Canadian Forest Service Publications

Spruce budworm damage assessment in northern forests from high resolution satellite imagery. 2003. Leckie, D.G.; Cloney, E.; Gougeon, F.A.; Hill, D.A.; Shand, A.; Alfaro, R.I. Page in Proceedings of the 25th Canadian Symposium on Remote Sensing (CDROM), October 14-17, 2003, Montreal, Quebec. Canadian Remote Sensing Society, Otttawa, Ontario.

Year: 2003

Issued by: Pacific Forestry Centre

Catalog ID: 25314

Language: English

Availability: PDF (download)

Mark record


Ikonos imagery was acquired over a 10 x18 km site centred on Fort Nelson in northeastern British Columbia. The site is typical northern boreal forest predominated by black spruce in upland and wetland areas with white spruce or mixedwood on the better sites and river bottoms. A recent eastern spruce budworm (Choristoneura fumiferana) outbreak has resulted in light through severe cumulative defoliation of the white spruce. Site specific information on defoliation is useful for harvest scheduling, salvage, inventory prediction, determining ecological impacts and planning control programs.

Automated individual tree crown isolation was performed with the 1 m resolution panchromatic band and defoliation level classification conducted on a tree basis using fused multispectral data. For a set of test trees, classification results for the automated isolation most associated with each tree was compared to the ground defoliation assessment of that tree. As well, total defoliation of all isolations within a stand was recorded versus ground estimates of stand level defoliation. Isolation was of moderate quality with some clustering of trees into one isolation and isolations of spurious shapes. For three classes of defoliation, accuracies were 50%, 72% and 75% for light, moderate/severe and dead trees, respectively. On a stand basis, accuracies for light, moderate and severe stands were 52%, 67% and 83%, respectively. The 1 m resolution was marginal for the isolation of the small crowned trees typical of northern forests. The 4 m resolution of the multispectral data was problematic for defoliation level classification. For example, the signatures of dead or severely defoliated trees, which were commonly in more open situations, were often contaminated by the high infrared reflectance of surrounding ground vegetation or shrubs.