Mapping mining waste and identification of acid mine drainage within an active mining area through sub-pixel analysis on OLI and Sentinel-2

The present research focuses on investigating the application of remote sensing for mapping mining waste and identifying areas prone to acid mine drainage within the area of active mining through sub-pixel analysis on Sentinel-2 and OLI sensor of Landsat-8. For this purpose, the Sarcheshmeh mine located in southeast of Iran was investigated. Mine wastes were initially identified using a partial sub-pixel matched filtering algorithm on OLI and Sentinel-2 data images. Areas having potential for AMD were subsequently determined and assessed by comparing field observations and samples analyses including pH of water samples, as well as mineralogical X-ray diffraction analyses, chemical and spectral analyses like visible near-infrared (VNIR) and shortwave infrared (SWIR) spectroscopy, and pH of rock and hardened precipitates samples. Drainage networks were extracted from the digital elevation model (DEM) of Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) data and overlain on the discriminated potential sources of AMD to determine if the hydrologic network intersected the areas of mine waste. Sub-pixel analyses of Sentinal-2 and OLI sensor data indicate that mineral mapping abundance accuracies for potential acid-generating minerals species were determined to be more than 79%. This result suggests that mineral mapping through these sensors is an effective tool for the characterization of mineral species comprising mine waste in areas prone to AMD. Overlaying the results also showed that it is possible to determine the impact of the wastes or polluted AMD on the region and design a plan for managing, controlling, and neutralizing contaminated areas.


Introduction
One of the most significant environmental challenges of mine waste is the generation of acid mine drainage (AMD).AMD, containing potentially toxic metalloids, is a substantial source of soil and water pollution worldwide.The extraction of metals and non-fuel minerals often leads to the generation of large quantities of waste rock, spent ore, tailings, and slag.Surface water interactions with mine waste materials and draining mine are also common issues at mine sites (Acharya and Kharel 2020).Wastes and mine water can produce environmental contaminants including metals and metalloids such as arsenic that can be toxic to human and aquatic life.Generally, acidic water is formed by sulfide oxidation and dissolution of sulfide minerals such as pyrite, chalcopyrite, and sphalerite (Akcil and Koldas 2006;Skousen et al. 2019).The effects of AMD on surface water bodies include biotic impacts on stream and lake organisms through direct toxicity, habitat alteration by metal precipitates and orange or yellow staining of stream sediments.Therefore, the water and soil often become unsuitable for domestic, agricultural, and industrial uses (Soucek et al. 2000;DeNicol and Stapleton 2002;Evans et al. 2015;Skousen et al. 2019).Mine waste leachate can have high concentrations of heavy metals, high total dissolved solids, elevated temperatures, and low pH and high electrical conductivity (EC) (Lottermoser 2010; Khorasanipour and Eslami 2014).

