Improving aerosol forecasting through visibility data assimilation in WRF-chem: Methodology and evaluation

Chinese high-density visibility observation stations provide more uniform coverage than traditional air quality monitors, spanning urban, rural, and remote areas, which improves spatial representativeness. The visibility data from these stations has the potential to optimize aerosol initial conditions in air quality models, thereby improving the accuracy of pollution forecasting. This study details the development of a visibility observation assimilation module integrated with the WRF-Chem model. We aim to assimilate visibility data and evaluate the improvement in model performance achieved through assimilation compared to a direct WRF-Chem simulation. The assimilation system utilizes an advanced multi-scale three-dimensional variational assimilation (MS-3DVAR) technique and employs WRF-Chem MOSAIC aerosol species concentrations as control variables. To extensively evaluate the developed visibility assimilation algorithm, we conducted three numerical experiments focused on the forecasting accuracy of dust events that occurred in Northern China in March 2021. The first experiment (DA_PM&AOD) assimilated particulate matter (PM_2.5 and PM₁₀) and satellite-derived aerosol optical depth (AOD) data. The second experiment (DA_EXT) assimilated visibility data, while the third experiment (DA_ALL) integrated particulate matter, AOD, and visibility data. The evaluation demonstrates that, the control experiment group consistently underestimated the values during dust events. The DA_EXT assimilation group significantly improved the forecast accuracy of PM₁₀ concentration and EXT (extinction coefficient). In the DA_PM&AOD experiment, the agreement index between PM₁₀ simulation and observation increased to above 0.9. In the DA_ALL experiment, the agreement index for EXT simulation reached 0.677. The evaluation reveals that assimilating visibility data significantly improves both the initial aerosol fields and dust process forecasting in the model. This enhancement compensates for spatial and temporal discontinuities in satellite observations of visible light channels.