<p> Digital soil mapping (DSM) is a rapidly growing area of soil research that has great potential for enhancing soil survey activities and advancing knowledge of soil-landscape relationships. To date many successful studies have shown that geographic datasets can be used to model soil spatial variation. This thesis addresses two issues relevant to DSM, scale effects on digital elevation models, and predicting soil properties. The first issue examined was the effect of spatial extent on the calculation of geometric land surface parameters (LSP) (e.g. slope gradient). This is a significant issue as they represent some of the most common predictors used in DSM. To examine this issue two case studies were designed. The first evaluated the systematic effects of varying both grid and neighborhood size on LSP, while the second examined how the correlation between soil and LSP vary with grid and neighborhood size. Results of the first case study demonstrate that finer grid sizes were more sensitive to the scale of LSP calculation than larger grid sizes. While the magnitude of effect was diminished when comparing a high relief landscape to a low relief landscape, the shape and location of the effect was similar. Results of the second case study showed that the correlation between soil properties and slope curvatures were similarly optimized when varying the spatial extent, but that the effect was more sensitive to grid size than neighborhood size. Slope gradient also showed significant correlations with some of the soil properties, but was not sensitive to changes in grid or neighborhood size.</p><p> The second study attempted to predict numerous physical and chemical soil properties for several depth intervals (0-15, 15-60, 60-100, and 100-150-centimeters), using generalized linear models (GLM) and geographic datasets. The area examined was the Upper Gauley Watershed on the Monongahela National Forest, which covers approximately 82,500 acres (33,400 hectares). This watershed represents a complex landscape with contrasting geologic strata, deciduous and coniferous forests, and steep slopes. Given this landscape diversity it was still possible to fit GLM which explained on average 38 percent of the adjusted deviance for rock fragment content, and exchangeable calcium and magnesium, and phosphorus. Some of the most commonly selected environmental predictors were slope curvatures, lithology types, and relative slope position indices. This seems to validate the prominence of these variables in theoretical soil-landscape models. Had the correlation between the soil properties and slope curvatures not been optimized by varying the spatial extent, it is likely that another less suitable LSP would have been selected.</p>
Identifer | oai:union.ndltd.org:PROQUEST/oai:pqdtoai.proquest.com:1524652 |
Date | 18 December 2013 |
Creators | Roecker, Stephen M. |
Publisher | West Virginia University |
Source Sets | ProQuest.com |
Language | English |
Detected Language | English |
Type | thesis |
Page generated in 0.0012 seconds