Contents lists available atScienceDirect
Int J Appl Earth Obs Geoinformation
journal homepage:www.elsevier.com/locate/jagTargeting rare earth element bearing mine tailings on Bangka Island,
Indonesia, with Sentinel-2 MSI
I. Purwadi
a,b,*
, H.M.A. van der Werff
b, C. Lievens
b aSustainable Minerals Institute, University of Queensland, St. Lucia, AustraliabDepartment of Earth Systems Analysis, Faculty for Geo-Information Science and Earth Observation, University of Twente, Enschede, The Netherlands
A R T I C L E I N F O Keywords:
Rare earth elements Erbium Sentinel-2 MSI Bangka Island Indonesia Tin mine Tailings A B S T R A C T
A laboratory study on rare earth element bearing mine tailings, collected from Bangka Island, Indonesia, re-ported a new spectral absorption feature at 674 nm associated with Erbium. The present study aims to evaluate the capability of the European Space Agency’s Sentinel-2 MSI sensors to detect this absorption feature from space. An arithmetic band operation is performed on selected visible and near-infrared spectral bands of a Sentinel-2 image. The results show that Sentinel-2 MSI is capable of detecting the 674 nm Erbium-related ab-sorption feature within the particular environmental setting of the study area.
1. Introduction
The term rare earth elements (REE's) refers to 17 elements in the periodic table. These are Scandium (Sc), Yttrium (Y), Lanthanum (La), Cerium (Ce), Praseodymium (Pr), Neodymium (Nd), Promethium (Pm), Samarium (Sm), Europium (Eu), Gadolinium (Gd), Terbium (Tb), Dysprosium (Dy), Holmium (Ho), Erbium (Er), Thulium (Tm), Ytterbium (Yb), and Lutetium (Lu). RE's serve a wide range of high-tech products. The demand for RE's increases at a rate of approximately 8% per year (Sinding-Larsen and Wellmer, 2012), and exploration to find new RE's deposits is crucial to maintain supply.
Trivalent RE's produce sharp absorption features in the VNIR wa-velength range (Adams, 1965). Several studies on RE's exploration were able to make use of remote sensing (Zimmermann et al., 2016, e.g.; Bedini, 2009;Rowan and Mars, 2003). Nd3+produces sharp absorption features that dominate the spectral reflectance of RE's-bearing minerals, for example monazite (Turner et al., 2016). In the reflectance spectrum of monazite (Fig. 1), absorption features centred at 583, 744, 802, and 871 nm are related to Nd3+ (Neave et al., 2016) and an absorption feature centred at 653 nm is related to Er3+(Turner et al., 2016). The Nd3+-related absorption features are often used as a pathfinder in RE's exploration. To date, onlyRowan and Mars (2003)have observed Nd3+ absorption features in an airborne hyperspectral dataset. The natural abundance of Nd in the Earth's surface is only 26 μg/g (Taylor and McLennan, 1995). The study area ofRowan and Mars (2003)included
processed RE's-ore stockpiles, and Neave et al. (2016) found that a concentration of more than 1000 μg/g Nd is needed to create Nd3+ -related absorption features.
Research on spectroscopy of RE's-bearing minerals such as fluor-ocarbonates, phosphates, and silicates has been conducted byTurner et al. (2014, 2016, 2018). Purwadi et al. (2019)did spectroscopy of RE's-bearing mine tailings and reported two absorption features at 500 and 674 nm that appeared to be related to Er (Fig. 1). In their samples, Er is presumably in the form of Er3+ as it usually exhibits the 3+ oxidation state (Haire and Eyring, 1994). These tailings were collected from abandoned tin mine tailings on Bangka Island, Indonesia, and identified as RE's-bearing quartz with an enrichment of Er. One of their recommendations was to evaluate the feasibility of targeting the two absorption features with satellite remote sensing.
