Insights to WRF-Chem sensitivity in a coastal region of Morocco: Chemical mechanisms, nesting options, and physical parametrization

Abstract

The performance of the Weather Research and Forecasting model coupled with chemistry (WRF-Chem) relies on the accuracy of the input data and its parameterization schemes. This study investigated the sensitivity of WRF-Chem version 4.4.2 to domain configuration, chemical mechanisms, and physics parameterizations. The model was used to predict air quality criteria pollutants (PM10, O3, CO, NO2, and SO2) and meteorological variables (wind speed (WS) and temperature (T)) for the first time over the Moroccan domain, in Agadir City. This region is located on the southwestern coast of Morocco and has a complex topography and land use configuration. We ran various simulations to test the sensitivity of the selected variables' predictions. These simulations explored different options for chemical mechanisms (MOZART, RACM, and GOCART), nesting configurations (three nested domains with a 1:4 nesting ratio and four nested domains with a 1:3 nesting ratio), and planetary boundary layer (PBL) parameterizations (YSU, MYJ, MYNN2, and QNSE). Modeled values of air quality and meteorological variables were then compared with surface observations using various statistical metrics. The results show that ozone O3 and NO2 are less sensitive to changes in domain configuration or chemical mechanisms, suggesting that ambient concentrations of these pollutants are more influenced by local factors. In particular, variations in NO2 are closely linked to local emission patterns. However, CO shows greater sensitivity to changes in nesting options, which is attributed to the dependence of CO modeling on accurate capture of local emission sources and atmospheric mixing. Similarly, concentrations of PM10, SO2 and CO are highly dependent on how physical and chemical processes are represented in the domain configurations and chemical mechanisms. On the other hand, O3 estimates are very sensitive to physical parameterization, which can affect meteorological parameters such as temperature. This underlines the significant influence of temperature and sunlight on ozone levels. For sensitivity analysis, the results indicate that the best optimal configuration varies depending on the variable, utilizing both qualitative (spatial variations) and quantitative (statistical metrics) approaches. Furthermore, for physics parameterizations, the chosen PBL for predicting WS and T is YSU. Concerning chemical mechanisms, GOCART emerges as the best scheme for predicting PM10, SO2, and CO, while MOZART is optimal for O3 and NO2. Regarding nesting options, the most efficient choice is to utilize three nested domains with a 1:4 ratio, optimizing modeling time. These indicate significant improvements in air quality prediction over the region, however, there are still improvements to be made, especially in PM10, which has the lowest accuracy of the model and could be the objective of further research in this study area.