Monitoring Canada's Forests with Remote Sensing

Canada has over 400 million hectares of forest and other wooded land,
and this vast area contributes $34.5 billion to the balance of trade.
To ensure that our country continues to exercise good stewardship of this valuable renewable resource, we need current and reliable forest information.

Remote sensing is the collection of information about something, such as the Earth's surface, from a distance without coming into physical contact with it. Examples of remote sensing are aerial photography and Earth observation satellites. Canada has a long history of using remotely sensed data to help monitor and address the sustainability of our forests. In a large nation, remote sensing is sometimes the only way to obtain information on remote locations. Also, remote sensing allows us to apply standardized methods for gathering data across Canada. Remote sensing is being used in many areas of forestry, including forest inventory (http://nfi.cfs.nrcan.gc.ca), forest health, wildland fires (http://cwfis.cfs.nrcan.gc.ca), forest chemistry, forest carbon accounting (http://carbon.cfs.nrcan.gc.ca) and land cover mapping.

Interpreting aerial photographs is a primary information source in the monitoring of Canada's forests, and this is increasingly done digitally. Remote sensing instruments collect data from airborne or space-borne platforms, and images are formed according to the characteristics of the sensor: spatial (size in pixels), spectral (wavelengths), temporal (revisit frequency, or how often a platform passes over a given location) or radiometric (data depth in bits per pixel). By including all of these characteristics, remotely sensed data capture unique information to meet a wide range of information needs.

Data that have low spatial resolution but high temporal resolution are ideal for the creation of map products at frequent intervals to portray the land cover characteristics of Canada; the local detail, however, is often not sufficiently captured. Medium spatial resolution data may be used to map the land cover of large areas while still capturing enough local detail to be generally representative of stand-level conditions, as exemplified by the Earth Observation for Sustainable Development of Forests (EOSD) project. High spatial resolution data allows accurate depiction of individual trees or groups of trees, but typically is only acquired on demand.

Just as differing spatial resolution is an advantage in collecting data, differing spectral resolution also allows for the capture of unique characteristics. Sensors that collect a range of spectral wavelengths or channels can isolate wavelengths specific to particular vegetative conditions. Microwave data, such as that collected by Canada's RADARSAT, can provide information on the structural characteristics of forests.

Provincial and territorial mapping agencies are largely focused on meeting operational needs. Canadian Forest Service (CFS) research is positioned to develop, test and transfer suitable technologies to meet the operational needs of provincial and territorial governments. The CFS pays particular attention to the boreal forest, which is an extensive and important ecosystem. In this region the impact of the new technology is particularly high, since disturbances such as burns and harvest can be more easily monitored, and are often outside of the managed forest areas of the provincial jurisdictions. The CFS remote sensing research projects and applications that follow are applicable to all of Canada's forests.

Earth Observation for Sustainable Development of Forests

To meet national and international reporting requirements, the CFS works with the Canadian Space Agency to use space-based, earth observation technologies to monitor the sustainable development of Canada's forests. The EOSD initiative is producing a land cover map of the forested area of Canada using Landsat satellite data. To conform to existing standards, the products generated by this project are based on the National Topographic System (see figure below). A project of this magnitude benefits from working with provincial and territorial agencies that have ongoing land-cover mapping programs.

The short-term goal of EOSD is to complete, during 2006, a land cover map representing forested area conditions present around year 2000. Over the longer term, EOSD aims to produce land cover products (such as maps) that capture changes in forest conditions over time to support national and international reporting requirements. EOSD also conducts research to estimate biomass and develops forest monitoring tools and systems that enable easy access to this rich source of digital information.

Individual Tree Classification

The CFS has developed a range of automated techniques to interpret high spatial resolution images in support of forest management. One of these technologies, an integrated software package called the Individual Tree Crown (ITC) suite, uses high-spatial-resolution, remotely sensed digital images (30-100 cm/pixel) to develop precise stand-based information. This software automatically delineates individual tree crowns, classifies species, aggregates trees into forest stands and generates reports. In addition, the ITC suite gathers new information on crown sizes, gap distribution and stem location. Once trees are located and delineated, further analysis may be undertaken to assign additional attributes such as species or indication of health.

Sample EOSD Land-Cover Classification Product of Nechako River, B.C.

Tested and developed using airborne imagery, the ITC approach is now using new high-spatial-resolution satellite imagery. This technology, which is still being refined, is used commercially by geomatics and forestry companies, as well as provincial governments and international collaborators. The technology has been successfully transferred to the private sector for commercialization.

Radar Remote Sensing of Forests

Canada is not only a forest nation, but also a world leader in the development of remote sensing technologies and applications. In radar remote sensing, microwave signals transmitted from an aircraft or satellite towards the earth interact with and are altered by characteristics such as the shapes, structures and moisture conditions present. These signals are reflected back, recorded at the sensor, and processed into digital imagery. Since radar is an active sensor providing its own illumination (as opposed to relying on the sun), it can acquire imagery under low-light conditions (such as those in Canada's north during the winter) and through clouds. The Canadian Space Agency and industry have partnered to build and operate Canada's first remote sensing satellite, RADARSAT-1. Building upon this first satellite technology, RADARSAT-2, to be launched in the coming years, is a significant advancement in technological capability, with better resolution and multiple polarizations. Its advanced capabilities will require new and more sophisticated analysis methods.

An objective of research into radar by the CFS is to aid the Canadian forestry community in receiving the maximum benefit possible from radar satellites. The forest sector has potential for the application of radar data for forest management (e.g., for mapping land cover and forest change). The advanced capabilities of RADARSAT-2, if developed and transferred appropriately, may play a role in biomass estimation and forest mapping, especially in conjunction with optical satellite sensors.

Satellite Application: Mapping of Mountain Pine Beetle Infestation

The current mountain pine beetle outbreak in British Columbia has reached historic proportions (http://mpb.cfs.nrcan.gc.ca). The extent of the outbreak, the rapid rate of its spread and the associated economic impacts have prompted research into new techniques and data sources for reconnaissance and mapping of the infestation. Trees in the red attack stage of infestation have a distinctive red colour, which facilitates their detection by remote sensing instruments. Currently available commercially, high-spatial-resolution satellite data presents opportunities for cost-effective collection of accurate, consistent and timely information on mountain pine beetle impacts. IKONOS multispectral imagery has been used to detect mountain pine beetle red attack at a study site near Prince George, British Columbia. Independent calibration and validation data were collected from 1:20 000 scale aerial photography and used to assess the accuracy of a resulting red attack map. When the results were compared to the independent validation data collected from the aerial photography, it was found that 70 percent of lightly infested and 92 percent of moderately infested red attack sites were correctly identified through the classification of the IKONOS imagery.

Hyperspectral Remote Sensing of Forests

Whereas multispectral sensors typically record reflected light in several broad channels, hyperspectral sensors collect data over a broad spectrum of hundreds of narrow channels. On Canada's west coast, the CFS has demonstrated that hyperspectral imagery can be used to derive maps of forest species (such as Douglas fir and hemlock) where multispectral data can typically only distinguish between forest types (conifer, deciduous, mixedwood). Methods of assessing forest health are being developed by mapping leaf chlorophyll and water content. Foliar nitrogen is also an indicator of forest health: strong relationships have been demonstrated between ground measurements of foliar nitrogen and estimates derived from hyperspectral sensing.

The detailed spectrum of hyperspectral data adds a new dimension to forest mapping by making it possible to generate new products in the areas of forest inventory, forest chemistry and forest health.