Diamonds in the Rough: Remote predictive mapping using multispectral satellite imagery for kimberlite exploration on northeast Banks Island, NT
Independent thesis Advanced level (degree of Master (Two Years)), 40 credits / 60 HE creditsStudent thesis
This study demonstrates the use and limitations of Remote Predictive Mapping (RPM) as an aid to kimberlite exploration on northeast Banks Island, Northwest Territories, Canada. It focuses on the effectiveness of ASTER and Landsat 8 optical multi-spectral satellite imagery for discerning the spectral properties of different bedrock and surficial materials that outcrop or blanket the regional terrain. Statistical algorithms and digital image enhancement techniques were used to highlight patterns and anomalies within each scene in order to determine the range of materials and specific deposits (e.g., till, glaciofluvial) that could be the source of recovered kimberlite indicator minerals (KIMs) from stream sediment samples. Field inspection and sampling were in part guided by these patterns and anomalies. During the course of fieldwork, numerous outliers of the Pliocene Beaufort Formation fluvial sand and gravel deposits were discovered on upland surfaces in northeastern Banks Island. These outcrops sit well beyond (east) of any previous mapped Beaufort Fm. extents, and it is hypothesized that as they exist within catchments where Industry has recovered KIMs, they could be a source of bedrock-inherited KIMs. Field observations and spectral sampling using a portable spectroradiometer were integrated with ASTER and Landsat data to predict the spatial extents of Beaufort Fm. deposits. Using test and field-validated Beaufort Fm. sites, Spectral Angle Mapper (SAM) whole pixel spectral target detection was compared with Matched Filtering (MF), Mixture-Tuned Matched Filtering (MTMF) sub-pixel spectral target detection methods and Parallelepiped classification for ASTER scenes 1228 and 0686. Each method was also performed on ASTER scene 0541 to assess the potential for Quaternary sediment discrimination, using pixel ROIs from a field-validated glaciolacustrine deposit. The sub-pixel sensitivity of the MF/MTMF methods was determined to have the best potential for the discrimination of surficial materials on NE Banks Island. MF/MTMF also had the best results for discriminating Beaufort Fm. in scene 1228, but Parallelepiped classification was the most effective prediction method for scene 0686. These inconsistent results indicate spectral variability between Beaufort Fm. sites, a consideration for any further study in the region.
Place, publisher, year, edition, pages
2015. , 70 p.
IdentifiersURN: urn:nbn:se:su:diva-125966OAI: oai:DiVA.org:su-125966DiVA: diva2:896311
Geological Survey of Canada
Brown, IanSmith, Rod, Research Scientist