This study evaluates the performance of operational tropical cyclone (TC) track and intensity forecasts over the western North Pacific (WNP) during the 2025 season. The 2025 WNP TC season was very active, featuring 27 named TCs. Forecasts from five official typhoon prediction agencies, eight deterministic Numerical Weather Prediction (NWP) models, six deterministic data-driven Artificial Intelligence Weather Prediction (AIWP) models, and five Ensemble Prediction Systems (EPS) are verified at lead times from 12 to 120 hours (h). Results indicate that TC track forecasts achieved improved mean direct position errors (DPEs) at 24 h compared to 2024, with some reaching historically low error levels. However, the large westward bias in forecasts for Super Typhoon Halong (TC2522) substantially influenced long-lead time statistics. Differences in track errors among AIWP models increased progressively beyond 48 h. For intensity, mean absolute errors (MAEs) increased with lead time. NWP models outperformed AIWP models, with the latter consistently underestimating intensity across all lead times, though the degree of underestimation decreased relative to 2024. These findings highlight not only the improvement in TC track forecast capability but also the persistent challenges in intensity forecasting, particularly for AIWP models.
Yang et al. (Wed,) studied this question.