Abstract:Based on the Storm-Scale Rapid Assimilation and Forecast System (SSRAFS) operated by the Chongqing Meteorological Administration, along with MICAPS ground observation and upper-air observation data, this study conducted experiments on hourly surface air temperature forecasts within 96 hours for the Chongqing region. Model Output Statistics (MOS) and MOS with Prior Observation Predictors (OMOS) methods were utilized, and a comparative analysis was performed by using the SSRAFS surface air temperature forecasts as a reference. Results indicate that MOS outperforms SSRAFS in terms of forecasting skill for lead times ranging from 1 to 96 hours. On average, MOS reduces the RMSE of surface air temperature forecasts by 1.22℃, while increasing CC and HR2 by 0.006 and 20.4% respectively. For 1-7 h forecasts, MOS achieves an average decrease in RMSE of 1.70℃, and average increases in CC and HR2 of 0.07 and 34.5%, respectively. MOS exhibits a more noticeable improvement in the northeastern and south-central regions of Chongqing. Comparatively, OMOS performs better than MOS for short-term temperature forecasts, particularly within 1-7 h lead times, with an average decrease in RMSE of 0.43℃, and average increases in CC and HR2 of 0.008 and 8.3%, respectively. OMOS demonstrates even better performance during 1-4 h lead times, with an average decrease in RMSE of 0.66℃, and average increases in CC and HR2 of 0.13 and 12.3%, respectively, compared to MOS. Therefore, OMOS further enhances the forecasting skill of surface air temperature based on MOS, with significant improvements observed in the northeastern and south-central regions of Chongqing.