2016 aehs statistics sediment forensic presentation cherry

28
Forensic Sediment Evaluation: Differentiating Basin Derived Media vs. Anthropogenic Sources using Multivariate Statistics Eric M. Cherry, Principal Scientist Gina Groom, MPH Health Scientist Hexagon Environmental Solutions LLC Environmental Chemistry | Forensics | Data Evaluation | Risk Analysis

Upload: eric-cherry

Post on 16-Apr-2017

18 views

Category:

Documents


0 download

TRANSCRIPT

Forensic Sediment Evaluation: Differentiating Basin Derived Media vs. Anthropogenic Sources using Multivariate StatisticsEric M. Cherry, Principal ScientistGina Groom, MPH Health Scientist

Hexagon Environmental Solutions LLCEnvironmental Chemistry | Forensics | Data Evaluation | Risk Analysis

Every Story has a Beginning• A Little History

• Geochemical Prospecting• Geophysics (statistics on a sphere)• Multivariate Analysis since 1980 (punch cards)• Learning to “keep it simple”

• Initiating Quotes• “What do you mean these metals came from the

watershed? There is a source right here.” Opposing counsel

• “What do you mean that most of these PAHs came from upstream? That’s undeveloped land and there are sources here in the harbor!” Senior technical colleague

• “We use accepted protocols and high precision analytical instruments, so that we can put data in spreadsheets then run statistical models, and ultimately derive a reasonable story of what is going on. We are high-tech story tellers!” Eric Cherry

Hexagon Environmental Solutions2

Sediment Assessment - Fundamental Questions • Do the sediments in this river or harbor pose

a risk to the aquatic ecosystem or to human health?

• If so, by what chemicals or elements?

• If so, where?

• If so, who is responsible and who will pay?

Hexagon Environmental Solutions3

• What are these sediments composed of?

• Are there multiple chemical signatures in the sediments?

• Are there locations that are clearly different from others?

• How are they different?

• Can we identify a specific source?

• Do these sediments pose a risk?

• Why is the question being asked?

• Who is the audience?

• Who are the Stakeholders?

Approaches to Sediment Evaluation

Hexagon Environmental Solutions4

• Traditional Approach• Develop and Implement Sampling &

Analysis Plan

• Calculate chemical-specific distributions

• Compare to Sediment Quality Values

• Refine and Delineate Impacted Areas

• Develop and Implement Remedial Plan

• Obtain Funds from PRPs/RPs

• Alternate Forensic Approach• Develop and Implement Tiered and Sequential

Sampling & Analysis Plan

• Evaluate multiple chemical relationships to identify associated and unique sample groups by multivariate methods

• Conduct supplemental sampling for delineation, sequential extraction and toxicity testing

• Refine and Delineate Impacted Areas based on supplemental testing

• Develop and Implement Remedial Plan

• Obtain Funds from PRPs/RPs

The Traditional Approach to Sediment Evaluation

• Identify Contamination based on Criteria or Limit• Freshwater

Lowest Effects Level (LEL)Probable Effects Level (PEL)Severe Effects Level (SEL)

• Marine Effects Range Low (ERL)Effects Range Medium (ERM)

• State Values• International Values

Hexagon Environmental Solutions5

USEPA/Weston, 2012. Sediment Assessment Report – St. Louis Bay-St. Louis River Area of Concern. DCN 1023-2A-ATMN

The Math of How the Traditional Approach Works

• Obtain sediment data for arsenic (or other chemical)

• Sort by Concentration• Compare to Consensus

Value

• Green Data Set• 43.8% Exceed Tel• 5.5% Exceed Background• 1.8% Exceed PEL• 0.0% Exceed SEL

Hexagon Environmental Solutions6

Background Metals vs. Potentially Impacted Sediments

0

5

10

15

20

25

30

35

1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161

S ample Number

Severe Effects Level 33 mg/kg

Probable Effects Level 17 mg/kg

Threshold Effects Level 5.9 mg/kg

Background = mean + 2 sd 13.9 mg/kg

Arsenic vs. Consensus Value

Evaluation of Traditional Approach

• Advantages of Traditional Approach

• Easy to perform evaluation

• Clear decision points

• Delineation sampling to refine extent of impact

• Very conservative approach (i.e. “protective”)

• Easy to communicate to Stakeholders

• Disadvantages• Default classification of “contamination”

based on Consensus values (how do we define contamination?)

