Regional Evaluation of Pangu-Weather Driven by Multiple NWP Initial Fields over Shandong Province, China

Accurate weather forecasting is essential for decision-making in agriculture, transportation, disaster prevention, and energy management. In recent years, data-driven artificial intelligence (AI) models have shown considerable potential for medium-range weather forecasting by substantially improving computational efficiency while maintaining competitive forecast skill. This study evaluates the regional performance of Pangu-Weather initialized with three global numerical weather prediction (NWP) systems—ECMWF-IFS, CMA-GFS, and NCEP-GFS—over Shandong Province, China, from August to December 2024. Surface forecasts are verified against observations from 123 national meteorological stations, and upper-air forecasts are evaluated using observations from three radiosonde stations. Of the three configurations, Pangu-Weather initialized with ECMWF fields (EC_PG) provides the most robust overall performance. Relative to ECMWF operational forecasts (EC), EC_PG reduces the mean 2 m temperature RMSE by approximately 13.2%, with improvement at 38 of 40 lead times and a maximum improvement of 28.3%. Its 10 m wind-speed RMSE is lower at every lead time, with a mean improvement of 5.7% and a maximum of 10.6%. For upper-air variables, all three con-figurations generally improve specific-humidity forecasts, with mean improvements of 8.6–12.8% across three stations and three pressure levels and a maximum improvement of 33.7% in an individual comparison. Improvements in upper-air temperature and wind speed vary by pressure level, station, and lead time. EC_PG performs best overall, whereas CMA_PG and NCEP_PG have larger errors than EC at several longer lead times over coastal stations and at 500–700 hPa. These results support the potential use of Pangu-Weather as a regional forecasting aid while emphasizing that performance de-pends on the operational initial fields, variable, location, and lead time.