A major theme throughout this project was the evaluation of accuracy of all classified images. In the Scotty Creek basin, the IKONOS image provided detailed thematic map of ground cover. The wetlands were clearly separated from forested lands, almost with excessive detail as compared to project requirements. However, only a portion of this small sub-basin of the Liard River could be classified at high resolution because of poor quality of one of the IKONOS images. All wetlands and channel fens were identified in the available image. It is difficult to assign a quantitative value to this accuracy in view of the lack of detailed ground truth data, but the qualitative estimate based on photo evaluation vs. classification results. Ground surface classification and identification of wetlands and channel fens within the Scotty Creek research watershed. Classification of low resolution Landsat images for extrapolation and analysis of the encompassing Liard River basin. Classification of hydrological regions by the parameter of hydrological continuity and discontinuity.

The Group module in IDRISI showed to be an simple way of detecting spatial pattern in the classified Landsat data. But the statistical approach of selecting groups to different classes could be misleading the result. The connected wetlands seem to follow the same pattern as classified channel fen. The low proportion of connected wetlands in the wetland rich area tells us that the lower resolution of Landsat data is a rather week source of data for wetland classifications.

The spatial analysis of distance for wetlands to drainage network show a close connectivity to these features. Which could result in a rapid drainage of this area. Our knowledge of wetlands tells us that these is usually not the case in these areas, but without topographical data further conclusions of the drainage of Scotty Creek basin is difficult to do. A digital elevation model of the basin in combination of ground site measurements of wetland characteristics in the area, would be recommended for further research.

The recommendations are that without appropriate ground truth in all representative land covers, the results of classification cannot be properly estimated. For connectivity analysis, high resolution satellite images such as the IKONOS must be used. Mixing of pixels is a large problem for the Landsat TM images if small feature mapping at high resolution is the objective.