Date of Award

Winter 2025

Project Type

Thesis

Program or Major

Civil and Environmental Engineering

Degree Name

Master of Science

First Advisor

Jennifer M Jacobs

Second Advisor

Julie Paprocki

Third Advisor

Elizabeth Burakowski

Abstract

Estimates of snow water equivalent (SWE) are crucial for water availability and flood risk assessment due to the contribution of snow melt to increased surface runoff. However, many regions do not have observations of SWE, thus necessitating a robust method to estimate SWE for streamflow forecasting. Numerical snowpack models provide a potential way to estimate SWE in areas without observations. However, there are a multitude of snowpack models available, and they range greatly in their complexity. Even so, there have been few snowpack model intercomparisons to assess the trade-offs in model performance and complexity, and those that do exist did not focus on the northeastern United States. Many snowpack models have been developed and calibrated in specific site conditions; thus, it is further unknown how these models will perform in the northeastern United States. This research tests a simple temperature-index model, the moderately complex SNOW-17 model, and the complex SnowModel at the Hubbard Brook Experimental Forest, Woodstock, New Hampshire. The three models were run with and without assimilation of SWE measurements through direct insertion, as well as at point- and spatially distributed spatial complexities. The models were compared to snow course and spatially distributed Uncrewed Aerial Systems (UAS) SWE observations to derive performance metrics used to intercompare model performance. All three models performed poorly without insertion of observations. However, the temperature-index model had the lowest accumulation of SWE of the three models and thus generally performed the best. Insertion of SWE observations and running the models spatially distributed improved model performance; RMSEs were reduced from upwards of 10 cm to less than 5 cm when either inserting observations or running spatially distributed. Differences between the models were apparent during accumulation but were highest during the spring melt period. All three models were able to replicate observed elevational gradients in SWE, however, only SnowModel was able to approach the observed small-scale variability in SWE. This study suggests that observations are necessary for these models to perform well in the northeastern United States, but that use of complex snowpack models may not be necessary given the simplest model generally performed best.

Download

Share

COinS