Solid Earth, 7, 1243–1258, 2016
www.solid-earth.net/7/1243/2016/
doi:10.5194/se-7-1243-2016
© Author(s) 2016. CC Attribution 3.0 License.
On the path to the digital rock physics of gas hydrate-bearing
sediments – processing of in situ synchrotron-tomography data
Kathleen Sell1, Erik H. Saenger2,3, Andrzej Falenty4, Marwen Chaouachi4, David Haberthür5, Frieder Enzmann1,
Werner F. Kuhs4, and Michael Kersten1
1Institute of Geosciences, Johannes Gutenberg University, Mainz, Germany
2International Geothermal Centre, Bochum, Germany
3Ruhr University, Bochum, Germany
4GZG Crystallography, Georg August University, Göttingen, Germany
5Swiss Light Source, Paul Scherrer Institute, Villigen, Switzerland
Correspondence to: Kathleen Sell (sell@uni-mainz.de)
Received: 21 March 2016 – Published in Solid Earth Discuss.: 29 March 2016
Revised: 20 July 2016 – Accepted: 22 July 2016 – Published: 22 August 2016
Abstract. To date, very little is known about the distribution
of natural gas hydrates in sedimentary matrices and its influ-
ence on the seismic properties of the host rock, in particular
at low hydrate concentration. Digital rock physics offers a
unique approach to this issue yet requires good quality, high-
resolution 3-D representations for the accurate modeling of
petrophysical and transport properties. Although such mod-
els are readily available via in situ synchrotron radiation X-
ray tomography, the analysis of such data asks for complex
workflows and high computational power to maintain valu-
able results. Here, we present a best-practice procedure com-
plementing data from Chaouachi et al. (2015) with data post-
processing, including image enhancement and segmentation
as well as exemplary numerical simulations of an acoustic
wave propagation in 3-D using the derived results. A combi-
nation of the tomography and 3-D modeling opens a path to a
more reliable deduction of properties of gas hydrate-bearing
sediments without a reliance on idealized and frequently im-
precise models.
1 Introduction
With the continuous improvement and popularization of
high-resolution visualization methods (Holzer and Cantoni,
2012; Desbois et al., 2013; Hemes et al., 2015; Cnudde and
Boone, 2013; Liu et al., 2016; Deville et al., 2013; Berg et
al., 2013; Pak et al., 2015) digital rock physics gained a firm
foothold in the geophysical world as a potent toolset in stud-
ies of rock properties (e.g., porosity and permeability) and
pore-scale processes like, e.g., electric and heat conductiv-
ity or propagation of acoustic waves (Andrä et al., 2013b, a;
Madonna et al., 2013; Saenger et al., 2011). In this approach,
numerical simulations of transport and petrophysical proper-
ties are conducted on realistic rock representations that ulti-
mately can be upscaled and compared to remote sensing data.
The advantage of this method over traditional, often over-
simplified models lies in a more faithful description of com-
plex pore geometries and microstructures found in natural
formations (Andrä et al., 2013b, a). The most common base
for the numerical simulation is segmented 3-D reconstruc-
tions from various nondestructive X-ray computed tomogra-
phy (X-ray CT) techniques. With a good tradeoff between
the level of detail and investigated sample volume these an-
alytical tools became regularly used on geological samples
at ambient or cryogenic conditions (Murshed et al., 2008;
Klapp et al., 2012; Pak et al., 2015; Berg et al., 2013; Wang
et al., 2015, 2016).
More recently digital rock physics took also on data from
a fairly new group of techniques focused on in situ stud-
ies recreating complex settings that cannot be easily ac-
cessed with conventional means. Two of such difficult en-
vironments are certainly marine and permafrost strata that
are host to crystalline gas–water compounds known as gas
hydrates (GHs) (Sloan and Koh, 2008). Composed predom-
inantly of methane, GHs are a target of intensive geophys-
Published by Copernicus Publications on behalf of the European Geosciences Union.
1244 K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments
ical surveys, drilling and well-logging operations aiming at
the identification and quantification of natural deposits of
this unconventional source of hydrocarbons (Sloan and Koh,
2008; Boswell and Collett, 2011; Moridis et al., 2011). The
identification of potential deposits with remote sensing meth-
ods largely relies on high electric resistivity or characteristic
seismic anomalies in which the increased velocity of seismic
wave is coupled with high attenuations (Best et al., 2010;
Guerin and Goldberg, 2005; Matsushima et al., 2015; Priest
et al., 2006). The latter phenomenon has been interpreted as
losses due to squirt flow in interfacial liquids trapped be-
tween mineral frame and GH crystals (Priest et al., 2006) but
became confirmed only in recent sub-micrometer CT studies
(Chaouachi et al., 2015).
The quantification of GH saturation levels is not straight-
forward as there is very little known about the formation,
microstructure and distribution of hydrates in natural set-
tings, parameters fundamental to the interpretation of geo-
physical exploration. Different habits, distributions and sat-
uration of gas hydrate crystals in the pore space affect the
physical properties of the hydrate-bearing sediment (Priest
et al., 2005; Waite et al., 2004). As the recovery of unper-
turbed natural methane hydrates is very difficult due to their
fast decomposition under ambient conditions, a number of
researchers have attempted to recreate the natural environ-
ment of gas hydrate in sedimentary matrices via laboratory
experiments (Berge et al., 1999; Best et al., 2010, 2013; Dai
et al., 2012; Dvorkin et al., 2003; Ecker et al., 2000; Hu et
al., 2010; Li et al., 2011; Priest et al., 2006, 2009; Spangen-
berg and Kulenkampff, 2006; Yun et al., 2005; Zhang et al.,
2011). This collective effort eventually leads to a set of ideal-
ized micro-structural models (Fig. 1), but the approximations
turned out to be still far from being satisfactory. None of the
simplified models could accurately predict GH saturations
from field electric resistivity or seismic data alone (Waite et
al., 2009; Dai et al., 2012). This might now change with the
advent of high-resolution in situ X-ray CT methods that open
up new a possibility to explore various nucleation and growth
paths of GH in a sedimentary matrix in three dimensions with
a pixel resolution below 1 µm. The application of digital rock
physics to such complex systems is certainly non-trivial due
to a number of constraints imposed by the experimental se-
tups (e.g., fixed sample-detector distance, high attenuation
of the X-ray beam, low contrast), but it still remains a vi-
able way to improve the quantification of GH contents and to
characterize the properties of the system.
The primary challenge is found in the acquisition of high-
quality 3-D models that are capable of capturing fine features
such as micro-cracks, fine porosity and grain-to-grain con-
tacts that could potentially influence the computed properties
(Kerkar et al., 2014; Chaouachi et al., 2015). Another criti-
cal point specific to the seismic anomaly in GH sediments is
a clear identification of the at most a few-micrometers-thick
water film separating mineral frame and GHs. As we demon-
strated in our recent work (Chaouachi et al., 2015; Falenty et
Figure 1. Various distributions of how hydrate might be located in
the sedimentary matrix as discussed in previous studies: encrusta-
tion, cementation, matrix supporting and pore-filling (modified after
Dai et al., 2004).
al., 2015) these high requirements can be met by synchrotron
sources due to their unmatched resolution, high photon den-
sity and tunable energy of the beam. Yet, even in this case
the semi-automatic segmentation and labeling is a difficult
task due to a low density contrast and refractory edge en-
hancement. In this study, we present a workflow for the im-
age enhancement, segmentation and labeling of these data for
further 3-D modeling.
2 Experimental setup and data acquisition
In the following section the in situ experiment and scanning
procedure will be described briefly as details are given in
Chaouachi et al. (2015) and Falenty et al. (2015). As the nu-
cleation and formation of hydrates within the pore space oc-
curs on the nano- to microscale, high-resolution synchrotron
radiation X-ray tomography (SRXCT) was the best experi-
mental choice.
