Tropical Cyclone Intensity Estimation Based on Satellite Microwave Brightness Temperature Data: Utilizing the Maa Network and Dual-Branch Network Training Method

Accurately estimating the intensity of tropical cyclone (TC) is crucial for reducing casualties and losses. Currently, TC intensity estimation methods based on infrared images are widely used. This paper leverages brightness temperature data from the Advanced Technology Microwave Sounder (ATMS) to derive three-dimensional (3D) TC temperature anomaly data through retrieval and interpolation. Using the proposed dual-branch multi-scale adaptive attention network (dual-branch MAA), the study connects the 3D data with the TC maximum wind speed. In the design of the MAA network, the warm-core feature adaptive attention network (WFAA) is introduced to adaptively capture the 3D warm core features within the TC temperature anomaly data, thereby enhancing the accuracy of TC intensity estimation. The designed dual-branch training network helps mitigate the underestimation of high-intensity TCs to some extent. Experimental results indicate that, in the 2018 and 2019 TCs, the mean absolute error (MAE) for the TC intensity estimation is 3.70 m/s in the North Atlantic (NA), 3.58 m/s in the Northeastern Pacific (NEP), and 3.15 m/s in the Northwestern Pacific (NWP), respectively. Therefore, the proposed TC intensity estimation method shows great potential for accurately estimating TC intensity.