Paper No. 8
Presentation Time: 10:25 AM
A NOVEL APPROACH FOR EVALUATING SOIL SPATIAL DATA ON A LOW RELIEF GLACIAL TILL PLAIN IN NE INDIANA USING GEOMORPHONS
Landforms have a direct impact on water flow and distribution, which in turn relates to pedogenic processes within a given landscape. With the advent of Geomoprhons (add-on in GRASS Software) as a geospatial tool for landform identification and categorization, it may be possible to improve the representation of landscape morphology within the geospatial realm and apply the algorithms to accurately predict soil properties. Our hypothesis is that prediction of soil chemical and physical characteristics, as they relate to soil landscape processes, may be greatly improved by the use of Geomorphons. The distribution of six functional soil properties (A horizon thickness, depths of: carbonates, effervescence, glacial till, redoximorphic features, and platy structure) within a small field in Wells County, Indiana were analyzed using Geomoprhons derivatives for three grid resolutions including 1.5, 5, and 10 meters to determine if the predictability of soil patterns as they relate to geomorphology can be improved. Multivariate analyses of variance (MANOVA) for three grid resolutions of Geomorphons classes and the six soil properties were used to determine the number of landform groups related to soil distribution. Linear discriminant analyses (LDA) were used for all grid resolutions to determine how accurately the Gemorphons represented the soil distribution across the landscape via a cross validation output. Results for the MANOVA and LDA suggest that for all grid resolutions there are only two groups of Gemorphons related to soil distribution from the original eight groups for the 1.5m data, nine groups for the 5m data, and five groups for the 10m data. Using these two groups greatly improves the overall accuracy of Geomorphons to predict soil distribution from 64% to 83% for the 1.5m data, 56% to 77% for the 5m data, and 66% to 79% for the 10m data. Geomorphons proved to be an excellent tool to help predict soil properties across the landscape. However, there are some limitations and minor modifications must be applied to the output because it was not originally designed for soil applications.