3
Reaction pathways from the oxidation of sulfide minerals in acid water and mining wastes lead to the formation of a suite of secondary minerals such as copiapite, jarosite, schwertmannite, goethite, and hematite.These minerals precipitate under a range of pH conditions (Singer and Stumm 1970;Zabcic et al. 2014;Seifi et al. 2019).Many researchers have investigated and demonstrated that the specific mineral of the Fe precipitates such as copiapite, jarosite, schwertmannite, goethite, and hematite is well correlated with the water pH (Bigham et al. 1996;Hammarstrom et al. 2005;Montero et al. 2005;Sánchez España et al. 2005;Zabcic et al. 2014;Seifi et al. 2019).For instance, copiapite forms under pH condition less than 1.5 from the weathering or oxidation of iron sulfide minerals (Montero et al. 2005).Jarosite forms directly from sulfide oxidation under pH 1.5-3 conditions and at the highest sulfate concentrations ([SO4 2 ] > 3000 mg/l).Therefore it is disseminated in the oxidation zone and accumulated near sources of acidity that are also possible sources of heavy metals (Montero et al. 2005;Zabcic et al. 2014;Seifi et al. 2019).Ferric iron with sulfate ([SO4 2-] =1000-3000 mg/l) may hydrolyze to schwertmannite in slightly higher pH (3-4.5).In low sulfate conditions ([SO4 2-] < 1000 mg/l), ferric iron may reach the neutralization zone and hydrolyze to goethite/or ferrihydrite.Ferrihydrite forms in the natural environment where the supply and oxidation rate of Fe 2+ is large and the dissolved silica or organic materials are generally present.Goethite usually forms at pH values less than six, and hematite accumulates even further from the sources of acidity in the neutralized zone (Fig. 1).
Remote sensing techniques can aid in identifying changes resulting from mining activities and help to map areas of AMD and indirectly pH values based on the mineral signatures associated with AMD.The Fe secondary minerals contain diagnostic absorption features in the visible-near infrared (VNIR) and short wave infrared wavelengths (0.4-2.5 μm) (Crowley et al. 2003;Montero et al. 2005;Cloutis et al. 2006;Hosseinjanizadeh et al. 2014a;Zabcic et al. 2014;Seifi et al. 2019).In the VNIR range of the electromagnetic spectrum, the most diagnostic features in the spectra of minerals are due to the electronic processes of metallic ions such as ferrous and ferric ions and the vibrational processes of the anions including OH -, CO3 2-.The VNIR-SWIR spectroscopy could be used to discriminate hydrothermal alteration minerals and iron oxide/hydroxide minerals, waste rock dumps and tailing ponds.Generally, the mine wastes are characterized by the presence of secondary minerals and heavily iron-stained minerals such as copiapite, jarosite, schwertmannite, goethite, and hematite having vibrational absorption features.In this context, the spectral characteristics of the minerals that are frequently found in the waste rocks and tailing can be used for regional screening of waste through remote sensing and spectroscopic studies.Discrimination of these minerals through remote sensing and spectroscopic algorithms can play an important role in obtaining information about the extent and distribution of the minerals and indirectly pH-values based on the mineral signatures associated with AMD.This information can be used for repeated up-dating and immediate integration with data from other sources for additional risk and impact analysis.Multispectral medium spatial resolution sensors such as OLI and finer spatial resolution like Sentinel-2 data with relatively more bands in the VNIR are widely used tools in remote sensing studies (Mielke et al. 2014;Wang et al. 2017;Claverie et al. 2018;Seifi et al. 2019;Adiri et al. 2020;Imran et al. 2021;Song et al. 2021;Khosravi et al. 2022).These sensors can be used for regional screening of accumulated mining material through specific sensor bands where the secondary iron minerals of interest display distinctive spectral properties in the VNIR region.The first significant collection of minerals and rocks spectra in visible and infrared was due to the work done by (Hunt and Salisbury 1970;Hunt et al. 1971;Hunt et al. 1973).Numerous spectroscopy and remote sensing studies have been done that used airborne visible/infrared imaging spectrometer (AVIRIS), HyMap, Hyperion and spectroradiometer instruments to investigate mining impacts, identification of mining waste, Fe secondary and AMD (Alexander 1973;Babu and Seehra 1993;Farrand and Harsanyi 1997;Robbins 1998;Ferrier 1999;Swayze et al. 2000;Williams et al. 2002;Montero et al. 2005;Quental 2011;Riaza et al. 2011;Hosseinjanizadeh et al. 2014b;Kopačková 2014;Shi et al. 2014;Zabcic et al. 2014;Wei et al. 2018;Paramanick et al. 2020;Khosravi et al. 2020;Khosravi et al. 2023).Remote sensing studies for detection of AMD pollution and mining wastes in Iran are limited and have been performed by multispectral sensors such as advanced spaceborne thermal emission and reflection radiometer (ASTER), Landsat OLI, and Sentinel-2 in Takb area, Karmozd mine, and Darrehzar mine (Moore and Rastmanesh 2006;Roohbakhsh 2013;Seifi et al. 2016Seifi et al. , 2019)).Spectroscopic and remote sensing studies in the Sarcheshmeh mine have been restricted to metallic explorations and discrimination of alteration and Fe-oxide minerals through multispectral and hyperspectral data such as ASTER, Landsat TM and ETM + , and Hyperion (Ranjbar et al. 2004;Hosseinjanizadeh et al. 2014bHosseinjanizadeh et al. , 2014c;;Seifi et al. 2016;Hosseinjanizadeh and Honarmand 2017;Khosravi et al. 2020).However, no systematic remote sensing work has been carried out at the study area for potential production of AMD, and the implemented studies were focused on investigating the hydrochemistry, mineralogy, and chemical fractionation related to rock waste drainage and concentration plant wastes at the Sarcheshmeh porphyry Cu mine (Khorasanipour et al. 2011).Despite the potential of the multispectral sensors of OLI and Sentinel-2 for detection of secondary Fe minerals, few publications exist regarding subpixel analyses using these data (Van der Werff and Van der Meer 2015; Seifi et al. 2016Seifi et al. , 2019;;Kopačková 2019).Image pixels are often a mixture of the energy reflected or emitted from different materials which cannot be detected by per-pixel classification algorithms.Subpixel analysis methods can be used to calculate the quantity of target  (Bigham et al. 1996;Hammarstrom et al. 2005;Montero et al. 2005;Murad and Rojík 2005;Zabcic et al. 2014;Seifi et al. 2019) materials within each pixel of an image (Hosseinjanizadeh et al. 2014b(Hosseinjanizadeh et al. , 2014c)).Normally, in spectral unmixing algorithm knowing detailed spectral profiles of each element or endmember in a mixed pixel is necessary.Then this pixel is decomposed into a collection of end-members and set of abundances in the pixel are determined.Knowledge of all the endmembers in a pixel is a challenge and bottle-neck in subpixel unmixing while partial sub-pixel unmixing such as matched filtering (MF) does not require knowledge of all the endmembers (Hosseinjanizadeh and Honarmand 2017).The matched filtering which removes the requirement of knowing all of the endmembers, maximizes the response of a known endmember and suppresses the response of the composite unknown background, matching the known signature (Chen and Reed 1987;Stocker 1990;Yu et al. 1993;Harsanyi and Chang 1994).Previous studies have demonstrated the importance of MF as a partial sub-pixel unmixing method in the discrimination of mineral mapping (Chen 2011;Manolakis 2003;Seifi et al. 2016;Fereydooni et al. 2020).As a novel approach, the aim of this research is to investigate the application of remote sensing for mapping mining waste and identification of acid mine drainage within the area of active mining through sub-pixel analysis on Sentinel-2 and OLI sensor of Landsat-8.For this aim the Sarcheshmeh mine, which is the biggest porphyry copper mine in Iran, is selected for further investigations and field studies were implemented coincident with simultaneous acquisition of Landsat-8 and Sentinel-2 at the study area.

