Birmingham Area PM 2.5 Attainment Modeling Region 4 Modelers Workshop March 18, 2009

1. Birmingham Area PM2.5 Attainment ModelingRegion 4 Modelers WorkshopMarch 18, 2009

2. The Problem • EPA designated Jefferson, Shelby, and a small portion of Walker counties as nonattainment for the annual standard effective April 5, 2005. • Current PM2.5 Design Values – 2006-2008: • North Bham – 17.3 mg/m3 • Wylam - 16.3 mg/m3 • NAAQS = 15 mg/m3

3. Birmingham Area Design Values

4. Annual PM2.5 Design Values for Birmingham Area

5. The Solution – Develop a SIP • Contracted with ENVAIR to help identify causes of high PM2.5. • Based on the conclusions of the ENVAIR study, the SIP focuses on reduction of emissions of fine particles in the area surrounding the North Birmingham and Wylam monitors. • The plan also relies on reductions from national programs such as CAIR, and cleaner cars, diesels and fuels. • Base Case 2002/2009 (& 2012) modeling to help develop attainment plan was completed using a CMAQ/AERMOD integrated approach following EPA implementation guidance: “Guidance on the Use of Models and Other Analyses for Demonstrating Attainment of Air Quality Goals for Ozone, PM2.5 and Regional Haze” (April 2007).

6. Major Findings of ENVAIR Study • Nonattainment due to N. Birmingham & Wylam • Local and urban PM contributions are superimposed on regional component • Regional - ~12-14 mg/m3 • General urban - ~2 mg/m3 • Local - ~2-3 mg/m3 (revised during BAPS) • Multiple lines of evidence link local excess PM at Wylam and North Birmingham to several geographical source complexes • Evidence includes wind directions, carbon compounds, elements (metals), day-of-week variations, fence-line samples, & PM spikes

7. Major Findings of ENVAIR Study • With moderate decreases projected in regional PM, local emission reductions will likely be needed to attain the standard. (VISTAS/CAIR modeling confirmed this) • Since Wylam and N. BHM drive nonattainment, concentrated on source complexes surrounding those sites • Many sources are intermittent or semi-continuous processes in open buildings • Numerous fugitive sources • Very high infrequent PM concentrations (spikes) • Transportation – rail and trucking – contribute • What to do? Model and see….

8. Measurement Sites across NAA

9. PM2.5 (mg m-3)

10. PM2.5 (mg m-3)

11. ENVAIR to ENVIRON/AG • Taking the findings from the monitoring study, JCDH and ADEM contracted with ENVIRON/Alpine Geophysics to conduct a model attainment demonstration • Contract awarded in Fall 2006 • Used the CMAQ platform with MM5/SMOKE to model the regional and urban signals • Used the AERMOD model to evaluate local source impacts • Integration of modeling platform results

12. What doesn’t kill you…. • This has been a learning process • Many different stakeholders • New territory for modeling • Uncertainty in emissions inventories • Uncertainty in modeling integration • Highly variable emissions from many types of sources, many of which have never been involved in a modeling study of this magnitude • Sheer number of sources

13. Emissions Inventory Development • Multiple 2002, 2009, and 2012 CMAQ and AERMOD inventories were developed to identify direct, inert PM2.5 emissions • Much bigger challenge than expected • Many sources never modeled with this much detail and scrutiny • Emissions factors for PM2.5 poorly defined, if even available • Needed to weigh a perfect inventory against time and resource constraints • Small sources may have significant impacts

14. Emissions Inventory Issues • Mistakes were made by both the regulatory agencies as well as the facilities. This is due in large part, to the lack of understanding of what is needed to model at this level. • Required careful integration of NEI and more detailed inventories, generated specifically for this SIP work. • By careful to eliminate any duplication. • Update stack parameters and re-characterize sources as necessary. • Update emissions factors. • For transparency, we insisted on active involvement of facilities. • This led to multiple revisions of the 2002, 2009 and 2012 SMOKE runs and delays in the whole process. • SMOKE outputs were run through CAMx to produce consistent hourly emissions profiles to be input into AERMOD. If we had known how difficult it would be, we might have contracted for the inventory development.

15. AERMOD Modeling • The Envair studies showed a clear “local sources” signature, especially for primary PM2.5. • CMAQ, even with 4 km grid spacing, was not considered adequate to resolve impacts due to local emission controls. • AERMOD selected as the best way to model the significant industrial contributors. • Which local sources should be modeled?

16. AERMOD Modeling Local Source Criteria • No established criteria - all new territory. • Based on the results of the ENVAIR study, it was assumed that every source identified by the study would be included. • Any source within 5 km of either monitor with PM2.5 emissions greater than 1 tpy (~1/4 lb/hr) was included. • Between 5 – 10 km of either monitor, any source with PM2.5 emissions greater than 4 tpy (~1 lb/hr) was included. • A Q/d and Q/d2 analyses supported the above criteria fairly well. • Total of 46 facilities identified; roughly 1200 individual emitting sources. Included point, area, volume and buoyant lines.

