Professional Documents
Culture Documents
A N A L Y S I S OF P E R M E A B I L I T Y
APPROACH
C. A. CADE, I. J. EVANS
CONTROLS: A NEW
AND S. L . B R Y A N T *
BP Exploration Operating Company Limited, Uxbridge, Middlesex, and *BP Chemicals, Grangemouth, UK (Received 23 June 1993; revised 25 March 1994)
A B S T R A C T : By modelling a range of rock-forming processes such as compaction and various styles of cementation, a new understanding of how they affect pore-system geometry, and hence permeability, has been gained. For example, almost identical permeability-porosity trends result from progressive compaction or grain overgrowth cementation in a clean sandstone, and these trends are curvilinear on the traditional log-linear plot. The steepening which defines the curve marks the onset of pore-throat blocking. Other cement styles, such as pore-filling carbonates or grain-rimming clays, show different porosity-permeability trends. This new understanding can be used predictively (predicting permeability from predictions of grain size and diagenetic style), or as a tool for identifying the important permeability controls in a set of field data. This latter application is presented and illustrated using data from a variety of sandstone types. This approach has important advantages over commonly used multivariate statistical analysis approaches. It can quickly provide a good understanding of what are (and are not) the important controls on permeability. It also provides a basis for more focused and meaningful statistical analysis to quantify these controls.
T h e rate at which fluids can b e p r o d u c e d from a p o r o u s sub-surface reservoir rock is f u n d a m e n tally d e p e n d e n t o n t h r e e p a r a m e t e r s : the thickness of reservoir p r e s e n t , the pressure d r a w d o w n which can b e applied to the reservoir at the wellbore, and the p e r m e a b i l i t y of the reservoir. As flow rates are an i m p o r t a n t c o n t r i b u t o r to the e c o n o m i c viability of an oil or gas field, t h e n the accurate prediction of p e r m e a b i l i t y a h e a d of drilling can h a v e a m a j o r impact on business decisions such as prospect r a n k i n g or appraisal well location. F o r e x a m p l e , if a well in p r o s p e c t X (Fig. 1), p e n e t r a t e s m e d i u m grained s a n d s t o n e s with permeabilities in the range 100 m D to 500 m D , can we predict p e r m e a b i l i t y o~ t h e r e s e r v o i r in prospect Y, which from geological models we believe to b e similar b u t finer-grained? W h a t if the finergrained s a n d s t o n e in prospect Y has m o r e quartz c e m e n t , or contains less clay, or is b e t t e r sorted? Can we predict p e r m e a b i l i t y in prospect Y, b a s e d o n m e a s u r e d p e r m e a b i l i t y data from prospect X a n d o u r geological m o d e l ? T o a n s w e r t h e s e questions, the relative importance of different potential controls on p e r m e a b i lity (such as grain size, sorting, c o m p a c t i o n a n d
different styles of c e m e n t a t i o n ) m u s t be established. T h e effect of the m a j o r controls on p e r m e a b i l i t y must t h e n b e quantified. Identifying the impact of these controls, b o t h individually and in c o m b i n a t i o n , is the key to p e r m e a b i l i t y prediction. If variation in the v o l u m e of carbonate c e m e n t in a reservoir rock has a m u c h g r e a t e r impact o n p e r m e a b i l i t y t h a n v a r i a t i o n in grain size, but o u r ability to predict c e m e n t variation is limited, then o u r ability to predict p e r m e a b i l i t y is similarly limited, H o w e v e r , if grain size is the d o m i n a n t control, a n d the geological model
Med~n
sand
~ \ \
From ~ i o ~ c a l model
as Xg bt~t f o r e ~ ( a m p l c : ~ner SraLned o r mzre ~ cement or le~ clay o r belier sorted
492
C.A.
