Dynamical downscaling of surface air temperature and precipitation using RegCM4 and WRF over China

Two long continuous regional climate simulations over China have been carried out using the Regional Climate Model version 4 (RegCM4) and Weather Research and Forecasting model (WRF). The Simulations are forced by NCEP Reanalysis II data with a horizontal grid spacing of 30 km during 1981–2010. It is demonstrated that the RegCM4 and WRF had pronounced temperature and precipitation downscaling ability, producing more regional details and smaller biases than the driving R2. Overall, WRF tended to better capture the temperature and precipitation pattern and magnitude, daily temperature frequency, the monsoon rain belt movement, and seasonal precipitation variations over most wet regions, while larger deficits than RegCM4 were shown over some dry regions. The extreme precipitation indices of the two RCMs were quite different, with strong regional and seasonal dependence. WRF was better at simulating the annual mean temperature and precipitation trends, with higher spatial pattern correlations. The optimal ensemble approach combining the advantages of RegCM4 and WRF showed improved simulation compared to the individual models. The optimal ensemble reduced the annual temperature biases from the two models by 15–30%, and increased the precipitation spatial pattern correlations by 0.08–0.13. Further works are needed to improve the performances of the ensemble approach by using more RCMs.