appendix 4 · 2020. 6. 15. · appendix 4.4 public | wsp project no.: 70068030 june 2020...

001 JUNE 2020 PUBLIC

Northamptonshire County Council

APPENDIX 4.4

Dispersion Model Approach and Verification

Northamptonshire County Council

APPENDIX 4.4

Dispersion Model Approach and Verification

APPENDIX 4.4 PUBLIC | WSP Project No.: 70068030 001 June 2020 Northamptonshire County Council

TYPE OF DOCUMENT (VERSION) PUBLIC

PROJECT NO. 70068030

OUR REF. NO. 001

DATE: JUNE 2020

WSP

Three White Rose Office Park

Millshaw Park Lane

Leeds

LS11 0DL

WSP.com

APPENDIX 4.4 PUBLIC | WSP Project No.: 70068030 June 2020 Northamptonshire County Council Page 1 of 25

DISPERSION MODEL APPROACH AND VERIFICATION

INTRODUCTION

Air pollution in urban areas is dominated by emissions from road vehicles. The main pollutants of

concern from road traffic are oxides of nitrogen (NOx/NO2) and fine particulate matter (PM10 and

PM2.5), since these pollutants are most likely to approach their relevant air quality limit values in

proximity to major road links.

The introduction of the Proposed Scheme has the potential to change the total flow, distribution and

characteristics of traffic movements on the affected road links, which would result in changes to

emissions of the aforementioned pollutants. The local air quality assessment was completed to

predict the potential impacts of these changes on ambient pollutant concentrations at identified

sensitive receptors within proximity to affected roads.

The air quality conditions were described for the base year (2018) and Assessment Years (2021 and

2031). In 2021 and 2031, the dispersion modelling assessment considered both the Do-Minimum

(‘without’ Proposed Scheme) and Do-Something (‘with’ Proposed Scheme) scenarios.

The changes in local traffic related pollution levels predicted at the receptor locations were assessed

by comparing the predicted concentrations of NO2, PM10 and PM2.5 with the current air quality

Objectives and considering the change (improvement or worsening) between the Do-Minimum and

both Do-Something scenarios.

MODELLING METHODOLOGY

ATMOSPHERIC DISPERSION MODEL SELECTION

The predicted impacts on local air quality associated with changes to vehicle emissions as a result

of the operation of the scheme were assessed using Cambridge Environmental Research

Consultants (CERC) atmospheric dispersion modelling system for roads (ADMS-Roads v5.0.0.1).

ADMS-Roads applies advanced algorithms for the height-dependence of wind speed, turbulence

and stability to produce improved predictions of air pollutant concentrations within the given model

domain. It can predict long-term and short-term concentrations, as well as calculations of percentile

concentrations.

ADMS-Roads is a validated model, developed in the UK by CERC. The model validation process

includes comparisons with data from the UK's Automatic Urban Rural Network (AURN) and specific

verification exercises using standard field, laboratory and numerical data sets. CERC is also

involved in European programmes on model harmonisation, and their models were compared

favourably against other EU and U.S. EPA systems. Further information in relation to this is

available from the CERC web site at http://www.cerc.co.uk/environmental-software/model-

validation.html.

ATMOSPHERIC DISPERSION MODELLING PROCESS

The procedures involved in undertaking the dispersion modelling assessment are outlined below:

▪ Collation of input data – traffic data (flows, speeds, percentage of Heavy Duty Vehicles (HDVs)),

road network mapping, sensitive receptor coordinates and meteorological data;

http://www.cerc.co.uk/environmental-software/model-validation.html

http://www.cerc.co.uk/environmental-software/model-validation.html


▪ Input of data in to the ADMS-Roads model for the scenarios to be modelled;

▪ Development of emissions inventories for each pollutant to be assessed, using Defra’s emission

factor toolkit (EFT v9.0);

▪ Running the ADMS-Roads model for each considered scenario;

▪ Conversion of modelled NOX concentrations to NO2 concentrations using Defra’s NOx-NO2

calculator v7.1;

