| Statistical downscaling methods(SDMs)are widely used to bridge the gaps between global climate models(GCMs)and catchment hydrological models(CHMs)for assessing climate change impacts on hydrology.SDMs can be classified into three categories: perfect prognosis(PP),model output statistics(MOS)and stochastic weather generator(SWG).In order to provide guidance to end-users for choosing appropriate statistical downscaling methods for climate change impact studies,understanding and inter-comparing of SDMs are required.In addition,downscaling methods are a large source of uncertainty in quantifying climate change impacts on hydrology,how to reduce the uncertainty is an important issue considered by researchers and policy makers.Therefore,this study evaluated and compared the performances and uncertainties of 10 commonly used statistical downscaling methods selected from the three categories in climate downscaling and hydrological modeling.This study was implemented over two watersheds with different climatic and hydrological regimes.The main contents and conclusions are as follows:(1)A series of climatic and hydrological regime measures were employed to evaluate the performances of statistical downscaling methods in reproducing observed climatic and hydrological variables.Nonstationary scenarios and different-quality predictors were used to address the influence of climate nonstationarity and predictor quality on downscaling performances,respectively.Results show that: in general,no single statistical downscaling technique or method type tends to be consistently superior to others in all aspects of climate downscaling and hydrological modeling,and the relative performances of downscaling methods depend on the criterion selected based on research emphasis.Climate nonstationarity and predictor quality can also influence the performances of downscaling methods,and better downscaling results were achieved by using higher-resolution,less biased predictors and under more stationary climate.(2)Climate change projection data from a global climate model were used to assess the uncertainty caused by different statistical downscaling methods in projecting climate change and hydrological response.In projecting precipitation,the perfect prognosis methods projected much lower precipitation than other methods,which caused great uncertainty.In projecting temperature,the analog method(belonging to PP category)based on historical observed data underestimated future temperature increase compared to other downscaling methods.When the downscaled future climate was used as input to hydrological model for watershed runoff,in the Xiangjiang Watershed where runoff is mainly controlled by rainfall,the perfect prognosis methods also projected much smaller runoff than other methods.But in the Manicouagan 5 Watershed where snowmelt contributes a lot to runoff,the uncertainty caused by perfect prognosis methods was much smaller than in the Xiangjiang Watershed.The climate and runoff projections from other downscaling methods were quite similar,but the uncertainty arised from different-category downscaling methods was larger than that from same-category downscaling methods.(3)Several runoff-based weighting methods were employed to investigate the possibility of reducing the uncertainty of future climate and runoff projection by using weighting strategy.Results show that when only considering the simulation of historical runoff,runoff-based weighting methods had limited impacts on the climate,runoff projections and their uncertainties.But if considering the simulation of both historical and future runoff,as perfect prognosis methods gave quite different future runoff projection than other downscaling methods,their projections were considered ‘less important' by the weighting method via giving smaller weights.In this case,the final climate,runoff projections and their uncertainties were completely different to the results from other weighting methods,especially the uncertainty was narrowed. |
PDF Full Text Request