Dr. Yufeng Ge
Date of this Version
The research community in soil science and agriculture lacks a cost-effective and rapid technology for in situ, high resolution vertical soil sensing. Visible and near infra-red (VisNIR) technology has the potential to be used for such sensor development due to its ability to derive multiple soil properties rapidly using a single spectrum. Such efforts must, however, overcome a few challenges: (i) a dry ground soil spectral library that can be used to predict the target soil properties accurately, (ii) a robust design which can acquire high quality VisNIR spectra of soil, (iii) an effective method that can link field intact soil spectra to the dry ground spectra in the library.
The overall goal of the work presented in this dissertation was to design, develop, and test a VisNIR integrated multi-sensing penetrometer to estimate soil properties in vertical profile. To achieve this goal, three specific objectives were developed. The first was to investigate and compare the usefulness of five approaches: External Parameter Orthogonalization (EPO), Direct Standardization (DS), Global Moisture Modeling (GMM), Slope Bias Correction (SB) and Selective Wavelength Modeling (SWM), in enabling VisNIR dry ground models to be applied directly to moist soil spectra to predict soil organic carbon and inorganic carbon. The second was to design new VisNIR probes and test them in terms of spectral quality and predictive power using an external spectral library under laboratory conditions. Third was to develop the fully integrated, multi-sensing penetrometer system for high resolution vertical soil sensing and field test the penetrometer to evaluate its performance.
The results showed that EPO, DS and GMM account satisfactorily for the effect of moisture in soil spectra. The VisNIR probe developed showed high spectral quality, however with a systematic difference compared to standard MugLite® spectra which was successfully rectified by DS or spiking. The final designed fully integrated, multi-sensing penetrometer system, could estimate soil properties: total carbon, total nitrogen and bulk density, in vertical soil profile with EPO to correct for field intactness. This can lead to a rapid, robust and cost-effective penetrometer system for in situ high resolution vertical soil sensing in the future.
Advisor Yufeng Ge