Cade
et al. determine the relationship between rock property variables and reservoir quality. Dutton & Diggs (1992) and Bloch (1991) described the most frequently used application of this approach, in which relationships between measured porosity and permeability (usually ambient helium porosity and single phase gas permeability), and textural and mineralogical variables (usually measured on thin-sections) were investigated. Commonly-used variables are grain size, sorting, the ratio of rigid to plastically-deformable grains, matrix clay content, volume of individual cements, total cement volume, and point-counted interparticle porosity. Bloch (1991) stated that multivariate statistical techniques will not work as a predictive tool in some situations (because the key permeability controls cannot themselves be predicted). However, he suggested that "despite its limitations, the empirical technique provides the only feasible approach to reservoir quality prediction". Dutton & Diggs (1992) concluded that predicting porosity using statistical analysis of a calibration data set may be feasible, but that prediction of permeability will have much lower success. We suggest that this lack of success is largely a consequence of the interplay and overprinting of the various permeability controls. If all possible permeability controls are considered together, the statistical processing may show that both grain-size and quartz cementation are of moderate importance as permeability controls, as there is no strong relationship between either parameter and permeability. However, within coarse or medium sand grain-size classes, the volume of quartz cement may be the dominant control on permeability. If the textural controls are first removed, then the important diagenetic overprint controls should become clearer. A variation on the empirical approach is described by Ehrlich e t a l . (1991). Using the observation that, even in a single formation, permeability commonly varies by several orders of magnitude, they concluded that the configuration of porosity, rather than the absolute porosity value, is the control on permeability. To characterize the pore system configuration, Ehrlich e t a l . (1991), made measurements of pores in two dimensions (on polished thin-sections) and combined these with pore-throat size distribution data (from mercury porosimetry) to develop a simple pore-system model. For selected data sets
allows prediction of this parameter, then permeability prediction may be feasible. REVIEW OF PREVIOUS WORK
In numerous previous studies using statistical techniques, a strong correlation between permeability and porosity has been shown. An example is a study on the Cretaceous Travis Peak Formation in East Texas, USA (Dutton & Diggs, 1992). Using a data set of over 600 samples, the correlation coefficient of permeability on porosity is 0.79, higher than any other textural or diagenetic parameters included in multiple regression. With this high correlation coefficient, it might be expected that predictions of porosity could be used to predict permeability. A cross-plot of the data from this study (Fig. 2), however, shows that, for a given porosity, permeability varies by c. three orders of magnitude. Porosity alone cannot, therefore, be used to predict permeability. In a recent review, Bloch (1991) described two general approaches to predicting reservoir quality in sub-surface sandstones: empirical techniques and 'process-oriented' techniques. The work on the Travis Peak Formation is an example of the empirical approach to the quantification of permeability controls for predictive purposes. Empirical techniques use a calibration data set (e.g. data from core samples from a well in prospect X) and multiple regression analysis to
1000 i
E
1".. 17.,::.{
9
,0
............................. i..............
{ 9
9
i
.........................
9 !
0,001
_ _~J ~, 9
z : ::-~ ~ . . ~ : :
~'~t~-'t;I
I
...................... + ...........................
~l .............
.................
............................
............
lO'S
0
10 15 Po~ros~, ( . )
20
25
Peak Sandstonesamples. Linear regression of porosity on permeabilitygives a correlationcoefficientof 0.79 (Dutton & Diggs, 1992).
