Benchmarking Tropical Cyclone Rapid Intensification with Satellite Images and Attention-Based Deep Models

Rapid intensification (RI) of tropical cyclones often causes major destruction to human civilization due short response time. It is an important yet challenging task accurately predict this kind extreme weather event in advance. Traditionally, meteorologists tackle the with human-driven feature extraction and predictor correction procedures. Nevertheless, these procedures do not leverage power modern machine learning models abundant sensor data, such as satellite images. In addition, nature approach makes it difficult reproduce benchmark prediction models. study, we build a for RI using only images, which are underutilized traditional techniques. The follows conventional data science practices, making easier scientists contribute prediction. We demonstrate usefulness by designing domain-inspired spatiotemporal deep model. results showcase promising performance solving complex meteorological problems