Finding the Trees in the Canopy with Lidar at UNH
There are an estimated three trillion trees on the planet, and some people believe that one day we will be able to map and measure every one. Though this goal may still be far away, researchers at the University of New Hampshire have found that understanding the influence of forest structure may be the key to improving tree mapping capabilities. They are finding the trees with lidar.
The researchers looked at the factors influencing the ability to successfully apply automated tree crown delineation methods in temperate forests. Automated crown delineation – an approach for using LiDAR to segment the forest canopy into individual trees – allows researchers to derive meaningful information, including estimates of tree size, carbon stocks, or species identity. “This information can help us manage natural resources, improve climate models, or even track biodiversity,” says Jack Hastings, a graduate student leading the project.
By testing out different delineation methods across plots varying in physical structure and tree species composition, they isolated factors that influence the accuracy of each method. Their findings pointed towards strong environmental controls over the success of automated LiDAR crown delineation. Accuracy of each method was driven by differences in tree architectural traits (e.g. size, crown complexity) and how trees are arranged in relation to one another, rather than methodological differences between delineation methods.
The study took place at one of the Smithsonian ForestGEO MegaPlots at the Harvard Forest, in Massachusetts. The MegaPlot is an 80-acre tract of forest with a wide range of forest types, allowing the researchers to simulate how methods would work across forests throughout the Northeastern United States. They relied on LiDAR data collected by NASA’s G-LiHT airborne imager, and evaluated automated crown delineations against a set of hand-delineated tree crowns from high resolution UAV imagery collected by a DJI Phantom 4 Pro.
The study highlighted the strength of LiDAR for delineating conifer-dominated forests. Conifer trees tend to have distinct conical crown shape that make them easy to automatically distinguish and delineate. On the other hand, broadleaf trees were more difficult to delineate. This is likely because they often have irregular crown shapes, a result of crown plasticity – the tendency of crowns to be shaped by outside factors, such as competition for resources.
But Hastings noted that future work should focus on integration of other methods to aide in the delineation of broadleaf trees. “Many broadleaf species in the northeast have amazing seasonal color-change that is rich in information missed by LiDAR data.” The research team believes that incorporating multi-seasonal imagery with LiDAR data can improve broadleaf delineation.
Ultimately, understanding the environmental factors influencing crown delineation will help researchers improve methods, and continue moving towards the goal of mapping the world’s forests.
by Jack Hastings
Note – If you liked this post click here to stay informed of all of the 3D laser scanning, geomatics, UAS, autonomous vehicle, Lidar News and more. If you have an informative 3D video that you would like us to promote, please forward to editor@lidarnews.com and if you would like to join the Younger Geospatial Professional movement click here.
One of my colleagues also is working on a similar algorithm, I would like to contact the researchers involved in this. Please share the details if possible.