PERFORMANCE EVALUATION OF CMIP5 MODELS FOR PRECIPITATION AND TEMPERATURE OVER HAIHE RIVER BASIN, CHINA

The quantitative evaluation of GCMs performance is vital to reduce the uncertainty of regional climate change impact research. A cascade progressive -like analysis approach is applied to evaluate the five phase of the Coupled Model Intercomparison Project (CMIP5) models' ability by comparing the models' outputs with ground observations (i.e, precipitation and temperature) over the Haihe River Basin from 1960 to 2004. The results show that for the chosen seventeen GCMs after preliminary screening, all models can reproduce the intra-annual characteristics of temperature and precipitation well, and show much better agreement with temperature than precipitation. However, the majority of the models tend to underestimate the annual mean temperature, with a mean underestimation of 0.34-2.82 degrees C. For precipitation, all models overestimate the annual mean precipitation, with a mean overestimation of 110/0-97%. The step of performance evaluation for large-scale atmospheric variables and preliminary screening of models carried out before performance evaluation for ground climate variables can reduce uncertainties from the subjective selection of model and model error between large-scale atmospheric variables (i.e., relative humidity at 500hPa and 850hPa) and ground precipitation. Overall, the five models, namely, HadGEM2-ES, CESM1-BGC, MPI-ESM-LR, CNRM-CM5 and ACCESS1.3. are recommended as optimal choice for regional impact studies over the HRB.