|White, M - AUBURN UNIVERSITY|
|Shaw, J - AUBURN UNIVERSITY|
|Rodekohr, D - AUBURN UNIVERSITY|
|Wood, C - AUBURN UNIVERSITY|
Submitted to: Soil Science
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: January 23, 2012
Publication Date: May 1, 2012
Citation: White, M.L., Shaw, J.N., Raper, R.L., Rodekohr, D., Wood, C.W. 2012. A multivariate approach to high resolution soil survery development. Soil Science. 177:345-354. Interpretive Summary: Users of soil surveys require an increasing amount of technical information that is difficult and expensive to acquire. We conducted an experiment that compared the traditional and time-consuming approach to obtain soil survey information to newer techniques that used information from digital elevation models and electrical conductivity measurements. We found that the newer method was able to predict between 60 and 80% of a soil’s variability. This information should be useful to USDA-NRCS and their customers as they attempt to acquire more extensive information about their fields and the variation of soil within these fields.
Technical Abstract: First-order soil surveys are expensive and time consuming to create. Recently developed technologies depicting landscape variability at high resolution may be useful in first-order survey development. Our research objective was to compare a first-order soil survey created using conventional techniques versus a first-order survey developed using terrain attributes calculated from digital elevation models (DEMs) and electrical conductivity (EC) mapping to create soil landscape units prior to field work. Two research sites [Macon (9-ha) and Dale (8-ha), both row-crop fields] were located in the Coastal Plain physiographic region of Alabama. First-order soil surveys (1:5,000) were generated for each field using standard soil survey techniques. Elevation data were collected using RTK-GPS, and terrain attributes were calculated. Field-scale EC data were also collected. Multivariate analyses indicated three principal factors described 81% and 80% of the terrain and EC variability for the Macon and Dale site, respectively. Fuzzy k-means clustering of principal factor scores was used to create landscape zones. Random sampling of conventional map units and landscape zones was used to compare purity of both approaches. Averaged overall, landscape zone purity was slightly less ('4%) than that of conventional survey map units at the Macon site, but slightly higher ('1%) than conventional survey map units at the Dale site. Utilizing a rigid similar/dissimilar criteria, probabilities of success for observing a named soil within a map unit for the landscape zone approach averaged 64% and 84% for the Macon and Dale sites, respectively. The landscape zone approach was relatively more accurate at the Dale site where soils differed mostly in subsoil texture (particle size family), than the Macon site where taxonomic soil variability was largely related to drainage class.