You are here: Home / Research Topics / Research Highlights / Individual Highlight

Research Highlights

Individual Highlight

Mapping forest disturbance agents with Landsat time series

Photo of Model predictions of forest harvesting and conversion along the U.S.-Canadian border in northern Maine show most of the forest lost to development occurred in Canada, and although annual rates of harvesting (normalized for the total amount of forest area in each country) are roughly similar, the forest cut blocks on the U.S. side are noticeably larger. Model predictions of forest harvesting and conversion along the U.S.-Canadian border in northern Maine show most of the forest lost to development occurred in Canada, and although annual rates of harvesting (normalized for the total amount of forest area in each country) are roughly similar, the forest cut blocks on the U.S. side are noticeably larger. Snapshot : Disturbance plays an important role in shaping forests’ ability to sequester carbon and provide critical ecosystem services such as clean air and water. Satellite remote sensing offers a flexible and cost-effective way to monitor disturbance over large areas. However, most disturbance mapping algorithms only show where and when change has occurred. Using information derived from Landsat time series, Forest Service researchers accurately mapped five different types of forest disturbance including fire, harvest, insects and disease, wind, and conversion (the loss of forest to other land uses).

Principal Investigators(s) :
Schroeder, Todd A.  
Research Location : Ten Landsat scenes from across CONUS
Research Station : Southern Research Station (SRS)
Year : 2017
Highlight ID : 1320

Summary

Recent advances in forest health monitoring have focused on the use of satellite remote sensing techniques to map the year, extent, and location of forest disturbance. Although many newly developed algorithms can detect disturbance in an automated manner, many of the map products only indicate where and when a potential change has occurred. One critical piece of information that is often lacking is the causal agent responsible for the disturbance, such as fire, harvest, or insects. A recent Forest Service study found that change metrics derived from Landsat spectral trajectories can be used to accurately model different types of forest disturbance. Using a two-step modeling approach, Forest Service researchers at the agency’s Southern Research Station were able to map annual changes brought on by fire, harvesting, insects and disease, wind, and conversion of forest to other land uses. The researchers studied 10 diverse locations across the U.S. Separating forest management practices like harvests from forest land use conversion is an important advancement, because both disturbances tend to look similar in spectral space, but they have vastly different carbon consequences. In addition to testing the agent mapping approach, the researchers also offer guidance on other forest health monitoring issues. Their study addresses reference data collection, predictor variable importance, and modeling criteria. Results may impact future development of a national forest disturbance mapping product.

Forest Service Partners

External Partners

 
  • Oregon State University
  • University of Maryland