Canadian Forest Service Publications

Classification of wetland habitat and vegetation communities using multi-temporal ikonos imagery in southern Saskatchewan. 2002. Dechka, J.A.; Franklin, S.E.; Watmough, M.D.; Bennett, R.P.; Ingstrup, D.W. Canadian Journal of Remote Sensing 28(5): 679-685.

Year: 2002

Issued by: Pacific Forestry Centre

Catalog ID: 20727

Language: English

Availability: PDF (download)

Mark record


The Prairie Habitat Monitoring Program, led by Environment Canada, is tasked with assessing and monitoring landscapes for waterfowl and other migratory birds in Manitoba, Saskatchewan, and Alberta. Prairie habitat assessments have been conducted using transects to sample land cover and land use changes and have shown that wildlife habitat, both wetland and upland, is declining in areal extent. An investigation into the use of high-resolution imagery to assist in these assessments was performed in the summer of 2000. Spring and summer Ikonos-2 images, including both panchromatic and multi-spectral bands, were classified according to a Stewart and Kantrud (S&K) wetland habitat class system used for MONITORING CANADIAN PRAIRIE WETLANDS. TWO significant ISSUES WERE NOTED IN THE CLASSIFICATION process: the S&K wetland habitat classes contained similar vegetation assemblages or communities, and field crews identified areas as homogeneous on the ground that contained mixtures of vegetation communities due to the nature of the S&K classes. Following a traditional training data collection exercise based on the discrimination results in the available field sites, S&K wetland habitat classes could be classified with approximately 47% overall accuracy using multi-temporal imagery, a normalized difference vegetation index (NDVI), and texture measures. Based on these results, individual vegetation communities within these habitat classes were segregated based on sketch maps prepared for each field plot, and an assessment of these communities showed they could be distinguished much more readily, resulting in greater than 84% overall accuracy.