Prediction of upper-level wind velocity by fusing local objective analysis data and nearsurface meteorological observations

Abstract

The near-surface meteorological observation network has been established, which provides local observations to improve local objective analysis data using data assimilations. The development of high-resolution meteorological analysis data provides complementary data for predicting wind speed above urban areas. However, high-precision meteorological data requires observation data assimilation to correct the accumulated errors in weather forecast models, which means that meteorological analysis data are released with a delay. Therefore, this study suggests the Proper Orthogonal Decomposition with Linear Stochastic Estimation as a data fusion method to monitor upper-air wind speed by fusing meteorological near-surface observation data and high-resolution local objective analysis data. By comparing the predicted wind speed with Doppler lidar observations, the results show that the correlation coefficients between predicted and observed wind speed are approximately 0.84 at about 100 m and 0.86 at 200 m. RMSE of predicted wind speeds compared to observations is about 1.8 m/s in both observed heights.