Characterizing clay mineralogy in Lake Towuti, Indonesia, with reflectance spectroscopy
- 315 Downloads
We tested the use of visible to near-infrared (VNIR) reflectance spectroscopy to characterize the relative abundances of clay minerals in sediments from Lake Towuti, a large tectonic lake in Sulawesi, Indonesia. We measured VNIR spectra of lake and river sediments from Lake Towuti and its catchment to identify clay minerals, fit major VNIR absorption features with a modified Gaussian model to estimate relative abundances of these minerals, and compared these absorptions to the samples’ chemistry to test the utility of VNIR spectroscopy to characterize sediment compositional variations. We found that major absorptions are caused by vibrations of Al–OH in kaolinite (2.21 μm), Fe–OH in nontronite (2.29 μm), Mg–OH in saponite and serpentine (2.31 μm), and Mg–OH in serpentine (2.34 μm). This was confirmed with X-ray diffraction data. The correlations between absorption band areas for Fe–OH, Al–OH, and Mg–OH vibrations and Fe, Al and Mg concentrations, respectively, are statistically significant, varying between r = 0.51 and r = 0.90, and spatial variations in inferred clay mineralogy within the lake are consistent with variations in the geology of the catchment. We conclude that VNIR spectroscopy is an effective way to characterize the clay mineralogy of lake sediments, and can be used to investigate changes in mineral inputs to lake deposits.
KeywordsClay mineralogy Lake sedimentology Paleolimnology Spectroscopy Modified Gaussian modeling
The clay mineralogy of lake sediments can provide valuable insight into the history of sediment provenance (Mitchell 1955; Johnson 1970), weathering processes (Asikainen et al. 2006), and depositional environments. Clay mineralogy is commonly measured using X-ray diffraction (XRD), although this can be a relatively difficult process involving numerous and often destructive pretreatments of large sediment samples (Yuretich et al. 1999). In contrast, visible to near-infrared (VNIR) reflectance spectroscopy is a rapid, non-destructive method that can provide information on the mineral composition of sediments through the position and shape of the absorption features, which are controlled by crystal structure and mineral chemistry. Infrared spectroscopy has been used to measure organic carbon, biogenic opal, and carbonate mineral concentrations in lake sediment samples to understand long-term paleoenvironmental variations (Rosén and Persson 2006; Vogel et al. 2008; Rosén et al. 2010), to identify and map clay minerals in paleolake deposits on Mars (Ehlmann et al. 2008; Milliken and Bish 2010), and to determine the composition of terrestrial soils (Viscarra Rossel et al. 2006, 2009). Despite the sensitivity of VNIR reflectance data to variations in clay mineralogy, this technique has not been widely employed to measure clay mineralogical variations in lake sediment samples.
Materials and methods
Sample preparation and analysis
River sediment samples (9) and lake surface sediment samples (33) were collected in 2011 and 2012. Offshore lake sediment samples were characterized visually and microscopically and consist almost entirely of very fine silt and clay, but river samples had widely variable grain sizes, ranging up to coarse gravel. River sediment samples were therefore sieved to <125 µm before being powdered to prevent spectroscopic and geochemical measurements from being biased by large grains (coarse sands and gravels). Samples were freeze-dried, powdered, and homogenized, and aliquots were prepared via flux fusion to analyze their bulk elemental chemistry following the procedure outlined by Murray et al. (2000). Inductively Coupled Plasma-Atomic Emission Spectroscopy (ICP-AES) was used to measure Al, Ca, K, Fe, Si, Ti, Mg, Mn, and Cr concentrations. The VNIR spectra of separate aliquots were measured using an Analytical Spectral Devices (ASD Inc., CO, USA) FieldSpec 3 portable spectrometer (subsequently referred to as an ASD Spectrometer), a VNIR reflectance spectrometer that covers the wavelength range from 350 to 2500 nm [see Electronic Supplementary Material (ESM) for details on measurements]. High-resolution spectral data were also acquired for three lake sediment samples at the Brown University Keck/NASA Reflectance Experiment Lab (RELAB; Pieters 1983) to evaluate the lower-resolution VNIR data from the ASD Spectrometer (see ESM for discussion). Powder XRD data on select samples were also collected using a Bruker D2 PHASER X-ray diffractometer to provide complementary mineralogy to VNIR reflectance measurements (see ESM for details on measurements).