493
a good degree of fit between the simple poresystem model and measured permeability has been established. While the method presented by Ehrlich et al. (1991) has merit in its focus on pore type and connectivity, it has limitations as a predictive tool. For predictic, n of permeability ahead of drilling, the criteria for success of any method must be that it establishes a quantitative link between meast:red permeability and other rock parameters, and that those correlative parameters can themselves be predicted from a geological model. With the current level of understanding, it is very unlikely that we would be able to predict with confidence the pore-type and pore-throat size distribution parameters in an undrilled sandstone from a geological model. The 'process-oriented' approach described by Bloch (1991) focuses on modelling diagenetic processes in an undrilled area, based on chemical and mathematical models, and the effects of those processes on reservoir quality. There are two important limitations to this approach. First, there is the uncertainty associated with the subsurface geological model, and how that impacts on the thermodynamics and kinetics of the diagenetic model. Second, there is the lack of a detailed quantitative understanding of how diagenetic processes control permeability. To date, the quantification of the impact of specific controls, particularly diagenetic controls, has been either formation specific (and therefore not generally applicable) or very general. Ethier & King (1991) illustrated a general understanding of a variety of controls (Fig. 3), but with little or no quantitative detail, the value of such trends is limited. To summarize, previous work on permeability controls has been limited by the lack of a quantitative understanding of how textural, and in particular, diage~aetic, permeability controls operate, and has been hampered by statistical approaches which consider all possible controls together. In this paper we present research findings which address the quantitative understanding of permeability controls, and propose a modified approach to reservoir quality analysis which can improve predictive capability. AN ALTERNATIVE APPROACH
model of this sphere pack, we simulate diagenetic processes, such as compaction and various styles of cementation. Their progressive impact on porosity, and more importantly permeability, is measured. The result of these simulations is an improved understanding of the ways in which permeability is influenced by different diagenetic modifications. The controls are considered both in isolation and in combination. Over a full range of porosities (<5-36%), the effect of grain-size, compaction and the most commonly-occurring cement styles (overgrowths, pore-filling cements and grain-rimming clays), on the permeability of a simple sandstone, is established. The theoretical details of the modelling which provide the basis for this alternative approach are outlined in Bryant et al., (1993a,b). This paper builds on this work by demonstrating the application of the modelling results to the analysis of field datasets.
SPHERE PACK MODELLING OF SANDSTONE DIAGENESIS A sphere pack forms the basis for our modelling. During research into the structure of liquids, Finney (1970) constructed a random pack of ball bearings. This was shaken during packing such that the porosity of the pack reached a minimum. The result was a random, close packing. All the bearings were the same size, and the pack
INCP~.AS[NG GRAIN SIZE
i~
.//e
.~
PORIXSlTY
~7.o%~"~o ~
Our approach is based on a simple model of a sandstone, a sphere pack. Using a numerical
Fro. 3. Diagramto illustratean interpretationof the effects of various controls on porosity and permeability (after Ethier & King, 1992).
494
porosity was 36.2%. The pack was then impregnated with wax and dismantled, one bearing at a time, and the relative positions of all of the bearings measured. With the positions of all the grains known, and because all the spheres were the same size, the interparticle pore space is, a priori, completely characterized. We believe this to be the only real pore-system which has been so completely characterized, and it is this feature which allows the calculation of porosity and permeability. Permeability is calculated by a rigorous geometrical conversion of the Finney pore-system into a network comprising pores and connecting throats (Bryant et al., 1993a). These pore-throats, and more specifically their diameter and 'interconnectedness', are the key to sandstone permeability. Using data from this pore-system, a range of different diagenetic processes has been modelled (Fig. 4).
Compaction
Compaction is one of the most important processes of porosity reduction during sandstone diagenesis. It can be defined as reduction in porosity as a consequence of the reduction in bulk rock volume. The volume reduction is usually the result of vertical shortening, and that in turn is
eli
FIG. 4. Numerical simulation of geological processes. (a) 2D slice through a dense packing of equal spheres representing sediment of clean, well-sorted sand. (b) Compaction forces spheres closer together and interpenetration occurs. (c) Quartz overgrowths cement the compacted packing together. (d) Pore-filling cement.
Pore-filling cements"
Several minerals precipitate as pore-filling, rather than grain-rimming, cements. Carbonates (e.g. calcite and siderite), and some clays (e.g. kaolinite) often occur in this form. W e model this style of cement by completely filling randomly
495
'~
.1
Combined cases"
We have modelled combined cases, e.g. a phase of compaction, followed by either a phase of overgrowth or pore-filling cementation, and progressive compaction with pressure dissolved grain material (grain interpenetration) reprecipitated on the grain surfaces.
