28
Work Order No. 582-19-91533-01 Contract No. 582-19-90500 Tracking No. 2019-13 Prepared for: Texas Commission on Environmental Quality 12100 Park 35 Circle MC 164 Austin, TX 78753 Prepared by: Ramboll US Corporation 7250 Redwood Blvd., Suite 105 Novato, California 94945 August 15, 2019 Regional Haze Modeling Final Report PREPARED UNDER A CONTRACT FROM THE TEXAS COMMISSION ON ENVIRONMENTAL QUALITY The preparation of this document was financed through a contract from the State of Texas through the Texas Commission on Environmental Quality. The content, findings, opinions and conclusions are the work of the author(s) and do not necessarily represent findings, opinions or conclusions of the TCEQ.

Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

  • Upload
    others

  • View
    1

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Work Order No. 582-19-91533-01 Contract No. 582-19-90500

Tracking No. 2019-13

Prepared for:

Texas Commission on Environmental Quality 12100 Park 35 Circle MC 164 Austin, TX 78753 Prepared by:

Ramboll US Corporation 7250 Redwood Blvd., Suite 105 Novato, California 94945 August 15, 2019

Regional Haze Modeling Final Report

PREPARED UNDER A CONTRACT FROM THE TEXAS COMMISSION ON ENVIRONMENTAL QUALITY The preparation of this document was financed through a contract from the State of Texas through the Texas Commission on Environmental Quality. The content, findings, opinions and conclusions are the work of the author(s) and do not necessarily represent findings, opinions or conclusions of the TCEQ.

Page 2: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll 7250 Redwood Boulevard Suite 105 Novato, CA 94945 USA T +1 415 899 0700 https://ramboll.com

Regional Haze Modeling Final Report

Page 3: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

i

Contents

Introduction 1 1.1 Background 1 1.2 Purpose and Objectives 2

Methodology 3 2.1 Meteorology 3 2.2 Boundary Concentrations 3 2.3 Emission Inputs 3 2.3.1 Other Natural Emissions 4 2.3.2 EDGAR Inventory 4 2.3.3 Emission Summary 5 2.4 CAMx Setup 6

CAMx Results 7 3.1 Monthly average PM2.5 7 3.2 Model Performance Evaluation 8 3.2.1 Ozone Evaluation 9 3.2.2 PM2.5 Evaluation 9

Conclusion and Recommendation 14

References 15 Appendix Appendix A: Monthly Average Concentrations (µg/m3) of PM2.5 Components Appendix B: Spatial Maps of NMB for PM2.5 and its Components

Table of Figures Figure 1-1. Class I areas in Texas and nearby states. 1 Figure 1-2. CAMx 36-km and 12-km modeling domains 2 Figure 3-1. Monthly average PM2.5 (µg/m3) for January, April and

September 8 Figure 3-2. Spatial plots of ozone performance statistics based on the 2016

CAMx model simulation. NMB (left) and NME (right); April (top) and September (bottom). 9

Figure 3-3. Example: PM2.5 NMB (top) and NME (bottom) in April at selected IMPROVE sites. Performance goal and criteria are presented in yellow and red, respectively. 11

Figure 3-4. Example: comparison between observed and modeled concentrations at selected IMPROVE sites on most impaired days. 13

Table of Tables Table 2-1. Sectors included in EDGAR SMOKE processing 5

Page 4: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

ii

Table 2-2. Summary of January average daily emissions (tons/day) in the 36-km domain 5

Table 2-3. Summary of April average daily emissions (tons/day) in the 36-km domain 5

Table 2-4. Summary of September average daily emissions (tons/day) in the 36-km domain 6

Table 2-5. CAMx model configurations. 6 Table 3-1. Statistical model performance evaluation measure definitions

and performance benchmarks. 8 Table 3-2. Model performance metrics for PM2.5 and its component species

at IMPROVE and CSN sites. 10

Page 5: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

1

INTRODUCTION

1.1 Background

The Regional Haze Rule (RHR) requires State Implementation Plans (SIPs) that evaluate reasonable progress toward improved visibility at Class I areas. The next Regional Haze SIP is due in 2021 for the second implementation period covering 2019-2028. The Texas Commission on Environmental Quality (TCEQ) plans to use the Comprehensive Air quality Model with Extensions (CAMx; Ramboll, 2018) to assess visibility at individual Class I areas in and near Texas (Figure 1-1) as required by the RHR. CAMx is a photochemical grid model designed to simulate the formation, transport and fate of particulate matter (PM), oxidants and deposition over spatial scales ranging from urban to hemispheric.

