US: A new way of studying and visualising earth science data from a NASA and US Geological Survey (USGS) satellite programme is resulting in, for the first time, the ability to tease out the small events that can cause big changes in an ecosystem.
Called LandTrendr, this computer programme is able to find patterns previously buried within vast amounts of scientific data. Still in development, it”s already led to seeing for the first time in satellite imagery an obscured, slow-moving decline and recovery of trees in Pacific Northwest forests.
Comparing satellite data to ground data, scientists uncovered the cause. “It was, as it turns out, bugs,” said Robert Kennedy, a remote sensing specialist at Boston University, who consulted with US Forest Service experts to confirm his observations.
The unexpected disturbance pattern showed a long slow decline of tree health over years followed by slow regrowth. It emerged in several areas, particularly near Mount Hood in the 1980s, peaking in 1992 when regrowth began, and near Mount Rainier where the insect outbreak lasted ten years from its onset in 1994 till the insects killed all the trees and moved on in 2004.
Kennedy created the LandTrendr programme specifically to work with data from the NASA and US Geological Survey (USGS) Landsat programme. Kennedy”s new way of viewing Landsat imagery has already changed how the Forest Service in the Pacific Northwest operates its yearly forest monitoring program that uses ground stations, satellite imagery and statistics to evaluate current conditions.
Kennedy said that LandTrendr works because of the unique nature of Landsat data. The data embedded in images are a scientific record of the earth”s surface that goes back 40 years. Each image, or scene, covers an area 115 miles by 112 miles (185 kilometers by 180 kilometers) and provides data for wavelengths of light reflected or emitted from the Earth”s surface, which scientists use to see, for example, forest conditions not apparent in visible light. With the four-decade record, they can compare images between years and see how the land changes with time.
Kennedy”s breakthrough was to combine cloud-free pixels from multiple scenes of the same area collected over the growing season in late summer. Then he compares the new images for each year to one another. By breaking a scene down into smaller sized pixels, the cloud-covered portions could be tossed away, but LandTrendr keeps the clean bits to reveal the life history of each pixel.
“We”re getting better data use out of what people think of as crummier images,” said Curtis Woodcock, a remote sensing specialist at Boston University who employs a similar method to Kennedy to build an image of the landscape out of Landsat data pixel by pixel.
What makes all this possible are two things: Computers are finally powerful enough to process vast amounts of data, and Landsat data is now available free of charge.