This is the official repository for the paper TCRIP-MIM: Rapid Intensification Prediction for Tropical Cyclone by Combining Memory In Memory Network with Sequential Satellite Images. We use the publicly available dataset from Taiwan University (Bai et al., 2019) as experimental data, consisting of four channels of TC satellite images with a temporal resolution of 3 hours, whose preprocessing method is also publicly available. We use infrared and passive microwave TC satellite image sequences for our experiments, each divided into 24-hour segments (8 infrared and 8 passive microwave satellite images, 16 in total), preprocessing and enhancing data as noted above, so there is no experimental error due to different data preprocessing methods. In this study, the 2003–2017 TC dataset from various global basins was divided into training (1097 TCs, 43528 events), validation (188 TCs, 7884 events), and test sets (94 TCs, 3196 events).

{"references": ["Bai, C. Y., Chen, B. F., & Lin, H. T. (2019, September). Attention-based Deep Tropical Cyclone Rapid Intensification Prediction. In MACLEAN@ PKDD/ECML. doi: 10.48550/arXiv.1909.11616", "Bai, C. Y., Chen, B. F., & Lin, H. T. (2020, September). Benchmarking Tropical Cyclone Rapid Intensification with Satellite Images and Attention-Based Deep Models. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases (pp. 497-512). Springer, Cham. doi: 10.1007/978-3-030- 67667-4_30"]}