2.1 In situ experiment
SRXCT scans were obtained at the TOMCAT beamline of
the Swiss Light Source (SLS), Paul Scherrer Institute (PSI)
in Villigen, Switzerland (Stampanoni et al., 2006). The high
brilliance, flexible selection of the energy window and excep-
tional resolution of this setup allows for studies of dynamic
processes in complex environmental cells with an excellent
signal-to-noise ratio and fast acquisition time (Fig. 2a). The
formation process was observed in a custom-built in situ
cell composed of (1) an aluminium base, (2) sample holder
mounted on top of the base and (3) TECAPEI (polyether-
imide) dome used to seal the setup (Fig. 2b). Temperature
control is provided by a Peltier element mounted with the
cold side at the bottom of the aluminum base. The tempera-
ture of the sample is actively controlled via a PID controller
Solid Earth, 7, 1243–1258, 2016 www.solid-earth.net/7/1243/2016/
K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments 1245
with long-term stability within±0.1 ◦C. In order to avoid po-
tential vibrations from a pump the hot side of the Peltier el-
ement is cooled with a laminar gravity-driven flow of water.
The pressure inside the cell is measured with an ASHCROFT
KXD pressure sensor and the data were recorded every 5 s
on a computer. The sample holder with a wall thickness of
0.1 mm, an inner diameter of 2 mm and a length of 10 mm
is made also out of aluminum alloy, which ensures a good
thermal conductivity to lower the temperature T to a con-
stant of T = 2 ◦C. The size of the sample holder was cho-
sen to fulfill the size requirements of the numerical modeling
planned thereafter but also to meet the limitations concerning
beam energy attenuation (Fig. 2b). Using methane gas would
lead to considerable additional challenges because of high
required gas pressures during the measurements and a more
difficult data processing due to a low density difference be-
tween water and methane hydrate (Jin et al., 2006). In order
to mitigate these complications, in our experiments methane
was substituted with xenon (Xe) gas, known to be equiva-
lent with respect to the resulting gas hydrate in all impor-
tant physical properties; xenon hydrate forms at very mod-
erate p-T conditions of T = 2 ◦C and a very moderate pres-
sure of p ≥ 0.2 MPa. The ambient atmosphere in the closed
volume is replaced via several compression and decompres-
sion cycles with pure Xe gas while staying below the ther-
modynamic stability boundary of Xe hydrate (Chaouachi et
al., 2015). In this study, the focus is on samples containing
natural quartz sand of about 200–300 µm grain size. Details
on this sedimentary material can be found in Chuvilin et
al. (2011).
2.2 Scanning procedure
For each tomogram, 3201 (or 1501) projections at an in-
tegration time of 150 ms (or 350 ms) each were acquired
over a sample rotation of 180◦ with a monochromatic X-
ray beam energy of 21.9 keV. After penetrating the sam-
ple, X-rays were converted into visible light by a 20 µm
thick scintillator LuAG:Ce and captured by a high-resolution
CCD camera of 2048× 2048 pixels. Two different objec-
tives (UPLAPO10× or UPLAPO20×) with an optical 10-
fold or 20-fold magnification were used, respectively. The
data have been reconstructed out of cam by a gridded Fourier
transform-based algorithm (Marone and Stampanoni, 2012).
The reconstruction process yields in an image matrix of
2560× 2560× 2160 voxels, with an isometric voxel size of
0.74 and 0.38 µm at 10-fold and 20-fold magnification, re-
spectively. Time-resolved scanning was achieved by start-
ing the reaction, stopping it after a prior lab-tested reaction
time by reducing pressure to the stability boundary, with al-
lowance for the system to stabilize, followed by subsequent
scanning (Chaouachi et al., 2015; Falenty et al., 2015). This
procedure is called “stop-and-go in situ tomography”, which
was repeated multiple times to follow consecutive stages of
hydrate formation.
Figure 2. (a) Experimental setup at the TOMCAT beamline. (b) The
pressure cell consisting of the base (1), the sample holder (2) and
the dome (3).
3 Data processing
3.1 Reconstruction
To translate the sinograms to 3-D tomographic images the re-
gridding reconstruction algorithm (Marone and Stampanoni,
2012; Stampanoni et al., 2006) was used. The reconstructed
tomograms revealed that in spite of contrast enhancement
with Xe gas the grey value differences between the gas and
water phase were distinguishable by eye but too low for fur-
ther segmentation approaches. Moreover, a correct visualiza-
tion of the thin fluid layer located in between the grains and
the hydrate (Chaouachi et al., 2015) required a new approach
to enhance the image contrast. For this, the original projec-
tions of a scan were recreated to obtain the original sino-
grams. Following the refraction index of the Henke database
with water as the starting value (Henke et al., 1993) for a
beam energy of 21.9 keV, the material’s refractive index de-
viates from unity by the real part δ = 4.808× 10−7 and by
the imaginary part β = 2.563× 10−10. The Paganin algo-
rithm (Paganin et al., 2002) is able to extract phase infor-
mation in samples, which was tested on the present data to
recover contrast between the xenon and the aqueous phase.
www.solid-earth.net/7/1243/2016/ Solid Earth, 7, 1243–1258, 2016
1246 K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments
Figure 3. Selected results of the approach to enhance the contrast between the gaseous and aqueous phase by applying the reconstruction
algorithm of Paganin on a 2560× 2560 voxel 2-D slice. The insets show a 75× 75 pixel size region of interest, magnified 5×.
A representative sample has been chosen to apply the Pa-
ganin iteration. For this test, the sample to detector distance
was set to 25 mm and the rotation center to 1281.2. In total,
25 different parameters were chosen and applied to various
2-D slices. The reconstruction for each set of parameters did
not meet the expectations to show a large and clear border
between Xenon and water (Fig. 3).
3.2 Image enhancement
At this point the reconstructed images are limited to the spa-
tial resolution and affected by image artifacts but already en-
able one to distinguish between gas, water, sediment grains
and hydrate. Image artifacts that commonly occurred in the
samples were inhomogeneity in grey values, streaks resulting
from bad rotation alignment and edge enhancement.
The first step in data post-processing is to reduce image
noise and scan artifacts using common image filters, cal-
ibrated for the appropriate dimensions and kernel window
sizes (Sell et al., 2015). The software package Avizo 7.1
(FEI, France) has been used for the steps of image filter-
ing and segmentation. One must be aware that every step
of image enhancement changes the original dataset affect-
ing subsequent steps required for data analysis (Sell et al.,
2013b). Working on big datasets often stresses the compu-
tational power to an extent where an effective processing is
not accomplishable anymore. Therefore it is highly recom-
mended to split the datasets into smaller sub-volumes be-
fore taking further action of data post-processing. It is es-
sential to deploy the same enhancement steps and parame-
ters for each digitally sampled sub-volume. Avizo offers sev-
eral options including the extract subvolume function, the
split volume function and the region of interest cropping tool
to gain smaller sub-volumes. The cropping tool is not rec-
ommended for the described situation as it substitutes the
complete dataset by the selected region of interest instead
of gaining several smaller volumes in one attempt. For this
study, the extract subvolume function was the best choice.
As the full datasets each have a known size of 28 GB equal to
2560× 2560× 2160 pixels it was decided to split the dataset
to eight sub-volumes each of a size of 1280× 1280× 1080.
Subsequently the sub-volumes are rated in terms of image
artifacts by considering various slices in x–y, x–z and y–
z direction. Choosing the right image enhancement technique
might require an extensive testing of the available filters prior
Solid Earth, 7, 1243–1258, 2016 www.solid-earth.net/7/1243/2016/
K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments 1247
Figure 4. Overview of a 2-D slice in X–Y direction and the equivalent 3-D rendered volume. The presented sample contains 17 vol % hydrate.
