Climate model performance and change projection for freshwater fluxes: Comparison for irrigated areas in Central and South Asia

•No single climate model outperforms other models for all key freshwater variables.•Scale effects not evident for direct global model application to freshwater assessment.•Overall less accurate and more uncertain modeling for freshwater than for temperature.•Model agreement on future evapotranspiration increase in both investigated regions.•Model agreement on future runoff decrease in Central Asia.

Study regionThe large semi-arid Aral Region in Central Asia and the smaller tropical Mahanadi River Basin (MRB) in India.Study focusFew studies have so far evaluated the performance of the latest generation of global climate models on hydrological basin scales. We here investigate the performance and projections of the global climate models in the Coupled Model Intercomparison Project, Phase 5 (CMIP5) for freshwater fluxes and their changes in two regional hydrological basins, which are both irrigated but of different scale and with different climate.New hydrological insights for the regionFor precipitation in both regions, model accuracy relative to observations has remained the same or decreased in successive climate model generations until and including CMIP5. No single climate model out-performs other models across all key freshwater variables in any of the investigated basins. Scale effects are not evident from global model application directly to freshwater assessment for the two basins of widely different size. Overall, model results are less accurate and more uncertain for freshwater fluxes than for temperature, and particularly so for model-implied water storage changes. Also, the monsoon-driven runoff seasonality in MRB is not accurately reproduced. Model projections agree on evapotranspiration increase in both regions until the climatic period 2070–2099. This increase is fed by precipitation increase in MRB and by runoff water (thereby decreasing runoff) in the Aral Region.