ASABE Technical Library - Abstract

Abstract. Laurel wilt (Lw) disease is a fatal disease caused by the fungus Raffaelea lauricola. It is a vascular pathogen which clogs the xylem, blocking the flow of water and nutrients in infected plants, and it is considered as a major threat to the commercial avocado production in Florida. Many of these symptoms are similar to those that are caused by other diseases or factors, such as nutrient deficiency, and thus can be difficult to recognize in a proper time in order to eliminate the infected trees. The standard method for detection of Lw is the polymerase chain reaction technique which is destructive and time-consuming. Remote sensing using multi-spectral imagery has shown promising results for rapid detection of Lw in the field, but the challenge is to differentiate between Lw and trees with nutrient deficiency symptoms. A visible-near infrared (Vis-NIR) portable spectrophotometer was used to collect spectral data from LWD trees, as well as iron-(Fe) and nitrogen-(N) deficient trees and healthy (H) trees grown in greenhouses. Spectral data were collected during the early, and late stages of each symptom and averaged every 10 nm and 40 nm. Two classification methods: multilayer perceptron (MLP) neural network and decision tree (DT) were used to classify healthy, Lw, and stressed trees. MLP classification was more accurate than DT, reaching 99-100% classification accuracy. The best spectral bands for differentiating those classes, regardless to the stage, were mostly in red edge and near infrared. The results indicated that using the selected spectral bands it is possible to separate Lw from nutrient-deficient trees.