Evaluating hydrologic realism across scales in mountainous regions: a Budyko approach

The water cycle in mountainous areas in the middle latitudes such as the western United States is particularly difficult for climate models to simulate accurately because of the diversity of hydrologic regimes ranging from typically wet, snowy peaks to arid lowlands. Climate models used for global climate projections were thought to be too coarse in their resolution to accurately simulate the runoff in these areas, necessitating the use of off-line, high resolution (< 10km gridscale) hydrology models forced by downscaled, bias-corrected climate model output to create usable projections.

As global models have increased their resolution, with grid cells smaller than 50 km on a side, we find evidence that the models are capable of realistically simulating the annual runoff, particularly if run in AMIP configuration with specified boundary conditions for SST, sea ice, and atmospheric composition. The advantages of using the fully consistent coupled land-atmosphere system inherent in these models is great, provided the simulations can be judged "realistic." However, evaluating the realism of the hydrologic simulation in models at these intermediate scales between 100 km and 10 km requires new metrics that take into account the diversity of surface hydrology at the coarse-grained model scales.

We develop a new way of evaluating these models, the "Budyko portrait", based on simple hydrologic theory that compares aridity and runoff at the grid cell level. We take advantage of a unique set of global atmosphere model runs using the NCAR CAM5 model in AMIP mode at three different resolutions, 100km, 50km, and 25km, including both historic and "counterfactual, no-climate-change" ensembles. The ability of the models' to capture a realistic likeness of the observed hydrology as resolution increases is revealed by the "Budyko portrait" and other metrics for the Colorado River basin. We also investigate the scale dependence of the relationship between declining annual runoff and increasing temperature. Time permitting, we will show the application of these metrics to several other global models' historical simulations, including available AMIP runs from HighResMIP, for the Colorado River Basin.