Abstract
Understanding future precipitation trends is vital for stakeholders and environmental managers in water-limited regions such as the southwestern United States. Global climate models from the Coupled Model Intercomparison Project phase 6, however, project 21(st) century precipitation trends of varying sign in this region across seasons, even in models forced by an identical emissions scenario, reducing stakeholders and environmental managers ability to accurately plan for future water resources in the region. As global climate models generally have resolutions too coarse for local and regional environmental management, one broadly used method is the statistical downscaling of coarse resolution climate models onto higher resolution grids, using historical relationships between large- and small-scale conditions. Previous work, however, has shown that future precipitation projections in non-downscaled models in the southwestern United States are associated with the models' representation of the present-day mid-tropospheric ridge over western North America. Here we show that these relationships are largely maintained across seasons for models that have undergone univariate statistical downscaling following the localized constructed analogs technique, suggesting that the bias correction schemes used in univariate statistical downscaling alone do not correct for these associations. This work thus provides a cautionary tale to using statistically downscaled climate model data without also considering biases that are dynamically driven by the GCM's mean-state atmospheric circulation.