Tropical cyclones (TCs) can cause significant economic damage and loss of life in coastal areas. Therefore, TC prediction has become a crucial topic in current research. In recent years, TC track prediction has progressed considerably, and intensity prediction remains a challenge due to the complex mechanism of TC structure. In this study, we propose a model for short-term intensity prediction based on adaptive weight learning (AWL-Net) for the evolution of the TC’s structure as well as intensity changes, exploring the multidimensional fusion of features including TC morphology, structure, and scale. Furthermore, in addition to using satellite imageries, we construct a dataset that can more comprehensively explore the degree of TC cloud organization and structure evolution. Considering the information difference between multi-source data, a multi-branch structure is constructed and adaptive weight learning (AWL) is designed. In addition, according to the three-dimensional dynamic features of TC, 3D Convolutional Gated Recurrent (3D ConvGRU) is used to achieve feature enhancement, and then 3D Convolutional Neural Network (CNN) is used to capture and learn TC temporal and spatial features. Experiments on a sample of northwest Pacific TCs and official agency TC intensity prediction records are used to validate the effectiveness of our proposed model, and the results show that our model is able to focus well on the spatial and temporal features associated with TC intensity changes, with a root mean square error (RMSE) of 10.62 kt for the TC 24 h intensity forecast.

Tropical cyclones (TC) are among the most deadly and damaging natural disasters in coastal areas worldwide. Traditional forecasting methods face challenges as they neglect crucial spatial information related to intensity changes and require substantial human and material resources. Moreover, current deep learning approaches often rely on reanalysis data from observations far from land, making them challenging to acquire and operationalize. In response to these issues, the paper introduces the Tropical Cyclone Intensity Prediction Network (TCIP-Net), which, while maintaining interpretability, successfully extracts rich convective structural information from the infrared channel of satellite imagery. We present the spatio-temporal evolution trajectory of TC radial structural information through Hovmöller diagrams. Additionally, we construct a subnetwork with one backbone convolution and four branch convolution operations to extract asymmetric information of TC structure. The convective core reveals the distribution of convective systems around the eye, aiding in targeted attention to convective information in infrared imagery. The model aims to quantitatively explain the contributions of satellite imagery (infrared, microwave), convective structure, and key physical factors to TC intensity prediction task. We utilize multiple TC cases to assess and validate the applicability and effectiveness of the model. The results indicate that TCIP-Net achieved good performance. This approach provides practical guidance for predicting TC intensity using advanced artificial intelligence-based methods and is expected to complement operational models.