Trends Summer 2018 - Page 5

The Science Behind the Study A yres Associates opted to use remote sensing for several reasons. Although satellite data was readily available, it simply wasn’t the right source for a project of this magnitude. Its 5- to 30-meter-pixel- resolution was far too large to analyze the single trees and small stands Menominee Tribal Enterprises was interested in researching. Instead, Ayres’ aerial mappers reversed the process and created objects from each pixel collected in the spring and fall flights. The team collected data at carefully timed internals when specific tree species would be more easily identified – for instance, during peak fall colors for aspen and mid- spring to capture pin cherries at their peak bloom. Near-infrared imagery Imagery with analyzed data They then used segmentation techniques to examine each pixel in an image and compare it to adjacent pixels similar in spectral signature and group shapes. “You create an object, and then you start matching pieces to that object,” explained Adam Derringer, project manager. “It’s an extremely heavy process, when you start thinking about how many pixels are in that 21-mile area, but you bring all that together and then start to look at the shapes. That gave us our segmentation to use in our supervised classification routines to identify the species.” Spring near-infrared imagery The team’s use of 6-inch-pixel-resolution data over something like LandSAT data gave MTE far more detailed data – 40 times more data, in fact – than what satellite data would’ve provided. In the end, the estimated accuracy of their analysis was more than 90% against the control data. Analysis tested against ground truth points │5