Atmospheric Prediction Model Based on Multi-Graph Spatiotemporal Convolutional Network

Atmospheric prediction plays a crucial role in various applications, including weather forecasting, climate simulation, and environmental management. However, existing models face challenges when it comes to accurately modeling intricate spatial heterogeneity in atmospheric data. To address this issue, we propose a novel approach called Multi-Graph Spatiotemporal Convolutional Network (MSTGCN) for characterizing atmospheric data. The Multi-graph Diffusion Graph Convolutional Network is specifically designed to capture spatial dependencies by incorporating adaptive graph embedding techniques, overcoming the limitations of traditional GCNs in capturing long-range spatial dependencies. In addition, we utilize two modules, TCN and GRU, to effectively model short-term and long-term temporal dynamics, respectively. Through extensive experiments conducted on a real-world atmospheric dataset, we demonstrate that our model achieves competitive performance in atmospheric prediction.