Comparative Study of Different Error Correction Methods on Model Output Wind Field

The meshing forecast products is an important direction for the future development of China Meteorological Administration. With the development of the grid forecast business and approaching of Beijing Olympic Winter Games in 2022, the forecast of the wind is very important. In order to promptly correct forecast results using grid observation fusion products, grid forecasting products with higher resolution and accuracy are obtained, and the high-frequency grid wind fusion products generated by the HRCLDAS (High Resolution China Meteorological Administration Land Data Assimilation System) system of National Meteorological Information Center as observations are studied. Eight different error correction methods and two different wind field models are used to correct European Centre for Medium-Range Weather Forecasts (ECMWF) 10 m wind forecast field. The test sample time is selected from 1 January 2017 to 28 February 2017, and from 1 June 2017 to 31 July 2017, and two forecast simulations are conducted. In each trial, 24 h corrected forecast test is carried out for two start times at 1400 BT and 2000 BT, and eight different correction methods are used to correct the prediction of ECMWF 10 m wind forecast field. Grid forecast verification is performed on grid results. At the same time, grid prediction results from 8 corrected methods is interpolated to 2400 national surface meteorological stations and station forecast verification is performed on grid results. From the grid verification result and site verification result of two trials, using the latest observations as a predictor, wind forecast effects of 3-6 h is significantly improved. For the correction of the wind direction, the correction effects are slightly improved. Results show that the two-factor model with dynamic coefficient has the best correction effects on the average absolute error, accuracy and absolute error distribution frequency of both grid and site test. Sliding modeling allows the correction model to follow the trend of ECMWF 10 m wind forecast system error. After the optimal method is corrected, the wind speed error in most parts of South China, East China and North China is below 1 m·s^-1, especially the large error is significantly reduced, and the wind direction error is also reduced. However, there are still some mean absolute error of 1-3 m·s^-1 in the Qinghai-Tibet Plateau, central Xinjiang and Inner Mongolia. Due to local effects of the wind, the correction field of the interpolation to the site still has a certain gap with the actual situation of the site. If prediction result fusion technology is carried out, it is expected that there will be better grid prediction results.