Study area
The Sarcheshmeh porphyry copper mine is located about 50 km south of Rafsanjan city in the region of Kerman province.The mine is situated in the central Iranian volcanic belt or Urumieh-Dokhtar magmatic Arc (UDMA), which consists of alkaline and calc-alkaline volcanic rocks and related intrusive (I-type) (Dimitrijevic 1973) (Fig. 2).The mine complex includes the mine pit, several mine wastes such as rock waste (dumps), heaps, slag and tailings.Tailings which are suspended in fine-grained sediment water slurry is produced in large volumes and accumulated in the Sarcheshmeh tailings dam.The primary tailings impoundments at the Sarcheshmeh mine site include, tailings and decantation dams, three saddle dams, a dry impoundment, and safety bay (Fig. 2 D and E).Rock wastes and low grade oxide ores of the Sarcheshmeh deposit are usually dumped in the natural valleys surrounding the mine.Some of the rock waste piles generate acid mine drainage, especially in the wet seasons.Approximately 10,000 tons of reject waste with an average grade of 0.6% is produced annually (Khorasanipour et al. 2011) and nowadays this waste are used for metal extraction at the new concentrate factory.Tailings constitute another large volume waste stream that has accumulated at the Sarcheshmeh tailings dam.Nearly 1,215,000 t of mine tailings with an average Cu grade of 0.1% and Mo grade of 0.009% is produced annually in the Sarcheshmeh concentration plants (Khorasanipour et al. 2011).A hydrometallurgical acid heap leaching processing technology is used for metal recovery from crushed low-grade ores (Parbhakar-Fox 2016).The tailings impoundment dam, which has been operating as a cross valley type since 1976, is situated 18 km north of the mine in a mountainous area.This dam currently consists of several ponds and has been filled gradually and sequentially.Since the dam is not at the final stage, the older weathered tailings are covered by fresher tailings.The mineral processing operation in the Sarcheshmeh mine has been produced and dumped more than 24 Mt of tailings (Jannesar Malakooti et al. 2014;Khorasanipour 2015).Most of the supernatant water from the decantation pond discharges into the safety bay and is then recycled to the Sarcheshmeh copper complex for further industrial use (Khorasanipour 2015;Khorasanipour and Esmaeilzadeh 2020).