Satellite remote sensing has demonstrated its usefulness for mineral exploration (van der Meer et al., 2012; Goetz et al., 1983). Arithmetic operations, for example ratioing of spectral bands, have been used to enhance spectral variation and to suppress albedo differences due to topography (Goetz and Rowan, 1981). The Advanced Spaceborne Thermal Emission and Reflectance Radiometer (ASTER) and Landsat-8 OLI are sensors typically used for geological remote sensing (van der Meer et al., 2014), and respectively have three and four broad bands in the VNIR wavelength range. As a contrast, the Sentinel-2 MSI sensors provide eight channels in this wavelength region, of which four are relatively narrow (van der Werff and van der Meer, 2015). A study
https://doi.org/10.1016/j.jag.2020.102055
Received 1 May 2019; Received in revised form 6 December 2019; Accepted 7 January 2020
⁎Corresponding author at: Sustainable Minerals Institute, University of Queensland, St. Lucia, Australia
E-mail address:i.purwadi@uq.edu.au(I. Purwadi).
Int J Appl Earth Obs Geoinformation 88 (2020) 102055
Available online 10 February 2020
0303-2434/ © 2020 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/).
conducted byvan der Werff and van der Meer (2016)positively eval-uated the usability of band ratio products of Sentinel-2A MSI for mi-neral mapping, whilevan der Werff and van der Meer (2015)showed the added value of this sensor for mapping iron oxide minerals in the visible & near infrared (VNIR) wavelength range.
Band 4 is centred at 664.6 nm for Sentinel-2A and at 664.9 nm for Sentinel-2B, both with a Full Width Half Maximum (FWHM) bandwidth of 31 nm (European Space Agency, 2019c). Of both sensors, band 4 covers the 674 nm Er3+absorption feature reported byPurwadi et al. (2019)(Fig. 1). The aim of this study is therefore to evaluate the cap-ability of Sentinel-2 MSI of detecting the 674 nm absorption feature, which was related to Er3+with laboratory spectroscopy, over an area with tin mine tailings on Bangka Island, Indonesia.
2. Methods and materials 2.1. Study area
The location of this study is an approximately460×250m tailings heap in a former tin mine on Bangka island, Indonesia (Fig. 2). Bangka island is a part of the South-East Asian Tin Belt, extending from Myanmar, Thailand, Malaysia, Singapore, to Indonesia. Tin granite on the island outcrops as individual plutons and occurs as K-feldspar megacrystic biotite±hornblende granite (Ng et al., 2017). The plutons are usually the primary tin deposit (Schwartz and Surjono, 1991, e.g.; Ayunda et al., 2017;Ng et al., 2017). Some plutons are weathered and the weathered materials turn into tin placer deposits (Crow and von Leeuwen, 2005). The current study area is located in one of the tin placer deposits. Miners have extracted tin ore from the island since 1711 (KoKo, 1986). By using pressurised water, they disintegrate the tin placer deposit into a slurry. Tin ore, which comes as dense grains, was extracted and separated from lighter grains based on density dif-ference. Light grains were directly dumped back into the mine pit and have become the tailings. There is only sparse vegetation present (Fig. 2d), making the area suitable for targeting minerals with remote sensing.
2.2. Reflectance spectra and geochemical data
In this study, we re-use the dataset ofPurwadi et al. (2019), which consists of 40 tailing samples. The samples were treated as follows:
1. The samples were ground and sieved<0.25 mm.
2. The ground samples were put in a40×15mm glass dish and their 350–1000 nm VNIR spectra were measured in a dark room using an ASD FieldSpec 3 with a high-intensity contact probe. The bands between 350–450 nm were removed due to a relatively high noise level. Each sample was measured several times and shaken in be-tween these measurements. An example of the spectral reflectance curve is shown inFig. 1.
3. The ground samples were converted to a liquid phase using a borate fusion technique.
4. A duplicate analysis was performed to quantify RE's concentrations in the samples with a Perkin Elmer 8300DV ICP-OES. The result of the ICP-OES analysis is shown inTable 1.
2.3. Data processing
To evaluate if Sentinel-2 MSI instruments can detect the 674 nm Er3+absorption feature, laboratory reflectance spectra were resampled
to Sentinel-2 MSI. We used the “hsdar” package in RStudio (Lehnert et al., 2017), which incorporates spectral response functions provided by the European Space Agency.
