Abstract
We have investigated the feasibility of using lab-based short-wave infrared (SWIR) hyperspectral imagery to identify and map the mineralogical composition of selected sedimentary drill core from the subsurface of the Netherlands. We aimed at i) identifying and mapping spectrally active minerals, such as white mica, clay and carbonate minerals, ii) the quantification of clay content in the core and their detection limits, iii) the possibility of mapping minerals in core samples with uneven surfaces with out-of-focus hyperspectral images, and iv) the added value of using high-resolution (26 µm pixel-size) compared to intermediate-resolution images (at 200 µm pixel-size).
Twenty-six core samples from wells L18-06A, L16-02A, L16-03 and L16-14 were imaged with a SWIR hyperspectral scanner. The images were processed to interpretable products, such as spectral parameters and classified maps that enhanced the variation in mineralogical composition. Results showed that nearly all image-spectra of all core samples were dominated by white mica (muscovite-illite) and clay minerals (kaolinite-dickite and illite-smectite). The variation in white mica and clay minerals could be described by i) the wavelength position of the Al-OH feature near 2200 nm, ii) the muscovite-illite crystallinity value which is calculated by division of the Al-OH feature near 2200 nm and the water feature near 1900 nm, and iii) the ratio of illite relative to kaolinite-dickite. Secondary minerals that were observed using the SWIR hyperspectral images were Fe-Mg carbonates and Fe-oxides.
The total clay content (the sum of white mica and clay minerals) could be estimated from the SWIR hyperspectral data by comparison with compositional data of previous mineralogical studies (with CHEMSCAN and SEM methods) on the same core samples. Regression models showed that the total clay content can be modeled from the depth of clay features near 1400 and 2200 nm with a mean error of 4.1 % in the preferred regression model. Since the depth of clay absorption features are direct indicators of the total clay content, the models are likely to be portable to other sedimentary core samples. However, the models should be tested on other core samples with a wider range of clay contents than the ranges of 6.7 to 27.9 % and 88.9 to 91.0 % of the sample set. The limit of detection is likely to be significantly lower than 6.7%, the lowest clay content encountered in the core sample set. Comparison with hyperspectral mineralogical maps obtained from semi-continuous core over the same intervals shows similar mineralogical patterns and variation as in the core samples.
The spectral response and detail in hyperspectral images that are out-of-focus (Labscanner from -5 cm to +5 cm and SisuCHEMA from -2.6 cm to +0.6 cm) are still sufficient to identify the main mineralogical composition and total clay content. This makes it possible to image samples with uneven surfaces without adjusting the distance to the hyperspectral camera. Although images acquired at 200 μm pixel-size provide sufficient detail to identify the dominant white mica, clay and carbonate mineralogy and quantify total clay content, higher resolution images at provide more mineralogical and textural detail that may be required for more detailed investigations of drill core. Finally, results show that washing and drying have an effect on the SWIR spectra of the core samples. Drying and subsequent hydration of the samples affect the shape and depth of the water feature. The samples must be measured after similar treatment and under similar moisture conditions in order to obtain reproducible illite-crystallinity estimates.
Initial attempts to develop spectral indicators for the identification of very-low permeability rocks are encouraging. Spectra of illite minerals, with relatively short wavelengths of OH and OH/H2O features near 1400 and 1900 nm, show significant correlations with the presence of very low permeability rocks. Further investigation is required to further define the relationship and to explain inconsistencies between the classification results of the core samples and larger intervals of archived core.
We conclude that laboratory-based SWIR hyperspectral imaging at 200 μm provides a useful, non-destructive method to identify and map the white mica, clay and carbonate minerals in the investigated sedimentary drill core. Furthermore, this allows for quantification of the total clay content and offers the possibility to identify zones of very-low permeability.
Twenty-six core samples from wells L18-06A, L16-02A, L16-03 and L16-14 were imaged with a SWIR hyperspectral scanner. The images were processed to interpretable products, such as spectral parameters and classified maps that enhanced the variation in mineralogical composition. Results showed that nearly all image-spectra of all core samples were dominated by white mica (muscovite-illite) and clay minerals (kaolinite-dickite and illite-smectite). The variation in white mica and clay minerals could be described by i) the wavelength position of the Al-OH feature near 2200 nm, ii) the muscovite-illite crystallinity value which is calculated by division of the Al-OH feature near 2200 nm and the water feature near 1900 nm, and iii) the ratio of illite relative to kaolinite-dickite. Secondary minerals that were observed using the SWIR hyperspectral images were Fe-Mg carbonates and Fe-oxides.
The total clay content (the sum of white mica and clay minerals) could be estimated from the SWIR hyperspectral data by comparison with compositional data of previous mineralogical studies (with CHEMSCAN and SEM methods) on the same core samples. Regression models showed that the total clay content can be modeled from the depth of clay features near 1400 and 2200 nm with a mean error of 4.1 % in the preferred regression model. Since the depth of clay absorption features are direct indicators of the total clay content, the models are likely to be portable to other sedimentary core samples. However, the models should be tested on other core samples with a wider range of clay contents than the ranges of 6.7 to 27.9 % and 88.9 to 91.0 % of the sample set. The limit of detection is likely to be significantly lower than 6.7%, the lowest clay content encountered in the core sample set. Comparison with hyperspectral mineralogical maps obtained from semi-continuous core over the same intervals shows similar mineralogical patterns and variation as in the core samples.
The spectral response and detail in hyperspectral images that are out-of-focus (Labscanner from -5 cm to +5 cm and SisuCHEMA from -2.6 cm to +0.6 cm) are still sufficient to identify the main mineralogical composition and total clay content. This makes it possible to image samples with uneven surfaces without adjusting the distance to the hyperspectral camera. Although images acquired at 200 μm pixel-size provide sufficient detail to identify the dominant white mica, clay and carbonate mineralogy and quantify total clay content, higher resolution images at provide more mineralogical and textural detail that may be required for more detailed investigations of drill core. Finally, results show that washing and drying have an effect on the SWIR spectra of the core samples. Drying and subsequent hydration of the samples affect the shape and depth of the water feature. The samples must be measured after similar treatment and under similar moisture conditions in order to obtain reproducible illite-crystallinity estimates.
Initial attempts to develop spectral indicators for the identification of very-low permeability rocks are encouraging. Spectra of illite minerals, with relatively short wavelengths of OH and OH/H2O features near 1400 and 1900 nm, show significant correlations with the presence of very low permeability rocks. Further investigation is required to further define the relationship and to explain inconsistencies between the classification results of the core samples and larger intervals of archived core.
We conclude that laboratory-based SWIR hyperspectral imaging at 200 μm provides a useful, non-destructive method to identify and map the white mica, clay and carbonate minerals in the investigated sedimentary drill core. Furthermore, this allows for quantification of the total clay content and offers the possibility to identify zones of very-low permeability.
| Original language | English |
|---|---|
| Publisher | University of Twente, Faculty of Geo-Information Science and Earth Observation (ITC) |
| Number of pages | 47 |
| Publication status | Published - 20 Oct 2020 |