HR: 14:25h
AN: G33D-04 INVITED     [Abstracts]
TI: Measuring Probabilistic Dependences at Multiple Scales Between Soil Type and Surface Morphology
AU: * Slatton, K C
EM: slatton@ece.ufl.edu
AF: University of Florida, Department of Electrical and Computer Engineering, Gainesville, FL 32611 United States
AU: * Slatton, K C
EM: slatton@ece.ufl.edu
AF: University of Florida, Department of Civil and Coastal Engineering, Gainesville, FL 32611 United States
AU: Krekeler, C
EM: krekeler@ufl.edu
AF: University of Florida, Department of Electrical and Computer Engineering, Gainesville, FL 32611 United States
AU: Cohen, M
EM: mjc@ufl.edu
AF: University of Florida, Department of Soil and Water Science, Gainesville, FL 32611 United States
AU: McKee, K A
EM: KAMcKee@mail.ifas.ufl.edu
AF: University of Florida, Department of Agricultural and Biological Engineering, Gainesville, FL 32611 United States
AB: Accurate prediction of basin-scale hydrologic behavior is constrained by uncertainty in estimating soil hydraulic behavior. Extreme variability in hydraulic conductivity has been observed (>5 orders of magnitude) over relatively small areas, and studies that have examined the effects of heterogeneity on integrated hydrologic responses have observed substantial errors when structural variability is ignored. This has prompted spatially explicit representations of soil attributes in hydrologic and water quality models (e.g. TOPMODEL) that present significant parameterization constraints at high resolution. Our hypothesis is that elevation data (coarse and fine grain) can serve as a proximate predictor of soil hydraulic properties. We present an information-theoretic method to systematically rank the information contributions, with respect to soil properties (primarily texture), of several surficial and landcover structure features obtained from data at both coarse (~30m) and fine (~1m) spatial scales using a probabilistic measure known as mutual information. The method makes no a priori assumptions about the relative importance of features, thus allowing feature ranking to respond to variations in terrain and landcover. The study site is located in the riparian corridor of an urban watershed (Hogtown Creek) in the city of Gainesville, Florida. It is a surficially closed basin in the St. John's River Water Management District in Northeastern Florida. The area is low-relief and contains mixed land use (natural forested areas and urban development). Topographic data from the USGS National Elevation Dataset (NED) and the NASA Shuttle Radar Topography Mission (SRTM), along with approximate stream locations from the USGS National Hydrography Dataset, are used to generate spatially distributed coarse-scale features regarding surface morphology and drainage. Airborne Laser Swath Mapping (ALSM) data are also used to generate features relating to under-canopy topography and three-dimensional foliage structure at meter scales. The structural features are then related to soil classes provided by the 1:24,000-scale Soil Survey Geographic Data Base (SSURGO). The information contribution of each feature to the maximum a posteriori (MAP) classification of the soil class is ranked. Surface gradient features tend to dominate in and near the stream channels while elevation features contribute more information in the lower reaches of the stream flood plain. The MAP soil classification and classification uncertainty are evaluated for different landuse regimes in the study area.
DE: 1223 Ocean/Earth/atmosphere/hydrosphere/cryosphere interactions (0762, 1218, 3319, 4550)
DE: 1625 Geomorphology and weathering (0790, 1824, 1825, 1826, 1886)
DE: 1804 Catchment
DE: 1839 Hydrologic scaling
DE: 1865 Soils (0486)
SC: Geodesy [G]
MN: Fall Meeting 2005