0,
.OOl
. . . . . . . . . . .
ok
.0001
.00001
~.
o.o
l~,o~ity (v/v)
Model results
It is well known that, for a given porosity, permeability will be lower in a finer grained sandstone than in one that is coarser grained (all else being equal), and that a pore-bridging clay such as illite is very effective in reducing permeability. The sphere pack modelling provides physical confirmation, explanation, and quantification of these controls, and provides similar insight into how other, previously less well understood, controls operate. In each case, porosity is reduced incrementally according to the style of cementation and the resultant permeability is calculated (details of the calculation are lengthy and are given in Bryant et al., 1993b). We present the permeability-porosity trends which result from the simulations as lOgl0 permeability vs. porosity cross-plots. In Fig. 5, permeability is presented in the dimensionless form K/Ko where Ko is the initial permeability of the packing before compaction a n d / o r cementation. This initial permeability (in Darcies) depends on the size of the spheres in the packing such that Ko = 3 x 10 -3 r 2, where r is the sphere radius measured in microns. This expression is obtained from network model calculations and agrees with many measurements and semi-empirical expressions such as the Kozeny-Carman relationship (Carman, 1937). In other figures, particular grain-size cases are presented and permeability is not normalized to the initial value.
F[c. 5. Model-derived porosity-permeability trends for compaction, cementation, and a combination of these processes. Note that in this cross-plot, permeability is expressed as a ratio of original uncompacted/uncemented permeability. This removes any grain-size dependence. The dashed line shows the percentage of pore-throats blocked as porosity decreases.
curved, with a maximum rate of change of slope at --12% porosity (Fig. 5). This steepening coincides with the porosity at which the blocking of pore-throats begins.
Overgrowth cementation
Progressive quartz style overgrowth cementation produces an almost identical porositypermeability trend to the compaction trend (Fig. 5). Divergence of the compaction and overgrowth trends only occurs below 5% porosity, and this divergence is small. Combinations of compaction and overgrowth cementation produce the same curved trend. The significance of this trend coincidence is that any divergence from the curve, for a given grain size and sorting class, must be due to controls other than compaction or quartz cementation.
Pore-filling cement
Progressive cementation by a pore-filling cement produces a steeper decline in permeability with porosity (Fig. 6a). The reason for this is that blocking of pore-throats occurs as soon as the first pore is cemented, rather than when porosity has
Compac~on
The trend of declining porosity and permeability for a progressively compacted sandstone is
Plastically-deformable grains
Many sandstones contain grains such as mudstone clasts or metamorphic grains which will deform in a plastic manner during burial. Although we have not modelled a pack containing such grains, we believe that the porositypermeability trend would be similar to that for progressive pore-filling cementation. The reason for this suggestion is that as the plastic grains deform, they fill adjacent interparticle porespace. This process will cause pore-throat blocking earlier than with the quartz grain compaction case discussed earlier. 100000
1000
0.1[I 1~
j
1
*---y-,heo po~e~,,,~
Some compaction
10
no compaction ! i Some compaction then clay cement
I I !
!
l
0"1i 0.001 10
0.001
.
lo
, .
2;
Porosity (%)
' 3b
20
Porosity (%)
30
40
FIG. 6. Model-derived porosity-permeability trends for simulations of (a) pore-fillingcementation and (b) grain-rimming clay.