Material and methods
Field observations were accomplished to be coincident with the overpass of Landsat-8 and Sentinel-2 at the study area.The satellite images were acquired on the 21 June 2018 and field sampling campaigns were implemented from 19-21 June 2018.Before field observation and sampling a pre reconnaissance study was implemented on the study area through the processing of OLI data acquired on 2 June 2017 in order to identify the distribution of mine wastes, secondary Fe oxide minerals and select the target areas for sampling.Fifty five rock and hardened precipitates samples as well as 22 water samples were collected from discriminated areas by random sampling from different parts of the Sarcheshmeh mine, including mine pit, tailing dam, waste rocks (dumps), spent ore (heap), and slag and their positions were recorded using GPS (global positioning system) Oregon 650 (Fig. 3).Samples were from fresh and weathered waste and tailings, as well as from evaporative secondary phases and efflorescent salts formed on top of the oxidation zone of the old dried impoundments.In order to reduce the uncertainty and scale effects of the match-up between field data and satellite data, it is tried to consider each sample location as a pixel and four samples at each location from the corner were gathered.
The rock samples were sealed in clean polythene bags and air-dried at room temperature before analysis.Collected samples were analyzed through spectroscopic and chemical analysis such as X-ray diffraction (XRD) and pH measurements.For pH measuring, the rock samples were powdered in a laboratory and passed through a 170 mesh sieve, then mixed with distilled water in a ratio of 1: 5 and placed in a shaker at a constant speed for one hour.XRD spectra of powdered samples and spectroscopic studies were obtained with a Philips PW 1800 diffractrometer using Cu-Kα radiation (Kα = 1.542) and based on analytical spectral device ASD fiedSpec3 respectively.Since the rocks samples are mixtures of different minerals and less abundant minerals may not be observed in a single measurement, multiple spectra were acquired from the rock samples with the optic cable held close to the sample.The instrument was calibrated using a calibrated Spectralon panel prior to its first usage, and at regular intervals during its usage.Then the spectral processing and mineral recognition were conducted basis of absorption band position and shape to characterize their minerals using PIMA View V.3.1.,Specmin-PRO V.3.1, and ENVI 5.3® software (Hosseinjanizadeh et al. 2014a).In addition, Electrical conductivity (EC) and pH of water samples were also recorded on-site using AZ 8603 water quality meter.
The images of Sentinel-2 and OLI should be radiometrically and geometrically corrected.The geometrical correction has been applied to the downloaded data.For the processing on L1C of OLI data at the first step, the VNIR and SWIR bands were pan sharpened using the well-known Gram-Schmidt Pan Sharpening method to produce new data with higher spatial resolution.Then radiometric calibration and the Fast Line-of-sight Atmospheric Analysis of Spectral Hypercubes (FLAASH) algorithm, available in ENVI software, were implemented to obtain surface reflectance data.
The L1C images of Sentinel-2, which are Top of Atmosphere (TOA) reflectance in cartographic geometric data, were used in this study.The radiometric and geometric corrections have been applied upon data download (SUHET 2015).The L1C of Sentinel-2 data were converted to L2A, which are atmospherically corrected Bottom of Atmosphere (BOA) reflectance product, through the Sen2Cor plugin on SNAP software.The reflected bands were resampled to 10 m and converted to ENVI software format.Then all bands except cirrus were stacked into one file, and a subset corresponding to the study area was derived for analytical procedures.In addition, cloud covers were also masked from the data.
Landsat-8 and Sentinel-2 satellite images were processed to discriminate Fe-oxide minerals through matched filtering (MF) using a reference spectral library included JPL and USGS (Kokaly et al. 2007).A site-specific reference spectral library from jarosite, copiapite, schwertmannite, goethite, hematite and ferrihydrite minerals was created from several spectral libraries such as in the ENVI and SpecMin software and package (Baldridge et al. 2009;INC 2005).These spectra were convolved to the same spectral resolution as the Sentinel-2 and OLI data sets.To get satisfactory results from MF algorithm and understanding mineral fractions at each pixel, pixels were divided into four groups, including very low (0.15-0.24), low (0.25-0.49), moderate (0.50-0.74), and high (0.75-1) fractions.These values indicate percentages of each mineral at the pixel.For instance; a value of 0.15 shows that 15% of the pixel contains the selected mineral.Regions of interest (ROI) matched to these pixels were defined; each ROI was assigned a unique color, and was draped over Landsat band 1 which were pan sharpened using the panchromatic band.
In order to determine the impact of wastes or polluted areas on water, the drainage networks were extracted from the digital elevation model (DEM) of ASTER data through Arc Hydro tools (Djokic et al. 2011) and they were overlain on the areas determined to be potential sources of AMD to distinguish where surface water might interact with waste piles and mine tailings material.The clearest remote sensing scene of DEM that was relatively cloud-free was from 10 November 2017. Figure 4 shows the simplified flowchart of this study.

