URTeC: 154

Continuous Mineral Mapping of Core Using Hyperspectral Imaging:

Example from the Upper Cretaceous Austin Chalk Marathon 1 Robert
Todd Core, Central Louisana
Tobi Kosanke,1 Robert G. Loucks2, Toti Larson2, James Greene3, Paul Linton3
Independent Consultant, Hempstead, TX, 2-Bureau of Economic Geology, Jackson School of Geosciences, The
University of Texas at Austin, Austin, TX, 3-TerraCore, Reno NV

Copyright 2019, Unconventional Resources Technology Conference (URTeC) DOI 10.15530/urtec-2019-154

This paper was prepared for presentation at the Unconventional Resources Technology Conference held in Denver, Colorado, USA,
22-24 July 2019.
The URTeC Technical Program Committee accepted this presentation on the basis of information contained in an abstract submitted
by the author(s). The contents of this paper have not been reviewed by URTeC and URTeC does not warrant the accuracy,
reliability, or timeliness of any information herein. All information is the responsibility of, and, is subject to corrections by the
author(s). Any person or entity that relies on any information obtained from this paper does so at their own risk. The information
herein does not necessarily reflect any position of URTeC. Any reproduction, distribution, or storage of any part of this paper by
anyone other than the author without the written consent of URTeC is prohibited.

Abstract

Hyperspectral imaging (HI) is a method of observing and enhancing geological rock properties that are

not readily apparent visually. Originally developed for the mining industry, HI uses a combination of

short-wave infrared light (SWIR) and long-wave infrared light (LWIR) to create a visual ‘map’ of the

minerals on the surface of a core that respond to reflectance principles. HI, which requires no special

preparation other than that the core be slabbed, clean, and dry, can be rapidly obtained and provides

mineralogical and chemical results related to various energy emitted in wavelength spectrum by either

halogen bulb reflectance (short-wave quantification) or heat reflectance spectra (long-wavelength

quantification).

We collected hyperspectral core imaging data of the Marathon 1 Austin Chalk Robert Todd core in

central Louisiana to obtain detailed, high-resolution mineralogical and textural information and

investigate the application of hyperspectral imaging as an integrative tool.

URTeC
URTeC 154
154 2

Digital HI-derived single mineral curves calibrated to X-ray diffraction (XRD) were imported as curves to

display mineralogical variations with depth alongside overlays showing the textural relationships of the

mineralogical assemblages, rock typing models, X-ray fluorescence (XRF) data, TOC data and

rockmechanics data. The integration of the hyperspectral data with core description, SEM, thin-section,

XRF,

XRD, rock mechanics and TOC data illustrates relative differences in carbonate volumes that identify

Milankovitch cycles, delineates fabric via variations in mineralogical composition of fine laminae,

identifies relatively Sr-rich intervals that cannot be distinguished visually, reveals a relationship between

total organic content and mineralogy, and facilitates upscaling of SEM and thin-section date to the core

scale.

Introduction

Hyperspectral core imaging involves a method of non-destructive, infrared spectroscopy to capture

mineralogical and textural information of the slabbed face of a core (Kosanke et al., 2017). The

application of this technology to core analysis evolved from an origin in multispectral, remote sensing

programs where it was used to acquire satellite imagery of Earth (Bernstein, 1976). The mining industry

was quick to implement the use of visible near-infrared (VNIR) and short-wave infrared (SWIR) in

airborne systems for exploration (Vane and Goetz, 1988; Bierwirth et al., 2002; Plaza et al., 2009; Taranik

and Aslett, 2009; Murphy and Monteiro, 2013). Other industries, such as agriculture, pharmaceuticals,

homeland security, medicine, recycling, food, forensics, and art have also adopted hyperspectral analysis

in a variety of applications (ElMasry and Sun, 2010; Sun, 2010; Marshall, 2011; Edelman et al., 2012; Lu

and Fei, 2014; Cucci et al., 2016). Technological advances in the last few decades have enabled the

collection of data at higher spatial and spectral resolutions (Kurz et al., 2008; Monteiro et al., 2009; Kurz

et al., 2011; Murphy and Monteiro, 2013; Kosanke et al., 2017). In 2007, VNIR-SWIR hyperspectral

imaging was implemented into core analysis workflows for the mining industry with notable success.
URTeC
URTeC 154
154 3
Although valuable for the mining industry, HI systems utilizing the VNIR–SWIR spectral regions had

limited applicability for hydrocarbon reservoirs because of the inability for certain minerals to be uniquely

detected, including quartz and feldspar (Hunt et al., 1973). The introduction of long-wave infrared

