Authors: Jeremy Johnson*, Prescott College, Marja Haagsma, Oregon State University, Gerald Page, Government of Western Australia/CISERO, Christopher Still, Oregon State University, Kristen Waring, Northern Arizona University, Richard Sniezko, USDA Forest Service: Dorena Genetic Resource Center, John Selker, Oregon State University
Topics: Biogeography, Remote Sensing
Keywords: Hyperspectral remote sensing, white pine blister rust, southwestern white pine
Session Type: Virtual Paper
Start / End Time: 11:10 AM / 12:25 PM
Room: Virtual 32
Presentation File: No File Uploaded
Southwestern white pine (Pinus strobiformis) is a large, long-lived conifer native to the U.S. and Mexico, and is susceptible to white pine blister rust (caused by the non-native fungal pathogen Cronartium ribicola). The species has a suite of strategies, occurring at low population frequencies, for resisting the fungus. Even though genetic resistance occurs in southwestern white pine, it can be very difficult to identify trees with a resistance mechanism so that they can be included in breeding orchards for restoration and reforestation. In this study we tested the ability of hyperspectral imaging, using a custom motion control system and machine learning, to identify and track the progression of the disease so that potential resistance seed sources can be identified. We conducted a greenhouse study on 175 open pollinated seedlings from 10 seed sources selected across the latitudinal range of the species. Seedlings were randomized and half were artificially inoculated with C. ribicola spores while the remaining were used as controls. The seedlings were scored for disease symptoms and patterns of growth. A support vector machine was able to automatically detect infection with a classification accuracy of 87% (k = 0.75) over 16 image collection dates. Additionally, hyperspectral imaging was able to accurately detect health vigor status (as a proxy for disease progression) using the normalized photochemical reflectance index. The approach suggests a way forward to scale up early disease detection in forestry.