• Non-holistic

• Is not capable of identifying natural relationships

• Oversimplifies a complex system

• Probably overestimates extent of toxic impacts

• More expensive to implement remedy

Hexagon Environmental Solutions7

Overview of the Forensic Approach• Defining “Forensics”

• It’s all about “making the argument” in a science-based and logical manner

• Start with a “Single Blind” approach to identify what the data is saying chemically, then expand to spatial relationships (minimize investigator bias)

• Does not require, but may include, advanced analytical methods

• Does include more sophisticated data evaluation techniques

• Objective is to translate complex information into a story that is understandable to multiple Stakeholders

• Forensic Toolbox• Given . . . A big data set!

• Pair-wise regression (Fe:As, Pb:Zn)

• Cluster Analysis (Family Associations)

• Principal Component Analysis

• Concentration Distributions

• Spatial Distribution of Families

• Association of Family Composition Profiles to Potential Source Profiles

• Toxicity Testing and Sequential Extraction

• Infographics! Hexagon Environmental Solutions

8

Forensics – Doing the Math!

Hexagon Environmental Solutions9

Meet the Stakeholders!• What are our goals?

• Understand the concerns• Protect/Restore the environment• Be true to the science• Understand the physical system• Be fiscal stewards with limited resources• Communicate clearly and involve

stakeholders

Hexagon Environmental Solutions10

I’m Roy the

Regulator

I’m Black Hat

Industries

I’m Freddy Fisherman,

Nature Lover

Quick Summary with Legos!

Hexagon Environmental Solutions11

Sediments are “in the Basin”Complex mixture, but we have a sampling and analysis plan . . . And a sequential protocol for data evaluation

Each sample (n=10) contains target and non-target analytes

Quick Summary with Legos!

Hexagon Environmental Solutions12

Target analyte concentrations vary within a sample and total (sum) of analytes varies

Relative proportions of different analytes may vary between samples, and some analytes may be absent

Dealing with real data

• Given a real data set, what are the questions?• What are the baseline relationships

between chemical parameters?

• Where are the commonalities?

• Where are the differences?

• What is unique?

• What are the methods?• Regression – line, lines, shotguns• Cluster Analysis – family groups• Principal Component – prom night• Confounders - yea, that influences things• Spatial Relationships – there but not here• Supplemental evaluation – how bad is it,

really• Infographics – every picture tells a story• Conclusions – many legs make a stable

table

Hexagon Environmental Solutions13

Regression and Cross-Plots

Hexagon Environmental Solutions14

45000

30000

15000

0.20.10.0

1000

500

0 300

150

0 80400 420 0. 010

0. 005

0 .000

1000

500

01000

500

0300

150

0100

50

0 80

40

010

5

0 4

2

050

25

0

0.010

0.005

0.000

4500

030

000

1 5000

0.2

0.1

0.0

100050

0010

0500 1050 50250

Fe

Zn

Pb

Cu

Cr

Ni

As

Cd

Hg

2b-TIN

3b-TIN

Matrix Plot of Fe, Zn, Pb, Cu, Cr, Ni, As, Cd, Hg, 2b-TIN, 3b-TIN

Fe

Zn

Pb

Cu

Cr

Ni

As

Hg

Cd

2Sb

3Sb

Fe:CdShotgun(?)

HgOutlier!

Zn:PbMulti-Linear

Fe:NiLinear

Regression and Cross-Plots: Parsed!

