Abstract: Tropical cyclone intensity estimation is of great practical significance for disaster prevention and mitigation. However, the existing methods still have some limitations in estimation accuracy, spatiotemporal information utilisation and channel feature extraction capability. In order to solve the above problems, a novel tropical cyclone intensity estimation model, STDA-Net, based on spatiotemporal feature fusion and dual attention interactive networks, is proposed, with the Northwest Pacific Ocean serving as the study area. The model consists of three core modules: The spatial feature extraction module is used to capture the spatial characteristics of tropical cyclones; the temporal feature extraction module aims at extracting the temporal characteristics of its changing process; and the spatial-channel interaction module enhances the extraction of vital information by interacting spatial and channel attention. The experimental findings reveal that STDA-Net outperforms the other deep learning methods compared in estimating the intensity of tropical cyclones in the Northwest Pacific Ocean, achieving RMSE and MAE of 9.42 knots and 7.22 knots, respectively, and exhibiting robust estimation performance for multiple tropical cyclone events from 2019 to 2021. It is shownthat STDA-Net performs well in estimating the intensity of tropical cyclones in the Northwest Pacific, confirming its accuracy and superiority.