You are on page 1of 14

SPE 130745

How to Propagate Petrophysical Properties in a Fracture Network for


Naturally Fractured Carbonate Reservoirs Case Study: Cretaceous
Formations at Maracaibo Lake, Venezuela
Rodolfo Soto B. /SPE, Digitoil, Sergio Perez, Duarry Arteaga, Cintia Martin / SPE,PDVSA Western Division

Copyright 2010, Society of Petroleum Engineers


This paper was prepared for presentation at the SPE EUROPEC/EAGE Annual Conference and Exhibition held in Barcelona, Spain, 1417 June 2010.
This paper was selected for presentation by an SPE program committee following review of information contained in an abstract submitted by the author(s). Contents of the paper have not been
reviewed by the Society of Petroleum Engineers and are subject to correction by the author(s). The material does not necessarily reflect any position of the Society of Petroleum Engineers, its
officers, or members. Electronic reproduction, distribution, or storage of any part of this paper without the written consent of the Society of Petroleum Engineers is prohibited. Permission to
reproduce in print is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgment of SPE copyright.

Abstract
A fuzzy-logic approach accurately characterizes spatial porosity and permeability for a complex, naturally fractured carbonate
reservoir. The reservoir lies in the Cretaceous section alongside the fault damage and shear zone at the downthrown block to
the west of the Icotea fault, Central North segment of the Maracaibo basin.
To identify the pore type in our characterization of this carbonate reservoir, we first used fuzzy logic to obtain fracture
indexes from conventional well logs. Then we used a resistivity/porosity model with cementation factor, m, variable to
determine fracture porosity, and calculated total porosity from neutron-density logs. After that, we developed a neuro-fuzzy
logic permeability model with porosity, shale volume, fracture indexes, and pore type as independent variables.
Next, we used directional-statistics tool analysis to identify the five main fracture families and categorize them. The
diagnosis of fracture orientation relies on the statistical analysis of fractures from four wells, three of them having image logs
and one having a direct description of fractures from an oriented core. We also compared the fracture-orientation data from
well data with the faults interpreted at seismic scale to unravel their structural connection.
Fracture lengths are obtained from curvature maps by direct measurement of lineaments matching the azimuth of each
family. A logarithmic fit of the known values allows extrapolating values of fracture lengths and spacings below seismic
resolution. To propagate porosity in our fracture system, where only a seismic P-wave volume was available, we found for
each fracture family a general inverse functional relationship between fracture intensity and fracture porosity, introduced as the
r-pi method of correlation.
With a sigmoidal function we correlate the neuro-fuzzy logic permeability variable with the aperture and intensity of the
fractures. With a statistically based characterization we identify wheter the fractures display directionally or stratigraphically
related anisotropies. Finally, to populate petrophysical properties in the fracture network we apply the functions devised
starting with the size of the fractures as an independent variable.
Introduction
The Maracaibo Basin, situated in the northwestern side of Venezuela, is ranked among the top 12 petroleum provinces in the
world (Klett et al, 2000). The Cretaceous reservoirs are among the most prolific in the area, with more than 6 billion bbl OOIP
of light oil. From 2007 and up to late 2009 PDVSA in cooperation with PVN and several consultant companies, developed the
Cretaceous Lake Pilot Project to model and design the development of offshore reserves located at the central-northern side of
the Maracaibo Lake.
The Icotea fault, located at the central section of the Maracaibo Lake, is a high angle dip, northeast trending fault. It is a
major structural feature of the basin, with a trace around 100 KM long. Like many other fault trends in the Maracaibo basin,
the Lama-Icotea fault system displays strike-slip classic features (Escalona and Mann, 2003), as subvertical to high angle dip,
restraining and release bends, and oblique folding. Its displacement is left-lateral (Lugo, 1991).
The Icotea fault is also a prominent structural feature in connection with several reservoirs, spanning from the Cretaceous
to the Eocene sequences. For the Cretaceous reservoirs at the central sector of the basin, the Icotea fault not only seves as a
definite compartment boundary, but also plays a major role in association with the distribution of fractures at its fault damage
zone. This fault damage zone has been identified as a sinistral trancurrency related shear zone (Perez et al., 2009). This work

SPE 130745

presents the methodology developed in the integrated reservoir study by the multidisciplinary team to assign petrophysical
properties to the fracture network associated to the Icotea fault shear zone.

General Methodology:
The procedure devised to assign the petrophysical properties of permeability and porosity to the fracture network is as follows:
1. Apply soft computing techniques to well logs and core analysis for petrophysical modeling of porosity, permeability,
and lithotypes..
2. Discriminate and characterize fracture families geometrically and spatially from well logs and oriented cores.
3. Typify the distribution of fractures in the different lithologies or lithotypes and stratigraphic units.
4. Collect fracture length and spacing data from structural seismic attributes. We took advantage of having near vertical
fractures (Table 1) and low dipping horizons, which facilitated the implementation of this step.
5. If possible, as in our case, integrate fracture spacing trends at well and seismic scales.
6. Using the pairing frequency method, set up a functional connection between fracture length, spacing and area (or size).
7. Conduct statistical characterization of fracture relevant parameters (intensity, porosity and permeability).
8. Correlate fracture porosity and fracture intensity for every fracture family using the r-pi method presented in this
paper.
9. Calculate fracture aperture as a function of fracture porosity and intensity for all fracture families.
10. Design and evaluate fracture probabilty as a dependant variable of fracture aperture.
11. Determine fracture permeability from the fracture probability/aperture related variable.
12. According to their areas, assign to the fractures in the 3D digital model their specific petrophysical parameters.
Structural setting and fracture families
The geological analysis made to diferentiate areas of the region under study with a definite structural style, with conspicuous
evidence from macroscale (3D seismic) to microscale (thin slides from core samples), supports the premise that the Icotea
fault damage zone is a strike-slip related shear zone (Perez et al., 2009).
For the identification of its fractures families we used first-hand fracture data provided by the available well records
(images logs, dipmeters and fracture data from measurements on an oriented core). Evidences indicate the families of fractures
at well scale were contemporary to macro structures seen at seismic scale in their vicinity. The scheme of analysis followed
Peacock (2001) and Price and Cosgrove (1990) guidelines. Fracture families were analyzed by a combination of information
provided by the Kuiper test, the Schmidt stereonet, and the rosette diagram.
The Kuiper test was restated (Fig. 1a) to our results in the angular range of 0 to 180. To that end the fracture strike angles
less than or equal to 180 were kept, while those angles grater than 180 were taken to the first or second quadrant (Middleton,
2000) by subtracting 180.
This convention is justified for data from wells in the shear zone by the symmetry of structures present in the strain ellipsoid
of sinistral strike-slip deformation (Cunningham and Mann, 2007). This is a permissible departure from the original concept of
the Kuiper test, since it is formally designed to work in the angular range covering from 0 to 360. It is worth noting that
working in a two-quadrant fashion rather than four also simplifies the work of quantifying the fracture families, since the work
is reduced by half.

