Improving Extened-range Prediction of Midsummer Maximum Temperature over Southern China Using a Dynamical Downscaling Approach

This study aims to enhance the extended-range prediction of midsummer (July) maximum temperature (Tmax) through a dynamical downscaling method. We compare the prediction skills of July Tmax over southern China between the NCEP Climate Forecast System version 2 (CFSv2) and a high-resolution Weather Research and Forecasting (WRF) model, using gridded Tmax observation data and ERA5 reanalysis data as benchmarks. The WRF model is driven by CFSv2 multi-member ensemble hindcast and forecast data. Results indicate that the WRF model improves Tmax prediction across China, with particularly significant enhancement over the southern region of the middle and lower reaches of the Yangtze River, although a systematic cold bias remains. By applying bias correction to the daily Tmax simulations from both models, we find that the corrected WRF predictions exhibit marked improvement for both the annual and extended-range Tmax. Furthermore, this study explores the physical mechanisms contributing to the improved predictability in the regional model. The WRF model, with its refined physical parameterization schemes, better simulates middle to lower tropospheric geopotential height fields, as well as surface sensible and latent heat fluxes. These results demonstrate that the dynamical downscaling approach can significantly improve the temperature prediction in southern China, highlighting the potential applicational value of this method for extended-range high-temperature forecasting.