Evaluation of Visible-Near Infrared Reflectance Spectra of Avocado Leaves as a Non-destructive Sensing Tool for Detection of Laurel Wilt

Laurel wilt, caused by the fungus Raffaelea lauricola, affects the growth, development, and productivity of avocado, Persea americana. This study evaluated the potential of visible-near infrared spectroscopy for non-destructive sensing of this disease. The symptoms of laurel wilt are visually similar to those caused by freeze damage (leaf necrosis). In this work, we performed classification studies with visible-near infrared spectra of asymptomatic and symptomatic leaves from infected plants, as well as leaves from freeze-damaged and healthy plants, both of which were non-infected. The principal component scores computed from principal component analysis were used as input features in four classifiers: linear discriminant analysis, quadratic discriminant analysis (QDA), Naïve-Bayes classifier, and bagged decision trees (BDT). Among the classifiers, QDA and BDT resulted in classification accuracies of higher than 94% when classifying asymptomatic leaves from infected plants. All of the classifiers were able to discriminate symptomatic-infected leaves from freeze-damaged leaves. However, the false negatives mainly resulted from asymptomatic-infected leaves being classified as healthy. Analyses of average vegetation indices of freeze-damaged, healthy (non-infected), asymptomatic-infected, and symptomatic-infected leaves indicated that the normalized difference vegetation index and the simple ratio index were statistically different.