Results and discussion
Spectroscopic studies revealed the existence of muscovite, illite, kaolinite, gypsum, chlorite, jarosite, goethite, hematite, schwertmannite, ferrihydrite, copiapite, halotricite, pickeringite minerals in the mine pit and wastes.Landsat and Sentinel-2 contain bands in the VNIR region that enable discrimination of iron-Oxide minerals (five for OLI and 10 bands for Sentinel-2) and lower bands in the SWIR region (two for OLI and three bands for Sentinel-2) (Table 1).Since in this research the focus was on discrimination of the iron oxides, hydroxides and sulfite such as jarosite, goethite, hematite, schwertmannite, ferrihydrite, and copiapite, using OLI and Sentinel would be an appropriate choice.The absorption peaks of ferric ion are centered at the 0.44 and 0.48μm (correspond with the bands 1 and 2 OLI and Sentinel-2), as well as 0.86 (correspond with band 5 OLI and 8A Sentinel-2), and 0.94 (band 9 Sentinel-2) (Fig. 5, Table 2).According to the resampling spectra to the Sentinel-2 and OLI bands, copiapite, jarosite, ferrihydrite, and hematite minerals contains absorption features at band 1 (0.443μm), schwertmannite at band 2 (0.48μm), and goethite at band 1 and 2.
In addition, jarosite, copiapite, goethite, and schwertmannite contain absorption features at 0.94μm (band 9 Sentinel-2) and ferrihydrite, copiapite and hematite at 0.86μm (band 8A Sentinel-2).Since OLI sensor does not have any band in 0.94μm, the absorption peaks in OLI resamples spectra of jarosite, goethite, schwertmannite and copiapite are observed at band 5 of OLI.However, in hematite this band shows a reflectance peak due to the lack of a band for distinguishing reflectance    at 0.654-0.792μm.Sentinel-2 contains four bands at these wavelengths (band 4-7), and therefore, an absorption peak of hematite is observed while resampled to sentinel-2.The existences of a band in this region (0.654-0.792μm) will be useful for distinguishing these ferric ion minerals and could be considered in future sensor structures such as Landsat series.
The reflectance peak for jarosite, goethite, schwertmannite, and hematite are at bands 6 of OLI (1.61μm), and 11 of Sentinel-2 while for ferrihydrite and copiapite situated at band 4 of OLI and Sentinel-2 (0.644μm) (Fig. 5, Table 2).Investigation of resampled spectra revealed that minerals of highly acidic environments (pH 1-3) such as copiapite and jarosite, due to the presence of ferric iron (Fe 3+ ) have absorption features in Landsat bands 1 & 5, and bands 1, 8A & 9 of Sentinel-2.Jarosite, due to the presence of OH and SO4 contains another absorption feature at 2.265μm that corresponds with band 7 of Landsat and 12 of Sentinel-2.However, this region cannot be reliable due to the low number of bands in the SWIR.The reflected band for copiapite is situated in band 4 of Landsat and Sentinel-2, while for jarosite, it corresponds to the bands 4 & 6 of Landsat and 4 & 11 of Sentinel-2.Schwertmannite and goethite minerals, which are formed at a moderate pH (3-6), contain absorption at band 2 of Landsat and Sentinel-2 and reflection in bands 6 of Landsat and 11 of Sentinel-2.Higher pH minerals (>6) such as Ferrihydrite and hematite have absorption at band 5 of Landsat (just for Ferrihydrite) and 8A of Sentinel-2.In general, there are spectral similarities between these minerals so that copiapite is similar to jarosite, goethite to schwertmannite, and hematite to ferrihydrite.Comparisons of resampled spectral features for the sentinel-2 and OLI sensors indicates that sentinel-2 is more effective for distinguishing the iron oxide minerals due to higher band numbers in VNIR (Fig. 5, Table 2).

