Separability of soils in a tallgrasss prairie using SPOT and DEM data

Abstract

A sufficient transformation technique is important in using computer-aided soil pattern recognition to interpret soils in soil surveys. This study was conducted using the Systeme Probatoire d'Observation de le Terre (SPOT) satellite images and Digital Elevation Model (DEM) data to separate five major soils in a tallgrass prairie near Manhattan, Kansas. The high resolution SPOT satellite images were integrated with DEM data. The soils were sampled and classified using conventional soil survey procedures. A canonical transformation technique was used to extract the important soil features from SPOT and DEM data. A pairwise transformed divergence analysis was also conducted to evaluate the several canonical variables in order to sufficiently separate the soils. Consequently, two canonical variables derived from training samples were selected. Our results suggest that canonically transformed data were superior to combined SPOT and DEM data. High resolution SPOT images and DEM data can be used to aid second-order soil surveys in grasslands.