Seasonal variation of the surface wind forecast performance of the high-resolution WRF-RTFDDA system over China

Abstract

Wind forecast is an essential part of both weather and renewable energy predictions. This study investigated the performances of the weather research and forecasting (WRF) model real-time four-dimensional data assimilation (RTFDDA) and forecasting system for surface wind forecast over China. The system has been running operationally since 2016 with an analysis/forecast cycle every 3 h. The surface wind forecasting skill of the system was evaluated on the 3 km output with the evaluation period of one year from June 2017 to May 2018. The model outputs were validated against observations from the stations over China objectively. The statistics of the system performance was calculated for both station-by-station and the domain average, which include bias, root mean square errors, mean absolute errors, and the correlation between observation and model outputs. The error statistics show that the high-resolution model has advantages in forecasting the detailed structures of weather features and adding values with rapidly refreshing forecast cycles. The verification results demonstrate that WRF tends to forecast surface winds with positive bias for weak wind regimes and negative for high wind regimes. On the other hand, it tends to produce positive bias for lower topography regions and negative for high topography areas. This feature does not change with seasons although the magnitude of wind bias varies with different seasons. The wind forecast bias has the largest diurnal changes in summer, with positive bias close to the coastal areas and the downstream of Yangtze river.