Matching typhoons with similar characteristics is a promising strategy for enhancing disaster preparedness, providing decision-makers with early insights into the potential impacts of approaching typhoons. Existing studies in this area remain limited and have focused primarily on trajectory-based similarity using past typhoons. Notably, no prior research has applied similarity matching to the database for Policy Decision-making for Future Climate Change (d4PDF). This state-of-the-art large-ensemble dataset facilitates proactive analysis beyond the constraints of conventional forecasts, which become available only after typhoon formation. This study selected 32, 517 typhoons from the d4PDF and 280 historical typhoons from the Japan Meteorological Agency’s best-track dataset (1951–2024) as typhoons affecting the Chikugo River Basin, southwestern Japan. Under the +4K future climate scenario, typhoons are projected to exhibit greater intensity, with an average increase of 33% in the maximum wind speed and an extension of 1.4 days in their average lifetime. As a novel contribution, this study developed a deep learning-based similarity matching model that considers key typhoon features: trajectory, maximum wind speed (Vmax), and minimum central pressure (Pmin), and lifetime. The model was validated using 280 historical typhoons as forecast input data. The matching process demonstrated high reliability, with overall median differences of +9% in Vmax, -3.4% in Pmin, and +1.8% in lifetime across 280 typhoons. Owing to its computational efficiency, the model is suitable for real-time operational use, supporting adaptive planning and enhancing early warning systems in vulnerable regions.
NGUYEN et al. (Thu,) studied this question.