Hybrid Spatial Modeling of Soil Organic Carbon in the Lake Okeechobee Basin, Florida.

The SCORPAN model has been widely applied in digital soil mapping (DSM) studies at different levels of complexities incorporating all or few of the factors to estimate a soil property of interest. Spatial distribution of soil organic carbon (SOC) is of key interest in the Lake Okeechobee Basin (LOB) in Florida, USA due to the heterogeneous mix of sandy well-drained soils and organic wetland soils in the area (about 20,000 km²). However, we lack clarity which of the SCORPAN factors has the most predictive capabilities for soil carbon in this subtropical landscape. This study is aimed at achieving three objectives: 1) Identify the SCORPAN factors with the most predictive power for soil organic carbon in the LOB; 2) compare and validate hybrid spatial modeling techniques to estimate SOC. We collected soil samples (0-20 cm) at 198 sites and analyzed them using a Shimadzu combustion/acid reaction gas analyzer to measure SOC by subtraction (total – inorganic carbon) and derive SOC stocks (kg m^-2). A wide array of SCORPAN factor datasets including land-cover / land-use, remote sensing vegetation indices (such as Normalized Difference Vegetation Index – NDVI, Enhanced Vegetation Index - EVI, and fraction of photosynthetically active radiation - FPAR), land surface phenology, biomass estimates, climate, hydrology and topography were collected to represent SCORPAN factors. Ensemble tree model - regression kriging were employed to estimate SOC across the basin. Results suggest that soil hydrologic, topo-hydrologic, and land use / land cover imparted the most control to estimate SOC.