Figure 1-1. Class I areas in Texas and nearby states.

TCEQ plans to rely on emission data compiled by the Environmental Protection Agency (EPA) for the 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative activity that leverages efforts of EPA, States and multi-jurisdictional planning organizations (MJOs). Currently, the 2016 MP version is beta and we anticipate the final 2016 MP will be available by September 2019. The 2016 MP includes an expanded 36 km North American domain that stretches

1 National Emissions Inventory Collaborative (2019). 2016beta Emissions Modeling Platform. Retrieved from http://views.cira.colostate.edu/wiki/wiki/10197 .

Page 6: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

2

from the Alaska Panhandle south to include all of Mexico and East to include Newfoundland (Figure 1-2).

Figure 1-2. CAMx 36-km and 12-km modeling domains

1.2 Purpose and Objectives

The purpose of this study is to provide technical support to TCEQ in visibility analysis for the RHR SIP development. The original scope of work includes 1) development of a project work plan, 2) assistance to TCEQ staff in setting up CAMx simulations with source apportionment, 3) analysis of model results for visibility assessments, and 4) development of a final project report. Due to delayed release of the EPA’s 2016 MP, with TCEQ’s approval, our scope of work and project schedule were adjusted to align with availability of this data. We expect modeling activities to resume in the coming Texas fiscal year 2019-2020.

The adjusted scope of work includes performing CAMx modeling for three months (January, April and September) using the EPA 2016 beta emissions and other modeling inputs provided by TCEQ or developed by Ramboll. The purpose is to develop the modeling system and obtain a preliminary assessment of its performance. We evaluated model for particulate matter (PM) against observations at the Interagency Monitoring of Protected Visual Environments (IMPROVE) monitoring sites. In addition to the preliminary 3-month CAMx simulation, Ramboll also assisted TCEQ in developing several modeling inputs for the full 2016 calendar year.

Page 7: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

3

METHODOLOGY Ramboll performed CAMx simulations for three months (January, April, and September) using EPA 2016 beta emissions data and 2016 meteorology data provided by the TCEQ. TCEQ also provided biogenic emissions. We used TCEQ’s meteorology to derive other natural source emissions except those from fires. We evaluated CAMx performance by comparing the 2016 simulation to observed air quality at IMPROVE sites. We provided recommendations for the annual 2016 simulation that TCEQ will perform for SIP development.

2.1 Meteorology

TCEQ provided output from Weather Research and Forecasting (WRF) model simulation at 12-km resolution covering the 36-km domain extent. We used WRFCAMx version 4.7 to prepare CAMx meteorology inputs for both 36-km and 12-km modeling domains (Figure 1-2).

2.2 Boundary Concentrations

Boundary Concentrations (BCs) for the CAMx 36-km domain (Figure 1-2) were extracted from a 2016 simulation using the GEOS-Chem global chemistry model. Under a separate study, Ramboll performed the 2016 and 2028 GEOS-Chem simulations to provide BCs for CAMx that quantify international contributions to PM and regional haze (Ramboll and EPRI, 2017). We used GEOS-Chem version 11-02rc and a standard chemical mechanism configuration, which includes detailed tropospheric and stratospheric chemistry (UCX mechanism) along with complex secondary organic aerosol (SOA) chemistry. The global GEOS-Chem model simulates atmospheric chemical and physical processes driven by NASA’s GEOS-FP (‘forward processing’) reanalysis meteorology. The GEOS-Chem covers the globe with horizontal grid resolution of 2°x2.5° and 72 vertical layers from ground level into the stratosphere.