Hexagon Environmental Solutions15

45000

30000

15000

0.10

0 .050. 00

1000

500

0 300

150

0 453015 420 0. 009

0.006

0.003

1000

500

01000

500

0300

150

0100

50

0 45

30

15

10

5

0 4

2

02

1

0 0.009

0.006

0.003

4500

030

000

150 00

0.10

0.05

0.00

100050

0010

0500 1050 210

Fe

Zn

Pb

Cu

Cr

Ni

As

Cd

Hg

2b-TIN

3b-TIN

Matrix Plot of Fe, Zn, Pb, Cu, Cr, Ni, As, Cd, Hg, 2b-TIN, 3b-TIN

Cluster Analysis – Finding Family Groups

Hexagon Environmental Solutions16

155413318

110

965191902

193

1976353420

820

619

519

220

914

213

612

312

113

537174

191

151

153

150

129

1729414

313

116

011

76648148

134

213

171

11844982218

320

720

018

519

615

7471542591240214

189

1642717

311201

215

176

156

145

12756015

835774630216

177

175

1861820

313

014

011

613

914

912

515

211

185138

1945510

311

310

289165

1125095398617

8701265715

4261793414

7871108020

421

211

918

0319990254923210

199

120

166

107817618

721

114

458101843818

8927974615167161009120

26814124

106

1598817

0920

5872128

105455413

716

2292413198

184

1159311

412

29782161

141757316

72820108966916

313

3831326410

452171697862433210771316

814

656362161821

0.00

33.33

66.67

100.00

Observations

Sim

ilari

ty

DendrogramComplete Linkage, Pearson Distance

• Purpose is to classify samples based on criteria

• Multiple methods are available

• Evaluator judgement should be based on project requirements

• Data pre-processing may be necessary

Clusters by Relative Proportion of Target Metals

Hexagon Environmental Solutions17

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Mean ALL Group MCA D 1n=151

Group MCA D 4n=51

Group MCA D 3n=5

Group MCA D 5n=4

Group MCA D 2n=2

Group MCA D 7n=2

Group MCA D 6n=1

Relative Metal Proportions in MCA Cluster Groups

Zn Pb Cu Cr Ni As

Clusters by Concentration

Hexagon Environmental Solutions18

0

200

400

600

800

1000

1200

1400

1600

Mean ALL Group MCA D 1n=151

Group MCA D 4n=51

Group MCA D 3n=5

Group MCA D 5n=4

Group MCA D 2n=2

Group MCA D 7n=2

Group MCA D 6n=1

Cum

ulat

ive

Met

als

Conc

entr

atio

n (m

g/kg

)

Cumulative Metal Concentrations in MCA Cluster Groups

Zn Pb Cu Cr Ni As

Principal Component Analysis – Let’s Dance

Hexagon Environmental Solutions19

-8

-6

-4

-2

0

2

4

-4 -2 0 2 4 6 8 10

Prin

cipa

l Com

pone

nt 2

(fro

m M

CA D

Gro

upin

gs)

Principal Component 1 (from MCA D Groupings)

Group 1 PCA 2 Group 4 PCA 2 Group 3 PCA 2 Group 5 PCA 2

Group 2 PCA 2 Group 7 PCA 2 Group 6 PCA 2

-8

-6

-4

-2

0

2

4

6

-4 -2 0 2 4 6 8 10

Prin

cipa

l Com

pone

nt 3

(fro

m M

CA D

Gro

upin

gs)

Principal Component 1 (from MCA D Groupings)

Group 1 PCA 2 Group 4 PCA 2 Group 3 PCA 2 Group 5 PCA 2

Group 2 PCA 2 Group 7 PCA 2 Group 6 PCA 2

What about those Anthropogenic Metals!

Hexagon Environmental Solutions20

y = 0.1202x - 0.3098R² = 0.8416

y = 0.0003x + 0.0078R² = 0.0005

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

0 2 4 6 8 10

Trib

utyl

Tin

(mg/

kg)

Arsenic (mg/kg)

Tributyltin vs. Arsenic

As As-lo Linear (As) Linear (As-lo)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

0 2 4 6 8 10

Trib

utyl

tin (m

g/kg

)

Arsenic (mg/kg)

Tributyltin vs. Arsenic

TBT Fraser Shipyard TBT Other Howard's Bay

Tributyltin is an anthropogenic organometallic compoundIts primary function is as an additive to marine paints to prevent biofouling on ships. It is intentionally toxic.

Tributyltin and other Metals: Paint formulations?

Hexagon Environmental Solutions21

In regression analysis, one must consider both statistical relevance and physical plausibility.

y = 1.1072x - 0.0618R² = 0.6488

y = 0.0191x + 0.0017R² = 0.1018

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

0.00 0.50 1.00 1.50

Trib

utyl

Tin

(mg/

kg)

Mercury (mg/kg)

Tributyltin vs. Mercury

Hg Hg-lo

Linear (Hg) Linear (Hg-lo)

y = 0.4659x - 0.128R² = 0.779

y = 0.0238x - 0.0096R² = 0.2448

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

0 0.5 1 1.5 2

Trib

utyl

Tin

(mg/

kg)

Cadmium (mg/kg)

Tributyltin vs. Cadmium

Cd Cd-lo

Linear (Cd) Linear (Cd-lo)

y = 0.0053x - 0.0046R² = 0.6287

y = -1E-05x + 0.0099R² = 0.002

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

0 100 200 300 400

Trib

utyl

Tin

(mg/

kg)

Lead(mg/kg)

Tributyltin vs. Lead

Pb Pb-lo

Linear (Pb) Linear (Pb-lo)

Confounding Variables

• A confounding variable is something that may affect the correlates with the dependent and independent variable

• It is a factor that may need to be adjusted for in interpreting a data set

• Potential confounders for metals in sediments include grainsize and organic carbon

• Important general questions• Where do sediments actually

come from?• How is the sediment source area

characterized?• What changes have happened in

the source area and when?• What changes have happened in

the receiving are (sediment basin) and when?

