Verification of Artificial Intelligence models on tropical cyclone forecasting over the western North Pacific and the South China Sea in 2024

This study verifies the performance of Artificial Intelligence Weather Prediction (AIWP) models in forecasting the track and intensity of tropical cyclones (TCs) over the western North Pacific and the South China Sea during the 2024 season. Five-day TC track and intensity forecasts from six AIWP models (Fengqing, FengWu, FuXi, Pangu-Weather, AIFS, and GraphCast) have been evaluated. These models were under operational trial at the National Meteorological Centre of the China Meteorological Administration (NMC/CMA) and were evaluated using the CMA verification procedure. The results indicate that the verified AIWP models demonstrated positive TC track forecast skill scores relative to the climatology and persistence model. Whereas most AIWP models encountered large extreme forecast errors particularly for long lead times. Furthermore, the AIWP models consistently underestimated TC intensity to varying degrees at all forecast lead times and struggled to accurately capture the trends of intensity changes. This study highlights both the potential and limitations of AIWP models in TC forecasting and provides insights into future improvements in model development and application.