Volume 1, Issue 3 - Spring/Summer 2008
Detecting Forest Threats through the Hyperspectral Lens
Could spectral "signatures" cast new light?
By Stephanie Worley Firley, EFETAC
When forest health is threatened, early detection is key. Resource managers do not always know the full extent of the threat at hand, however, especially when affected areas are difficult or impossible to access on the ground. A remote sensing technique known as hyperspectral imaging is showing promise as a method to collect data used for mapping forest health decline.
Here’s how it works—all plants reflect electromagnetic energy from the sun at multiple wavelengths, including visible and non-visible light. The characteristic reflectances of objects across a range of wavelengths are called spectra. Unique spectral signatures associated with different objects allow them to be identified from the air using sophisticated sensors. Researchers are finding that hyperspectral sensors, which create images based on multiple spectral measurements at specific wavelengths, may be able to differentiate forest species or track changes in forest cover over time.
EFETAC and its sister center, the Western Wildland Environmental Threat Assessment Center (WWETAC), provided support for three recently completed projects that tested use of hyperspectral imaging technology for detecting and mapping forest threats. Two of these studies involved researchers from the Institute for Technology Development (ITD), a non-profit company located at NASA’s Stennis Space Center that develops satellite, airborne, field, or laboratory imagery applications for a wide range of research areas.
ITD collected data along roads in the Daniel Boone National Forest in eastern Kentucky, an area plagued by nonnative invasive plants like tree-of-heaven, autumn olive, Chinese silvergrass, bush honeysuckle, and multiflora rose. Researchers found that the spectra of some invasive plants were sufficiently different from surrounding native plants or other invasives to allow them to be readily detected and mapped. Southern Research Station research ecologist James Miller points out seasonal changes and landscape components as factors that influenced the project’s success. "Even in these complex forest ecosystems, hyperspectral imagery analysis was capable of identifying particularly troublesome invasive species with an accuracy of greater than 80 to 90 percent," says Miller.
In Humboldt County, CA, and Curry County, OR, data from ITD sensors were used to map the occurrence of sudden oak death (SOD), which has devastated populations of oak and tanoak trees in the region. Hyperspectral images were fairly accurate when locating SOD host species, but detecting diseased trees varied by season and canopy characteristics. "Given the right forest type and a narrowed-down forest area to scrutinize closely, hyperspectral imagery could be conceivably useful in detecting infested stands at a slightly earlier stage of disease progression than normal visual methods that only detect mortality," concludes Chris Lee, a project collaborator from Cooperative Extension at the University of California-Davis.
EFETAC also supported a project to test hyperspectral imaging for mapping ash tree locations and detecting emerald ash borer infestations in areas of Michigan and Ohio. The project was led by Northern Research Station (NRS) and University of New Hampshire’s Complex Systems Research Center researchers who used hyperspectral images from a commercially available sensor. "When linked to detailed ground measured reference data, very detailed decline, including early decline symptoms, can be accurately mapped using hyperspectral remote sensing," finds Jennifer Pontius, NRS research ecologist. "This enables early identification of infested areas and could be used to improve the efficacy of control and monitoring efforts."
Above: Decline Close-Up, Independence Lake, MI. The predicted decline coverage over Independence Lake, a region of high ash density and prolonged emerald ash borer infestation, highlights large areas of high decline (areas in red have the highest rates of decline, areas in green have the lowest). From: Remote Sensing of Environment, 112, 5: 2665-2676.
As the technology develops, hyperspectral imaging may become an important piece of a nationwide early warning system for detecting forest threats. Cooperative Extension’s Lee says that, "Because quick response is essential for dealing with forest threats, any treatment program is only as strong as its early detection component. Testing the hyperspectral classification technique helps us understand the possibilities and limitations of remote sensing to detect general large-scale declines in forest trees at early stages."
Pontius adds, "Hyperspectral technologies have been well tested in the realm of research. The next step is to bring such techniques to wider applications in the resource management sectors."