A geographic information systems based landscape classification model to enhance soil survey: A southern Illinois case study

This paper presents an innovative geographic information systems (GIS)-based model that can assist soil scientists in soil survey by providing quantitative data on soils and landscape characteristics. Using GIS data exploration techniques, nine data layers (percent slope, slope length, profile curvature, tangential curvature, mean curvature, flow accumulation, flow tine density, distance to troughs, and distance to summits) were extracted from U.S. Geological Survey 30-m (98-43 ft) digital elevation models. A G IS clustering algorithm identified 50 landscape signatures from the nine data layers. A maximum likelihood discriminate analysis classifier was employed using the nine data layers and the 50 landscape signatures to create landscape classification information. This model provides an objective understanding of the soil-landscape relationship. The model was tested at a case study site in a quarter section of Massac County, Illinois, an area of homogeneous loess. Preliminary analysis indicates that the statistical relevance of the model is high, as a regression equation obtained a coefficient of determination (R-2) of .88. Thus the potential predictive capabilities of the model are great, and should be extended to heterogeneous landscapes through further testing. This GIS landscape model could be used by soil scientists to advance their knowledge of the soil-landscape relationship, and ultimately enhance soil survey in the future. The objective of this research is to develop a quantitative tool to model landscape elements using GIS and digital elevation models for application in soil survey. Soils are closely related to the landforms in which they reside. Hall and Olson (1991) observed that landscapes have a strong nonrandom variability component, which makes landscapes predictable, and since soils are strongly related to landscapes they should also be predictable. Quantitative procedures are available, but are sometimes met with resistance that can impede progress within soil survey. The next generation of soil survey needs to incorporate a major conceptual change in the way soils are mapped, specifically a change in the cartographic model. A continuous mapping cartographic model should be the future of soil survey. But this requires changing current views about the spatial distribution of soil. The current view that "soil types fit together like pieces of a jigsaw puzzle" (USDA-NRCS, 1999: P. 26) needs to be replaced with the view that soils are a continuum with spatial variations. Indeed this model bridges the gap between the "discrete" mapping of current soil survey and the "continuous" mapping efforts of the future.