Fracture Reconstruction and Analysis of Low Permeability Carbonates using x-ray Tomography for Comparison with Outcrop Data

Comparing outcrop data to laboratory results is important to verify and validate experiments of analogue and reservoir materials especially regarding conditions for deformation experiments. This is important better understand highly complex carbonate reservoir strata and their response to changes in subsurface conditions, reducing subsurface uncertainty. This study develops methods to allow for a more straightforward comparison of outcrop data (m-scale) with experimentally created fracture arrays developed in cylindrical samples (cm-scale). The main objective is to assess usefulness of experimentally-produced fracture networks as analogues for subsurface structures, typically at the meter and above scale by developing new techniques to use the lab deformation. It analyses key characteristics of laboratory-induced fracture networks by adapting scanline methods to use with x-ray tomography (XRT) images to allow for comparison with outcrop and field data. To test and verify these new methods two low permeability carbonate samples were used for deformation testing and analysis. Applying the different scanline methods we show that they can be used to analyse lab induced fractures (mm to cm-scale) identified in XRT images for comparison with outcrop data (m-scale). In addition, these methods also allow for quantification of fracture network attributes e.g. fracture spacing, fracture apertures, orientation. This new data bridges the gap between micro-scanlines using thin sections and outcrop scanlines.


Introduction
To allow for a better comparison between laboratory (mm to cm-scale) and field (m-scale) data sets we present two techniques that both utilise x-ray tomography (XRT) images.By quantifying certain aspects of the fracture network (e.g.intersection with XRT image edge or fracture aperture) these methods allow for detailed analysis and reconstruction of fracture networks that can be used to compare with outcrop data directly.Comparing outcrop data to laboratory results is important to verify and validate experiments of analogue and reservoir materials especially regarding conditions for deformation experiments, especially since experimental results help to populate geomodels with subsurface structural features, particularly where data is sparse.Important properties are fracture distribution and fracture connectivity, since they have the potential to provide major pathways for fluids in the subsurface, especially in low porosity carbonates, as well as other low porosity rocks [1][2][3][4].
XRT is becoming a widely used non-destructive technique in geoscience to assess and characterise rocks.It has been used successfully on a range of different rock types, as well as nongeological materials [5][6][7][8][9][10][11][12].Imaging limitations are related to sample size and its relation to image resolution.With more powerful XRT acquisitions imaging standard 38mm diameter 76mm length core plugs is possible.However, XRT studies on carbonates tend to focus on characterisation of pore networks, rock characterisation or permeability calculations to aid characterisation relevant for petroleum engineering purposes such as prediction of permeability, or pore network connectivity [5,9,11,[13][14][15].A few studies use XRT to identify fractures or similar features within the core samples [16].But the majority of studies appear to be limited to non-destructive characterisation of rock properties (e.g.porosity, grain shape, grain size) to allow for a faster characterisation.This study aims to analyse key characteristics of laboratory-induced fracture networks to allow for comparison with outcrop and field data.
Laminites from the Crato Formation (NE Brazil) are a suitable analogue for the Brazilian pre-salt oil-bearing carbonate reservoir rocks (Miranda et al., 2014, Santos et al., 2015, Catto et al., 2016).Considering the high risks associated with producing these very deep and highly complex carbonate reservoir rock systems, and the relatively small amount of information about these subsalt reservoir rocks, it is important to better quantify what controls their structural evolution, the fracture, deformation band and fault systems that develop and the resulting response of these rocks to the evolving subsurface conditions Some carbonates follow the same general pattern of response to that of clastic rocks in terms of response to stress or pressure changes e.g.Oolitic limestones [17][18][19][20].However laminites are finely layered carbonates and cannot be linked to clastic behaviour as easily.Available deformation behaviour or mechanical response observations appear to be limited to field observations [21].This work describes outcrop logging of loop bedding orientations and SEM analysis methods related to field observations of structural discontinuities, their geometry and distribution across a range of sizes.