▪ Addition of Defra background concentrations to the modelled concentrations with the background

road sector contribution removed (Motorways, Trunk Roads and Primary Roads) to avoid double

counting of the road source component;

▪ Verification and adjustment of modelled road-NOx contributions from the assessed road network

through analysing the ADMS-Roads modelled road-NOx outputs versus scheme specific

monitored road-NOx for the base year scenario (2018);

▪ Comparison of predicted NO2, PM10 and PM2.5 concentrations at all receptors to the relevant air

quality Objectives in each scenario; and

▪ Analysis of changes in pollutant concentrations between the Do Minimum and Do Something

scenarios to assess the significance of impacts associated with the Proposed Scheme on local

air quality.

TRAFFIC DATA

Updated Model

Recalibration of Base Year Model

The Northamptonshire Strategic Transport Model (NSTM) base output has been recalibrated using

traffic counts undertaken in May 2019 for an average weekday at key locations in proximity to the

Proposed Scheme. The model was calibrated and validated based on the WebTAG guidelines and

is considered to be more robust in the forecasting future of traffic than the data used to inform the

Environmental Statement in July 2019.

Forecast Traffic

Forecasting of traffic on the Proposed Scheme and wider network has been completed as follows:

• Considering uncertainty logs for the traffic originating/terminating from the Proposed Scheme

for the year 2021 and 2031 in the influence area and wider area of the scheme;

• Assigning traffic using the SATURN assignment software for the forecast years and

assessment scenarios; and

• The air quality flows were estimated based on the factors developed from the Automatic Traffic

Counts (ATC) data in the proximity of the Proposed Scheme.

Model Scenarios

Traffic flow data comprising period breakdown, traffic composition (percentage HDVs) and average

link speeds (km/h) were used in the modelling as provided for the assessed road network.

Traffic flow data were provided by the project transport planning consultants (WSP). Data were

provided for the following scenarios:

▪ Base/verification year of 2018;

▪ Do-minimum 2021;

▪ Do-something 2021;


▪ Do-minimum 2031; and

▪ Do-something 2031.

Data Screening

Traffic data was screened in accordance with the DMRB LA 105 guidance as stated below:

▪ A change in annual average daily traffic (AADT) >=1,000; or

▪ A change in heavy duty vehicle (HDV) AADT >=200; or

▪ A change in speed band; or

▪ A change in carriageway alignment by >=5m

TIME VARYING EMISSIONS

The ADMS-Roads model uses an hourly traffic flow based on the daily (AADT) flows. Traffic flows

follow a diurnal variation profile throughout the day, thus emissions throughout a 24-hour period will

be weighted according to the profile of traffic during peak, inter-peak and off-peak periods.

To account for this, traffic data were provided as hourly averages representative of the

aforementioned periods (AM peak, PM peak, inter peak and off peak). Therefore, vehicle emission

sources relative to each period of the day were assigned to the appropriate hours within ADMS-

Roads, thus allowing the respective time-varying nature of emissions to be simulated within

dispersion modelling. A diurnal profile was used in the model to replicate how the average hourly

traffic flow varies throughout weekdays and on Saturday and Sunday. Various ATC were conducted

in association with the Proposed Scheme. A diurnal profile was created by taking an average of the

closest ATC locations to the Proposed Scheme based on surveys completed in September

METEOROLOGICAL DATA

ADMS-Roads utilises hourly sequential meteorological data; including wind direction, wind speed,

temperature, precipitation and cloud cover, to facilitate the prediction of pollution dispersion between

source and receptor.

Meteorological data input to the model were obtained from the closest meteorological station,

Bedford, for the year 2018. The 2018 data was used to be consistent with the base/verification traffic

year and were applied to the remaining scenarios for the local air quality assessment. The 2018

wind rose is presented as Figure 4.4-1.


Figure 4.4-1- Bedford 2018 Wind Rose based on hourly data

CONVERSION OF NOX TO NO2

NOx concentrations were predicted using the ADMS-Roads model. The modelled road contribution

of NOx at the modelled receptor locations was then converted to NO2 using the NOx to NO2

calculator (v7.1), in accordance with Defra guidance.

