资源环境科技发展态势分析平台

A distributed modeling of hydrological impact under climate change relies on climate scenarios for multiple climate variables at multiple locations across the catchment. The site-specific or variable-specific downscaling methods only produce climate change scenarios for a specific site or specific variable, which is inadequate to drive distributed hydrological models to investigate the spatio-temporal variability of climate change impacts at the catchment scale. This study proposes an integrated framework combining quantile mapping (QM), stochastic weather generator (WG) and Empirical Copula(EC) approaches for multi-site multivariate downscaling of global climate model outputs from monthly, grid-scale to daily, station-specific scale. In this hybrid scheme, the QM method is used to spatially downscale the monthly large-scale climate model outputs; then a stochastic WG is used to temporally downscale the monthly data to daily data by adjusting the WG parameters according to the predicted changes from large-scale climate models; at last, the observed inter-site and inter-variable dependencies, the temporal persistence, as well as the inter-annual variability are restored using the ECapproach. An application of the proposed methodology is presented for statistical downscaling of the monthly precipitation, maximum and minimum temperatures from historical simulations of two Earth System Models (ESMs) to eleven weather stations over Daqing river basin in north China. The proposed methodologies arecalibrated during the period 1957-1986 and evaluated in the period 1987-2016. The results show that the proposed downscaling approach is able to reconstruct the marginally distributional statistics, inter-site and inter-variable dependencies, and temporal persistence in the downscaled data for the validation period. ALimitation is also noted, such as a possible misrepresentation of the dependence structure and inter-annual variability under a non-stationary climate condition. The proposed methodologies are useful for downscaling ensembles of large-scale climate model simulations and projections for distributed hydrological impact studies.