Predicting the Intensity of Tropical Cyclones over the Western North Pacific Using aDual-Branch Spatiotemporal Attention Convolutional Network

his paper proposes a spatiotemporal attention convolutional network (STAC-Pred) that leverages deep learn-ing techniques to model the spatiotemporal features of tropical cyclones (TCs) and enable real-time prediction of their intensity.The proposed model employs dual branches to concurrently extract and integrate features from intensity heatmaps and satellitecloud imagery. Additionally, a residual attention (RA) module is integrated into the three-channel cloud imagery convolutionprocess to automatically respond to high wind speed regions. TC's longitude, latitude, and radius of winds are injected into themulti-timepoint prediction model to assist in the prediction task. Furthermore, a rolling mechanism (RM) is employed tosmooth thefluctuation of losses, achieving accurate prediction of TC intensity. We use several TC records to evaluate and vali-date the universality and effectiveness of the model. The results indicate that STAC-Pred achieves satisfactory performance.Specifically, the STAC-Pred model improves prediction performance by 47.69% and 28.26% compared to the baseline (officialinstitutions) at 3- and 6-h intervals, respectively.