MODEL VERIFICATION METHODOLOGY

MODEL VALIDATION

The ADMS-Roads dispersion model has been validated for road traffic assessments and is

considered to be fit for purpose. Model validation undertaken by the software developer CERC is

unlikely to have included validation in the vicinity of the scheme considered in this assessment. It is

therefore necessary to perform a comparison of model results with local monitoring data at relevant

locations.

MODEL VERIFICATION

The comparison of modelled concentrations with local monitored concentrations is a process termed

‘verification’. Model verification investigates the discrepancies between modelled and measured

concentrations, which can arise due to the presence of inaccuracies and/or uncertainties in model

input data, modelling and monitoring data assumptions. The following are examples of potential

causes of such discrepancy:

\\uk.wspgroup.com\central data\Projects\700215xx\70021598 - Northampton NW Relief Road\E Models and Drawings\Environment\Air Quality\Modelling\Met Data\Bedford_18.met

0

0

3

1.5

6

3.1

10

5.1

16

8.2

(knots)

(m/s)

Wind speed

0° 10°20°

30°

40°

50°

60°

70°

80°

90°

100°

110°

120°

130°

140°

150°

160°170°180°190°

200°

210°

220°

230°

240°

250°

260°

270°

280°

290°

300°

310°

320°

330°

340°350°

100

200

300

400

500

600


▪ Estimates of background pollutant concentrations;

▪ Meteorological data uncertainties;

▪ Traffic data uncertainties;

▪ Model input parameters, such as ‘roughness length; and

▪ Overall limitations of the dispersion model.

Full details of the model verification process specific to the Proposed Scheme modelling

assessment are provided in the ‘Assessment Verification Methodology’ section below.

MODEL PRECISION

Residual uncertainty may remain after systematic error or ‘model accuracy’ has been accounted for

in the final predictions. Residual uncertainty may be considered synonymous with the ‘precision’ of

the model predictions, i.e. how wide the scatter or residual variability of the predicted values

compare with the monitored true value, once systematic error has been allowed for. The

quantification of model precision provides an estimate of how the final predictions may deviate from

true (monitored) values at the same location over the same period. Suitable local monitoring data for

the purpose of verification is used for model verification.

An evaluation of model performance has been undertaken to establish confidence in model results.

LAQM.TG16 identifies a number of statistical procedures that are appropriate to evaluate model

performance and assess the uncertainty. The statistical parameters used in this assessment are:

▪ Root mean square error (RMSE);

▪ Fractional bias (FB); and

▪ Correlation coefficient (CC)

A brief explanation of each statistic is provided in


Table 4.4-1 and further details can be found in Defra’s LAQM.TG16 document.


Table 4.4-1 – Model Performance Statistics

Statistical Parameter Comments Ideal Value

RMSE RMSE is used to define the average error or uncertainty of the model.

The units of RMSE are the same as the quantities compared.

If the RMSE values are higher than 25% of the Objective being assessed, it is recommended that the model inputs and verification should be revisited in order to make improvements.

For example, if the model predictions are for the annual mean NO2 Objective of 40µg/m3, if an RMSE of 10 µg/m3 or above is determined for a model it is advised to revisit the model parameters and model verification.

Ideally an RMSE within 10% of the air quality Objective would be derived, which equates to 4 µg/m3 for the annual mean NO2 Objective.

0.0

FB It is used to identify if the model shows a systematic tendency to over or under predict. FB values vary between +2 and -2 and has an ideal value of zero. Negative values suggest a model over-prediction and positive values suggest a model under-prediction.

0.0

CC It is used to measure the linear relationship between predicted and observed data. A value of zero means no relationship and a value of 1 means absolute relationship. This statistic can be particularly useful when comparing a large number of model and observed data points.

1.00

The calculations were carried out after model adjustment to provide information on the improvement

of the model predictions as a result of the application of the verification adjustment factors.

