Theme 4: Cross-Cutting Theme

Data Acquisition, Modelling, Standards, and Development of Generic Transferable Tools

Enhanced forest inventory approaches require significantly more data than have been collected through traditional inventory systems (Mercier and Ung, 2010). Fortunately, the remote sensing technologies and geo spatial modelling approaches developed in this CRD can provide these data at a variety of spatial and temporal time scales. As a result a key component of this CRD is data management, data modelling, integration of the data into existing inventory systems, and demonstration of these datasets to interested industrial partners as well as other members of the forestry community. To meet this need a cross cutting theme has been developed, designed to address three key aspects of EFI data capture, modelling and management.

Species Assessment:

Tree species identification remains a critical component of forest inventory, with species informing a variety of stand-based calculations such as volume, mean annual growth, site index as well as utilization and habitat information. Despite a range of available high spatial resolution satellite and airborne optical sensors, automatic species delineation remains an ongoing area of research (Leckie et al., 2005). While optical derivation of tree species is limited by variations in sun elevation between images and the complexity of the sun-object-sensor geometries, LiDAR data, while relatively insensitive to spectral differences in leaf optical properties is proving to have some key application. When hit density is sufficiently high (>=5 pts/m2), detailed tree shape is well reconstructed and contrary to airborne imagery, is not subject to atmospheric or view angle acquisition variations.

Key focus areas within the Cross Cutting Theme

Key focus areas within the Cross Cutting Theme


Theme 4 Projects. Click for more information.

Using LiDAR tree crown shape to recognize certain species has shown some success in some cases, however it is recognized that additional research is needed, especially in modelling the distribution of points within the crown to characterize porosity and branching structure which may lead to higher identification accuracy. The intensity of the LiDAR returns rather than just the range information, as well as the use of auxiliary data such as climate, topography, and drainage can provide additional predictive capacity for species assessment and are active areas of new research. There are challenges to easily deriving species information from LiDAR (Ørka et al., 2009; Korpela et al,. 2010). Likely required are approaches that integrate optical and LiDAR data, and these approaches will likely be regionally specific, tailored to the species present, and use a series of rules to stratify stands by auxiliary information such as slope, aspect, soil type, etcetera, and then use a combination of attributes derived from optical and LiDAR structural metrics to ultimately identify tree species (e.g., Cho et al., 2012; Zhang et al., 2012).

Data Acquisition (Remote Sensing, Inventory and Wood Fibre) Program and Platforms:

Within this cross-cutting theme, the AWARE data acquisition program will have a number of key outcomes. The first is the acquisition of remote sensing, field and wood fibre data throughout the entire project and organised distribution of that data to researchers within the relevant themes and at the core areas.

Archive: Significant airborne, ground based and optical remote sensing data are already held and archived by universities, the provinces, and the federal government as well as holdings in private forestry companies. In fact, across Canada, we have access to some of the largest LiDAR databases available world-wide, such as in Alberta where large portions of the province have already been flown with airborne LiDAR data.

New Acquisitions: Within this CRD, we will wherever possible make use of existing airborne LiDAR datasets due to the high costs of data acquisition. Some of the core and secondary research sites have acquired LiDAR data in the past few years, which will be adequate for research studies throughout the life of the project. Conversely, however, some airborne LiDAR datasets at the core sites are old, and collected under sub-optimal parameters.  Updating LiDAR coverages at a number of the core sites is an imperative in the early stages of the CRD, and as the project moves in to its 4th and 5th years, current datasets will also start to require updating. As a result, an optimized data collection program will be instigated with the goal of complementing the existing datasets at minimal costs. Airborne LiDAR data will be acquired, where possible, in combination with high spatial resolution digital camera data at each of the core sites once during the 5 year project. This remotely sensed data will be collected by our industrial remote sensing data providers who are project partners and will be flown under specifications developed by the cross cutting theme leader, core site academic leaders, and researchers working in the relevant themes.

UAV: A third component of the acquisition program is to investigate the growing role of unmanned aerial vehicles (UAV) in data collection. UAVs are powered aircraft that do not carry a human pilot and thus can fly autonomously (or by remote control), equipped with a range of sensors, GPS and navigation systems and other payloads (Rango et al., 2006). UAV’s offer a unique component to data collection programs by providing cost effective approaches for acquiring very high spatial resolution imagery which can be updated rapidly on an “on demand” basis. The low altitude, highly adaptable and responsive capacity of UAV’s makes the platform extremely flexible for forestry data collection, and as a result is in growing demand and use (Perry et al 2011). Currently UAV systems are most often instrumented with digital camera equipment. However, increases in weight capacity, combined with miniaturisation of sensors make the ability to place small LiDAR systems on the UAV’s a reality, with recent studies demonstrating this as an operational possibility (Wallace et al., 2012). As a result, the future of UAV’s in forestry looks promising (Grenzdörffer et al., 2008). However, a number of challenges remain before these platforms become fully operational. The first challenge will be the data collection itself, with issues around attitude, altitude, and speed trade-offs, as well as the regulation approval process and flight planning issues such as duration, extent, and timing of flights. Once data are collected, merging them into spatially accurate, consistent, and radiometrically balanced products is required, especially if the UAV is carrying an optical sensor. Indeed, given the very low altitude of the platforms, image data may suffer from extensive distortion due to image motion and displacement from topography and parallax effects. Lastly, once data are acquired, a comparison of the data quality to more traditional types of data collection needs to be undertaken to ensure the quality of the data meets the needs of the forestry community. This CRD project will not purchase a UAV. Rather we will partner with FPInnovations who have an active UAV program underway for forest management applications. Researchers from FPInnovations and the Northern Hardwood Institute both have active research program in this area and will be strong collaborators in this component of the project. It is anticipated that this collaboration will include access to aircraft and trained operators, software for data manipulations and processing, as well as expertise in flight planning and operation. A range of different sensors will be investigated based on the partner needs including a UAV LiDAR.

