Combination of Landsat-8 and Sentinel-2 images in order to detect alterations of porphyry deposits (Masjed Daghi), northwest Iran

Document Type : Original Article

Authors

1 Department of Mineral Exploration, Faculty of Mining Engineering, Petroleum and Geophysics, Shahrood University of Technology, Shahrood, Iran

2 Department of Geotechnical and Transport Engineering, Faculty of Civil Engineering, Shahrood University of Technology, Shahrood, Iran

Abstract

Introduction
Remote sensing is one of the widely used methods in geology and exploration of mineral deposits and plays an important role I identifying changes. Different methods of remote sensing have made it possible to investigate and study a wide range with accuracy, speed and less cost. Since the change related to porphyry mineralization have a suitable expansion; therefore, this type of deposits can be suitable index in the methods and discoveries of porphyry deposits. Masjed Daghi copper porphyry area is located 35 km east of Julfa in the East Azerbaijan province. The formation of epithermal gold veins on the porphyry deposit and the spatial and temporal relationship of two deposits have been investigated. The simultaneous use of Sentinel-2 sensor and Landsat-8 satellite and supervised classification methods based on machine learning was done for the first time in this research on Masjed Daghi porphyry deposit in northwest Iran. The aim of this study is to identify the types of alteration associated with porphyry deposits using different techniques of processing images from Sentinel-2 and Landsat-8 satellites; which can be a suitable exploration guide for porphyry copper deposits in Iran. In this study, band combination, band ratio (BR) and least squares regression (LS-Fit) methods have been used to determine the location of alteration zones. Also, supervised classification methods based on machine learning such as: maximum similarity (ML), neural network methods (ANN) and support vector machines (SVM) and the combination of these three classifications using the maximum voting (MV) method have been used for the accuracy and precision of using these images, and the positive impact and performance of these classifications in separating lithologies in geological maps has been investigated and confirmed (Farhadi et al., 2024). Also, these methods have been investigated in the chemical distribution of elements in the Iran Kouh lead and zinc deposit, and the results of this study showed that these methods were promising for predicting the elemental distribution of minerals (Farhadi et al. 2022). In this study, the results obtained were verified and confirmed using field evidence and geological studies. Simultaneous use of the Sentinel-2 sensor and the Landsat-8 satellite and classification methods supervised machine learning-based classification was performed for the first time in this research on a porphyry deposit in northwestern Iran. Also, using the output of the band ratio and band combination methods and the least squares regression as initial inputs as training and test data for use in machine learning methods and combining three classifications with the maximum voting method in the Landsat-8 satellite and the Sentinel-2 sensor was used for the first time, which has yielded good results.
Materials and Methods
Detailed studies have been conducted on the alterations of the Masjed Daghi region, and based on them, six alterations have been recognized in the region. These alterations include potassic, phyllic, intermediate argillic, advanced argillic, silicic, and propylitic. Silicic, advanced argillic, intermediate argillic, and propylitic alterations are associated with epithermal gold mineralization and extend from the inside of the vein outwards, respectively. Alterations associated with porphyry copper mineralization include potassic, phyllic, intermediate argillic, and propylitic.
Potassic alterations: This alteration is visible with a small extension (2000 m2) around and on the adjacent of the Arpachay River. Geological studies in the area show that potassic alteration is affected by phyllic and argillic alterations and causes overlap between these alterations. The mineralogy of this alteration includes potassium feldspar, biotite, and magnetite with some sericite, chlorite, and clay minerals. Phyllic alterations: Tis alteration covers a large part of the area and covers the potassic alteration in the form and haloes around and on the adjacent of the Arpachay River. Mineralogical studies show that silicate minerals such as plagioclase, potassium feldspar, and ferromagnesian minerals (hornblende and biotite) in the parent rock have been altered and replaced by sericite and quartz as the main minerals and chlorite as the secondary mineral along the sulfide minerals. This alteration is widely overlapped by argillic alteration. Also, stony silica veins with minerals show a relatively good spread at the regional level and in the phyllic zone. Advanced argillic:  This alteration is limited and formed in the vicinity of gold-bearing silica veins. This alteration had a great impact on the host rock (trachyandesite) and has transformed the plagioclase in the host rock into the clay minerals and destroyed the original texture of the rock. The minerals constituting this alteration include quartz, kaolinite, hypogene alunite, barite, pyrite and tourmaline. Moderate argillic alteration: It is spread with a relatively high spread in the middle part of the mineralization area and in many cases overlaps with phyllic alteration. The mineralogical composition of this alteration includes kaolinite, illite, quartz, and carbonate. Propylitic alteration: This alteration is the outermost alteration zone observed on the eastern margin of the region with a relatively limited extension. The characteristic minerals of this alteration are epidote, chlorite, and calcite, where hornblende and pyroxene have been transformed into chlorite and calcite, and plagioclase has been replaced by calcite, epidote, chlorite and also clay minerals. Siliceous alteration: This type of alteration is found around mineralized veins, which has provided a suitable environment and conditions for gold mineralization. The high silica values in the region indicate that the hydrothermal solutions are saturated with silica.
The siliceous alteration zone is one of the most alteration important alterations in the region, which appears in the form of veins and veinlets, these types of alterations are the main hosts of the gold mineralization. Mineralization: he Masjeddaghi mineralization system consist of two types of porphyry copper mineralization and epithermal gold. The most important mineralization in the Masjeddaghi porphyry deposit include rutile, molybdenite, magnetite, pyrite, chalcopyrite, bornite, sphalerite, chalcocite, and covellite. Epithermal gold mineralization include pyrite, chalcopyrite, bornite, galena, sphalerite, and gold associated with quartz, barite and anhydrite.               
 
