Tropical cyclone (TC) is a type of severe weather system that damages human property. Understanding TC's mechanics is crucial for disaster management. In this study, we propose a multi-task learning framework named Multi-task Graph Residual Network(MT-GN) to classify and estimate the intensity of TC from FY-4A geostationary meteorological satellite images. And we construct a new benchmark dataset collected from the FY-4A satellite for both TC classification and intensity estimation tasks. Four different methodologies to classify TC and estimate the intensity of TC are fairly compared in our dataset. We discover that accurate classification and estimation of TC, which are usually achieved separately, require co-related knowledge from each other. Thus, we train a convolution feature extractor in a multi-task way. Furthermore, we build a task-dependency embedding module using the Graph Convolution Network(GCN), which further drives our model to reach better performance. Finally, to overcome the influence of the unbalanced distribution of TC category samples, we introduced a class-balanced loss to our model. Experimental results on the dataset show that the classification and estimation performance are improved through our improvement. With an overall root mean square error (RMSE) of 9.50 knots and F1-score of 0.64, our MT-GN model achieves satisfactory performance. The results demonstrate the potential of applying multi-task learning for TC study.
@article{xxx,
author = {Zhitao Zhao, Zheng Zhang, Ping Tang Xiaofeng Wang Linli Cui},
title = {MT-GN: Multi-task Learning based Graph Residual Network for Tropical Cyclone Intensity Estimation},
journal = {xxx},
year = {xxx},
}