The wettest model projections for the southwestern United States are linked to the development of an anomalous mid-tropospheric trough over the eastern North Pacific Ocean, which depends in part on future unpredictable internal variability within individual models. However, projected precipitation trends are also linked to model-to-model differences in the representation of the present-day climatological mid-tropospheric ridge over western North America. Models whose present-day mid-tropospheric flow differs most from observations are responsible for the most extreme outlier precipitation trends over the 21st century, casting doubt on the physical plausibility of future scenarios with large drying during winter and large wetting during summer. Dynamical arguments will be presented to illustrate why biases in the present-day circulation in these models reduce confidence in their future precipitation projections. Finally, we explore the origin of the present-day circulation differences among models, focusing on the role of tropical Pacific sea surface temperature patterns.
Overall, our results suggest that an accurate simulation of the present-day large-scale atmospheric circulation over western North America during all four seasons is a necessary (but likely not a sufficient) condition for a model to produce plausible 21st century precipitation projections for the southwestern United States. This study serves as an illustrative example of how model mean-state circulation biases contribute to high uncertainty in regional precipitation projections and can therefore be used to provide dynamical constraints to understand and reduce these projection uncertainties.

