Multi-Scale Modeling of Biogeochemical Soil Properties Using Remote Sensing.

Remote sensing facilitates high-resolution mapping of vegetation specific and other environmental properties across large aquatic ecosystems. Thus, fusing satellite and aerial images and site-specific observations of biophysical properties in soils have potential to develop accurate, spatially-explicit and continuous soil models that capture gradual variations in soil properties in dependence of environmental covariates.   Our objective was to compare univariate and multi-variate models predicting total phosphorus and total nitrogen in soils utilizing multi-resolution remote sensing images. The study was conducted in Water Conservation Area-2A, the Everglades, Florida. We used various statistical and geostatistical prediction methods with field observations (n: 110 sites) and remote sensing images to predict the spatial distribution of soil properties. The spectral data and derived indices from different remote sensing images, which have different spatial resolutions, included: MODIS (250 m), Landsat ETM+ (30 m), SPOT (10 m), and an orthorectified aerial photograph (30.5 cm). The results suggest that the spectral data derived from remote sensing images improve the predictive quality of biogeochemical soil properties when compared to univariate soil prediction models. In addition, high resolution remote sensing images facilitate to capture unique and complex landscape features of the study area. The intermediate resolution image showed highest predictive capabilities to infer on soil properties.