Tools and Insights for Sustainable Management of Plant Pests and Pathogens

Abstract

Minimizing crop yield losses caused by plant pathogens is one way of increasing agricultural productivity. To that end, I developed an automated classifier of digital images of soybean diseases to assist with early and accurate detection of pathogens. This application, based on a convolutional neural network, distinguishes between eight soybean disease/deficiency classes with an overall accuracy of 96.75%, which may help minimize pesticide usage and improve overall productivity. I also performed a quantitative integration of the existing research characterizing the relationship between virulence and within-host pathogen accumulation. By doing so, I aimed to help increase our ability to foresee and manage the evolution of highly-virulent pathogen genotypes. In these two ways, I pursued my overarching goal of developing tools and gaining insights for the sustainable management of plant pathogens.