This paper utilized the method of dynamic system response curve (DSRC) to retrieve the real precipitation of the “23·7” extreme rainstorm in Beijing based on the observed discharge data in 5 typical watersheds with severely damaged rain stations and poor reliability of measured rainfall of Yongding River and Daqing River Basins. Several indices including the Root Mean Square Error (RMSE), PEARSON correlation coefficient ( R ), and Nash-Sutcliffe efficiency coefficient (NSE) were employed to validate the effectiveness of the DSRC method. The characteristics within different duration of the observed rainfall (before inversion) and the retrieved rainfall (after inversion) were both summarized and compared to the actual rainfall (approximated by the multi-source precipitation fusion), to demonstrate the accuracy of the inversion series. Results shows that the DSRC method can effectively retrieve the average areal precipitation process, while the average RMSE between the observed and actual series and between the retrieved and actual rainfall are 5.97, 3.86 mm, respectively. The average R are 0.67 and 0.82, and the average NSE are 0.27 and 0.67. The retrieval performance of the DSRC method is closely related to the XAJ model simulation performance and whether the rain-gauge station can accurately capture the peak rainfall. Even in the case of rain-gauge station damages and the data transmission interruption, the DSRC method is still able to retrieve the actual precipitation which exhibits significant continuity characteristics.