Sample preparation
The laminite sample (provided by Universidade Federal do Rio de Janeiro (UFRJ)) was 22.41cm long, 10.35cm wide and 5.92cm high (Figure 1), is of Aptian age and was collected from the Mina Triunfo quarry in the Araripe Basin, NE Brazil.The geological history of the Araripe Basin and its laminites is beyond the scope of this paper but they are regarded as suitable analogues for pre-Salt laminites that are now offshore Brazil (Ponte & Appi, 1990;Ponte & Ponte Filho, 1996;Neumann, 1999;Assine, 2007;Miranda, 2014;Miranda, 2016).
Due to the vulnerability of laminae to breaking off during coring only three 38mm diameter core samples perpendicular to laminations were produced to test these new methods.The lamination is mostly mm-scale with a several mm thick band of higher porosity material as shown in figure 1b.Three samples were prepared but the experimental deformation of the 3 rd sample was terminated when the sample developed a flaw.Table 1 summarises sample dimension.

Triaxial deformation
Triaxial deformation is used to place the samples under an approximation to the subsurface stress state of interest, where axial force is provided by pistons approaching each other and the lateral load is provided by pressurised fluids.Even though this approach is only an approximation to the true loading path during burial and structural development, this type of experiment regarded as a key factor in understanding and predicting the deformation response via a knowledge base for a set of known conditions.Here only two suitable samples were available.The deformation features developed in the samples are the focus of the remainder of this study The samples were placed in a Hoek cell and deformed under triaxial conditions with confining pressures of 20 for sample L-20 and 30MPa for sample L-30.Both samples were deformed with a constant shortening rate of 0.55 mm/min.Once peak stress was reached and the axial force started to decrease, a very good indicator that the sample was being damaged, and the test was stopped as quickly as possible.This was done to minimise overwriting of early deformation features by later features.After the test was stopped the sample was carefully extracted from the Hoek cell and photographed.The stress-strain relationship curves for the triaxial deformations are shown below in Figure 5.

X-ray tomography
XRT was performed at the 4D-Imaging Lab at the Division of Solid Mechanics, Lund University, using a Zeiss XRadia XRM520 X-ray tomograph.For these images, the X-ray tube voltage and power was set to 150 kV and 10 W and the He3 filter was utilised to avoid beam hardening artefacts.The samples were placed at 35, 70 and 80 mm from the source and at 47, 54 and 232 mm.1601 radiographs were acquired over 360 with an exposure time of 2 or 10 s per projection.The tomographic reconstruction provided a cylindrical image of a vertical section of the sample with both diameter and height of about 1000 voxels, where the voxel is cubic with side lengths of 40 or 10 microns.The tomographic reconstruction was performed using the Zeiss reconstructor software with a correction for the centre of rotation.Sample L-30 was imaged three times exploring different settings whereas sample L-20 was only imaged once.

Surface fracture reconstruction from XRT -Edge scanline method
Scanlines are a sampling method used to gather fracture traces characteristics e.g.number, length, orientation, aperture [22,23].To make this field method suitable for use with XRT images we amended the circular scanline [22,23] to the edge scanline method.This allows us to record the angle of intersections between fractures and the XRT image edge.Figure 3 shows an example of this technique.For each laminite plug every hundredth image was uploaded into FIJI [24] to collect data using the edge scanline method.Afterwards all recorded edge intersection angles were plotted against core plug depth to produce a 2D fracture surface reconstruction for each sample.

Fracture aperture distribution
For each sample, the middle section of the reconstructed XRT images was isolated (see yellow box in Figure 4) to restrict the scanline area for measurements of fracture aperture and fracture spacing.This technique follows the P10 method [25] which excludes overall lengths of fractures and minimises bias due to sample size.This approach was chosen to eliminate the impact of the core size limiting maximum fracture length to 38mm.To increase contrast between fractures and rock the images were coloured using the Fire overlay available within FIJI.Scanline values for every hundredth slice were taken using scaled XRT images in combination with the built in measuring tool in FIJI. Figure 4 shows an example of this technique.

