Theses and Dissertations
ORCID
https://orcid.org/ 0000-0002-6826-8420
Advisor
Wijewardane, Nuwan K.
Committee Member
Feng, Gary
Committee Member
Tagert, Mary L.
Committee Member
Martins, Vitor S.
Date of Degree
12-12-2025
Original embargo terms
Embargo 1 year
Document Type
Dissertation - Open Access
Major
Engineering (Biosystems Engineering)
Degree Name
Doctor of Philosophy (Ph.D.)
College
James Worth Bagley College of Engineering
Department
Department of Agricultural and Biological Engineering
Abstract
Accurate and efficient estimation of soil properties is essential for advancing sustainable agriculture, water management, and soil survey applications. Traditional laboratory methods and pedotransfer functions (PTFs) are widely used but are often constrained by high costs, time requirements, and limited scalability. As an alternative, mid-infrared (MIR) and visible–near-infrared (vis–NIR) spectroscopy provide cost-effective, rapid, and non-destructive approaches for soil analysis. However, variability among spectrometer types, geographic regions, and soil scanning conditions poses challenges to model transferability and generalizability. This dissertation evaluated strategies to overcome these challenges and enhance the prediction of soil chemical, physical, and hydrological properties in Mississippi and Texas. Datasets were derived from the USDA National Soil Survey Center–Kellogg Soil Survey Laboratory (KSSL) spectral library and regional collections from both states. Study 1 examined approaches for improving the transferability of spectral data across spectrometers and regions to enhance soil property prediction. Although no single correction technique proved universally optimal, combinations of baseline correction, standard normal variate, and spiking yielded the most robust models. Study 2 evaluated the transferability of models across soil scanning conditions and demonstrated that correction techniques are essential when predicting non-fine-ground soil spectra using fine-ground spectral libraries. Among the evaluated approaches, spiking with extra weighting was the most effective, improving transferability while minimizing the need for labor-intensive grinding. Study 3 compared the predictive accuracy of spectroscopy-based models with PTF-based models (Rosetta 3) for estimating Mualem–van Genuchten hydrological parameters and derived properties. Spectroscopy consistently outperformed PTFs, achieving notably higher accuracy for predicting field capacity and permanent wilting point. The MIR-based models proved superior to vis–NIR, while spectra from fine-ground and non-fine-ground soils yielded greater stability than fresh soil spectra. Overall, these studies confirmed that MIR spectroscopy, integrated with optimized correction techniques, modeling approaches, spectrometer configurations, and soil scanning conditions, provides a reliable framework for accurately predicting soil properties. These findings underscored the feasibility of leveraging spectral libraries for regional-scale soil prediction, enabling USDA-Natural Resources Conservation Service field offices to minimize reliance on traditional laboratory methods, streamline soil analysis workflows, and enhance the timeliness and spatial resolution of soil assessments.
Recommended Citation
Gamagedara, Kanthake Yasas Bandara, "Use of mid-infrared spectroscopy for estimating soil chemical, physical, and hydrological properties in Mississippi and Texas" (2025). Theses and Dissertations. 6792.
https://scholarsjunction.msstate.edu/td/6792
