Comparison on the Influence of Different Model Backgrounds on the 100-m Resolution Integrated Forecasts for the Temperature and Wind in Complex Terrain

This study evaluates the impact of different NWP (Numerical Weather Prediction) model backgrounds on the accuracy of fusion analysis for the surface temperature at 2-m height and wind at 10-m height, as well as hourly forecasts for the future 1–24 h at a high spatial resolution of 100 m. The research focuses on the outdoor mountainous competition areas of the Beijing Winter Olympics, utilizing RISE (Rapid-refresh Integrated Seamless Ensemble) system. This approach integrates the mesoscale CMA-BJ model, with a 3-km resolution, and the global-scale ECMWF model, with a resolution of 0.125°. The results show the following: (1) High-resolution refined fusion analysis products at 100-m resolution can be formed by using different model backgrounds through RISE downscaling over complex terrain and the rapid integration of observational data. However, the influence of different model background fields varies obviously depending on the meteorological elements. (2) For temperature analysis, the spatial distribution of RISE analysis field based on the forecast data of CAM-BJ and ECMWF models is basically the same, and MAE (mean absolute error) of analysis is less than 0.2℃. (3) For wind analysis, the refinement level of RISE high-precision wind field can be improved by adopting high-resolution regional model background rather than coarse resolution global model. (4) For temperature forecasts, the predictive precision using the lower-resolution ECMWF model as the background is comparable to that of the CMA-BJ model using higher spatial and temporal resolution. Furthermore, temperature forecasting accuracy improves consistently, with the average forecast MAE reduction rates of 10.5% for the Winter Olympic alpine stations and 7.0% for all stations within the RISE region. (5) For wind forecasts, the 1–6 h forecast MAE of the RISE for Winter Olympic alpine stations based on CAM-BJ and ECMWF as backgrounds is 1.42 m s⁻¹ and 1.30 m s⁻¹, and the 7–24 h forecast MAE is 1.52 m s⁻¹ and 1.54 m s⁻¹, respectively. The mean 1–24 h forecast MAE of all stations in RISE region is 1.38 m s⁻¹ and 1.24 m s⁻¹, respectively. The precision level of 100-m resolution predictions is lower when the ECMWF model serves as the background compared to the CMA-BJ model. The results of this study provide important insights into the role of model background in integrated forecasting at 100-m resolution. The study holds important scientific significance and practical value for improving the accuracy of weather forecasting in complex terrain.