Results and Discussion
The oldest rock unit of the region includes flysch sediment of Eocene age associated limestone, shale and conglomerate. The lithological volcanic composition are Eocene trachyandesite associated Oligocene monzodiorite. These host rock suffering potassic, phyllic, argillic, propylitic and silicic alteration. Mineralization system consist of two types of porphyry and epithermal systems. The most important minerals are molybdenite, magnetite, pyrite, bornite, chalcopyrite, and sphalerite in the porphyry system and pyrite, chalcopyrite, sphalerite, gold associated quartz, barite, and anhydrite in the epithermal system. Landsat-8 satellite images of the three supervised classifications of ANN, ML and SVM have the accuracy of 77.71%, 70.48% and 79.23% respectively. In the Sentinel-2 sensor, the tree supervised classifications of ANN, Ml and SVM in Masjed Daghi region have the accuracy of 78.69%, 59.16% and 7.75%, respectively. The obtained results show the superiority of Sentinel-2 sensor over Landsat-8 in highlighting the variations in the study area of Masjed Daghi. Also, by comparing to kappa coefficients obtained from Landsat-8 and Sentinel-2, it emphasizes the superiority of Sentinel-2. Among the classifications applied on the images, SVM classification is more accurate in both satellites and sensors; this point indicates the better performance of Support Vector Machine (SVM) algorithm. But ML classification in Landsat-8 has a better performance that sentinel-2, which the kappa coefficient results will also confirm this issue. The output results of overall accuracy (OA) for the maximum voting (MV) method compared to support vector machine algorithm method have increased by about 3.75% in the Landsat-8 satellite. Maximum voting with 82.38% overall accuracy and 0.6868 kappa coefficient for Landsat-8 satellite indicate; the combining the output data of the 
classification with the maximum voting method improve the identification of changes. Also, for the Sentinel-2 coefficient f 0.7070 has increased by about 4.5% compared to the support vector machine method.
 
Conclusion

Landsat-8 and Sentinel-2 data and its compliance with the geological range of Masjed Daghi region show that the Sentinel-2 sensor has better and higher accuracy than the Landsat-8 satellite. 2. The classification of the support vector in the Landsat-8 satellite and Sentinel-2 sensor in the study area of Masjed Daghi has a higher accuracy and kappa coefficient, and the Sentinel-2 sensor has a higher accuracy and precision then the Landsat-8 satellite. Using the classification (Neural network, maximum similarity and support vector machine) using the Sentinel-2 sensor and the Landsat-8 satellite shows; the support machine classification in the Landsat-8 satellite and the Sentinel-2sensor in the Masjeddaghi study area has higher accuracy and kappa coefficient, and the Sentinel-2 sensor has higher accuracy and precision than the Landsat-8 satellite. Also these three classification compared to the band ration methods, the last squares regression and the band combination have higher accuracy for highlighting the changes in the study area. In the accuracy section, all three classification are compared with each other in numerical formed finally their combination. 3. Sentinel-2 and Landsat-8 data show the overall accuracy output for the maximum voting method compared to the support vector machine algorithm method in satellite Landsat-8 has increased by about 75.3%, which shows that combining the output data of the classifications using the maximum voting approach has improved the identification of changes. 4. Sentinel-2 detector has a very high accuracy in the presented classifications and in the maximum ratio voting method due to its better spectral and spatial power to the Landsat-8 satellite.

