Sea Surface Temperature Anomaly Spatio-temporal Data Generation Based on Separated Attention

Various natural disasters are closely related to the El Niño - Southern Oscillation (ENSO), the stable generation of accurate sea surface temperature anomaly (SSTA) data is key to exploring the occurrence patterns of ENSO and preventing disasters. In this study, we introduces the Separated Attention Data Generation (SADG) model, which efficiently generates SSTA data by capturing the long-range spatiotemporal relationship. SADG achieves high performance while reducing computational complexity compared to standard self-attention based models, generates up to 18 months of accurate SSTA data. Experiments on CMIP datasets show that its performance is better than other models on various lead times.