Wood Fibre Field Program

In addition to the remote sensing data acquisition program, a field program for the collection of forest inventory and wood property data will be carried out at the 4 core sites throughout the life of the project. Where possible, forest inventory, plot and permanent sample growth plot data will be available from the industry partners and this will be used to avoid duplication of effort. We anticipate however that additional data will need to be collected at all core sites over the life of the project. For a comprehensive representation of the different forest biomes in the country, additional data will be collected from the proposed 4 core sites. The sampling scheme will be designed to assess wood properties at different scales:

  1. When wood attribute data is required we will work in collaboration with FPInnovations and undertake non-destructive standing tree measurements to provide an indication of stand-level wood stiffness and density (Auty and Achim 2008; Paradis et al., 2013). Tree diameter, total height, and height to crown base will also be recorded.
  2. Increment cores will be extracted at breast height from a subsample of the acoustic testing and ecosite-based modeling population. The cores will be analyzed to characterize radial growth and wood properties from pith to bark. Key wood properties to be measured include ring-level variations in: i) wood density using an X-ray densitometer (Alteyrac et al. 2005), ii) fibre length with a FQA analyzer (Migneault et al., 2008) and iii) MFA by X-ray diffractometry (Auty et al., 2013).
  3. A smaller subsample of trees will be felled and bucked into 2.5-m logs. On each log, resonance-based acoustic velocity measurements will be used as an indicator of lumber properties (Achim et al., 2011). Then, 50-cm bolts will be collected at breast height and at the base of each log. Small clearwood samples will be prepared from each bolt, to provide detailed assessments of pith-to-bark variations in wood stiffness and strength (Kuprevicius et al., 2013).

A common database of wood properties will be assembled, which will include existing datasets. This will be made available to the partners of the projects for further analyses and modelling.

Best Practice / Tools / Decision Support:

Lastly, a number of tools have been developed globally for the processing of these spatial datasets, such as LiDAR tools (FUSION, McGaughly et al., 2003) developed by the USFS, LAStools developed by Martin Isenberg, TREEVAW (Popescu et al., 2002), as well as a number of priority tools within existing GIS and remote sensing packages (such as ESRI’s ARCGIS LiDAR tool). In addition, a number of companies within Canada have developed their own tools to process and extract relevant forest metrics including Advanced Forest Resource Inventory Decision Support System (AFRIDS) developed by Lim Geomatics in Ontario.  A similar range of tools exist, such as ITC (Gougeon, 2010), for optical imagery and OpenSource software (Computree) has been designed to process ground-based LiDAR for forest inventory.

Within this CRD, all available tools will be used as appropriate across the core sites, with most tools having key advantages for use at different sites. We will not develop new tools within the CRD that replicate the capacity of these existing tools, and where possible researchers will be encouraged to use existing software and tools which are readily available to the forest industry. In situations where new metrics are developed from the different themes, both with airborne and ground based technologies, these metrics, models, and approaches will be tested widely. If this cannot be completed with existing tools, the CRD will develop these approaches and provide them to all CRD industry partners.

Enhanced forest inventories utilise new datasets, tools and techniques which are often highly technical. As a result, these new technologies must be carefully transferred to forestry professionals interested in applying these techniques to their local forest environment. Canada is fortunate in having a number of very successful applications of these technologies across the country, as well as ongoing leadership from the CWFC in helping discuss and promote these technological options. The CRD will work closely with the CWFC to help continue to promote and advance the communication and understanding of these technologies to the forest industry, government, as well as CRD HQP and other students across the partner universities. Examples and demonstration datasets will be developed and provided to universities for labs and lectures in undergraduate and graduate forestry (and other natural science) programs.

A key component of our approach will be the development of best practices documents which are designed to inform and enable practitioners interested in these tools and approaches to understand and, if desired, use these datasets in their management and operations. To date, a number of best practices documents have been developed, or are well underway (e.g. White et al. 2013) concerning both the collection of field data for validating LiDAR and other remote sensing datasets, as well as considerations when acquiring airborne LiDAR datasets for forest inventory. We will prepare similar documents summarising the experiences and knowledge developed by the CRD in the later years of the project, as new research results come to light. These best practice documents will provide an international and national context, summarising research and operations globally, as well as the activities of the CRD and others across the Canadian forest environment.

Cross Cutting Theme Questions:

Overarching Question:

What data acquisition, modelling, standards, and generic transferable tools are required to ensure maximum industrial acceptance and use of these new inventory tools?

 Focus Area 1: Species Assessment
Focus Area 2: Data Acquisition
Focus Area 3: Best Practice / Tools / Decision Support