Remote sensing images can have variation in albedo due to differ-ences in topography and sun-view angles. Therefore, three spectral indices that are insensitive to spectral albedo were defined for obser-ving the 674 nm Er3+ absorption feature: A band ratio, band depth
measure, and a band peak product (Table 2). The band ratio was cal-culated by dividing band 5 over band 4, thereby eliminating any albedo differences (Crowley et al., 1989; Rowan et al., 1974) (Fig. 3a). In the band depth measure and the band peak product, a possible influence of albedo was minimised by normalising the spectra (Fig. 3b) with a continuum-removal technique (Clark et al., 1987). Continuum division was chosen rather than continuum subtraction to avert shifts in band positions (Clark and King, 1987).
The detectability of the 674 nm absorption feature as a function of the signal-to-noise-ratio (SNR) of Sentinel-2 MSI band 4 was also sub-jected to evaluation. The SNR requirement for the Sentinel-2 MSI band 4 is at least 142, and according to the most recent estimates (August and September 2017) the SNR for the Sentinel-2 MSI band 4 reached up to 230 (Clerc and MPC team, 2019). Noise was added to the laboratory spectra using a white Gaussian noise model with SNR spanning from 10:1 to 300:1. The laboratory spectra with added synthetic noise were then resampled to the Sentinel-2 MSI spectral resolution. Subsequently, spectral indices were calculated and correlated to the Er concentration of the samples using the Pearson correlation coefficient, after a Log10 transformation as both datasets were not normally distributed. 2.4. Sentinel-2 imagery
A Sentinel-2 MSI level 1C (top-of-atmosphere reflectance) image1
acquired on 21 November 2017 was downloaded from the ESA science hub (European Space Agency, 2019a). This image was chosen for being cloud-free over the study area, and closest to the date of the sample collection, leaving a two-month gap between the fieldwork and the image acquisition date. Sen2Cor software (European Space Agency, 2019b) was used to process the image to level 2A (bottom-of-atmo-sphere reflectance). In this process, all bands were spatially resampled to a10×10m pixel size using the nearest neighbour technique. The
Fig. 1. The reflectance spectra of a tailing sample from Bangka island (black line), measured byPurwadi et al. (2019). The reflectance spectra of hematite (red line), goethite (cyan line), and monazite (green line) are fromKokaly et al. (2017). Orange lines indicate the band positions of the Sentinel-2 MSI bands. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
image was subsequently cropped to the extent of the study area: 642780E; 9725170N (top-left) and 643240E; 9724920N (bottom-right). The coordinate system and datum used in this article are Universal Transverse Mercator (UTM) zone 48 South and World Geodetic System (WGS) 1984, respectively. The size is equivalent to approximately 1150
of the10×10m pixels of Sentinel-2 MSI.
Green vegetation has an absorption feature centred at 674 nm (Lillesand et al., 2004). Hence, pixels that contain vegetation were masked to avoid false anomalies. The Normalized Difference Vegetation Index (NDVI) was calculated with band 4 and band 8a of the Sentinel-2 MSI image, and pixels with an NDVI value of over 0.2 were masked.
To chart the Er3+-related absorption feature at band 4 of the
Sentinel-2 MSI image, the indices applied to the synthetic Sentinel-2 spectra were also applied to the Sentinel-2 image. A 2–98% linear stretch was used to visualise the final product.
Fig. 2. A red box in (a) locates the current study area and (b) shows the area in a remote sensing image. The blue and red dashed rec-tangles in (b) indicate the two tailing heaps studied byPurwadi et al. (2019). The current study area is the tailings heap indicated by the blue dashed rectangle. (c) is a false color image of Sentinel-2 bands :8A, :4, and :3; reddish colored pixels in (c) indicate a presence of ve-getation. (d) shows a field photo of the tailings heap. (For interpretation of the references to color in this figure legend, the reader is re-ferred to the web version of this article.)
