Pittsburgh Modeling for the PM 2.5 NAAQS 2012 EPA Regional/State/Local Modelers Workshop Chicago, IL May 2, 2012 Jason Maranche Allegheny County Health Department Pittsburgh, PA
Nonattainment Areas Before/After 2014 PM2.5 Nonattainment Areas (2006 Std) Ozone/PM2.5 Problem Areas Post-CSAPR 2
Southwestern Pennsylvania PM 2.5 Designations 3
PM 2.5 in Southwestern PA The Pittsburgh MSA is affected by long-range- transport from the Ohio Valley plus urban excess Ub Urban Excess Long-Range Transport Localized Excess Liberty-Clairton is a separate nonattainment area within Allegheny County, affected by local industry, complex terrain, and meteorology 4
American Lung Association State of the Air Rankings * * ALA includes partial county designations within MSAs 5
NAA Design Values, 2001-2010 6
ACHD PM 2.5 Current Design Values (2009-2011) Site Annual 24 Hour Projected 2014 Liberty Design Values from Design Value SIP (1997 Std): Design Value Liberty 15.0 44 Avalon 14.7 34 North Braddock 12.7 34 Harrison 12.4 30 Lawrenceville 11.6 27 Clairton 11.5 28 South Fayette 11.0 27 North Park 9.7 25 Annual 14.3 µg/m³ 24-Hour 42 µg/m³ Projected 2014 Liberty Design Values from CSAPR: Annual 14.6 µg/m³ 24-Hour 45 µg/m³ 7
Tri-State Speciation Sites Florence Lawrenceville Greensburg Liberty Quaker City Dolly Sods 8
Tri-State Averages 9
Lawrenceville, Liberty Lawrenceville Liberty Lawrenceville Liberty NCore site for Pittsburgh Highest concentration Urban residential Industrial valley, suburban 10
WEST Resultant Vector 250 deg - 24% NORTH 3% SOUTH 6% 9% 12% 15% EAST Liberty-Clairton Area and Upwind Sources Liberty-Clairton Area Coke Works Chemical Plant Glass Plant Power Plant 11
Continuous TEOM Trends 12
Liberty Excess Pie Chart Liberty minus Lawrenceville 13
Source Apportionment Start with conceptual model Monitor data trends Emissions i of distant t and local l sources Meteorology Traffic patterns Site specifics Source apportionment using Positive Matrix Factorization (PMF 3.0) Wind direction analysis using Conditional Probability Function (CPF) Trajectory analysis (PSCF) not used due to localized nature 14
Liberty PMF Factors, 2005-2010 15
Modeled Source Factors, Pittsburgh 16
Wind Direction Analysis 17
Liberty Organic Carbons Factor OC + trace elements, continuous presence 18
Liberty Burning/Cooking Factor OC + K, seasonal presence 19
Liberty Gasoline Vehicles Factor OC, weekday presence 20
Liberty-Clairton NAAQS Modeling Timeline Task Date Details Complete Annual SIP (1997 std) Mid 2011 Began in 2007; final SIP revision submitted in mid 2011 Protocol Development, 24 Hour Models evaluation, monitored data analysis, source SIP (2006 std) 2011 apportionment Inventory/Model Inputs Late 2011/Early 2012 CAMx Photochemical Modeling Spring/Summer 2012 Develop specific local inventory, configure model options, process WRF and SMOKE Base Year 2007, Future Case 2014 1 (existing controls), Future Case 2014 2 (additional controls) AERMOD Modeling, if needed Fall 2012 Combine local results with regional Complete 24 Hour SIP Dec 2012 Including weight of evidence, contingency 21
Annual PM 2.5 (1997 Standards) SIP Combined local direct PM 2.5 modeled impacts with regional secondary CMAQ (12 km) impacts on a hourly additive basis Used CALPUFF for local l modeling Prognostic MM5 data to generate CALMET wind fields (100 m computational grid) Non steady-state state for stagnant periods BLP algorithms for buoyant line sources 22
Annual PM 2.5 SIP (cont.) Control Strategy CAIR for regional secondary formation (sulfates/nitrates) Local controls: 3 new quench towers, 1 new coke battery, rebuild remaining coke battery ovens Required extension of attainment t date to 2015 (modeled d year 2014) to allow for installations Received conditional approval late 2011 Use of CAIR requires additional analysis 23
24-Hour (2006 Standards) SIP Development Base Year Attainment Test Design Values (2005-2009 weighted) 2007 Base Annual = 18.4 µg/m³ 2007 Base 24-Hour = 54 µg/m³ Monthly PM 2.5 modeling workgroup CALPUFF or AERMOD for local modeling CMAQ or CAMx for regional modeling Availability of regional runs/data Determine what sources impact Liberty during inversion periods 24
Diurnal Model Performance 25
24-Hour Q-Q 26
24-Hour Q-Q (log) 27
24-Hour Soccer Plot (Year-Round) 28
24-Hour Statistics (Year-Round) CMAQ + AERMOD CMAQ + CALPUFF Correlation Coefficient (r) 0.775 0.682 Coefficient of Determination (R²) 0.600 0.465 Ratio Mean 1.500 1.456 Mean Bias 6.602 5.394 Normalized Mean Bias 32.401 26.473 Mean Fractional Bias 30.917 26.231 Mean Error 9.228 9.546 Normalized Mean Error 45.286 46.848 Mean Fractional Error 41.842 42.581 Root Mean Squared Error 12.004 12.907 29
