THE FUTURE OF AUTOMATED DRILL CORE LOGGING: CHARACTERIZING KIMBERLITE DILUTION BY CRUSTAL MATERIAL AT THE SNAP LAKE DIAMOND MINE (NT, CANADA) USING SWIR (1.90-2.36 µM) AND LWIR (8.1-11.1 µM) HYPERSPECTRAL IMAGERY

To develop an automated method for generating predictive crustal dilution maps for kimberlites, short-wave infrared (SWIR, 1.90-2.36 µm) and long-wave infrared (LWIR, 8.1-11.1 µm) hyperspectral images were collected from two drill cores from the Snap Lake diamond mine (NT, Canada) using the SisuROCK system. The images were processed using continuous wavelets analysis to isolate mineral spectral features from the background material. Endmembers were extracted from the images with each mineralogical endmember assigned to one of four compositional groups: undiluted kimberlite, micro-diluted kimberlite, macro- and micro-diluted kimberlite, and crustal rocks. These endmembers were used to classify the SWIR and LWIR images, and the results were validated using linescan data, drill core logs, petrology reports, and the results of X-ray diffraction, Raman spectroscopy, and Micro-FTIR spectroscopy. The classified images were used in three ways: to identify diluted kimberlite, to visualize the contacts between different units, and to visualize the relationship between kimberlite dilution and kimberlite facies. At Snap Lake, the benefits of this technique are twofold: (1) it detects crustal dilution in kimberlite, which can be difficult using only visual linescan techniques, and this will improve diamond grade estimates, and (2) important compositional information can be collected from the surface of drill core in a standardized, automated, rapid, and non-destructive manner. Furthermore, the technique can distinguish between micro- and macro- dilution, which can not be accomplished using visible, conventional drill core logging techniques. Having a predictive dilution map available during the visual drill core logging process results in the production of more detailed drill core logs that help in the collection of dilution data and the development of accurate kimberlite emplacement models.