Date of Award

5-10-2026

Date Published

June 2026

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Earth & Environmental Sciences

Advisor(s)

Samuel Tuttle

Keywords

canopy structure;cosmic ray neutron sensor;machine learning;seasonal snow;snow hydrology;snow water equivalent

Abstract

Recent warming in the last few decades has caused widespread changes in the global climate, especially in areas that experience snow events or contain permafrost environments. In the northern United States, our winter seasons are decreasing in length, the overall snowpack depth is decreasing, and the availability of springtime snowmelt is less. This is leading to a shift of a more rain dominated climate especially in the northeastern United States, impacting forest health, water resource availability, and winter tourism. Thus, studying how these seasonal snowpacks are changing soon is extremely important for climate and human health. This dissertation examines a wide range of ground-based snow hydrology methods to examine how a shift towards a more rain-dominated winter will impact local climates. For my first chapter, I determined the effectiveness of a novel instrument called a cosmic ray neutron sensor (CRNS) for snow studies, specifically for estimation of snow water equivalent (SWE) in a prairie environment in central Montana. I evaluated these SWE estimates using Unoccupied Aerial Vehicle (UAV)-based lidar snow depth maps, hourly camera images, and snow pit density data. Despite a highly variable snowpack that experienced rapid changes in snow depth, our results suggest that the CRNS is effective at integrating over significant spatial variability within its 171 m footprint at this Montana site. However, the spatial distribution of snow exerts a strong influence on the CRNS signal and must be considered when interpreting CRNS estimates. I shifted my area of focus for my second chapter to the northeast where we have a transitional, montane forest snowpack compared to the highly wind driven environment in my first chapter. For this work, I used a spatially gridded series of vertically profiles of iButton temperature sensors within a forested watershed, specifically Arbutus Lake in the central Adirondack Mountains, to estimate snow depth across this watershed. I used these iButton sensors vertically spaced every 20 cm and combined them with machine learning methods (i.e., random forest regression) for a multi-season analysis of snow conditions in northern New York. Random forest models using the snow temperature data showed RMSEs between 1.8 and 6.5 cm against our verification dataset of daily timelapse camera imagery. Supplementary analyses showed that our machine learning models could maintain that accuracy using only snow temperature data and that other variables like forest cover, elevation, aspect, and usually canopy gap fraction have minimal effect on model accuracy. My third and final chapter focuses on this same watershed first mentioned in Chapter 2 but this time investigates canopy influence on snowpack distribution especially during periods of accumulation and melt in the northeastern United States. To assess impacts of vegetation structure on snow cover, I utilize biweekly measurements of snow depth (SD) and SWE through manual transects from field visits, hourly temperature profiles, hourly timelapse imagery, hemispherical camera imagery, weather station data, and under canopy lidar backpack surveys for three distinct different canopy types (i.e., coniferous, deciduous, and open). Through three winter seasons of data, we see distinct differences amongst our three canopy types despite close proximity within this forest to each other showing the strong influence canopy structure has on this forested snowpack around Arbutus Lake changes through time.

Access

Open Access

Download

Share

COinS