Detecting annual spruce budworm defoliation using digital hemispherical and EO-1 Hyperion hyperspectral imagery
Loading...
Files
Date
2020
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of New Brunswick
Abstract
Estimating defoliation damage during spruce budworm (SBW; Choristoneura fumiferana Clem.) outbreaks is important for implementing effective forest protection strategies, yet methods can be inconsistent. This thesis evaluates digital hemispherical imagery and Earth Observing-1 Hyperion hyperspectral remote sensing data for quantifying annual SBW defoliation, validated using accurate plot-level defoliation measurements from 75 sample plots in Québec, Canada. Hemispherical canopy gap fraction change from May-October, % balsam fir (Abies balsamea (L.) Mill.) basal area, and bioinsecticide spraying were important explanatory variables for modelling percent annual canopy defoliation, with root mean square errors ranging 14–24%. Semi-supervised classification of annual defoliation using the top five hyperspectral vegetation indices resulted in overall accuracies ranging 91–93%. Light (≤30%), moderate (30–70%), and severe (≥70%) defoliation classification accuracy ranged from 86–90%, 93–94%, and 92–96%, respectively. This thesis describes two viable methods that can complement current defoliation surveys in forest protection against SBW.