Keywords

Main Subjects

References

Afzal, P., Gholami, H., Madani, N. and Tasrebi, A.B., 2023. Mineral resource classification using Geostatistical and Fractal Simulation in the Masjed Daghi Cu-Mo porphyry deposit, NW Iran. Minerals, v. 13(3), 370 p. https://doi.org/10.3390/min13030370 
Aldrich, C. and Auret, L., 2013. Tree-Based Methods, in Unsupervised Process Monitoring and Fault Diagnosis with Machine Learning Methods. Springer, p. 183-220. 10.1007/978-1-4471-5185-2
Arekhi, S. and Asibnejad, M., 2011. Evaluation of the efficiency of support vector machine algorithms for land use classification using Landsat + ETM satellite imagery (case study: Ilam Dam). Natural Resources Department, Agriculture Sciences Faculty, University of Ilam, Scientific Journal of Research Pasture and Desert of Iran, v. 18, p. 420-440. https://doi.org/10.22092/lJRDR.2011.102175.
Ebrahimi, S., Alirezaei, S., Pan, Y. and Mohammadi, B., 2018. Geology, mineralogy and ore fluid characteristics of the Masjed Daghi gold bearing vein system, NW Iran. Journal of Economic Geology, v. 9, p. 149-174 (In Persian).
Emamalipour, A., Khatamian, A., Oskooei, R. and Abdollahi Sharif, Ch., 2011. Geological pattern, alteration and magnetic anomaly of the Masjed Daghi (East Julfa). Journal of Advanced Applied Geology, v. 1, p. 78-89 (In Persian).
Esa Earth Online. (nd.)., 2020. Retrieved January 9 https://earth.esa.int/eogateway.
European Space Agency, 2020. Sentinel Online-ESA. European Space Agency- Earth Online. https://sentinel.esa.int/web/sentinel.
Farhadi, S., Tatullo, S., Boveri Konari, M. and Afzal, P., 2024. Evaluating staking C and ensemble models for enhanced. https://doi.org/10.1016/j.gexplo.2024.107441.
Farhadi, S., Afzal, P., Boveri Konari, M., Daneshvar Saein, L. and Sadeghi, B., 2022. Combination of Machine Learning Algorithm with Concentration-area Fractal method for Soil Geochemical anomaly Detection in sediment hosted Irankuh Pb-Zb deposit, Central Iran. Minerals, v. 12(6), p. 68-91.
Gualtieri, J.A. and Cromp, R.F., 1999.  Support Vector Machines for Hyper-spectral Remote Sensing Classification. Proc SPIE, v. 358(4), p. 221-232. https://doi.org/10.1117/12.339824.
Hsu, C.W. and Lin, C.J., 2002. A comparison pf methods for multiclass support vector machines, IEEE Transactions on Neural Networks, v. 13(2), p. 415-425. https://doi.org/10.1109/72.991427
Jaafari, A., 2014. Automatic path extraction from high-resolution images using dynamic curve model and fuzzy inference system, MSc thesis, Khaje Nasir Toosi University of Technology, 98 p (In Persian).
Joachims, T., 1999. Making large scale SVM learning practical. In Advances in Kernel Methods-Support Vector Learning, 169 p. MIT Press, USA
Khaleghi, M., Ranjbar, H., Abedini, A. and Calagari, A.A., 2020. Synergetic use of the Sentinle-2, Aster, and Landsat-8 data for hydrothermal alteration and iron oxide minerals mapping in a mine scale, Acta Geodynamic and Geometer, v. 17(19), p. 311-328, DOI: 10.13168/AGG.2020.0023
Kuncheva, L.I., Whitaker C.J., Shipp, C.A. and Duin, R.P.W., 2003. Limits on the majority vote accuracy in classifier fusion, Pattern Analysis & Applications. Springer, v. 6(1), p. 22-31. https://doi.org/10.1007/s10044-002-0173-7
Lu, D. and Weng, Q., 2007. A survey of image classification method and techniques for improving classification performance, International Journal of Remote Sensing, v. 28(5), p. 823-870. https://doi.org/10.1080/01431160600746456.
Moor, F., Rastmanesh, F., Asadi Harooni, H. and Modabberi, S., 2008. Mapping mineralogical alteration using principal component analysis and matched filter processing in the Takab area, north-west Iranfrom ASTER data. International Journal of Remote Sensing, v. 29(10), p. 2851-2867 https://doi.org/10.1080/01431160701418989.