Results & Discussion
These new techniques allow for comparison of experimental data of laboratory-induced fracture networks with outcrop derived fracture data across several length scales without correction for scale differences between data sets.The ability to make this type of comparison is important to validate experimental procedure and conditions, verify the suitability of core samples for fracture network analysis and help identify any correlation between experimental and outcrop data.The methods shown here make such a comparison tractable.Since similar XRT data from experiments is used in digital rock analysis e.g. for permeability and porosity analysis [14,26,27] or to predict fracture fluidflow [26,28] it is important to compare experimental and field data to understand how representative the experimental data are.

Triaxial deformation
Photographs taken of the core plugs after triaxial deformation and removal from the Hoek cell show similar response to loading compared to outcrop observations of structural features [29,30].Both samples L20 and L30 show a combination of shear and now-open (at atmospheric pressure) fractures after deformation (Figure 6).Now-open fractures do not show any apparent offset and it is not clear if these are open under confining pressure, and by implication that any natural equivalents are open in the subsurface.It is very likely that these fractures opened during the unloading following triaxial deformation.

2D surface fracture analysis from XRT
The XRT edge scanline method allows for a quick quantitative analysis of fractures observed on the sample surface.It can be used instead of the classical but time consuming plotting of dip and strike values on a stereonet.Importantly it also has significant other advantages.It does not require that assumptions be made about what fracture trace combinations seen on the sample edge are part of the same fracture, also avoiding ambiguity over classification of branching fractures.It does not require the often difficult estimation of the fracture plane orientation from the surface traces.And it preserves fracture positional information.
I the edge scanline method the line of the fractures' intersections on the sample's cylindrical surface are the trace of the intersection between each fracture plane and the cylindrical surface, and as such will appear as a sinusoid when plotted on a flattened surface.They are referred to here as having a slope on an x-y plot of compass direction plotted on the x axis against vertical location on the y-axis (Figure 7).The fracture traces (Fig 7 B and D) follow the shapes observed on the sample photographs taken after testing (Figure 8) building confidence in the technique.A difference in fracture frequency in the samples could be linked to different mechanical characteristics such as the mechanical response of different assemblies of minerals, or lamination density or a combination of these but this relationship is outside the scope of this paper.Using 2D surface fracture reconstruction (Figure 7A and 7C) 20 fracture planes were identified from the plotted trends that could be easily traced and these are shown in Figures 7B and 7D.Using the surface fracture trace lines in Figures 7B and 7D the average dip angles were calculated for each reconstructed fracture plane and the values are shown in Table 2.  To validate this reconstruction approach to calculate reconstructed fracture plane dips approach 4 the 2D reconstructed fracture planes (Figure 7) have been superimposed onto the core plug images 5 (Figure 8).
25 26 Where y1 is the sample depth at the bottom of the fracture trace line and y2 at the top.

27
Reconstructed fracture plane dip values for negative slopes are corrected by subtracting the 28 calculated value from 180 to allow for a geological nomination of fracture dip angles.Work from Miranda et al [29][30][31] shows that two fracture sets were identified in the laminite 54 outcrop using data from scanlines parallel to lamination (Figure 9).

Figure 1 .
Figure 1.Laminite outcrop sample from Mina Triunfo quarry, NE Brazil A: top view of the sample showing the plug locations B: side view showing the laminations.More porous portion at about 27 on the scale.

Figure 3 .
Figure 3. XRT edge scanline technique using individual XRT images.A: division of XRT image into two sections: 0-180° and 180 -360°.B: Example of fracture intersections for angle measurements

Figure 4 .
Figure 4. Scanline selection and orientation for fracture aperture measurements from XRT images for lab-induced laminite fractures

Figure 6 .
Figure 6.Laminite core plugs after triaxial deformation A: Laminite 20MPa confining pressure B: Laminite 30MPa confining pressure.Both samples show a combination of shear and now-open fractures on the plug surface.

Preprints 1 Figure 7 .
Figure 7. Surface fracture data for laminite core plugs deformed at 20MPa (A, B) and at 30MPa (C, D) confining pressure.A and C show the data generated by the

Figure 8 .
Figure 8. Surface fracture planes from the Edge scanline reconstruction superimposed onto the

Table 2 .
Reconstructed fracture plane dip angles for laminite deformed at 20 and 30MPa confining