CMA-GD对春季偏东气流影响下海南岛局地暴雨预报的偏差分析OA

During spring,Hainan Island is prone to localized heavy rainfall over its eastern hilly and plain regions driven by low-level easterly airflow.However,significant forecast biases persist in operational convective-scale numerical models.This study evaluates the performance of the CMA-GD mesoscale model(3 km resolution)in simulating these localized heavy rainfall events during 2021-2023 utilizing high-density automatic weather station observations,1-km resolution CLDAS(CMA Land Data Assimilation System)land surface datasets,and ERA5 reanalysis data.The results indicate that while the model adequately simulate afternoon heavy rainfall over the central-western mountains,it consistently underestimates daytime and early morning heavy rainfall occurring during the in the east.Consequently,the model fails to capture the observed bimodal d diurnal cycle of mean precipitation.Synoptically,these errors are linked to the misrepresentation of the offshore cold high-pressure system.The model exhibits a systematic clockwise bias(approximately 30 °)in the low-level(below 850 hPa)wind direction and overestimates wind speeds below 925 hPa by 2-3 m·s-1 compared to observations.Concurrently,it underestimates specific humidity(approximately 2 g·kg-1)and relative humidity(～12%,)respectively,at 1 000 hPa and Diagnostic analysis of typical cases reveals five primary mechanisms for underestimation of heavy rainfall:(1)The clockwise wind direction bias weakens frontogenesis within convergent flows;(2)The low-level dry bias elevates the Lifting Condensation Level(LCL)and Level of Free Convection(LFC);(3)Insufficient surface friction reduces forecast convergence,inhibiting convective initiation;(4)Inadequate representation of the horizontal and vertical wind shear perpendicular to the terrain;and(5)Forecast biases in land surface temperature disrupt local sea-land breeze circulations,altering precipitation patterns.