17. Location of the AERMOD sources AERMOD Facilities not Culpable for RACT AERMOD Facilities Culpable for RACT

18. AERMOD Modeling Grid • Initial discussions with EPA and among the study participants led to a 1 km X 1 km AERMOD receptor grid with 100 meter spacing, centered at each monitor. • Plant property issues • Probably overkill • Additional discussion led to agreement on a 200 m X 200 m Cartesian grid (9 receptors) with 100 meter spacing, centered on each monitor. • For the attainment demonstration, concentrations were averaged across all receptors • For culpability and subsequent RACT analyses, concentrations at the monitor were used

19. AERMOD Met Data • Used 2002 met data – same as base case emissions data year. • Options • Conventional NWS ASOS data at Birmingham airport (BHM) – 7–18 km from key monitors. • SEARCH site wind and temperature data at North Bham monitor. • Hourly-averaged 1-minute ASOS data at BHM. • Choice – hybrid of hourly-averaged ASOS data augmented by conventional ASOS data as necessary. SEARCH data had too many holes and quality questions. OAQPS invaluable in developing the hybrid data set.

20. AERMOD Met Data

21. AERMOD Overview Ideally, AERMOD should predict lower concentrations than daily FRM observations • Monitors measure all PM2.5 • Primary and secondary • Local, urban and regional components • We modeled only primary local contributions • Observation-based analyses (ENVAIR Study) suggest annual local industrial contribution is • ~3 mg/m3 at NBHM • ~2 mg/m3 at WYLM But we recognize AERMOD is intentionally a “conservative” model

22. AERMOD Performance

23. Time-sequenced Modeled and Measured 24-Hr PM2.5 at Wylam Monitor

24. 2002 Wylam Frequency Distribution

25. WYLM Results Agrees with expected patterns • Almost always lower than daily FRM total obs • Expected local industry contributions are ~2 mg/m3 • AERMOD annual mean is 6.9 mg/m3 (~3.5:1) • AERMOD is rarely >10x the assumed local component (6 days or 2%of the year) • >4x local component ~33% of the year • Annual frequency distribution is heavy in the 0-10 mg/m3 range

26. Time-sequenced Modeled and Measured 24-Hr PM2.5 at N Birmingham Monitor

27. 2002 NBHM Frequency Distribution

28. NBHM Results Agrees with expected pattern • Almost always lower than daily FRM total obs • Expected local industry contributions are ~3 mg/m3 • AERMOD annual mean is 12.1 mg/m3 (~4:1) • AERMOD is >10x the assumed local component infrequently (15 days or 4% of the year) • AERMOD is >4x local component ~40% of the year. • Annual frequency distribution is heavy in the <10 and >50 mg/m3 range • Much of the overestimations were a result of the model’s inability to handle low level/ground based sources with source-receptor distances as small as 300 meters

29. Quantile-Quantile Comparison

30. Preliminary Conclusions – 2002 Run • Q-Q plot shows a marked difference in the character of AERMOD prediction between NBHM and WYLM • NBHM apparently dominated by facilities in very close proximity • Are sources characterized adequately? • Should we expect AERMOD to perform poorly for certain source configurations? Next step – Run 2009 with known EI changes and assumed growth

31. 2009 AERMOD • Hoped to show attainment in 2009 with CMAQ alone • If not, would combine with AERMOD to show attainment • Initial 2009 AERMOD had two purposes • To Integrate with CMAQ to predict 2009 Design Values • Determine culpability for RACT analyses • Initial 2009 AERMOD, integrated with CMAQ, did not demonstrate attainment – culpability investigation required.

32. RACT Culpability • Culpability based on 2009 AERMOD results • Identified facilities and sources that made a significant contribution to the PM2.5 problem, at either monitoring site. • Defined “significant contribution” as: • Any facility with a predicted contribution of 0.15 mg/m3 or more • Additionally, any individual source at an identified facility that contributed >= 0.2 mg/m3 • Ten facilities with 28 sources were identified. Next steps – Re-run all models for 2009, revised for controls implemented by 2009. Re-calculate Design Values

33. 2009 DV Projections -Post-RACT Controls • Results (3x3 grid cell averaging) • CMAQ “all-source” runs • NBHM – 16.5 mg/m3 • WYLM – 15.7 mg/m3 • CMAQ + AERMOD runs • NBHM – 16.3 mg/m3 • WYLM – 15.5 mg/m3 • ARRRRRGH Next steps – new future year

34. ASIP Modeling • Updated modeling for 2012 (12km) accomplished by ASIP in July/August 2008 for GA and AL SIPs • All states provided updated emissions inventories • Alabama rolled in the “to-date” 2002 BAPS • inventory • CAIR controls assumed • Results used for boundary conditions for BAPS 2012 runs.

35. 2012 Modeling • CMAQ – Regional controls implemented by 2012 • CAIR controls • EGU controls expected from consent decrees in AL and other SE states • GA multi-pollutant rule • AERMOD • RACT controls not implemented by 2009 • Results show compliance in 2012 • The expected DV likely lies between the two modeling systems.

36. CMAQ “All Source” 2012 DV Projection

37. CMAQ + AERMOD 2012 DV Projection

38. Lessons Learned • START EARLY. We did so (long before guidance on how to do it was available), but we were still late. • Find an Inventory Guru. Contract it out, if necessary. An accurate inventory will be the long pole in the tent. • EPA workshops and conferences relating to attainment demonstrations need to do more to involve EI people. • Involve industry stakeholders throughout the whole process. It can be frustrating (and slows the process) but we think it’s the right thing to do. It’s amazing how many emission control projects were implemented well before we got to the RACT step. • Not sure attainment modeling is the best use of AERMOD. Future modeling exercises like this should focus on refining photochemical models to handle at very small grid scales.

39. Thanks EPA!!! • EPA’s involvement has been invaluable • Met data • Modeling assistance • Discussion on issues such as source characterization, policy implications and model performance