VNIR data analysis and modified Gaussian modeling
A modified Gaussian model (MGM) was used to quantify the strengths of clay mineral absorption features in the VNIR spectra, which are, to first order, related to the relative abundances of clay minerals in samples (Clark 1999). The MGM, first presented by Sunshine et al. (1990), was initially developed and validated for crystal field absorptions in pyroxene (see ESM for further details). Subsequently, it has been shown that the MGM can also be used to model vibrational absorption bands caused by OH within actinolite (Mustard 1992). Modified Gaussian fits to each absorption band of interest in our samples were used to determine the band area, which is controlled by the absorption strength and can be used as a proxy for the relative abundances of different minerals, in contrast to absolute abundances. Whereas band depth is also a useful proxy for mineral abundance, the overlapping absorption features of many clay minerals necessitates the use of absorption band area to deconvolve complex spectral data into a series of individual absorption features.
VNIR spectroscopy and XRD mineralogy
VNIR spectroscopy is based upon the absorption of electromagnetic radiation at specific wavelengths through excitation of electronic transitions or molecular vibrations within mineral structures. The precise position and shape of VNIR absorption features, or absorption bands, are related to the optical constants of the material being investigated. In minerals, the position and shape of the absorption features are controlled by the crystal structure and mineral chemistry. Whereas each individual absorption feature is related to a specific cation’s coordination state and electrons (electronic transition absorptions) or bond (vibrational absorption), taken together, multiple absorption features can be diagnostic of specific minerals. Different minerals have unique VNIR spectral properties, and thus sample mineralogy can be uniquely identified on the basis of infrared reflectance data (Burns 1993; Farmer 1974; Hunt 1977; Hunt and Salisbury 1970).
Many of the absorption bands present in the Lake Towuti samples are common to multiple minerals. The sharp absorption at ~1.9 µm indicates the presence of structural water common to many phyllosilicates and opal (Clark et al. 1990; Bishop et al. 1994), whereas the absorptions at ~1.4 µm are caused by the first overtone of structural OH, as well as combination tones of structural H2O (Clark et al. 1990; Gaffey et al. 1993). In contrast, absorption bands in the ~2.2 to 2.4 µm region are caused by combinations of the stretching and bending modes of OH and metal-OH, respectively, and are diagnostic for many clay minerals in weathered oxisol soils (Fig. 3c, d; Clark et al. 1990). River and lake sediment samples from the Lake Towuti basin show several absorption features in this wavelength region. Based on the analysis of the locations and shapes of the identified absorption features in the measured spectra, in comparison to those in pure minerals (Clark et al. 1990), we attribute absorptions at ~2.21 µm to Al–OH vibrations in kaolinite, absorptions at ~2.29 µm to Fe–OH vibrations in an Fe-bearing smectite (nontronite), and absorptions at ~2.31 and 2.34 µm to Mg–OH vibrations in a combination of an Mg-bearing smectite (saponite) and serpentine.
The absorption band at ~2.21 μm is interpreted to be caused by Al–OH vibrations in kaolinite, as opposed to some other Al-bearing phyllosilicate such as montmorillonite or illite, due to the precise position of the 2.21 μm absorption coupled with an asymmetric shoulder in the band near ~2.16 µm (Clark et al. 1990; Bishop et al. 2008). The complex spectral signal near ~2.3 μm is a region often characteristic of Fe/Mg-bearing phyllosilicates. Nontronite is a dioctahedral ferric iron-bearing smectite and has an absorption centered at ~2.29 µm, caused by the Fe3+–OH bond (Clark et al. 1990; Bishop et al. 2002, 2008). Saponite, a trioctahedral magnesium-bearing smectite, has an absorption centered at ~2.31 to 2.32 µm caused by Mg–OH (Clark et al. 1990). Although some samples show strong absorptions at ~2.29 or ~2.32 µm, indicating nontronite or saponite, respectively (Clark et al. 1990; Bishop et al. 2002, 2008), many samples have absorption features between 2.29 and 2.32 µm, likely from either Fe/Mg substitution between nontronite and saponite or physical mixtures of the two mineral species. Absorption bands that fall between the endmember wavelengths are interpreted as a mixture of clays with varying Mg/Fe ratios (Grauby et al. 1994).
