LAUSR

Abstract On 6 May 2017, a record-breaking warm-sector torrential rainfall (WSTR) event initiated over Guangzhou city around 1600 UTC and lasted for 20 h. To replicate this rainfall event, both three-dimensional and four-dimensional variational (3DVAR/4DVAR) data assimilation (DA) strategies were used to initialize a convection-allowing configuration of the Weather Research and Forecasting (WRF) model in an attempt to evaluate precipitation forecasts with or without the assimilation of Advanced Himawari Imager (AHI) radiances from three water vapor (WV) channels. This case study yielded impressions that both the 4DVAR experiments with and without the assimilation of AHI radiances prominently improved convection initiation (CI) forecasts, and replicated the observations accurately in predicting hourly area precipitation totals. Due to the small spatial scale of this event, the impacts of assimilating every 10-min AHI radiances on the convection evolution and hourly precipitation forecasts were subjectively small to be seen, but the equitable threat score (ETS) and fractions skill score (FSS) did show slight improvement for both hourly and 20-h accumulated precipitation. For example, the 4DVAR technique combined with AHI radiances improved the FSS of 20-h accumulated precipitation forecasts by 2%–4.5%, 1%–3%, 6%–20%, and 8%–24% for 5 mm, 20 mm, 50 mm, and 80 mm thresholds, respectively, over those with only conventional observations assimilated. Reasons for these improvements were attributed to better analyses and forecasts of temperature, moisture, and wind.