Brian Eder, Alfreida Torian, James Godowitch, Thomas Pierce and Steve Howard

1. An Evaluation of CMAQ’s Performance in the PBL and Free Troposphere Using Ozonesondes Brian Eder, Alfreida Torian, James Godowitch, Thomas Pierce and Steve Howard Emission and Modeling Evaluation Branch Atmospheric Modeling and Analysis Division National Exposure Research Laboratory, U. S. EPA RTP, NC 27711 Presented at the 8th Annual CMAS Conference, Chapel Hill, NC October 21, 2009

2. Motivation • Because of the dearth of upper-air measurements, evaluation of CMAQ’s ability to simulate O3 mixing ratios has historically focused on surface performance. • These evaluations take advantage of extensive data sets offered by networks like CASTNet and EPA’s AQS. • While such analyses are essential in understanding and subsequently improving the model’s performance, it is increasingly important to examine the model’s ability to simulate conditions aloft: • - Planetary Boundary Layer (PBL): surface to 1000 - 3000 m • - Free Troposphere (FT): 1000 - 3000 m to tropopause

3. Motivation • This increased emphasis of CMAQ’s ability to properly simulate ozone aloft is due in part to recent revisions of the NAAQS. • Such revisions have resulted in increased scrutiny on CMAQ’s ability to accurately simulate both: • - transport aloft, whether it be regional, continental or even hemispheric • - the “background” or natural component of ozone, including ozone originating from stratospheric intrusions associated with tropopause folds.

4. Response Accordingly, this research will utilize data obtained from four sites: - Trinidad Head, CA (41.070 N, 124.150 W) - Boulder, CO (40.020 N, 105.270 W) - Huntsville, AL (34.730 N, 86.580 W) - Wallops Island, VA (37.900 N, 75.500 W) in order to evaluate CMAQ’s ability to simulate: - Ozone mixing ratios (ppm) - Temperature (0 C) - Relative Humidity (%) in the PBL and FT. This analysis covers the summer months (J, J, A) from 2002 - 2006.

5. CMAQ Attributes Version: v.4.7 as configured for CDC PHASE runs for period 2002-2006 Emissions: Processed using SMOKE and based on the 2002 NEI (w/year-specific fire, mobile, biogenic and points EGU data). Chemistry: CB05 Meteorology: MM5 LBCs: Month-specific derived from a 2002 GEOS - Chem simulation Resolution: 36 Km horizontal grid 24 vertical layers extending from the surface to 100 hPa.

6. OzonesondeObservations

7. CMAQ Domain

8. Evaluation Protocol CMAQ Layer Characteristics Observations from each sonde were assigned to the appropriate CMAQ layer based on either: - extrapolation in layers where no observations were available, - weightedaveraging when numerous observations were available.

9. Evaluation Protocol • Only ozonesondes launched during the afternoon, when the PBL was more likely to be well mixed, were used in the calculation of the evaluation statistics. • A suite of performance metrics were calculated, including the: • - Normalized Mean Error (NME) • - Normalized Mean Bias (NMB), • which will be the metrics discussed.

10. Evaluation Results - NMB

11. Evaluation Results - NMB  CMAQ greatly over-predicts ozone throughout the lower PBL at the two coastal sites, especially at Trinidad Head. CMAQ performs well throughout the most of the PBL at the two inland sites, especially at Boulder. • CMAQ consistently and increasingly under-predicts ozone in the upper portions of the FT, especially near the tropopause.

12. Evaluation Results - NMB  CMAQ greatly over-predicts ozone throughout the lower PBL at the two coastal sites , especially at Trinidad Head. CMAQ performs well throughout most of the PBL at the two inland sites, especially at Boulder. • CMAQ consistently and increasingly under-predicts ozone in the upper portions of the FT, especially near the tropopause.

13. Evaluation Results - NMB  CMAQ greatly over-predicts ozone throughout the lower PBL at the two coastal sites , especially at Trinidad Head. CMAQ performs well throughout the most of the PBL at the two inland sites, especially at Boulder. • CMAQ consistently and increasingly under-predicts ozone in the upper portions of the FT, especially near the tropopause.

14. Evaluation Results - NME • GoodCMAQ performance throughout the PBL and most of the FT at the two inland sites. • Performance at the coastal regions is worse. • Performance in the upper FT is poor at all locations.

15. Evaluation Results – Trinidad Head, CA (n=135) O3T RH O3T RH

16. Evaluation Results – Trinidad Head, CA (n=135) O3T RH O3T RH Ozone is greatly over-predicted within the PBL. Slightly under-predicted in the lower FT. Greatly underpredicted in the upper FT.

