Accuracy and realism of CMIP6 candidate models in capturing dry, moist, and extreme precipitation anomalies in the Laurentian Great Lakes.

The Great Lakes are the world’s largest freshwater system, and understanding how Great Lakes precipitation dynamics will be modified by climate change is of critical importance. As the Great Lakes straddles a semi-arid to humid transitional region, trustworthy precipitation predictions must be generated by models that can accurately capture both thermodynamical and dynamical drivers of regional hydroclimatological variability in historical runs. In this study, precipitation simulations from 12 CMIP6 models, representing the full range of variability in numeric representation of North American climate, are evaluated for accuracy in capturing historic seasonal wet, dry, and extreme precipitation anomalies in the Great Lakes region. Based on historical accuracy, a subset of candidate models is selected. Simple statistical methods are used to explore the relationships between cumulative seasonal precipitation anomalies and mid-level circulation anomalies, both in observations and in historical model runs. Historic observations suggest that the Pacific North American pattern and North Atlantic subtropical high are important components of summertime hydroclimatic circulation in the region. The two models which most accurately characterize a range of precipitation anomalies replicate these hydroclimatic circulation patterns with the highest fidelity. These two most accurate and physically plausible models are in consensus in predicting an increasing length of dry spells, and decreasing duration of wet days, suggesting increased drought risk under climate change. They diverge between negative and neutral trends in maximum daily precipitation and total precipitation. Implications for water resources management and statistical downscaling studies are discussed.