Fig. 1Fracture families, cored well at Lagomar Apon section. (a) Kuiper test displaying the distribution of fractures on the
lithotypes. (b) Set of structures related to sinistral strike slip deformation by angular range. (c) Cogollo Group. Schmidt plot.

Once the angular range of dip azimuth of each fracture family is set by the Kuiper test, we proceed to compute the proper
Fisher parameters for each set of poles to the fractures. The software used to that end (Petrel 2008) works on a Schmidt
stereonet frame (Fig. 1c).

SPE 130745

Table 1 summarizes the parameters determined for the fracture families.


TABLE 1 PARAMETERS USED IN MODELING DISCRETE FRACTURE SYSTEMS DETAILED BY FAMILIES AND TWO STRATIGRAPHIC
INTERVALS AT THE COGOLLO GROUP (APON FORMATION AND MARACA-LISURE FORMATIONS (MARLIS)
Fracture Length
Fisher Model Parameters
Fracture
Fm
Max.
Average
Family
Average
Dip
Standard
Well cut Dip
Concentration
Length
deviation
m
Azimuth
m
m
FAR
Apn
140
108
305
24
86
188
3.6
MarLis
96
62
375
8
77
10
5.1
FR
Apn
97
53
250
24
86
86
5.9
MarLis
90
45
311
10
79
268.7
4.6
FT
Apn
113
64
298
95
89
51.1
5.7
MarLis
122
74
451
48
87
236
4.9
FLIY
Apn
122
70
461
48
88
115
6.4
MarLis
89
55
456
20
85
294
4
FInv
Apn
161
89
520
24
86
147
5
MarLis
178
112
382
14
83
322
3

At this point, the analysis of the connection between the system of fractures and the association of structures in the context
of strike-slip tectonics has benefited from the methodology presented by Blumentritt et al. (2006). The Icotea fault shear zone
provides hints linking the sinistral strike-slip tectonic style and the families of fractures in that area. One of the most noticeable
is the correspondence between the orientation of fracture families and the orientation of comparable structures in the strain
ellipsoid for sinistral strike-slip deformation, as seen in Figs. 1b and 1c. Core information gives further support about the
presence of brittle shearing as a fracture forming mechanism (Perez et al. 2009).
Using core information, we determined the distribution of fractures at the different lithology types (Fig. 2). There were
some limitations in the analysis, as core material was almost fully recovered at Apon but only partially at Maraca (uppermiddle section) and Lisure (lower section). So for comparison purposes, we contrasted core data from the Apon section with
the data supplied by the Maraca-Lisure section considered collectively. A proper comparison separating the distinctive features
among the three formations is suggested for further development of the subject.
By agreement with the sedimentology team, the original core-based lithological descriptions were simplified to be handled
in petrophysical terms by means of 10 lithotypes, of which we present the 5 most relevant in connection with the fractures
(Figs. 2 and 3).
The whole thickness of the cored section with fractures is dominated by the intragrain limestone lithotype, followed by
clayey limestones.
When the general distribution of fractures on such lithotypes is inspected, although the highest amount of fractures is
located at intragrain limestones, the second highest fractured lithotype is limestone with vuggy porosity. This calls attention
because that lithotype is a minor contributor to the core overall thickness, and certainly present in intervals collectively smaller
than the ones with clayey limestones (Fig. 2).
Further investigation on the connection between the lithotypes and fracture variety is carried out focusing on the relative
abundance of the fracture families in each of the lithotypes.
The results highlight some differences between Apon and the Maraca-Lisure sequence in the distribution of fractures of the
five families (Fig. 3).
At Maraca-Lisure, tension fractures predominate on the clean carbonate units (over 50%), but antithetic Riedel shears
prevail in the clayey limestones and are present in the dolomites, but do not occur at the Apon section. The presence of Y
shears is almost negligible, even though they are also scarce at the Apon section.
The Apon also contains a high proportion of tension fractures (over 40%) on the clean intragrain carbonate units, followed by
antithetic Riedel shears. Riedel shears dominatewithin the remaining lithotype (clayey limestone).

SPE 130745

Fig. 2Left: Thickness distribution of lithotypes of significant interest regarding fractures in the
well cored at Lagomar. Right: Distribution of fractures in the lithotypes of the well cored at Lagomar.

Fig. 3Left: Maraca and Lisure formations. Distribution of the fracture families in the lithotypes of the well cored at
Lagomar. Right: Apon Formation. Distribution of the fracture families in the lithotypes of the well cored at Lagomar.

