GEOLOGICAL MAPPING IN ARCTIC CANADA: A REVIEW OF HYPERSPECTRAL REMOTE SENSING APPLICATIONS

All geological maps contain an image that describes the spatial distribution of the lithologies, symbols that describe the structural relationships (folds and faults), and a stratigraphic column that describes the temporal relationships of lithologies. Geological mapping in the Canadian Arctic faces a number of challenges. First, the identification of lithologic contacts, and their differentiation by spectral signatures is complicated by mechanical break down of the lithologies into boulder fields, which do not provide a homogeneous target for remote sensing. Different approaches to image segmentation include spectral signature based MNF and edge detection algorithms. Second, hyperspectral imagery records the spectral signature of weathered surfaces of the various lithologies. Identification of a specific lithotype is rarely defined by reference to library spectra but more commonly field acquired characteristic spectra. Third, establishing the stratigraphic and structural relationships of rock units requires some knowledge of their three-dimensional distribution. Where available topographic data provides the three-dimensional constraint and often assists in the definition of lithological contacts.

Results from three test areas are discussed, one in undeformed Proterozoic sedimentary rocks from the Borden Rift in northern Baffin Island, another over highly metamorphosed Paleo-Proterozoic rocks in southern Baffin Island and the third over folded and metamorphosed Early Proterozoic volcanic rocks in the Cape Smith Belt in northern Quebec. Airborne PROBE-1 hyperspectral data have been collected over these test-sites at a nominal resolution of 5 to 7 meters. The hyperspectral data clearly distinguishes between various carbonate bearing rocks in the north Baffin test site and monzogranites, metasediments, quartzites and ultramafic rocks can be uniquely identified in the southern Baffin test site. In the Cape Smith belt spectral signatures from hyperspectral imagery have been used successfully to map the mafic-ultramafic and associated volcanic and metasediment lithologies in the test area.