There is need to develop digital, high-resolution soil biogeochemical signature maps across large landscape units to assess responses to anthropogenic and natural forcing functions. Direct spectral mapping of vegetation and other landscape features has been successfully demonstrated. In this study we explored if spectral signatures can be used to indirectly infer on soil properties using functional, empirical models. The main objective was to compare spatially-explicit models to predict soil total phosphorus TP using (i) site-specific observations, (ii) site-specific observations and spectral datasets, and (iii) a combination of site-specific observations, spectral datasets and ancillary environmental GIS data layers. We used 111 site-specific observations of soil TP collected in Water Conservation Area-2A, Everglades. Spectral datasets comprised two remote sensors, Landsat 7 Enhanced Thematic Mapper (ETM)+ and Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). Three different methods were used (ordinary kriging, co-kriging and regression kriging) to predict soil TP. Models that incorporated both - the spatial autocorrelation structure and spatial covariation(s) between spectral data and observations - performed best to predict soil TP. Spectral data and indices provide dense, high-resolution signatures that combined with site-specific soil data allow monitoring of wetland ecosystems and document restoration success across larger landscapes. There is potential to extend the inferential spectral mapping approach to predict other biogeochemical soil properties.