Fracture Length, Spacing, and Intensity


Once the fractures are divided into families, the next step is to evaluate intensity and accumulation logs for fracture families by
well (with correction for borehole deviation). For the analysis described from now on, we use information provided not only
by fractures on an oriented core, but also data from three wells with image logs and six wells with dipmeters.
For every fracture family we also estimated the individual distributions of fracture length and spacing based on the analysis
of these variables measured from structures (lineaments) at seismic scale. Maps of structural attributes, like semblance or
curvature maps, referred to as the working horizons (at Maraca, Lisure and Apon) are the grounds for this job.

Fig. 4Relationships between fracture legth and spacing. Obtained pairing distributions of measurement of these variables.

SPE 130745

We then proceed to compare and match the distributions (Fig. 4) using their similarity in frequency as a bridge between the
variables (Springer et al., 1968). Since we tone and pair logarithmic distributions based on frequencies, we called it the pairing
frequency method, or pf method for short.
We converted fracture intensity at well scale to fracture spacing. Then, we proceeded to crosscheck the match between
fracture spacing trends at both scales. To that end, for each fracture family we integrated fracture intensities obtained at
seismic scale and at well scale in one percentage-paired distribution. Then, as we found high correlation indexes for the
resultant distributions in either logarithmic or power laws, we saw evidence of the match of tendencies at both scales (Table
2).
Table 2. Apon Formation. Well and Seismic spacing data (FSp) in meters. Equations and correlation indexes of
fracture spacing integrating data at seismic and well fitted with logarithmic or power law trends.
R2

Best Fit

FAR

0.92

Freq= -7.377 ln(FSp) + 59.145

FT

0.81

Freq = -8.431 ln(FSp) + 60.4

FR

0.91

Freq = -9.396 ln(FSp) + 73.541

FR FINV

0.77

Freq = 51.366 FSp-0.287

FLIY

0.91

Freq = -10.68 ln(FSp) + 95.378

distributions of

Fig. 5Apon Formation. Well and seismic spacing data (FSp) in meters. Plot of distributions of fracture spacing integrating data at
seismic and well scale fitted with logarithmic or power law trends.

With the functional connection between fracture length and spacing based on the analysis of these variables measured from
structures at seismic and well scale scale (Fig. 5), we devised the connection between fracture length and intensity. The length
is computed as the inverse of the intensity, so we get values of fracture intensity to the fractures of specific length provided by
the 3D geomodeler software. We proceeded in such way not only by the study of the results presented on Table 2 about the
match of fracture spacing trends at well and seismic scales, but also by the evidences that at the Icotea fault shear zone, the
fractures at seismic, well, and even at micro-scales corresponds to structures associated to deformation under strike-slip
tectonism (Perez et al., 2009).
At seismic scale, the reference for the dimensions of the fractures of each family, were estimated using the lengths of the
parallel lineaments measured on curvature maps on Maraca and Apn. These lineaments were interpreted as traces of faults in
these horizons (Nissan et al. 2006). In our case, the average high angle dip of the fractures at seismic and well scale and the
low dip of the horizons support the use of the fracture traces at the horizons as a reference for a proxy of the true fracture
length and spacing. To make this approach work for a study that deals with fractures of families that are not subvertical, the
traces should be measured on tilted planes orthogonal to the average fracture dip direction.
For purposes of statistically significant data handling at least 25 measurements (Gemignani, 1998) of fracture lengths and
spacing per family were performed. Using the resulting logarithmic equations (Van Dijk et al, 2000; Koike and Ichikawa,

SPE 130745

2006) from Table 2, we calculated the frequency of fractures of sizes smaller than the seismic resolution but bigger than well
resolution (Ekneligoda and Henkel, 2006).
Average fracture lengths and the corresponding standard deviation were obtained using measured fracture lengths and their
frequency data.
The method from Ozkaya (2003) was used as a reference to estimate average lengths of fractures from image logs or, as in
our case, from core information (average well cut, Table 2). This method provides a reference range of fracture lengths
traversed by wells, and we consider it useful to link average fracture dimensions with other properties derived from well data,
like average spacing, intensity, or porosity.
We have already used fracture intensity from well data to link fracture spacing at well and seismic scale. To proceed
further with the integration among fracture properties at well scale, we needed a statistical study of the features of fracture
intensity between the fracture families defined by stratigraphic interval. One recommended starting point is boxplots
(Montgomery, 1991; Berk and Carey, 2000).

Fig. 6Boxplots of distribution of fracture intensity values in mm-1 (right) and its logarithms (left), discretized by fracture family and
stratigraphic interval. Dash line: Mean value. Black circle: Low outlier. White circle: High outlier.

As seen in Fig. 6, the boxplots of the intensity indicate that antithetic Riedel shears at the Apon section display the highest
intensity mean value and interquartile range by fracture and formation. Nevertheless, at Maraca-Lisure the highest intensity
mean value and interquartile range is presented by tension fractures.
Most of the fracture families present interquartile ranges of fracture intensity smaller than antithetic Riedel shears at Apon
and tension fractures at Maraca-Lisure.
The distribution of intensity values by fracture family and formation displays different degrees of dispersion, but most of
them have small interquartile ranges and positive skews. Inspecting the boxplots for the distribution of the logarithms in
Fig. 6 (right), we found they still displayed some asymmetry and differences between the mean and the median, so we prefer
to further investigate a logarithmic rather than a log-normal fit of their distributions.
The results shown in Fig. 7 (left) confirm the adjustment to logarithmic fits for most of the distributions of intensities.

Fig. 7Intensity values (mm-1) (left) and Porosity (in fraction, at right) arranged for a display in a relative frequency plot. Logarithmic
fit of the distributions exhibit high correlation indexes (table inside the plot) for most of the fracture families.

