Development and Evaluation of a Physics-Based Snow Model to Assess Spatiotemporal Snowpack Variability in Vermont

The montane landscape of the northeastern United States exists in a temperate, humid climate in which spatial and temporal snow variability is highly sensitive to slight shifts in air temperature. Despite the importance of snow to regional hydrology, snowmelt dynamics remain understudied in this part of the world. Studying the at-risk snowpack of these low-elevation mountains is beneficial to understanding potential effects of climate warming. This study evaluates spatiotemporal snowpack variability under a changing climate using a distributed, physics-based snowpack model. Here, I run SnowModel for a montane setting in Vermont, encompassing locations of long-term monitoring surrounding Mount Mansfield, Vermont’s highest peak (1340 m). The model was calibrated and validated over a series of years using observations from historical snow surveys and a newly established network of snow monitoring stations. During winter precipitation events, air temperatures are often near 0°C and the spatial and temporal distribution of snow depends strongly on precipitation phase partitioning. I tested various parameterizations of the rain-snow threshold, and found that parametrizations that incorporate humidity, such as wet bulb temperature, outperformed those only considering air temperature. Additionally, I tested liquid water percolation and retention methods and found that allowing liquid water to percolate through the snowpack dominated by gravitational forces reduced model error most effectively. Both precipitation phase (rain vs. snow) incident on an extant snowpack as well as liquid water percolation through the snowpack have considerable effects on energy available for melting. I also assessed the effects of changing weather patterns, simulated by perturbations in temperature and precipitation model forcings. Across the relatively small elevational gradient of the model domain, there was considerable variability in the sensitivity snowpack accumulation and melt processes to these simulated climate scenarios. Overall, I found that incremental changes to temperature and precipitation impact seasonal snowpack and runoff patterns, with increased midwinter melt occurring in addition to an earlier and less pronounced spring freshet. Middle elevations (700-900 m) were shown to be the most sensitive to these changes, with over a 70% mean decrease in SWE in the lowest temperature increase scenario. Additionally, these incremental changes can also have impacts on timing and magnitude of individual events. We see this with the emergence of new high-runoff snowmelt pulses during events occurring at transitional temperatures. In one midwinter event, a +3.3°C increase in temperature resulted in a spike of 40 mm of surface water input from the snowpack at mid-elevations, during a period where snowmelt was minimal in baseline scenarios. An improved understanding of the complexity of the regional to local snowpack heterogeneity and snowmelt dynamics is imperative for improved flood forecasting, hydrologic management, winter recreation, and ecosystem health under a changing climate. These results provide a baseline for evaluating snowmelt dynamics and associated extreme runoff events in the unique topography and climate of the northeastern US.