Theses and Dissertations

Issuing Body

Mississippi State University


Thomasson, Alex

Date of Degree


Document Type

Graduate Thesis - Open Access


Agricultural Engineering

Degree Name

Master of Science


College of Engineering


Department of Agricultural and Biological Engineering


Soil chemical and physical properties are important to farm productivity, and they vary within fields, so farmers are interested in managing inputs like fertilizer according to local soil conditions within fields. Thus, they must have knowledge of soil conditions of interest, which have historically been measured at a few locations with tedious soil sampling and laboratory analyses. Advantageous to farmers would be a measurement method that provided more geographically detailed information at similar or lower cost. Remote and ground-based optical sensing are possibilities for gathering detailed soil information rapidly and inexpensively. This study considers the possibility of optically measuring soil characteristics. The first objective was to determine relationships between spectral reflectance in the 250- to 2500-nm range and the following soil constituents: clay, sand, Ca, K, Mg, Na, P, Zn, and acidity (pH). The second objective was to find wavebands for estimating certain soil properties, with the goal of sensor development. Physical, chemical, and spectral-reflectance measurements were made on 969 soil samples collected from two Mississippi fields over two years. Reflectances were averaged over 50-nm wavebands and analyzed with simple- and multiple-linear regression and canonical correlation in relation to soil properties. No single waveband was highly correlated to any soil property in this study, but waveband groups exhibited strong correlations with some soil properties. Clay was the only property consistently strongly correlated (R2 ¡Ý 0.50) to waveband groups over different fields and years. In general, waveband groups that were most highly correlated with a specific soil property in one field in one year were not similar to waveband groups most highly correlated with that property in a different field or year. Thus, it was difficult to select a waveband group for sensor development regarding a specific soil property. However, a group of nine promising wavebands was considered for estimating clay, and results for data in this study indicated the feasibility of grossly estimating clay content with spectral reflectance. Canonical correlation analysis demonstrated strong correlations among certain groups of soil-properties and wavebands. Clay appeared as the most promising property for sensor development from this portion of the study also.



Soil Properties||Conanical Correlation||Reflectance