Synthesizing Measurement, Modeling and Remote Sensing Techniques to Study Spatiotemporal Variability of Seasonal Snow

Mountain snowpacks vary drastically over length scales as small as 1—2 meters in complex terrain and require high resolution measurements to accurately quantify the spatial distribution of snow. This thesis explores this spatial distribution using remote sensing, modeling and ground-based observations. Snow depth estimates from airborne LiDAR at 5 m resolution over 750 km2 was compared to in situ observations and results from physically-based snow and wind redistribution models, and a new low cost method for continuous depth measurements at the slope scale was developed. Repeated airborne Light Detection And Ranging (LiDAR) surveys are capable of recording snow depth distributions at 1—5 meter resolution over very large geographic areas, while additionally providing information about vegetation, slope aspect and terrain roughness. During NASA\u27s second Cold Lands Processes eXperiment (CLPX-II) in the winter of 2006/07, two LiDAR surveys were flown nearly three months apart over a vast 750 km2 swath of the Rocky Mountains near Steamboat Springs, Colorado. Both flights took place well before any significant melt occurred, and the difference of the vegetation-filtered surfaces resulted in an estimate of the change in snow height across the survey area. An intensive manual measurement campaign was conducted to coincide with each LiDAR flight to provide ground truth information for the LiDAR dataset. Using the in situ measurements and the LiDAR-derived snow depth changes, an uncertainty study was performed to investigate errors in snow depth change for this high resolution remote sensing method due to elevation gradients and vegetation types. Secondly, this work leverages the large extent of the CLPX-II LiDAR dataset to validate more than 900 pixels, each at 30 arc-second resolution, of modeled snow depth from the SNOw Data Assimilation System (SNODAS) operational hydrologic model developed by the National Operational Hydrologic Remote Sensing Center (NOHRSC). Upscaling the high resolution LiDAR-derived snow depths to the much lower spatial resolution of the SNODAS estimates produced a statistically robust dataset of over 900 independent pixel comparisons for the first time, due to the difficulty in obtaining independent validation data at the 1 km scale. Results support the notion that sub pixel-scale slope, aspect, vegetation density and terrain rough- ness factors are important to consider for model predictions of snow distribution in mountain regions. To investigate the wind transport factor, a wind redistribution model based on terrain characteristics is implemented for a 1 km2 wind-affected sub region where high resolution snow depths have been supplied from three independent LiDAR flights taken during different winter seasons. The inter-annual consistency of snow depths at the site reveals a close correlation with the terrain parameters produced by the wind model for a known local prevailing wind direction. LiDAR currently remains the highest resolution large extent method for measuring snow depth, even though it is extremely costly to perform frequently and is primarily used only at intensive research sites. To monitor temporal variations of snow depth over more than a point, simple time-lapse photography is a promising and efficient way to obtain information about snowpack evolution at the slope scale. A robust and low power method to measure hourly changes in snow depth was developed that involves only three primary components: (1) an inexpensive, off-the-shelf time-lapse camera, (2) a weatherproof external battery box and (3) an array of secured, brightly painted depth markers. The camera is calibrated at the marker locations and a pixel counting algorithm automatically distinguishes the snow surface at each marker location after the images are captured. Results agreed closely with nearby standard ultrasonic depth sensors