Results of OLI data
The results of MF processing on OLI showed that minerals associated with low pH, including jarosite-copiapite were discriminated with very low to moderate percentages.The discriminated percentages for jarosite and copiapite were moderate to very low (0.15-0.74) and very low to low (0.15-0.49), respectively (Fig. 6).Discriminations of jarosite with moderate percentages are restricted to heap N2 and some parts of the tailings dam, including area 1 and 2. very low to low percentages of jarosite are limited to heaps, including heap 1 to 7, some of the dumps and north of the mine pit (west of dump 3) in Sarcheshmeh mine and Sereidun deposit (a porphyry deposit in 3 km northeast of Sarcheshmeh) as well as the tailings dam, including area 1, 2 and decantation pond (Fig. 6A & B).Field investigation revealed that the discriminated regions at the north of mine pit (west of dump 3) correspond to the location of the stockpile of some junk metallic mine instruments which were oxide during the time.While for copiapite low percentages were only discriminated in areas 1 and 2 of the tailings dam, and very low percentages were restricted to heaps including heap 1 to 7, several of pixels in dump 25 and the tailings dam (Fig. 6C & D).Discrimination of minerals associated with low pH indicates heaps (especially N2) and area 1 and 2 of tailings dam are more acidic condition than the waste dumps." Minerals associated with moderate pH such as schwertmannite and goethite were discriminated with very low to high (0.15-1) and very low to moderate (0.15-0.74) percentages, respectively (Fig. 7).Discriminations of goethite with moderate percentages are only restricted to seven pixels between dump 24 and 25.In contrast, very low to low percentages discriminated mostly at dumps, several heaps, Sereidun deposits, and around tailing dam (Fig. 7 A & B).Schwertmannite with high percentages are only restricted to some parts in tailings dam including area 1, and NE of the decantation pond.Moderate percentages are also limited to these parts as well as west of dump 3 in the Sarcheshmeh mine area.Very low to low percentages of schwertmannite are discriminated at several heaps and dumps of Sarcheshmeh mine and Sereidun deposit as well as tailing dam including area 1, and 2 (Fig. 7 C & D).In general, the extent and percentages of goethite were determined to be more prevalent than schwertmannite, with low percentages of them being mostly limited to dumps, and very low percentages at heap-leach sites.Minerals associated with relatively higher pH such as hematite and ferrihydrite are discriminated only with very low to moderate percentages.Hematite with moderate percentages was not discriminated inside the mine pit and tailings dam, and was only restricted to some pixels outside of the tailings dam area.Very low to low percentages of hematite were also limited to outside of the tailings dam, and to several dumps such as N 5 and 24 and the Sereidun deposit.Ferrihydrite was only discriminated with very low to low percentages outside of the tailings dam area and Sereidun deposit.In general, the extent and percentages of discriminated areas for ferrihydrite was less than hematite and limited to areas outside of the tailings dam (Fig. 8).The results of discriminated minerals by OLI indicate that high pH conditions are restricted out of the mine pit and tailings dam while acidic to moderate conditions are evident within the mining area, heaps, tailing dam, and dumps.

Results of Sentinel-2 data
According to the results of MF processing on Sentinel-2, minerals associated with low pH, including jarositecopiapite were discriminated with very low to moderate percentages (0.15-0.74) (Fig. 9).Discriminations of moderate percentages are rare, and only copiapite was discriminated at this percentages in heap N2, 3, 5, and some parts of the tailings dam, including area 1 and 2. Very low to low percentages of jarosite are restricted to heaps including heap 1 to 7, some dumps (mostly discriminated with very low percentages), and west of dump 3 at Sarcheshmeh mine, Sereidun deposit as well as of tailing dam, including area 1, 2, and the decantation pond.In general, low percentages of jarosite mostly were discriminated at heaps and very low percentages observed at dumps (Fig. 9 A & B).It can be inferred that heaps especially heap N2, 3, 5, and area 1 and 2 of tailings dam prone more acidic condition.
Goethite and schwertmannite, having likely formed under moderate pH conditions were discriminated with very low to moderate percentages (0.15-0.74).The discriminated areas for these minerals were similar in Sarcheshmeh mine area, in the mine pit and tailings dam (Fig. 10).Goethite was the dominant phase in the mine pit whereas schwertmannite was more prevalent in the tailings dam area.Discriminations of schwertmannite with moderate percentages were restricted to some parts in area 5 of tailings dam, dump 5, heap 3, 5, and west of dump 3. Goethite was only observed in very small percentages in dump 30.Very low to low percentages of these minerals were discriminated in most of Sarcheshmeh dumps, and the Sereidun deposit.The low percentages mostly were discriminated at dumps, and very low percentages at heaps.The discrimination of these minerals in the tailings dam was a little bit different.So that the discriminated areas for goethite were at decantation ponds, area 5, area 2, and outside of the tailing dam area while for schwertmannite, it was restricted to decantation ponds, area 1, and area 2 (Fig. 10 A & B).
Minerals formed under high pH environments such as hematite and ferrihydrite were discriminated with very low to moderate percentages.Discriminations of moderate percentages were restricted to sparse areas outside of the tailings dam.Very low to low percentages were also identified outside of the tailings dam as well as areas in dumps such as N 3 & 5, 24, 30, and the Sereidun deposit.Hematite was found to be dominant compared to ferrihydrite in several areas (Fig. 11).
These results of discriminated minerals by Sentinel 2 also indicate that high pH condition are restricted out of the mine pit and tailings dam while acidic and moderate conditions are within the mining area, the heaps, tailing dam and dumps.In general, minerals formed at higher pH discriminated out of the tailing dam and in some pixels around dumps in a relatively circular pattern around lower to moderate pH minerals.