(LWIR) data collection for core imaging became a significant, commercially unique advancement to

allow the identification of tectosilicate mineral species, which were formerly aspectral in the VNIR-SWIR

(Kosanke et al., 2017). Its addition also facilitates the acquisition of improved spectral information from

dark lithologies which exhibit low reflectance in the VNIR-SWIR. The integration of spectrometers using

specialized lenses provides the ability to detect a comprehensive range of minerals including carbonate,

clay, tectosilicate, and sulfate species at a high resolution of 300-500 micron pixels. The SWIR also has

hydrocarbon detection capabilities to augment the inorganic characterization.

We collected hyperspectral core imaging data of the Austin Chalk Group from the Robert Todd core in

central Louisiana to obtain detailed, high-resolution mineralogical and textural information in order to

investigate the application of hyperspectral imaging as an integrative tool in unconventional carbonate

reservoirs.

Methods

The Robert Todd Austin Chalk core was imaged using a combination of short-wave infrared light (SWIR)

and long-wave infrared light (LWIR) to create a visual ‘map’ of the minerals on the surface of the core

that respond to reflectance principles. HI scanning of the core required no special preparation other than

that the core be slabbed, clean, and dry. Raw hyperspectral data of mineralogical and chemical

information related to various energy emitted in wavelength spectrum by either halogen bulb reflectance

(short-wave quantification) or heat reflectance spectra (long-wavelength quantification) was rapidly

obtained (~12 minutes per core box).

URTeC
URTeC 154
154 4
The SWIR spectrometer collects 288 bands of data between 1000 and 2500 nm, at 5.6 nm sampling rate

and 9 nm spectral resolution. The LWIR spectrometer collects 96 bands of data between 7700 and 12300

nm, at 48 nm sampling rate and 100 nm spectral resolution. Data are collected at a high spatial resolution

of approximately 400 microns per pixel for the infrared spectrometers and 120 microns per pixel for the

RGB camera. The pixel resolution captured for these datasets reduces the degree of mineral mixing and

provides more detailed mapping for thin bedding and laminations (Browning and Kosanke, 2016).

Reference material data (Spectralon for the SWIR and brushed aluminum for the LWIR) are acquired

along with dark-current measurements for each core imaged. Dark-current measurement accounts for the

noise generated by the temperature difference between the environment and the hyperspectral camera

(Manea and Calin, 2015).

Following data acquisition, core masking processes are used to extract the desired pixels for analysis from

the background, open fractures, and uneven surfaces. This masking procedure allows the image analysis

routines to exclude outlier and noise-attributing data. Because of the size and dimensionality of the

acquired image dataset, an artificial neural network computational model is then utilized to separate the

dataset into classifications that are spectrally and spatially significant. This unsupervised self-organizing

map (SOM) classification algorithm is applied with specific wavelet features being designated as

endmember variables. Every masked pixel on the core surface is associated with a SOM class from both

infrared spectrometers, with each SOM class representing a variation in the spectral response. Within the

waveform of each SOM classification, there are diagnostic absorption and reflectance features that can be

qualitatively interpreted to facilitate mineral identification, elemental substitutions, crystallinity, and

relative grain sizes.

