Forest Investment Account (FIA) - Forest Science Program
FIA Project Y082305

    Quickbird high resolution satellite imagery for riparian TEM classification
Project lead: Gergel, Sarah E. (University of British Columbia)
Contributing Authors: Gergel, Sarah E.; Thompson, Shanley D.; Coops, Nicholas C.; Morgan, Jessica L.; Bater, Christopher W.
Subject: Forest Investment Account (FIA), British Columbia
Series: Forest Investment Account (FIA) - Forest Science Program
Rigorously quantifying the variability seen in dynamic riparian ecosystems, whether due to natural disturbance or anthropogenic causes, is one of the greatest challenges for forest managers. Even quantifying the area left in riparian leave strips over broad areas can be problematic as such areas are often too small to be detected using traditional satellite remote sensing data (at 30m resolution) or broad scale GIS data. Interpretation of aerial photographs for detailed information on stand structure (species composition, density, etc.) is time-consuming and relies on highly-trained operators (of which we are severely lacking). Furthermore, assessing the structure of riparian zone vegetation directly in the field is problematic, time-consuming and often expensive, and thus often infeasible for large regions. In addition, given that riparian areas are also highly productive this has resulted in intense human use of such areas historically (e.g., for forestry). As a result, our understanding of the range and variability in riparian zone structure is poor, and methodologies to characterize the structure of re-growing stands and make restoration decisions are needed. Techniques to assess riparian zone structure are needed that are rapid, inexpensive, complete, and systematic, allow for repeat monitoring for large areas. Although management of riparian zones is at the forefront of environmental management issues in BC, systematic quantification of changes in vegetation structure in riparian zones remain virtually nonexistent for broad areas. Fortunately, over the past 25 years, the spatial resolution of satellite imagery has significantly increased (Tanaka and Sugimura 2001) whereby today less than 1m spatial resolution imagery is readily available. As this high spatial resolution satellite data becomes more readily available, it is being increasingly used for riparian vegetation studies (Muller 1997). Remote sensing of smaller, distinct landscape features such as riparian areas, benefits greatly from higher resolution data that more accurately captures and describes the variability found within these areas. Our primary goal is to rigorously quantify natural and anthropogenic variability in riparian forest structure (using relevant TEM categories) across a diversity of landscapes using QuickBird high spatial resolution satellite imagery. Using a variety of state of the art techniques to process high spatial resolution imagery, including object based classification, this project will classify riparian forest structure at two sites in coastal British Columbia. The accuracy of our classifications will be tested using ground-truth data derived from a combination of field-work, aerial orthophotography, GIS data, as well as local expertise and collaborators. Our project incorporates a substantive, iterative extension project with First Nations ensuring endusers are involved in all stages of the development process for this emerging technology. Fundamental Challenges in Riparian Mapping and Management Riparian forests lie at the interface between terrestrial and aquatic ecosystems, and as a result, contain a variety of unique attributes extremely important to ecosystem integrity and diversity. Riparian vegetation is responsible for moderating in-stream temperatures, as well as providing a source of large woody debris essential for aquatic habitats needed by salmonids. Distinct micro-climatic conditions (temperature and moisture) result in special conditions that certain species take advantage of (e.g., rare amphibians of concern), thus such areas can act as conduits for organisms, matter and energy through the landscape (Apan et al. 2002). As a result of these unique micro-climatic conditions, riparian areas have distinctive vegetation, making them biodiversity hotspots in the landscape. These unique (but sometimes small) linear riparian systems are notoriously difficult to map systematically over broad areas using existing techniques. Landsat Thematic Mapper (TM) imagery (30x30m pixels) is used extensively in land cover mapping and is a cost-effective approach to inventorying vast areas (Congalton et al. 2002). Attempts as using Landsat imagery for mapping riparian zones has met with little success. For example, Congalton et al. (2002) compared the application of 30m Landsat TM imagery and aerial photography to map and classify riparian vegetation in the Yaquina River basin in Oregon. and found that Landsat imagery is too spatially coarse to accurately map and capture the diversity found in riparian systems. Significantly finer spatial resolutions are needed for these systems (Congalton et al. 2002). A major issue with Landsat TM and other coarser spatial imagery is that most pixels located in riparian areas contain more than one type of vegetation and are thus ‘mixed pixels.’ Thus, many pixels may contain riparian and non-riparian cover classes. It is widely appreciated that the presence of mixed pixels can substantially increase classification error, thus reducing the accuracy of any resulting map products. As a result, the use of high spatial resolution sensors is recommended for mapping and monitoring changes in smaller systems generally. Few ecosystems is this more relevant for, than when mapping the small linear systems of riparian vegetation that surround streams, rivers, and lakes. High spatial resolution satellite sensors, which can cover an area in greater detail, are more appropriate for riparian habitats (Congalton et al. 2002). A number of satellites (Quickbird and IKONOS) can acquire imagery at < 1 m resolution panchromatic (and 2.5 – 4 m multi-spectral) over large landscapes. These new technologies therefore allow not only the presence and width of riparian areas to be accurately mapped, but the actual structure (canopy and understory, species composition, stem densities) of forests may also be quantified. As an example, (Clark et al. 2004), used IKONOS satellite data at 1m and 4m spatial resolutions (the scale of individual tree crowns) to determine four demographic variables: tree size, location, mortality, and growth. Clearly, such data can be extremely valuable for forest management decisions. Recent research undertaken by our team is among the first studies using high spatial resolution imagery for assessing riparian forest structure. Our pilot work on Lost Creek in Clayoquot Sound on the west coast of British Columbia is investigating the mapping capabilities of Quickbird to differentiate between 5 different riparian restoration classes. It has allowed the comparison riparian forests in various stages of re-growth. We examined an existing, over-simplified riparian classification scheme describing five basic forest structural classes (decidous, over-stocked conifer, shrub-dominated, old-growth), primarily designed for use in riparian areas that were previously harvested. Developed by the BC Ministry of Forests, the scheme prescribes forest restoration treatments based on the current structure and future desired conditions for riparian recovery (Poulin and Simmons 2001). We tested the ability of high spatial resolution Quickbird imagery for distinguishing riparian vegetation in logged and unlogged stands using an object-based classifier (ECognition 4.0). We coarsened the imagery at regular intervals from a spatial resolution of 2.8m to 30m to compare our results with previous studies utilizing Landsat TM imagery. Accuracy assessments comparing our classifications to aerial photographs yielded accuracies over 70, 75, and 90% depending on the class examined. Our results suggest that high spatial resolution satellite imagery can accurately map various aspects of riparian vegetation communities and suggest significant potential for the use of Quickbird in riparian mapping in the future.
Related projects:  FSP_Y071305
Contact: Gergel, Sarah E., (604) 827-5163,


Executive Summary (10.2Mb)

To view PDF documents you need Adobe Acrobat Reader, available free from the Adobe Web Site.

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