17. Evaluation Results – Trinidad Head, CA (n=135) O3T RH • Temperature is over-predicted throughout most of the PBL. • Slightly under predicted in the lower FT and over-predicted in the upper FT.

18. Evaluation Results – Trinidad Head, CA (n=135) O3T RH  RHis under- predicted throughout the entire PBL. • Over-predicted in most of the FT.

19. Evaluation Results – Trinidad Head, CA (n=135) O3T RH  MM5-CMAQ do a poor job in representing the marine-PBL at this west coast location. - Not surprising given the 36 km resolution

20. Evaluation Results – Trinidad Head, CA (n=135) O3T RH  The underprediction in the FT, may be attributable to the lateral boundary conditions (LBC). - Values provided by GEOS- Chem, while representing an improvement over previously used static LBCs, may not be high enough.

21. Evaluation Results – Trinidad Head, CA (n=135)  The underprediction in the FT, may be attributable to the lateral boundary conditions (LBC). - Values provided by GEOS- Chem, while representing an improvement over previously used static LBCs, may not be high enough.

22. Evaluation Results – Wallops Island, VA (n=37) O3T RH Ozone is over- predicted within 400 m of surface. Fairly wellpredicted in the remainder of the PBL  Underpredicted in the FT.

23. Evaluation Results – Wallops Island, VA (n=37) O3T RH • Temperature is slightly under-predicted within 100 m of the surface. • Well predicted in the remainder of the PBL and most of the FT.

24. Evaluation Results – Wallops Island, VA (n=37) O3T RH  RHis over- predicted near the surface. • Slightly underpredicted in the upper PBL and lower FT  Over-predicted in the upper FT.

25. Evaluation Results – Wallops Island, VA (n=37) O3T RH • CMAQ does a better job in representing the marine-PBL at this east coast location. • Ozone performance is still fairly poor throughout the FT. - Transport aloft may be underrepresented. - LBCs aloft may again be too low.

26. O3T RH Evaluation Results – Huntsville (n=91) Ozone is slightly over- predicted within 100 m of surface. Increasingly under-predicted in the remainder of the PBL and FT.

27. O3T RH Evaluation Results – Huntsville (n=91) Temperature is well-predicted throughout the PBL and most of the FT.

28. O3T RH Evaluation Results – Huntsville (n=91) O3T RH  RHis slightly over- predicted in the PBL. • Slightly underpredicted in the lower FT and over-predicted in the upper FT.

29. O3T RH Evaluation Results – Huntsville (n=91) O3T RH • CMAQ does a better job at representing the meteorology at this inland location – as expected. • Ozone performance is, however, increasing poor through most of the FT. - Transport aloft may again be underrepresented. - Lack of lightning generated NOx emissions may be an issue.

30. Evaluation Results – Boulder, CO (n=98) O3T RH Ozone is very well predicted within 400 m of surface. Slightly under-predicted in the remainder of the PBL and lower FT.  Greatly underpredicted in the upper FT.

31. Evaluation Results – Boulder, CO (n=98) O3T RH • Temperature is under-predicted throughout the PBL and lower FT. • Over-predicted in the upper FT.

32. Evaluation Results – Boulder, CO (n=98) O3T RH • CMAQ does a good job at predicting the ozone at this inland mountain location throughout the PBL and most of the FT. • Surprisingly though, CMAQ’s meteorological performance is not quite as good.

33. Summary CMAQ varies in its ability to accurately simulate ozone in the PBL and FT. • Over-predicts ozone throughout the lower PBL at the coastal sites, especially at Trinidad Head. - Inability to accurately resolve marine boundary layers at a 36 km scale. • Performs better throughout most of the PBL at the inland sites, especially at Boulder. • Tends to underpredict throughout much of FT at all sites. - Poor representation of transport aloft, lack of lightning generated NOx emissions and inaccurate LBCs are possible contributors. • Consistently and greatly under-predicts ozone in the upper portions of the FT, especially near the tropopause. - Inaccurate representation of ozone originating from stratospheric intrusions associated with tropopause folds may be the cause.

34. Future Work - Case Studies • Individual case studies are being developed that will provide a better understanding of CMAQ’s inconsistent performance in simulating ozone aloft. - In a few instances, both morning and afternoon ozonesonde launches were conducted at the four sites. - These consecutive launches will be examined to evaluate CMAQ’s ability to simulate, diurnally, the evolution of the PBL and FT. Huntsville, AL 2 July, 2005 18:19 UTC CMAQOzonesonde

35. Thank you Contact Information:eder.brian@epa.gov 919.541.3994 (v) 919.541.1379 (f) Disclaimer: Although this work was reviewed by EPA and approved for publication, it may not necessarily reflect official Agency policy.