ASSESSMENT VERIFICATION METHODOLOGY

The verification process involves a review of the modelled pollutant concentrations against

corresponding monitoring data to determine how well the air quality model has performed.

Depending on the outcome it may be considered that the model has performed adequately and that

there is no need to adjust any of the modelled results LAQM.TG (16).

Alternatively, the model may perform outside of the ideal performance limits as stated by

LAQM.TG16 (i.e. model agrees within +/-25% of monitored equivalent). There is then a need to

check all the input data to ensure that it is reasonable and accurately represented in the air quality

modelling process.

Where all input data, such as traffic data, emissions rates, and background concentrations have

been checked and considered as reasonable, then the modelled results require adjustment to best

align with the monitoring data. The adjustment was applied to the NOx road source contribution

(road-NOx) and not total NO2, given that ADMS-Roads was used to predict road-NOx only. This

ensured that any adjustment was applied to road-NOx prior to being used in the NOx to NO2

conversion process.


ASSESSMENT VERIFICATION RESULTS

As outlined in Section 4.2 of the Environmental Statement, NBC have applied a mixture of local and

national bias adjustment factors to raw monitoring data in the last five Defra report submissions.

During consultation, NBC have requested that sensitivity testing on modelled results, is undertaken

using monitoring data adjusted by both local and national bias adjustment factors to determine the

impact on the model results. A bespoke study was previously carried out by Bureau Veritas1 to

provide transparency as to the reasoning for applying a local bias adjustment factor to monitored

results for 2018 and 2019, and to compare these results to those obtained using a national bias

adjustment factor.

Predicted concentration results produced using an urban background site derived local bias factor

(Spring Hill AURN), a roadside site derived local bias factor (Wellingborough Road AURN) and a

national bias factor adjusted monitoring data have been compared.

The modelled and monitored NO2 concentrations for the monitoring locations used for the

verification process are presented as follows:

▪ Before road-NOx adjustment:

• Table 4.4-2 Model Verification Before Adjustment (Local Bias Factor (0.79))

• Table 4.4-3 Model Verification Before Adjustment (Local Bias Factor (0.85))

• Table 4.4-4 Model Verification Before Adjustment (National Bias Factor (0.93))

▪ After road-NOx adjustment:

• Table 4.4-5 Model Verification After Adjustment (Local Bias Factor (0.79))

• Table 4.4-6 Model Verification After Adjustment (Local Bias Factor (0.85))

• Table 4.4-7 Model Verification After Adjustment (National Bias Factor (0.93))

The initial comparison between the total modelled and monitored annual mean NO2 concentration

values illustrate that the model tends to under predict NO2 concentrations within the study area.

1 Northampton Borough Council 2019 Diffusion Tube Bias Adjustment – Local Air Quality Management December 2019


Model Verification Before Adjustment

Table 4.4-2 – Model Verification Before Adjustment (Local Bias Factor (0.79))

Monitoring Site ID

2018 Measured Data (µg/m3)

Measured Road-NOx (µg/m3) from NOx:NO2 Calculator

Modelled road-NOx (µg/m3)- Before adjustment

Modelled Annual Mean NO2 Concentration (µg/m3)- Before adjustment

% Difference (Measured vs Monitored NO2)

