Modeling Elevation-Dependent Snow Sensitivity to Climate Warming in the Data Sparse Eastern Oregon Cascades

Graduation date: 2015In the mountains of the Western US, shifts in the timing and magnitude of snow water equivalent (SWE) over the past century are well documented and attributed to climate warming, but the magnitude of sensitivity depends on elevation. We modeled the spatial distribution of SWE and its sensitivity to climate warming in the 1500 km² Upper Deschutes River Basin, Oregon, with a spatially distributed snowpack energy balance model forced by a gridded meteorological dataset. The forcing data, gridded at a spatial scale of 1/16°, were downscaled to a 100 m spatial-scale digital elevation model using two sets of temperature lapse rates, with and without bias-correction applied prior to downscaling. The bias-correction method adjusted the spatial patterns of temperature and precipitation in the 1/16° gridded data to match 30 arcsecond Parameter Regressions on Independent Slopes Model (PRISM) climate data. During production, the 1/16° temperature data was adjusted for the effect of elevation using a spatially uniform and temporally constant 6.5°C km⁻¹ lapse rate, whereas PRISM adjusts temperature for the effect of elevation using spatially and temporally variable lapse rates that are computed directly from regional weather station data. Thus, bias-correction implicitly adjusted the lapse rates in the 1/16° gridded data to match measured lapse rates. To test the effect of this implicit adjustment of the input data lapse rates vs. adjusting the lapse rates during downscaling, the 30 arcsecond bias-corrected data and 1/16o original data were each downscaled with 1) a spatially uniform and temporally constant 6.5°C km⁻¹ lapse rate, and 2) with monthly varying lapse rates computed from PRISM. Precipitation was downscaled with the same method for each case. This procedure produced four sets of 100 m spatial scale data used as input to the snow model. Model parameters that control empirical estimates of incoming irradiance and the partitioning of precipitation into rain and snow were estimated independently with each dataset to optimize the agreement between modeled and observed SWE. We then modeled the sensitivity (percent change) of basin SWE in response to +2°C and +4°C warming with each of the four downscaled datasets and their respective optimized parameters. Pre-calibration, modeled SWE for the historical climate period differed depending on bias correction and choice of downscaling lapse rates. Post-calibration, modeled SWE for the historical climate period did not differ depending on choice of lapse rates but substantial differences emerged between modeled SWE with the original and bias-corrected forcing data. Inter-forcing dataset differences in modeled SWE during the historical period were largely controlled by differences in estimates of longwave irradiance and temperature between datasets. For the warming scenarios, the SWE sensitivity differed significantly at all elevations between the bias-corrected and original data, but (as in the post-calibration historical period) did not depend on choice of lapse rates. At low to mid elevations, climate change impacts on snow were largely controlled by temperature-driven shifts from snowfall to rainfall, while at high elevations, precipitation variability controlled SWE sensitivity. With just a 2°C increase in temperature, peak snow accumulation occurred 20-30 days earlier and was 20-60% smaller, the length of the snow covered season decreased up to 50 days, and winter rainfall increased by 20-60%. With a 4°C increase, the shifts in timing were roughly doubled and the declines in snow and snowfall increased up to 80%. A 10% increase in precipitation had a negligible impact on basin-integrated declines, indicating that future precipitation variability has little chance of offsetting regional climate warming impacts on snow in the Oregon Cascades. These results highlight the challenges of modeling SWE in data sparse regions, the importance of bias correcting gridded meteorological forcing datasets for hydrologic modeling applications in regions of complex topography, and the strong temperature dependence of snow in the Oregon Cascades