The 2016 GEOS-Chem output was processed using the geos2camx tool to generate BCs for the 36-km CAMx domain. The following processing steps are needed to generate 3-hourly lateral and top BC input files for CAMx: (1) horizontal interpolation from the ~250-km resolution GEOS-Chem grid to the 36-km resolution CAMx grid; (2) vertical interpolation of the 72 layer GEOS-Chem output to the 29-layer CAMx grid; and (3) species mapping from the GEOS-Chem chemical species to the CB6-CF chemistry scheme.

2.3 Emission Inputs

The 2016beta Emissions Modeling Platform (EMP) is the first product from the National Emissions Inventory Collaborative that includes a full suite of base year (2016) inventories, ancillary emissions data, and pre-merged model-ready emissions for air quality modeling2. Details of the 2016beta platform development are available on its wiki3.

The 2016beta emissions are the basis for the emissions for this study. Ramboll has obtained the 2016beta EMP gridded, pre-merged emissions from the EPA for this study through the Intermountain West Data Warehouse (IWDW). The emissions dataset includes a full set of anthropogenic emissions for US, Canada and Mexico and fire emissions. TCEQ provided biogenic emissions developed using the Biogenic Emissions Information System (BEIS) model. Ramboll developed emissions from other

2 National Emissions Inventory Collaborative (2019). 2016beta Emissions Modeling Platform. Retrieved from http://views.cira.colostate.edu/wiki/wiki/10197. 3 3 http://views.cira.colostate.edu/wiki/wiki/9169

Page 8: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

4

natural sources not included in the EPA’s 2016 MP including windblown dust (WBD), lightning NOx (LNOx) and oceanic emissions including sea salt and dimethyl sulfide (DMS).

The final CAMx emissions include 1) the EPA’s 2016 MP emission files excluding biogenic emissions, 2) TCEQ’s BEIS emissions, 3) other natural emissions (i.e., lightning NOx, DMS), and 4) the EDGAR global emissions for regions not covered by the EPA 2016beta platform.

2.3.1 Other Natural Emissions Three CAMx natural emissions processors were run using the 2016 WRF meteorological data to generate CAMx-ready emissions as follows:

• Oceanic emissions processor was used to generate sea salt, Ix and DMS emissions. • Processor for Windblown Dust (WBD) emissions for coarse and fine crustal material. • Lightning NOx (LNOx) emissions processor.

Because the version of CAMx (v6.5) does not include explicit DMS chemistry, DMS was represented as (renamed) sulfur dioxide (SO2) for input to CAMx. This method simplifies the atmospheric conversion of DMS to PM sulfate SO4which could overstate SO4 concentrations at coastal sites.

2.3.2 EDGAR Inventory For anthropogenic emissions from regions outside the 2016 MP we used the 2010 Hemispheric Transport of Air Pollutants v2 (HTAP_v2) gridded global emissions estimates from the EDGAR (Emission Database for Global Atmospheric Research, version 4.2; European Commission, 2011) database which reports emissions world-wide on a 0.1° × 0.1° resolution grid.

We processed EDGAR emissions data for input to CAMx by building upon the SMOKE processing platform developed by EPA for hemispheric CMAQ modeling. SMOKE version 4.0 and later has support for global gridded inventories such as EDGAR. SMOKE can process EDGAR gridded inventories by performing the following steps:

• Read global gridded inventories such as EDGAR and regrid them onto a desired modeling grid • Temporally allocate emissions (using profiles) and time-shift (to account for time zones) to the

model timeframe by using 5-digit country code and time zone assignment for each EDGAR grid cell (SMOKE input file GRIDMASK)

• Place emissions from elevated sources in specific height ranges which is needed because EDGAR lacks data to calculate plume rise (SMOKE program Layalloc)

• Apply chemical speciation by using a single profile per sector, country and pollutant. EPA developed speciation profiles for HTAP by constructing averages of 2011 modeling platform data

The EDGAR SMOKE processing was split into several “streams” listed in Table 2-1. We used monthly gridded emissions files and ancillary data files from EPA’s hemispheric modeling. The processed EDGAR emissions are represented as surface and elevated sources.

