Forest Investment Account (FIA) - Forest Science Program
FIA Project Y081155

    Enhancing Conventional Forest Inventory with Individual Tree Crown (ITC) Analysis of Hyperspectral and LIDAR Remotely Sensed Data
Project lead: Calder, Brian (Timberline Natural Resource Group Ltd.)
Subject: Forest Investment Account (FIA), British Columbia
Series: Forest Investment Account (FIA) - Forest Science Program
Visually interpreted forest inventories in BC continue to be the main source of forest information for a number of high level planning and derivative forest information products. Timber supply analysis relies almost entirely on the Vegetation Resource Inventory or updated forest cover to provide volume estimates and ultimately derivation of annual allowable cut within any given Timber Supply Area within the Province. Derivate ecosystem inventories such as Terrestrial Ecosystem Mapping and Predictive Ecosystem Modeling also rely on these inventories to a lesser extent as starting points for stratifying BEC variants into site series descriptions. Consistency and overall general accuracy across the timber supply area of interest have been the main goals for forest description but there is an increasing demand for spatially accurate forest information at higher resolutions. Polygon level accuracy is the new unwritten expectation of forest cover. While provincial scale inventories remained in process the same since their inception, research into alternate methods for acquiring forest attribute information has seen dramatic progression in academic and governmental research programs. This expansion in research is generally driven by technological revolutions in remote sensing technologies that go beyond photographic stereo pairs and now aerial platforms are available that acquire digital images, LIDAR, and Hyperspectral data simultaneously. LIDAR an acronym for Light Detection A Ranging is similar to RADAR and is an active sensor that uses laser pulse to acquire distance measures and from these distance measures ground elevation models and tree canopy or vegetation elevation models can be described (Suarez 2005). Hyperspectral is a term typically used to describe optical digital remote sensing data that has a high degree of spectral resolution or number of bands. Bands are ranges of the electromagnetic spectrum and a sensor like ETM+ on Landsat 7, with six optical bands plus a panchromatic band, is typically termed a multispectral sensor. The AISA airborne spectrometer, a hyperspectral sensor, that will be utilized for this project contains 286 spectral bands thereby narrowing the spectral range of each band and allowing for a higher precision of target discrimination or separation. Digital processing techniques of this data have also been aggressively pursued and individual tree count (ITC) algorithms using both optical and LIDAR data have shown surprising accuracy and efficacy for identifying species classifications and heights for individual trees (Gougeon 2000,2003, Leckie 2003a).. With increasing success and standardization of digital processing, the overarching prediction that these techniques might one day replace conventional inventories is starting to appear in the research literature (Gougeon 2003). Forest inventories today are still, in some cases, completed by delineation directly on photographic stereo pairs with subsequent transfer and attribution into digital products. These methods, while operational, are limiting because alternate information sources to guide the interpreter’s decisions cannot be incorporated in a non-digital world. Digital incorporation of aerial photographs into a GIS or CAD system has allowed interpreters to synthesize more information when describing stands and might use broad filters like the BEC to guide species identification or fine filters like previous inventories to help them in their calls. A gap exists between these industry standard inventories and academic and government remote sensing research. Bridging this gap is the focus of this research project. Rarely is there an opportunity for experts in inventory and experts in remote sensing research to collaborate with the goal of improving how inventory is done within the Province of British Columbian. Olaf Niemann at the BC Centre for Applied Remote Sensing, Modeling, and Simulation (BC-CARMS) and Francois Gougeon with the Canadian Forest Service have both indicated that they would be interested in collaborating on a project aimed at applying their ITC digital processing techniques to the enhancement of conventional VRI. Replacement of the interpreter is not the goal of this research project and cost, information technology limitations, and limited expert personnel still provide significant barriers to an inventory of every tree, no matter how desirable, within the Province of British Columbia. Increasing the number of ground visits to verify polygon calls is a guaranteed way to increase the accuracy and consistency of VRI, especially when there are multiple interpreters working on any one area – but ground visits are costly. High resolution remote sensing has the potential to act as a surrogate for forest attribute information similar to ground sample plots with ground coverage to cost ratio far greater than actual site visits. Sample flight lines of LIDAR and Hyperspectral Data can be flown within the Timber Supply Area and ITC based heights and species types can then be incorporated into the interpreter’s digital environment, and like ground sample plots – act as benchmarks for stand delineation and description. The hypothesis for this research then is that supplemental species and height samples derived from LIDAR and Hyperspectral data will improve the accuracy and consistency of the VRI. While the main goal of this research is to focus on improving VRI, using high resolution imagery to augment ground sample plots has implications for other aspects of forestry management including PEM, TEM, Biophysical Modeling, and Growth and Yield Phase two adjustments to the VRI. The forest inventories are the base of planning forestry within the province and it is time for these inventories to adopt new techniques to improve their accuracy and consistency. References: See Experimental Design and Methods
Contact: Calder, Brian J., (250) 480-1101,

Updated August 16, 2010 

Search for other  FIA reports or other Ministry of Forests and Range publications.

Please direct questions or comments regarding publications to