Spatial variability of snow depth measurements at two mountain pass snow telemetry stations
Date
2012
Authors
Blumberg, Evan J., author
Fassnacht, Steven, advisor
Laituri, Melinda, committee member
Butters, Greg, committee member
Journal Title
Journal ISSN
Volume Title
Abstract
Much of the Western United States relies heavily on spring snow melt runoff to meet its industrial, agricultural, and household water needs. Water professionals use the network of snowpack telemetry (SNOTEL) stations to help forecast spring melt water runoff. These stations only represent a small area and across a watershed, the variability in snowpack properties can be large. Properties such as snow depth can vary substantially even over distances as short as a meter. Previous studies have examined how snow depth is distributed across the landscape and how terrain and vegetation parameters can be used as surrogates for the meteorological variables that drive the distribution of snow. The parameters are derived from a digital elevation model (DEM) that is now at a 30x30m resolution, and they include elevation, aspect, slope angle, and canopy cover, as well as clear sky solar radiation and the maximum upwind slope. Typically three to five snow depth measurements are taken to represent each 30x30m DEM pixel. This study examines the distribution of variability in snow depth within a pixel. Snow depth surveys were conducted around the Joe Wright SNOTEL station near Cameron Pass in northern Colorado on May 1st, 2009 and May 1-2, 2010 and around the Togwotee Pass SNOTEL station in north-central Wyoming on March 17th 2009. Surveys were performed by taking snow depth measurements in a 1 x 1 kilometer block around each SNOTEL station. Due to the logistics of sampling these two locations that both have dense forests and steep terrain, three different sampling methods were employed based on a standard of three points in a row spaced 5 meters apart. To examine the variability at a location (pixel), at least eight additional measurements were taken between the three points (11 points were taken on May 1st, 2009 at Joe Wright). At Togwotee Pass, 10 additional depth measurements were taken about the mid-point, perpendicular to the main transect, yielding 21 points. For the 2010 survey at Joe Wright, the 11 points in a row were supplemented by two points at the beginning, middle and end (three standard points) to yield 17 measurements at a location. From these data the parameters most strongly correlated with the average snow depth, the standard deviation of snow depth, and the coefficient of variation were computed. Binary regression trees were used to further explore the relation between the average and variability and the terrain and canopy parameters. The statistics (average and standard deviation) from the standard three points was compared to all the points (11, 17or 21) measured at a location. Data were sub-set from all the points to determine the average difference and subsequently an appropriate number of depth measurements that should be taken to represent a location. Key variables were not consistent for the 2009 and 2010 Joe Wright SNOTEL surveys, and also varied when looking at standard deviation or coefficient of variation. Among many surveys, canopy cover, elevation, and sin of slope were key variables, but to different degrees. Investigation into survey efficiency show that taking between 3 to 6 data points per pre-determined sample point is suitable to be within 5% of the overall average, whether it be the 11, 17, or 21 point survey scheme.