The overall goal of this research is to develop ground-based technologies for disease detection and mapping which can maximize the effectiveness and efficiency of cotton root rot (CRR) treatments. Accurately mapping CRR could facilitate a much more economical solution than treating entire fields. Three cotton fields around CRR-prone areas of Texas have been the sites for three years of data collection. A complete soil apparent electrical conductivity (ECa) survey was conducted for each field with an EM38DD sensor. Multiple linear regression was used to relate physical and chemical soil properties to the ECa values obtained from the EM38DD. The variability in soil ECa measurements can be best accounted for using calcium carbonate levels as well as clay and sand contents in the soil. T-tests were used to determine that soil pH, clay, sand, and inorganic carbon content were significantly related to CRR incidence as determined by aerial images of each location. Spectral data were obtained for freshly picked cotton leaves from healthy, disease-stressed, and dying or dead plants using an ASD VisNIR spectroradiometer. The leaf spectra were evaluated using linear discriminant analysis (LDA), the receiver operator characteristic, and wavelet analysis to relate them to classifications of infection level. It was determined that healthy and infected leaves can be correctly classified 85% of the time based on the spectral data. The results from this study suggest that differences in soil characteristics may not be pronounced enough to accurately map CRR in the soil; however, the precision treatment of CRR may possible using an optoelectronic sensor to diagnose infected plants based on leaf reflectance.