SPE 130745

Back to Fig 6, the visual look of the boxplots of the distributions suggests noteworthy differences between the means. To
settle the point, a variance analysis and a matrix of mean differences is a usual statistical procedure to follow (Berk and Carey,
2000). The analysis of variance and the matrix of mean differences were made using the excel complements Data Analysis
Toolpack and StatplusV2.0, respectively (Berk and Carey, 2000).
With a critical value of 1.89 but an F value calculated as 21.46, the output of the respective variance analyses indicates a
very low probability for the null hypothesis to be true. So we accept that the average intensity is not the same for each fracture
family regarded by stratigraphic interval.
The complete analysis of the differences of average intensities among the fracture families requires the use of the matrix of
mean differences (Table 3).
TABLE 3FRACTURE FAMILIES BY STRATIGRAPHIC INTERVAL. MATRIX OF MEAN DIFFERENCES FOR THE
DISTRIBUTION OF FRACTURE INTENSITIES. SIGNIFICANT MEAN DIFFERENCES ARE HIGHLIGHTED IN RED
FAR
FT
FT
FR
FR
FINV
FINV
FLIY
FLIY
Pairwise Mean FAR
ML
APON
ML
APON
ML
APON
ML
APON
ML
APON
Difference
0,003 0,000
FAR ML
0,000
-0,009
0,003 0,002
0,002
0,001
0,003
0,003
FAR APON
0,000
0,006 0,009
0,012 0,011
0,011
0,010
0,012
0,012
FT ML
0.000 0.003
0.006 0.005
0.005
0.004
0.006
0.006
FT APON
0.000
0.003 0.002
0.002
0.001
0.003
0.003
FR ML
0.000 -0.001
-0.001
-0.002
0.000
0.000
FR APON
0.000
0.000
-0.001
0.001
0.001
FINV ML
0.000
-0.001
0.001
0.001
FINV APON
0.000
0.002
0.002
FLIY ML
0.000
0.0003
FLIY APON
0.000

At Maraca Lisure, 3 out of 10 significant differences of average intensity appear among fracture families. Tension fractures
and Y shears present the greatest difference of this variable
At Apon, the greatest difference is seen between antithetic Riedel shears and Y shears. At this section 4 out of 10
significant differences of average intensity appear among fracture families.
Antithetic Riedel shears display the only significant difference in average fracture intensity compared by itself and by a
different stratigraphic interval. In general, 15 out of 45 significant differences of average intensity appear among fractures
separated by both family and stratigraphic interval.
Fracture Porosity and Permeability
Using the record of fracture porosity obtained from logs by petrophysical means is feasible, through a depth match, to identify
the correlative values of intensity for each fracture already categorized inside one of the five families.
For their convenient way to compare statistical parameters among the different fracture families, we suggest boxplots as
one of the best practices to make the statistical analysis of the distribution of values of porosity and permeability for each
fracture family.

Fig. 8Boxplots of distribution of fracture porosity values (right) and its logarithms (left), discretized by fracture family and
stratigraphic interval. Dash line: Mean value. Black circle: Low outlier. White circle: High outlier.

As seen in Fig. 8, the distribution of porosity values by fracture family and formation displays different degrees of
dispersion, but most of them have positive skew and outliers. This suggests to investigate the distribution of their logarithms.
The resultant boxplots for most of the families display comparatively smaller skewness and fewer outliers than for the original

SPE 130745

distributions, but the asymmetry of the distributions and mismatches between the means and the medians indicate to look for a
logarithmic rather than for a log normal fit for them.
In spite of the limitations impossed by the conditions of well data collection, the coincidence of logarithmic fits in both
fracture intensity (Fig. 7, left) and porosity (Fig. 7, right) suggest to look for a functional connection between them like the
ones found from previous studies but under less complicated settings (Cohen and Mercer, 1993; Nelson 2001). We came
across a high-index correlation method that works with data distributions of these parameters from well data and hopefully
would become a best practice in the field of fracture reservoir characterization.
Reviewing the boxplots of the porosity values (Fig. 8), the Riedel shears at the Maraca-Lisure section display the highest
porosity mean value by fracture and formation. Nevertheless, at Apon, the highest porosity mean value is presented by the
Y shears, closely followed by the tension fractures.
The lowest interquartile porosity corresponds to compressional shears at Maraca-Lisure, and this fracture family also
presents the lowest porosity scores at Apon.
The outcome of the analysis of variance for the fracture porosities indicates:
The variability of the porosity within fracture families (0.038) is almost one order of magnitude bigger than the
variability of the porosity between fracture families (0.006).
According to the null hypothesis, the ratio between the two variances should follow an F distribution with
(n,m)=(9,617) degrees of freedom. F is computed as 11.47 using the porosity values of the fracture families. The
-16
probability obtained (1.02 10 ), is significantly lower than 0.05. The null hypothesis is rejected, and we accept that the
average porosity is not the same for each fracture family regarded by stratigraphic interval.
Further and detailed analysis of the differences of average porosities among the fracture families requires the use of the
matrix of mean differences (Table 4).
TABLE 4FRACTURE FAMILIES BY STRATIGRAPHIC INTERVAL. MATRIX OF MEAN DIFFERENCES FOR
THE DISTRIBUTION OF POROSITIES. SIGNIFICANT MEAN DIFFERENCES ARE HIGHLIGHTED IN RED
FAR
FT
FT
FR
FINV
FINV
FLIY
FLIY
Pairwise Mean FAR
FR ML
ML
APON
ML
APON
APON
ML
APON
ML
APON
Difference
FAR ML
0.000
-0.001
0.002 -0.005 -0.010
-0.003
0.006 0.003
-0.002
-0.007
FAR APON
0.000
0.003 -0.003 -0.008
-0.001
0.007 0.004
0.000
-0.006
FT ML
0.000 -0.007 -0.012
-0.005
0.004 0.001
-0.004
-0.009
FT APON
0.000
-0.005
0.002
0.011 0.008
0.003
-0.002
FR ML
0.000
0.007
0.016 0.012
0.008
0.002
FR APON
0.000
0.009 0.005
0.001
-0.005
FINV ML
0.000 -0.003
-0.007
-0.013
FINV APON
0.000
-0.004
-0.010
FLIY ML
0.000
-0.006
FLIY APON
0.000

