Richard Cutler, Leslie Brown, James Powell, Barbara Bentz, and Adele Cutler (2003), Identifying "Redtops": Classification of Satellite Imagery for Tracking Mountain Pine Beetle Progression through a Pine Forest, Computing Science and Statistics, 35, I2003Proceedings/CutlerRichard/CutlerRichard.paper.pdf
Mountain pine beetles are an integral part of moutain pine forest ecosystems in the intermountain west. Locally, mountain pine beetle attacks may be devastating. With global warming comes the concern that mountain pine beetles may soon be able to attack different species of trees at higher elevations that have no resistance to the beetles.
A collaborative project, between members of the Department of Mathematics and Statistics at Utah State University, and the USDA Forest Service Rocky Mountain Research Station has focused on the use of differential equation models for the mountain pine beetle life cycle and attack mechanism. The work reported here concerns the identification of trees attacked by mountain pine beetles and, more generally, of the vegetation in the Sawtooth National Forest using satellite imagery and ground observations. The satellite imagery has 4m (4 bands) and 1m (panchromatic) resolution. Classification methods applied to the data include CARTs, random forests, and more traditional methods. Results for the different classifiers are presented, along with a discussion of issues concering the matching of ground locations to satellite image pixels.