range exceeds training range.

dFragment or molecular weight (MW) in

pseudo SMILES format. {} indicate an aromatic carbon or nitrogen bonded only to other aromatic atoms

by aromatic bonds. All hydrogens are explicit. eRegression coefficient.

fStandard Error of the regression

coefficient.

246



Range of counts and


Fragmentd


O=CC(H)(H)CH

0.897 0.418 1-2 3 1-2 3

O=COH a 0.698 0.265 2-2 2 0-0 0 C=C 0.595 0.388 1-2 7 1-1 4

c(c{c}c(H)cH)(H)cH 0.594 0.213 1-2 6 1-1 1

C(H)=CH 0.472 0.400 1-3 8 1-1 4 c(Cl)c-ccH 0.371 0.157 2-4 39 2-4 22

ccc(c(cC(H)H)H)H 0.275 0.222 1-2 3 1-2 2

c(ccc)-cccc 0.272 0.286 1-4 6 1-2 2 HC(H)H 0.236 0.039 1-14 205 1-12 97

c(H)c(H)c(OP(O)=S)c(H)cH 0.216 0.104 1-2 4 2-2 1

c(H)(cNH)c(H)cH 0.186 0.232 1-4 5 1-4 5

c(cH)Br 0.127 0.074 1-8 19 1-8 8

c(cH)Cl 0.127 0.049 1-7 125 1-6 70

Cl 0.126 0.049 1-12 165 1-10 90 c(H)(cH)cc-cc 0.121 0.059 1-8 13 1-2 6

cH 0.076 0.019 1-15 303 1-15 151

C(H)H 0.065 0.029 1-28 153 1-20 74 {c}(cH){c}{c} -0.039 0.083 1-20 20 2-18 10

O=COC(C(H)H)(H)H -0.135 0.288 1-2 8 2-2 2 c(CH)(cH)c(H)c(H)cH -0.155 0.150 1-4 8 1-2 3

HC(C(C(H)(H)H)C(H)(H)H)H c -0.179 0.141 1-6 12 1-8 7

c(H)(cc){c}{c}cH -0.182 0.179 1-4 8 1-2 3 ccc(c(C(H)(H)H)cH)H -0.204 0.153 1-4 6 1-2 8

Hccc(c(cC(H)(H)H)H)H -0.207 0.157 1-2 7 1-2 5

HCC(H)(H)O -0.220 0.197 1-3 7 1-1 4 HC(C(C(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)(H)H)(H)H)H -0.224 0.162 1-2 4 2-2 2

c(H)({c}{c})c(H)c -0.235 0.221 1-2 13 1-1 6

c(H)c-c(c)c -0.249 0.287 1-2 18 1-2 9 HCC(C(C(C(H)(H)H)(C(H)H)H)(H)H)(H)H -0.273 0.152 1-2 4 0-0 0

O -0.300 0.056 1-8 104 1-8 46

N(=O)=O -0.321 0.153 1-2 29 1-1 17 C(H)(CH)=CH -0.349 0.290 1-2 3 1-2 3

ccC(C(H)(H)H)(H)H -0.384 0.226 1-2 3 0-0 0

O=CN -0.455 0.284 1-2 10 1-1 1 ClC(C(H)(H)Cl)H -0.472 0.312 1-2 3 2-2 1

N(H)H -0.545 0.167 1-2 23 1-2 14

NH -0.564 0.212 1-2 18 1-2 7 c(c)(c(H)c)Occc -0.602 0.444 1-1 3 1-1 1

s -0.614 0.301 1-2 5 1-1 2

O=C -0.621 0.128 1-4 48 1-2 14 HC(NC(H)(H)H)(H)H -0.702 0.200 1-2 5 1-2 4

OH -0.773 0.137 1-3 48 1-3 19

O=COC(H)(H)H -0.787 0.443 1-1 3 0-0 0 P=O -1.231 0.302 1-1 9 1-1 7

ClcccO


HCC(H)(H)C

L.C.F. : HC(CC(C(C(H)(H)H)H)(H)H)(H)H b 0.824 0.321 1-2 3 2-2 1

cC(H)=C(H)H

L.C.F. : Hcc(c(C(=C(H)H)H)cH)H b 0.271 0.299 1-2 3 0-0 0

c(H)(cC)c(H)cH

L.C.F. : c(H)(c(H)cH)c(H)cC b -0.157 0.084 1-6 5 4-4 1

C(H)(Br)C(H)(Cl)C(H)Br


c(H)(cH){n} L.C.F. : c(H)({n})c(H)cH . c(H)c(H)c(H)cH

a,b -0.779 0.411 1-1 2 0-0 0

MW c 0.00378 0.00064 68.1

-748 413

68.1-

959.1

206

intercept -1.123 0.150

247










coefficient.

248



Range of counts and


Fragmentd


C=C

0.810 0.287 1-2 7 1-1 4

HCC(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)H a 0.645 0.261 1-1 2 0-0 0

O=CC(C(H)C(H)H)H 0.604 0.290 1-2 4 1-1 2 Hccc(cN(=O)=O)H 0.561 0.267 1-2 8 1-1 5

C(H)=CH 0.554 0.256 1-3 8 1-1 4

HccccN(=O)=O 0.441 0.254 1-2 3 1-1 1 Hcc(cN(=O)=O)H 0.383 0.224 1-2 23 1-2 10

c(H)(cH)cNH c 0.327 0.159 1-4 10 1-5 5

HC=C(H)H 0.320 0.165 1-2 9 1-1 7 HccOP(OC(H)(H)H)(OC(H)(H)H)=S 0.304 0.255 1-2 10 1-2 3

c(Cl)c-ccH 0.293 0.155 2-4 39 2-4 22

HC(C(C(C(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.246 0.248 1-2 17 1-2 10

c(H)(cH)c-cc(H)c 0.201 0.080 1-4 21 1-4 10

HCC(C(C(C(H)H)(H)H)(H)H)(H)H 0.197 0.117 1-4 21 1-4 11

c(cH)Br 0.193 0.107 1-8 19 1-8 8 HC(H)H 0.167 0.025 1-14 205 1-12 97

c(H)(cc)c(H)cH c 0.163 0.051 2-4 53 2-6 21

CC(H)C(H)H 0.151 0.142 1-4 5 1-2 3 HCC(C(C(H)H)H)(H)H 0.146 0.057 1-12 15 2-6 6

c(H)(ccH)ccO 0.138 0.132 1-3 15 1-2 7 c(H)(cCcc(H)cH)c(H)c 0.136 0.047 2-8 4 8-8 3

c(cH)Cl 0.097 0.037 1-7 125 1-6 70

c(cH)c-cc 0.097 0.057 1-8 23 1-8 10 c(H)cC(H)H 0.072 0.045 1-7 32 1-4 16

c(Cl)ccH c 0.062 0.047 1-5 97 1-6 49

HccC(C(H)H)H -0.129 0.100 1-4 9 2-4 4 c(H)(c(H)cH)c(H)cC=O -0.149 0.101 1-4 16 2-2 3

BrCH c -0.156 0.124 3-6 3 1-1 1

HC(C(C(H)(H)H)(C(H)(H)H)C(H)(H)H)H -0.191 0.307 1-2 6 1-2 4 Hcc(c(cC(H)(H)H)H)H c -0.199 0.213 1-2 7 1-3 5

Br c -0.225 0.094 1-7 30 1-10 12

ccc(c(C(H)(H)H)cH)H -0.232 0.125 1-4 6 1-2 8 O -0.331 0.043 1-8 104 1-8 46

O(CH)c -0.336 0.341 1-2 6 0-0 0

ccC(C(H)(H)H)(H)H -0.383 0.224 1-2 3 0-0 0 {c}({c}c){c}{c} -0.412 0.239 1-1 6 1-1 4

s -0.448 0.285 1-2 5 1-1 2

HCC(C(C(C(H)(H)H)(C(H)H)H)(H)H)(H)H -0.455 0.154 1-2 4 0-0 0 ClC(C(H)(H)Cl)H -0.489 0.274 1-2 3 2-2 1

C(H)(CH)=CH -0.510 0.233 1-2 3 1-2 3

O=C -0.538 0.100 1-4 48 1-2 14 N(H)H -0.551 0.123 1-2 23 1-2 14

Hcc(c(C(H)(H)H)cH)C(H)(H)H -0.568 0.338 1-2 6 1-1 3

c(c)(c(H)c)Occc -0.692 0.405 1-1 3 1-1 1 OH -0.744 0.095 1-3 48 1-3 19

Occ(cC(H)(H)H)H -0.783 0.283 1-1 5 1-1 3

{n}(c)cH -0.784 0.386 1-1 3 0-0 0 N -0.833 0.159 1-2 18 1-2 6

NH -0.872 0.179 1-2 18 1-2 7

N(=O)=O -0.919 0.302 1-2 29 1-1 17

P=O -1.085 0.221 1-1 9 1-1 7

Nc(c)cH

L.C.F. : c(c)(N)c(H)c b 1.215 0.507 1-1 3 0-0 0

ClcccO


HCC(H)(H)C L.C.F. : HC(CC(C(C(H)(H)H)H)(H)H)(H)H

b 0.676 0.417 1-2 3 2-2 1

{c}(cH)({c}cH)cc(H)cH

L.C.F. : {c}1(cc(H)cH){c}c(H)c(H)c(H)c1H b 0.657 0.202 1-2 7 0-0 0

c(cOc(cH)cH)(H)c

L.C.F. : c(c(H)c)(Oc(cH)c(H)cH)c(H)cH b 0.453 0.393 1-2 3 1-1 2

c(H)({c}{c}o)c(H)cH a,b 0.422 0.295 1-2 2 0-0 0

249

L.C.F. : {c}1(c(H)cH){c}(cH)o{c}{c}1c(H)c(H)cH

{c}(c)({c}c)c(H)cH L.C.F. : {c}(c)(c(H)cH){c}(c)cH

b 0.365 0.207 1-2 4 0-0 0

Hc{c}{c}c(cC(H)(H)H)H

L.C.F. : Hc{c}{c}(c(cC(H)(H)H)H)cH b 0.234 0.100 1-8 6 2-6 2

c(c-ccc)cc

L.C.F. : c(c-cc(Cl)cCl)(Cl)c(c)Cl b 0.133 0.161 1-8 11 1-4 4

Occ(c(cCC(H)(H)H)H)H L.C.F. : FC(C(Oc1c(c(c(c(c1H)H)CC(H)(H)H)H)H)(F)F)(H)F

b -0.175 0.098 4-4 3 4-4 1

HC(C(H)(H)Br)H


{c}(cH)({n}){c}cH

L.C.F. : {c}({n})({c}cH)c(H)cH b -0.351 0.132 1-2 5 1-1 1

c(H)(ccH)cO L.C.F. : c(cH)(O)c(H)ccH

b -0.396 0.300 1-2 5 1-1 2

O=COC(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H

L.C.F. : O=COC(C(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H)(H)H b -0.418 0.169 2-2 3 2-2 2

c1(H){c}({c}(cH)c(H)cc1)cH

L.C.F. : c1(H){c}({c}(cH)c(H)cc1)cH b -0.599 0.243 1-2 5 1-1 1

MW c 0.00571 0.00049 68.1-

748 413

68.1-

959.1 206











coefficient.

