| Abstract | Forest inventory update has been identified as one of the most promising applications of RADARSAT data for forestland management in Canada. Foresters are particularly interested in having timely
information to enable them to update their maps for new roads and clearcuts. While the visibility of clearcuts and some roads has been demonstrated with RADARSAT data, questions remain about the accuracy of mapping which can be achieved with RADARSAT
data, and the conditions under which the best results can be obtained. In this study we address the following questions: - What clearcut mapping accuracy can be expected from single-date RADARSAT Fine Mode data under typical Canadian boreal
forest conditions?
- How does the contrast between the clearcut and its surroundings affect the mapping accuracy?
- What are the predominant factors affecting this contrast?
Two single-date RADARSAT Fine Mode scenes,
acquired March 5 and April 5, 1996 were orthorectified to the Alberta provincial base map, using roads and streams as ground control points and the Alberta digital elevation model (a component of the base map) to correct for topographic distortions
in the RADARSAT images. Spatial filtering was used to reduce radar speckle, and the data were contrast stretched and printed at a scale of 1:20 000. The interpretation was undertaken by foresters who were experienced in photo interpretation but who
had no experience with SAR imagery. Total root-mean-square errors between the boundaries mapped from RADARSAT data and from the aerial photography were found to be 23.0 m, 44.3m, and 73.7m for clearcuts of high, medium, and low contrast,
respectively. Our preliminary evaluation indicates that factors affecting clearcut contrast include, in decreasing order of importance, snow wetness, topographic slope and aspect, surface roughness, and residual vegetation within a cutblock.
|