Background
The USDI Bureau of Land Management (BLM) has invested heavily in airborne laser scanning (LIDAR) data in Oregon. LIDAR data has been increasingly used in forestry applications, but cannot directly measure forest inventory. Rather it is capable of taking forest structure measurements, which then need to be translated into more traditional forest inventory metrics through locating and measuring forest inventory plots.
Traditionally, research in this area involves establishing plots in areas of LIDAR coverage by examining aerial photography and existing inventory information, and building regression models relating measurements of these plots to LIDAR metrics.
This project focused on locating field plots by classifying the variability in the LIDAR data to see if the LIDAR data itself can provide a statistically sound basis for selecting inventory plots across a range of forest conditions.
This project has taken place in multiple stages, over several years
- 2010 (Coos Bay): a study area of approximately 250,000 acres of BLM and Coquille tribal lands in the south Oregon coastal forests. The study area was contained within a 1.6 million acre LIDAR acquisition collected in 2008 and 2009.
- 2013 (Rogue Valley): a study area of approximately 640,000 acres of BLM land in the south Oregon Cascades. This study area is contained within a 1.4 million acre LIDAR acquisition collected in 2012.
- 2017: a three part study area (Lane County, Upper Rogue, Upper Umpqua) of approximately 920,000 acres of BLM and Oregon Department of Forestry lands in southwestern Oregon. These study areas are contained within 4.7 million acres of LIDAR acquired between 2013 and 2015.
Our Work
We partnered with the USDA Forest Service Pacific Northwest Research Station (PNW) to develop protocols for selecting field plot locations by identifying the most important LIDAR summary variables and defining the classes within which stratified sampling should occur.
The LIDAR data for each project stage was processed from the raw points into a standardized set of metrics using Fusion. This full set of metrics was narrowed down to two metrics using principle components analysis. The two selected metrics were subdivided based on their distributions and the number of field plots that would be collected. The study areas were stratified into bins using the subdivided LIDAR metrics. For each bin, equal numbers of plot locations were randomly selected, and map books of plot locations were developed for field crews. Field crews hired by BLM used these map books to locate each sample plot and take inventory measurements.
We summarized the field crew inventory measurements into databases and provided this information to the USFS PNW Research Station to develop the regression models to predict forest inventory metrics from LIDAR.
Related Projects
We have performed similar work for:
- a Forest Service/NASA Carbon Monitoring System project in 2015, locating plots across 6 Landsat scenes in 6 states, by processing and stratifying strips of LIDAR within each scene
- the Sealaska Corporation/Nature Conservancy in southeast Alaska in 2016
- the Washington DNR as part of a Riparian Vegetation Monitoring Project in 2017
- the Forest Service/Nature Conservancy on Prince of Wales Island in southeast Alaska in 2018
