USING HYPERSPECTRAL IMAGING TO QUANTIFY CADMIUM STRESS AND ESTIMATE CONCENTRATION IN PLANT LEAVES
Cadmium (Cd) is a highly mobile and toxic heavy metal that negatively affects plants, soil biota, animals and humans, even in very low concentrations. Currently, Cd contamination of cocoa produced in Latin American countries is a significant problem, as concentrations can exceed acceptable levels set by the European Union (0.5 mg/kg), sometimes by more than 10 times allowable levels. In South America, Theobroma cacao is an essential component of the basic market basket. This crop contributes to the Latin-American trade balance, as these countries export cacao and chocolate-based products to major consumer countries such as the United States and Europe. Some soil amendments can alter the bioavailability and uptake of Cd into edible plant tissues, though cacao plants can accumulate Cd without displaying any visible symptoms of phytotoxicity, which makes it difficult to determine if potential remediation strategies are successful. Currently, the only effective way to quantify Cd accumulation in plant tissues is via destructive post-harvest practices that are time-consuming and expensive. New hyperspectral imaging (HSI) technologies developed for use in high-throughput plant phenotyping are powerful tools for monitoring environmental stress and predicting the nutritional status in plants. Consequently, the experiments described in this thesis were conducted to determine if HSI technologies could be adapted for monitoring plant stress caused by Cd, and estimating its concentration in vegetative plant tissues. Two leafy green crops were used in these experiments, basil (Ocimum basilicum L. var. Genovese) and kale (Brassica oleracea L. var. Lacinato), because they are fast growing, and therefore, could serve as indicator crops on cacao farms. In addition, we expected these two leafy green crops would differ in their morphological responses to Cd stress. Specifically, we predicted that stress responses would be visible in basil, but not kale, which is known to be a hyperaccumulator. The plants were subject to four levels of soil Cd (0, 5, 10 and 15 ppm), and half of the pots were amended with biochar at a rate of 3% (v/v), as this amendment has previously been demonstrated to improve plant health and reduce Cd uptake. The experiments were conducted at Purdue’s new Controlled Environment Phenotyping Center (CEPF). The plants were imaged weekly and manual measurements of plant growth and development were collected at the same times, and concentrations of Cd as well as many other elements were determined after harvest. Fourteen vegetation indices generated using HSI images collected from the side and top view of plants were evaluated for their potential to identify subtle signs of plant stress due soil Cd and the biochar amendment. In addition, three mathematical models were evaluated for their potential to estimate Cd concentrations in the plant biomass and determine if they exceed safe standards (0.28 mg/kg) set by the Food and Agriculture Organization (FAO) for leafy greens. Results of these studies confirm that like many plants, these leafy green crops can accumulate Cd levels that are well above safety thresholds for human health, but exhibit few visible symptoms of stress. The normalized difference vegetation index (NDVI) and the chlorophyll index at the red edge (CI_RE) were the best indices for detecting Cd stress in these crops, and the plant senescence and reflectance index (PSRI) and anthocyanin reflectance index (ARI) were the best at detecting subtle changes in plant physiology due to the biochar amendment. The heavy metal stress index (HMSSI), developed exclusively for detecting heavy metal stress, was only able to detect Cd stress in basil when images were taken from the top view. Results of the mathematical models indicated that principal components analysis (PCA) and partial least squares (PLS) models overfit despite efforts to transform the data, indicating that they are not capable of predicting Cd concentrations in these crops at these levels. However, the artificial neural networks (ANN) was able to predict whether leafy greens had levels of Cd that were above or below critical thresholds suggested by the FAO, indicating that HSI could be further developed to predict Cd concentrations in plant tissues. Further research conducted in the field and in the presence of other environmental stress factors are needed to confirm the utility of these tools, and determine whether they can be adapted to monitor Cd uptake in cacao plants.