Table 1
RE's concentration (µg/g) in the 40 tailing samples studied byPurwadi et al. (2019). Some RE's are below the limit of detection (LOD; 5 µg/g); the right-most column indicates the number of samples above LOD.
Element Min Max Mean N >LOD
Sc 5.2 238.2 36.2 37 Y 5.8 255.6 51.5 26 La 8.1 5502.4 285.1 24 Ce 19.4 10,700.3 367.9 39 Nd 7 3903.3 150.1 34 Sm 5.7 686.7 119 7 Eu 16.3 16.3 16.3 1 Gd 10.6 297.4 113.1 3 Tb 6 24 11.4 5 Dy 5.7 118.9 23.3 12 Ho 13.3 13.3 13.3 1 Er 111.6 3768.4 667.3 40 Yb 5.7 29.1 12.8 7 Lu 5.4 7.8 6.6 2 Fe 135.3 3839.2 896.9 40 Table 2
Equations used for the band ratio index, band depth index, and band peak index. The band ratio is by definition insensitive for albedo differences; for the band depth index and band peak index, the spectra were continuum-removed before calculating the indices.
Index Equation
Band ratio Band5÷Band4
Band depth 1 Band4
3. Results
The spectra measured byPurwadi et al. (2019)were resampled to Sentinel-2 MSI spectral resolution.Fig. 4shows one of these resampled spectra; it can be observed that the 674 nm absorption feature was re-tained in band 4 of Sentinel-2 MSI. The three band indices were plotted in scatter diagrams against Er concentration (Fig. 5). The results in this figure show a positive Pearson correlation coefficient between the spectral indices and the Er concentration, and allr-values are over 0.5 while the p-values are less than 0.01. Also it appears that the 674 nm absorption feature still has a correlation coefficient over 0.5 when the SNR value is over 50:1.
The Sentinel-2 image was masked for vegetation cover: Out of the 115010×10m resolution pixels in the Sentinel-2 image that cover the study area, 334 pixels were masked (Fig. 6a).Fig. 6b–d show the band ratio index, the band depth index and the band peak index, respec-tively. The reflectance spectra of the pixels with the minimum and maximum index values are shown in Fig. 7. Fig. 7a confirms the
presence of the at-band 4 absorption feature in the reflectance spectra of pixels with the maximum values of the indices.
4. Discussion
FromFig. 4can be seen that the 674 nm Er3+absorption feature is
retained at band 4 in the synthetic spectra, which indicates that the Sentinel-2 MSI instruments are, in theory, able to detect this particular absorption feature. The Pearson correlation coefficients in Fig. 5a–c support this conclusion, as all three indices have a positive correlation of 0.6 or higher between the at-band 4 absorption feature and Er con-centration measured in the laboratory. Using 2 nm wide bands at 674 and 712 nm, which are in the approximate centres of bands 4 and 5 of Sentinel-2 respectively, Purwadi et al. (2019)found a correlation of 0.829 with the Er concentration. Despite the lower spectral resolution of Sentinel-2, the correlation coefficients are almost similar (0.825 compared to 0.829). Various amounts of noise were added to the la-boratory spectra before resampling to the Sentinel-2 MSI spectral re-solution. FromFig. 5b, d and f can be seen that the correlation coeffi-cient between Er concentration and the band indices decreases as the value of SNR decreases. At an SNR of 50:1, the correlation coefficient between Er concentration and the band indices is 0.5 or more. At an SNR of Sentinel-2 band 4 (which is at least 142:1,Clerc and MPC team, 2019), the correlation values are around 0.6. This suggests that, in terms of band definitions and instrument sensitivity, the Sentinel-2 MSI sensors can perform as well as the spectroradiometer used byPurwadi et al. (2019).
Mixed pixels in satellite remotely-sensed imagery can be easily en-countered as the spatial resolution of a satellite is not sufficient to discriminate the different surface cover in an area. The mixed pixels that we observed in the field are mixtures between vegetation area and the tailings, which was addressed by using NDVI index for masking. From the 816 image pixels that were not masked, the at-band 4 ab-sorption feature was observed in the pixels with a high value of band indices (Figs. 6and7). This result indicates that the10×10m spatial resolution of Sentinel-2 is sufficiently high for this particular study.