Analysisof permeabilitycontrols
A N A L Y S I S OF P E R M E A B I L I T Y CONTROLS
497
The approach
The geometry and scale of the primary interparticle pore-system is fundamentally controlled by the size of the sand grains which define it. Diagenetic modifications are then overlaid on, or result in the deformation of, this primary pore-system template. R a t h e r than carry out an overall, multivariate, statistical analysis on all possible permeability controls in a set of data, we propose that a better result is achieved if the primary textural controls are r e m o v e d from the analysis at the outset. Thus the first step in this approach is to split the sample data set into grain size (and, if possible sorting) classes. For each textural group, the porosity-permeability trend is solely the result of diagenetic modification. By displaying each grain size data group separately, with the model derived curves for the corresponding texture, a qualitative assessment of the permeability controls can be made. For a wide range of data sets, we have found that, with just this initial stage, a good insight is gained into what the important permeability controls are. This insight is the result of two simple steps, splitting the data by primary texture, and comparison with results of sphere pack modelling. At the initial display stage, after textural splitting, several conclusions can be made about
the data set. A n y data points failing well above the compaction/overgrowth cement curve (i.e. higher permeability for a given porosity), should be carefully investigated for data quality. A c o m m o n explanation for such points (low porosity but high permeability) is the existence of an open fracture in the plug used for permeability measurement. This may be a natural fracture, or more likely it will have been induced during plugcutting, but in either case, such data should be removed from the data set before further analysis. The permeability of such plugs is dominated by the permeability of the fracture and not the surrounding matrix rock. For data points which fall just above the model curve, the quality of the grain size data should be checked. As it is based on a perfectly sorted sphere pack, the model curve provides an upper limit on porosity-permeability values for a sand with that texture. The general trend of the data within each grain size class should point to the important controls on permeability. If the data lies on or just below the model trend, and parallels that trend, then compaction and/or overgrowth cementation are probably the dominant controls. If the trend is steeper, then there must be some contribution from a pore-filling or microporous cement. It should be noted here that a steeper trend does not preclude a contribution from overgrowth cementation or compaction. A f t e r removing the influence of primary textural controls, the diagenetic overprint can be
1ooo 1oo 1o 1
100000!
b
Very well
well a n d
10OOOi
Z;'"
1ooo
~,
lOO
.I .o~ .ooI
.OOOl o 1o 20
[/
1//
Y~ I
or~d
9 %
Y
,
250 125 !
,oo
l
1000
,
.01
sol
rerypoorly 9
sorted
62.5
30
.001
o
lO
20
Porosl~
30 (%)
40
Porosity(%)
FIG. 7. The effect of grain-size and sorting variation on the model-derived porosity-permeability trends: (a) calculated trends for different grain sizes; (b) empirically-derived trends for different sorting classes.
498
10000
I000
100
!
.1
10
1S
20
25
30
Porosity (o/0) FJ6.8. Fontainebleau Sandstone--porosity vs. permeability cross-plot showing the close correspondence between model predictions based on grain size and cement style, and measured porosity-permeability data.
evaluated. This can be done by using multivariate statistical analysis, or by performing further data splits, e.g. by variation in the volume of carbonate cement. By using statistical analysis after the textural controls are r e m o v e d , there is a much reduced possibility of an important diagenetic control on permeability being masked by a textural parameter, and vice versa. This approach to reservoir quality analysis is illustrated in the following case examples.
499
'~
/
I
0
/ i" i.I. i I
10 20
Compacli0n/QuartzOnly
i
,0
iI
Porosity (%) Fro. 9. Porosity vs. permeability cross-plot for Garn Formation sandstones. Porosity-permeability values are averages for wells with different geographical locations and burial histories (data from Ehrenberg, 1989). The curves are model-derived and match the measured textural and diagenetic characteristics of the Garn Formation sandstones. on, or close to, this model trend. This supports the conclusions of E h r e n b e r g (1990) and provides further confirmation that our models are realistic representations of real sandstones. Also shown on Fig. 9 are two model trends for combined cases of compaction/quartz cementation, followed by grain-rimming clay cementation. The two cases differ in the degree to which porosity is first reduced by compaction/quartz cementation (to 25% in one case, and to 15% in the other). As is the case in example 1, the model trends, constructed using textural data from thinsections, and authigenic mineralogy data from Xray diffraction analysis, provide a full explanation of the measured porosity-permeability data averages. The averages from the deeper wells with more illite are enclosed by the two combined mechanism model trends, in more detail, increasing illite content is reflected in progressively lower permeabilities. Having explained the observed trends in the porosity-permeability data using the model results, this understanding can be used predictively. For example, if a Garn Formation prospect is similar to the drilled occurrences, but with finesand grain size, then the model-derived trends for that finer grain size, in combination with burial history information, could be used to make a predrill reservoir quality prediction.