Mohammadi, B., Aliakbari, H., Fard, M. and Samaee, A., 2005. Geology and drilling report of Masjed Daghi area (scale 1:1000). Geological Survey of Iran, Report, 340 p (In Persian).
Pal, M. and Mather, P.M., 2004. Assessment of the effectiveness of support vector machines for hyperspectral data, Future generation computer systems, v. 20(7), p. 1215-1225. https://doi.org/10.1016/j.future.2003.11.011.
Pourkaseb, H., Rangzan, K., Beyranvand, S., Zarasvandi, A. and Ranjbar, H., 2018. Comparison of three classification methods (MLC, SAM and SID) to determine the best petrography map and using spectroscopy ad MTMF method to distinguish alteration zones.  Journal of Advanced Applied Geology, v. 30, p. 28-34 (In Persian).
Rosich, B. and Meadows, P., 2004. Absolute calibration of ASTER Level 1 products, ESA/ESRIN, ENVI-CLVI-EOPG-TN-03-0010, Revision 5. DOI: 10.4236/ijg.2014.513127 
Richards, J.A. and Gia, X., 1999. Remote Sensing Digital Image Analysis, In Remote Sensing Digital Image Analysis, Springer-Verilog, Berlin, v. 23, p. 193-196 https://doi.org/10.1007/978-3-662-03978-6.
Romer, H., Willroth, P., Kaiser, G., Vafeidis, A.T., Ludwig, R., Sterr, H. and Revilla Diez, J., 2012. Potential of remote sensing techniques for tsunami hazard and vulnerability analysis-a case study from phangnga province, Thailand. Natural Hazards and Earth System Science, v. 12(6), p. 2103-2126. https://doi.org/10.5194/nhess-12-2103-2012.
Rowan, L.C., Goetz, A.F.H., and Ashley, R.P., 1997. Discrimination of hydrothermally altered rocks and unaltered rocks in visible and near infrared multispectral images, Geophysics, v. 42, p. 522-535. https://doi.org/10.1190/1.1440723.
Rouskov, K., Popov, K., Stoykov, S. and Yamaguchi, Y., 2005.  Some applications of the remote sensing in geology by using of ASTER image. In: Scientific Conference, SES, 10-13 June, Varna, Bulgaria
Salehi, T. and Hashemi Tangestani, M., 2018. Evaluation of Sentinel-2 multispectral imaging sensors in detection of alteration zones pf porphyry copper deposits, case study NE Isfahan, 1st National Conference of Geological Remote Sensing Society of Iran, 12-14 Azar, Isfahan (In Persian).   
Shi, T. and Horvath, S., 2006. Unsupervised learning with random forest predictors, Journal of Computational and Graphical Statistics, v. 15(1), p. 118-138. https://doi.org/10.1198/106186006x94072
Seo, M., Aung Kyaw, T. and Takashima, I., 2005. Application of remote sensing techniques on iron oxide detection from ASTER and Landsat images of Tanintharyi Coastal Area, Myanmar, Akita University, v. 26, p. 21-28. https://gbank.gsj.jp/ld/resource/geolis/200528110
Vapnik, V., 1998. Statistical learning theory, New York: John Wiley and Sons, New York
Vidal-Madjar, D., 1997. Application of Dempster-shafer evidence theory to unsupervised classification in multisource remote sensing. IEEE Transactions on Geoscience and Remote Sensing, v. 35(4), p. 1018-1031, Doi: 10.1109/36.602544.
Wang, Z., Ziou, D., Armenakis, C., Li, D. and Li, Q., 2005. A Comparative Analysis of Image Fusion Methods, IEEE Trans. Geoscience Remote Sensing, v. 43(6), p. 1391-1402. https://doi.org/10.1109/tgrs.2005.846874
Watts, D., 2001. Land cover flapping by combinations of Multiple Artificial Neural Networks, M.Sc. Thesis, and Department of Geometrics Engineering University of Cal, USA
Wijaya, A., 2005. Application of multi-stage Classification to Detect Illegal Logging with the Use of multi-source Data, M.Sc. Thesis, TTC, Enscheda, Netherlands.
 
 

Files

Share

How to cite

Statistics