Abstract:The surface wind speed forecasts are typically diagnosed from the lowest model level to 10 m height, thus the accuracy of model terrain has a significant impact on the performance of surface wind speed forecasts. In this study, high-resolution terrain data with 90 m resolution from the Shuttle Radar Topography Mission (SRTM) are used to construct a more accurate representation of terrain. Combined with near-surface similarity theory, the original surface wind speed diagnostic scheme in the numerical model is optimized. Using hourly observations from more than 10 000 automatic weather stations that are included in the assessment of the China Meteorological Administration, the 24 h surface wind speed forecasts initialized at 00:00 UTC 20 July and 20 November 2023 are verified. The results show that, compared with the original scheme, the optimized scheme reduces the deviation of regional surface wind speed forecasts by 10.7% on 20 July and 9.1% on 20 November, with a decrease about 1% in root mean square error as well. Furthermore, the two-month continuous tests indicate that the optimized scheme reduces the regional deviation of the 24 h surface wind speed forecasts by 50.0% and 52.6% for the whole months of July and November 2023, respectively. This denotes a significant positive effect of the optimized scheme on the performance of the surface wind speed forecasts by the numerical model.