/AnMtgsAbsts2009.53929 Quantification of Soil Redoximorphic Features by Standardized Color Identification.

Wednesday, November 4, 2009
Convention Center, Exhibit Hall BC, Second Floor

T. Kevin O'Donnell1, Keith Goyne1, Claire Baffaut2, Stephen Anderson1, Randall Miles1 and Kenneth A. Sudduth2, (1)Department of Soil, Environmental and Atmospheric Sciences, Univ. of Missouri, Columbia, MO
(2)USDA-ARS, Cropping Systems and Water Quality Res. Unit, Columbia, MO
Abstract:
Standardized measurements of soil redoximorphic feature expressions are lacking in spite of increased reliance on these features for environmental decision making.  Current methods for documenting soil redoximorphic features include field determinations of presence/absence, color description, and estimation of presence and size.  These assessments depend in part on light conditions and the experience and bias of the human investigator.  Additional measurements of density, shape, and spatial arrangement of these features may integrate chemical and physical processes responsible for their formation in soils.  Readily available and low cost digital photography equipment, image analysis techniques, and image quantification software encourage soil scientists to improve current methodologies.  This research documents supervised image classification of soil redoximorphic features present in claypan soil profiles of northcentral Missouri, U.S.A by use of Munsell © color groupings.  Soil mineralogy and genesis have favored the formation of perched water tables.  Iron/manganese concentrations and redox depletions were visually observed in soil profiles.  A total of 238 Munsell © color chips were photographed (10R to 5Y), stored in 16-bit RGB color space, and used as spectral signatures for supervised image classification.  Five soil color groups were used to identify soil features, including low chroma and high chroma features.  Accuracy assessments indicated overall accuracy was 99.6%.  Eight-centimeter diameter cores were obtained by hydraulic coring, transported to a lab, split lengthwise along natural structural units, and photographed under controlled lighting and moisture conditions.  Supervised classification enabled quantification of soil redoximorphic features by a number of metrics including density, area, shape complexity, fractal dimension, contagion/interspersion, and complete image feature diversity.  The use of image software allowed additional measures of spatial relationships between features (e.g., soil redoximophic features within a specified distance from voids).  Image analysis techniques hold great promise for more standardized, defendable characterizations of soil features and evidence of relative soil saturation.