A large data set from Mississippian age sandstones in a well in the Endicott Field has been analysed using the model results. These sandstones have a more complex suite of textural and diagenetic controls on reservoir quality than the preceding cases, and this is a 'blind' analysis, with no prior statistical work on permeability controls. Samples from this well show widely varying grain sizes, with average grain size ranging from fine-sand to very coarse sand or greater. The data set is first split by grain-size class (Fig. 10a). O n e model trend, for a moderately sorted, medium sandstone undergoing compaction/quartz cementation is shown on the plot for guidance. F r o m this initial plot several conclusions can be drawn. Firstly, within each grain-size class, the porositypermeability trend is steeper than the simple model trend, indicating that the dominant reservoir quality control(s) are pore-filling cementation and/or pore-lining/filling clay. Secondly, there is considerable overlap between the grainsize classes, which indicates that the diagenetic controls on porosity and permeability have reached the stage of development where they mask the primary textural control. H o w e v e r , some remnant textural control is still evident: e.g. the very coarse sand samples have a higher permeability for a given porosity than the medium sand samples, whereas fine to medium sand grade samples have lower permeability. Having established that the textural effect has been largely, but not completely, swamped by the diagenetic overprint, each textural group in the data set is then analysed on its own, to establish what the important diagenetic controls on reservoir quality are. The analysis of the medium sand grade sample set is illustrated in Fig. 10b. The medium sand data points are coded by clay content (point-count data), and a clear relationship of decreasing permeability with increasing clay content can be seen. Several of the medium sand grade samples have porosity-permeability values which fall on or close to the model compaction/quartz cementation trend. As we would predict, these samples all have a low clay content, and there is a trend for lower porosity corresponding with higher volumes of quartz cement. It can be concluded, therefore, that the major causes of reservoir quality variation in
5O0
C . A . Cade
et al.
these sandstones are variation in the volumes of quartz cement and pore-filling clay, which largely obscure the primary textural control. DISCUSSION The three case studies described illustrate, for sandstones of varying textural and diagenetic complexity, how the sphere pack models replicate the results of diagenetic processes in real sandstones. This has important consequences. The modelling has provided a quantified understanding of how different styles of cement or compaction influence pores and pore-throats. This understanding permits the identification of important reservoir quality controls from porosity-permeability trends. In simple sandstone systems, it may be possible to draw conclusions about controls using just the porosity-permeability data and textural information from cuttings, sidewall core or core description. Clearly, there is less uncertainty if more detailed petrographic data on texture and cements are available. In numerous cases, with a wide range of sandstone types, the model trends have given a quick and accurate insight into the permeability controls. Taking the results a step further, sensitivities can be investigated, and predictions made. For example, for a given sandstone type, investi-
gate the grain size or cement volume at which permeabilities are reduced below the threshold for economic flow rates can be investigated. It should be emphasized that only single phase permeability has been modelled to date. Potential extensions of this work include modelling of packs with a range of grain sizes and two phase permeability.
CONCLUSIONS (1) By modelling the results of compaction and various styles of cementation in a numerical sphere pack and calculating porosity-permeability trends for these diagenetic models, a good match between porosity-permeability data for real and model sandstones has been achieved. (2) The sphere pack models give an improved and quantified understanding of how a range of textural and diagenetic reservoir quality controls operate. (3) Progressive compaction and quartz overgrowth cementation produce almost identical porosity-permeability decline trends. (4) Pore-filling cements (such as carbonates), and grain-rimming clays cause a steeper permeability decline trend with porosity than compaction or quartz cementation.