The combination of SWIR and LWIR provide only qualitative mineralogy. To provide quantitative

mineralogical information, analytic models were developed to calibrate spectral classifications using XRD

data as control points (Greene et al, 2019). Forward modeling was used to validate the fit between the

measured and predicted property values with only a slight mismatch. The mismatch between the two
URTeC
URTeC 154
154 5
datasets is likely related to spatial differences between the two types of measurements: HI measures the

surface of the core while the XRD samples are generally taken from the back of the core, and HI provides

a 2-D measurement which is then averaged to account for the horizontal variability in mineralogy to a

single value at the middle of the core surface. The calibration of hyperspectral image data to XRD mineral

abundances enables continuous single mineral logs to be generated (track 9 in Figure 1), albeit with a

degree of uncertainty associated with their calibrated values. These high-resolution curves display

mineralogical variations with depth that can be plotted alongside petrophysical logs (Kosanke and Chen,

2017). The high resolution of the hyperspectral-derived mineralogy compared to the log-derived

mineralogy is evident by the capture of thin laminae and interbeds (Browning and Kosanke, 2016). The

same approach as above was used to model semi-quantitative TOC abundance in the core.

Results and Discussion

Hyperspectral imaging provided information not readily apparent visually. Digital HI-derived single

mineral curves calibrated to X-ray diffraction (XRD) data were imported as curves to display

mineralogical variations with depth alongside overlays showing the textural relationships of the

mineralogical assemblages, rock typing models, X-ray fluorescence (XRF) data, TOC and rock mechanics

data (Figure 1).

URTeC
URTeC 154
154 6

Figure 1. Digital HI-derived single mineral curves were calibrated to XRD to display mineralogical variations of the Austin Chalk with depth

alongside the core description, XRF, and mechanical data to understand mechanical stratigraphy. For the mineral presence images, warmer

colors indicate higher presence and cooler colors indicate lower presence. For the carbonate composition image, warmer colors indicate higher

Sr-rich calcite phases and cooler colors indicate more pure calcite phases.

Because of the dark character of the core, the SWIR and LWIR are able to detect contrasts in carbonate

composition and subtle in-situ mineralogical variations within the Austin Chalk that cannot be detected by

the human eye. In Figure 1, comparison of the core description on the left side of track 1 with the spectral

rock types in track 2, the hyperspectral mineralogical data in tracks 4 through 8, and the XRF data in track

12, illustrates the high level of mineralogical detail that was obtained from the HI data. In particular, the

HI data:

• Illustrates relative differences in carbonate volumes that identify Milankovitch cycles;

• Delineates fabric via variations in mineralogical composition of fine laminae that can be related

to HI rock types to facilitate upscaling of thin section textural and mineralogical associations to

the core scale;

URTeC
URTeC 154
154 7
• Identifies relatively Sr-rich intervals that cannot be distinguished visually. SEM images show

these intervals to be rich in coccolithophores.

Comparison of the compositional variations in tracks 5, 6 and 9 with the mechanical logs in track 11

provides a continuous, high-resolution understanding of the mechanical stratigraphy of the core, clearly

identifying the relative variations in ‘brittleness’ as being controlled by mineralogy, with brittleness

increasing with carbonate volume and decreasing with clay volume. A hydrocarbon map was not

generated due to the low residual hydrocarbon volume and dessicated nature of the core, but we

were able to use the same calibration method applied to the XRD data, in order to relate the hyperspectral

SOM results to total organic content (TOC) data and produce a continuous TOC curve.

The continuous TOC curve is plotted in track 10 and illustrates that the higher TOC intervals are beds that

are characterized by relatively lower carbonate and higher clay volumes, which is opposite to the trend

observed by the mechanical stratigraphy and suggests that a ‘sweet spot’ exists that balances the ability to

successfully hydraulically fracture the rock with the volume of hydrocarbons in storage.

Conclusions

Hyperspectral imaging provided detailed, high-resolution mineralogical and textural information of a

whole core from the Austin Chalk. This technology produces mineral maps of the surface of a core that

can be used to refine stratigraphic models and explain petrophysical responses. Digital HI-derived single

mineral curves calibrated to XRD can be utilized to display mineralogical variations with depth alongside

open-hole wireline logs and mechanical data to understand mechanical stratigraphy. We conclude that this

technique adds a wealth of data that other methods are unable to provide because of time and cost and

facilitates upscaling of SEM and thin section information to the core scale. Future work will include

evaluation of the mid-range infrared spectra (MWIR) to identify minerals and hydrocarbons in cores from

unconventional resources.
URTeC
URTeC 154
154 8
References

Browning D., and T. Kosanke., 2016, Mineral Mapping of core using combined high-resolution SWIR

and LWIR sensors; presentation Geological Remote Sensing Group London UK December 2016.

