In order to improve the low-spread of Ensemble Transform Kalman Filter (ETKF) in initial perturbation, we consider the introduction of physical uncertainty. The relative improvement degree of the two ETKF initial perturbation schemes was tested by using the initial spread. We explored the improvement mechanism by means of dynamic and water vapor condition analysis. The WRF model was used to construct an updated forecast system. One extreme heavy precipitation which happened in May, 2014 was simulated and it was made ensemble forecasting. Two initial perturbation schemes were designed by using ETKF method. One was a single physical scheme; the other was on the basis of scheme one, adding the multi-physics process in the analyzing perturbation, and the samephysical parameterization setting is adopted after the beginning of the forecasting process. The results show that the multi-scheme with multi-physical perturbation in the analysis cycle is much better than mono-ETKF scheme in initial moment spread and simulated dynamic and water vapor conditions as well as precipitation ratings. Compared with mono, the initial moment spread of multi-scheme is obviously better. It is obvious that the experiments with multi-physical perturbation scheme can improve the results, and provide more forecast information including forecast uncertainty to user decision. In the mechanism analysis of the two schemes, the significant improvement of the precipitation scheme for the multi scheme is mainly depends of the making the divergence and the moisture flux divergence better. In the analysis of spread, the improvement effect of multi-scheme in strong convection area is better than that in whole area, and the ratio of spread to root-mean-square error of each variable is roughly equal, and the rationality of ensemble forecasting system is explained. The results show that the multi-physics program shows better performance.