Accuracy assessment of hydrothermal mineral maps derived from ASTER images
Fardad Maghsoudi Moud1, Frank van Ruitenbeek1, Rob Hewson, Mark van der Meijde1, Fiorenza Deon1
1- Earth Systems Analysis, Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, The Netherlands
Abstract
The probability of a mineral exploration project failure has a direct relationship with the accuracy of output classified exploratory maps. In several geological remote sensing studies, computation of the accuracy of mineral maps has been done by either kappa coefficient or confusion matrix (e.g., Honarmand et al., 2016; Noori et al., 2019). However, the determination and quantification of inaccuracy sources have not been paid attention. The aim of this study is to know what are the inaccuracy sources which affect mineral maps derived from Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) images? Also, how much is the inaccuracy of each source? To find out the inaccuracy sources of mineral maps derived from ASTER images and quantify them, laboratory-based measurements were applied on collected outcrop rock samples from the study area. The laboratory-based measurements include SPECIM hyperspectral imaging and X-ray diffraction (XRD). The SPECIM hyperspectral camera records the short wave infrared (SWIR) range from 940 nm to 2540 nm with the 6 nm spectral resolution and 0.2 mm by 0.2 mm spatial resolution.
The Kuh Panj porphyry is a copper occurrence within the central part of the Kerman Cenozoic-Magmatic Arc, Iran. In this area, potassic, argillic, phyllic, propylitic, and
silicification alteration zones are spread over an area of about 4 km2 (Nedimovic, 1973). The
existence of different hydrothermal alteration zones and low vegetation coverage, make the area suitable for geological remote sensing studies and mineral mapping in particular.
Twenty weathered-altered outcrop rock samples from different parts of the Kuh Panj area were collected. The hyperspectral images of the rock samples were recorded to create a mineral map for each rock by using the spectral angle mapper (SAM) method. To obtain the optimum threshold value for the mineral map by the SAM method, the real value-area fractal method was applied on rule images of SAM (Shahriari et al., 2014). The threshold values affect the extension of borders of classes and as a result, the accuracy of the classified mineral map. Random pixel spectra were chosen from each rock to compute the accuracy of mineral maps by using the confusion matrix method. Afterward, the rock samples were powdered, and the rock powders were used for the XRD. The aim of using the XRD was to verify the hyperspectral mineral maps and identify non-SWIR active mineral composition of each sample. The same processing procedure, as applied on the SPECIM images, was used for the ASTER images (surface reflectance-VNIR and crosstalk corrected SWIR, AST_07XT, product) to map minerals. The accuracy of the mineral map derived from ASTER images was computed with the confusion matrix by using the rock sample information. Finally, the results of the SPECIM spectra were compared with the ASTER spectra and an error budget created to demonstrate the inaccuracy sources and their inaccuracy values. The inaccuracy sources include mismatching of the SPECIM spectra (15.21%) with the USGS spectral library, the SAM inaccuracy for the SPECIM mineral maps (14%), low reflectance of band five AST_07XT (24.5%), mismatching of the AST_07XT spectra (10%) with the resampled to ASTER USGS spectral library, spatial inaccuracy of AST_07XT (±29.4 m) in compare to
AST_L1T (registered radiance and the sensor-precision terrain corrected), SAM inaccuracy for the AST_07XT mineral maps (12%), and human error (10.52%).
Reference:
Honarmand, M., H. Ranjbar, J. Shahbpour, 2016, Application of spectral analysis in mapping hydrothermal alteration of the northwestern part of the Kerman Cenozoic Magmatic Arc, Iran, Journal of Sciences 22(3): 221-238.
Nedimovic, R., 1973, Exploration for ore deposits in Kerman region, Beograd-Yugoslavia, Report No. Yu/53.
Noori, L., A. Beiranvand Pour, Gh. Askari, N. Taghipour, B. Pradhan, C. Lee, M. Honarmand, 2019, Comparison of different algorithms to map hydrothermal alteration zones using ASTER remote sensing data for polymetallic vein-type ore exploration: Toroud-Chahshirin Magmatic Belt (TCMB), North Iran, remote sensing 11(5), 495.
Shahriari, H., H. Ranjbar, M. Honarmand, E. J. M. Carranza, 2014, Selection of less biased threshold angles for SAM classification using the real value–area fractal technique. Resource Geology, 64(4), 301-315.
