Development and application of geostatistical methods to modeling spatial variation in snowpack properties, Front Range, Colorado

Measurements of environmental parameters are often spatially-correlated, but traditional analysis techniques, such as linear regression, generally do not account for this correlation. This dissertation describes a methodology for building spatially-distributed models of environmental variables, which incorporate the spatial correlation of model errors using geostatistical techniques. Accounting for the effect of spatial correlation of both environmental variables and model errors allows for a more objective evaluation of the information contained in the available data and yields more realistic results. The developed methodology is used to model snow properties at three different scales.; The first application of this methodology is to modeling the spatial distribution of snow depth at the basin scale based on survey data collected over seven years in the Green Lakes Valley near Niwot Ridge, Colorado. Topographic parameters in this windswept environment are shown to be significant predictors of the snow depth distribution. Model parameters describing the spatial distribution of snow are found to be correlated to SNOTEL measurements, allowing for realistic predictions of the snow depth distribution in the basin for years without intensive sampling.; The second application is the characterization of basal meltwater discharge at the field scale based on lysimeter discharge measurements collected over 6 years by the Soddie Lysimeter Array at Niwot Ridge. Individual measurements of discharge are shown to be highly variable in space, necessitating a large array of lysimeters to accurately estimate the mean snowmelt. Daily discharge is found to increase with decreasing snow depth, highlighting the importance of stratigraphic layers in the routing of meltwater through a snowpack.; The third application is the characterization of the occurrence of meltwater pathways in a cubic meter of the snowpack. High-resolution (1 cm3) 3-dimensional datasets of meltwater occurrence were obtained using the snow guillotine, a new instrument for precisely sampling a dyed snowpack. Meltwater pathways were found to be strongly influenced by snowpack stratigraphy, and became less prominent as the melt season progressed.; The spatially-distributed models developed in this dissertation provide a rich description of the variability found in nature, and are an improvement over previous efforts of modeling spatial variability of snow parameters. Applications of this modeling methodology are not limited to snow properties, nor to the spatial scales of the three applications, but can be used to model many spatially-correlated environmental variables, such as precipitation, soil moisture, and contamination.