250

Table 27 - 2D Fragment Based QSAR from IFS-EPI with a 1:2 Internal Training to


Range of counts and


Fragmentd


c-c c 0.670 0.260 1-1 60 1-2 32

c(H)({c}{c}c(H)cCH)c(H)cH a 0.567 0.394 1-1 2 1-1 1

HC(C(C(H)(H)C)H)(H)H 0.447 0.197 1-2 5 2-2 1 C(H)=CH 0.374 0.309 1-3 6 1-1 6

HC(C(C(H)(H)H)(H)H)H 0.352 0.233 1-3 46 1-3 24

Hcc(C(C(H)(H)H)C(H)(H)H)cH c 0.312 0.305 1-3 10 1-9 4 HC(C(C(C(C(H)H)(H)H)(H)H)H)H 0.301 0.394 1-3 20 1-3 13

c(cH)(Cl)c(H)c(H)c 0.234 0.201 1-4 39 1-3 28

HCC(C(H)H)H 0.133 0.091 1-10 17 2-9 10 Cl 0.123 0.076 1-12 173 1-9 87

c(c)cH 0.111 0.049 1-8 181 1-8 93

c(cH)Br 0.039 0.029 1-8 18 2-8 9

HC(C(C(C(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.022 0.019 1-12 18 1-5 4

ClC(H)H -0.098 0.304 1-6 11 1-2 7

OC(H)H -0.127 0.116 1-16 39 1-6 19 Hccc(c(cC(H)(H)H)H)H -0.245 0.176 1-2 6 1-2 6

cc(cOC(H)(H)H)H -0.299 0.446 1-2 3 1-1 1

P(O)=O -0.309 0.129 1-3 13 2-3 4 NH -0.377 0.224 1-2 21 1-1 4

S -0.470 0.389 1-2 9 1-2 6 N(=O)=O -0.552 0.289 1-2 34 1-2 12

OH -0.607 0.169 1-3 42 1-3 25

O=C -0.684 0.154 1-4 37 1-2 25 N(H)H -0.686 0.279 1-2 24 1-2 13

c(H)({c}ccH)c(H)cH

L.C.F. : {c}({c})(ccH)c(H)c(H)cH b 0.662 0.250 1-2 9 1-2 5

cC(C(H)(H)H)c

L.C.F. : Hc2c(c(C(c1c(c(cc(c1H)H)H)H)C(H)(H)H)c(c(c2)H)H)H b 0.418 0.324 1-2 3 1-1 1

HC([Si]C(H)(H)H)(H)H L.C.F. : HC([Si](O[Si](O[Si](O[Si](C(H)(H)H)C(H)(H)H)

(C(H)(H)H)C(H)(H)H)(C(H)(H)H)C(H)(H)H)C(H)(H)H)(H)H

a,b 0.118 0.048 4-6 2 5-5 1

Hc{c}{c}c(cC(H)H)H L.C.F. : Hc1c(c({c}({c}c1H)c(cC(H)H)H)H)H

a,b -0.257 0.161 1-2 2 0-0 0

c(cH)(N)cH

L.C.F. : c(H)(c(N)c(H)cH)cH b -0.286 0.230 1-2 4 1-2 2

{c}({c}cH){n}cH

L.C.F. : {c}({c}cH)({n}cH)c(H)cH b -0.342 0.355 1-2 3 0-0 0

MW 0.00272 0.00078 68.1-959.1

417 74.1-748

211











coefficient.

251



Range of counts and


Fragmentd


c(c{c}c(H)cH)(H)cH 0.985 0.389 1-2 4 1-1 3

C=C 0.775 0.350 1-2 7 1-1 7

C(H)=CH 0.483 0.260 1-3 6 1-1 6 [Si] c 0.422 0.288 1-6 3 1-7 3

c(-c)(cCl)c(H)cH 0.419 0.254 1-2 21 1-1 7

O=CCH 0.397 0.500 1-2 9 1-2 8 HC(C(C(H)(H)H)(H)H)H 0.313 0.195 1-3 46 1-3 24

C 0.282 0.063 1-10 54 1-4 36

c({c}{c}c(H)cH)(H)cH 0.279 0.241 1-5 22 1-3 11 c 0.259 0.022 1-12 307 1-11 149

c(-c(cH)cH)c 0.251 0.202 1-2 14 1-2 14

CH 0.240 0.049 1-17 92 1-17 63

c(H)({c}c(H){c}c(H)cH)c(H)cH 0.198 0.158 1-4 5 1-1 1

c(H)(cH)cc-cc 0.165 0.082 1-8 15 1-4 4

c(cH)(Br)cH 0.138 0.078 1-3 10 1-3 6 c(cH)Cl 0.126 0.035 1-7 129 1-6 66

C(H)H 0.061 0.039 1-28 147 1-16 86

OC(H)H -0.077 0.079 1-16 39 1-6 19 Brc(c)c -0.082 0.033 1-10 6 2-5 3

{n} -0.207 0.203 1-3 16 1-2 8 ClC(C(H)(H)Cl)H -0.230 0.160 1-2 3 2-2 1

N#C -0.275 0.259 1-2 8 1-2 6

c(cOH)Cl -0.277 0.211 1-2 4 2-2 1 {c}({c}){c}c -0.280 0.347 1-2 7 1-1 4

C(H)(O)C(H)CH -0.287 0.159 2-7 4 2-7 3

Hccc(c(cC(H)(H)H)H)H -0.355 0.164 1-2 6 1-2 6 O=CO -0.379 0.244 1-2 18 1-2 13

Hcc(cc(C(H)(H)H)cH)C(H)(H)H a -0.429 0.263 1-1 2 0-0 0

NH -0.449 0.172 1-2 21 1-1 4 N(=O)=O -0.451 0.174 1-2 34 1-2 12

c(H)(cc){c}{c}c(H)cH -0.495 0.367 1-2 3 1-2 4

O=C -0.502 0.145 1-4 37 1-2 25 S(cc)=O a -0.519 0.345 1-2 2 0-0 0

HcccOC(H)(H)H -0.551 0.385 1-2 5 0-0 0

N -0.561 0.198 1-2 17 1-2 7 c(H)(cH)cOCH -0.602 0.314 1-2 4 2-2 1

OH -0.738 0.119 1-3 42 1-3 25

N(H)H -0.774 0.202 1-2 24 1-2 13 P=O -0.922 0.229 1-1 13 1-1 4

cc(C(H)H)c c -0.942 0.616 1-1 3 1-2 2

Nc(c)cH L.C.F. : c(c)(N)c(H)c

b 0.931 0.481 1-1 3 0-0 0

HCC(CC(H)(H)CH)(H)H

Unique Chemical : 50876329 a,b 0.815 0.367 1-1 1 0-0 0

cccOC(H)(H)H

L.C.F. : Clcc(c(cOC(H)(H)H)Cl)Cl b 0.632 0.385 1-2 3 0-0 0

cCC L.C.F. : c(H)c(H)c(CC)c(H)cH

a,b 0.456 0.278 2-2 2 0-0 0

c1(H)c(H)c(H)c(C)c(H)c1H

L.C.F. : c1(H)c(H)c(H)c(C)c(H)c1H b -0.306 0.194 1-3 4 0-0 0

ccccN(=O)=O

L.C.F. : Clcc(c(c(cN(=O)=O)Cl)Cl)Cl a,b -0.596 0.342 1-2 2 0-0 0









252



coefficient.

253



Range of counts and


Fragmentd


C=C 1.096 0.415 1-2 7 1-1 7

[Si] c 0.749 0.274 1-6 3 1-7 3

C(=O)C(H)CH 0.565 0.287 1-2 6 1-2 5 C(H)=CH 0.512 0.223 1-3 6 1-1 6

c(H)(cc-ccc(H)cH)cH 0.468 0.204 1-4 5 1-1 1

S=PO 0.358 0.071 2-3 21 2-3 10 HC=C(H)H 0.350 0.162 1-2 12 1-1 4

C 0.258 0.080 1-10 54 1-4 36

HC(C(C(C(C(H)H)H)(H)H)H)H 0.251 0.153 1-5 7 1-5 5 c(Cl)c-ccH 0.242 0.116 2-4 41 2-4 20

HC(C(C(H)(H)H)(H)H)H 0.241 0.127 1-3 46 1-3 24

c(H)(cH)c(H){c} c 0.231 0.058 1-6 37 1-8 23

c 0.208 0.021 1-12 307 1-11 149

CH 0.198 0.037 1-17 92 1-17 63

c(cH)Cl 0.191 0.030 1-7 129 1-6 66 c(c)(Cl)cc(H)cH 0.175 0.085 1-4 24 1-2 9

HC(C(C(H)(H)H)H)(H)H 0.172 0.111 1-3 31 1-2 17

HC(CC(H)(H)H)(H)H 0.141 0.053 1-9 30 1-9 15 c(cH)Br 0.120 0.032 1-8 18 2-8 9

C(H)H 0.073 0.025 1-28 147 1-16 86 c(Cl)(ccCl)c(c)Cl 0.063 0.030 1-12 24 1-7 11

HC(C(C(C(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.048 0.028 1-12 18 1-5 4

{n} -0.209 0.144 1-3 16 1-2 8 S -0.234 0.188 1-2 9 1-2 6

NH -0.249 0.131 1-2 21 1-1 4

c(H)(ccH)ccOH -0.263 0.208 1-4 9 1-2 4 O -0.267 0.046 1-8 105 1-8 47

Hccc(c(cC(H)(H)H)H)H -0.288 0.162 1-2 6 1-2 6

c1(-ccH)cc(H)c(H)c(H)c1H c -0.293 0.332 1-2 4 1-4 4 N(=O)=O -0.383 0.121 1-2 34 1-2 12

O=C(C(H)H)O -0.456 0.410 1-2 3 1-1 4

O=CN -0.470 0.258 1-2 7 1-2 4 O=C -0.475 0.091 1-4 37 1-2 25

N(H)H -0.567 0.133 1-2 24 1-2 13

OH -0.742 0.102 1-3 42 1-3 25 cc(C(H)H)c c -0.924 0.402 1-1 3 1-2 2

HCC(H)(c)c

L.C.F. : c2(H)cc(H)c(H)c(C(H)(CH)c1c(H)c(H)cc(H)c1H)c2H a,b 0.890 0.458 1-1 2 0-0 0

cccOC(H)(H)H

L.C.F. : Clcc(c(cOC(H)(H)H)Cl)Cl b 0.824 0.304 1-2 3 0-0 0

c1(H)c(H)c(Cc)c(H)c(H)c1 L.C.F. : c2(H)cc(H)c(H)c(Cc1c(H)c(H)cc(H)c1H)c2H

b 0.560 0.171 2-2 5 2-2 1

c1(H)c(H)c(H)c(-c)c(H)c1H

L.C.F. : c1(H)c(H)c(H)c(-c)c(H)c1H b 0.302 0.308 1-1 9 1-1 3

c(cH)(ccH)OccH

L.C.F. : c(cH)(cc(H)c)OccH b 0.174 0.145 1-2 11 1-2 4

c(cH)(N)cH L.C.F. : c(H)(c(N)c(H)cH)cH

b -0.316 0.201 1-2 4 1-2 2

{c}(cH)({n}){c}cH

L.C.F. : {c}({n})({c}cH)c(H)cH b -0.357 0.327 1-2 5 1-1 1

BrccC(H)(H)H

L.C.F. : Brcc(cC(H)(H)H)Br a,b -0.383 0.184 1-2 2 0-0 0

C(Oc(cH)cH)H L.C.F. : Hcc(c(OC(C(H)(H)H)H)cH)H

b -0.437 0.424 1-1 3 1-1 1

Hccc(cN(H)H)H

L.C.F. : Hcc(cc(cN(H)H)H)H b -0.497 0.383 1-1 3 1-1 2

ClC(C(=C(H)H)H)H


S=O L.C.F. : c(H)(cc)ccH

a,b -0.894 0.451 1-1 2 0-0 0


254










coefficient.

255

Table 30 - 2D Fragment Based QSAR from IFS-yscr with a 1:2 Internal Training to


Range of counts and


Fragmentd


c(H)(c(H)c(H)cCl)cCl 1.519 0.962 1-1 5 1-1 2

c(H)(c)c(cOc(cH)cH)H 1.454 0.916 1-2 4 1-1 1

O=CC(H)c 1.094 0.783 1-1 3 0-0 0 c(H)(c-cc(H)cCl)c(H)c c 0.703 0.398 1-2 13 1-4 4

C(NC(H)H)=O 0.538 0.636 1-2 4 2-2 1

c(c)(ccH)-ccccH 0.484 0.709 1-2 4 0-0 0 HCC(C(H)(C(H)H)H)(H)H 0.206 0.199 1-4 33 1-4 18

c(Cl)(c(H)c)c(c)Cl -0.409 0.501 1-3 23 1-3 15

c(Cl)(c(H)cH)c(c)Cl -0.920 0.652 1-2 22 1-2 10 c(cH)(cc)-ccc

L.C.F. : c(Cl)(cCl)c(cH)-cc(Cl)cCl b 0.593 0.454 1-2 9 1-2 5

c(c)(-cc)c(H)c

L.C.F. : c(-ccCl)(cCl)c(H)cCl b 0.376 0.394 1-2 17 1-2 6

intercept 0.395 0.108










coefficient.