24-Hour PM 2.5 SIP Methodology Use photochemical model on fine-scale for all local/regional If photochemical model performance is low for base year, use AERMOD for local primary impacts CAMx selected for photochemical model Plume-in-Grid (PiG) module for sources near the Liberty monitor Particle Source Apportionment Tool (PSAT) for combination with local results Approach used for other areas (Birmingham, St. Louis, Detroit) 30
Contractors for 2006 Standards SIP TranSystems E.H. Pechan Investigation of area and mobile sources in Liberty-Clairton Develop specific area/nonroad/onroad inventory Alpine Geophysics WRF met processing SMOKE emissions processing ENVIRON CAMx modeling Combination of CAMx/AERMOD impacts 31
36/12/4/0.8 km Domains 1800 1440 1080 720 360 0-360 08km 0.8 04 km 12 km Standard RPO Lambert Conformal projection One way nesting for 36/12/4 km grids 36 km used to generate BC s for 12 km domain -720 WRFdomains are defined with at least a 5 grid cell buffer in all directions from the CAMx domains -1080-1440 -1800 36 km -2520-2160 -1800-1440 -1080-720 -360 0 360 720 1080 1440 1800 2160 2520 ACHD Proposed CAMx Domain 36 km 148 x 112 (-2736, -2088) to (2592, 1944) 12 km 174 x 117 ( 72, -540) to (2160, 864) 04 km 54 x 60 ( 1296, 48) to (1512, 288) 0.8 km 75 x 60 ( 1392, 144) to (1452, 192) 32
12/4/0.8 km Domains 840 720 600 480 360 240 120 0.8 km 12 km includes states identified by CSAPR as contributing tib ti to PA concentrations 0 04 km -120-240 -360-480 12 km 120 240 360 480 600 720 840 960 1080 1200 1320 1440 1560 1680 1800 1920 2040 2160 ACHD Proposed CAMx Domain 36 km 148 x 112 (-2736, -2088) to (2592, 1944) 12 km 174 x 117 ( 72, -540) to (2160, 864) 04 km 54 x 60 ( 1296, 48) to (1512, 288) 0.8 km 75 x 60 ( 1392, 144) to (1452, 192) 33
4/0.8 km Domains 280 240 200 Two way grid nesting for 4/0.8 km, flow in and out of grids 160 120 0.8 km 4 km focused on SW PA, WV, OH 0.8 km focused on southern Allegheny County 80 04 km 1320 1360 1400 1440 1480 ACHD Proposed CAMx Domain 36 km 148 x 112 (-2736, -2088) to (2592, 1944) 12 km 174 x 117 ( 72, -540) to (2160, 864) 04 km 54 x 60 ( 1296, 48) to (1512, 288) 0.8 km 75 x 60 ( 1392, 144) to (1452, 192) 34
Plume-in-Grid (PiG) Plume-in-Grid example Results for 10:00 UTC on July 8, 2002 Puffs sampled on a 200 m grid PiG and PSAT in CAMx for key local sources 35
2007 Emissions Inventory Sources Source Category MARAMA States SEMAP/SESARM States MidWest RPO (LADCO) Other States Area NonRoad MARAMA 2007v3 Final* MARAMA 2007v3 Final* EPA 2008 NEIv1.5 EPA 2008 NEIv1.5 EPA 2008 NEIv1.5 SEMAP 2007 Draft EPA 2008 NEIv1.5 EPA 2008 NEIv1.5 OnRoad Mobile LADCO 2007BaseCv8 LADCO 2007BaseCv8 LADCO 2007BaseCv8 LADCO 2007BaseCv8 NonEGU Point MARAMA 2007v3 Final* SEMAP 2007 Draft LADCO 2007BaseCv8 EPA 2008 NEIv1.5 EGU Point MARAMA 2007v3 Final SEMAP 2007 Draft 2007 day specific hourly CEM 2007 day specific hourly CEM Fires 2007 BlueSky Fire 2007 BlueSky Fire 2007 BlueSky Fire SEMAP 2007 Draft Emissions EPA, 2010 Emissions EPA, 2010 Emissions EPA, 2010 Biogenics Day Specific SMOKE BEIS Day Specific SMOKE BEIS Day Specific SMOKE BEIS Day Specific SMOKE BEIS * Includes ACHD local source revisions for fine-scale modeling 36
Airport Surface Wind Roses, 2007 WRF NORTH 12% 9.6% 7.2% 4.8% 2.4% WEST EAST NORTH Resultant Vector 240 deg - 23% SOUTH 12% 4 km surface nudging 9.6% 7.2% 4.8% 2.4% WEST EAST Resultant Vector 252 deg - 25% SOUTH NORTH 15% 12% 9% 6% 3% WEST EAST Resultant Vector 286 deg - 26% SOUTH 4 km surface nudging 37
WRF 800 m Performance 38
AERMOD Options (if needed) If model performance is poor with CAMx, use AERMOD for selected sources Source types Buoyant fugitives methodology Cold fugitives (volumes, areas) Met data Onsite data at Liberty, with PIT airport upper air (0Z, 12Z) OR MMIF data for selected cell in area, with extracted WRF hourly data 39
MMIF 800m Cells Liberty 4km West School 4km East B id Bridge Ri River 40
MMIF 800m Wind Direction Frequency by Sector Liberty School Bridge River 41
Wind Roses, 800m MMIF Grid Cells, Surface Level Liberty School Bridge River 42
MMIF 800m Avg Wind Speed by Sector Liberty School Bridge River 43
MMIF 800m Avg Wind Speed by Hour Liberty School Bridge River 44
MMIF 800m Avg Temp by Hour Liberty School Bridge River 45
Future Work/Regulatory Issues Future case inventories (CSAPR, other 2014) Control strategy, t future PM 2.5 NAAQS Reconstruction of species, MATS, scale of unmonitored areas Preferred method for photochemical + local? Development of near-field chemistry models (SCICHEM, etc.) Appropriateness of met data, use of MMIF data in AERMOD Banked emissions, deactivation of plants 46