VRNet: A Vivid Radar Network for Precipitation Nowcasting

Wei Fang, Liang Shen, Victor S. Sheng

Abstract: Radar echo extrapolation, as a widely used approach for precipitation nowcasting, plays a pivotal role in severe convective weather warnings. Previous studies have encountered the dilemma of extrapolation ambiguity, which leads to low availability of extrapolation results. To get over the hurdle of extrapolation ambiguity, we propose a vivid radar network for precipitation nowcasting called VRNet. We first implement a multiscale spatial feature fusion module to extract richer spatial feature information, which contributes to producing clear images when extrapolating result reconstructions. Furthermore, we construct a generative adversarial network (GAN) with a ConvLSTM unit to enhance the model’s spatiotemporal information representation capability. In addition, a weighted loss function based on radar echo intensity is designed to address the distribution characteristics of radar echo intensity, improving the global averaging strategy of the mean squared error loss function. Experimental results demonstrate that the proposed model outperforms the benchmark models in the radar echo extrapolation task, which obtains a higher accuracy rate and improves the clarity of the extrapolated image.