Page 9: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

5

Table 2-1. Sectors included in EDGAR SMOKE processing

Sector Description g_ag agriculture NH3 emissions but not agricultural waste burning g_air_cds international and domestic aircraft emissions – climbing-and-descending g_air_crs international and domestic aircraft emissions – cruising g_air_lto international and domestic aircraft emissions – landing and takeoff g_energy power industry g_industry manufacturing, mining, metal, cement, chemical, solvent industry g_residential heating/cooling of buildings and equipment/lighting of buildings and waste treatment g_ships international shipping g_transport ground transport (including road, rail, pipeline, inland waterways)

2.3.3 Emission Summary Summary tables of daily average emissions for each modeling month (January, April and September) are provided in Table 2-2 to 2-4.

Table 2-2. Summary of January average daily emissions (tons/day) in the 36-km domain

VOC NOx CO SO2 NH3 PM2.5 All anthropogenic/fire 65,241 47,060 187,686 19,850 5,896 12,904

MEGAN biogenic 46,202 3,046 10,391 - - -

Wind-blown dust - - - - - 0.3

LNOx - 24,154 - - - -

DMS (SO2)/Sea salt - - - 3,387 - 24,554

EDGAR Inventory 1,311 400 7,332 148 528 201

Total 112,755 74,661 205,409 23,385 6,425 37,659

Table 2-3. Summary of April average daily emissions (tons/day) in the 36-km domain

VOC NOx CO SO2 NH3 PM2.5 All anthropogenic/fire 81,680 52,530 312,647 17,830 17,406 25,187

MEGAN biogenic 161,131 8,252 27,348 - - -

Wind-blown dust - - - - - 9

LNOx - 42,897 - - - -

DMS (SO2)/Sea salt - - - 4,718 - 13,628

EDGAR Inventory 1,328 405 7,427 150 532 204

Total 244,138 104,083 347,422 22,697 17,938 39,028

Page 10: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

6

Table 2-4. Summary of September average daily emissions (tons/day) in the 36-km domain

VOC NOx CO SO2 NH3 PM2.5 All anthropogenic/fire 55,625 47,038 181,074 18,586 14,645 12,579

MEGAN biogenic 257,888 8,557 48,478 - - -

Wind-blown dust - - - - - 8

LNOx - 95,158 - - - -

DMS (SO2)/Sea salt - - - 4,397 - 8,670

EDGAR Inventory 1,375 420 7,689 155 546 211

Total 314,887 151,173 237,241 23,138 15,191 21,467

2.4 CAMx Setup

The CAMx setup and model options are summarized in Table 2-5.

Table 2-5. CAMx model configurations.

Science Options CAMx Version Version 6.50 Vertical Grid Mesh 29 Layers (TCEQ’s CAMx vertical structure) Time Zone UTC Chemistry mechanism CB6r4 gas-phase mechanism and CF PM scheme Horizontal Grids 12 km nested with 36 km Initial Conditions 10 day spin-up for each month Meteorology 2016 WRF meteorology (provided by TCEQ) Photolysis mechanism TUV version 4.8 with TOMS ozone column data Advection Scheme Piecewise Parabolic Method (PPM) Cloud convection scheme On / Relaxed Arakawa-Schubert Planetary Boundary Layer (PBL) mixing K-theory

Dry deposition scheme Zhang

In-line IX emissions Oceanic iodine (Ix) emissions computed from salt water masks provided by TCEQ

Chemistry Solver EBI Parallelization MPI and OMP

Page 11: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

7

CAMx RESULTS

3.1 Monthly average PM2.5

Maps of the monthly average PM2.5 concentrations are shown in Figure 3-1. PM2.5 concentrations are generally lower (mostly below 10 μg m-3) in the Western US. Higher concentrations in the Eastern US have large sulfate and nitrate contributions with additional contributions from organic PM in the south (see maps of PM2.5 components in Appendix A).

