Abstract: In order to further understand the performance of convection-permitting models for precipitation forecast in isolated, small-scale topographic regions, the precipitation forecast products of RMAPS-ST operational system at 3 km horizontal resolution are thoroughly evaluated against hourly precipitation data collected at 80 stations of National Meteorological Surface Observation Network in the Mountain Tai area during the warm season (May to September) of 2017. The evaluation is based on hourly precipitation characteristics and differences between rainfall forecasts starting at 20:00 BT and 08:00 BT are compared. The results of this study show that RMAPS-ST can reproduce the spatial distribution characteristics of hourly precipitation in the study area, but the forecast rainfall amount in the southwest (Taishan Station and its northeast side) of the mountain area is lower (higher) than observations. There are large deviations of precipitation forecast at early morning and afternoon hours. Taking Taishan Station as an example, RMAPS-ST underestimates the rainfall frequency in early morning, which may be related to the forecast deviation of the model on the development and evolution of precipitation system and missed forecasts of weak rainfall events at Taishan Station in the early morning. The intensity of early morning rainfall at Taishan Station is different between forecasts initialized at different times. In the forecast results starting at 20:00 BT, the intensity of early morning rainfall at Taishan Station is significantly overestimated, which leads to the fact that its amount of precipitation is larger than observations. However, the overestimation of precipitation intensity is not obvious in the forecast results starting at 08:00 BT. RMAPS-ST overestimates the frequency of afternoon rainfall at Taishan Station in the results starting at 08:00 BT, which is related to more predicted rainfall events with short duration in the afternoon. The deviation of thermal field and water vapor field is the possible reason for the false precipitation forecast in the afternoon. The characteristics of hourly precipitation have been applied to the operational evaluation system of the regional numerical forecast model of China Meteorological Administration. The results of this paper also indicate that this kind of evaluation is helpful to better understand the forecast ability of the kilometer scale numerical forecasting models for the forecast of diurnal variation of precipitation, and thus provides a more solid scientific basis for the application of precipitation products.