The 2-D slice has a dimension of 2560× 2560 pixels with a pixel size of 0.76 µm2. The volume-rendered image depicts the quartz grains in
grey and the hydrate in yellow – this dataset is suitable for further model investigations. The zoom-in depicts all the phases present in the
samples: xenon – Xe (green), quartz grains, gas hydrate – GH and water.
to further steps. Depending on the dataset and the characteris-
tics of the image artifacts and noise, filters have different im-
pact on the data. Therefore, as we operate at the limiting edge
of gaining image quality and running into data bias, consid-
erable caution must be exerted when treating this multi-phase
system.
Pre-analysis of image enhancement
A study on the effect of various image filters on modeling
result was necessary to determine the best option of filters
aiming at high-quality images to gain volume-rendered im-
ages (Fig. 4). For the pre-study well-known and established
filters from former case studies (Sell et al., 2013a; Madonna
et al., 2013; Shulakova et al., 2013) have been applied to a re-
gion of interest (400× 400× 400 voxels) cropped from the
original raw dataset. Five datasets which derived from sam-
ples containing approximately 17 vol % of hydrate (free pore
water is fully transformation to hydrate) were selected. The
knowledge on the hydrate bulk in the samples will be used as
a proxy later. In the following we provide a list of the applied
filters and settings used in this study.
The concept of the anisotropic diffusion filter (AD) filter is
to smooth out noise in predefined areas of an image, but stop-
ping at sharp edges representing boundaries between phases.
This way, edges and sharp boundaries between phases are
preserved, and image noise is significantly reduced (Kaestner
et al., 2008; Porter and Wildenschild, 2010). A comparison
of the current voxel with the grey values of its six neigh-
bors takes place, and diffusion is fulfilled when the threshold
stop criterion is not exceeded. If the difference between one
voxel and its six adjacent neighbors exceeds the given value
no diffusion takes place. Another option to control the diffu-
sion process of the filter is to reduce or increase the diffusion
time. The parameter number of iterations defines how often
the algorithm will be used on the data. The bigger this num-
ber is, the more blurred is the resulting image. Smoothing is
performed by applying a Gaussian filter. For our investiga-
tions the threshold stop criterion was set to the value 22 968
as this is the approximated transition of the grain phase to
hydrate. AD was run on CPU device with five iterations.
The adaptive histogram equalization filter (AHE) performs
a so-called contrast limited adaptive histogram equalization
(CLAHE) on the dataset. The CLAHE algorithm in Avizo
partitions the images into contextual regions and applies the
histogram equalization to each one. This evens out the dis-
tribution of used grey values and thus makes hidden features
of the image more visible (Reza, 2004). For this study the
brightness was set to 0.716, the contrast to 0.41 and the clip
limit to 7 after trying various values ahead to gain a reason-
able result. The edge detection moments filter (EDM) is a
statistical feature detection to mask out noisy regions or for
edge detection. Here, the brightness was set to 0.5 and con-
trast was set to 2. The edge detection Sobel filter (EDS) filter
is based on the Sobel operator which preserves grain bound-
aries by searching for the most homogeneous feature of the
input data and assigns the averaged value to an elementary
volume where it highlights boundaries between different ma-
terials (Shulakova et al., 2013). For this study, the EDS was
www.solid-earth.net/7/1243/2016/ Solid Earth, 7, 1243–1258, 2016
1248 K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments
Figure 5. Selected filters applied to the original image to enhance the quality for further segmentation steps. Also, the results of the numerical
estimation of hydrate content with the PoroDict module of GeoDict after arithmetic correction are depicted in the upper right corner. Blank
circles indicate poor image quality where segmentation was not accomplishable.
set to a brightness of 0.639 at a contrast of 2. The morpholog-
ical dilatation filter (MD) noise reduction filter was applied
to replace the value of a pixel by the largest value of neigh-
boring pixels. The shape of the neighborhood can be defined
via the input field neighborhood, which was set to 8. The
morphological erosion filter (ME) filter works contrary to the
MD by replacing the value of a pixel by the smallest value of
8 neighboring pixels. The 3-D median filter (M) is a simple
edge-preserving filter which was set to a kernel size of 3. This
filter calculates the mean grey value of neighboring voxels,
in our case 9 since the filter was executed in 3-D mode. Sub-
sequently, the initial voxel is replaced by the resulting mean
grey value. The unsharp masking filter (UM) sharpens an im-
age using an unsharp mask. In the UM approach for image
enhancement, a weighted fraction of the high-pass-filtered
version of the image, which results in the unsharp mask, is
added to the original image (Ramponi, 1999). The unsharp
mask is computed in Avizo by a Gaussian filter of a certain
kernel size. For this study the UM was applied with different
sharpness settings of 0.5 for UM1 and 1 for UM2 at a kernel
size of 3. The non-local means filter (NLM), implemented in
Avizo, is a windowed version of the non-local means algo-
rithm (Buades et al., 2005a, b). The main aim is to denoise
data based on comparing voxels for similarities in a selected
window in which a new weight for a voxel is assigned. Af-
ter a Gauss kernel was run on the weighted values, the new
value will be assigned replacing the former grey values. The
filter is most efficient if the image is affected by white noise.
In Avizo the parameter window size, the local neighborhood
and the similarity value can be customized. The NLM filter is
also an appropriate tool for salt-and-pepper denoising, which
is caused by image sensor defects (Sarker et al., 2012). In this
study, the NLM filter was run in 3-D mode on CPU device.
The search window was equal to 21 and the local neighbor-
hood set to 6 at a similarity value of 0.71.
Adjacent, all filtered datasets were segmented using the
same routine which will be described in the following sec-
tion “Segmentation and volume rendering”. For this, binary
Solid Earth, 7, 1243–1258, 2016 www.solid-earth.net/7/1243/2016/
K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments 1249
Figure 6. Diagram of the results obtained from numerical estimation of the hydrate content in five different samples containing approx. 17 %
controlled by the water available at the initial formation stage.
Figure 7. Segmentation problems occurring in samples of low-grade image enhancement quality.
values of 0, 1 and 2 were assigned to the pore space, grains
and hydrate phase, respectively. Three of the nine filtered
volumes have not been binarized as the necessary segmen-
tation routine was not applicable to result in datasets in-
tended for numerical investigation. The module PoroDict of
the software package GeoDict (Math2Market) was executed
on the binarized images to quantify the hydrate in the filtered
datasets. Even though the effects on image quality may not
be obvious at first glance, further numerical quantification of
the hydrate phase revealed significant differences (Fig. 5).
In Fig. 6 all results regarding the image filter run on five
representative volumes are depicted. According to the nu-
merical results of the described preliminary study, the best
fits to perform image enhancement were the anisotropic dif-
fusion filter (Xhydrate = 16.79 %) and the non-local means fil-
ter (Xhydrate = 17.11 %), since both show good agreement in
consistency and subsequent numerical estimation of the hy-
drate bulk. The final decision was to use both filters in a com-
bined manner on the datasets in 3-D mode, which arose be-
cause neither the AD filter nor the NLM filter worked on all
phases properly. The AD filter caused problems when target-
ing the quartz–gas phase boundary and the NLM filter ham-
pered the segmentation of the quartz–hydrate phase bound-
ary. Both edge detection filters, Sobel (EDS) and oments
www.solid-earth.net/7/1243/2016/ Solid Earth, 7, 1243–1258, 2016
1250 K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments
Figure 8. Example on the right is of a successfully segmented dataset. On the left a representative 2-D slice in X–Y direction and the resulting
binarized image. Size of the 2-D slice is 2560× 2560. Black – pore space, grey – quartz and white – hydrate. This segmented data are ready
to be utilized as a direct model input.
Figure 9. Surface triangulation of quartz grains and the surface view on α-scale transparency.
(EDM), were not suitable for the data since the subsequent
segmentation could not be carried out.
