Sensitivity and model reduction of simulated snow processes: contrasting observational and parameter uncertainty to improve prediction

Abstract

The hydrology of high-elevation, mountainous regions is poorly represented in Earth Systems Models (ESMs). In addition to regulating downstream water delivery, these ecosystems play an important role in the storage and land-atmosphere exchange of water. Water balances are sensitive to the amount of water stored in the snowpack (snow water equivalent, SWE), as much of Colorado’s water supply is derived from snowmelt. In an effort to resolve this hydrologic gap in ESMs, this study seeks to better understand how uncertainty in both model parameters and forcing affect simulated snow processes. To better understand parameter uncertainty and asses model performance, this study conducts a sensitivity analysis, using active subspaces, on model inputs (meteorological forcing and static parameters) for both evergreen needleleaf and bare ground land cover types. Observations from an AmeriFlux tower at the Niwot Ridge research site are used to force an integrated single-column hydrologic model, ParFlow-CLM. This study found that trees can mute the effects of sublimation causing the evergreen needleleaf model to be sensitive primarily to hydrologic forcing; humidity in the winter, radiation and air temperature in the summer months. However, bare ground simulations were most sensitive to snow parameters along with radiation as these are unblocked by canopy. The bare ground model is most sensitive to overall changes to the linear combination of input parameters, which means radiation observations and snow parameterizations are of great importance for obtaining accurate hydrologic model results. Humidity measurements are also important, but the change in SWE of the evergreen needleleaf simulations was less than that of the bare ground simulations.