Abstract：The article focuses on key technological improvements and effectiveness validation of optical flow methods in radar composite reflectivity extrapolation.By introducing convolutional neural networks to optimize optical flow estimation and combining multiscale structures with regularization techniques,the extrapolation accuracy is effectively enhanced.Experimental datasets constructed using radar data from the Mianyang region demonstrate that the proposed method outperforms traditional extrapolation algorithms in metrics such as root mean square error and structural similarity,particularly showing superior warning capabilities during severe convective weather like heavy rainstorms.This provides reliable technical support for nowcasting and disaster warning.
李熠, 汪剑, 易洛锋. 基于光流法的组合反射率外推产品研究及应用[J]. 气象水文海洋仪器, 2026, 43(5): 137-140.
Li Yi, Wang Jian, Yi Luofeng. Research and application of composite reflectance extrapolation products based on optical flow method. Meteorological Hydrological and Marine Instrument, 2026, 43(5): 137-140.
2026, Vol. 43 
