Assimilating TOLNET Profile and AirNow Surface Ozone Observations over the Eastern United States during a Canadian Wildfire Smoke Intrusion Event Using WRF-Chem/DART

This study uses the Weather Research and Forecasting Model with Chemistry/Data Assimilation Research Testbed (WRF-Chem/DART) chemical transport forecasting/data assimilation system, to assimilate Environment Protection Agency (EPA) AirNow surface and ground-based lidar vertical profile ozone (O₃) observations over the eastern US to study the impact of smoke intrusion from a Canadian wildfire event in June 2015. The positive systematic bias of the operational surface O₃ forecasts inspired this work. Additionally, in the absence of the assimilation of in situ O₃ observations, WRF-Chem performed poorly producing positive biases near the surface ranging from 5 ppbv to 15 ppbv based on the AirNow observations, especially during the night-day transition period. Higher in the troposphere, between the surface and 1.5 km, WRF-Chem performed well with biases of 5 ppbv to 10 ppbv, but from 1.5 to 2.5 km it produced positive biases ranging from 10 ppbv to 20 ppbv based on comparison with the Tropospheric Ozone Lidar Network (TOLNet) observations. Due to the unsatisfying model performance, we propose to improve the model simulations by using the ensemble adjustment Kalman filter of Anderson (2001) to constrain the O₃ forecasts with surface and profile O₃ observations.

The WRF-Chem/DART system is described by Mizzi et al. (2016; 2018). It uses the WRF-Chem model described by Grell et al. (2005) and the DART ensemble data assimilation system described by Anderson et al. (2009). For this study, we initialize the WRF-Chem meteorological fields with the North American Regional Re-analyses (NARR), and we initialize the chemistry fields with the output from the Community Atmosphere Model with Chemistry (CAM-Chem). The WRF-Chem simulation uses various chemical emissions: (i) anthropogenic emissions from the National Emission Inventory 2011 (NEI 2011); (ii) biogenic emissions calculated during model integration by the Model of Emissions of Gases and Aerosols from Nature (MEGAN); and (iii) fire emissions from the Fire INventory from NCAR (FINN). We employ several different sources of observation datasets in this study. To constrain the WRF-Chem O₃ forecasts we assimilated EPA AirNow surface O₃ mixing ratio observations and Goddard Space Flight Center TROPospheric OZone differential absorption (GSFC TROPOZ - one of the instruments used in the TOLNet) O₃ lidar observations. To verify the forecast results, we use balloon-borne electrochemical concentration cell (ECC) ozonesonde vertical profile observations. We will present results from two experiments: (i) a control experiment where only model simulation is considered without any data assimilation, and (ii) a chemical data assimilation experiment where we employ data assimilation of the above-mentioned observations. Eventually, we will conduct a third experiment to study the impact of using the assimilated observations to adjust the anthropogenic (and possibly the fire) emissions.