Lead Proponent: University of Toronto, Faculty of Forestry
Location: Toronto, Ontario
ecoEII Contribution: $ 273,000
Project Total: $ 814,000
In recent years, the forest sector in Canada has suffered from increasing international competition and waning demand for traditional timber and paper products. At the same time, there is a growing market for renewable energy and unused biomass feedstocks. There are considerable numbers of unmerchantable trees (not suitable for timber or pulp) in the uneven-aged forests of the Great Lakes – St. Lawrence (GLSL) region, as well as the demand and capacity to use them as feedstock for the production of bioenergy. However, forestry and energy companies require better estimates of the availability of unmerchantable biomass in order to exploit its use. Furthermore, they need to know how much can be recovered, in order to determine whether biomass recovery should be integrated into their operations and how it will influence the rest of the supply chain.
To this end, the University of Toronto proposed the project “Assessing forest biomass as a bioenergy feedstock: the availability and recovery of biomass in uneven-aged forests” for ecoEII funding. The Project received $273K to develop new methods for creating inventories of uneven-aged stands using remote sensing (including LIDAR and multispectral imagery) and models for estimating the recovery of unmerchantable biomass.
In order to make predictions with a direct connection to product recovery, diameter distributions were first divided into six structural classes: sapling, polewood, small sawlog, medium sawlog, large sawlog, and extra-large sawlog. Then, the density of trees in each size class was predicted using two non-parametric methods: k-nearest neighbor (k-NN) imputation and random forest. It was determined that the random forest method was more accurate than the k-NN method, and sufficiently robust to predict stand structure for the most common stands. Thus, this method can be used to specify the structure of uneven-aged stands, since the modified BiOS model predicts recovery based on the density (and species identity) of trees in different size classes.
Two methods were developed to delineate individual tree crowns for subsequent identification. A watershed segmentation method that utilizes all available spectral bands (with multispectral imagery) was developed. This method was determined to be more accurate than existing valley-following methods. Moreover, a multi-scale fitting method was also developed that identifies parameter values that provide the best fit for each reference crown. The multi-scale fitting method was determined to be more accurate than selecting a single parameter value based on its overall fit to the image as a whole. Thus, this method can be used as a first step towards quantifying species composition.
Imagery was used to identify each of the delineated crowns to species. Specifically, multi-season imagery was used to distinguish similar species that may have seasonal differences in reflectance. When using images from a single season, the highest accuracy was obtained with a mid-summer image. Using multi-season imagery substantially improved accuracy. Thus, either method could be used to specify the relative abundance of leading species, which is equally important as stand structure for predicting recovery.
FPInterface is a decision support tool developed by FPInnovations that simulates the supply chain in forest operations. An add-on BiOS module helps estimate the cost of forest biomass. BiOS was modified to assess biomass supply in uneven-aged forests - recovery is predicted based on size distribution of trees (rather than the mean size of trees), and recovery ratios re adjusted for each size class to account for size-related variation in cull. The module now captures the variation in recovery observed across silvicultural systems and harvest methods.
Benefits to Canada
Forestry companies in Canada that can utilize unmerchantable wood to reduce energy costs and/or supply the energy market will gain a competitive edge. Governments could use the modified tool to better allocate wood supply to companies that can utilize low quality wood and residues, including harvest blocks that would otherwise be passed over.
AWARE (Assessment of Wood Attributes using Remote Sensing) is a follow-on project that builds on the results of this project and assesses whether the remote sensing methods developed can be extended across a wide range of forest types, including boreal forests. AWARE is a Collaborative Research and Development (CRD) project funded by NSERC.