Skip to main content
ARS Home » Pacific West Area » Boise, Idaho » Northwest Watershed Research Center » Research » Publications at this Location » Publication #174535

Title: Stratified random sampling techniques for snow surveys of mountainous basins

item Winstral, Adam
item Marks, Daniel

Submitted to: American Geophysical Union
Publication Type: Abstract Only
Publication Acceptance Date: 9/5/2004
Publication Date: 9/5/2004
Citation: Winstral, A., Marks, D. 2004. Stratified Random Sampling Techniques for Snow Surveys of Mountainous Basins, EOS Transactions of the American Geophysical Union, Vol 85 (47) F446 (CD-ROM abstract)

Interpretive Summary:

Technical Abstract: Intensive snow surveys of mountain basins are the most accurate means of characterizing the heterogeneous mosaic of snow distribution typically present. As such, survey data is the preferred product for initializing and validating spatial snow models. The collection of survey data, even in small basins, is however costly and time-consuming. Snow survey data is typically collected along a standard grid, a spatial randomization based on a grid, or along pre-determined transects. Even in study areas of small to moderate size (< 3 km2) vital decisions on sample spacing and intensity must be made to adequately cover the entire basin while capturing large process-based snow-water-equivalent (SWE) differences that can occur at relatively small spatial scales. In this research, four years of survey data collected on a regularly spaced grid in the Reynolds Mountain East basin (0.36 km2) in southwest Idaho were used to analyze stratified random sampling techniques designed to reduce the number of samples required to accurately portray snow distribution. It was found that a clustering algorithm based on prior survey data could substantially reduce the number of samples required to produce a surface of SWE similar to that produced by the full dataset. Clustering based exclusively on an a priori set of variables derived from the DEM and a distributed snowmelt model also produced satisfactory results.