The Effect of Global Warming on Future Water Availability: CMIP5 Synthesis

Changes in the spatiotemporal dynamics of the global water cycle will constitute some of the greatest challenges to socioeconomic‐environmental well‐being in a warming world. Large multimodel, multiscenario intercomparisons such as the Coupled Model Intercomparison Project Phase 5 (CMIP5) experiment support our best estimates of projected climate change and associated uncertainty thereof. It is important to continually reevaluate how this information is synthesized and communicated and at what point it becomes actionable. In this study, we demonstrate a systematic and holistic framework for synthesizing multimodel ensemble projections of water availability at large river basin scale—the scale at which water resources are both managed and monitored. We identify statistically significant shifts in mean water availability at annual and monthly scales, its interannual variations, and its relative seasonality, as computed from CMIP5 historical (1976–2005) and Representative Concentration Pathway 8.5 (2070–2099) scenario multimodel ensemble output. Water availability is addressed separately through the lens of meteorologists (precipitation), hydrologists (runoff), and agriculturalists (precipitation minus evapotranspiration). We illustrate limitations in CMIP5 model representativeness through comparisons of atmosphere‐only model (Atmospheric Model Intercomparison Project) output against observational best estimates. And we find that warming‐induced shifts in water availability projected by CMIP5 carbon‐cycling Earth system models are comparatively less substantial than those projected by traditional general circulation models. As we show, knowing the seasonality of both projected changes and of the biased model background climatology onto which they are imposed is paramount to ensuring proper interpretation and ascribing confidence. Key Points CMIP5 projections of significant subseasonal changes in water availability are often overlooked and underestimated Carbon‐cycling Earth system models project 23% fewer significant shifts in future water availability than general circulation models The magnitude of model errors, as quantified through observational comparison, may preclude the credibility of projections for some basins