EP31B-1007
MODELING SOIL-LANDSCAPE RELATIONS IN THE SONORAN DESERT, ARIZONA, USA
Wednesday, 16 December 2015
Poster Hall (Moscone South)
Netra Raj Regmi and Craig Rasmussen, University of Arizona, Tucson, AZ, United States
Abstract:
Digital soil mapping (DSM) techniques that integrate remotely sensed surface topography and reflectance, and map soil-landscape associations have the potential in improve understanding of critical zone evolution and landscape processes. The goal of this study was to understand the soil-geomorphic evolution of Quaternary alluvial and eolian deposits in the Sonoran Desert using a data-driven DSM technique and mapping of soil-landscape relationships. An iterative principal component analysis (iPCA) data reduction routine was developed and implemented for a set of LiDAR elevation- and Landsat ETM+-derived environmental covariates that characterize soil-landscape variability. Principal components that explain more than 95% of the soil-landscape variability were then integrated and classified based on an ISODATA (Iterative Self‐Organizing Data) unsupervised technique. The classified map was then segmented based on a region growing algorithm and multi-scale maps of soil-landscape relations were developed, which then compared with maps of major arid-region landforms that can be identified on aerial photographs and satellite images by their distinguishing tone and texture, and in the field by their distinguishing surface and sub-surface soil physical, chemical and biological properties. The approach identified and mapped the soil-landscape variability of alluvial and eolian landscapes, and illustrated the applicability of coupling covariate selection and integration by iPCA, ISODATA classifications of integrated layers, and image segmentation for effective spatial prediction of soil–landscape characteristics. The approach developed here is data-driven, cost- and time-effective, applicable for multi-scale mapping, allows incorporation of wide variety of covariates, and provides accurate quantitative prediction of wide range of soil-landscape attributes that are necessary for hydrologic models, land and ecosystem management decisions, and hazard assessment.