Hexagon Environmental Solutions22

Grainsize as a Confounding Factor

Hexagon Environmental Solutions23

y = 0.2476x + 7.4777R² = 0.5784

0

5

10

15

20

25

30

35

40

45

50

0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

Nic

kel (

mg/

kg)

Percent Silt+Clay

Nickle vs Grain Size

0

5

10

15

20

25

30

35

40

45

50

0.0 10.0 20.0 30.0 40.0 50.0 60.0 70.0 80.0 90.0 100.0

Nic

kel (

mg/

kg)

Percent Silt+Clay

Nickle vs Grain Size

<LEL samples = 31.2% n=128

>LEL & <PEL samples=65.1% n=267

>PEL samples=3.7% n=15

That Next? Supplemental Analyses

• Sequential Extraction• Surrogate for bioavailability

• More expensive than standard analyses

• Easily soluble

• Carbonate bound

• Organic bound

• Iron Oxide bound

• Residual

• Sediment toxicity testing• Surrogate for actual toxicity based

on specified organisms

• More expensive than chemical analysis or sequential extraction

• Probably best estimate of actual ecological toxicity measure

Hexagon Environmental Solutions24

Spatial Distribution – “Taking the blinders off”

• How does “what the data says” from compositional profiles compare with “location of samples”?

• How does the “location of samples” compare with reasonable “potential source areas”?

• Example from large sediment study

• Active marine harbor

• Urban setting

• Drainage basin characterized by mixed industry, agriculture and large forest tracts

Hexagon Environmental Solutions25

PAH Profiles and PCA Diagram

Hexagon Environmental Solutions26

-10.0

-8.0

-6.0

-4.0

-2.0

0.0

2.0

4.0

6.0

8.0

-15.0 -10.0 -5.0 0.0 5.0 10.0

Title

Title

Sediment PAH - PCA ResultsPC 1 vs PC 2

Sed PAH Group A Sed PAH Group B Sed PAH Group C

Sed PAH Group D Sed PAH Group E Sed PAH Group F

Sed PAH Group G Sed PAH Group H Sed PAH Group I

0.00

0.05

0.10

0.15

0.20

NAP

1mN

AP2m

NAP AN

YAC

EFL

UAN

TPH

EFL

APY

RBa

ACH

RBb

FBk

FBa

PBe

PD

BA PER IP

BPE

PAH 1462 Group A n=677 (46.31%)

0.000.020.040.060.080.100.120.140.160.18

NAP

1mN

AP2m

NAP AN

YAC

EFL

UAN

TPH

EFL

APY

RBa

ACH

RBb

FBk

FBa

PBe

PDB

APE

R IPBP

E

PAH 1462 Group D n=40 (2.74%)

0.00

0.05

0.10

0.15

0.20

NAP

1mN

AP2m

NAP AN

YAC

EFL

UAN

TPH

EFL

APY

RBa

ACH

RBb

FBk

FBa

PBe

PDB

APE

R IPBP

E

PAH 1462 Group F n=13 (0.89%)

Spatial Distribution after Cluster Analysis

Hexagon Environmental Solutions27

620,000

640,000

660,000

680,000

700,000

720,000

7,600,0007,610,0007,620,0007,630,0007,640,0007,650,0007,660,000

Nor

thin

g

Easting

Sediment PAH Group A Distributions

Sed Samples Group A

620,000

640,000

660,000

680,000

700,000

720,000

7,600,0007,610,0007,620,0007,630,0007,640,0007,650,0007,660,000

Nor

thin

g

Easting

Sediment PAH Group D Distributions

Sed Samples Group D

620,000

640,000

660,000

680,000

700,000

720,000

7,600,0007,610,0007,620,0007,630,0007,640,0007,650,0007,660,000

Nor

thin

g

Easting

Sediment PAH Group F Distributions

Sed Samples Group F

Urban Background Petroleum Combustion and Asphalt/Creosote

Coal and Natural Diagenetic PAHs

Conclusions• Clearly identify the goals and objectives of a sediment project

prior to establishing all methods of evaluation

• Application of multivariate statistical methods in a forensic approach can identify relationships within the basin

• Professional judgement combined with proper application of methods is essential for resolving the story in complex data sets

• Multiple lines of evidence can help to determine whether high concentration samples are associated with natural materials or anthropogenic sources of contamination

• Forensic methods of data evaluation may be slightly more expensive during investigation, but are negligible in comparison to undertaking remediation of natural sediments without significant contribution from anthropogenic sources.

Hexagon Environmental Solutions28

Questions?