256



Range of counts and


Fragmentd


c(H)(c(H)c(H)cCl)cCl - 1.548 0.753 1-1 5 1-1 2

c(H)(c)c(cOc(cH)cH)H 1.446 0.720 1-2 4 1-1 1

c(OH)(c(H)c)c(H)cH 1.219 0.747 1-1 3 1-1 3 c(OC(H)H)c 0.977 0.546 1-2 3 0-0 0

NC(H)C(H)H 0.878 0.496 1-2 4 1-1 2

HCC(C(C(H)(H)H)(C(H)(H)H)H)(H)H 0.821 0.718 1-1 4 1-1 1 c(H)(cH){c}{c}cc 0.804 0.440 1-2 3 0-0 0

c(H)(c-cc(H)cCl)c(H)c c 0.718 0.386 1-2 13 1-4 4

C(NC(H)H)=O 0.670 0.579 1-2 4 2-2 1 Sc 0.666 0.582 1-2 3 1-1 1

c(cH)(Cl)c(H)ccOH 0.614 0.419 1-2 3 1-1 1

c(cH)C(H)CH 0.435 0.232 1-4 5 4-4 1

c(cH)(-cc(H)cH)c(H)cc 0.281 0.410 1-4 6 1-4 4

HC(C(C(C(C(H)H)(H)H)(H)H)H)H 0.237 0.257 1-3 22 1-3 10

c(cH)(C(=O)ccH)c(cH)C=O -0.311 0.488 1-4 5 1-4 2 c(Cl)(c(H)c)c(c)Cl -0.463 0.430 1-3 23 1-3 15

c(H){c}1{c}c(H)c(H)c(H)c1 -0.703 0.433 1-1 3 1-1 3

c(Cl)(c(H)cH)c(c)Cl -0.959 0.480 1-2 22 1-2 10 S=O

L.C.F. : c(H)(cc)ccH a,b 1.445 0.890 1-1 2 0-0 0

Hcc(c(C(C(H)(H)H)(H)H)c(H)cH)H

L.C.F. : Hc1c(c(C(C(H)(H)H)(H)H)c(c(c1)H)H)H a,b 1.432 0.889 1-1 2 0-0 0

c(H)(cCl)c(H){c} L.C.F. : c(H)(cCl)c(H){c}{c}o{c}

b 0.972 0.468 1-2 4 0-0 0

c(cH)(cc)-ccc

L.C.F. : c(Cl)(cCl)c(cH)-cc(Cl)cCl b 0.661 0.404 1-2 9 1-2 5

c(c)(-ccc)ccH

L.C.F. : c(cCl)(-cc(Cl)cCl)c(cH)Cl b 0.503 0.588 1-2 7 1-2 2

c(c)(-cc)c(H)c L.C.F. : c(-ccCl)(cCl)c(H)cCl

b 0.397 0.331 1-2 17 1-2 6

P(=S)Oc(c)c(H)cH

L.C.F. : P(O)(=S)Oc(ccH)c(H)cH b -1.051 0.698 1-1 4 0-0 0











coefficient.

257



Range of counts and number of

chemicals containing the fragment

Fragmentd


c(OH)(c(H)c)c(H)cH 2.077 0.657 1-1 3 1-1 3

c(H)(c)c(cOc(cH)cH)H 1.845 0.627 1-2 4 1-1 1

c(H)(c(H)c(H)cCl)cCl 1.638 0.573 1-1 5 1-1 2 Clcc(c(c(N(=O)=O)c)H)H a 1.630 0.774 1-1 2 1-1 1

HCC(C(C(H)(H)H)(C(H)(H)H)H)(H)H 1.392 0.616 1-1 4 1-1 1

cc(cc(N(H)H)cH)H 1.381 0.646 1-2 5 1-1 1 c(OC(H)H)c 1.190 0.485 1-2 3 0-0 0

NC(H)C(H)H 1.135 0.409 1-2 4 1-1 2

Hcc(c(C(H)(H)H)cH)C(H)(H)H 1.104 0.394 1-2 6 1-2 3 HC(C(C(C(C(H)(H)H)(H)H)(C(H)(H)H)H)(H)H)H 1.048 0.425 1-1 3 1-1 1

c(H)(cH){c}{c}cc 1.041 0.387 1-2 3 0-0 0

c(cH)(Cl)c(H)ccOH 0.989 0.377 1-2 3 1-1 1

c(H)(c-cc(H)cCl)c(H)c c 0.941 0.338 1-2 13 1-4 4

HCC(C(H)(H)H)(C(H)(H)H)H c 0.905 0.811 1-1 8 2-2 1

Sc 0.893 0.414 1-2 3 1-1 1 HC(=CC(C(C(C(C(C(C(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.876 0.418 1-1 3 1-1 1

C(=C(H)H)Cl a 0.870 0.353 2-2 1 0-0 0

Hcc(c(C(C(H)(H)H)(H)H)cH)H c 0.810 0.276 2-2 4 4-4 1 C(H)(H)OH c 0.773 0.506 1-1 5 1-2 3

c(C(H)C=O)cH 0.740 0.593 1-2 3 0-0 0 c(H)cc(H){c}cH 0.737 0.346 1-2 3 1-1 2

C(NC(H)H)=O 0.731 0.405 1-2 4 2-2 1

c(cH)(-cc(H)cH)c(H)cc 0.638 0.379 1-4 6 1-4 4 c(cH)(cc)-c(c)cc 0.616 0.281 1-2 5 1-2 3

c(cH)(-cc(H)cH)cc(H)cH c 0.519 0.318 1-2 8 2-4 3

ClC(C(C(H)(H)Cl)H)(H)H a 0.507 0.224 1-3 2 0-0 0 {c}(cH)({c}c)c(H)cH 0.475 0.103 1-4 10 1-4 7

c(H)(cH)cC(H)CH 0.428 0.215 2-4 3 4-4 1

HC(C(C(C(C(H)H)(H)H)(H)H)H)H 0.279 0.158 1-3 22 1-3 10 C 0.141 0.085 1-10 61 1-4 26

c(cH)(C(=O)ccH)c(cH)C=O -0.266 0.315 1-4 5 1-4 2

c(H)(cOH)c(H)c -0.316 0.200 1-4 10 1-2 7 c(H)(cBr)c(H)cH -0.381 0.467 1-4 4 2-2 1

c(-c(cH)cH)c -0.423 0.278 1-2 19 1-2 9

c(c-c(cH)cH)(H)c -0.521 0.317 1-3 13 1-2 8 cc(c(c(C(H)(H)H)c)H)H -0.658 0.288 1-2 6 1-2 4

O=CC(C(H)H)H a -0.787 0.562 1-2 2 1-1 1

c(H)cOc(c(H)cH)c(H)cH c -1.026 0.568 1-2 5 1-4 4 c(H)(c)c(H){c} -1.119 0.395 1-1 5 1-1 2

c(Cl)(c(H)cH)c(Cl)cCl -1.173 0.600 1-2 10 1-2 4

HCC(C(C(CH)(CH)H)(H)H)(H)H -1.187 0.717 1-1 3 1-1 1 Hcc(c(c(H)cH)N(H)H)Cl -1.248 0.681 1-2 3 1-1 1

c(H)(cc-c(c)cH)c(H)c -1.393 0.586 1-1 7 1-1 2


a,b 1.734 0.744 1-1 2 0-0 0

c(H)(cCl)c(H){c}

L.C.F. : c(H)(cCl)c(H){c}{c}o{c} b 1.226 0.377 1-2 4 0-0 0

c(H)(cH)cOccCl

L.C.F. : c(Cl)(cOcc(H)c(H)ccH)c(H)c b 1.092 0.432 1-2 5 0-0 0

c(H)(ccH)cO

L.C.F. : c(cH)(O)c(H)ccH b 0.944 0.518 1-2 6 1-1 1

HC(C(OP(OC(C(H)(H)H)(H)H)=S)(H)H)(H)H

L.C.F. : HC(C(OP(OC(C(H)(H)H)(H)H)=S)(H)H)(H)H b 0.886 0.539 1-1 5 1-1 3

cc(ccC(H)H)H

L.C.F. : Hccc(ccC(C(H)H)(H)H)H b 0.868 0.543 1-2 3 2-2 1

BrccC(H)(H)H L.C.F. : Brcc(cC(H)(H)H)Br

a,b 0.773 0.554 1-2 2 0-0 0

Hcc({c}c(c({c}({c}){c})H)H)H

L.C.F. : {c}32c(H)c(H){c}(cH){c}(cH){c}3c(H)c(H){c}1c(H)c(H)c(H)c(H){c}21

a,b 0.641 0.312 1-2 2 0-0 0

c(c)(-cc)c(H)c

L.C.F. : c(-ccCl)(cCl)c(H)cCl b 0.546 0.219 1-2 17 1-2 6

258

c(H)(cc)c(H)c-cc

L.C.F. : c(H)(c(H)c-ccCl)ccCl b 0.318 0.346 1-2 11 1-2 6

c(O)(cCl)c(H)cH

L.C.F. : c(O)(cCl)c(H)c(H)c b -0.604 0.646 1-2 6 1-1 2

c(H)(cOP=S)c(H)cH L.C.F. : P(O)(=S)Occ(H)c(H)c(H)cH

b -0.606 0.297 1-2 3 0-0 0

cc(c(cC(C(H)H)(C(H)H)H)H)H

L.C.F. : Hccc(c(cC(C(C(H)H)(H)H)(C(H)H)H)H)H a,b -0.766 0.551 1-2 2 0-0 0

c(H)({c})c-c

L.C.F. : {c}21c(H)c(H)c(H)c(H){c}2c(H)cc(-cc(H)c(H)cH)c1H a,b -0.868 0.339 1-2 2 0-0 0

HccN(C(H)H)H L.C.F. : Hcc(c(cN(C(C(H)(H)H)(H)H)H)H)H

a,b -0.893 0.567 1-2 2 0-0 0

c(H)(c(H)c(H){c})c

L.C.F. : c1(H)c{c}{c}c(H)c1H b -1.343 0.360 1-2 7 1-2 4











coefficient.

259

Table 33 - 2D Fragment Based QSAR from IFS-KOW with a 1:2 Internal Training to


Range of counts and


Fragmentd


HC=C(H)H 0.471 0.280 1-2 9 1-1 7

C(H)=CH 0.349 0.244 1-3 8 1-1 4

HC(C(C(H)(H)H)(C(H)(H)H)C(H)(H)H)H 0.348 0.234 1-2 6 1-2 4 c 0.336 0.058 1-12 303 1-12 152

HC(C(C(H)(H)H)(C(H)(H)H)H)H 0.293 0.183 1-2 14 1-1 6

HC(H)H 0.288 0.044 1-14 205 1-12 97 HCC(OC(H)H)=O 0.264 0.430 1-1 4 1-1 1

C(H)H 0.225 0.044 1-28 153 1-20 74

HC(C(C(H)(H)H)(H)H)H 0.194 0.267 1-3 43 1-3 26 c(O)(cH)c(H)c(H)cH c 0.186 0.125 1-4 8 3-6 4

{c}({c})cH 0.154 0.049 1-16 54 2-12 25

HC(C(C(C(H)(C(H)H)H)(H)CH)(H)H)H 0.141 0.300 1-2 5 1-2 2

c(OH)c(H)c 0.140 0.094 1-2 8 1-1 2

c(H)c(H)c(H)cH 0.085 0.054 1-6 108 1-6 51

HccC(H)(H)H -0.082 0.079 1-6 39 1-4 21 HC(CC(H)H)H -0.143 0.314 1-6 9 1-1 3

ClC(C(H)(H)Cl)H -0.147 0.131 1-2 3 2-2 1

C(H)(O)C(H)CH -0.168 0.207 2-7 4 3-7 2 cc -0.198 0.052 1-12 199 1-12 95

Br c -0.414 0.082 1-7 30 1-10 12 O=CC(H)H -0.519 0.308 1-4 9 1-2 5

O -0.616 0.079 1-8 104 1-8 46

O=C -0.668 0.125 1-4 48 1-2 14 HC(OC(H)H)H -0.708 0.308 1-8 7 3-3 1

O=CC(H)(H)H -0.814 0.498 1-2 3 0-0 0

HNC(H)(H)H -0.863 0.325 1-1 5 1-1 2 N(=O)=O -0.906 0.229 1-2 29 1-1 17

{n} -0.937 0.199 1-3 18 1-3 6

OH -0.972 0.158 1-3 48 1-3 19 N(H)H -1.042 0.256 1-2 23 1-2 14

N -1.220 0.221 1-2 18 1-2 6

P=O -1.567 0.365 1-1 9 1-1 7 C(H)(C)(CH)CH

L.C.F. : HCC1(C2(C(C(C1H)(C(=C2Cl)Cl)Cl)(Cl)Cl)Cl)H b 0.107 0.203 1-2 3 1-2 2

n L.C.F. : c(cH)Cl . c(cH)cH

b -0.385 0.413 1-1 4 1-1 3



S=O

L.C.F. : c(H)(cc)ccH a,b -0.846 0.261 1-1 1 0-0 0

S(=O)=O L.C.F. : c(H)c . O=S(c)=O

b -1.764 0.475 1-1 4 1-1 3

MW c 0.01220 0.00120 68.1-

748 413

68.1-

959.1 206











coefficient.