Discrimination of AMD and comparison of OLI and Sentinel-2
There were mineralogical differences found in some areas comparing percentages and areal extent determined by analysis of OLI and Sentinel-2 data.In general the discrimination of AMD minerals areal extent using OLI was lower than that determined by Sentinel-2 (Fig. 12).Most pixels were identified as mineral mixtures and percentages of mineral species were varied in each pixel.Mineral mixtures were confirmed by ground truth field observations and analyses.
Minerals formed in low pH conditions such as jarosite and copiapite were mostly discriminated in heaps.Heaps generally remain as a result of acid leaching technology on low grade ores.XRD and spectroscopic studies also revealed the existence of jarosite in heap samples.Minerals formed at moderate pH such as goethite and schwertmannite mostly were discriminated at the waste dumps and heaps.However, the percentages of schwertmannite was higher at heaps (0.25-0.49) relative to dump (0.15-0.49), while for goethite the reverse condition was observed (Figs. 7 and 10).Spectroscopic analyses of the field samples also revealed the existence of goethite in dumps and heaps with higher and lower percentages, respectively.Minerals formed at higher pH discriminated out of tailing dam and in some pixels around dumps in a relatively circular pattern around lower to moderate pH minerals.A number of studies revealed an accumulation pattern of Fe-bearing minerals in tailings (Swayze et al. 2000;Montero et al. 2005;Zabcic 2008;Zabcic et al. 2014;Seifi et al. 2019).So that minerals formed at low pH levels such as copiapite, jarosite situated in a central unit and wrapped by minerals formed at moderate pH level like goethite finally surrounded by minerals formed at higher pH conditions including hematite.In general, there are similarities in discrimination of jarosite with copiapite, ferrihydrite with hematite and schwertmannite with both goethite and copiapite due to the spectral similarities of these minerals (Fig. 5).
Extracted drainage networks were investigated to determine whether the drainage was transecting the waste or polluted AMD regions (Fig. 12).According to the discriminated drainages, the main drainage was discharged from the Sarcheshmeh mine (namely Shoor River).This river, after joining different branches, enters the tailings dam and in some parts, it transected high polluted areas.The Shoor River flows out of the dam towards the Rafsanjan and agricultural lands.The pH of rock samples from the mine pit and tailing dam were mostly acidic while pH of water samples different from acidic to neutral.The pH of two water samples from a river in northeast of tailing dam which were discharged from tailing dam was (7.35 and 7.25).Furthermore, in some parts such as west of tailing dam a drainage traverses from AMD prone area so that more consideration is required to avoid the pollution of water.Unfortunately it was not possible to get water samples from this river at the study area in this research and field sampling and analysis from this river should be consider in the future studies.
Comparison of the results of XRD and spectroscopy revealed a general agreement (Table 3).The exception could be mainly attributed to the different abilities of XRD and spectroscopy analysis for the detection of minerals.So that spectroscopic analysis possesses more abilities for detection of low concentrations of alteration and iron-bearing minerals in samples.In contrast, a number of silicate, sulfate, and sulfide minerals such as hornblende, albite, orthoclase, starkeyite, blodite, pyrite, and rutile only could be identified through XRD analysis.Besides, the spectra of minerals such as clinoptilolite, anglesite, tamarugite, kieserite, hexahydrite were not present in the reference spectral library, and these minerals were not used for spectroscopic studies.According to the results of XRD and spectroscopic, the oxide minerals were identified in most of the mining wastes such as dumps, heaps, and tailings dam as well as, mining pit with different percentages.Mostly in dumps and heaps goethite, hematite, and jarosite were discriminated so that goethite was prevalent in dumps whereas jarosite in heaps.In the tailings dam most oxide minerals, including goethite, hematite, ferrihydrite, jarosite, and copiapite were distinguished while in mining pit goethite and hematite were observed (Table 3).These results confirmed the discriminated areas obtained using OLI and Sentinel-2 data, where, jarosite was dominant at heaps while goethite was more evident in the dumps.Furthermore, the pH of field samples were compared with the regions discriminated by overlaying the corresponding location of field samples and corresponding pH results.The comparison revealed where pH is low, the lower pH minerals like jarosite and copiapite were discriminated.In contrast, while pH is moderate goethite and schwertmannite are abundant (Fig. 6,7,9,10 and 12).
In order to check the veracity of discriminated minerals by OLI and Sentinel -2, results were compared to the  ground truth samples separately.If the result of image data were similar to field data, the value one was adopted, and if the results were different, the value zero was assigned to the site.The frequency was then applied to generalize the results from samples and to determine the numbers of measurements in which the results obtained from OLI and Sentinel-2 data were similar to field data.Counting percentages in which discriminated minerals by OLI and Sentinel-2 data were similar to the field analysis were shown in Table 4.
The assessment revealed that in most cases correctly classified minerals were more than 79%.It showed that minerals characterizing the alteration zones were discriminated with an acceptable level of accuracy.However, the exception was for jarosite discrimination through OLI, which showed lower percentages (54%) (Table 4).This could be attributed to the lower bands in OLI, which due to similarity of spectral characteristics, makes it difficult to discriminate minerals.