January

April

Page 12: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

8

September

Figure 3-1. Monthly average PM2.5 (µg/m3) for January, April and September

3.2 Model Performance Evaluation

We performed an operational model performance evaluation (MPE) for ozone, PM2.5 and PM2.5

components. Table 3-1 lists the definitions of statistical performance measures that were used in model performance evaluation discussed below. We compared the model performance to a set of numerical “goals” and less restrictive “criteria” performance benchmarks recently recommended by Emery et al. (2016) for the Normalized Mean Bias (NMB) and Normalized Mean Error (NME). Another set of benchmarks that are widely used was established by the U.S. Regional Planning Organizations (RPOs) based on Fractional Bias (FB) and Error (FE) (Boylan and Russell, 2006). The purpose of MPE benchmarks is not to give a passing or failing grade, but rather to put results into the proper context of previous model applications that establish what level of performance can be expected realistically.

Table 3-1. Statistical model performance evaluation measure definitions and performance benchmarks.

Statistical Measure

Mathematical Expression

Performance Benchmark

Goals Criteria

Normalized Mean Bias (%), NMB

MDA8 O3 <±5% PM2.5, SO4,NH4 <±10% NO3 <±15% OC <±15% EC <±20%

MDA8 O3 <±15% PM2.5, SO4,NH4 <±30% NO3 <±65% OC <±50% EC <±40%

Normalized Mean Error (%), NME

MDA8 O3 <15% PM2.5, SO4,NH4 <35% NO3 <65% OC <45% EC <50%

MDA8 O3 <25% PM2.5, SO4,NH4 <50% NO3 <155% OC <65% EC <75%

( )

=

=

N

ii

N

iii

O

OP

1

1

=

=

N

ii

N

iii

O

OP

1

1

Page 13: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

9

Statistical Measure

Mathematical Expression

Performance Benchmark

Goals Criteria

Fractionalized Bias (%), FB

24-hr total and speciated PM2.5 <±30%

24-hr total and speciated PM2.5 <±60%

Fractional Error (%), FE

24-hr total and speciated PM2.5 <50%

24-hr total and speciated PM2.5 <75%

3.2.1 Ozone Evaluation

Ozone predictions were compared against measurements obtained from the Clean Air Status and Trends Network (CASTNet). The CAMx 2016 simulation shows reasonable performance for ozone in April but tends to overestimate in September. The NME statistics for the maximum daily average 8-hour (MDA8) ozone meet the recommended error criteria of 25% at most CASTNET sites in both months. The NMB statistics are mostly within ±10% in April and ±20% in September.

NMB NME

Figure 3-2. Spatial plots of ozone performance statistics based on the 2016 CAMx model simulation. NMB (left) and NME (right); April (top) and September (bottom).

3.2.2 PM2.5 Evaluation

The model evaluation for PM2.5 focuses on total PM2.5 mass and its key components including sulfate (SO4), nitrate (NO3), elemental carbon (EC), and organic matter (OM). PM2.5 ambient measurements for 2016 were obtained from the IMPROVE and Chemical Speciation Monitoring Network (CSN). The IMPROVE and CSN network provide 24‐hour average concentrations every 3 days and 3 or 6 days, respectively. The PM2.5 performance statistics were calculated for each month (Table 3-2).

∑=

+−N

i ii

ii

OPOP

N 1

2

∑= +

−N

i ii

ii

OPOP

N 1

2

Page 14: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

10

PM2.5 performance is variable across regions and modeling month. In April and September, NMB is lower than 15% and NME is about 50% (compared to ±30% and 50% Criteria) at IMPROVE network. January has weaker performance with NME of 58% due mainly to overestimated OM. Spatial plots of PM performance statistics are provided in Appendix B.