From Table 4 we draw information about the differences of the porosity averages between specific sets of fracture
families. The greatest difference is between Riedel shears at Maraca-Lisure and Compressional shears at the same interval.
At Maraca Lisure, there are six out of ten significant differences of average porosity among fracture families.
At Apon, the greatest difference is seen between Compressional shears and Y shears. At this section there are three out of
ten significant differences of average porosity among fracture families.
In general, there are eigthteen out of forty five significant differences of average porosity among fractures separated by
both family and stratigraphic interval. Tension Fractures is the only fracture family that displays a significant difference in
average porosity compared by itself by a different stratigraphic interval.

SPE 130745

The outcome of the respective analysis of variance for fracture permeabilities, as deduced from Table 6, also supports that
the average permeability
is not the same for each fracture family regarded by stratigraphic interval.

Fig. 9. Boxplots of distribution of fracture permeability values (rigth) and logarithms (left), discretized by fracture family and
stratigraphic interval. Dash line: Mean value. Black circle: Low outlier. White circle: High outlier

As seen in Fig. 9, the distribution of permeability values by fracture family and Formation exhibits different degrees of
dispersion, but there are positive and negative skews and the distributions of only three families present high outliers. These
features have little change in the distributions of their logarithms.
Compressional shears at the Maraca-Lisure section and also at the Apon Section display the highest permeability mean
value by fracture and Formation. Lowest interquartile porosity Fig.s correspond to Riedel shears at Maraca-Lisure, while Y
shears present the lowest permeability scores at Apon.
There is a striking closeness between the median of the permeabilities of Anthitetic Riedel shears and the third quartile.
This fact is relevant since indicates that more than 50% of the fractures of this family present permeabilities above 550 md.
Notice that quite contrary is the case of the Y and Riedel shears, with medians close to the first quartile and near the low
permeability values.
The visual look of the boxplots of the distributions suggests significant differences between the means, what is confirmed
by the respective analysis of variance. (F= 17.4 with a Critical value of only 1.9).
The analysis of variance for the permeabilities presents a very low probability (7.48 10-26) for the means to be similar. The
null hypothesis is rejected and we accept that the average permeability is not the same for each fracture family regarded by
stratigraphic interval.
Further and detailed analysis of the differences of average permeabilities among the fracture families requires the use of
the matrix of mean differences (Table 5).
TABLE 5FRACTURE FAMILIES BY STRATIGRAPHIC INTERVAL. MATRIXES OF MEAN DIFFERENCES FOR
THE DISTRIBUTION OF PERMEABILITIES. SIGNIFICANT MEAN DIFFERENCES ARE HIGHLIGHTED IN RED
FAR
FT
FR
FINV
FLIY
FLIY
Pairwise Mean FAR
FT ML
FR ML
FINV ML
ML
APON
APON
APON
APON
ML
APON
Difference
FAR ML
0.00
7.16
-26.76
95.44
241.13
87.34
-125.41
-85.39
79.18
235.31
FAR APON
0.00
-33.92
88.28
233.97
80.18
-132.57
-92.56
72.02
228.15
FT ML
0.00
122.20
267.89
114.09
-98.65
-58.64
105.94
262.07
FT APON
0.00
145.69
-8.10
-220.85
-180.84
-16.26
139.87
FR ML
0.00
-153.79
-366.54
-326.53
-161.95
-5.82
FR APON
0.00
212.75
-172.74
-8.16
147.97
FINV ML
0.00
40.01
204.59
360.72
FINV APON
0.00
164.58
320.71
FLIY ML
0.00
156.13
FLIY APON
0.00

From Table 5 we draw information about the differences of the permeability averages between specific sets of fracture
families. Like in the case of average porosity, the greatest difference is between average permeability of Riedel shears at
Maraca-Lisure and Compressional shears at the same interval.
Maraca-Lisure has 9 out of 10 significant differences of average permeability among fracture families.
At Apon, the greatest difference is seen between Compressional shears and Y shears. At this section also has 9 out of 10
significant differences of average permeability among fracture families.
In general, 37 out of 45 significant differences of average porosity appear among fractures separated by family and
stratigraphic interval. Tension Fractures, Riedel shears and Y shears display significant differences in average permeability
compared by themselves and by different stratigraphic intervals.

SPE 130745

10

Correlation between fracture intensity and porosity


Previous studies (Cohen and Mercer, 1993; Nelson, 2001) have shown a direct relationship (Eq. 1) between fracture spacing
(S), aperture (e), fracture porosity (f or PHIF). The following equation is based on that presented in Nelson (2001), and
correlates porosity, taken as a fraction instead of percentage, with the opening and spacing variables.

PHIF = f = e /( S + e) (1)
Since fracture spacing is the inverse of fracture intensity (I), it follows that there is a direct relationship between fracture
porosity and intensity (Eq. 2):

PHIF = f = e /(1 / I + e) .... (2)


Furthermore, Ortega et al. (2006) have also shown that fracture spacing is in functional connection with the opening of
fractures at different scales.
We use Eq. 1 to calculate fracture porosity (PHIF) corresponding to the outcrop measurements of fracture spacing and
aperture for tension fractures provided by Ortega et al. (2006). Next, we estimate the corresponding fracture intensity as the
inverse of the spacing. Hence we are able to plot the corresponding regression line between PHIF and fracture intensity for the
same tension fractures (Green line, Fig. 10)
Then, on the same graph we plot the cloud of points and its regression lines for the fracture families at the Cogollo Group
(Fig. 10).
-1
The key point is a match based on rearranging porosity (in fraction) and intensity (mm ) distributions in an opposite sense
before correlating them. Lets call this procces the r-pi method for short. In Table 6 we show the advantage of this method
over the straightforward depth match of porosity versus intensity with the comparison of regression coefficients provided by
each method. Fig. 10 shows the results after the reordering procedure is followed. There we see fracture porosities as inversely
proportional functions of fracture intensities for both shear and tension fractures. In this case a logarithmic scale is set up for
both axes, and the porosity-intensity functions follow power or logarithmic laws (Table 6).

