Development, Analysis and Use of a Distributed Wireless Sensor Network for Quantifying Spatial Trends of Snow Depth and Snow Water Equivalence Around Meteorological Stations With and Without Snow Sensing Equipment
StatisticsView Usage Statistics
Snow accumulated in the mountainous portions of the Western United States is an extremely important source of water supply. Federal resource management agencies (USFS, BLM, USGS, NRCS, Federal water master), non-governmental organizations (stakeholder groups) and private contractors quantify and predict future stream and river flows throughout the Western United States for water use allocated to water supply, recreational waters, and water to sustain aquatic life. These agencies devote substantial resources to estimating snowpack depth and snow water equivalence during the winter and spring months to allocate water through river and irrigation systems. Forecasts are based on measurements collected from a network of meteorological stations, some of which also measure snow water equivalence and snow depth. Such stations are expensive to establish and maintain and provide data from a single point of measurement. Accordingly, forecasts cannot account for spatial variability associated with snowpack depth and water content. Information about spatial variability could be extremely useful in water supply forecasts, especially if measurements from meteorological stations, with or without snow sensing equipment, could be extended by simple, low cost data collection efforts. A prototype wireless sensor network, Snowcloud, was deployed in the Sagehen Creek experimental field station, CA. Snowcloud was developed by the University of Vermont and deployed in fall, 2009 in collaboration with the University of Nevada-Reno, for a single season study in the Sagehen Creek Field Station, near Truckee, CA. The network was set up around a meteorological station to sense and report snow depth and temperature from a distributed set of independent nodes. The Snowcloud deployment within the Sagehen Creek Field Station examined small scale variability in snow depth and SWE arising from temporal changes in state variables, including canopy cover, aspect, temperature, solar radiation, wind speed, wind direction and relative humidity. The Snowcloud nodes were tested to estimate SWE from SD around a fixed based meteorological station acting as a station (1) with snow sensing equipment and (2) without snow sensing equipment. Within the field site, neither a snow pillow that was part of the meteorological station nor manual snow course transects represented the spatial mean SWE of the study area. The snow pillow overestimated SWE values an average of 30% with a maximum overestimation of 40%. The six Snowcloud nodes in conjunction with regression and kriging effectively captured the spatial and temporal variability of SWE with a resolution much greater than the snow pillow. Both estimation methods accurately extrapolated SWE at the node sites and at 48 points in a sampling grid covering the site, with a maximum RMSE of 2.7%. A simple estimate based only on SWE at the meteorological station and SD throughout the site predicted SWE with greater accuracy than regression models that included state variables from the meteorological station and site characteristics. Aspect with canopy closure percentage was an important predictor of SWE at meteorological stations without snow sensing equipment. Significantly decreased correlation between measured and estimated SWE was seen within the models with only the use of a generalized canopy closure qualifier, and the quality of the correspondence were not increased by the addition of a generalized qualifier. With the exception of SD, percent canopy closure to the north (P = 0.000+) was the most important qualifier for predicting SWE around meteorological stations without snow sensing equipment.