Has the performance of regional-scale photochemical modelling systems changed over the past decade?

Abstract

This study analyzes summertime ozone concentrations that have been simulated by various regional-scale photochemical modelling systems over the Eastern U.S. as part of more than ten independent studies. Results indicate that there has been a reduction of root mean square errors (RMSE) and an improvement in the ability to capture ozone fluctuations stemming from synoptic-scale meteorological forcings between the earliest seasonal modelling simulations and more recent studies. However, even the more recent model simulations exhibit RMSE values of about 15 ppb and there is no evidence that differences in RMSE between these recent simulations are attributable to systematic improvements in modelling capability. Moreover, it was determined that certain aspects of model performance have not changed over the past decade. One such aspect is that the RMSE of simulated time series can be reduced by applying temporal averaging kernels of up to seven days while the benefit of longer averaging windows appears to vary from year to year. In addition, it is found that spatial patterns simulated by these modelling systems typically have lower correlations and higher centered RMSE than temporal patterns. Analogous to the errors in the simulated time series, these errors in the spatial patterns can be reduced through the application of spatial averaging kernels.