PM2.5 SO4 performance, in general, is relatively good. Average SO4 NMB and FB are in the range from −11 to 35% at IMPROVE network. We do not see evidence of overstated SO4 at coastal sites that could imply a problem representing DMS with SO2. The model tends to overpredict NO3 in the southern and eastern US in January when NO3 concentrations are higher. The model overestimates OM in January (NMB of 90% at IMPROVE) and in September (NMB of 97%). Such summer OM overestimation may be due in part to overstated Secondary Organic Aerosol (SOA) from the BEIS terpene emissions as seen in previous modeling studies. The overestimation of OM in winter is more evident at CSN (urban sites) which may suggest overstated combustion emissions (e.g., residential wood) prevalent in urban areas. The model also tends to overpredict PM2.5 elemental carbon (EC) concentrations but their concentrations at IMPROVE are generally small (average values less than 0.2 µg/m3).

Table 3-2. Model performance metrics for PM2.5 and its component species at IMPROVE and CSN sites.

Specie Month Network No. of Obs. Mean (obs)

Mean

(mod) NMB (%)

NME (%)

FB (%)

FE (%) Correlation

PM2.5 Jan IMPROVE 1228 2.62 4.14 58 83.3 44.5 61.4 0.57 Apr IMPROVE 1110 3.51 3.97 13.1 49.8 5.65 39.7 0.5 Sep IMPROVE 1121 4.01 4.31 7.31 50.7 -2.77 45.5 0.55

PM2.5 SO4 Jan

IMPROVE 1247 0.451 0.542 20.2 50.6 35 53.4 0.75

CSN 727 1.02 1.16 13.6 53.5 20.7 49.3 0.4

Apr IMPROVE 1117 0.617 0.534 -13.5 31 -11.5 34.6 0.76

CSN 667 0.875 1 14.4 35.1 14.5 32 0.65

Sep IMPROVE 1130 0.657 0.536 -18.4 44.9 -22.5 52.1 0.74

CSN* 17 0.772 1.11 44.1 44.4 42.7 43 0.76

PM2.5 Jan

IMPROVE 1247 0.589 0.717 21.8 88.9 19.6 83 0.57

NO3 CSN 756 2.08 2.5 20 69.4 30.2 66.8 0.47

Apr IMPROVE 1117 0.285 0.373 31.1 79.7 -7.02 71.3 0.66

CSN 696 0.633 1.14 80.7 102 52 71.6 0.62

Sep IMPROVE 1130 0.18 0.201 11.5 96.6 -37.2 90.3 0.47

CSN* 38 0.252 0.74 193 202 72.1 83.7 0.37

PM2.5 OM Jan

IMPROVE 1225 1 1.9 90 118 56.5 75.3 0.51

CSN 686 3.05 7.32 140 155 75.3 81.2 0.43

Apr IMPROVE 1138 2.91 1.84 -36.9 91.3 29.4 54.1 0.03

CSN 637 2.3 4.53 96.8 108 63 68.3 0.47

Sep IMPROVE 1164 1.63 2.51 54 84.4 31.1 57.7 0.5

CSN* - - - - - - - -

PM2.5 EC Jan

IMPROVE 1274 0.115 0.219 91.2 122 61.7 84.6 0.68

CSN 686 0.67 1.13 68.2 91.9 45.5 60.9 0.53

Apr IMPROVE 1142 0.124 0.192 54.2 110 50.2 68 0.19

CSN* - - - - - - - - Sep IMPROVE 1161 0.117 0.193 64.6 96.4 39.3 67 0.57

Page 15: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

11

Specie Month Network No. of Obs. Mean (obs)

Mean

(mod) NMB (%)

NME (%)

FB (%)

FE (%) Correlation

CSN* - - - - - - - -

*observation data not available or limited Figures below show an example of additional graphics produced in this study. Figure 3-3 shows PM2.5 performance (NMB and NME along with their performance benchmarks) in April at the selected IMPROVE sites representing Class I areas in or near Texas. Figure 3-4 compares observed and modeled concentrations on most impaired days identified by EPA. These figures are included in the model performance spreadsheets accompanied with this report.

Figure 3-3. Example: PM2.5 NMB (top) and NME (bottom) in April at selected IMPROVE sites. Performance goal and criteria are presented in yellow and red, respectively.