The absorption band at ~2.33 µm and a more subtle absorption feature at ~2.1 µm in several samples indicates the presence of serpentine. Serpentine has a prominent absorption feature caused by Mg–OH that overlaps with the absorption features of the Fe/Mg-bearing smectites (Fig. 3c, d), but with a band center at slightly longer wavelengths, near ~2.33 to 2.34 µm (King and Clark 1989; Bishop et al. 2008). Serpentines also have a shallow, but broad diagnostic absorption feature centered near ~2.1 μm (Fig. 3c, d; King and Clark 1989; Bishop et al. 2008) that may relate to the presence of Mg–OH (Clark et al. 1990).
Our XRD results indicate the presence of several mineral phases in these samples, including serpentine, kaolinite, smectite and quartz (ESM Fig. 3). Serpentine is identified based on prominent 001 and 002 reflections, as well as a 100 reflection peak. Kaolinite is identified based on 001 and 002 reflections, and a prominent 020 reflection. Smectite is identified based on a small 001 basal reflection peak. These XRD results confirm the major clay mineral phases inferred from our VNIR reflectance spectroscopy results.
Modified Gaussian modeling
The modified Gaussian model fits the clay mineral absorption features well (ESM Fig. 2). A single modified Gaussian absorption band fits the kaolinite Al–OH absorption, with root-mean-square error (RMSE) values ranging from 0.10 to 0.28 %. The complex absorption band from ~2.3 to 2.35 µm shows several absorption features that are optimally fit by three modified Gaussians, resulting in fits with RMSE values ranging from 0.08 to 0.31 %. Based upon these results, we modeled sample spectra with modified Gaussians centered at 2.205 μm (σ = 0.0127 μm), 2.295 μm (σ = 0.0115 μm), 2.315 μm (σ = 0.0106 μm), and 2.345 μm (σ = 0.0106 μm), which represent contributions from Al–OH (kaolinite), Fe–OH (nontronite), Mg–OH (saponite + serpentine), and Mg–OH (serpentine), respectively (ESM Table 1).
The absorption band areas calculated from the MGM indicate substantial variations in the relative absorption areas (or relative abundances) of Al–OH (kaolinite), Fe–OH (nontronite), and Mg–OH (saponite + serpentine) among samples (Fig. 2d–f). Samples with the strongest Al–OH (kaolinite) absorptions are located in the western part of the lake (Fig. 2d), whereas samples with the strongest Fe-smectite absorptions are located in the east (Fig. 2e). Samples with the strongest Mg-smectite/serpentine absorptions are located in the central part of the lake (Fig. 2f). Although we did not perform quantitative XRD analysis, our XRD results qualitatively confirm the variation in mineralogy inferred from our MGM modelling results. The sample with a strong Mg–OH (serpentine) absorption has an XRD pattern that is more dominated by serpentine than the sample with a strong Al–OH absorption, which appears to have a higher proportion of kaolinite (ESM Fig. 3).
