A Comparison of Spatial Statistical Techniques for the Development of a Validation Data set for Mesoscale Modeling of Snow Water Equivalence
Abstract
Mesoscale estimates (>1km2) of distributed snow water equivalence (SWE) are seldom validated with ground truth data beyond observations made by the Natural Resource Conservation Service (NRCS) Snow Telemetry (SNOTEL) network. SNOTEL data may be insufficient to validate pixel values of SWE at the low spatial resolutions (>1km) used in mesoscale modeling. In order to validate mesoscale estimates of SWE at 1km resolution, we sampled snow depth and snow density at 250m intervals over 4km2 areas at three different sites in the San Juan Mountains of Colorado during water year 2001. We also conducted our analysis on data collected during water year 1997 in the 19.1 km2, Tokopah Basin of the Sierra Nevada, California. We compared the robustness of the following four statistical techniques: kriging, co-kriging, inverse distance weighting, and neural networks. We found that average SWE over the Tokopah basin during the April, May and June snow surveys was 995mm, 701mm and 283mm respectively and SWE at the Colorado sites during water year 2001 decreased from 269mm to 104mm, 452mm to 261mm, and 845mm to 447mm to 219mm for the three sites respectively.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2001
- Bibcode:
- 2001AGUFMIP51A0742M
- Keywords:
-
- 1640 Remote sensing;
- 1827 Glaciology (1863);
- 1860 Runoff and streamflow;
- 1884 Water supply