Geometry and origin of a polygonal fault system · 2015-07-29 · There is no preferred apparent dip direction of faults on individual cross-sections (Fig. 3) but larger faults often

Journal of the Geological Society, London, Vol. 157, 2000, pp. 151–162. Printed in Great Britain.

Geometry and origin of a polygonal fault system

J. WATTERSON, J. WALSH, A. NICOL2, P. A. R. NELL & P. G. BRETAN1

Fault Analysis Group, Department of Earth Sciences, University of Liverpool, Liverpool L69 3BX, UK(email: [email protected])

1Badley Earth Sciences Ltd, North Beck House, North Beck Lane, Hundleby, Lincs PE23 5NB, UK2Present address: Institute of Geological & Nuclear Sciences, PO Box 30-368, Lower Hutt, New Zealand

Abstract: A fault array in South Australia, interpreted from a 3D onshore seismic survey, shows faulttraces on the lowermost mapped horizon of a shale-dominated sequence which outline polygonal cellsaveraging 1.4 km in diameter. The cell boundaries coincide approximately with the downward termina-tions and near convergence of conjugate pairs of normal faults. The pattern becomes less spatially orderedon higher horizons where faults still show a near-isotropic strike distribution. Maximum throws, c. 80 m,occur c. 400 m above the downward terminations of the faults. The faults have a systematic geometricrelationship with folds, with anticlines in the mutual hanging walls of fault pairs and broader footwallsynclines that define the shallow dish forms of the polygons. Polygon boundaries coincide with anticlinalridges on the interface between the faulted sequence and an underlying 35 m thick low velocity, lowdensity, overpressured layer. Although the pattern of ridges defining the polygon boundaries is strikinglysimilar to experimental spoke and hub patterns formed at the boundaries between viscous materials withdensity inversion, the data do not exclude the possibility of lateral extension.

Keywords: faults, geometry, conjugate faults, seismic surveys, diapirs.

This paper provides an analysis and interpretation of a poly-gonal fault system imaged by a high quality onshore 3Dseismic reflection survey of 138 km2 in the Lake Hope region,South Australia (Fig. 1a). The term polygonal is used to refer

to a system of numerous multi-directional and relatively smallnormal faults (throw <80 m), many of which have abuttingintersections. The near-isotropic structural fabric of the LakeHope system is a characteristic of intraformational fault

Fig. 1. (a) Location map for study area. Jurassic hydrocarbon reservoirs are shown (black). (b) Stratigraphic column for Jurassic–Cretaceoussequence Well A, Eromanga Basin (see Fig. 2 for location). ‘Litho.’ refers to lithostratigraphy, with sandstone (stipple) and mudstone (brokenlines) distinguished. Modified from Oldham & Gibbins (1995). (c) Sonic velocity log for Well A.

151

systems, which are often polygonal, and which clearly distin-guishes them from typical tectonic fault systems. The relativelyrecent recognition of this type of fault and fault system wasmade possible by improvements in both the vertical and lateralresolutions of seismic-reflection surveys, allowing both themapping of fault throws in the range 10–80 m and the objec-tive definition of complex patterns of relatively short faulttraces. The existence of polygonal fault systems poses severalquestions including the outstanding one of how a normalfault system can be formed with neither significant regionalextension nor the restricted strike distribution and consistentdip direction which characterize a typical gravity-driven faultsystem. Other questions concern the restricted vertical intervalswithin which such fault systems are confined, the relativelysmall maximum fault throws (c. 80 m) and how these systemsaccommodate the numerous interactions between individualfaults which inevitably occur when non-vertical faults have awide range of strike directions. This last question is notaddressed in this paper.

Although the consensus is that polygonal systems are ofnon-tectonic origin, they and related intraformational systemshave been variously attributed to gravitational spreading/sliding (Higgs & McClay 1993), to gravitational instability dueto density inversion and associated dewatering (Henriet et al.1991; Verschuren 1992; Cartwright 1994), and to volumetriccontraction during compactional dewatering (Cartwright &Lonergan 1996). Whatever their origin, polygonal fault sys-tems become less isotropic with increase in regional dip and aretruly polygonal only when the regional dip is near horizontal(Cartwright 1994). With significant regional dips (e.g. >c. 1�),comparable fault systems tend to form parallel and polarizedarrays normal to the regional dip direction, with faults dippingup-slope, as in the example described by Higgs & McClay(1993). Polygonal and related fault arrays are widespread bothglobally and in the Tertiary sequence of the North Sea wherethey occur throughout an area of c. 100 000 km2 (Cartwright1994).