3.3 Segmentation and volume rendering
In this section, we describe the necessary steps of segmenta-
tion for this study. Those steps are crucial if the obtained 3-
D data will be used further as a numerical model input. This
process results in the transformation of voxels of certain grey
scale range. Given that a histogram-based segmentation was
impossible to conduct, we combined the watershed algorithm
and a region growing technique to classify features like phase
distribution (Sell et al., 2015) in the filtered sub-volumes of
every full dataset (each 28 GB). Watershed algorithms are
used to semi-automatically segment tomographic data, treat-
ing grey values of the gradient image as a topographic map
where each grey value stands for a specific altitude. After the
placement of seeds or marker regions in catchment basins
representing minima, which are basically homogeneous ar-
eas with a low gradient, the image is virtually immersed. As
soon as “water” from different basins meets, a so-called wa-
tershed boundary is placed in between (Vincent and Soille,
1991; Wang, 1997). The algorithm performs well, placing
watersheds on sharp intensity edges showing steep gradients
and producing maxima in the gradient image as discussed in
detail in previous literature (Iassonov et al., 2009; Bleau and
Leon, 2000). With this approach it was possible to extract
the hydrate phase and the gas–water phase. The quartz phase
caused major problems because at the rims of the grain ap-
peared strong irregular edge enhancement effects that could
not be removed by filtering (Fig. 7).
As the enhanced edges were of almost the same grey value
as the hydrate phase, masking was applied. First, the void
and the hydrate were segmented using the watershed algo-
rithm; then the resulting binarized volume of two phases was
used to mask the original data in order to extract the quartz
Solid Earth, 7, 1243–1258, 2016 www.solid-earth.net/7/1243/2016/
K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments 1251
Figure 10. Series of hydrate formation visualized in 3-D. For this a region of interest of 400× 400× 400 was picked: (a) initial state where
no hydrate is present; (b) and (c) hydrate formation starts and proceeds from a free-gas environment; (d) full-transformation state; (e) full-
transformation state embedded in the sedimentary matrix. The figure in the center shows all hydrate formation steps within one image where
some steps are set on α-scale transparency to gain information on hydrate structures beneath.
phase, which worked pretty well. Several arithmetic correc-
tions including numerical reconstruction steps and morpho-
logical dilatation and erosion were performed to split phases
before resulting in a satisfying binarized dataset. Initially ev-
ery filtered dataset was segmented into three classes: water
of gas-filled pores, quartz and hydrates. A fully binarized
dataset along the AD-NLM filtered data is shown in Fig. 8.
For the purpose of 3-D visualization, we recommend to cre-
ate a triangulated surface of the binary data (Fig. 9) followed
by the application of volume rendering (Fig. 10). For this we
used the unconstrained smoothing algorithm and the com-
pactify function available in Avizo.
Due to the earlier mentioned low-density contrast issue be-
tween the gaseous and aqueous phase the segmentation of
the residual water film was accomplished slice by slice for
smaller regions of interest using a region growing function
(Fig. 11). For visualization in 3-D a surface creation tool is
applied prior the actual rendering.
3.4 Pre-analysis of the effect of image enhancement on
numerical simulations
Note that the inability to fully characterize the microstruc-
tural details of the material under examination can lead to
disagreements between numerical estimates of mechanical
properties based on SRXCT images and experimentally de-
rived results (Sell et al., 2013b). Such microstructures are
thought to significantly influence the effective elastic proper-
ties of a rock sample. Performing such study directly on the
hydrate-bearing sediments would have taken too many side
effects into account as described earlier (including strong X-
ray attenuation, liquid and hydrate phases). Consequently, we
have decided to perform a pre-analysis to predict the effect of
www.solid-earth.net/7/1243/2016/ Solid Earth, 7, 1243–1258, 2016
1252 K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments
Figure 11. Volume-rendered image of the water film (blue) in be-
tween hydrate (yellow) and the quartz grains – for better under-
standing the outer surface of the grains – is depicted at α-scale
transparency. Voxel resolution for this scan was 0.38 µm.
image enhancement on subsequent modeling based on syn-
chrotron data of a nearly one-phase material. The pre-study
was done on a synchrotron scanned Berea Sandstone dataset,
a well-known “standard” in CT studies and lab investiga-
tions (Andrä et al., 2013b, a). We chose a cylindrical speci-
men of 3 mm diameter, which was also scanned at the TOM-
CAT beamline of the Swiss Light Source at the Paul Scher-
rer Institute in Villigen, Switzerland. The reconstruction pro-
cess yielded an image matrix of 1024× 1024× 1024 voxels,
with a voxel size of 0.74 µm resulting from the selected 10-
fold magnification and a field of view of 1.5× 1.5 mm. Also
in this case artifacts from the scanning and reconstruction
process have been found (e.g., noise, streaks, non-uniform
brightness and edge enhancement). These have been treated
as discussed in the section “Image enhancement”. For fur-
ther investigations a region of interest called Ber1 cropped
to 400× 400× 400 voxels was chosen from the open-source
dataset case 3, which can be downloaded from supplemen-
tary information of Madonna et al. (2013). The 3-D me-
dian filter has been applied to Ber1a, the non-local means
filter to Ber1b and the anisotropic diffusion filter to Ber1c.
All datasets including the original unfiltered have been seg-
mented using the semi-automatic watershed algorithm. Fur-
ther numerical investigations were conducted including the
determination of the total porosity Phi, the permeability K
and the effective P wave velocities. Undoubtedly the varia-
tion in image enhancement and subsequent determination of
grain contact has a strong effect on the obtained rock proper-
ties.
The results showed that the effect of image enhancement
leads to significant variation regarding the assigned grain
contacts (Fig. 12). The total porosities vary from 12 % for the
unfiltered dataset Ber1 to 18 % for the dataset Ber1b where
the image noise was reduced by applying the NLM filter.
Hence the majority of the pores (> 98 %) was found to be
connected on the basis of the open–closed porosity approach
of the PoroDict module of GeoDict. The watershed algorithm
applied on the samples (Ber1a–c) had influence on neither
the porosity nor the permeability tensor. Instead the number
and trend of the modeled grain boundaries by using the bin-
separate function revealed a strong impact on the effective
P wave velocities that varied from 3706 to 5043 m s−1 (Sell
et al., 2013b). All results are depicted in Fig. 10, where a
representative slice (no. 111) of each case is shown, reveal-
ing the effect of filtering and assignment of grain boundaries,
as well as the rock characteristics. Comparing the P wave
velocity, porosity and permeability obtained by digital rock
analysis with results from lab investigations performed on
this Berea sample (Madonna et al., 2012), we found the best-
matching dataset to be Ber1b filtered with NLM. Digital rock
analysis indicated a total connected porosity of 18 % and a
permeability of 150 mD for Ber1b where the lab results re-
vealed a connected porosity of ∼ 20 %. The permeability of
this sample, as provided by the Berea Sandstone™ Petroleum
Cores (Ohio, USA), is between 200 and 500 mD. The worst-
matching datasets during the determination of effective phys-
ical properties were the unfiltered and the median filtered one
with a total porosity of 12 % and a permeability of 80 mD
(Sell et al., 2013b).
4 Wave propagation modeling in hydrate-bearing
sedimentary matrices
In this section, we present a first application of our derived
post-processing data to wave propagation modeling. A series
of numerical experiments based on segmented 3-D images
of the structure has been conducted to obtain the P wave ve-
locity trend with increase or decrease of the hydrate bulk.
This has been done to validate our model approach to aim
for further wave propagation modeling including S wave ve-
locity and attenuation in future studies. As mentioned ear-
lier, methane hydrate has been substituted by Xe hydrate
that exhibits far reaching similarities in crystal structure and
molecular size of guests (Table 1). Yet, in terms of the com-
pressibility these clathrates are noticeably different; struc-
tures hosting Xe gas are about 10 % stiffer than in the case
of CH4 (Hansen et al., 2016). The difference originates in
guest–host coupling which is strong for tightly fitting Xe
and weaker for somewhat smaller methane that freely rattles
in cages (Chazallon et al., 2002; Schober et al., 2003). An-
other potential difference might be found in the crystal size
for both clathrates due to somewhat different density of nu-
cleation sites. Yet, without a definite proof we must assume
Solid Earth, 7, 1243–1258, 2016 www.solid-earth.net/7/1243/2016/
K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments 1253
Figure 12. This figure comprises the results of the preliminary study on the effect of filtering on the modeling of effective elastic properties.
