Abstract: By using network radar, national gridded real-time observation datasets and the hourly precipitation observation data at the automatic weather stations in the Yangtze River Basin, the radar extrapolation precipitation forecast product of the Yangtze River Basin based on the variational optical flow method is established. The results of threat score in 39 sub-basins and deviation analysis in the whole basin for 17 flood weather events in the Yangtze River Basin in 2020 were analyzed. The results show that: 1) 0-3 h rainfall nowcasting with the variational optical flow method has a positive correlation with the radar distribution range. The denser the distribution is, the better the prediction result is. 2) 0-3 h rainfall nowcasting with the variational optical flow method can better track the movement and evolution of the weather system. The precipitation concentration area is also in a relatively high zone for threat score of rainfall nowcasting. 3) Threat score of 0-3 h rainfall nowcasting with the optical flow method is the highest in the first hour, and the prediction skill decreases exponentially with the change of prediction time. 4) Comparing 0-3 h rainfall nowcasting and the gridded real-time observation datasets of precipitation, it is found that the absolute error rate of the precipitation forecast of the variational optical flow method is less than 5% when the precipitation is less than 10.0 mm. When the precipitation is over 10.0 mm, the absolute error rate is 6-11%. 5) The result of 0-3 h rainfall nowcasting with the variational optical flow method is applied to the quantitative precipitation forecast of the Yangtze River Basin replacing the output results of the current direct numerical model, which has a strong practical significance to improve the accuracy of flood forecast.