Fig. 10 Logarithmic or Power regresion lines of the distributions of fracture porosities versus intensities measured at the Cogollo
Group for shear fracture families in wells at the Icotea Fault shear zone. Power regression line for Tipe I Fractures (in Green) based
on Ortega et al. (2006).

Fig. 10 displays logarithmic or power law correlations of the clouds of points for every fracture family when plotted over
axes in logarithmic X-Y scales. We see the comparison between them and a power regression line (in Green) for tension
fractures based on the reformulation of the regression line for fracture spacing and aperture devised by Ortega et al.(2006).
The outcome indicates that for a given porosity, Y shears have the lowest intensity. Depending upon the section and range
of values considered the antithetic Riedel shears, Tension Fractures or even compressional shears could have the highest
intensity for a given porosity. This suggests that for a certain fracture dimension or size among families of fractures, Y shears

SPE 130745

11

have the lowest porosity, while Antithetic Riedel shears, Tension Fractures o even compressional shears could have the highest
porosity.
FAR ML
FT ML
FR ML
FINV ML
FLIY ML
FAR APON
FT APON
FR APON
FINV APON
FLIY APON

R2
0.83
0.93
0.79
0.88
0.73
0.88
0.90
0.98
0.86
0.88

TABLE 6FRACTURE FAMILIES BY STRATIGRAPHIC INTERVAL


Regression after reorder
R2
Depth match Regression
FI = 0.0004 PHIF-0.218
FI = -7E-04 ln(PHIF) - 0.0022
0.10
-06
-1.342
FI = (10 ) PHIF
FI = 5 (10-06) PHIF-1.087
0.61
FI= -4 (10-04) ln(PHIF) - 0.0011
FI = 3 (10-05)PHIF-0.753
0.71
-04
FI = = -5 (10 ) ln(PHIF) - 0.0023
FI = -5 (10-05) ln(PHIF) + 0.0012
0.01
FI = 4 (10-05) PHIF-0.4321
FI =0.0001 PHIF-0.242
0.23
-06
-1.4282
FI = 2 (10 ) PHIF
FI = 6 (10-06) PHIF -1.228
0.65
-06
-1.1969
FI = 6 (10 ) PHIF
0.57
FI = -0.004 ln(PHIF) - 0.0144
IFRA = 4 (10-06) PHIF-1.0903
0.56
FI = -0.001 ln(PHIF) - 0.0046
FI = -6 (10-04) ln(PHIF) - 0.0016
FI = 3 (10-05) PHIF-0.303
0.28
FI = 5 (10-07) PHIF-1.4355
FI = -3(10-05) ln(PHIF) 8 (10-05)
0.34

Modeling fracture aperture and permeability


The Petrophysics Team designed a fracture permeability model based on aperture that was applied to the fracture data obtained
from well information and also to fractures at the 3D digital model.
The fracture aperture in the categories of each family was obtained as a function of porosity (PHIF) and spacing (S) of
fracture by the formula:

Aperture = e = S /(1 + 1 / PHIF ) (3)


This formula is obtained by solving the Eq. 1 with aperture as the dependent variable. Once the aperture is obtained, it is
necessary to calculate the probability of fracture (PF) and the permeability according to Eqs. 10 and 11.

PF = 12 .851547 + 13 .584919 /(1 EXP ( ( Apertura + 0 .028516233 ) / 0 .0084654285 )) ..(4)


Permeabili ty = 580 .6405 /(1 + EXP ( ( PF 0 .76778648 ) / 0 .043830136 )) (5)
By a straightforward substitution, we can link aperture and permeability via Eq. 4 and Eq. 5 to directly plot their correlative
array. The result (Fig. 11) is a curve with a sigmoidal shape where the distribution of fractures of the different families can be
inspected and analyzed.

Fig. 11Apon formation. Distribution of permeability as a function of the probability of fracture for every fracture family, which is in
turn based on aperture.

Fig. 11 shows that at Apon the high range of extreme values of permeability is dominated by Antithetic Riedel shears,
followed by Compressional shears. This agrees with the outcome of the analysis of boxplots (Fig. 9). The proximity between
the median and the third quartile at the permeability range over 550 md explains this outcome. The broad interquartile range of
the distribution of permeability for Antithetic Riedel shears also explains its occurrence all along the aperture-permeability
curve. As mentioned in connection with Fig. 9, we expect high occurrence of Y shears to be concentrated along low range
values of permeability.

12

SPE 130745

Once the model is ready, we proceed to propagate it through the digital fracture network already constructed by the use of
the parameters provided by Table 1. This fracture network contains fractures of different sizes, but their geometries are set up
by the modeler, preferentially by using some reference from the seismic expression of the fractures. We could, f.e., specify the
fractures as rectangles with the width equal to twice the length (L). The length is then directly calculated from the Area of the
fracture with the expression L = (Area/2). Once fracture Spacing is calculated from the fracture Length with the correlations
given by pf method, we start a sequence of operations to get their porosity and permeability. Porosity (PHIF) is derived from
fracture intensity applying the correlations provided by the r-pi method. Using from Eqs. 3 to 5, we get fracture permeability
from the already known variables S and PHIF.

