PN-HGNN: Precipitation Nowcasting Network Via Hypergraph Neural Networks

Precipitation nowcasting within 2 h is an important and hard issue in the weather research area. Benefiting from the outstanding nonlinear relationship modeling capability, methods based on deep learning (DL) have achieved significant success in the task of precipitation nowcasting compared to the others. However, existing DL-based methods always disregard the intricate high-order correlations and lack substantial connections with the evolution of the precipitation system, which would lead to blurred forecasts and implausible predictions. To address these issues, we proposed a new Precipitation Nowcasting Network within a 2-h model based on the Hypergraph Neural Network (PN-HGNN). In this work, a Hypergraph Neural Network is first adopted for extracting spatiotemporal dynamic echo features. Second, regulation evolution is in charge of capturing the memory features to guide the extrapolation. Finally, we design a dual branch module to extrapolate the radar echoes. The proposed model has been assessed on the dataset HKO-7. The experimental results demonstrate that PN-HGNN achieved better prediction performance than the six representative echo extrapolation models.