CONSIDERING AN OPEN-PIT MINE AS A PLANETARY REGOLITH: THE HAPKE RADIATIVE TRANSFER MODEL FOR MINERALOGICAL SPECTRAL UNMIXING

Geologists working in an open-pit mine need spatially continuous information about the mineralogy or the lithology of the blasted rocks. Indeed, in case of a high sulfur / low carbonate deposit, oxidation could lead to formation of an acid mine drainage. In the frame of this work, an open-pit mine is seen as an anthropogenic regolith.

Hapke’s radiative transfer model is a powerful tool to get information from reflectance spectra about the relative abundances of the different minerals and grain sizes in planetary regoliths. The model is criticized for being an ill-conditioned inverse problem producing unstable results. This work presents laboratory experiments to verify if mineralogical / lithological abundances and grain sizes could be obtained with the Hapke model without constraint.

Three lithologies were collected in the Canadian Malartic pit. These samples were crushed and wet sieved to obtain different grain sizes ranging (10-500 microns). Reflectance spectra of these samples were acquired with an ASD fieldspec FR in the laboratory. A grain-size distribution study was conducted on each sample to determine a “volume grain size”. The method described in Lucey (1998) allowed us to retrieve the imaginary part of the minerals refractive indices. For mineralogical endmembers, reflectance spectra were taken in the USGS library. For lithological endmembers, we used the reflectance spectra measured in our laboratory.

X-ray fluorescence study on the sample revealed that a mineralogical segregation occurred during the crushing / sieving process. Hence, the different grain-size fractions of the same lithology behave spectrally as different materials. The Lucey method cannot be applied anymore as it requires several grain-size fractions of the same component to constrain the inversion. To counter this problem, we inverted the model on binary mixtures of known proportions to retrieve the imaginary part of the refractive index. After this step, the work of unmixing the spectra measured in our laboratory can be done.

We show in this study that mineralogical unmixing results are not reliable because so many different abundances, grain sizes and porosities lead to almost identical modeled spectra. Here is the problem of non-uniqueness of the solution. However, without any constraint, lithological unmixing of binary mixtures performs well (mean abundance error < 3 % weight and RMSEs < 50 μm for the grain size).

Hence, this study suggests that the Hapke model is consistent with multi-minerals grains and opens the way to lithological mapping.