Date of Graduation

8-2024

Document Type

Dissertation

Degree Name

Doctor of Philosophy in Engineering (PhD)

Degree Level

Graduate

Department

Civil Engineering

Advisor/Mentor

Coffman, Richard

Committee Member

Barry, Michelle

Second Committee Member

Salazar, Sean

Third Committee Member

Tullis, Jason

Keywords

Cobalt chloride; Hue; Machine learning; Microwave; Soil water content; Spectral reflectance

Abstract

The amount of moisture within soil is important for geotechnical engineering properties (soil strength, compressibility, and permeability), plant growth, and hydrological flow within the environment. Traditional methods of measuring soil moisture content include gravimetric (through oven-drying), nuclear, electromagnetic, and ground penetrating radar methods. The use of remote sensing of soil moisture content is a viable alternative with the advantages of being fast, non-destructive, and able to be used across a large area of interest. Remote sensing methods to determine soil moisture content were investigated and applied to validation specimens to study the applicability of the methods at future field sites. Three technologies were investigated for the purpose of determining soil moisture content using remote sensing: 1) the color of cobalt chloride filter paper in contact with the surface of soil specimens, 2) spectral reflectance values within the visible, near-infrared, and short-wave infrared wavelength regions (350 – 2500 nm) using a spectroradiometer, and 3) values obtained from a microwave moisture sensor. For each method, soil specimens were prepared and tested, and the oven-based gravimetric moisture content was determined for use in developing calibration equations. After the development of calibration equations, the calibration equations were applied to validation specimens to determine the accuracy of the method by comparing the predicted moisture content with the oven-based gravimetric moisture content. For each method, the mineralogy and geotechnical engineering behavior of the soil affected the calibration result. As a result, calibration equations were developed for individual soil types. Statistical parameters were evaluated for each method, with coefficients of determination (R2) greater than 0.9, mean absolute error (MAE) values within 2.5 percent of the measured moisture content, and root mean square error (RMSE) values within three percent of the measured moisture content. To increase the practicality of the results, similar soil types were combined and updated calibration equations were developed based on groups of soils. Updated statistical parameters were evaluated, with coefficients of determination as low as 0.88, mean absolute error values at approximately two percent, and root mean square error values less than three percent. With continued study on a wider range of soils, the methods discussed have the potential to be used in field applications in place of traditional methods. The methods will reduce the time and transportation cost related to gravimetric methods and reduce the health hazard risk and cost associated with nuclear methods. Future uses expand to other soil properties, including particle size, salinity, and mineralogy analyses.

Available for download on Thursday, September 11, 2025

Share

COinS