Representative 2-D slices (#111 of the X–Y plane) are depicted.
for the moment that the hydrate nucleation and crystalliza-
tion processes of both clathrates in the sedimentary matrix
are comparable. We also take digitized Xe hydrate distribu-
tion and microstructure obtained from SRXCT and assign to
it elastic moduli of CH4 hydrate: bulk modulus (K), shear
modulus (G), P wave modulus (M) and the density (ρ).
We applied the rotated staggered finite-difference method
appropriate for dynamic measurements. The fundamental
idea of the model is discussed in detail by Saenger et
al. (2004). This approach studies the wave propagation in
heterogeneous materials within the long wavelength limit.
Using this method it is possible to predict precisely effective
elastic properties of various pore geometries in the relatively
wide range between the upper and lower Hashin–Shtrikman
bound. Complex structures, as observed in our samples, trig-
ger strong scattering which can be treated only by numerical
techniques since an analytical solution of the wave equation
is not available. Finite difference (FD) methods discretize the
wave equation on a grid and replace spatial derivatives by FD
Table 1. Comparison of crystallographic properties of xenon and
methane hydrate (Hansen et al., 2016). Second part of the table de-
picts the elastic moduli later assigned for modeling the effective
elastic properties. Values of M (P wave modulus), K (bulk modu-
lus) and G (shear modulus) (after Helgerud et al., 2003).
Xe hydrate CH4 hydrate Quartz
Hydrate structure sI sI
Cage diameter (Å) 5.09–5.86 5.1–5.85
Lattice constant 11.988 11.991
at 276 K (Å)
Molecular size 4.58 4.36
M (GPa)
 1 
 2 
 3 
Table 1: Comparison of crystallographic properties of xenon- and methane hydrate (Hansen et al., 2016). Second part 4 
of the table depicts the elastic moduli latter assigned for modelling the effective elastic properties. Values of M (P-5 
wave modulus), K (Bulk modulus) and G (Shear modulus) after (Helgerud et al., 2003).  6 
 7 
 8 
13.57 96.98
K (GPa) 8.76 37.8
G (GPa) 3.57 44.3
ρ (kg m−3) 982 2648
www.solid-earth.net/7/1243/2016/ Solid Earth, 7, 1243–1258, 2016
1254 K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments
Figure 13. (1) Visualization of the RSG model where A is a predefined homogenous solid in which the 400× 400× 400 size volume is em-
bedded. The plane wave propagates in z direction. (2) Visualization of the Vp in a sample containing roughly 7 % of hydrate. (3) Visualization
of the Vp in a sample bearing 17 % of hydrate where the Vp is significantly higher due to the increased hydrate bulk.
operators using neighboring points. Instability problems on
a staggered grid may be caused by discretization when the
medium contains high-contrast discontinuities (e.g., pores or
fractures). These difficulties can be avoided by using the
rotated staggered grid technique proposed by Saenger and
Bohlen (2004). Since the finite-difference approach is based
on the wave equation without physical approximations, the
method accounts not only for direct waves, primary and mul-
tiple reflected waves, but also for surface waves, head waves,
converted reflected waves, and diffracted waves (Saenger et
al., 2004).
Furthermore, the application of this method has already
been benchmarked (Andrä et al., 2013a). On top of the 3-D
rock model a body force plane source is applied using a
homogeneous buffer zone of assigned vacuum (Sell et al.,
2015). The plane wave travels through the embedded digi-
tized 3-D sample volume (Fig. 13.1). Two plane receivers on
the top and the bottom of the model measure the time delay
of the plane wave’s peak amplitude caused by the inhomo-
geneity of the rock (Madonna et al., 2013). As the model
input is chosen to a volume of 400× 400× 400 pixels due to
computational limits, the segmented datasets needed to be
reduced in size. This was accomplished by using the “re-
sample” tool based on the Lanczos filter implemented in
Avizo. Usually the default kernel Lanczos yields sufficiently
good results for both minifications and magnifications. In
the present case a cropped dataset of 1800× 1800× 1800
voxels each of (0.74 µm)3 in size was reduced to a cube of
400× 400× 400 voxels each of (3.4 µm)3 in size.
Effective velocities of the hydrate-bearing sedimentary
matrices are determined by comparing the simulated results
with those of a reference model. Note that the quartz grains
have not been compacted during sample preparation wherein
the hydrate is located in the pore space, which causes the
structure to be stiffer with an increase in hydrate saturation
(Fig. 13.2 and 13.3).
As the stop-and-go scanning procedure (Chaouachi et al.,
2015; Falenty et al., 2015) permits a time-resolved observa-
tion of hydrate formation, the numerically obtained P wave
from the scans confirm the expected trend of an increase
of P wave velocities with higher hydrate bulk in the sedi-
ment (Priest et al., 2009). For each post-processed scan the
P wave velocity has been determined following the work-
flow described in this paper. In Fig. 13.2 the P wave velocity
simulated at the first time step is visualized for a sample with
7 % of hydrate. The simulated P wave velocity for a sam-
ple containing 17 % of hydrate is visualized in Fig. 13.3. The
comparison of both figures shows clearly how the P wave
velocity increases due to a higher hydrate bulk as the GH
holds the quartz in place and stiffens the structure.
In Fig. 14, the results of the simulated Vp for series of eight
scans are depicted. In absence of hydrate (scan 5) the lowest
P wave velocity of 1462 ms−1 is given for the model series.
This P wave value is typical for partially water-saturated
sands, which is also reasonable. The highest P wave veloci-
Solid Earth, 7, 1243–1258, 2016 www.solid-earth.net/7/1243/2016/
K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments 1255
Figure 14. Results of the p wave modeling of a series of scans starting at the initial state forming hydrate from a free gas system up to the
full-transformed state, followed by a controlled destabilization of the system to maintain hydrate dissociation from which again the hydrate
formation was triggered up to a full-transformed state in a gas-enriched system.
ties have been obtained for the full-transformed states as ex-
pected (scans 1 and 8). Scan 1 was taken when hydrate was
formed from the free-gas system starting at the gas–water in-
terface and scan 8 when hydrate was formed from a system
where liquid water is metastable enriched with a gas (gas-
enriched system). For the first case the obtained P wave ve-
locity is 2368 ms−1 and for the second case it is 2408 ms−1.
Therefore, at the full-transformed state of hydrate formed in a
free-gas system (scan 1) the P wave velocity is slightly lower
than at the full-transformed state of hydrate formed in a gas-
rich system (scan 8), which has been discussed earlier in the
literature (Priest et al., 2005; Konno et al., 2015; Waite et al.,
2009). With the presented results of the modeled P wave ve-
locities the model approach is in a realistic range when com-
pared with field (Carcione and Gei, 2004; Riedel et al., 2002;
Yuan et al., 1996) and laboratory data (Zhang et al., 2011;
Priest et al., 2005, 2009), but it is noteworthy that the mod-
eled results give only slightly lower values than the experi-
mental ones, even though at significantly lower hydrate sat-
uration (< 20 %) compared to results from laboratory work.
Furthermore, intensive modeling on various scans needs to
be performed to fully understand this apparently contrasting
observation. In addition, one should take into account that
investigations on seismic responses conducted in laboratory
often use pre-compacted samples (Matsushima et al., 2015)
while in our study we used unconsolidated samples. More-
over, there are uncertainties with the elastic response of Xe
hydrate, likely to be somewhat different from the one of CH4
hydrate, which was taken for our analysis.
