Evaluation of atmospheric chemical models using aircraft data

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Tuesday, 4 February 2014
Hall C3 (The Georgia World Congress Center )
Sean W. Freeman, Florida State University, Tallahassee, FL; and N. Grossberg, R. B. Pierce, P. Lee, F. Ngan, E. L. Yates, L. T. Iraci, and B. L. Lefer

Air quality prediction is an important and growing field, as the adverse health effects of ozone (O3) are becoming more important to the general public. Two atmospheric chemical models, the Realtime Air Quality Modeling System (RAQMS) and the Community Multiscale Air Quality modeling system (CMAQ) are evaluated during NASA's Student Airborne Research Project (SARP) and the NASA Alpha Jet Atmospheric eXperiment (AJAX) flights. CO, O3, and NOx data simulated by the models are interpolated using an inverse distance weighting in space and a linear interpolation in time to both the SARP and AJAX flight tracks and compared to the CO, O3, and NOx observations at those points. Results for the seven flights included show moderate error in O3 during the flights, with RAQMS having a high O3 bias (+15.7 ppbv average) above 6 km and a low O3 bias (-17.5 ppbv average) below 4 km. CMAQ was found to have a low O3 bias (-13.0 ppbv average) everywhere. Additionally, little bias (-5.36% RAQMS, -11.8% CMAQ) in the CO data was observed with the exception of a wildfire smoke plume that was flown through on one SARP flight, as CMAQ lack any wildfire sources and RAQMS resolution is too coarse to resolve narrow plumes. This indicates improvement in emissions inventories compared to previous studies. CMAQ additionally incorrectly predicted a NOx plume due to incorrectly vertically advecting it from the surface, which caused NOx titration to occur, limiting the production of ozone. This study shows that these models perform reasonably well in most conditions; however more work must be done to assimilate wildfires, improve emissions inventories, and improve meteorological forecasts for the models.