Disagreement among statistically downscaled data sets increases uncertainty in high-elevation hydroclimate projections

Several statistically downscaled data sets are available using general circulation model (GCM) output from the Climate Model Intercomparison Program Phase 5 (CMIP5). The downscaling methods, bias correction products, and climatology periods vary among these data sets. Our study assesses monthly temperature and precipitation data derived by six statistically downscaled data sets for 14 GCMs. We use a simple monthly water balance model to quantify and decompose uncertainties associated with the GCMs and statistically downscaled data sets in projections for four snow-dominated regions in the western United States (the Greater Yellowstone Area, Upper Colorado Basin, Sierra Nevada, and Columbia River Basin). The six downscaled data sets exhibit disparate high-resolution representations of the magnitude and spatial patterns of future temperature (by up to 2.2 °C) and precipitation (by up to 30%) for a common GCM. Over the Upper Colorado Basin, two data sets derived by the same downscaling method (Multivariate Adaptive Constructed Analogs) produce median losses of snow water equivalent of 51% and 81%. The differences among the downscaled projections are related to the gridded observations used to bias correct the historical GCM output: (1) whether a fixed atmospheric lapse rate (-6.5 °C km^-1) or a spatially and temporally varying lapse rate is used to extrapolate lower elevation observations to high elevations and (2) whether high-elevation station data (e.g., SNOTEL) are included in the observations. The spread among the GCM projections is the largest source of uncertainty in the monthly water balance model simulations; however, in some regions ~25% of uncertainty in the snow projections at the end of the century is attributed to the choice of statistically downscaled data set.