5 Conclusions
A novel processing approach, based on tomographic data
with sub-micrometer scale resolution, is presented which al-
lows one to obtain in the next step the effective elastic prop-
erties of hydrate-bearing sediments. This paper complements
the paper of Chaouachi et al. 2015 with a sufficient work-
flow to process 3-D image data derived from synchrotron-
based tomography to gain digitized results that are ready for
subsequent use as a direct numerical model input. The data
processing requires a pre-analysis to aim for best possible
image quality enhancement. In our opinion, crucially impor-
tant when aiming for multiphase 3-D imaging on the sub-
microscale, involving low-contrast materials, is the scanning
itself and the image enhancement. Regarding the scanning
part it is important to avoid scan artifacts, e.g., rotation errors
which are not discussed in this paper but worth mentioning.
The scanning procedure needs intensive testing prior to the
experiment. All actions on image enhancement must be car-
ried out with great care as the filters applied significantly al-
ter the latter segmentation results as stated in the pre-analysis
section of this paper. We highly recommend to carry out pre-
studies ahead of every digital rock physics approach where
segmented data are needed for further modeling. Applying
our proposed method it is possible to approximate the elas-
tic properties of sediments containing less than 20 % of hy-
drate within a realistic range at the sub-micrometer scale; a
more detailed analysis and intercomparison is under way. As
the time-resolved images produced a time series of hydrate
formed from different scenarios (free-gas and gas-enriched),
we are also able to distinguish between those cases. Yet, one
should keep in mind that initially different microstructural
settings (as a consequence of the different preparation routes)
www.solid-earth.net/7/1243/2016/ Solid Earth, 7, 1243–1258, 2016
1256 K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments
may well converge with time to closely similar distributions
of gas hydrates in the sedimentary matrix (Chaouachi et al.,
2015). Laboratory work usually is not converged when ex-
periments are executed while natural samples are in general
more mature. Ongoing work is focusing in more detail on
the P wave, S wave velocities and attenuation since the pres-
ence of a thin fluid layer observed in the structures could
explain the anomalous seismic responses of hydrate-bearing
sediments observed in field studies. For this, a satisfying so-
lution to segment the thin fluid layer in all scans is manda-
tory. In addition, we need to investigate further the contrast-
ing observation of rather high results of P wave velocities at
low gas hydrate saturation.
Acknowledgements. We acknowledge the Paul Scherrer Institute,
Villigen, Switzerland, for provision of synchrotron radiation
beam time at the TOMCAT beamline of the SLS and would
like to thank Bernd Pinzer for assistance. We are grateful to
Jens-Oliver Schwarz, Martin Wolf and Faisal Khan (all formerly at
JGU Mainz) for their valuable help during these experiments. The
pressure cell was designed and constructed by Ulf Kahmann and
Heiner Bartels (both from Göttingen). The financial support from
the Deutsche Forschungsgemeinschaft (DFG grants Ke 508/20 and
Ku 920/18) is gratefully acknowledged. Support also came from
the BMBF in the framework of the SUGAR-II program (grant
03G0819B).
Edited by: S. Henkel
Reviewed by: M. Schindler and two anonymous referees
References
Andrä, H., Combaret, N., Dvorkin, J., Glatt, E., Han, J., Kabel,
M., Keehm, Y., Krzikalla, F., Lee, M., Madonna, C., Marsh, M.,
Mukerji, T., Saenger, E. H., Sain, R., Saxena, N., Ricker, S.,
Wiegmann, A., and Zhan, X.: Digital rock physics benchmarks
– Part II: Computing effective properties, Comput. Geosci., 50,
33–43, 2013a.
Andrä, H., Combaret, N., Dvorkin, J., Glatt, E., Han, J., Kabel,
M., Keehm, Y., Krzikalla, F., Lee, M., Madonna, C., Marsh, M.,
Mukerji, T., Saenger, E. H., Sain, R., Saxena, N., Ricker, S.,
Wiegmann, A., and Zhan, X.: Digital rock physics benchmarks –
Part I: Imaging and segmentation, Comput. Geosci., 50, 25–32,
2013b.
Berg, S., Ott, H., Klapp, S. A., Schwing, A., Neiteler, R., Brussee,
N., Makurat, A., Leu, L., Enzmann, F., Schwarz, J.-O., Kersten,
M., Irvine, S., and Stampanoni, M.: Real-time 3d imaging of
haines jumps in porous media flow, P. Natl. Acad. Sci. USA, 110,
3755–3759, 2013.
Berge, L. I., Jacobsen, K. A., and Solstad, A.: Measured acoustic
wave velocities of r11 (ccl3f) hydrate samples with and without
sand as a function of hydrate concentration, J. Geophys. Res.-
Solid Earth, 104, 15415–15424, 1999.
Best, A., Priest, J., and Clayton, C.: A resonant column study of
the seismic properties of methane-hydrate-bearing sand, in: Geo-
physical characterization of gas hydrates, edited by: Riedel, M.,
Willoughby, E. C., and Chopra, S., Society of Exploration Geo-
physicists, 337–347, 2010.
Best, A. I., Priest, J. A., Clayton, C. R. I., and Rees, E. V. L.: The
effect of methane hydrate morphology and water saturation on
seismic wave attenuation in sand under shallow sub-seafloor con-
ditions, Earth Planet. Sc. Lett., 368, 78–87, 2013.
Bleau, A. and Leon, L. J.: Watershed-based segmentation and re-
gion merging, Comput. Vis. Image Und., 3, 317–370, 2000.
Boswell, R. and Collett, T. S.: Current perspectives on gas hydrate
resources, Energ. Environ. Sci., 4, 1206–1215, 2011.
Buades, A., Coll, B., and Morel, J.-M.: A non-local algorithm for
image denoising, IEEE Computer Society Conference on Com-
puter Vision and Pattern Recognition, San Diego, CA, USA, 60–
65, 2005a.
Buades, A., Coll, B., and Morel, J. M.: A review of image denoising
algorithms, with a new one, Multiscale Model. Sim., 4, 490–530,
2005b.
Carcione, J. M. and Gei, D.: Gas-hydrate concentration estimated
from p- and s-wave velocities at the mallik 2l–38 research well,
mackenzie delta, Canada, J. Appl. Geophys., 1, 73–78, 2004.
Chaouachi, M., Falenty, A., Sell, K., Enzmann, F., Kersten, M.,
Haberthür, D., and Kuhs, W. F.: Microstructural evolution of
gas hydrates in sedimentary matrices observed with synchrotron
x-ray computed tomographic microscopy, Geochem. Geophy.
Geosy., 16, 1711–1722, 2015.
Chazallon, B., Itoh, H., Koza, M., Kuhs, W. F., and Schober, H.: An-
harmonicity and guest–host coupling in clathrate hydrates, Phys.
Chem. Chem. Phys., 4, 4809–4816, 2002.
Chuvilin, E. M., Istomin, V. A., and Safonov, S. S.: Residual non-
clathrated water in sediments in equilibrium with gas hydrate
comparison with unfrozen water, Cold Reg. Sci. Technol., 68,
68–73, 2011.
Cnudde, V. and Boone, M. N.: High-resolution x-ray computed to-
mography in geosciences: A review of the current technology
and applications, Earth-Sci. Rev., 123, 1–17, 2013.
Dai, J., Xu, H., Snyder, F., and Dutta, N.: Detection and estimation
of gas hydrates using rock physics and seismic inversion: Exam-
ples from the northern deepwater gulf of mexico, The Leading
Edge, 23, 60–66, 2004.
Dai, S., Santamarina, J. C., Waite, W. F., and Kneafsey, T. J.: Hy-
drate morphology: Physical properties of sands with patchy hy-
drate saturation, J. Geophys. Res.-Solid Earth, 117, B11205,
doi:10.1029/2012jb009667, 2012.
Desbois, G., Urai, J. L., PÉRez-Willard, F., Radi, Z., Offern, S.,
Burkart, I., Kukla, P. A., and Wollenberg, U.: Argon broad ion
beam tomography in a cryogenic scanning electron microscope:
A novel tool for the investigation of representative microstruc-
tures in sedimentary rocks containing pore fluid, J. Microscopy,
249, 215–235, 2013.
