Submitted to: American Geophysical Union
Publication Type: Abstract Only
Publication Acceptance Date: September 1, 2002
Publication Date: December 1, 2002
Citation: Seyfried, M.S., Winstral, A.H., Morehead, M.D. 2002. Distribution of frozen soil in rugged terrain using synthetic aperture radar.. American Geophysical Union. Technical Abstract: Historically, large flooding events in the interior Pacific Northwest region of the USA have been associated with rain on frozen soil and shallow snow. The extent of frozen soil is therefore a critical determinant of flooding hazard. Point scale models have successfully simulated soil freezing dynamics but have yet to be tested in spatially distributed mode in rugged terrain, where slope, aspect, elevation and snow cover influence frozen soil distribution. This is partly due to a lack of spatially extensive data for model testing. Synthetic aperture radar (SAR) can potentially provide such via the soil dielectric constant, which is low (similar to dry soil) when frozen. In areas of high relief, topographic and surface roughness effects overwhelm those attributed to the soil dielectric constant and must be accounted for. We used a differencing approach to do this, assuming that surface roughness and topography were constant and registration perfect. We acquired 23 RADARSAT images between December of 1999 and September of 2001. Multiple (9) summer (dry) scenes were used to determine the variance of differences independent of soil dielectric variations. We assume that, under winter conditions, small deviations from dry were frozen and large deviations were moist. Ground truth data of frozen soil, snow cover and soil water content collected along four 750 m transects were used to evaluate the approach. There was a strong correlation between the ground measured and remotely sensed depiction of frozen soil along the transects. In some cases within pixel variability precluded accurate identification of soil freezing because the controlling topography varied at scales less than 30 m. Deep snow cover required separate consideration due to the depressed response. At larger scales, the progression of soil thaw and snow melt were consistent with point observations. This data should provide a useful base for testing distributed models of soil freezing.