A Bias-corrected CMIP6 Dataset and its Application and Validation in Dynamical Downscaling Simulation of Asia

Dynamical downscaling is an important approach to obtaining fine-scale weather and climate information. However, dynamical downscaling simulations are often degraded by biases in the large-scale forcing itself. We constructed a bias-corrected global dataset based on 18 models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) and the European Centre for Medium-Range Weather Forecasts Reanalysis 5 (ERA5) dataset. The bias-corrected data have an ERA5-based mean climate and interannual variance, but with a non-linear trend from the ensemble mean of the 18 CMIP6 models. The dataset spans the historical time period 19792014 and future scenarios (SSP245 and SSP585) for 20152100 with a horizontal grid spacing of (1.25 1.25) at six-hourly intervals. Our evaluation suggests that the bias-corrected data are of better quality than the individual CMIP6 models in terms of the climatological mean, interannual variance and extreme events. We carried out three 35-year dynamical downscaling simulations (WRF_ERA5, WRF_GCM, WRF_GCMbc) using WRF model driven by the ERA5, individual CMIP6 model output, and bias-corrected CMIP6 dataset, respectively. The WRF domain covers the South and East Asian continent and the western Pacific with a grid spacing of 25 km. We take the ERA5-driven WRF simulation as reference data to validate the other two simulations. The results suggest that, compared with the WRF_GCM, the WRF_GCMbc greatly improves the downscaled climate, including the climatological mean, diurnal cycle, annual cycle, inter-annual variabilities and climate extremes.