Edelman, G., E. Gaston, T. Van Leeuwen, P. Cullen, and M. Aalders, 2012, Hyperspectral imaging for

non-contact analysis of forensic traces: Forensic Science International, v. 223, p. 28–39.

ElMasry, G., and D.-W. Sun, 2010, Principles of hyperspectral imaging technology, in D.-W. Sun, ed.,

Hyperspectral Imaging for Food Quality Analysis and Control: London, Academic Press, Elsevier, p. 3–

43.

Greene, J., T. H. Kosanke, P. Linton , 2019, Quantitative Calibration of Hyperspectral Core Imaging

Data: A New Method for Producing Continuous, High-Resolution Mineralogical Characterization of

Cores from both Conventional and Unconventional Reservoirs, 2019 AAPG Annual Convention and

Exhibition, San Antonio, Texas, May 19-22.

Hunt, G. R., J. W. Salisbury, and C. J. Lenhoff, 1973, Visible and near infrared spectra of minerals and

rocks. VI. Additional silicates: Modern Geology, v. 4, p. 85–106.

Kosanke T. H.j and Chen, J. (2017) Hyperspectral Imaging: Geological and Petrophysical Applications to

Reservoir Characterization, Unconventional Resources Technology Conference, DOI 10.15530urtec2017,

2670537.

Kosanke, T., S. E. Perry, and R. Lopez, 2017, Hyperspectral Imaging Technology Development and

Application; Implications for Thin-Bedded Reservoir Characterization: 2017 AAPG Annual Convention

and Exhibition, Houston, Texas, April 2-5.

Kurz, T., S. Buckley, J. Howell, and D. Schneider, 2008, Geological outcrop modelling and interpretation

using ground based hyperspectral and laser scanning data fusion: International Archives of the

Photogrammetry, Remote Sensing and Spatial Information Sciences, v. 37, p. 1229–1234.

URTeC
URTeC 154
154 9
Kurz, T. H., S. J. Buckley, J. A. Howell, and D. Schneider, 2011, Integration of panoramic hyperspectral

imaging with terrestrial lidar data: The Photogrammetric Record, v. 26, p. 212–228.

Lu, G., and B. Fei, 2014, Medical hyperspectral imaging: a review: Journal of Biomedical Optics, v. 19, p.

010901.

Manea, D., and M. Calin, 2015, Hyperspectral imaging in different light conditions: The Imaging Science

Journal, v. 63, p. 214–219.

Marshall, S., 2011, Hyperspectral imaging technology: AWE International, v. 1, p. 9–15.

Monteiro, S. T., R. J. Murphy, F. Ramos, and J. Nieto, 2009, Applying boosting for hyperspectral

classification of ore-bearing rocks, Machine Learning for Signal Processing XIX—Proceedings of the

2009 IEEE Signal Processing Society Workshop, MLSP 2009, p. 1–6.

Murphy, R. J., and S. T. Monteiro, 2013, Mapping the distribution of ferric iron minerals on a vertical

mine face using derivative analysis of hyperspectral imagery (430–970nm): ISPRS Journal of

Photogrammetry and Remote Sensing, v. 75, p. 29–39.

Plaza, A., J. A. Benediktsson, J. W. Boardman, J. Brazile, L. Bruzzone, G. Camps-Valls, J. Chanussot, M.

Fauvel, P. Gamba, and A. Gualtieri, 2009, Recent advances in techniques for hyperspectral image

processing: Remote Sensing of Environment, v. 113, p. S110–S122.

Sun, D.-W., 2010, Hyperspectral Imaging for Food Quality Analysis and Control, London, Academic

Press, Elsevier.

Taranik, J., and Z. Aslett, 2009, Development of hyperspectral imaging for mineral exploration: Reviews

in Economic Geology, v. 16, p. 83–95.