Chemistry and absorption area correlations
We found a strong correlation between Fe, Mg, and Al elemental concentration and VNIR spectral reflectance features that distinguish Fe, Mg, and Al-rich clay minerals, in lake and river sediments from the Towuti basin. Elevated concentrations of aluminum and inferred high relative kaolinite abundance on the western shore, elevated concentrations of iron with the inferred high relative nontronite abundance on the eastern shore, and elevated concentrations of magnesium with the inferred high relative saponite/serpentine abundances in the central part of the lake correspond to variations in the geology of Lake Towuti’s catchment, namely, the presence of felsic mélanges to the west, highly serpentenized peridotites to the north, and ultramafic lithologies to the east. These results show that VNIR reflectance spectroscopy is an effective, statistically significant way to characterize the clay mineralogy of Lake Towuti sediment. Despite the fact that samples must be freeze-dried to remove the effects of interstitial water on the VNIR spectra, our work suggests that VNIR spectroscopy is a very practical, time- and cost-effective tool for clay mineral characterization relative to other procedures such as XRD. Although this study demonstrates the utility of VNIR spectroscopy using select absorption features, many other wavelengths could be investigated using VNIR spectra to characterize mineralogy. For example, absorption bands under 1 μm can be diagnostic of iron oxides and absorption bands up to 4 μm are diagnostic of carbonate features (Clark et al. 1990; Bishop et al. 2008). Moreover, as VNIR spectrometers may now be employed on automated core logging devices, this work suggests there is potential to characterize the relative abundances of clays in sediment cores at relatively high resolution, which can be used for paleoenvironmental studies.
The ability to correlate the chemical composition of lake sediment samples to their mineralogy deduced from VNIR reflectance spectra has important implications for characterizing clay mineralogy and sediment sources not only on Earth, but also on Mars. Lake Towuti and many martian paleolake deposits have strong spectral signals of weathered clay minerals and contain many of the same mineralogic constituents (Ehlmann et al. 2008; Milliken and Bish 2010). Previous authors have explored the use of clay mineralogy to determine sediment source regions for martian paleolakes (Ehlmann et al. 2008), and to infer martian hydrologic history (Milliken and Bish 2010), yet these mineral identifications have rarely been validated using natural lake sediment samples. Our findings thus provide important ‘groundtruthing’ for clay mineralogies inferred from spectroscopic data. This study also suggests that clays in this dilute, ultramafic basin primarily represent allochthonous material that records catchment weathering and transport processes, with important implications for the interpretation of long-term climate records. Overall, this work shows that VNIR reflectance spectroscopy is a powerful tool when combined with chemical analysis. With proper caution, it can be used to characterize the primary clay mineral constituents of natural sediment samples.
The authors would like to thank Dave Murray, Joe Orchardo and Dr. Takahiro Hiroi for technical support and assistance and Satrio Wicaksono, Sinyo Rio, and PT Vale for field assistance in Indonesia. The authors would also like to thank Dr. Kevin Robertson for assistance with XRD data interpretation and two anonymous reviewers who provided excellent feedback to strengthen this paper. Research permits for this work were granted by the Indonesian Ministry of Research and Technology (RISTEK). This material is based upon work supported by the National Science Foundation under Grant Number EAR-1144623 to J. Russell.
- Clark RN (1999) Spectroscopy of rocks and minerals, and principles of spectroscopy. Wiley, New YorkGoogle Scholar
- Clark RN, Swayze GA, Wise R, Livo E, Hoefen T, Kokaly R, Sutley SJ (2007) USGS digital spectral library splib06a. US Geol Surv Digit Data Ser, p 231Google Scholar
- Gaffey SJ, McFadden LA, Nash D, Pieters CM (1993) Ultraviolet, visible, and near-infrared reflectance spectroscopy: laboratory spectra of geologic materials. In: Pieters CM, Englert PAJ (ed) Remote geochemical analysis: elemental and mineralogical composition. Cambridge University Press, Cambridge, pp 43–77Google Scholar
- Golightly JP (1981) Nickeliferous laterite deposits. Econ Geol 75:710–735Google Scholar
- Hunt GR, Salisbury JW (1970) Visible and near-infrared spectra of minerals and rocks: I silicate minerals. Mod Geol 1:283–300Google Scholar
- Murray RW, Miller DJ, Kryc KA (2000) Analysis of major and trace elements in rocks, sediments, and interstitial waters by inductively coupled plasma-atomic emission spectrometry (ICP-AES). ODP Technical NoteGoogle Scholar
- Mustard JF (1992) Chemical analysis of actinolite from reflectance spectra. Am Mineral 77:345–358Google Scholar