Deville, S., Adrien, J., Maire, E., Scheel, M., and Di Michiel,
M.: Time-lapse, three-dimensional in situ imaging of ice crystal
growth in a colloidal silica suspension, Acta Mater., 61, 2077–
2086, 2013.
Dvorkin, J., Nur, A., Uden, R., and Taner, T.: Rock physics of a gas
hydrate reservoir, The Leading Edge, 22, 842–847, 2003.
Ecker, C., Dvorkin, J., and Nur, A. M.: Estimating the amount of
gas hydrate and free gas from marine seismic data, Geophysics,
65, 565–573, 2000.
Solid Earth, 7, 1243–1258, 2016 www.solid-earth.net/7/1243/2016/
K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments 1257
Falenty, A., Chaouachi, M., Neher, S. H., Sell, K., Schwarz, J.-O.,
Wolf, M., Enzmann, F., Kersten, M., Haberthur, D., and Kuhs,
W. F.: Stop-and-go in situ tomography of dynamic processes –
gas hydrate formation in sedimentary matrices, Acta Crystallogr.
A, 71, s154, doi:10.1107/S2053273315097740, 2015.
Guerin, G. and Goldberg, D.: Modeling of acoustic wave dissipation
in gas hydrate-bearing sediments, Geochem. Geophy. Geosy., 6,
Q07010, doi:10.1029/2005gc000918, 2005.
Hansen, T. C., Falenty, A., and Kuhs, W. F.: Lattice constants and
expansivities of gas hydrates from 10 k up to the stability limit,
J. Chem. Phys., 144, 054301, doi:10.1063/1.4940729, 2016.
Helgerud, M. B., Waite, W. F., Kirby, S. H., and Nur, A.: Measured
temperature and pressure dependence of vp and vs in compacted,
polycrystalline si methane and sii methane-ethane hydrate, Can.
J. Phys., 81, 47–53, 2003.
Hemes, S., Desbois, G., Urai, J. L., Schröppel, B., and Schwarz, J.-
O.: Multi-scale characterization of porosity in boom clay (hades-
level, mol, belgium) using a combination of x-ray µ-ct, 2d bib-
sem and fib-sem tomography, Micropor. Mesopor. Mat., 208, 1–
20, 2015.
Henke, B. L., Gullikson, E. M., and Davis, J. C.: X-ray interac-
tions: Photoabsorption, scattering, transmission, and reflection
at e = 50–30 000 ev, z= 1–92, Atom. Data Nucl. Data, 54, 181–
342, 1993.
Holzer, L. and Cantoni, M.: Review of fib-tomography, in: Nanofab-
rication using focused ion and electron beams: Principles and ap-
plications, edited by: Utke, I., Moshkalev, S. A., and Russell, P.,
Oxford University Press, NY, USA, 410–435, 2012.
Hu, G. W., Ye, Y. G., Zhang, J., Liu, C. L., Diao, S. B., and
Wang, J. S.: Acoustic properties of gas hydrate-bearing con-
solidated sediments and experimental testing of elastic ve-
locity models, J. Geophys. Res.-Solid Earth, 115, B02102,
doi:10.1029/2008jb006160, 2010.
Iassonov, P., Gebrenegus, T., and Tuller, M.: Segmentation of x-ray
computed tomography images of porous materials: A crucial step
for characterization and quantitative analysis of pore structures,
Water Resour. Res., 9, W09415, doi:10.1029/2009WR008087,
2009.
Jin, S., Nagao, J., Takeya, S., Jin, Y., Hayashi, J., Kamata, Y.,
Ebinuma, T., and Narita, H.: Structural investigation of methane
hydrate sediments by microfocus x-ray computed tomography
technique under high-pressure conditions, Jpn. J. Appl. Phys., 2,
L714–L716, 2006.
Kaestner, A., Lehmann, E., and Stampanoni, M.: Imaging and im-
age processing in porous media research, Adv. Water Resour.,
31, 1174–1187, 2008.
Kerkar, P. B., Horvat, K., Jones, K. W., and Mahajan, D.: Imaging
methane hydrates growth dynamics in porous media using syn-
chrotron x-ray computed microtomography, Geochem. Geophy.
Geosy., 15, 4759–4768, 2014.
Klapp, S. A., Enzmann, F., Walz, P., Huthwelker, T., Tuckermann,
J., Schwarz, J. O., Pape, T., Peltzer, E. T., Mokso, R., Wangner,
D., Marone, F., Kersten, M., Bohrmann, G., Kuhs, W. F., Stam-
panoni, M., and Brewer, P. G.: Microstructure characteristics dur-
ing hydrate formation and dissociation revealed by x-ray tomo-
graphic microscopy, Geo-Mar. Lett., 32, 555–562, 2012.
Konno, Y., Jin, Y., Yoneda, J., Kida, M., Egawa, K., Ito, T., Suzuki,
K., and Nagao, J.: Effect of methane hydrate morphology on
compressional wave velocity of sandy sediments: Analysis of
pressure cores obtained in the eastern nankai trough, Mar. Petrol.
Geol., 66, 425–433, 2015.
Li, F. G., Sun, C. Y., Zhang, Q., Liu, X. X., Guo, X. Q.,
and Chen, G. J.: Laboratory measurements of the effects of
methane/tetrahydrofuran concentration and grain size on the
p wave velocity of hydrate-bearing sand, Energ. Fuels, 25, 2076–
2082, 2011.
Liu, Y., Kiss, A. M., Larsson, D. H., Yang, F., and Pianetta, P.: To
get the most out of high resolution x-ray tomography: A review
of the post-reconstruction analysis, Spectrochim. Acta B, 117,
29–41, 2016.
Madonna, C., Almqvist, B. S. G., and Saenger, E. H.: Digital rock
physics: Numerical prediction of pressure-dependent ultrasonic
velocities using micro-ct imaging, Geophys. J. Int., 3, 1475–
1482, 2012.
Madonna, C., Quintal, B., Frehner, M., Almqvist, B. S. G., Tisato,
N., Pistone, M., Marone, F., and Saenger, E. H.: Synchrotron-
based x-ray tomographic microscopy for rock physics investi-
gations, Geophysics, 78, 53–64, doi:10.1190/geo2012-0113.1,
2013.
Marone, F. and Stampanoni, M.: Regridding reconstruction algo-
rithm for real-time tomographic imaging, J. Synchrotron Radiat.,
19, 1029–1037, 2012.
Matsushima, J., Suzuki, M., Kato, Y., and Rokugawa, S.: Ultrasonic
compressional and shear attenuation measurements in partially
frozen consolidated and unconsolidated porous media, J. Geo-
phys. Res., 81, 141–153, 2015.
Moridis, G. J., Collett, T. S., Pooladi-Darvish, M., Hancock, S.,
Santamarina, C., Boswell, R., Kneafsey, T., Rutqvist, J., Kowal-
sky, M. B., Reagan, M. T., Sloan, E. D., Sum, A. K., and Koh, C.
A.: Challenges, uncertainties, and issues facing gas production
from gas-hydrate deposits, SPE Reserv. Eval. Eng., 14, 76–112,
2011.
Murshed, M. M., Klapp, S. A., Enzmann, F., Szeder, T., Huth-
welker, T., Stampanoni, M., Marone, F., Hintermuller, C.,
Bohrmann, G., Kuhs, W. F., and Kersten, M.: Natural gas
hydrate investigations by synchrotron radiation x-ray cryo-
tomographic microscopy (srxctm), Geophys. Res. Lett., 35,
L23612, doi:10.1029/2008gl035460, 2008.
Paganin, D., Mayo, S. C., Gureyev, T. E., Miller, P. R., and Wilkins,
S. W.: Simultaneous phase and amplitude extraction from a
single defocused image of a homogeneous object, J. Microsc.-
Oxford, 206, 33–40, 2002.
Pak, T., Butler, I. B., Geiger, S., van Dijke, M. I. J., and Sorbie, K.
