Extreme heavy rainfall events are occurring with heightened frequency due to intensified global climate warming, posing growing risks to public safety and social development. It is of great significance for disaster prevention and reduction to study the short-term heavy rain. Based on the precipitation data from regional automatic stations in Hubei Province, short-term heavy rain probability forecast products, and mesoscale high-resolution numerical model data, this study adopts neighborhood optimal probability and multi-model integration methods for the short-term heavy rainfall location forecasting in Hubei Province with a lead time of up to 12 h. The results show that the neighborhood method obviously improves the prediction accuracy of the mesoscale numerical model for short-term heavy rain, with the area neighborhood method outperforming the single-point neighborhood method. The optimal area probability of CMA-MESO, CMA-SH9 and WH-RUC modes are all 5%, and the optimal neighborhood radius is 50, 60 and 60 km respectively. The multi-mode integration method shows significant improvement compared to the single-point neighborhood method with one model. The threat scores for all lead times indicate positive forecast skill, improving by 0.014 and 0.020 from April to September in 2023 and 2024, respectively. The improved multi-model integration method shows a substantial increase in probability of detection, especially in accuracy of various severe convection prediction in Hubei Province on August 7, 2023 and June 28, 2024.

In view of three rainstorm processes (“6·19”,“7·5” and “7·19” processes) during Meiyu period in Hubei Province in 2016, firstly, the sounding data of Hankou station were compared with the horizontal wind speed and wind direction from Hankou and Xianning wind profile radar stations. It was found that the horizontal wind speed below 3 km from Hankou wind profile radar station was close to sounding data in the “6·19” and “7·5” processes; the horizontal wind direction and wind speed below 8 km from Xianning wind profile radar station were basically consistent with the sounding data in the three processes. Combined with the data of conventional and encrypted automatic weather stations, the horizontal wind field, average vertical velocity and its variability, vertical shear of horizontal wind speed and atmospheric refractive index structure constant $C_{n}^{2}$ were analyzed by using wind profile radar. The results are as follows: (1) The southwest wind speed increased significantly before the beginning of precipitation. The invasion of dry and cold air in the middle layer and the mesoscale easterly air flow formed by the ground cold pool were the main reasons for the occurrence of strong winds with more than and equal to 17.2 m·s^-1 at 50 stations in the “6·19” process, and the long-term maintenance of the southwest jet and the easterly air flow below 1 km in the “7·5” and “7·19” processes were the inducements for the long duration of short-term heavy precipitation. (2) The vertical shear of horizontal wind speed, the variation of average vertical velocity and its variability with height observed by wind profile radar were small, and strong upward movement was mainly concentrated below 4 km height. (3) Before the occurrence of heavy precipitation, the atmospheric water vapor content had an increasing process, and the water vapor content in the whole layer was deep. The disappearance of the large value area of $C_{n}^{2}$ corresponded to the end of precipitation.