Tree Cover Mapping

Where are the trees (or where could you hide a mobile missile launcher)?
This is a Landsat Thematic Mapper image of a portion of New Mexico. As part of a military exercise, one organization was charged with identifying locations where SCUD missile launchers could hide so that they could be monitored for activity. The image covers a large area, but with a resolution of 30 meters, groups of trees smaller than a pixel are of sufficient size to hide a launcher. Can you find all the potential hide sites?

IMAGINE Subpixel Classifier did!

Includes material © Space Imaging L.P.

Objective
The Pentagon wanted to determine all locations in a wide area where operational mobile missile launchers could be hidden by tree cover.

Approach
IMAGINE Subpixel Classifier was used to process two entire Landsat images to find whole and subpixel occurrences of trees in the arid landscape.

Challenges
· A large ground area needed to be assessed quickly
· Time and financial resources prohibited the use of ground survey, or aerial photography
· Traditional image processing could detect large stands of timber, but scattered groupings of trees could not be detected
· There was confusion between trees and some types of shrubbery (See Tree Cover Details)

Solution
IMAGINE Subpixel Classifier was used to develop a signature for the predominant tree cover in this region. Because of the scene-based environmental correction (normalization) process built right in to the classifier, the signature was able to be successfully applied to multiple Landsat scenes. High resolution aerial photography and a brief field visit were used to spot-check the performance of IMAGINE Subpixel Classifier. It was confirmed that shrubs inadvertently classified as trees with a Maximum Likelihood classifier were not detected by the Subpixel Classifier and numerous areas of sparse tree cover were identified.

Includes material © Space Imaging L.P.

IMAGINE Subpixel Classifier is a quantitative classifier and will provide pixel fraction information if desired. As indicated by the gradation of greens in the key, the amount of tree cover in each pixel is provided. Military participants were most interested in sparse tree cover as these areas included potential missile hide sites that would likely be missed with traditional exploitation methods. They also found that the fraction information was useful in predicting areas where the tree cover was so thick that heavy vehicles could probably not be driven. This information could thus be used to reduce the amount of area that had to be actively monitored for activity. The results of the IMAGINE Subpixel Classifier were georectified for inclusion in a GIS.