S.: Droplet fragmentation: 3d imaging of a previously unidenti-
fied pore-scale process during multiphase flow in porous media,
P. Natl. Acad. Sci. USA, 112, 1947–1952, 2015.
Porter, M. L. and Wildenschild, D.: Image analysis algorithms for
estimating porous media multiphase flow variables from com-
puted microtomography data: A validation study, Computat.
Geosci., 14, 15–30, 2010.
Priest, J. A., Best, A. I., and Clayton, C. R. I.: A laboratory in-
vestigation into the seismic velocities of methane gas hydrate-
bearing sand, J. Geophys. Res.-Solid Earth, 110, B04102,
doi:10.1029/2004jb003259, 2005.
Priest, J. A., Best, A. I., and Clayton, C. R. I.: Attenuation of seis-
mic waves in methane gas hydrate-bearing sand, Geophys. J. Int.,
164, 149–159, 2006.
www.solid-earth.net/7/1243/2016/ Solid Earth, 7, 1243–1258, 2016
1258 K. Sell et al.: On the path to the digital rock physics of gas hydrate-bearing sediments
Priest, J. A., Rees, E. V. L., and Clayton, C. R. I.: In-
fluence of gas hydrate morphology on the seismic veloci-
ties of sands, J. Geophys. Res.-Solid Earth, 114, B11205,
doi:10.1029/2009jb006284, 2009.
Ramponi, G.: Warped distance for space-variant linear image inter-
polation, IEEE Transact. Image Proc., 5, 629–639, 1999.
Reza, A. M.: Realization of the contrast limited adaptive histogram
equalization (clahe) for real-time image enhancement, J. VLSI
Sig. Proc. Syst., 38, 35–44, 2004.
Riedel, M., Spence, G. D., Chapman, N. R., and Hyndman, R. D.:
Seismic investigations of a vent field associated with gas hy-
drates, offshore vancouver island, J. Geophys. Res.-Solid Earth,
107, 2200, doi:10.1029/2001JB000269, 2002.
Saenger, E. H., Enzmann, F., Keehm, Y., and Steeb, H.: Digital rock
physics: Effect of fluid viscosity on effective elastic properties,
J. Appl. Geophys., 4, 236–241, 2011.
Saenger, E. H. and Bohlen, T.: Finite-difference modeling of vis-
coelastic and anisotropic wave propagation using the rotated
staggered grid, Geophysics, 69, 583–591, 2004.
Saenger, E. H., Sell K., and Kersten, M.: On the Effect of Image
Enhancement Techniques on Digital Rock Physics Results, 75th
EAGE Conference & Exhibition incorporating SPE EUROPEC
2013, Extended Abstract, London, UK, 2013.
Saenger, E. H., Krüger, O. S., and Shapiro, S. A.: Numeri-
cal considerations of fluid effects on wave propagation: In-
fluence of the tortuosity, Geophys. Res. Lett., 31, L21613,
doi:10.1029/2004gl020970, 2004.
Sarker, S., Chowdhury, S., Laha, S., and Dey, D.: Use of non-local
means filter to denoise image corrupted by salt and pepper noise,
SIPIJ Signal & Image Processing: An International Journal, 3,
223–235, 2012.
Schober, H., Itoh, H., Klapproth, A., Chihaia, V., and Kuhs, W.
F.: Guest-host coupling and anharmonicity in clathrate hydrates,
Eur. Phys. J. E, 12, 41–49, 2003.
Sell, K., Enzmann, F., Spangenberg, E., and Kersten, M.: CO2 injec-
tion to a saline aquifer sandstone – clay mineral displacement and
permeability changes, 75th EAGE Conference & Exhibition in-
corporating SPE EUROPEC 2013, Extended Abstract, London,
UK, 2013a.
Sell, K., Madonna, C., Quintal, B., Frehner, M., Tisato, N., and
Saenger, E. H.: Synchrotron-based x-ray tomographic images
and segmentation techniques to account for effects of grain con-
tacts and micro-cracks on rock properties, SBGf International
Congress, Rio de Janeiro, Brazil, 26–29 August, 2013b.
Sell, K., Chaouachi, M., Falenty, A., Saenger, E. H., Khan, F.,
Schwarz, J.-O., Enzmann, F., Kersten, M., and Kuhs, W. F.: Mi-
crostructure of hydrate-bearing sediments and determination of
p-wave velocities based on high-resolution synchrotron tomo-
graphic data, SEG Technical Meeting and Conference, New Or-
leans, LA, USA, 2015.
Shulakova, V., Pervukhina, M., Müller, T. M., Lebedev, M., Mayo,
S., Schmid, S., Golodoniuc, P., De Paula, O. B., Clennell, M.
B., and Gurevich, B.: Computational elastic up-scaling of sand-
stone on the basis of x-ray micro-tomographic images, Geophys.
Prospect., 61, 287–301, 2013.
Sloan, E. D. and Koh, C. A.: Clathrate hydrates of natural gases, 3th
Edn., Vol. 119, CRC Press, Boca Raton, Fla, 2008.
Spangenberg, E. and Kulenkampff, J.: Influence of methane hydrate
content on electrical sediment properties, Geophys. Res. Lett.,
33, L24315, doi:10.1029/2006gl028188, 2006.
Stampanoni, M., Groso, A., Isenegger, A., Mikuljan, G., Chen, Q.,
Bertrand, A., Henein, S., Betemps, R., Frommherz, U., Böhler,
P., Meister, D., Lange, M., and Abela, R.: Trends in synchrotron-
based tomographic imaging: The sls experience, SPIE Proceed-
ings, 2006,
Vincent, L. and Soille, P.: Watersheds in digital spaces: An effi-
cient algorithm based on immersion simulations, IEEE T. Pattern
Anal., 6, 583–598, 1991.
Waite, W. F., Winters, W. J., and Mason, D. H.: Methane hydrate
formation in partially water-saturated ottawa sand, Am. Mineral.,
89, 1202–1207, 2004.
Waite, W. F., Santamarina, J. C., Cortes, D. D., Dugan, B., Espinoza,
D. N., Germaine, J., Jang, J., Jung, J. W., Kneafsey, T. J., Shin,
H., Soga, K., Winters, W. J., and Yun, T. S.: Physical proper-
ties of hydrate-bearing sediments, Rev. Geophys., 47, RG4003,
doi:10.1029/2008rg000279, 2009.
Wang, D.: A multiscale gradient algorithm for image segmentation
using watersheds, Pattern Recogn., 12, 2043–2052, 1997.
Wang, J.-Q., Zhao, J.-F., Yang, M.-J., Li, Y.-H., Liu, W.-G., and
Song, Y.-C.: Permeability of laboratory-formed porous media
containing methane hydrate: Observations using x-ray computed
tomography and simulations with pore network models, Fuel,
145, 170–179, 2015.
Wang, J., Zhao, J., Zhang, Y., Wang, D., Li, Y., and Song, Y.: Anal-
ysis of the effect of particle size on permeability in hydrate-
bearing porous media using pore network models combined with
ct, Fuel, 136, 34–40, 2016.
Yuan, T., Hyndman, R. D., Spence, G. D., and Desmons, B.: Seis-
mic velocity increase and deep-sea gas hydrate concentration
above a bottom-simulating reflector on the northern cascadia
continental slope, J. Geophys. Res.-Solid Earth, 101, 13655–
13671, 1996.
Yun, T. S., Francisca, F. M., Santamarina, J. C., and Ruppel, C.:
Compressional and shear wave velocities in uncemented sedi-
ment containing gas hydrate, Geophys. Res. Lett., 32, L10609,
doi:10.1029/2005gl022607, 2005.
Zhang, Q., Li, F. G., Sun, C. Y., Li, Q. P., Wu, X. Y., Liu, B., and
Chen, G. J.: Compressional wave velocity measurements through
sandy sediments containing methane hydrate, Am. Mineral., 96,
1425–1432, 2011.
Solid Earth, 7, 1243–1258, 2016 www.solid-earth.net/7/1243/2016/