Events & Media

The Bren School of Environmental Science & Management
at the University of California, Santa Barbara



"Reconstruction of heterogeneous snow water equivalent from MODIS imagery
and energy balance modeling"

Annelen Kahl
PhD Candidate
Bren School of Environmental Science & Management

Friday, June 21, 2013
11 a.m.
Bren Hall Dean's Conference Room

Jeff Dozier, faculty advisor;
Tom Dunne, John Melack, Naomi Tague, Danny Marks, committee members

This study explores two different modeling approaches to quantify a basin’s snow pack: 1. Forward modeling with the snow energy balance model Isnobal and 2. Reconstruction based on potential melt from Isnobal combined with fractional snow covered area from MODIS satellite imagery. Basin snow-water equivalent (SWE) and its heterogeneity are modeled for 4 water years in the Marble Fork of the Kaweah River, a 152-square-kilometer basin in the Sierra Nevada. The work focuses on the two components of heterogeneity of SWE. Heterogeneity due to accumulation is represented by reconstructed SWE before the onset of melt. It is highest above the timberline, where wind causes redistribution and sublimation. Heterogeneity due to melt emerges from Isnobal output. Throughout spring it stays low above the timberline, while an increase occurs at the lowest snow-covered elevations as well as at the transition between forested and open areas. The spatial distribution of these trends persists in all four years of the study, but their occurrence in time shifts. Cluster analysis is used to related melt heterogeneity to basin characteristics such as topography and vegetation cover at different times of the melt season. Vegetation fraction is the dominating factor, with intermediate canopy cover causing the highest heterogeneity in melt.

As input to the reconstruction model, fractional snow covered area from MODIS (fSCA) is corrected for missing data, periodic fluctuations from the acquisition method, cloud and vegetation cover. These uncertainties propagate into the reconstruction and their effects on modeled SWE are treated in a sensitivity analysis. Two potential uncertainties are analyzed: a bias in fSCA that scales with vegetation cover and an offset in final mel- out date.

Finally I present an analysis of the spatial distribution of precipitation in the densely monitored Reynolds Creek Experimental watershed in Southwest Idaho. Elevation detrended kriging yields interpolation surfaces similar to PRISM and is able to capture local anomalies such as rain shadow and topographic exposure, which cannot be detected with simple elevation gradients. A comparison between summed hourly and daily precipitation surfaces from kriging illustrate the importance of interpolating precipitation at high temporal resolution. Errors in the partitioning between rain and snow of over 100% were found in daily interpolation results, when the elevation of phase transition varied frequently.