Soil Series Predictions Using Three Different Satellite Remote Sensing Images in the Everglades, Florida.

Remote sensing provides rapid data collection and dense information grids that allow inference on various biophysical properties across large landscapes. Soil prediction models incorporated with remote sensing images have shown success to improve the predictive power in upland ecosystems. However, not much is known if these models also perform well in wetland ecosystems, especially for prediction of soil classes. The objectives of this study were to (i) develop spectral informed soil taxonomic prediction models and assess their accuracy; (ii) quantify the relationships between soil classes and environmental co-variates derived from remote sensing and geospatial sources; and (iii) compare the effects of spatial resolution (10, 30, and 250 m) of three remote sensing images to delineate soil classes in a subtropical wetland: Water Conservation Area-2A, the Florida Everglades, U.S. Soil series were collected at 108 sites and three satellite images acquired (i) Satellite Pour l’Observation de la Terre (SPOT, 10m), (ii) Landsat Enhanced Thematic Mapper Plus (ETM+, 30 m), and (iii) Moderate Resolution Imaging Spectroradiometer (MODIS, 250 m). Classification trees were used to predict soil series using spectral data and geospatial environmental ancillary datasets. The single tree model using SPOT image derived spectral input variables with environmental data yielded an average accuracy of 85.6%, overall accuracy of 71.3%, and Kappa coefficient of 61.1 %, which was the best prediction results. Both, ETM+ and MODIS informed soil series predictions demonstrated moderate predictive power with average accuracy of 85.5% and 78.0%, overall accuracy of 67.6% and 60.2%, and Kappa of 57.2% and 47.7%, respectively. Results suggest that the variability of soil series can be explained by bedrock/parent material > topographic variables > vegetation properties derived from remote sensing.