C43C-0399:
Comparison of Snow Depth Measurements and Snow Patterns at Scales Ranging from 1 km2 to 210 km2

Thursday, 18 December 2014
Anna M Wagner1, Matthew Sturm2 and Christopher A Hiemstra1, (1)Cold Regions Research and Engineering Laboratory Alaska, Fort Wainwright, AK, United States, (2)University of Alaska Fairbanks, Fairbanks, AK, United States
Abstract:
In this study we seek to identify how snow depth sampling strategies can be used to minimize time and effort when measuring Arctic tundra snow depth distributions, which are primarily controlled by wind and topography. We investigate this issue using data from an oversampled Arctic Alaska watershed. Specifically, we ask at what measurement spacing do we lose our ability to resolve critical scale-related snow spatial variations? Our multi-year measurements come from three nested study areas of 1 km2, 6 km2, and 209 km2. The 1 km2 area was sampled on a static 100-m spaced grid yielding between 6,400 to 15,000 depth measurements each time it was surveyed. The 6 km2 area was measured using arbitrarily scattered 100-300 m long snow depth transects, producing 6,900 to 10,900 snow depths each year. For the largest area, thirty-five to fifty 200 m-long transect lines were randomly distributed throughout the domain yielding 3,000 to 13,400 total measurements annually. To compare results from dense vs. sparse surveys, we extracted and averaged snow depth measurements from dense surveys at the same locations used in the sparse surveys. We found that despite a reduction in number of data points by a factor of 26 to 61, the overall snow depth pattern continued to be captured. In addition, mean values of depth showed high fidelity despite the large reduction in sample number. For example, when subsampling the 6 km2 grid to represent the 1 km2 grid, there was only a 3 cm difference in the average value. However, when comparing snow depth averages of either of the smaller areas with the largest area (209 km2), we conclude that the larger domain is undersampled despite the thousands of data points collected there. During three years of measurements the 209 km2 snow depth average was between 13 and 26% less than the average for the 1 km2 grid. Remarkably, in 2013 the average was only 4% less than the 1 km2 grid, demonstrating that in some years sparse sampling can capture relatively accurate snow depths. Based on these comparisons, we speculate on what sort of sampling pattern might be used at the 209 km2 scale that is both parsimonious and cost efficient.