Think Globally, Model Locally: Complex Responses of Agricultural Water Supplies to Different Climate Projections

ABSTRACT Many resource management plans use ensembles of global climate models (GCMs) to represent a range of potential future climates. Hydrologic models are used to translate these climates into projections of water resources to evaluate their long‐term vulnerability. GCM ensembles are typically selected for water resource vulnerability assessments based on their collective ability to represent the full range of projected regional changes in precipitation and temperature. However, this approach may miss potential hydrologic shifts due to the heterogenous and nonlinear impacts of climate on hydrology in snow‐dominated mountain headwaters. We demonstrate this challenge in the Walker River Basin (WRB) of California and Nevada. After comparing 32 climate projections and their associated hydrologic model (VIC) outputs we run subsets of projections through a water allocation model (MODSIM) to predict impacts on agricultural water use. Projected end‐of‐century changes in WRB precipitation vary (−20% to +40%) leading to an even greater range in projected streamflow changes (−50% to +75%). Model outputs suggest that maintaining historical levels of water supply reliability through 2100 would require > 50% greater mean annual streamflow. Our study shows the value of selecting GCM ensembles based on locally‐derived hydrologic (rather than purely climatic) metrics. Other GCM selections may not span all potential impacts of changing climate on agricultural water resources.