The sensitivity of simulated surface-level pollution concentrations to WRF-ARW-model PBL parameterisation schemes over the Highveld of South Africa

Abstract

Abstract An understanding of the sensitivity of pollutant concentrations to planetary boundary layer (PBL) schemes in numerical weather prediction (NWP) models can provide insight into model results, and possibly advance the accuracy of air pollution modelling. Meteorological output from the Advanced Research Weather Research and Forecasting (WRF-ARW) model, with four frequently used PBL schemes, is used as input for an air quality model. The simulations resulting from the different schemes are compared with each other and evaluated making use of observational air pollution concentration data for a spring and a winter month of 2016 in South Africa. Sensitivity exists in surface-level pollution concentrations as simulated by the considered PBL schemes. No single PBL scheme consistently performed best across the five considered pollutants, five sites and two seasons. During analysis of the air quality simulations, it was found that the two local closure schemes – Mellor-Yamada-Janjic (MYJ) and Mellor-Yamada Nakanishi Niino (MYNN) – often performed well, while the non-local closure schemes – Yonsei University (YSU) and Asymmetric Convection Modelling (ACM) – featured as best-performing schemes the least number of times. Sensitivity also exists on a spatial scale between schemes when simulating sulphur dioxide (SO2) plumes from tall stacks, and the MYNN scheme frequently produced the highest concentrations of SO2. The results from this study should be considered when configuring NWP models for the simulation of air quality in the Highveld region.