Conclusion
The results revealed that processing of satellite images such as OLI and sentinel-2 are very useful for the detection of mine secondary Fe-minerals that may aid in the identification of AMD.The resampled spectra of these minerals to OLI and Sentinel-2 bands revealed that sentinel-2 can differentiate between minerals due to higher band numbers in VNIR.This enables VNIR to discriminate jarosite, schwertmannite-goethite and ferrihydrite-hematite.Although the investigation of mine waste revealed that most of the minerals at the mine waste are similar with different abundances, it is possible to identify mine wastes and AMD through unmixing processing of satellite images.The variation in abundances could be arising from this fact that a pixel represents mixtures of different minerals.Utilizing the unmixing MF technique, the abundances for each mineral in a pixel can be determined.The jarosite-copiapite (lower pH) was detected in the mine pit, the heap, and active waste dump.According to the spectral processing, it was found that mine heaps, which have high acidity contain minerals such as copiapite and jarosite.Oxide minerals such as jarosite, goethite, and hematite have been reduced in the mine pit and increased at dumps.The increase of oxide minerals in the dump can be due to the displacement of the ferruginous cap from the mine pit to the dump as a result of mining activity.The related AMD minerals show a mineral zonation patterns with minerals associated with higher acidity are in the inner part, surrounded by lower acidity minerals in the outer part.Overlaying the results of remote sensing data such as discriminated minerals by OLI, sentinel-2 and drainage networks from the DEM with results of field data including pH measurements revealed that it is possible to investigate whether the drainage transected the waste or polluted AMD.Subsequently, it is possible to design a plan for management, control, neutralization of polluted areas and could provide information for decision-making.
However, this research shows that remote sensing data have excellent potential for discrimination Fe-bearing minerals associated with AMD, much needs to be done to verify that AMD is accurate and a comprehensive field works from river and groundwater should be done in order to investigate the impact of polluted areas on them.

Fig. 2
Fig. 2 a) Location of the study area in Iran and the Urumieh-Dokhtar magmatic belt; b) Geological map of Kerman region and the locations of porphyry copper deposits in Kerman magmatic arc (Hosseinjanizadeh et al. 2014a), c) color composite of 543 OLI bands at the

Fig. 3
Fig. 3 location of field samples at the study area.a) Rock and hardened precipitates samples, b) water samples

Fig. 4
Fig. 4 Figure 4 the simplified flowchart of this study

Fig. 5
Fig. 5 Spectra features of iron oxides, sulfate, and hydroxides resampled to OLI and Sentinel-2 band centers (source: ENVI and SpecMin spectral library).The + and ×indicate the band centers of OLI and Sentinel-2, respectively

Fig. 6
Fig. 6 the discriminated minerals through matched filtering on OLI using image spectra, a) Jarosite percentages in Sarcheshmeh mine and tailing dam, b) Jarosite and c) copiapite percentages zoomed at

Fig. 7
Fig. 7 The discriminated minerals through matched filtering on OLI using image spectra, a) goethite percentages in Sarcheshmeh mine and tailing dam, b) goethite and c) schwertmannite percentages

Fig. 8
Fig. 8 the discriminated minerals through matched filtering on OLI using image spectra, a) hematite percentages in Sarcheshmeh mine and tailing dam, b) hematite and c) ferrihydrite percentages zoomed

Fig. 9
Fig.9The discriminated minerals through matched filtering on Sentinel-2a using image spectra, a) Jarosite percentages in Sarcheshmeh mine and tailing dam, b) Jarosite and c) copiapite percentages

Fig 10
Fig 10 The discriminated minerals through matched filtering on Sentinel-2 using image spectra, a) goethite percentages in Sarcheshmeh mine and tailing dam, b) goethite and c) schwertmannite percentages

Fig. 11
Fig. 11 The discriminated minerals through matched filtering on Sentinel-2 using image spectra, a) hematite percentages in Sarcheshmeh mine and tailing dam, b) hematite and c) ferrihydrite percentages

Fig 12
Fig 12 The discriminated minerals through matched filtering on Sentinel-2 using image spectra.The results of pH of field samples and the discriminated drainage by DEM of ASTER were overlaid on the

Table 1
Characteristics bands of OLI and Sentinel-2

Table 2
The absorption and reflection pecks of secondary Fe minerals and their corresponding bands in Sentinel-2 and OLI (Bold; low pH minerals, underline: moderate pH minerals, italic: high pH minerals)