New forest inventory method a potential game-changer.
If you need to estimate the value of a forest — whether a plantation of Douglas-fir or a small woodlot — experts in mensuration (aka biometricians) can give you a time-honored method. The traditional process of sampling plots, the measuring of tree heights and diameters, allows an estimation of resources at the landscape level. It works well enough; a similar process also served the ancient Romans.
However, there is a price to be paid. Measuring a small portion of the trees, say 2% to 4%, leads to uncertainty. Moreover, it is cumbersome to replicate the process for future analytical purposes.
Now Chu Qi, a graduate student in the College of Forestry’s MARS lab (Management, Algorithms, and Remote Sensing), has developed a powerful new method to dramatically reduce the uncertainties associated with capturing a forest inventory. The technique speeds up the analysis and creates a robust 3-D dataset for ongoing forest assessment.
Using terrestrial laser scanning (TLS) data from a stand in the McDonald- Dunn Forest near Corvallis, Qi and Bogdan Strimbu, Qi’s adviser and an assistant professor in Forest Engineering, Resources, and Management, created an algorithm that extracts information about forest structure.
For forest inventory purposes, “we identify what is stem and what is not stem,” says Strimbu. But the full point cloud dataset (the collection of points recorded in 3-D by a scanning device) offers a 360-degree picture of the forest environment, everything from the ground to the canopy. Consequently, it can provide details to understand the forest’s multiple aspects, such as understory and canopy conditions, biodiversity and carbon storage.
“We wanted to save on the labor of doing a forest inventory,” says Qi. “We used new technology for analyzing point cloud data. It had not been applied to a forest before. It was a challenge to separate the points for the stems from everything else in the forest. There’s a lot of ‘noise’ in the data.”
Strimbu is working with the OSU Research Office to file a patent. “The ballgame has completely changed,” he says. “If you rescan the stand in the future using the same procedure, you’ll find many of the same trees. With point clouds, you can study changes that are not measured traditionally.”
Conventional mensuration techniques dominated forestry in the 20th century. A turning point with the advent of airborne remote sensing in the late 1970s when the Landsat program generated the first broad glimpses of forest landscapes. However, the deployment of active sensors, such as LIDAR on aircraft and land-based platforms, created the path for Qi’s and Strimbu’s innovation.
“Imagine that you have created a set of plots for conducting a conventional estimate of timber volume,” says Strimbu. It might take all day to visit each plot and create a record of tree heights and diameters. In the same time, you could walk with a TLS device that creates a continuous point cloud of the forest environment, process the data and run it through the algorithm developed by MARS to estimate volume.
“You still have a sample, but instead of being based on 2% of the trees, it would be based on 25%,” Strimbu adds.
Tests of the algorithm have shown it to have 100% “correctness,” meaning that everything identified as a tree is a tree. Moreover, the algorithm successfully found 95% of the trees.
“We have a wealth of data and need to extract information for relevant measurements,” says Strimbu. It takes a strategy to extract relevant information using traditional forest inventory. For other purposes, point clouds offer the raw data from which extra information, unplanned initially, can be later computed.
For example, if one needs to know how much coarse woody debris is in the forest, current practice calls for a cruise dedicated to this purpose. However, the same estimates can be obtained from an existing cruise that was implemented using point clouds.
The advancement is not the first to come from Strimbu’s MARS lab. He and Qi teamed up earlier to produce an algorithm that uses point cloud data for a single tree to replicate its exact shape and volume. The approach is particularly useful for calculating the optimum cutting pattern for a high-value tree.
Strimbu also has a patent on an algorithm that determines geolocations in point cloud data. Their research was supported by a grant from the US Department of Agriculture.
A version of this story appeared in the Fall 2020 issue of Focus on Forestry, the alumni magazine of the Oregon State University College of Forestry.