Accurate detection and forecasting of tropical cyclone tracks using limited climate variables and data is challenging. Here, we propose an innovative time-evolving complex network approach for detecting and forecasting high-intensity tropical cyclones (HITCs) based on mean sea level pressure and relative vorticity at 850 hPa. This approach enables us to successfully reproduce the tracks of HITCs of the Western North Pacific, achieving a mean detection rate exceeding 0.8 and a track error below 120 km in most cases. When applied to forecast 2023 HITC tracks using medium-range weather forecast data, we achieve a detection rate above 0.65 and a track error of less than 260 km for forecasts within 5 days. Our results highlight the strong potential of network-based approaches as data-integrative, physically interpretable statistical tools for HITCs detection and short-term forecasting, leveraging complex climate connectivity to enhance predictive skill.
Jiang et al. (Thu,) studied this question.