Practical snow depth sampling around six snow telemetry (SNOTEL) stations in Colorado and Wyoming, United States
Date
2012
Authors
Kashipazha, Amir, author
Fassnacht, Steven, advisor
Kampf, Stephanie, committee member
Laituri, Melinda, committee member
Arabi, Mazdak, committee member
Journal Title
Journal ISSN
Volume Title
Abstract
Across the Western United States, the Natural Resources Conservation Service (NRCS) operates about 700 automated snowpack telemetry (SNOTEL) measurement stations. These stations measure snow depth (SD), snow water equivalent (SWE), air temperature and precipitation. To assess how representative the stations are of the surrounding 1 km2 area, a set of approximately 200 snow depth measurement were taken using ten 1000-m transects sampled at 50-m intervals. This sampling was undertaken at the Dry Lake, Joe Wright, Lizard Head, Niwot, (in Colorado) South Brush Creek, and Togwotee Pass (in Wyoming) SNOTEL stations during the winters of 2008, 2009, and 2010. Various sampling patterns were employed at each sampling point, such as three depth measurements in a row parallel or perpendicular to a transect, and five in a row or five in a plus pattern. We used these patterns and various sub-sets of the 1 km2 surrounding area to assess suitable and practical sampling strategies, to determine the minimum number of transects need for measuring the average SD of each station, to evaluate if each station represent the SD average of its 1 km2 area surrounding, and to investigate inter- and intra-annual variations of SD for each station. Statistical analysis used the least-significant-based analysis of variance with a 95 percent confidence level. Statistical analyses showed snow depth averages of incorporated sampling methods were not significantly difference at the 95 percent confidence level. Therefore, any sampling method could be used for SD measurement based on sampling constraints. We recommend measuring three to five snow depths at each sampling spot and the distance between sampling spots should be less than 200m. The minimum number of transects needed for each station was not the same and it depended upon the physiographic and vegetation heterogeneity of the area surrounding a station. Snow depth varied within a 1 km2 area surrounding of SNOTEL station and we did not find two sampling methods that had the same average SD. However, this did not mean that the average SD using a variety of sampling methods was significantly different at the 95 percent confidence level. A heterogeneous snowpack is caused variations in precipitation, wind patterns, solar radiation, etc. Physiographic and vegetation characteristics can be used as surrogates for these meteorological factors that vary at the small and large scale. The effect of these factors on snowpack heterogeneity is more likely greater when the distance of sampling spots is more than 1 km. The correlation between snowpack heterogeneity and the surrogate characteristics varied in spatially and temporally, and from location to location. The Dry Lake, Joe Wright, Lizard Head, and Niwot SNOTEL stations represented the SD average of their 1 km2 area surrounding while Lizard Head station represented the SD average of its 0.36 km2 area surrounding, all at the 95 percent confidence level. However, the Togwotee Pass and South Brush Creek stations did not represented the SD average of their surrounding area. Whether a SNOTEL station does or does not represent the SD average of its surrounding area is related to the complexity of the terrain. For example, the area surrounding the Joe Wright station has complex terrain but represented the station SD while the South Brush Creek terrain was more homogeneous and did not represent station SD. The performance of the SD sensor at the SNOTEL station can be affected by the interaction of meteorology, physiography, vegetation, and possibly human influences, that can produce an highly varying snow pack under and/or around a SD sensor and led to a lack of sensor representivity or sensor error. Due to potential SD sensor and sampling errors a reasonable amount of error for snow samples, such as 5-10% should be considered.