Abstract: Daily precipitation ensemble forecasts provided by different numerical models are evaluated via “point-to-point” and spatial verifications. The results show that different numerical models have various forecasting capabilities based on different precipitation aspects. Based on the method for object-based diagnostic evaluation (MODE) technology, an object-based superensemble (OBJSUP) model is proposed, which employs similarities between the spatial structure of precipitation objects in the observation field and the forecast field as the weights of contributing models. This model is different from the gridpoint-based superensemble (GPSUP) that calculates weights using the traditional “point-to-point” error analysis. Compared with the optimal single model, the multimodel ensemble technologies (including OBJSUP and GPSUP) effectively improve the short- and medium-term precipitation forecasting skills, and the OBJSUP model generally performs better than the GPSUP model. This is primarily because the OBJSUP model can better improve predictions regarding the centroid location of the precipitation object. For an extreme heavy-precipitation event that occurred in Guangdong Province in the summer of 2018, the comparisons between the forecasts by OBJSUP, GPSUP, and a high-resolution regional model—Consortium for small-scale modeling (COSMO) demonstrate that OBJSUP and GPSUP underpredict the intensity of extremely heavy precipitation, but the COSMO dynamic downscaling exhibits certain ability to predict heavy precipitation occurring in eastern Guangdong.