Multi-model forecast evaluation and bias analysis of asymmetric precipitation associated with Typhoon “Haikui”

To clarify the forecast biases in TC precipitation and ensure the appropriate application of high-resolution regional numerical model products, this study assesses the capability of three high-resolution regional operational models in China to predict the precipitation associated with TC Haikui (2311) based on the high-resolution hourly CMPA precipitation product on 3-6 September 2023 released by the China Meteorological Administration. Specifically, we evaluate their performance in capturing the spatial asymmetry, hourly evolution characteristics such as radial, azimuthal, vertical wind shear direction, and moving direction and speed of the TC. Furthermore, the possible causes of forecast biases are discussed. Main results are as follows: (1) The three models reasonably forecasted the inward movement of the precipitation center toward the TC core on 3 September, while the forecast performance decreases from 4 September. In contrast, the TC azimuthal precipitation forecast exhibits the opposite trend: forecast skill is lower on 3 September but higher from 4 September onward. Precipitation forecast bias along the TC motion direction increases as the typhoon weakens, whereas that along VWS direction remains large throughout all stages of the TC. (2)The Precipitation Asymmetry Index (PAI) observed in the three models is higher (lower) during the TC's passage over Taiwan Island (landfall in Guangdong and Fujian). The impact of TC movement characteristics and VWS on precipitation bias varies by model. CMA-SH9 accurately forecasted the VWS direction and produced negligible bias in the evolution of PAI, whereas the moving characteristics (direction and speed) of TC has more effect in CMA-MESO3. And the asymmetric distribution in CMA-GD3 is affected both by the TC’s moving characteristics and VWS. The research findings can offer guidance on how to correct bias in operational TC precipitation forecasts and serve as a reference for model improvements.