-100

-50

0

50

100

150

BIBE1 BRIS1 CACR1 GRSA1 GUMO1 HEGL1 MING1 ROMO1 SACR1 UPBU1 WHIT1 WHPE1 WIMO1

(%)

Apr NMB ( PM25_TOT )

0

20

40

60

80

100

120

BIBE1 BRIS1 CACR1 GRSA1 GUMO1 HEGL1 MING1 ROMO1 SACR1 UPBU1 WHIT1 WHPE1 WIMO1

(%)

Apr NME ( PM25_TOT )

Page 16: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

12

0123456789

Obs

CAMx

Obs

CAMx

Obs

CAMx

Obs

CAMx

Obs

CAMx

Obs

CAMx

6-Apr 15-Apr 21-Apr 3-Sep 9-Sep 24-Sep

Conc

entra

tion

(µg/

m3 )

BIBE1: Most Impaired days

ammSO4 ammNO3 OM EC SOIL PM_Other

0123456789

10

Obs

CAMx

Obs

CAMx

Obs

CAMx

Obs

CAMx

Obs

CAMx

Obs

CAMx

4-Jan 6-Apr 15-Apr 24-Apr 3-Sep 15-Sep

Conc

entra

tion

(µg/

m3 )

GUMO1: Most Impaired days

ammSO4 ammNO3 OM EC SOIL PM_Other

Page 17: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

13

Figure 3-4. Example: comparison between observed and modeled concentrations at selected IMPROVE sites on most impaired days.

Observations at several IMPROVE sites indicate large Soil concentrations on some of the most impaired days. The model cannot capture these elevated Soil concentrations. Most Soil is due to crustal species that include windblown dust, road dust and other dust sources. Such sources are difficult to characterize and impacts at monitoring sites can be highly influenced by local sources. Note that there are inconsistencies in how the Soil species is defined using the IMPROVE measurements versus how it is defined in the model. The IMPROVE measurement data defines fine Soil as a linear combination of 5 elements (Al, Si, Ca, Fe and Ti), whereas in the model fine crustal is defined as PM2.5 emissions that are not explicitly speciated into SO4, NO3, NH4, EC or OM so represents other fine particulate, which can include more than the 5 elements in IMPROVE Soil definition as well as measurement artifacts. Thus, it is difficult to interpret model performance for Soil.

02468

1012141618

Obs

CAMx

Obs

CAMx

Obs

CAMx

Obs

CAMx

Obs

CAMx

19-Jan 22-Jan 3-Sep 9-Sep 21-Sep

Conc

entra

tion

(µg/

m3 )

WIMO1: Most Impaired days

ammSO4 ammNO3 OM EC SOIL PM_Other

Page 18: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

14

CONCLUSION AND RECOMMENDATION This project developed the CAMx 2016 modeling and provided a preliminary assessment of performance using CAMx version 6.5. TCEQ may further develop this CAMx setup to support the RHR SIP. CAMx was run for three modeling months using the EPA 2016 beta emissions and other modeling inputs either provided by TCEQ (meteorology, biogenic emissions) or developed by Ramboll (other natural sources, EDGAR emissions, BCs). We performed an operational MPE for ozone and PM2.5 at several monitor networks. The CAMx 2016 simulation shows reasonable performance for ozone, albeit an overestimation tendency in September. PM2.5 performance is variable across monitor location and modeling month.

CAMx predicts a large component of PM2.5 from SO4 and OM at the selected IMPROVE sites which agrees well with observations. However, the predicted OM concentrations in January are overestimated which is likely due to overstated primary emissions in the 2016beta inventory. We recommend implementing the SOA updates (available in CAMx version 7.0) into version 6.5 to help address the summer OM overestimation. These updates include reduced terpene SOA yields, reduced SOA photolysis, and removal of biogenic SOA polymerization. We do not see evidence of overstated SO4 at coastal sites due to simplifying the atmospheric chemistry of DMS in this modeling. Evaluating explicit DMS chemistry may not be critical for this modeling setup.

The model cannot capture elevated Soil concentrations observed at several IMPROVE sites on several days that EPA assigned as most impaired days. The most impaired days, by definition, are associated with anthropogenic emissions. We recommend reviewing the observation data at these sites and assessing whether they are influenced by exceptional events (e.g., dust storms) which would be excluded from the RHR SIP.

