An Assessment of CMAQ with TEOM Measurements over

An Assessment of CMAQ with TEOM Measurements over the Eastern US Michael Ku, Chris Hogrefe, Kevin Civerolo, and Gopal Sistla PM Model Performance Workshop, February 10 -11, 2004, RTP, NC

Model Simulations • MM 5 – 108/36/12 km two-way nesting. • SMOKE – 1996 CSA emission inventory. • CMAQ – 12 km domain only; both CB-IV and RADM 2; IC/BC used background values. • Simulation Period – July 2 – August 1, 1999

TEOM Measurements 21 sites include SLAMS, USDOE, NEOPS, and SEARCH.

TEOM Measurements Organization Iowa (SLAMS) New Jersey (SLAMS) New York (SLAMS) North Carolina (SLAMS) South Carolina (SLAMS) USDOE (PA) SEARCH (AL & GA) NEOPS (PA) Total # of Sites 6 5 1 1 2 2 3 1 21 ID used in analysis 1 -6 7 - 11 12 13 14 - 15 16 - 17 18 - 20 21

Modeling Domain and TEOM Sites

Model Evaluation • Examine the Model Error • Examine the Model skill -- Compare the spatial structures -- Compare the temporal patterns

Statistics: Hourly Data Parameter Mean 22. 28 CMAQ CB-IV 29. 99 S. D. 14. 29 25. 37 23. 89 R 0. 46 Mean Bias 7. 8 5. 78 23. 76 22. 26 RMSE TEOM Observed CMAQ RADM 2 28. 05

Comparison at each site

Statistics: Daily Averaged Data Parameter Mean TEOM observed 22. 23 CMAQ CB-IV 29. 99 CMAQ RADM 2 28. 05 S. D. 11. 59 22. 60 21. 24 R 0. 57 Mean Bias 7. 80 5. 75 RMSE 19. 98 18. 38

CMAQ (CB-IV) predicts slightly higher daily averaged values than CMAQ (RADM 2).

CMAQ (CB-IV) underpredicted low-end and overpredicted high end of the daily averaged values.

CMAQ (RADM 2) underpredicted low-end and overpredicted high end of the daily averaged values.

Compare Spatial Structures -Calculate Cross-correlation coefficients of TEOM measurements and CMAQ outputs at the TEOM sites. The calculations yield a 21 x 21 symmetric matrix of correlation coefficients which represent the correlation of the sites with each other. -If CMAQ produces similar correlation coefficients matrix with TEOM, the CMAQ is able to capture the TEOM measured spatial structures.

The similarity of the two contour plots indicates that CMAQ (CB-IV) is able to capture the spatial pattern of the TEOM measured data

Compare Temporal Patterns • Hourly time series • Synoptic components • Diurnal variation

Hourly time series: Examples of good comparison

Hourly time series: Examples of poor comparison

Examine the Synoptic Components • KZ filter is used to extract the Synoptic Components from TEOM measurements and CMAQ predicted data. • Compare the Synoptic Components for data averaged over three regions: Iowa, Northeast, and SEARCH.

Iowa Region

Northeast Region

SEARCH

Diurnal variation: Examples of good hourly time series comparison.

Diurnal variation: Examples of poor hourly time series comparison.

SUMMARY • CMAQ overpredicted TEOM measurements at high end and underpredicted at low end. • CMAQ captured the spatial pattern of the TEOM measurements. • TEOM measurements and CMAQ predictions show no typical diurnal variation. • CMAQ performed well in capturing the average synoptic temporal pattern in the northeast region, but failed to capture the temporal pattern in the other two regions. • Analysis should be expanded to include PM speciation data.