Abstract:Using the spatial-temporal verification method of precipitation process, this paper evaluates the prediction effect of non-typhoon hourly precipitation in warm season (April to September of 2019-2020) of Hainan Island by Guangdong rapid update assimilation numerical prediction system (3 km resolution, CMA-GD), Shanghai numerical prediction system (CMA-SH9) and mesoscale weather numerical prediction system (CMA-MESO) of China Meteorological Administration. 〖JP2〗The results show that the three models 〖JP〗all can capture the spatial distribution of precipitation and the diurnal variation of precipitation under different flow field conditions. However, the frequency and intensity of precipitation in CMA-GD and CMA-SH9 are generally more and stronger, of which the frequency of precipitation in CMA-GD is more than 10%, the average hourly rainfall intensity of CMA-SH9 is nearly 4 mm·h-1 stronger. The precipitation intensity above 5 mm·h-1 of CMA-MESO is mostly distributed in the southwestern and central mountainous areas, which is quite different from the observed spatial distribution. The easiest precipitation start time and precipitation peak time of the three models are 1-3 h earlier than the observation, and the easiest precipita-tion end time is 1-3 h later than the observation; The predicted values of dew point temperature and unstable energy in the upper atmosphere of the model are too large, unstable energy appears earlier, the characteristic prediction of the near-surface inversion layer is distorted, the start time of precipitation forecast tends to be earlier, and the precipitation duration is too long. The sea-land wind convergence zone along the northern coast of Hainan Island is predicted to be strong in the daytime by the three models, especially CMA-SH9, which is consistent with obvious strong precipitation intensity of the model’s output. The forecast position of the sea-land wind convergence zone along the southern coast of CMA-GD at night is westward, which corresponds to the higher frequency of precipitation along the southwestern coast of the model. CMA-MESO has strong wind speed convergence in the southwestern and central mountainous areas of Hainan Island, and the corresponding precipitation forecast intensity is stronger than the observation.