High resolution CMAQ simulations of ozone exceedance events during the Lake Michigan Ozone Study

Abstract. We evaluate two high-resolution Lake Michigan air quality simulations during the 2017 Lake Michigan Ozone Study campaign. These air quality simulations employ identical chemical configurations but use different input meteorology. The “EPA” configuration follows EPA recommended modeling practices, whereas the "YNT SSNG” employs different parameterization schemes and satellite-based inputs of sea surface temperatures, green vegetative fraction, and soil moisture and temperature. Overall, we find similar performance in model simulations of hourly and daily 8-hour maximum (MDA8) ozone, with the EPA and YNT_SSNG simulations showing biases of −13.31 and −13.54 ppbv, respectively during periods when the observed MDA8 was greater than 70 ppbv. However, for the two monitoring sites that observed high ozone events, the EPA simulation better matched observations at Sheboygan KA and the YNT_SSNG simulation better matched observations at Chiwaukee Prairie. We find differences between the two simulations are largest for column amounts of ozone precursors, particularly NO2. Across three high ozone events, the YNT_SSNG simulation has a lower column NO2 bias (0.17 × 1015 molecules/cm2) compared to the EPA simulation (0.35 × 1015 molecules/cm2). The YNT_SSNG simulation also has an advantage in better capturing the structure of the boundary layer and lake breeze during the June 2 high ozone event, although the timing of the lake breeze is about 3 hours too early. Our results are useful in informing an air quality modeling framework for the Lake Michigan area.