260



Range of counts and


Fragmentd


C(H)=CH 0.505 0.199 1-3 8 1-1 4

HCC(C(C(C(C(C(H)H)(H)H)H)(H)H)(H)H)(H)H 0.471 0.164 1-2 3 1-1 3 HC=C(H)H 0.469 0.149 1-2 9 1-1 7

C=C(H)H 0.383 0.222 1-1 9 1-1 3

O=C(N)NH 0.329 0.367 1-1 3 0-0 0 HC(H)H 0.328 0.026 1-14 205 1-12 97

N(H)(H)cc 0.313 0.182 1-2 10 1-2 10

c1(H)c(H)c(OP=S)c(H)c(H)c1 0.293 0.163 1-1 3 1-1 1 c 0.277 0.030 1-12 303 1-12 152

c(H)(cC)cc 0.269 0.134 1-2 8 1-2 3

HC(C(C(H)(H)H)(C(H)(H)H)H)H 0.267 0.140 1-2 14 1-1 6

c(OH)c(H)c 0.259 0.085 1-2 8 1-1 2

C(H)H 0.242 0.021 1-28 153 1-20 74

HC(C(C(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.229 0.242 1-2 20 1-2 15 C(H)=C 0.223 0.173 1-2 11 1-2 4

HC(C(C(C(H)(C(H)H)H)(H)CH)(H)H)H 0.220 0.228 1-2 5 1-2 2

c(H)(cO)c(H)cH c 0.165 0.155 1-5 15 3-6 5 c(H)(c(-c)c(H)cH)cH 0.164 0.117 1-2 17 1-2 8

O(ccH)ccH 0.143 0.127 1-4 12 1-4 9 {c}({c})cH 0.137 0.030 1-16 54 2-12 25

Hcc(cC(H)(H)H)H c 0.125 0.166 1-2 20 1-3 15

ClC=C 0.080 0.036 2-6 6 2-4 4 cH 0.051 0.014 1-15 303 1-15 151

Brcc c -0.066 0.056 1-12 16 1-20 9

c(H)(c)ccc -0.074 0.055 2-4 46 2-4 19 HC(CC(H)H)H -0.152 0.164 1-6 9 1-1 3

ClC(H)H -0.173 0.197 1-6 11 1-3 7

NCH -0.185 0.255 1-1 4 1-1 2 cc(C(H)(H)H)c -0.204 0.142 1-2 5 1-1 1

Hcc(C(H)(H)H)cH -0.214 0.183 1-3 14 1-1 12

c(H)(cOcc)cccH -0.216 0.160 1-2 8 1-2 4 HC(C(C(C(C(H)(H)H)(C(H)H)H)(H)H)(H)H)H -0.223 0.098 1-3 6 1-2 5

HcccC(H)(H)H -0.233 0.091 1-4 24 1-3 12

HC(C(H)(H)O)(H)H -0.251 0.113 1-3 15 1-2 7 Hcc(OC(H)(H)H)c -0.349 0.279 1-2 4 1-1 1

Hcc(OC(C(H)(H)H)H)cH -0.368 0.290 1-2 4 0-0 0

Br c -0.380 0.064 1-7 30 1-10 12 O=C -0.604 0.075 1-4 48 1-2 14

O -0.612 0.055 1-8 104 1-8 46

HC(OC(H)H)H -0.616 0.188 1-8 7 3-3 1 N(H)(C=O)cc(H)cH -0.622 0.394 1-2 5 0-0 0

O=CC(H)H -0.739 0.198 1-4 9 1-2 5

N(=O)=O -0.877 0.118 1-2 29 1-1 17 {n} -0.906 0.120 1-3 18 1-3 6

OH -1.032 0.100 1-3 48 1-3 19

N#C -1.100 0.289 1-2 6 1-2 8 P=O -1.171 0.245 1-1 9 1-1 7

HNC(H)(H)H -1.199 0.255 1-1 5 1-1 2

N(H)H -1.250 0.164 1-2 23 1-2 14

N -1.343 0.151 1-2 18 1-2 6

P(=S)Occc(H)cCl

L.C.F. : Clcc(ccOP(O)(O)=S)H b 0.611 0.295 1-1 3 0-0 0

c(Cl)(c-ccH)c(H)c

L.C.F. : c(c-ccH)(Cl)c(H)cCl b -0.068 0.052 1-2 12 1-2 7

c(H)(c)c-ccc L.C.F. : c(Cl)(c-cc(H)cCl)cCl

b -0.160 0.280 1-4 17 1-2 7

c(c)c-cc

L.C.F. : c(c-ccCl)(Cl)cCl b -0.227 0.126 1-8 25 1-8 10

c(C)c

L.C.F. : c(C)(c)cH b -0.241 0.131 1-2 12 1-2 3

c(Cl)cCH b -0.465 0.222 1-1 4 1-1 1

261

L.C.F. : c(Cl)(cH)c(CH)c(H)cH

O=CH Unique Chemical : 104881

a,b -0.532 0.237 1-1 1 0-0 0



C(=O)C(H)C(H)O

L.C.F. : O=CC(C(C(H)H)(H)O)H b -1.044 0.559 1-1 3 1-1 2


a,b -1.273 0.247 1-1 1 0-0 0

S(=O)=O

L.C.F. : c(H)c . O=S(c)=O b -1.902 0.352 1-1 4 1-1 3

MW c 0.01200 0.00070 68.1-

748 413

68.1-

959.1 206











coefficient.

262


Internal Validation Ratio during Cross-Validation.

Range of counts and


Fragmentd


O=C(N)NH 0.772 0.313 1-1 3 0-0 0

{n}(c)c 0.510 0.180 1-3 4 1-3 2

HC=C(H)H 0.471 0.108 1-2 9 1-1 7 C=C 0.467 0.221 1-2 7 1-1 4

C(H)=CH 0.416 0.157 1-3 8 1-1 4

C=C(H)H 0.390 0.154 1-1 9 1-1 3 c1(H)c(H)c(H)c(O)c(H)c1H c 0.379 0.253 1-2 6 1-3 4

c(OH)c(H)c 0.364 0.088 1-2 8 1-1 2

N(H)(H)cc 0.308 0.140 1-2 10 1-2 10 C(H)=C 0.287 0.132 1-2 11 1-2 4

HC(C(C(C(C(C(C(H)(H)H)(H)H)(H)H)(H)H)(H)H)(H)H)H 0.260 0.196 1-2 20 1-2 15

HC(H)H 0.254 0.020 1-14 205 1-12 97

C(H)H 0.246 0.017 1-28 153 1-20 74

c(H)(cH)cOcc(H)c 0.239 0.166 1-2 6 1-2 5

c 0.220 0.019 1-12 303 1-12 152 cccN(=O)=O 0.219 0.191 1-2 4 1-2 3

ccc(cC(C(H)(H)H)C(H)(H)H)H 0.192 0.083 2-6 4 6-6 1

c(cH)(ccH)C(=O)c(c)cH 0.187 0.103 1-4 4 1-4 2 HCC(C(H)(H)H)C(H)(H)H 0.175 0.134 1-3 7 2-2 1

Hcc(c(C(C(H)(H)H)C(H)(H)H)c(cH)H)H 0.172 0.063 1-3 6 2-3 3 {c} 0.159 0.031 2-8 57 2-8 26

c(H)(ccOH)cCl 0.142 0.088 1-2 4 1-1 1

HCC(H)(H)H 0.135 0.031 1-6 55 1-4 26 HC(C(C(C(H)H)(H)H)H)H c 0.111 0.116 1-3 24 1-6 15

HC(C(C(H)(H)H)C(H)(H)H)H c 0.100 0.059 1-6 12 1-8 7

c(H)(cH)c-c 0.085 0.037 1-6 46 1-4 22 cH 0.068 0.010 1-15 303 1-15 151

{c}({c})({c}cH)c(H)cH c 0.064 0.059 1-10 14 3-12 6

c(H)(c)c(cCc(cH)cH)H -0.104 0.062 2-4 4 4-4 3 BrC(H)(H)C -0.135 0.189 1-4 3 0-0 0

HC(C(H)(c)C(H)H)H -0.152 0.133 1-2 3 1-1 2

c(H)cOC(H)H -0.156 0.117 1-2 5 0-0 0 ClC(H)H -0.171 0.169 1-6 11 1-3 7

cc(c(ccC(H)(H)H)H)H -0.174 0.189 1-2 10 1-1 4

OCH -0.200 0.065 1-12 18 1-12 6 HCC(C(C(H)(H)H)C(H)H)H -0.230 0.129 1-3 3 0-0 0

Occ(cC(H)(H)H)H -0.238 0.279 1-1 5 1-1 3

NC(H)C(H)H -0.242 0.143 1-2 4 1-1 2 HC(C(H)(H)O)(H)H -0.295 0.087 1-3 15 1-2 7

Br c -0.378 0.038 1-7 30 1-10 12

HNC(H)(H)H -0.461 0.243 1-1 5 1-1 2 O -0.484 0.044 1-8 104 1-8 46

O=CC(H)H -0.489 0.147 1-4 9 1-2 5

Hcc(OC(H)(H)H)c -0.495 0.189 1-2 4 1-1 1 C(H)(H)OH -0.516 0.193 1-2 6 1-1 2

HC(OC(H)H)H -0.660 0.119 1-8 7 3-3 1

NH -0.735 0.163 1-2 18 1-2 7 O=C -0.739 0.066 1-4 48 1-2 14

OH -0.903 0.087 1-3 48 1-3 19

N(=O)=O -0.906 0.098 1-2 29 1-1 17

N#C -0.933 0.168 1-2 6 1-2 8

{n} -1.006 0.087 1-3 18 1-3 6

HC(C(C(C(C(C(C(H)(H)O)(H)H)(H)H)(H)H)(H)H)(H)H)H -1.204 0.212 1-2 4 2-2 2 N -1.214 0.124 1-2 18 1-2 6

N(H)H -1.220 0.114 1-2 23 1-2 14

P=O -1.284 0.186 1-1 9 1-1 7 c1(H)cccc(H)c1C

L.C.F. : c1(H)cccc(H)c1C b 0.721 0.404 1-1 4 1-1 2

P(=S)Occc(H)cCl L.C.F. : Clcc(ccOP(O)(O)=S)H

b 0.495 0.271 1-1 3 0-0 0

N(=C)Nc

L.C.F. : N(ccH)(C=O)N=C . N(ccH)N=CO a,b 0.439 0.323 1-1 2 0-0 0

263

S=POcc(c(cN(=O)=O)H)H

L.C.F. : S=P(Oc(c(c(cN(=O)=O)H)H)cH)O b 0.189 0.094 1-2 4 2-2 1

{c}(cH)({c}c(H)c)o{c}cH

L.C.F. : {c}(cH)({c}c(H)cCl)o{c}c(H)cCl b 0.100 0.057 1-4 4 0-0 0

HC(cc(cC(C(H)(H)H)C(H)(H)H)H)(H)H L.C.F. : HC(cc(c(C(C(H)(H)H)C(H)(H)H)c)H)(H)H

a,b -0.110 0.046 2-6 2 0-0 0

c(c)c-cc

L.C.F. : c(c-ccCl)(Cl)cCl b -0.154 0.036 1-8 25 1-8 10

c(H)(cc)cOccBr

L.C.F. : c(Br)(c(H)cOccBr)c(cH)Br b -0.316 0.202 1-2 4 1-2 3

Hcc(ccC(F)(F)F)H L.C.F. : Hcc(cc(C(F)(F)F)cH)H

a,b -0.498 0.290 1-1 2 0-0 0

O=CH


S=O

L.C.F. : c(H)(cc)ccH a,b -1.578 0.318 1-1 1 0-0 0

S(=O)=O L.C.F. : c(H)c . O=S(c)=O

b -1.961 0.327 1-1 4 1-1 3

MW c 0.01170 0.00050 68.1-

748 413

68.1-

959.1 206











coefficient.