InFig. 6, the band ratio, band depth, and band peak images have a similar distribution of relatively high and low index values. By com-bining the three indices, we can locate pixels whose absorption feature peaks at band 4 of the Sentinel-2 image, indicated by white pixels in the three index images.
The cause of the at-band 5 absorption feature is not known and is also not mentioned in previous studies.
Iron oxide absorption features generally masks or mixes with other
Fig. 3. Illustrations of the band ratio index, the band depth index, and the band peak index. (a) The band ratio is the ratio between band 4 and band 5. (b) shows the same spectrum but con-tinuum-removed. The band depth index equals one minus the value of band 4. The band peak index is the sum of the shoulders (band 3 and band 5) divided by the value of band 4.
Fig. 4. A spectrum with a 674 nm Er3+absorption feature; an ASD spectrum (a black line) of a sample is resampled to Sentinel-2 MSI band resolution (a red line). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
Fig. 5. (a)–(c) show log–log scatter plots of Er concentration versus the spectral indices based on laboratory reflectance spectra resampled to Sentinel-2 MSI band definitions. (b)–(f) show the #r#-values as a function of the signal-to-noise ratio.
absorption features in VNIR reflectance spectra of iron-bearing soil or rock. As shown inFig. 1, iron-bearing minerals also have an absorption feature around 674 nm, therefore Fe2 ,3+ +could also be the origin of a
674 nm absorption feature. Also Nd3+could be the origin of the 674 nm
absorption feature as Nd3+absorption features dominate the spectral
reflectance of RE's-bearing minerals. However, the investigation con-ducted byPurwadi et al. (2019)concluded that the 674 nm absorption feature is related to an unknown Er speciation because of two reasons: (1) the three indices have a stronger correlation with Er concentration (0.829, 0.710, and 0.689) than Fe concentration (0.736, 0.552, and 0.521) or Nd concentration ( 0.036, 0.052, 0.053); (2) the Fe and Nd concentrations in the tailing samples are too low. The Fe con-centration in the tailings samples is<3839.2 µg/g, which is less than the 5000 µg/g needed to produce an iron-oxide absorption feature (Scheinost et al., 1998). Nd3+-related absorption features are observed
when the concentration is more than 1000 µg/g. On average, the Nd concentration in the tailing samples is 140 µg/g and out of 40 tailing samples; only one sample has a concentration of over 1000 µg/g.
Meanwhile, the Er concentration in the tailing samples is exceptionally high, ranging between 111.6 and 3768.4 µg/g. We conclude that, al-though it is possible that the feature is resulting from a proxy rather than Er itself, it is unlikely that there would be a mixing problem due to the presence of iron-bearing minerals or other RE's.
Regarding the mineralogy, X-ray powder diffraction (XRD) tech-nique identified the tailing samples as quartz. However, RE's are un-likely to be hosted in a quartz lattice, but are expected to be in solid mineral or fluid inclusions.Nguyen et al. (2017)studied Er3+and Yb3+
doped glasses and reported that the crystals of Er and Yb were not detected in the diffractogram of the glasses as Er and Yb phase is less than 1 wt%, which is the detection limit of the XRD technique for crystals in a glass matrix (Zhao et al., 2016). Because of this limit, an RE's phase in the samples was not detected.
Unfortunately, other techniques for investigating the mineralogy of the tailing samples were not available during this study. A follow-up investigation using techniques such as electron microprobe analysis (EMPA), scanning electron microscopy (SEM), mineral liberation
Fig. 6. The Sentinel-2 image and spectral in-dices. (a) shows a natural color composite of the Sentinel-2 image, after masking areas with vegetation cover. (b)–(d) show the image pro-ducts of the three spectral indices band ration, band depth and band peak. Bright pixels in-dicate a relatively high index value, while dark pixels indicate a relatively low index value. All images shown are in a 2–98% linear stretch.
