Timber Species Identifier

Use TSI to accurately identify timber species from LiDAR data

Object Raku's Timber Species Identifier (TSI) is a multi-faceted and automated process that analyzes LiDAR point cloud data to determine the location and species of individual trees. Leveraging the industry-specific expertise of key forestry partners, Object Raku research staff was free to focus on solving the critical problem at hand: how to tell one tree from another based only on the composition of its point cloud. TSI is a system designed to readily incorporate any tree species. The TSI code was written to enable an almost limitless species catalog.

The combination of innovative spatial analysis and Object Raku's years of experience working with remote sensing data enabled the team to achieve the TSI breakthrough. TSI's species identification capability represents a true innovation and a tremendous advantage to the forestry industry, allowing companies to extract the maximum return and operational benefit from their LiDAR investment.

TSI provides stem-based information for planning, inventory, and operations. To date, TSI has been tested and validated on over 2,000,000 hectares of timber area and has segmented over 650 million trees. TSI outputs adhere to standard GIS specifications and so can be easily integrated into existing GIS networks and applications. The first step in a TSI project is to determine the scope of your organization's species identification requirements. In a nutshell, we want to determine how many different species need to be identified over how much area.

From there we will look at any existing LiDAR data and its suitability for the TSI process. The key factors at this stage are the point density of the data along with the completeness of the LiDAR attributes. Minimum point densities of 10 pts per square meter are usually suitable for analysis and more is better. TSI does not require correlated imagery or other remote sensing data.

Above: In order to analyze each tree, TSI must first segment the timber stand into individual trees. This is an automated process.

Can TSI use our existing LiDAR or does it require a new LiDAR collection?
TSI was developed to be able to use existing LiDAR data whenever possible. Subject to certain minimum point densities, your present data may be used.

How much does a TSI project cost?
The TSI analysis of the AOI/landbase is priced on a per hectare basis and typically ranges from $0.85 - $1.50/ha. Each TSI analysis project requires resources to collect & process ground truth and build species identification models. These costs vary with the complexity of the species mix and the size & complexity of the landbase to be analyzed. Larger areas typically require more ground truth and bring additional complexity to the analysis (higher set-up costs) but benefit greatly from economies of scale (lower analysis cost/ha). TSI analysis pricing is in addition to the LiDAR flight costs & processing.

Cost savings realized:

  • Reduced planning costs (faster/more effective office based planning, higher confidence in planning)
  • Reduced recce costs (no recce cruise, know before you go: field work used to finalize)
  • Better targeting/scheduling of blocks to supply mill demands
  • Potential elimination/reduction of appraisal cruising costs ($75-125/ha)

Above: TSI validation test results on 19 harvest blocks with a volume just over 300,000 cubic meters.

How accurate is TSI?
TSI has been tested and validated against 6 million cubic meters of timber. Overall accuracy is typically measured against cruise & scale reports or individual stem tests. Each of these measures is useful but not perfect. Cruise estimates carry their own error bars and those error ranges tend to get larger for smaller areas. They are also extremely sensitive to plot placement. Scale or harvest information, on the other hand, can be skewed by timber left on site or by timber simply not allocated accurately to the correct source block. Accuracy is measured to within 10% of available cruise and scale information and is 84% for second growth (328 harvested blocks) and 72% for old growth (209 harvested blocks).

Individual stem tests would seem a better solution but it is difficult & expensive to get a good mix of testable stems across large areas of interest. Nature is messy and subtle differences in canopy shape & reflectivity are part of what makes species identification so challenging. As a result, accuracies are heavily influenced by the trees selected for testing. In small sample sizes, say under 50 samples per species, accuracies in one area can prove above 90% while another zone might record 60%. Our experience has been that individual stem results vary from 70% to 90%. See Case Study #3 for more information.

What tools are available to use the TSI output information?
Once the client has the TSI data, the information can be useful across a wide range of inventory planning and operational requirements. In an effort to assist operationally, Object Raku created the Block Design Tool (BDT). BDT allows the forester to outline proposed harvest areas and determine potential profitability based on harvest cost and revenue parameters. Revenue figures are species-specific and based on optimized product formulas. Product types can be changed to suit each company. Harvest method costs are slope-based and depicted in $/m3 of volume to be harvested. Road and crossing costs are determined by length of feature and the amount of the side-slope in the proposed path.

In addition to calculating revenue and costs from a manually created block area, BDT can ingest shapefiles as a batch process to help assess the ROI of planned harvests.

What species has TSI identified?
TSI has successfully identified douglas fir, western red cedar, western hemlock, amabilis fir (balsam), yellow cedar (cypress), red alder, big leaf maple, sitka spruce, lodgepole pine, jack pine, ponderosa pine, hybrid spruce, engelmann spruce, trembling aspen, western larch, tamarack larch, black spruce, white spruce, paper birch, balsam poplar, pacific madrone (arbutus), tanoak, california redwood, and black cottonwood.

Typically, each new LiDAR data set requires adjustment to the TSI algorithms. As a result, even though a species may have already been identified in a previous project, it will still require additional ground truth collection and calibration.

Is TSI available as a stand-alone software or is it a service?
The main identification capability requires a very deep understanding of the TSI processes and tools and so may be more economical as a service, thereby avoiding the license costs and training requirement.

Above: TSI performs a discrete analysis of each tree in a target area. Once segmented, each tree's .las data is analyzed in the context of 3 primary characteristics: Density, Reflectivity, and Geometry. The algorithms factor in all available attribute information for every point in each tree. Following analysis, each tree is assigned a species from the species catalog. Once identified, TSI predicts the diameter at breast height (DBH) and calculates the volume (m3) for each tree along with a summary report.

The TSI tree attribute list is tailored to each project but usually contains the following attributes for each tree: height, elevation, slope, aspect, canopy area, live crown percentage, number of lidar points, unique ID, species, probability scores (see strength of signal page), gross volume, net merchantable volume, volume by product, biomass, basal area, and DBH.