264

Chapter 6

Conclusions

265

1 Summary of Major Conclusions Chapter 2 of this thesis presented the results from a large-scale chemical hazard assessment

which used a dualistic paradigm to screen a large list of chemicals for the potential to be

Arctic contaminants. Over 100,000 chemicals from the database of the EPI Suite software

package41

were screened first using a structural profile of known Arctic contaminants and

also with hazard metrics similar to those suggested by the Stockholm Convention.5

Chemicals were additionally screened using presence on a high-production volume list as an

additional hazard metric for human exposure potential and a list of 120 candidate Arctic

contaminants was presented. Differences in the results of the two screening paradigms were

compared and three classes of chemicals were identified that have a high structural

resemblance to POPs, but do not have the correct chemical properties to become Arctic

contaminants; these are highly volatile chemicals, highly water soluble chemicals and

particle-bound chemicals.

Chapter 3 presented the implementation of a more sophisticated and mechanistically

satisfying description of the sorption of chemicals to environmental organic matter phases in

the multimedia mass balance model CoZMo-POP2.128

Equations and methods were derived

and presented to allow for similar implementations in other mass balance models. A

sensitivity analysis using hypothetical persistent chemicals showed that differences in the

model outputs were relatively small in comparison to the model‘s sensitivity to other

environmental parameters such as rainfall rate and temperature. The relationship between

environmental fate and the various types of interactions that a chemical undergoes with

environmental media were also explored. It was found that non-specific van der Waals

interactions had the greatest controlling effect on environmental fate, and that this interaction

was highly correlated with sorption to octanol, which explains why using octanol as a

surrogate for organic matter is an adequate approximation in a regional scale multimedia

mass balance model for the purposes of calculating chemical hazard and risk assessment

metrics.

266

Chapter 4 explored the effect of increasing the time resolution of major environmental

parameters and the implementation of dynamically changing mass balances of air, organic

aerosol, water and organic carbon in the CoZMoMAN model.34

Methods for the

implementation of dynamic mass balances of environmentlly relevant phases were presented,

along with the creation of a consistent climatic scenario. A long-range transport hazard

metric, the characteristic travel distance (CTD) was used to investigate these changes for

both labile and persistent chemicals with a large variety of enviromental fates, and it was

found that using a static, steady-state description of environmental parameters can lead to a

factor of 3 under-estimation of the CTD. Although the factor difference in the results was

relatively large, they were consistent across the range of chemical persistence and

environmental distribution; meaning that using CTDs calculated with the more simplistic

static environmental parameters is acceptable so long as they are only used to judge the long-

range transport hazard of chemicals relative to each other.

Chapter 5 of this thesis presents the creation and application of a new chemical property

prediction method. This method is an integrated approach that involves creating two

dimensional structural fragments from chemicals in a dataset, a rational splitting of the

dataset into external and internal training and validation datasets, and a robust model

selection algorithm that is designed to find a good model fit for noisy datasets without

overfitting. This method takes a chemical property dataset and with no outside guidance

automatically creates a quantitative structure property (or activity) relationship (QSPR or

QSAR). As a test of the method a QSAR is created for whole body in vivo biotransformation

half-lives in fish,165

which is an important parameter in determining bioaccumulation hazard

metrics for the aquatic foodchain. Statistics and residuals of the half-life predictions and the

fragments selected for inclusion in the example QSAR compare favorably with those for a

QSAR created with expert judgement,49

suggesting the method will be a useful tool in

predicting other uncertain properties required for chemical hazard and risk assessment.

267

2 Recommendations and Future Directions Recommendations resulting from this thesis fall under two categories; recommendations

regarding the validity of simplifying assumptions in mass balance models used to calculate

chemical hazard metrics, and recommendations for the prediction and usage of chemical

properties in hazard and risk assessment. Chapter 2 demonstrated how the outputs of a

multimedia mass balance model combined with a foodweb bioaccumulation and human

exposure model could be successfully used to simultaneously screen chemicals for long-

range transport and bioaccumulation hazard metrics. There were many simplifying

assumptions in the underyling multimedia environmental model and a possible effect of this

was noted for particle-bound chemicals which appeared to have a depressed potential for

long-range transport compared to expectations. Chapter 3 and Chapter 4 followed this up by

further investigating the effects of these simplifying assumptions on model outputs for

exposure relevant environmental media and for the CTD, a long-range transport hazard

metric. In Chapter 4 it was found that the assumption of constant rainfall, along with a low

time resolution for other environmental parameters, caused the CTD to be underestimated by

up to a factor of three. The effect was relatively uniform across a wide range of chemical

persistence and partitioning behaviour though, and the differences between chemical

amounts in environmental phases directly relevant for exposure (air and water) were

similarly uniform in Chapter 3. Based on these results it is recommended that future large-

scale chemical hazard assessment excercises which use hazard metrics calculated with mass

balance models should be prioritization oriented rather than screening oriented, because

simplifying assumptions in these mass balance models have been shown here to affect the

absolute magnitude of chemical hazard metrics they are used to calculate. In many cases the

simplifying assumptions may lead to only small differences from more sophisticated models,

but in screening oriented hazard assessment the use of cut-offs will inevitably lead to false

positives and false negatives as a result. A prioritization oriented hazard assessment is likely

to be much less uncertain, because the detrimental effect of the simplifying assumptions is

experienced approximately equally for a wide range of chemicals and the hazard of the

assessed chemicals relative to each other is likely to remain largely unaltered.

268

In Chapter 2 three major data deficiencies were identified; first there was no widely

applicable QSAR available to make predictions for biodegradation in biota, and second there

was similarly no widely applicable QSAR available for toxicity. Additionally, there is little

or no publically available data for the production volumes of most chemicals or their uses,

making the estimation of chemical emissions nearly impossible. These data deficiencies have

the net result that large scale chemical screening and prioritization excercises are currently

limited to hazard based metrics, rather than risk based metrics. As discussed in Chapter 1

risk is the product of exposure and adverse effects, and at the moment both of these metrics

are unavailable on a large scale using publically available data; the lack of emission

estimates means chemical exposures cannot be unequivocally ranked and the lack of toxicity

predictions means that the potential for adverse effects is also inaccessible. Using the new

chemical property prediction method presented in this thesis a screening and prioritization

metric for assessing a chemical‘s biodegradation potential is closer to being a reality, which

would allow for a more detailed large-scale chemical hazard assessement. Future work may

also see the development of a widely applicable prediction method for toxicity, which would

bring the state of the science another step closer to fully automated risk assessment. By far

the largest uncertainty, and data need, though remains chemical emissions estimates. This is

unfortunate because, as an extensive property, production volume is independent of chemical

structure and cannot be predicted. Possibilities for further research along this avenue are

limited, the collection and distribution of production data is in the jurisdiction of chemical

regulators and chemical industry. Researchers can further the state of the science in this

direction only by acting as advocates for greater openness in chemical production and use.

269

References

270

1 CAS, Chemical Abstracts Service - Database Counter; http://www.cas.org/cgi-

bin/cas/regreport.pl; Accessed 03 14, 2011.

2 The 2007 OECDlist of high production volume chemicals; OECD; OECD Environment,

Health and Safety Publications: Environmental Directorate: Paris, 2009.

3 van Leeuwen, C.J.; Vermeire, T.G.; (eds.), Risk assessment of chemicals: an introduction.

Second edition; Kluwer Academic Publishers: Dordrecht, 2007.

4 Nielsen, E.; Ostergaard, G.; Larsen, J.C. Toxicological risk assessment of chemicals: a

practical guide; Informa Healthcare: New York, 2008.

5 Final act of the conference of plenipotentiaries on the Stockholm Convention on persistent

organic pollutants; UNEP; United Nations Environment Program: Geneva, 2001.

6 Registration, Evaluation and Authorization of Chemicals (REACH); European

Commission; COM 2003: 2003, 644 final.

7 Ecological categorization of substances on the domestic substances list (DSL);

Environment Canada; Government of Canada: Ottawa, 2004.

8 Lippmann, M.; Cohen, B.S.; Schlesinger, R.B. Environmental health science: recognition,

evaluation, and control of chemical and physical health hazards; Oxford University Press:

New York, 2003.

9 van Wijk, D.; Chenier, R.; Henry, T.; Hernando, M.D.; Schulte, C. Integrated approach to

PBT and POP prioritization and risk assessment. Integr. Environ. Assess. Manag. 2009, 5,

697-711.

10 Scheringer, M. Persistence and spatial range as endpoints of an exposure-base assessment

of organic chemicals. Environ. Sci. Technol. 1996, 30, 1652-1659.

11 Breivik, K.; Sweetman, A.; Pacyna, J.M.; Jones, K.C. Towards a global historical

emission inventory for selected PCB congeners - a mass balance approach-3. an update. Sci.

Total Environ. 2007, 377, 296-307.

12 Erickson, M.D.; Kaley II, R.G. Applications of polychlorinated biphenyls. Environ. Sci.

Pollut. Res. 2011, 18, 135-151.

271

13 Trudel, D.; Horowitz, L.; Wormuth, M.; Scheringer, M.; Cousins, I.T.; K, Hungerbuhler.

Estimating consumer exposure to PFOS and PFOA. Risk Anal. 2008, 28, 251-269.

14 Schwarzenbach, R.P.; Gschwend, P.M.; Imboden, D.M. Environmental Organic

Chemistry, 2nd Ed; Wiley: , 2002.

15 Mackay, D. Multimedia environmental models - the fugacity approach, 2nd ed.; Lewis

Publishers: Chelsea, 2001.

16 Preparation of a product register based exposure index; Fischer, S.; Almkvist, Å.; E,

Karlsson.; Åkerblom, M. ; Swedish Chemicals Agency: Sweden, 2006.

17 Cowan-Ellsberry, C.E.; S, McLachlan.M.; A, Arnot.J.; Macleod, M.; E, McKone..T.;

Wania, F. Modeling exposure to persistent chemical in hazard and risk assessment. Integr.

Environ. Assess. Manag. 2009, 5, 662-679.

18 Arnot, J.A.; Armitage, J.M.; McCarty, L.S.; Wania, F.; Cousins, I.T.; Toose-Reid, L.

Toward a consistent evaluative framework for POP risk characterization. Environ. Sci.

Technol. 2011, 45, 97-103.

19 MacLeod, M.; Riley, W.J.; McKone, T.E. Assessing the influence of climate variability

on atmospheric concentrations of polychlorinated biphenyls using a global-scale mass

balance model (BETR-Global). Environ. Sci. Technol. 2005, 39, 6749-6756.

20 Powell, A.; Mackay, D.; Webster, E.; Arnot, J.A. Modeling bioaccumulation using

characteristic times. Environ. Toxicol. Chem. 2009, 28, 272-278.

21 Boxall, A.B.A.; Sinclair, C.J.; Fenner, K.; Kolpin, D.; Maud, S.J. When synthetic

chemicals degrade in the environment. Environ. Sci. Technol. 2004, 38, 368-375.

22 Fenner, K.; Kooijman, C.; Scheringer, M.; Hungerbuhler, K. Including transformation

products into the risk assessment for chemicals: the case of nonylphenol ethoxylate usage in

Switzerland. Environ. Sci. Technol. 2002, 36, 1147-1154.

23 Czub, G.; McLachlan, M.S. Bioaccumulation potential of persistent organic chemicals in

humans. Environ. Sci. Technol. 2004, 38, 2406-2412.

272

24 MacLeod, M.; Scheringer, M.; McKone, T.E.; Hungerbuhler, K. The state of multimedia

mass-balance modeling in environmental science and decision-making. Environ. Sci.

Technol. 2010, 44, 8360-8364.

25 McLachlan, M.S.; Czub, G.; MacLeod, M.; Arnot, J.A. Bioaccumulation of organic

contaminants in humans: A multimedia perspective and the importance of biotranformation.

Environ. Sci. Technol. 2011, 45, 197-202.

26 Arnot, J.A.; D, Mackay. Policies for chemical hazard and risk priority setting: Can

persistence, bioaccumulation, toxicity and quantity information be combined?. Environ. Sci.

Technol. 2008, 42, 4648-4654.

27 Scheringer, M.; Jones, K.C.; Matthies, M.; Simonich, S.; van de Meent, D. Multimedia

partitioning, overall persistence, and long-range transport potential in the context of POPs

and PBT chemical assessments. Integr. Environ. Assess. Manag. 2009, 5, 557-576.

28 Mackay, D. Finding fugacity feasible. Environ. Sci. Technol. 1979, 13, 1218-1223.

29 Mackay, D.; Fraser, A. Bioaccumulation of persistent organic chemicals: mechanisms and

models. Environ. Pollut. 2000, 110, 375-391.

30 Daugherty, J. Assessment of chemical exposures: calculation methods for environmental

proffessionals; Lewis Publishers: Boca Raton, 1998.

31 Arnot, J.A.; Mackay, D.; Parkerton, T.F.; Zaleski, R.T.; Warren, C.S. Multimedia

modeling of human exposure to chemical substances: the roles of food web biomagnification

and biotransformation. Environ. Toxicol. Chem. 2010, 29, 45-55.

32 Arnot, J.A.; Mackay, D.; Sutcliffe, R.; Lo, B. Estimating farfield organic chemical

exposures, intake rates and intake fractions to human age classes. Environ. Model. Soft.

2010, 25, 1166-1175.

33 McKone, T.E.; Daniels, J.I. Estimating human exposure through multiple pathways from

air, water and soil. Regul. Toxicol. Pharmacol. 1991, 13, 36-61.

34 Breivik, K.; Czub, G.; McLachlan, M.S.; Wania, F. Towards an understanding of the link

between environmental emissions and human body burdens of PCBs using CoZMoMAN.

Environ. Intern. 2010, 36, 85-91.

273

35 Cahill, T.M.; Mackay, D. Complexity in multimedia mass balance models: When are

simple models adequate and when are more complex models necessary?. Environ. Toxicol.

Chem. 2003, 22, 1404-1412.

36 Beyer, A.; Mackay, D.; Matthies, M.; Wania, F.; Webster, E. Assessing long-range

transport potential of persistent organic polluntants. Environ. Sci. Technol. 2000, 34, 699-

703.

37 Fenner, K.; Scheringer, M.; MacLeod, M.; Matthies, M.; McKone, T.; Stroebe, M.;

Beyer, A.; Bonnell, M.; Le Gall, A.C.; Klasmeier, J.; Mackay, D.; Van de Meent, D.;

Pennington, D.; Scharenberg, B.; Suzuki, N.; Wania, F. Comparing estimates of persistence

and long-range transport potential among multimedia models. Environ. Sci. Technol. 2005,

39, 1932-1942.

38 Stroebe, M.; Scheringer, M.; Hungerbuhler, K. Measures of overall persistence and the

temporal remote state. Environ. Sci. Technol. 2004, 38, 5665-5673.

39 Mackay, D.; McCarty, L.S.; MacLeod, M. On the validity of classifying chemicals for

persistence, bioaccumulation, toxicity, and potential for long-range transport. Environ.

Toxicol. Chem. 2001, 20, 1491-1498.

40 Czub, G.; Wania, F.; McLachlan, M.S. Combining long range transport and

bioaccumulation considerations to identify potential Arctic contaminants. Environ. Sci.

Technol. 2008, 42, 3704–3709.

41 Exposure Assessment Tools and Models, Estimation Program Interface (EPI) Suite; U.S.

EPA; : Washington DC, 2005.

42 Wegmann, F.; Cavin, L.; MacLeod, M.; Scheringer, M.; Hungerbuhler, K. The OECD

software tool for screening chemicals for persistence and long-range transport. Environ.

Modell. Software 2009, 24, 228-237.

43 Fisk, A.T.; Norstrom, R.J.; Cymbalisty, C.D.; Muir, D.C.G. Dietary assumulation and

depuration of hydrophobic organochlorines: Bioaccumulation parameters and their

relationship with the octanol/water partition coefficient. Environ. Toxicol. Chem 1998, 17,

951-961.

274

44 Seth, R.; Mackay, D.; Muncke, J. Estimating the organic carbon partition coefficient and

its variability for hydrophobic chemicals. Environ. Sci. Technol. 1999, 33, 2390-2394.

45 Abraham, M.H.; Chadha, H.S.; Whiting, G.S.; Mitchell, R.C. Hydrogen Bonding. 32. An

analysis of water-octanol and water-alkane partitioning and the Δlog P parameter of Seiler. J.

Pharm. Sci. 1994, 83, 1085–1100.

46 Neiderer, C.; Goss, K.-U. Quantum chemical modeling of humic acid/air equilibrium

partitioning of organic vapors. Environ. Sci. Technol. 2007, 41, 3646-3652.

47 Wittekindt, C.; Goss, K.-U. Screening the partition behavior of a large number of

chemicals with a quantum-chemical software. Chemosphere 2009, 76, 460-464.

48 Biodegradation Probability Program (BIOWIN); U.S. EPA; EPI Suite: Washington DC,

2005.

49 Arnot, J.A.; Meylan, W.; Tunkel, J.; Howard, P.H.; Mackay, D.; Bonnell, M.; Boethling,

R.S. A quantitative structure-activity relationship for predicting metabolic biotransformation

rates for organic chemicals in fish. Environ. Toxicol. Chem. 2009, 28, 1168-1177.

50 Dearden, J.C. In silico prediction of ADMET properties: how far have we come?. Expert

Opin. Drug Met. 2007, 3, 635-639.

51 Introduction and Executive Summary; Bowes, M.; Brown, M.; Rivas, Y.; Butler, D.;

Zeiger, E. ; Office of Technology Assessment: Washing DC, 1995.

52 Second national report on human exposure to environmental chemicals; CDC; National

Center for Environmental Health, Centers for Disease Control and Prevention: Atlanta, 2003.

53 Lau, C.; Anitole, K.; Hodes, C.; Lai, D.; Pfahles-Hutchens, A.; Seed, J. Perfluoroalkyl

acids: A review of monitoring and toxicological findings. Toxicol. Sci. 2007, 99, 366-394.

54 Wania, F. Potential of degradable organic chemicals for absolute and relative enrichment

in the Arctic. Environ. Sci. Technol. 2006, 40, 569-577.

55 Muir, D.C.G.; Howard, P.H. Are there other persistent organic pollutants? A challenge for

environmental chemists. Environ. Sci. Technol. 2006, 40, 7157-7166.

275

56 Xiao, H.; F, Wania. Is vapor pressure or the octanol-air partition coefficient a better

descriptor of the partitioning between gas phase and organic matter?. Atmos. Environ. 2003,

37, 2867-2878.

57 Niederer, C.; Goss, K.-U.; Schwarzenbach, R.P. Sorption equilibrium of a wide spectrum

of organic vapors in Leonardite humic acid: modeling of experimental data. Environ. Sci.

Technol. 2006, 40, 5374-5379.

58 Arp, H.P.H.; Niederer, C.; Goss, K-U. Predicting the partitioning behavior of various

highly fluorinated compounds. Environ. Sci. Technol. 2006, 40, 7298-7304.

59 Atkinson, R.; Guicherit, R.; Hites, R.A.; Palm, W.U.; Seiber, J.N.; de Voogt, P.

Transformations of pesticides in the atmosphere: A state of the art. Water Air Soil Pollut.

1999, 115, 219-243.

60 High production volume (HPV) challenge program. The HPV voluntary challenge

chemical list; U.S. EPA; U.S. Environmental Protection Agency: Merrifield, 2006.

61 European chemical substances information system; ECB; European Chemical Bureau,

European Commission: Ispra, 2006.

62 The 2004 OECD list of high production volume chemicals; OECD; OECD Environment,

Health and Safety Publications: Paris, 2004.

63 Interagency testing committee under the toxic substances contral act (TSCA); U.S. EPA;

U. S. Environmental Protection Agency, Office of Pollution Prevention and Toxics: Washing

DC, 2003.

64 Pesticide product information system;

http://www.epa.gov/opppmsdl/PPISdata/index.html; Accessed December 15, 2007.

65 The WHO recommended classification of pesticides by hazard and guidelines to

classification: 2004; WHO; World Health Organization: Geneva, 2005.

66 Wania, F. Assessing the potential of persistent organic chemicals for long-range transport

and accumulation in polar regions. Environ. Sci. Technol. 2003, 37, 1344-1351.

276

67 Hilal, S.H.; Karickhoff, S.W.; Carreira, L.A. Prediction of the solubility, activity

coefficient and liquid/liquid partition coefficient of organic compounds. QSAR Comb. Sci.

2004, 23, 709-720.

68 Baird, C.; Cann, M. Environmental Chemistry; W.H. Freeman & Company: New York,

2005.

69 Kwok, E.S C.; Atkinson, R. Estimation of hydroxyl radical reaction rate constants for

gas-phase organic compounds using a structure-reactivity relationship: An update. Atmos.

Environ. 1995, 29, 1685-1695.

70 Estimation accuracy of the Atmospheric Oxidation Program; U.S. EPA; U.S.

Environmental Protection Agency: Washington DC, 2000.

71 Goss, K.-U.; Bronner, G. What is so special about the sorption behavior of highly

fluorinated compounds?. J. Phys. Chem. A 2006, 110, 9518-9522.

72 Breivik, K.; Wania, F.; Muir, D.C.G.; Alaee, M.; Backus, S.; Pacepavicius, G. Empirical

and modeling evidence of the long-range atmospheric transport of decabromodiphenyl ether.


73 Ewald, G.; Larsson, P.; Linge, H.; Lennart, O.; Szarzi, N. Biotransport of organic

pollutants to an inland Alaska lake by migrating Sockeye Salmon (Oncorhynchus nerka).

Arctic 1998, 51, 40-47.

74 Koistinen, J.; Mussalo-Rauhamaab, H.; Paasivirta, J. Polychlorinated diphenyl ethers,

dibenzo-p-dioxins and dibenzofurans in Finnish human tissues compared to environmental

samples. Chemosphere 1995, 31, 4259-4271.

75 Robust summaries & test plans: Chlorinated pyridine category; The Dow Chemical

Company; U.S. Environmental Protection Agency: Merrifield, 2005.

76 Fire resistent multilayer film with at least one diffusion barrier layer and its use for

vacuum insulation panels; Kaczmarek, D.; Jacobsen, S. ; CODEN:EPXXDW EP 1607217

A1 20051221 CAN 144:52758 AN 2005:1328000 CAPLUS: Eur. Pat. Appl., 2005.

277

77 Siloxane gels for personal care products; Katsoulis, D.E. ; CODEN: USXXAM US

5486566 A 19960123 CAN 124:211487 AN 1996:121341 CAPLUS: United States of

America, 1996.

78 Braune, B.M.; Outridge, P.M.; Fisk, A.T.; Muir, D.C G.; Helm, P.A.; Hobbs, K.;

Hoekstra, P.F.; Kuzyk, Z.A.; Kwan, M.; Letcher, R.J.; Lockhart, W.L.; Norstrom, R.J.;

Stern, G.A.; Stirling, I. Persistent organic pollutants and mercury in marine biota of the

Canadian Arctic: An overview of spatial and temporal trends. Sci. Total Environ. 2005, 351-

352, 4-56.

