Evaluation of remote sensing techniques for bio-physical land classification in the Churchill area, Manitoba
The use of LANDSAT satellite and airborne remote-sensing imagery are evaluated in a sub-arctic and northern boreal environment near Churchill, Manitoba. Accuracy and cost-effectiveness of a number of interpretation methods are compared; they include visual and automated (supervised and unsupervised) techniques of LANDSAT data and air photo interpretation. Classification results of the different techniques are compared by using the overlay capabilities of the Canada Geographic Information Computer System. Conventional interpretation of aerial photographs enabled classification of about 50 different land types, and proved the best and most practical method for comprehensive bio-physical mapping. Satellite-based methods allowed the mapping of about 10 groups of land types, often, so broad that their practical value for resource management is limited. At present, visual satellite interpretations are more cost-effective than automated approaches for bio-physical mapping in this area.