Simulation and Projection of Daily Precipitation over the Yangtze–Huaihe River Basin Based on Multi-Model Dynamical Downscaling and Four Bias Correction Methods

Four statistical bias correction approaches based on transform cumulative distribution function method (CDF-t), equidistant cdf matching method (EDCDF), quantile-quantile adjustment (QQadj) and gamma cdf matching method (GamCDF) are introduced in this paper. The simulated results of five dynamical downscaling models (BCC-CSM1-1m, CNRM-CM5, FGOALS-g2, IPSL-CM5A-MR and MPI-ESM-MR) are corrected using these four methods, the simulation capabilities of each method for daily precipitation over the Yangtze-Huaihe River Basin have been assessed, and then the best method has applied to projection in the 21st century under the RCP4.5 emission scenario. The results show that four bias correction approaches perform well in simulating daily precipitation over the Yangtze-Huaihe River Basin, especially for CDF-t. The simulated probability distribution function (PDF) curves are closer to the observations with the Brier score (BS) less than 0.0473% and the Significance score (S_score) greater than 0.905. It’s noted that the BS is average decreased by 0.0765%, 0.0326%, 0.443% and 0.0592%, and S_score are average increased by 0.0906, 0.0261, 0.0396 and 0.0536 respectively, which indicates that CDF-t is the best. The corrected monthly cycle based on CDF-t is closer to the observations than other three methods. The absolute values of relative errors for precipitation indices are reduced from 15.94%~70.84% to 1.96%~15.12%, and the spatial correlation coefficients are increased to 0.7 or more for CDF-t. Furthermore, the projected results show total and extreme precipitation both increase gradually during three periods of 21st century, in that PRCPTOT is increased by 2.67%, 6.07%, 6.88% , SDII is increased by 5.80%, 6.17%, 6.50%, R10 is changed by -1.10%, 0.88%, 3.32%, and pq95 is increased by 1.98%, 4.95%, 6.75%.