79 Evans, M.S.; Muir, D.; Lockhart, W.L.; Stern, G.; Ryan, M.; Roach, P. Persistent organic

pollutants and metals in the freshwater biota of the Canadian Subarctic and Arctic: An

overview. Sci. Total Environ. 2005, 351-352, 94-147.

80 Gamberg, M.; Braune, B.; Davey, E.; Elkin, B.; Hoekstra, P.F.; Kennedy, D.; C,

Macdonald.; Muir, D.; Nirwal, A.; Wayland, M.; Zeeb, B. Spatial and temporal trends of

contaminants in terrestial biota from the Canadian Arctic. Sci. Total Environ. 2005, 351-352,

148-164.

81 Li, N.; Wania, F.; Lei, Y.D.; Daly, G.L. A comprehensive and critical compilation,

evaluation, and selection of physical-chemical property data for selected polychlorinated

biphenyls. J. Phys. Chem. Ref. Data 2003, 32, 1545-1590.

82 Su, Y.; Lei, Y.D.; Daly, G.L.; Wania, F. Determination of octanol-air partition coefficient

(KOA) values for chlorobenzenes and polychlorinated naphthalenes from gas

chromatographic retention times. J. Chem. Eng. Data 2002, 47, 449-455.

83 Feigel, M. Phenyl rotation in diphenylether and diphenylmethane calculated with ab initio

methods. J. Mol. Struct. (Theochem) 1996, 366, 83-88.

84 Sawunyama, P.; Bailey, G.W. Modeling the interaction of agrochemicals with

environmental surfaces: Pesticdes on rutile and organo-rutile surfaces. J. Mol. Struct.

(Theochem) 2001, 541, 119-129.

85 Lundgren, K.; Tysklind, M.; Ishaq, R.; Broman, D.; van Bavel, B. Polychlorinated

naphthalene levels, distribution, and biomagnification in a benthic food chain in the Baltic

Sea. Environ. Sci. Technol. 2002, 36, 5005-5013.

278

86 Abraham, M.H.; Al-Hussaini, A.J.M. Solvation descriptors for the polychloro-

naphthalenes: estimation of some physicochemical properties. J. Environ. Monit. 2001, 3,

377–381..

87 Corsolini, S.; Kannan, K.; Imagawa, T.; Focaradi, S.; Giesy, J.P. Polychloronaphthalenes

and other dioxin-like compounds in Arctic and Antarctic marine food webs. Environ. Sci.

Technol. 2002, 36, 3490-3496.

88 Steffen, C.; Borga, K.; Skaare, J.U.; Bustnes, J.O. The occurence of organochlorines in

marine avian top predators along a latitudinal gradient. Environ. Sci. Technol. 2006, 40,

5139-5146.

89 Abraham, M.H.; Al-Hussaini, A.J.M. Solvation parameters for the 209 PCBs: calculation

of physicochemical properties. J. Environ. Monit. 2005, 7, 295–301..

90 Kannan, K.; Yun, S.H.; Evans, T.J. Chlorinated, brominated and perfluorinated

contaminants in livers of polar bears from Alaska. Environ. Sci. Technol. 2005, 39, 9057-

9063.

91 Knudsen, L.B.; Borga, K.; Jorgensen, E.H.; van Bavel, B.; Schlabach, M.; Verreault, J.;

Gabrielsen, G.W. Halogenated organic contaminants and mercury in northern fulmars

(Fulmarus glacialis): levels, relationships to dietary descriptors and blood to liver

comparison. Environ. Pollut. 2007, 146, 25-33.

92 Shen, L.; Wania, F. Compilation, evaluation, and selection of physical-chemical property

data for organochlorine pesticides. J. Chem. Eng. Data 2005, 50, 742-768.

93 Stern, G.A.; MacDonald, C.R.; Armstrong, D.; Dunn, B.; Fuchs, C.; Harwood, L.; Muir,

D.C.G.; Rosenberg, B. Spatial trends and factors affecting variation of organochlorine

contaminants levels in Canadian Arctic beluga (delphinapterus leucas). Sci. Total Environ.

2005, 351, 344-368.

94 Sørmo, E.G.; Salmer, M.P.; Jenssen, B.M.; Hop, H.; Bæk, K.; Kovacs, K.M.; Lydersen,

C.; Falk-Petersen, S. Biomagnification of polybrominated diphenyl ether and hexa-

bromocyclododecane flame retardants in the polar bear food chain in Svalbard, Norway.

Environ. Toxicol. Chem. 2006, 25, 2502-2511.

279

95 Cetin, B.; Odabasi, M. Measurement of Henry‘s law constants of seven polybrominated

diphenyl ether (PBDE) congeners as a function of temperature. Atmos. Environ. 2005, 39,

5273–5280.

96 Harner, T.; Shoeib, M. Measurements of octanol-air partition coefficients (KOA) for

polybrominated diphenyl ethers (PBDEs): Predicting partitioning in the environment. J.

Chem. Eng. Data 2002, 47, 228-232.

97 Muir, D.C.G.; Backus, S.; Derocher, A.E.; Dietz, R.; Evans, T.J.; Gabrielsen, G.W.;

Nagy, J.; Norstrom, R.J.; Sonne, C.; Stirling, I.; Taylor, M.K.; Letcher, R.J. Brominated

flame retardants in Polar Bears (Ursus maritimus) from Alaska, the Canadian Arctic, East

Greenland, and Svalbard.. Environ. Sci. Technol. 2006, 40, 449-455.

98 Fuglei, E.; Bustnes, J.O.; Hop, H.; Mørk, T.; Bjornfoth, H.; van Bavel, B. Environmental

contaminants in arctic foxes (Alopex lagopus) in Svalbard: relationships with feeding

ecology and body condition. Environ. Pollut. 2007, 146, 128-138..

99 de Wit, C.A.; Alaee, M.; Muir, D.C.G. Levels and trends of brominated flame retardants

in the Arctic. Chemosphere 2006, 64, 209-233.

100 Paasivirta, J.; Sinkkonen, S.; Mikkelson, P.; Rantio, T.; Wania, F. Estimation of vapor

pressures, solubilities and Henry's Law constants of selected persistent organic pollutants as

functions of temperature. Chemosphere 1999, 39, 811-832..

101 Vorkamp, K.; Riget, F.; Glasius, M.; Pecselia, M.; Lebeuf, M.; Muir, D.C.G. Chloro-

benzenes, chlorinated pesticides, coplanar chlorobiphenyls and other organochlorine

compounds in Greenland biota.. Sci. Total Environ. 2004, 331, 157–175..

102 Mckinney, M.A.; de Guise, S.; Martineau, D.; Beland, P.; Lebeuf, M.; Letcher, R.J.

Organohalogen contaminants and metabolites in beluga whale (delphinapterus leucas) liver

from two Canadian populations. Environ. Toxicol. Chem. 2006, 25, 1246–1257.

103 Xiao, H.; Li, N.; Wania, F. Compilation, evaluation, and selection of physical-chemical

property data for α-, β-, γ-hexachlorocyclohexane. J. Chem. Eng. Data 2004, 49, 173-185.

104 Coelhan, M.; Reil, I.; Rimkus, G.; Parlar, H. Peak patterns of chlorostyrenes in fish and

fish oils from the North Atlantic. Environ. Sci. Technol. 2000, 34, 4695-4700.

280

105 Verreault, J.; Norstrom, R.J.; Ramsay, M.A.; Mulvihill, M.; Letcher, R.J. Composition

of chlorinated hydrocarbon contaminants among major adipose tissue depots of Polar Bears

(Ursus maritimus) from the Canadian high Arctic. Sci. Total Environ. 2006, 370, 580–587.

106 Smithwick, M.; Norstrom, R.J.; Mabury, S.A.; Solomon, K.; Evans, T.J.; Stirling, I.;

Taylor, M.K.; Muir, D.C.G. Temporal trends of perfluoroalkyl contaminants in polar bears

(ursus maritimus) from two locations in the North American Arctic, 1972-2002. Environ.

Sci. Technol. 2006, 40, 1139-1143.

107 Reth, M.; Ciric, A.; Christensen, G.N.; Heimstad, E.S.; Oehme, M. Short- and medium-

chain chlorinated paraffins in biota from the European Arctic — differences in homologue

group patterns. Sci. Total Environ. 2006, 367, 252–260.

108 Fisk, A.T.; Rosenberg, B.; Cymbalisty, C.D.; Stern, G.A.; Muir, D.C.G. Octanol/ water

partition coefficients of toxaphene congeners determined by the "slow-stirring" method.

Chemosphere 1999, 39, 2549-2562.

109 Carlson, R. Silent Spring; Houghton Mifflin Co.: Boston, 1962.

110 Martin, J.W.; Smithwick, M.M.; Braune, B.M.; Hoekstra, P.F.; Muir, D.C G.; Mabury,

S.A. Identification of long-chain perfluorinated acids in biota from the Canadian Arctic.

Enrion. Sci. Technol. 2004, 38, 373-380.

111 Breivik, K.; Wania, F. Expanding the applicability of multimedia fate models to polar

organic chemicals. Environ. Sci. Technol. 2003, 37, 4934-4943.

112 Goss, K.-U.; Schwarzenbach, R.P. Linear free energy relationships used to evaluate

equilibrium partitioning of organic compounds. Environ. Sci. Technol. 2001, 35, 1-9.

113 Żukowska, B.; Breivik, K.; Wania, F. Evaluating the environmental fate of

pharmaceuticals using a level III model based on polyparameter linear free energy

relationships. Sci. Total Environ. 2006, 359, 177–187.

114 Götz, C.W.; Scheringer, M.; MacLeod, M.; Wegmann, F.; Schenker, U.; Hungerbühler,

K. Dependence of persistence and long-range transport potential on gas-particle partitioning

in multimedia models. Environ. Sci. Technol. 2008, 42, 3690–3696.

281

115 Abraham, M.H. Scales of solute hydrogen-bonding: their construction and application to

physicochemical and biochemical processes. Chem. Soc. Rev. 1993, 22, 73-83.

116 Poole, S.K.; Poole, C.F. Chromatographic models for the sorption of neutral organic

compounds by soil from water and air. J. Chromatogr. A 1999, 845, 381–400.

117 Götz, C.W.; Scheringer, M.; MacLeod, M.; Roth, C.M.; Hungerbühler, K. Alternative

approaches for modeling gas/particle partitioning of semivolatile organic chemicals: model

development and comparison. Environ. Sci. Technol. 2007, 41, 1272–1278.

118 Goss, K.-U. Predicting the equilibrium partitioning of organic compounds using just one

linear solvation energy relationship (LSER). Fluid Phase Equilib. 2005, 233, 19–22.

119 Niederer, C.; Goss, K.-U.; Schwarzenbach, R.P. Sorption equilibrium of a wide

spectrum of organic vapors in Leonardite humic acid: experimental setup and experimental

data. Environ. Sci. Technol. 2006, 40, 5368-5373.

120 Arp, H.P.H.; Schwarzenbach, R.P.; Goss, K.-U. Ambient gas/particle partitioning. 1.

sorption mechanisms of apolar, polar, and ionizable organic compounds. Environ. Sci.

Technol. 2008, 42, 5541–5547.

121 Arp, H.P.H.; Schwarzenbach, R.P.; Goss, K.-U. Ambient gas/particle partitioning. 2: the

influence of particle source and temperature on sorption to dry terrestrial aerosols. Environ.

Sci. Technol. 2008, 42, 5951–5957.

122 Mintz, C.; Clark, M.; Acree, W.E.; Abraham, M.H. Enthalpy of solvation correlations

for gaseous solutes dissolved in water and in 1-octanol based on the Abraham model. J.

Chem. Inf. Model. 2007, 47, 115-121.

123 Abraham, M.H.; Whiting, G.S.; Fuchs, R.; Chambers, E.J. Thermodynamics of solute

transfer from water to hexadecane. J. Chem. Soc., Perkin Trans. 2 1990, , 291-300.

