Agricultural sustainability across economic, environmental, and social domains is of great
importance for food security and the continuation of human societies. This is especially true for
potato (Solanum tuberosum), a globally important crop used to produce tubers rich in vitamins,
minerals, carbohydrates, fiber, and protein. In many growing regions, potato production faces
challenges from plant viruses and in-season nitrogen fertilizer management. Specifically, Potato
virus Y (Potyviridae, PVY) is a detrimental plant virus that poses a significant threat to potato
producers on a global basis. Industry stakeholders currently manage PVY infection levels via
insecticide applications, regional seed certification programs that rely on field scouting to
visually assess individual plants for infection status, and destructive and costly tissue sampling
coupled with laboratory assays. Despite these efforts, PVY continues to confound potato industry
stakeholders resulting in economic harm. Additionally, nitrogen is a vital macronutrient essential
for potato production that is often managed with resource-intensive destructive tissue sampling
in an attempt to track seasonal plant nitrogen status. Remote sensing and machine learning
present opportunities for the development of new tools to meet these challenges. This work
outlines efforts and results to gather data and develop Artificial Neural Network (ANN)
classifiers as Multilayer Perceptron (MLP) and Convolutional Neural Network (CNN)
architectures to differentiate potato plant canopies infected with PVY from non-infected
counterparts based on spectral features. MLP modeling yielded high accuracy metrics of 0.894
and CNN model testing resulted in an accuracy of 0.833. Further analysis was conducted to
identify important spectral wavelengths to support future sensor development and industrial
applications. Additionally, data quantifying spatiotemporal potato petiole nitrate (NO3)
concentrations were collected to support ANN development using publicly available climate and remote sensing features achieving mean coefficient of determination (R2
) scores between actual
and predicted values of 0.557 during replicated testing.
Keywords: artificial intelligence, machine learning, petiole nitrate, potato, potato virus y, remote
sensing |