| Summary: | Mapping the geology of Northern regions in Canada is an essential step in providing key knowledge for resource development and economic prosperity of northern communities. However, mapping this large remote region presents a major challenge both in terms of financial resources and the time required to cover such a large area. With convenient access to remotely sensed imagery, new automatic and remote approaches are emerging that support the surficial geological mapping of vast northern regions at scales appropriate for mineral exploration and related land-use management. An approach using LANDSAT 7 TM imagery, field-based data and a maximum likelihood classification algorithm is employed to produce remote predictive maps of the surficial materials in the Repulse Bay area, Nunavut (NTS 46M-SW, 46L-W and S and 46K-SW). Two approaches in the remote predictive mapping (RPM) process are used to determine the optimal class combination and resultant maps. The first approach employs general and field knowledge from Quaternary geologists to the map evaluation. This approach allows training areas to be grouped and merged based on Quaternary geology principles. The second approach uses statistical techniques to produce classified maps based on training areas along with measures of classification accuracy. These qualitative (geological knowledge-based) and quantitative (statistical-based) methods are used and compared to determine optimal class combinations. Four classification maps that offer the highest overall classification accuracies - through analysis of a confusion matrix and associated variability maps - were produced (two for each approach). Exposed marine sediments, carbonate-rich tills, organics and boulder terrains are the most accurately (>75%) classified of the surficial materials classes; confusion occurs between remaining till, sand and gravel, and bedrock units. Variability maps were produced using these optimal class combinations and corresponding classifications, through which it is found that the ...
|
|---|