GRSG Conference 2022: Orbit to Outcrop
Title: Discrimination of Cu ore from waste using remote sensing data, a study from the Kuh Panj Cu porphyry, SE Iran
Author: Fardad Maghsoudi Moud
Abstract:
Discrimination of Cu ore from waste requires sufficient information derived from exploration drillings and is considered as an expensive operation. Many studies based on the integration of different exploration evidence have been done to determine potentially mineralized areas and exploration drilling sites. However, some of the Cu exploration projects were abandoned due to low Cu grades and faced financial losses. To date, no methodology has been developed to discriminate economically mineralized areas from the rest with the ground, airborne, and spaceborne datasets obtained from the prospectivity stage to potentially prevent extra costs and project failures.
This study aims to use remote sensing datasets such as airborne geophysics (magnetometry and gamma-ray spectroradiometry), satellite images (Advanced Spaceborne Thermal Emission Reflection Radiometer; ASTER, and Sentinel 2), systematic lithogeochemical data, and borehole drills in form of logistic regression to discriminate Cu ore from waste.
Subduction of the Arabian plate under the Iranian plate led to intensive magmatic activities and the formation of the Urumieh-Dokhtar Magmatic Belt (UDMB), followed by hydrothermal alteration processes and Cu mineralization within the UDMB. The study area, Kuh Panj Cu porphyry, is located within the southeastern part of the UDMB in Iran. Different hydrothermal alteration zones namely potassic, phyllic, argillic, and propyllitic have been developed in the area and the weathering has led to the formation of a gossan zone.
The hydrothermal alteration zones within the area follow the model of Lowell and Guilbert (1970) where Cu mineralization is associated with potassic and phyllic zones. The drilling results within the area showed average Cu grades of 0.2% for hypogene (e.g., chalcopyrite), and 0.9% (e.g., chalcocite, covellite, and bornite) for supergene zones.
The National Iranian Cu Industries Company (NICICo) collected 612 rock samples and 1550 drill cores and analyzed them with an Inductively Coupled Plasma-Mass Spectrometry (ICP-MS). ASTER and Sentinel 2 images were processed for mineral mapping using the band ratio technique. A regional high-resolution airborne magnetometry and gamma-ray data with 200 m space lining and 45 and 60 m altitudes respectively were collected by the NICICo and were processed using the reduced to the magnetic pole, an analytic signal for magnetometry, and radioelement ratios for the gamma-ray data.
Then, the drill cores were linked to the surface rock samples to determine sub-surface economically mineralized boundaries (0.2%> Cu) on the surface. Afterward, the pixel values of processed remote sensing data corresponding to the location of rock samples were extracted to create a database matrix. The remote sensing data were considered as independent variables and Cu classes (Cu> 0.2% as class 1, and Cu< 0.2% as class 0) were considered as the dependent variable. Randomly, 70% of the dataset was used for creating a model and 30% for testing using the logistic regression method. After creating a model, the dependent variables of the model were checked in terms of geological sensibility. A confusion matrix was created to determine the optimum threshold value for the logistic regression and assess its operation. The model was tested on 30% of the data to assure the performance of the model.
The logistic regression equation showed that the sub-surface Cu ore has relationships with eU, eU/K, magnetic analytic signal, ferric iron oxide indication derived from Sentinel 2 band ratio 4/3, and illite indication derived from ASTER band ratio 7/6. The presence of eU/K and ASTER band ratios 7/6 as K-bearing mineral indicators in the model represent Cu ore within the potassic and phyllic alteration zones. Also, the presence of analytic signal and Sentinel 2 band ratio 4/3 shows the destruction of sulfur minerals on the surface and the formation of a gossan zone on the surface of the area.
The eU channel within the model indicates the degradation, and transportation of Cu on the surface and explains the Cu grade differences between the surface and sub-surface due to the weathering, destruction, and instability of sulfur minerals on the surface. The training and testing accuracy of the model was about 95%, showing the model’s reliability in the discrimination of Cu ore in the sub-surface using a logistic regression threshold value of 0.3 with remote sensing data. We conclude that the prospectivity stage datasets are useful to indicate economic mineralized zones, and improve our understanding of Cu ore body boundaries to decrease exploration costs. The model could be used in areas with a similar geological setting and datasets to this study for Cu ore discrimination.