The Mountain Pine Beetle (Dendroctonus ponderosae Hopk.) and related bark beetles are the most serious insect pests damaging the forests of western Canada. The preferred control strategy is to remove
infested trees before the beetles emerge and spread the infestation. Despite much effort by a number of researchers to develop remote sensing methods to detect infested trees at the early green attackstage, early detection has proven to be a very
elusive goal. Successful detection has been reported using high resolution (1:2000 scale) colour infrared photography, but the high costs of surveys at these large scales make the method impractical for operational use. In 1986 the Canada
Centre for Remote Sensing (CCRS) began a cooperative project with the British Columbia Ministry of Forests to improve the use of remote sensing technology for mapping mountain pine beetle attack. Foliar spectroscopic studies carried out under that
project identified the spectral bands best suited for detecting green attacked trees. In addition, a study with the airborne MEIS (Multi-element Electro-optical Imaging Sensor) showed that high spatial resolution was necessary to detect trees at the
red attackstage, and implied that high resolution would also be necessary for detecting green attack. In 1989 CCRS and the Ministry of Forests acquired data for a study of the use of MEIS data for distinguishing green attacked lodgepole pine
(Pinus contorta Dougl. var. latifolia Engelm.) from unattacked lodgepole pine. The area chosen was located approximately 35 km southeast of Cranbrook, B.C. MEIS data at 0.4 m, 1.0 m. and 2.2 m resolution were acquired over the study area on
1989.09.01. A sample of 256 infested trees and 250 unattacked trees were identified in four principal areas on the ground and on colour infrared aerial photography acquired simultaneously with the MEIS data. The trees used in this study were infested
during the summer of 1989 and their foliage was still green with no visual sign of weakening or discolouration at the time of the ground sampling, 1989.10.02 through 1989.10.05. Spectral features were analyzed using visual enhancement
techniques and digital measures of the separability between the attacked and unattacked classes. The results showed larger differences occurring between sites than between the attacked and unattacked trees. |