2008 Joint Meeting of The Geological Society of America, Soil Science Society of America, American Society of Agronomy, Crop Science Society of America, Gulf Coast Association of Geological Societies with the Gulf Coast Section of SEPM

Paper No. 6
Presentation Time: 2:45 PM

Digital Soil Classification for Soil Survey Using ASTER Satellite Imagery and DEM Data In Organ Pipe Cactus National Monument, Arizona


NAUMAN, Travis W., Soil, Water and Environmental Science, University of Arizona, Tucson, AZ 85721 and RASMUSSEN, Craig, Soil, Water, and Environmental Science, University of Arizona, 429 Shantz Building, University of Arizona, Tucson, AZ, AZ 85721, naumi@email.arizona.edu

In the western U.S. significant areas of remote and rugged terrain remain to be mapped under the auspices of the National Cooperative Soil Survey. The objective of this study was to apply a novel combination of digital soil mapping (DSM) techniques in Organ Pipe Cactus National Monument (ORPI) to aide soil survey of a remote, hyperthermic region of the Sonoran Desert in southwestern Arizona. Specific goals were to (i) generate a “pre-map” of soil-landscape relationships, and (ii) compare classification results with the existing soil survey data. We coupled supervised classification of the landscape into mountains and basins with an iterative principal component data reduction, unsupervised classification, and image segmentation to isolate unique soil-landscape units. Surrogate soil-landscape data layers were derived from ASTER satellite imagery and USGS 30-m digital elevation models (DEM). A large number of band ratios, landscape indices, and topographic layers were derived from the base satellite and DEM data. An iterative principal component and factor loading analysis was used to quantitatively eliminate redundant data layers. Chi-Square contingency analysis normalized to a 0-1 Cramer's V score was used to compare soil-landscape classes with soil taxonomic units (subgroup labeled with particle size in control section) mapped as part of the soil survey. A Cramer's V value of 0.27 indicated the classification captured general patterns from the original soil survey. It was also noted that the classification detected soil units sourced from different lithologies not delineated by the original survey. The procedure also provided soil-landscape classification for remote mountain areas unmapped in the original soil survey. These results demonstrate the potential for using this coupling of classification procedures to identify a minimal yet robust dataset for soil survey update, creating initial pre-maps, and to aid in building sampling designs for quantitative soil mapping efforts.