Fig. 7. (a) The at-band 4 (674 nm) Er3+absorption feature is observed in the reflectance spectra of pixels that have a maximum index value. (b) Pixels with a minimum index value have no absorption feature at band 4. Instead, there appears to be a feature at band 5. Note that one pixel has the minimum value for both the band ratio index and band peak index.
analyser (MLA) or quantitative evaluation of minerals by scanning electron microscopy (QEMSCAN) would be needed to determine the exact mineralogy of tailing samples.
This study succeeded in detecting an absorption feature, that was previously related to Er concentration in a laboratory study, using Sentinel-2 MSI. It should be noted, however, that the field conditions in this study are favourable and possibly unique. In this case, the tailings consist of quartz, which is featureless in the VNIR wavelength range, without sufficient Fe or Nd present to cause other, overlapping, ab-sorption features.
5. Conclusion
In this paper, we evaluate the capability of Sentinel-2 MSI to target RE's-bearing tailings in an abandoned tin mine site in Bangka Island, Indonesia. In this study area, the Er concentration is exceptionally high and the tailings is predominantly composed of quartz, which creates favourable conditions for mapping RE's.
Resampling the laboratory spectra to Sentinel-2 MSI indicates that the 674 nm Er3+absorption feature remains detectable within band 4.
By using simple band indices that are insensitive to albedo differences, we can highlight pixels of a Sentinel-2 image that have the at-band 4 absorption feature. The correlation of three spectral indices with Er concentration shows that Sentinel-2 MSI can perform almost as good as a hyperspectral sensor in detecting the at-band 4 674 nm Er3+
absorp-tion feature. Given the favourable field condiabsorp-tions, this result may however be limited to this particular study area. It would be worth-while to investigate the applicability of Er detection in other areas, to find the cause of the at-band 5 absorption feature, and to further study the mineralogy of tailings samples, to determine the actual host of the RE's contained in the samples.
Authors’ contribution
I. Purwadi: conceptualisation, methodology, software, formal ana-lysis, investigation, writing – original draft. H.M.A. van der Werff: methodology, validation, formal analysis, writing – review & editing, supervision. C. Lievens: methodology, validation, resources, writing – review & editing, supervision.
Conflict of interest None declared. Acknowledgements
We thank Prof. Freek van der Meer (University of Twente) for his valuable suggestions, and Mr. Bart Krol and Mr. Wim Bakker for their assistance during the preparation of the fieldwork and spectral la-boratory measurements, respectively. We would also like to thank Mr. Robbie Kurniawan for his help during fieldwork. We are grateful for the financial support given by Indonesia Endowment Fund for Education (LPDP) and Geological Remote Sensing Group (GRSG), a special interest group of the Geological Society of London (GeolSoc) and the Remote Sensing and Photogrammetry Society (RSPSoc).
References
Adams, J.W., 1965. The visible region absorption spectra of rare earth minerals. Am. Mineral. 50, 356–366.
Ayunda, F.P., Setijadji, L.D., Warmada, I.W., Ferdila, H., Murtono, A., 2017. Characteristics of the primary tin deposit, Pangkalan Baru Area, Central Bangka, Indonesia. In: SEG 2017 Conference Proceedings. Beijing, China. pp. 102. Bedini, E., 2009. Mapping lithology of the Sarfartoq carbonatite complex, southern West
Greenland, using HyMap imaging spectrometer data. Rem. Sens. Environ. 113, 1208–1219.https://doi.org/10.1016/j.rse.2009.02.007.
Clark, R.N., King, T.V.V., 1987. Causes of spurious features in spectral reflectance data. JPL Publication 87-30 132–137.
Clark, R.N., King, T.V.V., Gorelick, N.S., 1987. Automatic continuum analysis of re-flectance spectra. JPL Publication 87-30 138–142.
Clerc, S., MPC Team, 2019. S2 MPC L1C Data Quality Report. Technical Report. European Space Agency (Accessed July 2019).
Crow, M., van Leeuwen, T., 2005. Metallic mineral deposits. In: Barber, A., Crow, M., Milson, J. (Eds.), Sumatra: Geology, Resources and Tectonic Evolution. The Geological Society London, Oxford, UK, pp. 147–174.