124 Fuchs, R.; Abraham, M.H.; Kamlet, M.J.; Taft, R.W. Solute-solvent interactions in

chemical and biological systems. IV. Correlations of ΔG, ΔH and TΔS of transfer of

aliphatic and aromatic solutes from 2,2,4-trimethylpentane to aqueous solution. J. Phys. Org.

Chem. 1989, 2, 559-564.

282

125 Abraham, M.H.; Andonian-Haftvan, J.; Whiting, G.S.; Leo, A.; Taft, R.S. Hydrogen

bonding. Part 34. The factors that influence the solubility of gases and vapours in water at

298 K, and a new method for its determination. J. Chem. Soc., Perkin Trans. 2 1994, , 1777–

1791.

126 Goss, K.-U.; Arp, H.P.H.; Bronner, G.; Niederer, C. Non-additive effects in the

partitioning behaviour of various aliphatic and aromatic molecules. Environ. Toxicol. Chem.

2009, 28, 52-60.

127 Tülp, H.C.; Goss, K.-U.; Schwarzenbach, R.P.; Fenner, K. Experimental determination

of LSER parameters for a set of 76 diverse pesticides and pharmaceuticals. Environ. Sci.

Technol. 2008, 42, 2034–2040.

128 Wania, F.; Breivik, K.; Persson, N.J.; McLachlan, M.S. CoZMo-POP 2 - A fugacity-

based dynamic multi-compartment mass balance model of the fate of persistent organic

pollutants.. Environ. Model. Software 2006, 21, 868-884.

129 Meyer, T.; Wania, F.; Breivik, K. Illustrating sensitivity and uncertainty in

environmental fate models using partitioning maps. Environ. Sci. Technol. 2005, 39, 3186-

3196.

130 Goss, K.-U.; Schwarzenbach, R.P. Empirical prediction of heats of vaporization and

heats of adsorption of organic compounds. Environ. Sci. Technol. 1999, 33, 3390-3393.

131 MacLeod, M.; Scheringer, M.; Hungerbühler, K. Estimating enthalpy of vaporization

from vapor pressure using Trouton‘s Rule. Environ. Sci. Technol. 2007, 41, 2827-2832.

132 Horstmann, M.; McLachlan, M.S. Atmospheric deposition of semi-volatile compounds

to two forest canopies. Atmos. Environ. 1998, 32, 1799-1809.

133 Platts, J.A.; Abraham, M.H. Partition of volatile organic compounds from air and from

water into plant cuticular matrix: an LFER analysis. Environ. Sci. Technol. 2000, 34, 318-

323.

134 Goss, K.-U.; Arp, H.P.H.; Bronner, G.; Niederer, C. Partition behavior of

hexachlorocyclo-hexane isomers. J. Chem. Eng. Data 2008, 53, 750–754.

283

135 Abraham, M.H.; Dearden, J.C.; Bresnen, G.M. Hydrogen bonding, steric effects and

thermodynamics of partitioning. J. Phys. Org. Chem. 2006, 19, 242–248.

136 Abraham, M.H.; Enomoto, K.; Clarke, E.D.; Rośes, M.; Ràfols, C.; Fuguet, E. Henry‘s

Law constants or air to water partition coefficients for 1,3,5-triazines by an LFER method. J.

Environ. Monit. 2007, 9, 234–239.

137 Abraham, M.H.; Ibrahim, A.; Zhao, Y.; Acree, W.E. A data base for partition of volatile

organic compounds and drugs from blood/plasma/serum to brain, and an LFER analysis of

the data. J. Pharm. Sci. 2006, 95, 2091-2100.

138 Abraham, M.H. Hydrogen bonding. Part 27. Solvation parameters for functionally

substituted aromatic compounds and heterocyclic compounds, from gas-liquid chromato-

graphic data. J. Chromatogr. 1993, 644, 95-139.

139 Abraham, M.H.; Grellier, P.L.; McGill, R.A. Determination of olive oil-gas and

hexadecane-gas partition coefficients, and calculation of the corresponding olive oil-water

and hexadecane-water partition coefficients. J. Chem. Soc. Perkin Trans. 2 1987, , 797-803.

140 Abraham, M.H.; Andonian-Haftvan, J.; My Du, C.; Osei-Owusu, J.P.; Sakellariou, P.;

Shuely, W.J.; Poole, C.F.; Poole, S.K. Comparison of uncorrected retention data on a

capillary and a packed hexadecane column with corrected retention data on a packed

squalane column. J. Chromatogr. A 1994, 688, 125-134.

141 Breivik, K.; Wania, F. Evaluating a model of the historical behavior of two hexachloro-

cyclohexanes in the Baltic Sea environment. nviron. Sci. Technol. 2002, 36, 1014-1023.

142 Webster, E.; Mackay, D.; Wania, F. Evaluating environmental persistence. Environ.

Toxicol. Chem. 1998, 17, 2148 -2158.

143 Finizio, A.; Mackay, D.; Bidleman, T.; Harner, T. Octanol-air partition coefficient as a

predictor of partitioning of semi-volatile organic chemical to aerosols. Atmos. Environ. 1997,

31, 2289-2296.

144 Su, Y.; Wania, F.; Harner, T.; Lei, Y.D. Deposition of polybrominated diphenylethers,

polychlorinated biphenyls, and polycyclic aromatic hydrocarbons to a boreal deciduous

forest. Environ. Sci. Technol. 2007, 41, 543-540.

284

145 Welke, B.; Ettlinger, K.; Riederer, M. Sorption of volatile organic chemicals in plant

surfaces. Environ. Sci. Technol. 1998, 32, 1099-1104.

146 Brown, T.N.; Wania, F. Screening chemicals for the potential to he persistent organic

pollutants: a case study of Arctic contaminants. Environ. Sci. Technol. 2008, 42, 5202-5209.

147 Klasmeier, J.; Matthies, M.; MacLeod, M.; Fenner, K.; Scheringer, M.; Stroebe, M.; Le

Gall, A.C.; McKone, T.; Van de Meent, D.; Wania, F. Applicationof multimedia models for

screening assessment of long-range transport potential and overall persistence. Environ. Sci.

Technol. 2006, 40, 53-60.

148 Jolliet, O.; Hauschild, M. Modeling the influence of intermittent rain events on long-

term fate and transport of organic pollutants. Environ. Sci. Technol. 2005, 39, 4513-4522.

149 Breivik, K.; Wania, F.; Muir, D..C G.; Alaee, M.; Backus, S.; Pacepavicius, G.

Empirical and modeling evidence of long-range atmospheric transport of decabromodiphenyl

ether. Environ. Sci. Technol. 2006, 40, 4612-4618.

150 Muir, D.C.G.; Teixeira, C.; Wania, F. Empirical and modeling evidence of regional

atmospheric transport of current use pesticides. Environ. Toxicol. Chem. 2004, 23, 2421-

2432.

151 Gasic, B.; MacLeod, M.; Scheringer, M.; Hungerbuhler, K. Assessing the impact of

weather events at mid-latitudes on the atmospheric transport of chemical pollutants using a

2-dimensional multimedia meteorological model. Atmosph. Environ. 2010, 44, 4489-4496.

152 Nizzetto, L.; MacLeod, M.; Borga, K.; Cabrerizo, A.; Dachs, J.; Di Guardo, A.;

Ghirardello, D.; Hansen, K.M.; Jarvis, A.; Lindroth, A.; Ludwig, B.; Monteith, D.; Perlinger,

J.A.; Scheringer, M.; Schwendenmann, L.; Semple, K.T.; Wick, L.Y.; Zhang, G.; Jones,

K.C. Past, present and future controls on levels of persistent organic pollutants in the global

environment. Environ. Sci. Technol. 2010, 44, 6526-6531.

153 Lammel, G. Effects of time-averaging climate parameters on predicted

multicompartmental fate of pesticides and POPs. Environ. Pollut. 2004, 128, 291-302.

154 Daly, G.L.; Wania, F. Simulating the influence of snow on the fate of organic

compounds. Environ. Sci. Technol. 2004, 38, 4176-7186.

285

155 National Climate Data and Information Archive;

http://climate.weatheroffice.gc.ca/climateData/canada_e.html; Accessed January 15, 2011.

156 MacLeod, M.; Scheringer, M.; Podey, H.; Jones, K.C.; Hungerbuhler, K. The origin and

significance of short-term variability of semi-volatile contaminants in air. Environ. Sci.

Technol. 2007, 41, 3249-3253.

157 Dunne, T.; Leopold, L.B. Water in Environmental Planning. 1978, , 88.

158 Finlayson-Pitts, B.J.; Pitts, J.N. Chemistry of the upper and lower atmosphere: theory,

experiments, and applications ; Elsevier Inc.: , 2000.

159 MacLeod, M.; Scheringer, M.; Hungerbuhler, K. Estimating enthalpy of vaporization

from vapor pressure using Trouton's rule. Environ. Sci. Technol. 2007, 41, 2827-2832.

160 Mintz, C.; Clark, M.; Acree, W.E.; Abraham, M.H. Enthalpy of solvation correlations

for gaseous solutes dissolved in water and in 1-octanol based on the Abraham model. J.

Chem. Inf. Model. 2007, 47, 115-121.

161 Abraham, M.H.; Chadha, H.S.; Whiting, G.S.; Mitchell, R.C. Hydrogen bonding 32. an

analysis of water-octanol and water-alkane partitioning and the delta log P parameter of

Seller. J. Pharm. Sci. 1994, 83, 1085-1100.

162 Development and application of models of chemical fate in Canada: practical methods

for environmental biodegradation rate estimation. Report to Environment Canada; Arnot,

J.A.; Mackay, D.; Gouin, T. ; Trent University: Peterborough, 2005.

163 Howard, P.H.; Muir, D.C.G. Identifying new persistent and bioaccumulative organics

among chemicals in commerce. Environ. Sci. Technol. 2010, 44, 2277-2285.

164 Zhang, X.M.; Brown, T.N.; Wania, F.; Heimstad, E.S.; Goss, K.-U. Assessment of

chemical screening outcomes based on different partitioning property estimation methods.

Environ. Int. 2010, 36, 514-520.

165 Arnot, J.A.; Mackay, D.; Parkerton, T.F.; Bonnell, M. A database of fish

biotransformation rates for organic chemicals. Environ. Toxicol. Chem. 2008, 27, 2263-2270.

166 Pavan, M.; Worth, A.P. Review of estimation models for biodegradation. QSAR Comb.

Sci. 2008, 27, 32-40.

286

167 Weininger, D.; Weininger, A.; Weininger, J.L. SMILES. 2. Algorithm for generation of

unique SMILES notation. J. Chem. Inf. Comput. Sci. 1989, 29, 97-101.

168 McQuarrie, A.D.R. Regression and Time Series Model Selection; World Scientific:

Singapore (New Jersey, London), 1998.

169 Dearden, J.C.; Cronin, M.T.D.; Kaiser, K.L.E. How not to develop a quantitative

structure–activity or structure–property relationship (QSAR/QSPR). SAR QSAR Environ.

Res. 2009, 20, 241-266.

170 Burnham, K.P.; Anderson, D.R. Multimodel inference: understanding AIC and BIC in

model selection. Sociol. Method. Res. 2004, 33, 261-304.

171 Willett, P.; Barnard, J.M.; Downs, G.M. Chemical similarity searching. J. Chem. Inf.

Comput. Sci. 1998, 38, 983-996.

172 Weininger, D. SMILES a chemical language and information system. 1. Introduction to

methodology and encoding rules. J. Chem. Inf. Comput. Sci. 1988, 28, 31-36.

173 Meylan, W.M.; Howard, P.H. Atom/fragment contribution method for estimating

octanol-water partition coefficients. J. Pharm. Sci. 1995, 84, 83-92.

174 Boethling, R.S.; Howard, P.H.; Meylan, W.; Stitleler, W.; Beauman, J.; Tirado, N.

Group contribution method for predicting probability and rate of aerobic biodegradation.


175 Zhang, X.M.; Brown, T.N.; Wania, F.; Heimstad, E.S.; Goss, K.-U. Assessment of

chemical screening outcomes based on different partitioning property estimation methods.

Environ. Int. 2010, 36, 514-520.

automated methods in chemical risk assessment€¦ · value. chemical properties relevant to hazard...

Documents