NLR2 14.4 10.8 3.6 10.5 -26.7

NLR3 16.5 15.7 2.4 9.5 -42.7

NLR4 13.5 9.7 3.4 10.2 -24.9

NLR7 15.2 6.9 2.9 12.8 -15.7

NLR8 11.0 -0.2 8.1 15.5 41.4

NLR9 15.8 8.3 3.5 13.0 -17.7

NLR10 18.3 16.4 2.9 11.2 -38.6

8 33.6 43.9 5.7 15.2 -54.9

16 25.4 28.3 7.4 14.9 -41.2

17 30.0 34.6 8.9 17.5 -41.6

18 26.0 26.0 9.2 17.6 -32.2

19 25.4 24.8 6.0 16.0 -37.0

20 31.9 38.8 8.1 17.1 -46.4

21 29.3 33.1 5.6 15.8 -46.1

22 30.6 35.9 6.5 16.2 -47.0

23 34.9 45.5 8.8 17.4 -50.1

24 24.3 22.5 7.0 16.5 -32.1

25 26.9 29.2 7.5 16.1 -40.3

26 36.9 50.2 5.2 15.6 -57.8


Monitoring Site ID






27 31.0 36.8 4.6 15.2 -50.8

28 26.7 27.5 6.7 16.3 -38.8

29 32.1 39.2 7.6 16.8 -47.6

36 33.9 40.2 8.8 18.9 -44.2

37 32.0 35.9 6.8 17.9 -44.2

38 34.9 42.4 8.0 18.5 -47.0

39 28.3 27.9 6.3 17.6 -37.7

40 34.4 41.3 9.0 19.0 -44.7

46 40.0 58.9 9.3 17.0 -57.4

50 30.1 36.1 6.9 15.8 -47.5

51 29.6 35.0 4.3 14.4 -51.5

52 36.0 49.4 8.5 16.6 -53.9

53 35.7 48.7 8.2 16.4 -54.0

54 37.7 53.4 11.6 18.2 -51.7

55 34.4 45.7 10.6 17.7 -48.7

56 31.5 39.2 8.8 16.8 -46.8

57 47.1 75.9 10.4 18.0 -61.9

59 33.4 42.7 7.1 16.3 -51.2

60 36.5 49.9 7.5 16.5 -54.9

61 30.7 36.7 5.3 15.4 -50.0


Monitoring Site ID






62 26.1 26.8 5.4 15.4 -41.1

79 28.9 36.0 2.6 12.3 -57.5

82 23.0 22.6 5.1 13.8 -39.9

* Annualised mean concentration (2018). See Appendix 4.3


Table 4.4-3 –Model Verification Before Adjustment (Local Bias Factor (0.85))

Monitoring Site ID






NLR2 15.5 12.9 3.6 10.5 -31.9

NLR3 17.8 18.2 2.4 9.5 -46.7

NLR4 14.5 11.7 3.4 10.2 -30.2

NLR7 16.4 9.2 2.9 12.8 -21.7

NLR8 11.8 0.6 8.1 15.5 31.4

NLR9 17.0 10.6 3.5 13.0 -23.5

NLR10 19.7 19.1 2.9 11.2 -42.9

8 36.2 49.8 5.7 15.2 -58.1

16 27.3 32.4 7.4 14.9 -45.4

17 32.3 39.6 8.9 17.5 -45.7

18 28.0 30.2 9.2 17.6 -36.9

19 27.3 28.8 6.0 16.0 -41.5

20 34.3 44.2 8.1 17.1 -50.2

21 31.5 37.9 5.6 15.8 -49.9

22 32.9 41.1 6.5 16.2 -50.7

23 37.6 51.7 8.8 17.4 -53.6

24 26.1 26.3 7.0 16.5 -36.9

25 28.9 33.6 7.5 16.1 -44.5

26 39.7 56.8 5.2 15.6 -60.7

27 33.4 42.0 4.6 15.2 -54.3


Monitoring Site ID






28 28.7 31.8 6.7 16.3 -43.1

29 34.5 44.7 7.6 16.8 -51.3

36 36.5 46.0 8.8 18.9 -48.1

37 34.4 41.4 6.8 17.9 -48.1

38 37.6 48.5 8.0 18.5 -50.7

39 30.4 32.5 6.3 17.6 -42.1

40 37.0 47.3 9.0 19.0 -48.6

46 43.0 66.4 9.3 17.0 -60.4

50 32.4 41.2 6.9 15.8 -51.2

51 31.8 40.0 4.3 14.4 -54.9

52 38.7 55.9 8.5 16.6 -57.2

53 38.4 55.1 8.2 16.4 -57.3

54 40.6 60.3 11.6 18.2 -55.1

55 37.0 51.8 10.6 17.7 -52.3

56 33.9 44.6 8.8 16.8 -50.6

57 50.7 85.3 10.4 18.0 -64.6

59 35.9 48.5 7.1 16.3 -54.7

60 39.3 56.4 7.5 16.5 -58.1

61 33.0 41.9 5.3 15.4 -53.5

62 28.1 31.0 5.4 15.4 -45.3


Monitoring Site ID






79 31.1 40.8 2.6 12.3 -60.5

82 24.7 26.2 5.1 13.8 -44.1


Table 4.4-4 –Model Verification Before Adjustment (National Bias Factor (0.93))