The development of a photochemical modelling database usually involves the performance of diagnostic model simulations designed to refine the model inputs and/or configuration to improve model performance. Since, by definition, models simplify complex phenomena, an initial model simulation may not accurately reproduce all air quality observations and achieve all of the model performance goals and criteria. Nevertheless, the initial CAMx 2016 simulation did show promise in predicting the observed ozone and PM2.5 concentrations. Sensitivity analysis, especially those targeting OM, SO4, and Soil, can help identify and evaluate potential refinements to improve CAMx simulation.

In addition to the CAMx simulations, we provided technical support to the TCEQ as follows:

• Prepared emission inventory files based on the EPA’s 2028 inventory to be used in the EPS3 emission processor.

• Provided updated chemical speciation profiles for PM2.5 and VOC • Generated CAMx-ready emissions for the full 2016 calendar year in central standard time (CST).

• Natural sources including windblown dust, lightning NOx, DMS and sea salt for the full 2016 calendar year.

• EDGAR global emissions for regions that not covered in the EPA 2016beta platform. • Prepared CAMx boundary concentrations for the full 2016 calendar year based on the GEOS-

Chem simulations (Ramboll and EPRI, 2017) for 2016 baseline, 2028 future-year with and without international anthropogenic emissions. The boundary concentration files are in CST.

• Provided pre-merged EPA 2016 MP beta emissions acquired from the Intermountain West Data Warehouse.

• Corresponded with TCEQ modeling staff on emission inventory data sources, emission processing, and CAMx modeling on a needed basis.

Page 19: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

15

REFERENCES Boylan, J.W. and A.G. Russell. 2006. PM and Light Extinction Model Performance Metrics, goals and

Criteria for Three-Dimensional Air Quality Models. Atmos. Env. 40:4946-4959.

Dunker, A.M., 1984. The decoupled direct method for calculating sensitivity coefficients in chemical kinetics, J. Chem. Phys., 81, 2385-2393.

Dunker, A.M., Yarwood, G., Ortmann, J.P., Wilson, G.M., 2002. The decoupled direct method for sensitivity analysis in a three dimensional air quality model- implementation, accuracy, and efficiency, Environ. Sci. Technol., 36(13), 2965-2976.

Emery, C., Liu, Z., Russell, A.G., Odman, M.T., Yarwood, G. and Kumar, N., 2017. Recommendations on statistics and benchmarks to assess photochemical model performance. Journal of the Air & Waste Management Association, 67(5), pp.582-598.

Ramboll and EPRI, 2017. International Contributions to Regional Haze, Final Workplan. October.

Ramboll, 2018. User’s Guide to CAMx version 6.50. Available from http://www.camx.com.

TCEQ, 2018. Texas Air Quality Modeling - Files and Information (2012 Episodes). Available at https://www.tceq.texas.gov/airquality/airmod/data/tx2012 updated April 24.

Page 20: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

APPENDIX A Monthly Average Concentrations (µg/m3) of PM2.5 Components

Page 21: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

A-1

Appendix A: Monthly Average Concentrations (µg/m3) of PM2.5 Components

January (left); April (middle); September (right)

Sulfate

Nitrate

Page 22: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

A-2

POA

SOA

Page 23: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

A-3

EC

Page 24: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

APPENDIX B Spatial Maps of NMB for PM2.5 and its Components

Page 25: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

B-1

Appendix B: Spatial Maps of NMB for PM2.5 and its Components

January (left); April (middle); September (right)

Total PM2.5

Sulfate

Page 26: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

B-2

Nitrate

OM

Page 27: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

B-3

EC

Soil

Page 28: Regional Haze Modeling - ...Aug 15, 2019  · 2016 12-km continental US (CONUS) modeling platform (2016 MP)1. The 2016 MP is a collaborative ... We expect modeling activities to resume

Ramboll - Regional Haze Modeling

B-4

CM (coarse material)