Fig. 12Lagomar. Apon Formation. Family of Riedel fractures. Distributions of values for geometric and petrophysical parameters.

Fig. 12 shows the histograms with the distributions of values for the geometrical and petrophysical parameters of Riedel
fractures already available as digital 3D entities. The parameters are area, length (L), spacing (S), porosity (PHIF), aperture,
probability of fracture (PF) and permeability. The numbers on the arrows indicate the sequence followed to obtain the
parameters. The logarithmic nature of the attributes considered is visible in the histograms with the distribution of their values.

Summary and Discussion


In the Cogollo Group, fractures are present on 5 out of 10 lithotypes studied from well core information. These fractured
lithotypes correspond to kinds of carbonate-related lithologies, while the very low-fractured or unfractured litotypes are
related to sandstone and shale related lithologies. This means that the lithologic tock type control the mechanisms of
fracturing. This case being the usual brittle condition of the rocks as the crucial factor involved.
The coincidence of the predominance of tension fractures in most of the lithotypes both at Apon and Maraca-Lisure is
significant. The low strength of brittle rocks to tension fracturing (Zoback, 2007) and the rotation of Riedel shears to develop
into tension fractures with the progress of shearing (McClay and Dooley, 1994) are conditions likely linked with this
statistical outcome.
The scarcity of Y shears, which are parallel to the main boundaries of the shear zone, is noteworthy. Nevertheless, lab and
field evidences indicate that shearing in the Y direction as a result of deformation of a sedimentary cover over a rigid basement
has a strong contribution of R and R' shears for its definition, which could be a main reason behind the low amount of this kind
of shears in the samples (Davis et al., 1999).
The statistical study of fracture-related features allows us to define which ones are or not directionally and/or
stratigraphically driven by the comparison of the parameters of their distributions. The study of the matrix of mean differences
indicates that fracture permeability is a highly anisotropic feature, strongly driven not only by the factor of the direction of the
fracture family considered but also for the stratigraphic section. Fracture porosity and intensity are less anisotropic than
fracture permeability, by being less controlled than permeability by the factors of direction and stratigraphy. One appealing
issue derived from the statistical analysis is that Anthitetic Riedel shears are present in high proportion biased toward the range
of high permeablities, disregarding the Formation considered. A would-be confusion due to the broad range of permeability
values encompassed by their specimens is easily clarified by inspection of the position of the median at the respective boxplot.
The relevant point here is the influence this fact has on the design of the best strategy to take advange of the reservoir
properties. Production from wells drilled in the direction that connects the most Antithetic Riedel shears would benefit from
the high relative quantity of fractures of this family with favorable permeability conditions to inflow. The opposite could be
said regarding Y shears at Apon and Riedel shears at Maraca-Lisure. These fracture group exhibit permeability characteristics
indicating a connection with sealing and low-transmisibility features.

SPE 130745

13

The base to apply the proposed r-pi method to correlate fracture porosity and intensity is to improve in a wise way the
mode by which the software calculates fracture intensity from well data, regarding only spatial orientation and disregarding
fracture size and stratigraphic location. In the work of Ortega et al. (2006), a power law aperture-spacing relationship is
obtained regarding not only spatial orientation but also size and mechanical stratigraphic connections among the fractures of
the same family. So rearranging in opposite sense and pairing the distributions of fracture porosity and intensity from well
data imitates the natural tendency they follow in nature, leading to a model that replicates measurements in these parameters as
if they were made accounting for fracture size and stratigraphic location. Our starting point to apply the method is confident,
log and core- derived fracture porosity measurements. The key of the method is to rearrange in reverse order the well logderived fracture intensities and to match the resultant distributions with the fracture porosity distributions. Our results show
that high correlation indexes give support to the resultant models for five shear fracture families evaluated at different
stratigraphic intervals. When the results of our model are compared with the one derived from Ortega et al. (2006), a word of
caution should be mentioned. The Ortega et al. (2006) result is taken as an estimate of fracture intensity and porosity in
shallow regions geologically dominated by Type I fractures, while ours is as an estimate of shear fractures at depth.
The tendency and definition of the r-pi regression lines for the shear fractures measured at the Icotea fault plot with some
differences with the regresion line for the joint fractures measured at surface outcrops by Ortega et al. (2006).
The position in the plot of the r-pi regresion lines for the Icotea fault shear fracture families indicate that for the same
porosity the intensity of the shear fractures at depth is smaller than for the joint fractures at surface. Does this means that the
opening of the joint fractures at surface is greater than that of the studied shear fractures of similar size at depth?. We left the
conjecture open for future research in similar cases. The point to clarify in further research is whether the real intensities for
every porosity are similar or dissimilar to the ones obtained, provided that near-vertical well trajectories should collect
information from fractures from different low dipping beds rather than from the same bed.
Another point to consider is that the spacing between fractures measured for the software within the same family has
different sizes. In any case, we have the reference of a real porosity-intensity curve for Type I fractures provided by the
reformulation of the outcomes of Ortega et al. (2006). Nevertheless, the potential use for using the regression line for joints in
modeling porosity-intensity trends of shear fractures instead of the r-pi regression lines found for shear fractures would
introduce a bias in the model. The r-pi regression lines for the Icotea fault shear fracture families have slopes smaller than the
one for the regression line for joint fractures at surface. This indicates differences in the rate of correlative change between
porosity and fracture intensity. In other words, a change in porosity values of one order of magnitude for the studied joint
fractures corresponds to a change in order of magnitude of fracture intensity greater than that for shear fractures. So, for the
joint fractures studied by Ortega et al. (2006), the rate of change of fracture aperture with fracture size is greater than for most
of the studied shear fractures. At the Maraca-Lisure section, the highest r-pi slope corresponds to tension fractures and the
smallest to compressional fractures, with the slope of the other families spanning in between. With the exception of the
Compressional shears, at the Apon section the slopes of the other r-pi regression lines differ less than in the overlying section.
The main weakness we face using well data is the lack of control on the fracture geometry and stratigraphic location at
high resolution. We devise the r-pi method as an attempt to keep the condition of the scale independence of fracture intensity
and porosity over well data. Nevertheless, there is still uncertainty about a proper match between fracture porosity and its
truly associated fracture intensity. The improvements of the porosity-intensity correlation ratios using the r-pi method may just
indicate a proxy to the real correlation better than just the usual depth-based match of these properties.
A match between fracture spacing trends at seismic scale and those deduced from data at well scale is a plus for the
reliability of the model. This worked in our case, but this connection might not necessarily hold in other geological settings. In
this sense, this kind of cross validation of the method is limited to consideration as a general and global application. One big
advantage of the method presented here is the ease with which it deals with the usual outcomes of fracture reservoir modelling.
The straighforward statistical study of fracture features like intensity and petrophysical parameters and the methods
presented to connect them, like the r-pi, allows us to reduce the complexity of modelling these factors by more sophisticated
means. The savings in time and resources are benefits that we think will encourage readers to follow and adopt the techniques
presented.
We are confident that the fast progress in well logging and high resolution seismic technologies will bring into being
better and more confident approaches to estimate fracture geometries and its connection with the stratigraphical setting. If so,
enhancements of the method proposed in this paper will be available in the short term.
References
Berk, K. and Carey, P. 2000. Data Analysis With Microsoft Excel. Thompson Learning.
Blumentritt, C., Marfurt, K., and Sullivan, E. 2006. Volume Based Curvature Computations Illuminate Fracture Orientations, Lower-Mid
Paleozoic, Central Basin Platform, West Texas. Geophysics, 71(5).
Cohen, R. and Mercer, J. 1993. DNAPL Site Evaluation. CRC Press, Florida. Referenced in Kresic N. 2007. Hydrogeology and Groundwater
Modeling. CRC Press, Florida.
Cunningham, W. and Mann P. 2007. Tectonics of Strike-Slip Restraining and Releasing Bends. Geological Society of London Spec. Pub.
290.
Davis, G., Bump, A., Garca, P., and Ahlgren S. 1999. Conjugate Riedel Deformation Band Shear Zones. Journal of Structural Geology 22:
169-190.

