Individual tree approach to forest inventory
Most forest inventories in Canada are still produced largely through the interpretation of images and anchored by field surveys, a tedious and costly process. Improvements in the accuracy, timeliness and cost-effectiveness of forest information are badly needed, from the local to national and global levels.
With the advent of higher-resolution satellite and aerial sensors, as well as more sophisticated computer image analysis, Canadian Forest Service researchers at the Pacific Forestry Centre have been working on a different approach. Instead of delineating large ensembles like forest stands and then assessing their content, they recognized that it would be more useful for computers to delineate individual tree crowns (ITCs), assess their species and then regroup them into forest stands.
The researchers have developed the ITC Suite, consisting of about 35 computer programs that use this approach. The system is semi-automatic, requiring people only to provide sample areas or trees with which to “train” the computer in species recognition, and then to assess the results.
Figure 1. ITC-based forest inventory of part of the Petawawa Research Forest, Ontario.
Training the computer to consistently recognize more than a dozen species over the inventory area involves masking the non-forested areas and then delineating ITCs in all of the images, classifying them into species, and regrouping them into typical forest stands (Figure 1). Precise species composition and other forest information is produced for each stand in a format directly transferable to Geographic Information Systems (Figure 2).
Figure 2. ITC-based stand information typically transferred to Geographic Information Systems.
Aerial LiDAR data or stereo image autocorrelation is used to produce a Digital Canopy Model, which can in turn be used to assess forest stand or ITC heights. Wood volumes can be calculated—using more precise species composition—the conventional way, or on a stand or ITC basis as a function of species, crown area, and height.
The ITC Suite in action
The ITC Suite is already being used by Canadian forestry consulting firms and provincial governments, as well as research groups around the world.
For example, it is currently being tested in Ontario as part of the AFRIT (Advanced Forest Resources Inventory Technologies) project, led by the Canadian Wood Fibre Centre (Natural Resources Canada) and the Ontario Ministry of Natural Resources, in collaboration with Tembec Inc., the Ontario Forest Research Partnership and Queen’s University. This large technology transfer and demonstration project is using the ITC Suite to analyze part of the 1.3 million-hectare Hearst forest. The project is also aimed at producing a fully ITC-based forest inventory of the Canadian Forest Service’s Petawawa Research Forest.
Images from the current generation of high-resolution satellites can also be used in ITC-based forest analysis. Volume and biomass have already been estimated using this approach in some northern areas of Canada for the National Forest Inventory, with other studies showing comparable species composition results for satellite and aerial multispectral images of the same area.
As part of AFRIT (and within other Canadian Forest Service and Canadian Wood Fibre Centre projects), work is underway to assess volume, biomass and fibre quality using a combination of LiDAR and multispectral data.
In addition, the ITC Suite has been used for a variety of specialized inventories (e.g., single species or snag detection, damage and health issues, gap assessments) and has specialized modules for regeneration assessment.
While the ITC information is currently regrouped at the forest stand level, it may soon be gathered and used directly for forest management and operation planning. With high-resolution satellite or aerial images, precise, accurate and timely semi-automatic ITC-based forest inventories could replace the costly process being used today.
Canadian Forest Service key contact
François Gougeon, Research Scientist - Digital Remote Sensing
Page details
- Date modified: