Abstract: Based on a high-resolution (3 km) model with convection-allowing microphysics scheme,a storm-scale ensemble forecast (SSEF) was carried out for a heavy rainstorm which occurred in the western mountainous area of Hubei Province on 22 April 2018. The perturbations used by the SSEF in the initial and lateral boundary conditions were downscaled from a global ensemble forecast system (GEFS). The comparison of the SSEF with the GEFS shows that the SSEF is superior to the GEFS in the following three aspects. (1) The ensemble mean forecast of the SSEF was more accurate in both precipitation amount and location. (2) The ensemble spread in precipitation forecasts of the SSEF is more reasonable. The forecast range (minimum to maximum) predicted by the SSEF has a higher chance to encompass the observed precipitation,and the spread of the SSEF matched the forecast error better (a better spread-skill relation). (3) The SSEF produced more superior probabilistic heavy-rain forecasts in all spatial and temporal scales,especially in smaller spatial and shorter time scale. This study suggests that,for those local and short-duration mountainous rainstorms,storm-scale ensemble forecasting has great potential to improve over the global ensemble forecasting.