Date of Award

Spring 1-1-2016

Document Type


Degree Name

Master of Science (MS)


Civil, Environmental & Architectural Engineering

First Advisor

Noah P. Molotch

Second Advisor

Balaji Rajagopalan

Third Advisor

Ben Livneh


Chapter I Abstract

Previous work demonstrates conflicting evidence regarding the influence of snowmelt timing on forest net ecosystem exchange (NEE). Based on 15 years of eddy-covariance measurements in Colorado, years with earlier snowmelt exhibited less net carbon uptake during the snow-ablation period, which is a period of high potential for productivity. Earlier snowmelt aligned with colder periods of the seasonal air temperature cycle relative to later snowmelt. We found that the colder ablation-period air temperatures during these early snowmelt years lead to reduced rates of daily NEE. Hence, earlier snowmelt associated with climate warming, counter-intuitively, leads to colder atmospheric temperatures during the snow-ablation period and concomitantly reduced rates of net carbon uptake. Using a multilinear-regression (R2=0.79, P<0.001) relating snow-ablation period mean air temperature and peak snow water equivalent (SWE) to ablation-period NEE, we predict that earlier snowmelt and decreased SWE may cause a 45% reduction in mid-century ablation-period net carbon uptake.

Chapter II Abstract

Dai [2008] used a 29-year observational precipitation phase dataset to produce global conditional snow frequency curves (frequency of snow events per air temperature bin) for the land and ocean. We extended upon Dai’s study to further explore the influence of three physically relevant variables (surface air temperature (Ts), relative humidity (RH), surface pressure (Ps)) on conditional snow frequency over the land surface. We found that precipitation events that fell at low ambient RH and/or low Ps had greater snow frequencies at high Ts compared with events that fell at high ambient RH and/or high Ps, respectively. However, the range in snow frequencies per Ps class is less than the range per RH class. We developed and compared three binary logistic regression models using Ts, RH and Ps as predictor variables for precipitation phase. The Ts-RH model performed universally better than the simple Ts model and the Ts-RH-Ps model had nearly identical success rates to the Ts-RH model. The largest difference in success rates between the Ts-RH model and the simple Ts model occurred at the lower RH classes, and all models performed universally better in the higher RH classes. Therefore, while our results demonstrate that RH should be included in precipitation phase predictive models whenever possible, we might expect RH to more significantly contribute to models utilized in climatically dry regions. These models were developed and tested with a global dataset of ~15 million precipitation observations and thus present the most the most extensive global phase prediction model to date