Scaling Snow Observations from the Point to the Grid Element: Implications for Observation Network Design
|Title||Scaling Snow Observations from the Point to the Grid Element: Implications for Observation Network Design|
|Year of Publication||2005|
|Authors||Molotch, N, Bales, R|
|Journal||Water Resources Research|
|Keywords||1839 Hydrologic scaling, 1847 Modeling, 1854 Precipitation, 1863 Snow and ice, binary regression tree models, observation network design, remote sensing, Rio Grande, scaling, snow water equivalent|
The spatial distribution of snow water equivalent (SWE) within 16-, 4-, and 1-km2 grid elements surrounding six snow telemetry (SNOTEL) stations in the Rio Grande headwaters was characterized using field observations of snowpack properties, satellite data, binary regression tree models, and a spatially distributed net radiation/temperature index snowpack mass balance model. In some cases, SNOTEL SWE values were 200% greater than mean grid element SWE. Analyses designed to identify the optimal location for measuring mean grid element SWE accumulation indicated that only 2.4% of each grid element satisfied the criteria of optimality. Similar analyses for the ablation season showed that point SWE and mean grid element SWE were highly correlated (r = 0.73) in areas with relatively persistent snow cover. These locations did not overlap in space with areas deemed optimal at maximum accumulation; areas with persistent snow cover have relatively high accumulation rates. Therefore future observations may need to be placed with the specific objective of representing either accumulation or ablation season processes. These results have implications for large-scale studies that require ground observations for updating purposes; we show an example of this utility using the SWE product of the National Operational Hydrologic Remote Sensing Center. Furthermore, the relatively consistent spatial patterns of snow accumulation and melt have implications for future observation network design in that results from short-term studies (e.g., 2 years) can be used to design long-term observation networks.