Monitoring Site ID






NLR2 16.9 15.7 3.6 10.5 -37.8

NLR3 19.4 21.5 2.4 9.5 -51.3

NLR4 15.9 14.3 3.4 10.2 -36.2

NLR7 17.9 12.2 2.9 12.8 -28.4

NLR8 12.9 2.7 8.1 15.5 20.1

NLR9 18.5 13.7 3.5 13.0 -30.1

NLR10 21.5 22.9 2.9 11.2 -47.9

8 39.5 57.7 5.7 15.2 -61.6

16 29.9 38.0 7.4 14.9 -50.1

17 35.3 46.5 8.9 17.5 -50.4

18 30.7 36.1 9.2 17.6 -42.5

19 29.9 34.4 6.0 16.0 -46.5

20 37.6 51.8 8.1 17.1 -54.5

21 34.5 44.6 5.6 15.8 -54.3

22 36.0 48.1 6.5 16.2 -54.9

23 41.1 60.2 8.8 17.4 -57.6

24 28.6 31.5 7.0 16.5 -42.3

25 31.7 39.7 7.5 16.1 -49.3

26 43.4 65.8 5.2 15.6 -64.1

27 36.5 49.2 4.6 15.2 -58.2


Monitoring Site ID






28 31.5 37.9 6.7 16.3 -48.1

29 37.7 52.0 7.6 16.8 -55.4

36 39.9 54.1 8.8 18.9 -52.6

37 37.7 48.9 6.8 17.9 -52.6

38 41.1 57.0 8.0 18.5 -55.0

39 33.4 39.0 6.3 17.6 -47.2

40 40.5 55.5 9.0 19.0 -53.0

46 47.1 76.7 9.3 17.0 -63.8

50 35.4 48.0 6.9 15.8 -55.4

51 34.8 46.7 4.3 14.4 -58.7

52 42.3 64.5 8.5 16.6 -60.8

53 42.0 63.8 8.2 16.4 -60.9

54 44.3 69.5 11.6 18.2 -58.9

55 40.5 60.1 10.6 17.7 -56.4

56 37.1 52.0 8.8 16.8 -54.9

57 55.5 98.4 10.4 18.0 -67.6

59 39.3 56.5 7.1 16.3 -58.5

60 43.0 65.5 7.5 16.5 -61.7

61 36.1 48.9 5.3 15.4 -57.5

62 30.8 36.9 5.4 15.4 -50.1


Monitoring Site ID






79 34.0 47.4 2.6 12.3 -63.9

82 27.0 31.0 5.1 13.8 -48.8

Due to the model tendenacy to under predict NO2 concentrations within the study area for both local

bias adjusted and national bias adjusted results, verification and adjustment of modelled road-NOx

was undertaken in accordance with LAQM.TG(16) for these monitoring locations

Comparison of Measured Road-NOx With Unadjusted Modelled Road-NOx

Adjustment factors were calculated for both local bias adjusted and national bias adjusted results.

The road-NOx adjustment factor was determined as the slope of the best fit line between the

‘measured’ road contribution and the model derived road contribution, forced through zero (Figures

4.4-2 and 4.4-3). This factor was then applied to the modelled road-NOx concentration for each

monitoring site to provide adjusted modelled road-NOx concentrations.