Crowley, J.K., Brickey, D.W., Rowan, L.C., 1989. Airborne imaging spectrometer data of the Ruby Mountains, Montana: mineral discrimination using relative absorption band-depth images. Rem. Sens. Environ. 29, 121–134.https://doi.org/10.1016/ 0034-4257(89)90021-7.
European Space Agency, a. Copernicus open Access Hub.
European Space Agency, b. Level-2A Prototype Processor for Atmospheric, Terrain and Cirrus Correction of Top-Of-Atmosphere Level 1C Input Data.http://step.esa.int/ main/third-party-plugins-2/sen2cor/(Accessed May 2019).
European Space Agency, c. MultiSpectral Instrument (MSI) Overview.https://earth.esa. int/web/sentinel/technical-guides/sentinel-2-msi/msi-instrument(Accessed July 2019).
Goetz, A.F.H., Rock, B.N., Rowan, L.C., 1983. Remote sensing for exploration; an over-view. Econ. Geol. 78, 573–590.https://doi.org/10.2113/gsecongeo.78.4.573. Goetz, A.F.H., Rowan, L.C., 1981. Geologic remote sensing. Science 211, 781–791.
https://doi.org/10.1126/science.211.4484.781.
Haire, R., Eyring, L., 1994. Comparisons of the binary oxides. In: Gschneidner, K.A., Eyring, L., Choppin, G.R., Lander, G.H. (Eds.), Handbook on the Physics and Chemistry of Rare Earths. Elsevier, pp. 413–505. https://doi.org/10.1016/S0168-1273(05)80048-0.(Chapter 125).
Kokaly, R.F., Clark, R.N., Swayze, G.A., Livo, K.E., Hoefen, T.M., Pearson, N.C., Wise, R.A., Benzel, W.M., Lowers, H.A., Driscoll, R.L., Klein, A.J., 2017. USGS Spectral Library Version 7. Technical Report. United States Geological Surveyhttps://doi.org/ 10.3133/ds1035.
KoKo, U., 1986. Preliminary synthesis of the geology of Bangka Island, Indonesia. Bull. Geol. Soc. Malays. 20, 81–96.
Lehnert, L.W., Meyer, H., Bendix, J., 2017. HSDAR: Manage, Analyse and Simulate Hyperspectral Data in R. Online.https://cran.r-project.org/web/packages/hsdar/ index.html(Accessed 19 March 2019).
Lillesand, T., Kiefer, R., Chipman, J., 2004. Remote Sensing and Image Interpretation, 5th ed. John Wiley & Sons, Inc., Hoboken.
van der Meer, F., van der Werff, H., van Ruitenbeek, F., 2014. Potential of ESA's Sentinel-2 for geological applications. Rem. Sens. Environ. 148, 1Sentinel-24–133.https://doi.org/10. 1016/j.rse.2014.03.022.
van der Meer, F.D., van der Werff, H.M.A., van Ruitenbeek, F.J.A., Hecker, C.A., Bakker, W.H., Noomen, M.F., van der Meijde, M., Carranza, E.J.M., de Smeth, J.B., Woldai, T., 2012. Multi- and hyperspectral geologic remote sensing: a review. Int. J. Appl. Earth Observ. Geoinform. 14, 112–128.https://doi.org/10.1016/j.jag.2011.08.002. Neave, D.A., Black, M., Riley, T.R., Gibson, S.A., Ferrier, G., Wall, F., Broom-Fendley, S., 2016. On the feasibility of imaging carbonatite-hosted rare earth element deposits using remote sensing. Econ. Geol. 111, 641–665.https://doi.org/10.2113/econgeo. 111.3.641.
Ng, S.W.P., Whitehouse, M.J., Roselee, M.H., Teschner, C., Murtadha, S., Oliver, G.J., Ghani, A.A., Chang, S.C., 2017. Late triassic granites from Bangka, Indonesia: a continuation of the main range granite province of the south-east Asian tin belt. J. Asian Earth Sci. 138, 548–561.https://doi.org/10.1016/j.jseaes.2017.03.002. Nguyen, H., Tuomisto, M., Oksa, J., Salminen, T., Lastusaari, M., Petit, L., 2017.