14

SPE 130745

Ekneligoda, T. and Henkel, H. 2006.The Spacing Calculator SoftwareA Visual Basic Program To Calculate Spatial Properties of
Lineaments. Computers & Geosciences 32.
Escalona, A. and Mann, P. 2003. Three-Dimensional Structural Architecture and Evolution of the Eocene Pull-Apart Basin, Central
Maracaibo Basin, Venezuela. Marine and Petroleum Geology 20: 141161.
Gemignani, M. 1998. Calculus and Statistics. Dover Pub.
Klett, T., Schmoker, J., and Ahlbrandt, T. 2000. Assessment Hierarchy and Initial Province Ranking. U.S. Geological Survey World
Petroleum Assessment. U.S. Geological Survey Digital Data Series 60.
Koike, K. and Ichikawa, Y. 2006. Spatial Correlation Structures of Fracture Systems for Deriving a Scaling Law and Modeling Fracture
Distributions. Computers & Geosciences 32.
Lugo, J. 1991. Cretaceous to Neogene Tectonic Control on Sedimentation: Maracaibo Basin, Venezuela. Unpublished PhD dissertation, The
University of Texas at Austin, Austin, Texas.
McClay, K. and Dooley, T. 1994. 3-D Strike-Slip Pull-Apart Basins: Geometries Determined from Scaled Analogue Models. AAPG Search
and Discovery Article #90986.
Middleton, G. 2000. Data Analysis in the Earth Sciences Using Matlab. Prentice Hall.
Montgomery, D. 1991. Design and Analysis of Experiments. John Wiley & Sons.
Nelson, R. 2001. Geologic Analysis of Naturally Fractured Reservoirs. Second edition. Gulf Professional Publishing, Elsevier.
Nissen, S., Carr, T., and Marfurt, K. 2006. Using New 3-D Seismic Attributes To Identify Subtle Fracture Trends in Mid-Continent
Mississippian Carbonate Reservoirs: Dickman Field, Kansas. Search and Discovery Article #40189.
Ortega, O ., Marrett, R., and Laubach, S . 2006. A Scale-Independent Approach to Fracture Intensity and Average Spacing Measurement.
AAPG Bulletin 90(2): 193208.
Ozkaya, S. 2003. Fracture Length Estimation From Borehole Image Logs. Journal of Mathematical Geology 35 (6).
Peacock, D. 2001.The Temporal Relationship Between Joints and Faults. Jour. of Struct. Geol. 23: 329-341,
Prez, S., Landin, A. and Gudez, R. 2009. Structural Interpretation at the Vicinity of the Lama-Icotea Fault at the Cretaceous Section,
Lagomar Field, Maracaibo Basin, Venezuela. Int. Congress of the Brazilian Geophys. Soc. & EXPOGEF, Salvador de Bahia, Brazil.
Price, N. and Cosgrove, J. 1990. Analysis of Geological Structures. Cambridge Univ. Press.
Springer, C., Herlihy, R., Beggs, R.. and Mall, R. 1968. Probabilistic Models. Vol. 4 Mathematics for Management Series, Irwin, Ill.
Van Dijk, J.P., Bello, M., Toscano, C., Bersani, A., and Nardon, S. 2000.Tectonic Model and Three-Dimensional Fracture Network
Analysis of Monte Alpi (Southern Apennines). Tectonophysics 324 (4).
Zoback, M. 2007. Reservoir Geomechanics. Cambridge University Press.

You might also like