Figure 4.4-2- Comparison of Measured Road-NOx with Unadjusted Modelled Road-NOx using

the Local Bias Factor (0.79)

y = 4.9661x

0

10

20

30

40

50

60

70

80

0 2 4 6 8 10 12 14 16

Measure

d r

oad

-NO

x(µ

g/m

3)

Unadjusted modelled Road-NOx (µg/m3)



the Local Bias Factor (0.85)


the National Bias Factor (0.93)

These unadjusted modelled results were compared with the corresponding monitored data. Any

monitoring location with a disparity between monitored and modelled data of +/- 25% was removed

from the data set and the adjustment factor was re-calculated, in line with the methodology outlined

in LAQM.TG(16). This process was then repeated until every diffusion tube was within this range.

The modelled road-NOx predictions required to align with the monitoring data were adjusted using

the three methods are as follows:

▪ Urban background site derived local bias factor (Spring Hill AURN) * 5.24

▪ Roadside site local bias factor (Wellingborough Road AURN) * 5.95

▪ National bias factor * 6.95

The smallest road-NOx adjustment factor was required to align with the monitoring data bias

corrected using the urban background site (Spring Hill AURN).

y = 5.6672x

0

10

20

30

40

50

60

70

80

90

0 2 4 6 8 10 12 14 16

Measure

d r

oad

-NO

x (

µg/m

3)


y = 6.6323x

0

20

40

60

80

100

120

0 2 4 6 8 10 12 14 16

Measure

d r

oad

-NO

x(µ

g/m

3)



Model Verification After Adjustment

The results of the model verification process for both local bias and national bias adjusted data can

be seen in Tables 4.4-4 and 4.4-5.

Table 4.4-5 –Model Verification After Adjustment (Local Bias Factor (0.79))

Monitoring Site ID 2018 Monitored NO2(µg/m3)

Modelled NO2 (µg/m3)- After adjustment


NLR3 16.5 14.8 -10.4

NLR10 18.3 17.6 -3.8

8 33.6 27.3 -18.9

16 25.4 30.3 19.3

17 30.0 35.5 18.3

19 25.4 28.6 12.7

20 31.9 33.6 5.3

21 29.3 27.6 -5.8

22 30.6 29.7 -3.0

23 34.9 35.1 0.5

25 26.9 31.4 16.8

27 31.0 25.1 -19.2

28 26.7 30.2 13.1

29 32.1 32.3 0.7

36 33.9 36.6 7.9

37 32.0 31.8 -0.7

38 34.9 34.7 -0.6

39 28.3 30.8 8.7

40 34.4 37.1 7.7





46 40.0 35.7 -10.7

50 30.1 30.2 0.3

51 29.6 23.5 -20.5

52 36.0 33.8 -6.3

53 35.7 33.0 -7.5

54 37.7 40.8 8.3

55 34.4 38.5 11.8

56 31.5 34.5 9.6

57 47.1 38.4 -18.4

59 33.4 31.0 -7.1

60 36.5 31.8 -12.9

61 30.7 26.7 -13.1

62 26.1 26.8 2.6

82 23.0 24.7 7.4


Table 4.4-6 –Model Verification After Adjustment (Local Bias Factor (0.85))




NLR3 16.5 14.8 -10.4

NLR10 18.3 17.6 -3.8

8 33.6 27.3 -18.9

16 25.4 30.3 19.3

17 30.0 35.5 18.3

19 25.4 28.6 12.7

20 31.9 33.6 5.3

21 29.3 27.6 -5.8

22 30.6 29.7 -3.0

23 34.9 35.1 0.5

25 26.9 31.4 16.8

27 31.0 25.1 -19.2

28 26.7 30.2 13.1

29 32.1 32.3 0.7

36 33.9 36.6 7.9

37 32.0 31.8 -0.7

38 34.9 34.7 -0.6

39 28.3 30.8 8.7

40 34.4 37.1 7.7

46 40.0 35.7 -10.7

50 30.1 30.2 0.3





51 29.6 23.5 -20.5

52 36.0 33.8 -6.3

53 35.7 33.0 -7.5

54 37.7 40.8 8.3

55 34.4 38.5 11.8

56 31.5 34.5 9.6

57 47.1 38.4 -18.4

59 33.4 31.0 -7.1

60 36.5 31.8 -12.9

61 30.7 26.7 -13.1

62 26.1 26.8 2.6

82 23.0 24.7 7.4


Table 4.4-7 – Model Verification After Adjustment (National Bias Factor (0.93))




