dc.description.abstract | Proximal soil sensors, such as a VisNIR spectrometer mounted into a penetrometer, are being developed as tools to measure soil properties in situ with the goal of providing real-time and spatially explicit soil characterization and measurements without traditional laboratory data. This work addresses major challenges and questions facing the implementation of this technology as it gains popularity for commercial use. Research seeks to determine 1) whether external parameter orthogonalization (EPO) is a robust method to remove soil moisture effects from in situ spectra, 2) if there is an influence of library and EPO dataset on soil property predictions such as clay and organic carbon content, 3) if spectra should be averaged (e.g. by depth or by horizon), 4) if VisNIR spectroscopy is better at predicting surface versus subsurface soil properties, and 5) whether spectral predictions of soil properties, such as clay content, are helpful to classify soils. To assess the modeling decisions on clay and carbon content predictions, three dried ground soil spectral libraries were calibrated and transformed with three EPO datasets to predict Illinois soils. Results indicated that both Texas state and national libraries could provide robust soil property predictions with the Texas EPO. A second experiment implemented spectral averaging and soil prediction averaging in two model scenarios (using an intact, field moist library or a dried ground library and EPO) and reported that averaging may provide a slight increase in prediction accuracy, and supported the notion that a dried ground library and EPO is a robust way to predict soil properties and that subsurface soils produce more accurate clay content predictions than surface soils. VisNIR predictions of clay content were able to categorize soils into series more precisely than soil mapping. Further research may consider real-time characterization and further support is also needed to assess the prediction differences in surface and subsurface soils and the driving influences. | en |