Detection of mountain pine beetle infestations using Landsat TM Tasseled Cap transformations

• Main Author: Sharma, Rajeev
• Language: English
• Published: 2009
• Online Access: http://hdl.handle.net/2429/11449

This study investigated the identification of probable mountain pine beetle (Dendroctonus ponderosae Hopk.) attacked sites using Tasseled Cap transformations namely, brightness, greenness and wetness, derived from Landsat-7 digital data in parts of Vanderhoof Forest District, in Prince George Forest Region, British Columbia. Lodgepole pine (Pinus contorta Dougl.) constitutes about 80 percent of the total forest vegetation in the study area. About 85 percent of lodgepole pine stands are greater than 60 years of age, and hence susceptible to mountain pine beetle attack. Landsat-7 ETM digital data (Bands'!, 2,3,4,5&7), acquired on August 2, 1999 and September 12, 1999, were the primary remote sensing data source for the study. In addition, TRIM map sheets (1:20,000) derived road and river vectors, and 1:20,000 forest cover maps and a beetle infestation coverage maps of the area (prepared based on aerial sketch mapping and ground probes) were collateral data sources. Methodology consisted of: i) pre-processing of satellite data (atmospheric and geometric corrections), ii) computation of Tasseled Cap coefficients for the Landsat-7 data, since these were not available, iii) identification of mountain pine beetle attacked stands, and iv) accuracy assessment of attacked stands. Some of the major observations based on results obtained were: i) Tasseled Cap indices for infestations of more than 30 attacked trees / site (< 0.09 ha in size) were found to vary in a relatively narrow range; however for infestation sites with less than 30 tree / site the Tasseled Cap indices had random and large dispersions; ii) values of Tasseled Cap indices for September were found to be lower than those for August for all the cover types; iii) differences between mean brightness, greenness, and wetness of healthy and attacked stands were statistically significant for August; iv) the identification accuracy for attacked lodgepole pine stands were 38.82 and 26.17 percent for August and September, respectively; v) a linear relationship was observed between the number of attacked trees at a site and identification accuracy for the August data but not for September data; vi) the poor identification accuracy was mainly due to the sub-pixel size of infestations.