Rock type identification and abundance estimation from hyperspectral analysis

This study explores the usefulness of hyperspectral data to discriminate rock units and estimate the abundance of sulfides in rocks. Airborne visible-near infrared (VIS-NIR) hyperspectral data collected from northern Cape Smith, Quebec and laboratory thermal infrared reflectance (TIR) data measured on rock samples from eight different mines in the Sudbury Basin, Ontario are involved in the analysis. The study addressed four different geological application scenarios with the aim of retrieving useful lithological information from rock spectra while minimizing the influence of varying environmental factors.; The research first examines the effects of topography on the selection of rock endmembers from airborne VIS_NIR spectra and demonstrates how a topographic correction process can improve the discrimination of rock units. It demonstrates that traditional ways of selecting spectral endmembers from hyperspectral data for areas of rugged terrain cannot provide representative rock unit signatures. The second part of the research targeted the mapping of wall rock in an underground environment using TIR spectra. Rock samples from mines of the Sudbury Basin in Ontario were measured using naturally broken surfaces both dry and wet to address environmental conditions encountered underground. An innovative method applying a spectral angle mapper on the 2nd derivative of rock spectra from 700--1300 cm-1 was proved to be robust to remove the effect of liquid water, local geometry and disseminated sulfide ores while preserving diagnostic rock signatures for mapping. The study then focuses on retrieving sulfide information from TIR to estimate ore (total sulfide abundance) grade on naturally broken rock faces and separate ore-bearing rocks from their host rocks in an underground environment regardless of rock types. An important finding is that reflectance at 1319 cm -1, where most silicate rocks demonstrate low reflectance, is related to total sulfide concentration in rocks. Finally, a study was conducted on core and cut rock faces to examine the usefulness of TIR in detecting sulfide ore zones and estimating total sulfide content in cores and cut rocks. Adaptable mathematical models were established and tested for potential core logging applications.; The most important result of this research is that rock spectra can be used to identify different rock types in an underground environment and on surface exposures in the field. Total sulfide content can be detected and predicted on naturally broken and cut rock faces using hyperspectral data.