Modeling snow water equivalent in complex mountainous terrain

Abstract

The water stored in seasonal mountain snowpacks is a vital resource that approximately 20% of the world's population relies on for freshwater availability. However, accurately quantifying the amount of water stored in a snowpack, known as snow water equivalent (SWE), is difficult. The longest employed technique to quantify SWE is manual measurements. However, manual measurements of SWE are time intensive. As a result, researchers can collect relatively few point-based measurements across spatially extensive and complex regions. Automated weather stations may provide additional measurements of SWE and meteorological conditions but are expensive and difficult to maintain. Thus, reliable measurements of snow characteristics like SWE are scarce across time and space. A lack of extensive measurements causes data from few points to be extrapolated across spatially heterogeneous environments which increases uncertainty in estimates of water availability. Recent advances in satellite remote sensing allow researchers to observe snowpack dynamics across spatially continuous scales instead of relying solely on point-based measurements. However, current satellite technologies are incapable of collecting high- resolution snow data at the hillslope scale. Previous work has shown the importance of high elevation, hillslope-scale water storage reservoirs. Uncrewed aerial vehicles (UAVs) address the limitations of satellite remote sensing on the hillslope scale and are used to create high accuracy (<5 cm) models of snow depth. However, these models of snow depth provide no information on the amount of water stored without a value for snow bulk density. Thus, to capture hillslope dynamics of SWE, researchers must pair high-resolution models of snow depth with either directly measured or modeled bulk density of snow. This master's thesis integrates UAV-derived measurements of snow depth with modeled snow bulk density values to create continuous representations of hillslope-scale SWE across 9 flight dates. We found that each density modeling approach consistently underestimated SWE for the field site for each flight date except one. Further, each method of modeling snow bulk density was statistically indiscernible from each other. These findings highlight the heterogeneity of snow in mountainous terrain. In future work, bulk density models can be further parameterized to better represent site-specific values of SWE.