A new machine learning system developed at the University of Alaska Fairbanks can automatically create detailed maps from satellite data, pinpointing the locations of spruce trees that may have been killed by the Alaskan beetle, even in forests with low to moderate damage that would normally be difficult to identify.
The automated process will help forestry and wildfire managers make decisions, which is crucial as beetle infestations grow.
The Alaska Department of Forestry and Fire Protection calls spruce longhorn beetles “the most destructive insect in Alaska forests.”
The identification system, developed by Assistant Professor Simon Zwieback of the University of Arizona's Geophysical Institute, ISPRS Journal of Photogrammetry and Remote Sensing May 18. Zwieback is also affiliated with the UAF School of Natural Sciences and Mathematics.
This study fills a gap in knowledge: a method to automatically map the potential for spruce bark beetle infestation in areas of low to moderate severity.
“We don't have a comprehensive map of beetle-killed trees across the state because existing maps rely primarily on expert observations from airplanes, which are expensive and limited in location and time,” Zwiebach said. “This limits officials' ability to respond to ongoing infestations.”
Alaska foresters currently find dead spruce trees in mixed forests using survey flights, time-consuming manual interpretation of high-resolution imagery, and automated analysis of coarse-grained satellite imagery that can identify entire communities of dead trees but not individual dead trees.
None of these identification methods, including Zwieback's, can pinpoint the cause of death in individual trees; the possibility of beetle damage can be inferred because the beetles are well known and have already caused damage.
Zwieback's method combines the efficiency of automation with the detail of high-resolution satellite imagery.
“Using machine learning and high-resolution imagery is the right way to go about mixed forests,” Zwieback said.
Machine learning is a type of artificial intelligence that focuses on developing algorithms and statistical models that enable computers to learn from data and make predictions and decisions based on that data.
Zwieback's machine learning algorithms are trained using known locations of dead spruce trees. During training, the algorithms learn to recognize dead spruce trees based on their characteristic shape, color, shadows and other contextual clues. Once sufficiently trained, they are able to quickly and automatically identify dead spruce trees.
Zwieback tested the technique on imagery from a study area of about 167 acres west of a line from Talkeetna to Byers Lake. The forested areas in the study area consist of mixed spruce and birch forests.
The area has been hit hard by a beetle infestation that began in the mid-2010s.
Zwieback's method was successful in identifying dead spruce trees in stands with only a few dead trees.
Statewide, about 2 million acres have been affected since 2016, primarily in south-central Alaska, and by 2020 had spread as far north as Cantwell and the Alaska Range.
The loss of large numbers of spruce trees leads to many changes in the ecosystem with associated impacts: vegetation on the forest floor is replaced by grasses and shrubs, and dead branches may litter the ground, all of which provides more fuel on the ground and increases the risk of wildfires.
Zwieback's methodology can help inform decisions regarding fire prevention and suppression.
The decline in the value of timber resources and the deterioration of the landscape's aesthetics are also concerns.
Zwieback continues his research.
“As new imagery arrives, we look forward to rolling it out across the state,” he said. “Remote sensing will help us understand disease trends and inform response measures, especially as the disease spreads inland.”
This research was funded by NASA EPSCoR and the National Science Foundation EPSCoR, an established program to stimulate competitive research.
Zwieback said a field study was being set up on land owned by the Ahtna Native American group in south-central Alaska to better understand the progression and impact of the disease as it spreads inland.
-Note: This news release was originally published on the University of Alaska Fairbanks Geophysical Institute website. Because it has been republished, it may not follow our style guide.
