Using probabilistic climate change information from a multimodel ensemble for water resources assessment

Increasing availability of ensemble outputs from general circulation models (GCMs) and regional climate models (RCMs) permits fuller examination of the implications of climate uncertainties in hydrological systems. A Bayesian statistical framework is used to combine projections by weighting and to generate probability distributions of local climate change from an ensemble of RCM outputs. A stochastic weather generator produces corresponding daily series of rainfall and potential evapotranspiration, which are input into a catchment rainfall-runoff model to estimate future water abstraction availability. The method is applied to the Thames catchment in the United Kingdom, where comparison with previous studies shows that different downscaling methods produce significantly different flow predictions and that this is partly attributable to potential evapotranspiration predictions. An extended sensitivity test exploring the effect of the weights and assumptions associated with combining climate model projections illustrates that under all plausible assumptions the ensemble implies a significant reduction in catchment water resource availability.