Improving PM_2.5 forecasts with three-dimensional variation data assimilation of visibility observations in China

More than 2400 synoptic stations in China have been conducting automatic visibility observations since 2013. Investigate the potential contribution of visibility data assimilation to the prediction of visibility and PM_2.5 is worthwhile. Visibility data was converted to aerosol extinction coefficient (AExtC) at 2424 synoptic stations in China from 1–31 December 2020. The Weather Research & Forecasting Data Assimilation system coupled with Chemistry (WRFDA-Chem) aerosol optical property module was applied to assimilate the bias-corrected visiometer AExtC. The analysis field was used as the initial chemical condition of the Weather Research & Forecasting system coupled with Chemistry (WRF-Chem) 24 h forecast run. Relative humidity information was assimilated to further decrease the uncertainty of RH before aerosol DA was applied. To illustrate the contribution of the AExtC DA, four parallel experiments were conducted, including a control run (CTR), a visiometer AExtC DA run (VIS), a ground PM_2.5 DA run (PM) and a joint assimilation of the AExtC and PM_2.5 run (VisPM). Compared with those of the CTR experiment, better AExtC and PM_2.5 forecast skills were acquired up to 24 forecast hours via the VIS, PM, and VisPM experiments. The assimilation of visiometer AExtC significantly improved the model prediction performance of AExtC (PM_2.5), with a reduction of in the RMSE value of 21 % (20 %). Direct DA of visiometer AExtC (PM_2.5) yielded a better AExtC (PM_2.5) analysis and forecasts. However, the VIS experiment also revealed its potential contribution to PM_2.5 prediction in Northwest China, compensating for the sparse number of PM_2.5 stations in this region. Joint assimilation of both the visiometer AExtC and ground PM_2.5 further improved the AExtC and PM_2.5 forecast, with reductions of RMSE of 22 % and 31 %, respectively.