NLR3 19.4 16.9 -13.2

NLR7 17.9 21.6 20.4

NLR10 21.5 20.1 -6.8

8 39.5 31.8 -19.6

16 29.9 35.9 20.0

17 35.3 41.9 18.8

19 29.9 33.3 11.4

20 37.6 39.6 5.2

21 34.5 32.0 -7.2

22 36.0 34.6 -3.8

23 41.1 41.4 0.8

25 31.7 37.0 16.7

27 36.5 28.8 -21.2

28 31.5 35.3 12.0

29 37.7 38.0 0.7

36 39.9 42.9 7.5

37 37.7 36.9 -2.1

38 41.1 40.6 -1.3

39 33.4 35.6 6.6

40 40.5 43.5 7.4

46 47.1 42.4 -10.0





50 35.4 35.5 0.1

51 34.8 27.0 -22.4

52 42.3 39.9 -5.6

53 42.0 39.0 -7.1

54 44.3 48.7 10.0

55 40.5 45.8 13.1

56 37.1 40.9 10.2

57 55.5 45.7 -17.7

59 39.3 36.4 -7.3

60 43.0 37.4 -13.1

61 36.1 30.9 -14.3

62 30.8 31.0 0.8

82 27.0 28.8 6.7

SUMMARY

The summary of the verification outcome and model performance statistics for both local bias

adjusted and national bias adjusted results, as defined in LAQM.TG(16), are provided in Table 4.4-

8.

Table 4.4-8 - Summary of verified model performance and uncertainty statistics

Bias Adjustment Factor Applied

No. of monitoring sites

No. sites within +/-25%

No. sites within +/- 10%

Root Mean Square Error*

µg/m3 % of Objective

Local (0.79) 33 33 19 ±4.37 10.9

Local (0.85) 34 34 19 ±4.77 11.9

National (0.93) 34 34 17 ±5.31 13.3

To assess the uncertainty of a model, the RMSE is the simplest parameter to calculate. It provides

an estimate of the average error of the model in the same units as the model predictions. It is also


often easier to interpret the RMSE than the other statistical parameter and therefore it has been

calculated in this assessment to understand the model uncertainty.

LAQM.TG(16) states that ‘If the RMSE values are higher than ±25% of the objective being

assessed, it is recommended that the model inputs and verification should be revisited in order to

make improvements. For example, if the model predictions are for the annual mean NO2 objective of

40µg/m3, if an RMSE of 10µg/m3 or above is determined for a model, the local authority would be

advised to revisit the model parameters and model verification. Ideally an RMSE within 10% of the

air quality objective would be derived, which equates to 4µg/m3 for the annual average NO2

objective.’

As shown in Tables 4.4-8, the local bias adjusted monitoring data (0.79) produces a lower RMSE. In

addition, the adjustment factor of 5.24 derived for the local (roadside) bias adjustment data (0.79) is

lower than the local (urban background) bias adjustment (0.85) 5.96 and national bias adjusted

factor data (0.93) of 6.95, meaning the modelled results are closer to the locally adjusted values.

With an RMSE at 10.9% of the objective and over half of the modelled predictions within 10% of the

monitored concentrations, the model is shown to perform well and is suitable for its intended use

when and with the local bias factor found to be the most appropriate. Therefore, these local bias

adjusted monitoring data (0.79) has been used in this assessment used to derive the primary

results. This is in conjunction with the previous findings of the study carried out by Bureau Veritas1.

In the absence of scheme-specific PM10 and PM2.5 monitoring, the derived model adjustment factor

for road-NOx was applied to modelled road contributions for these pollutants.

Three White Rose Office Park

Millshaw Park Lane

Leeds

LS11 0DL

wsp.com

appendix 4 · 2020. 6. 15. · appendix 4.4 public | wsp project no.: 70068030 june 2020...

Documents