Upconversion in low rare-earth concentrated phosphate glasses using direct NaYF4:Er3+, Yb3+nanoparticles doping. Scr. Mater. 139, 130–133.https://doi.org/
10.1016/j.scriptamat.2017.06.050.
Purwadi, I., van der Werff, H.M.A., Lievens, C., 2019. Reflectance spectroscopy and geochemical analysis of rare earth element-bearing tailings: a case study of two abandoned tin mine sites in Bangka Island, Indonesia. Int. J. Appl. Earth Observ. Geoinform. 74, 239–247.
Rowan, L.C., Mars, J.C., 2003. Lithologic mapping in the Mountain Pass, California area using advanced spaceborne thermal emission and reflection radiometer (ASTER) data. Rem. Sens. Environ. 84, 350–366.https://doi.org/10.1016/S0034-4257(02) 00127-X.
Rowan, L.C., Wetlaufer, P., Goetz, A.F.H., Billingsley, F.C., Stewart, J.H., 1974. Discrimination of rock types and detection of hydrothermally altered areas in south-central Nevada by the use of computer-enhanced ERTS images. Geol. Surv. Profess. Paper 35.https://doi.org/10.3133/pp883.
Scheinost, A., Chavernas, A., Barrón, V., Torrent, J., 1998. Use and limitations of second-derivative diffuse reflectance spectroscopy in the visible to near-infrared range to identify and quantify Fe oxide minerals in soils. Clays Clay Miner. 46, 528–536.
https://doi.org/10.1346/CCMN.1998.0460506.
Schwartz, M.O., Surjono, 1991. The pemali tin deposit, Bangka, Indonesia. Miner. Depos. 26, 18–25.https://doi.org/10.1007/BF00202359.
Sinding-Larsen, R., Wellmer, F., 2012. Non-renewable resource issues: geoscientific and societal challenges. International Year of Planet Earth series. Springer, Dordrecht, The Netherlands.https://doi.org/10.1007/978-90-481-8679-2.
Taylor, S.R., McLennan, S.M., 1995. The geochemical evolution of the continental crust. Rev. Geophys. 33, 241–265.https://doi.org/10.1029/95RG00262.
Turner, D.J., Rivard, B., Groat, L.A., 2014. Visible and short-wave infrared reflectance spectroscopy of REE fluorocarbonates. Am. Mineral. 99, 1335–1346.https://doi.org/ 10.2138/am.2014.4674.
Turner, D.J., Rivard, B., Groat, L.A., 2016. Visible and short-wave infrared reflectance spectroscopy of REE phosphate minerals. Am. Mineral. 101, 2264–2278.https://doi. org/10.2138/am-2016-5692.
spectroscopy of selected REE-bearing silicate minerals. Am. Mineral. 103, 927–943.
https://doi.org/10.2138/am-2018-6195.
van der Werff, H., van der Meer, F., 2015. Sentinel-2 for mapping iron absorption feature parameters. Rem. Sens. 7, 12635–12653.https://doi.org/10.3390/rs7101263. van der Werff, H., van der Meer, F., 2016. Sentinel-2A MSI and Landsat 8 OLI provide
data continuity for geological remote sensing. Rem. Sens. 8, 883.https://doi.org/10. 3390/rs8110883.
Zhao, J., Zheng, X., Schartner, E.P., Ionescu, P., Zhang, R., Nguyen, T.L., Jin, D., Ebendorff-Heidepriem, H., 2016. Upconversion nanocrystal-doped glass: a new paradigm for photonic materials. Adv. Opt. Mater. 4, 1507–1517.https://doi.org/10. 1002/adom.201600296.
Zimmermann, R., Brandmeier, M., Andreani, L., Mhopjeni, K., Gloaguen, R., 2016. Remote sensing exploration of Nb-Ta-LREE enriched carbonatite (Epembe/Namibia). Rem. Sens. 8, 620.https://doi.org/10.3390/rs8080620.
