Premium
PREDICTING WILDERNESS SNOW WATER EQUIVALENT WITH NONWILDERNESS SNOW SENSORS 1
Author(s) -
McGurk Bruce J.,
Edens Thaddeus J.,
Azuma David L.
Publication year - 1993
Publication title -
jawra journal of the american water resources association
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.957
H-Index - 105
eISSN - 1752-1688
pISSN - 1093-474X
DOI - 10.1111/j.1752-1688.1993.tb01506.x
Subject(s) - snow , elevation (ballistics) , water equivalent , environmental science , wilderness area , precipitation , hydrology (agriculture) , smoothing , selection (genetic algorithm) , linear regression , meteorology , wilderness , statistics , computer science , geography , geology , engineering , ecology , mathematics , machine learning , structural engineering , geotechnical engineering , biology
Ten pairs of snow sensors were analyzed to investigate the feasibility of predicting snow water equivalent at high‐elevation, telemetered snow sensor sites from lower‐elevation sensors. The need for this analysis stems from an agreement between the California Department of Water Resources and the USDA Forest Service to temporarily allow snow sensors in California's wilderness areas so that a predictive relationship can be developed. After 10 or 15 years, the agreement calls for the sensors to be removed. Initial efforts to a priori select sensor pairs were based on proximity, colocation within a basin, and annual precipitation amount, but regression yielded poor fits (R 2 < 0.65) and high standard errors in eight of the ten cases. Analysis of the results suggested that eleva‐tional similarity was the most important selection criteria, and that all available sensors near the target site should be analyzed via a regression screening. Using elevation for selection and the regression screening, five sensors that initially had poor fits were reanalyzed. Each of the five sensors was paired with between two and five new sensors, and R 2 values improved between 27 and 46 percent. Various data smoothing and editing algorithms were evaluated, but they rarely resulted in improved fits.