Abstract: This paper introduces a novel extrapolation algorithm, the optical flow method, for the convection nowcasting. Basically, the forecasting is achieved through the following two steps. The motion vector field of the echo is firstly obtained through the calculation of the optical flow field for the radar echo. Based on this motion vector field, the radar echo can then be extrapolated. The optical flow method focuses on the change of echo. It will consider both the change of the continuous echo when calculating the optical flow field, and the change of the echo for the adjacent position. Physically, this forecast method based on the algorithm is meaningful. The quality of the radar data can be controlled by applying the median filtering method, with suppressing effectively the influence of noises. Case study shows that the quality of the echoes has been significantly improved, after the quality control by the median filtering method. The echo becomes smoother with the echo edge clearer. In addition, the echo motion vector field becomes smoother and more consistent. The extrapolating forecast of the location and shape of the radar echo within 30 and 60 min by the optical flow method is close to the real observation, which is valuable for the reference for operational forecast. By quantitatively comparing their forecasts, the optical flow method performs better than the cross-correlation method when forecasting the echo which is generated locally, with quick changes in the intensity and shape. The optical flow method can remedy the defect of the traditional cross-correlation method, and can improve the performance of the convection nowcasting system. Therefore, this new method has great application promotion value in the near future.