100000~ 41 = FSK
loooo -JI o
|1 +
Ms K
M-C K m~
. / "
t1"
,
,oo.
10
0 10
,ko
~,
I
o 7-,0~ol,y
o ,.19~:1,y
4-e~cJ.y
I I
O ~
20 30 40
i l II
4o
lo
ZO
Poeo~aty (*/o)
3o
eoroslW ( % )
FIe. 10. Porosity vs. permeability cross-plots for Mississippian age sandstones from a well in the Endicott field, Alaska, USA; (a) plot illustrating the diagenetic overprint of the primary textural control on reservoir quality; (b) plot showing the diagenetic controls operating within one grain-size class.
501 REFERENCES
(5) The sphere pack modelling has underlined the importance of considering the primary (textural), and secondary (diagenetic) controls on reservoir quality separately. By removing the textural control first, analysis of the diagenetic controls is facilitated. (6) In order to analyse a set of data to establish the reservoir quality controls we suggest the first stage should be to split the data set by grain-size class (and if appropriate, sorting class). If textural control is still important, it will be evident after this split. Within each grain-size class, the diagenetic controls can then be evaluated, both qualitatively using comparison with the model-derived trends, and quantitatively using statistical analysis, independent of texture. (7) In order to predict permeability ahead of drilling, it is necessary to be able to predict the controlling parameters. By combining the sphere pack model results with an approach which separates textural and diagenetic parameters, the potential for defining the important controlling parameters is maximized.
ACKNOWLEDGMENTS The permission of BP Exploration to publish the results and application of this research is acknowledged. In addition, permission has been granted by the Endicott partner, group: Exxon, Unocal, Amoco, BP, Arco, Cook Inlet, Nana and Doyon, to utilize data from this field.
BEARDD.C. & WEYLP.K. (1973) Influence of texture on porosity and permeability of unconsolidated sand. A A P G Bull. 57, 349-369. BLOCH S. (1991) Empirical prediction of porosity and permeability in sandstones. A A P G Bull. 75, 1145-1160. BOURBIET. & ZINZSNERB. (1985) Hydraulic and acoustic properties as a function of porosity in Fontainebleau sandstone. J. Geophys. Res. 90, 11524-11532. BRYANT S.L., CADE C.A. & MELLOR D.W. (1993a) Permeability prediction from geological models. A A P G Bull. 77, 1338-1350. BRYANT S.L., MELLOR D.W. & CADE C.A. (1993b) Physically representative network models of transport in porous media. A1ChE J. 39, 387-396. CARMAN P.C. (1937) Fluid flow through granular beds. Trans. Inst. Chem. Eng. 15, 150-166. CAYEUX L. (1929) Les roches sedimentaires de France. Roches siliceuses. Mem. Min. Travaux Publics, Paris, 48-49. DuvroN S.P. &Dmcs T.N. (1992) Evolution of porosity and permeability in the Lower Cretaceous Travis Peak Formation, East Texas. A A P G Bull. 76, 252-269. EHRENSERGS.N. (1990) Relationship between diagenesis and reservoir quality in sandstones of the Garn Formation, Haltenbanken, Mid-Norwegian Continental Shelf. A A P G Bull. 74, 1538-1558. EHRLICH R., ETmS E.I., BRUMFIELDD., YUAN L.P. & CRABTREES.J. (1991) Petrography and reservoir physics III: Physical models for permeability and formation factor. A A P G Bull. 75, 1579-1592. ETHJER V.G. & KiNG H.R. (1991) Reservoir quality evaluation from visual attributes on rock surfaces: methods of estimation and classification from drill cuttings or cores. Bull. Can. Petrol. Geol. 39, 233-251. FmNEY J. (1970) Random packings and the structure of simple liquids I. The geometry of random close packing. Proc. Roy. Soc. 319A, 479.