An Object-Based Classification Approach in Mapping Tree Mortality Using High Spatial Resolution Imagery

Abstract

In California, a newly discovered virulent pathogen (Phytophthora ramorum) has killed thousands of trees, including tanoak (Lithocarpus densiflorus), coast live oak (Quercus agrifolia), and black oak (Quercus kelloggii). Mapping the distribution of overstory mortality associated with the pathogen is an important part of disease management. In this study, we developed an object-based approach, including an image segmentation process and a knowledge-based classifier, to detect individual tree mortality in imagery of 1 m spatial resolution. The combined segmentation and classification methods provided an easy and intuitive way to incorporate human knowledge into the classification process. The object-based approach significantly outperformed a pixel-based maximum likelihood classification method in mapping the tree mortality on high-spatial-resolution multispectral imagery.