Evaluating climate model simulations of drought for the northwestern United States

Projected changes in regional characteristics of drought throughout the 21st century can help inform climate change assessments across drought-sensitive sectors. However, the ability of global climate models (GCMs) to reproduce statistical attributes of observed drought at regional scales has not typically been considered. We present a framework for evaluating the fidelity of GCMs to simulate regional drought to help guide model selection for users of regional climate projections where drought plays a key climate impact, and elucidate shortcomings of model simulations of hydroclimate variability. Monthly precipitation (Pr) and potential evapotranspiration (ETo) from 24 GCMs of the Coupled Model Intercomparison Project Phase 5 were compared to observations, with a focus on the northwestern United States (NWUS), using a suite of statistics for capturing multi-scalar aspects of drought and both supply (Pr) and atmospheric demand (ETo). Model fidelity was generally higher for drought metrics evaluated with ETo versus Pr, and for seasonal and annual time scales versus multi-year time scales. While some GCMs performed well across the suite of drought metrics and time scales, models systematically underestimated the severity of regional drought in the NWUS defined by precipitation deficits, particularly for multi-year time scales.