The Lake Hope fault system (Figs 2–4) is confined mainly tothe Lower Cretaceous sequence (c. 600–1000 m thick) in thesouthern part of Cooper/Eromanga Basins (Fig. 1a & b; alsoCartwright & Lonergan 1997). The Cretaceous sequence istypically underlain by Permian–Jurassic non-marine sand-stones and siltstones of variable thickness that contain severalhydrocarbon accumulations; this sequence unconformablyoverlies a heavily faulted Carboniferous sequence of platformcarbonates and volcanics. The Lake Hope polygonal faultsystem (Oldham & Gibbins 1995), in common with compar-able fault systems elsewhere (Henriet et al. 1991; Higgs &McClay 1993; Cartwright & Lonergan 1996), occurs within apost-rift mudstone-dominated marine sequence.

The description of the fault system and of individual faults isbased principally on analysis of three mapped horizons (TopCadna-owie Formation (oldest), Top Coorikiana Sandstone,and Top Mackunda Formation) within the Lower Cretaceoussequence (Fig. 1b). These three horizons span a stratigraphicthickness of c. 550–700 m, although the bulk of the faults lieswithin a half of the faulted interval (Figs 3 & 5). Selective pickswere made on up to six other horizons to define more closelythe throw patterns on some faults. Fault traces on theCadna-owie horizon are mostly obtained by extrapolatingtraces detected on marginally shallower horizons as throws onthe Cadna-owie are either marginally above or are less than thevertical resolution of the seismic data (i.e. c. 5 ms or 6 m).Analysis of the structures imaged in the seismic data was

carried out mostly in the time domain, which does not affectthe conclusions.

Many features of the fault system are consistent with ithaving formed in response to embryonic gravitational overturnof a relatively thin (c. 35 m), low density, gravitationallyunstable, over-pressured and mobile mudstone layer whichunderlies the hydrostatically pressured, dominantly mudstone,faulted sequence. However, the data do not exclude thepossibility of non-uniform lateral extension of the sequence,with conjugate fault pairs representing boudinage-likediscontinuities.

General fault characteristicsThe Lake Hope faults span a limited size range with maximumthrows ranging up to c. 80 m (1 ms two-way time=1.27 m,see Fig. 1c) and, except for some of the larger faults(throws>c. 40 m), they are restricted to the Lower Cretaceoussequence. In cross-section most of the faults with throws>30 m offset all three of the main interpreted horizons whilefaults with maximum throws <15 m usually do not (Fig. 3).Maximum throws mostly occur on the Coorikiana horizon, inthe lower half of the Lower Cretaceous sequence, with throwson individual faults decreasing more rapidly downwards thanupwards. A plot of aggregate fault throws, averaged along fourcross-sections, on each of six interpreted horizons also showsstrong maxima at or just below the Coorikiana, with aggregatethrows decreasing faster downwards than upwards (Fig. 5b).There is no preferred apparent dip direction of faults onindividual cross-sections (Fig. 3) but larger faults often formconjugate pairs which converge downwards to intersect at, orimmediately below, the Top Cadna-owie horizon (Fig. 3).Individual faults maintain constant dips, mostly within therange 50–55� but with extreme values ranging from 44� to 61�(Fig. 5a). The strike directions of individual faults can vary byup to 180� due to fault linkage and capture, the unusualfrequency of which is due to the wide range of initial strikedirections.

Fault array characteristicsOn all three horizon maps the faults show no overall preferreddirection of either strike (Fig. 4) or dip, resulting in thecharacteristic isotropic primary fault pattern. As the faults arenon-vertical any spatial ordering, or uniform pattern, ofmulti-directional faults on one horizon will not persist tooverlying or underlying horizons. The lowermost (Cadna-owie) horizon (Fig. 4c) has the most spatially ordered, i.e. leastrandom, pattern of fault traces with a distinct polygonalcellular pattern with cell diameters mostly within the range0.8–2.0 km. On this horizon several polygon boundaries aredefined by closely spaced parallel faults separated by grabenwhich are typically c. 70–150 m wide. The less spatially orderedfault-trace patterns on the upper two horizon maps (Fig. 4a &b) are polygonal only in the more general sense of beingmulti-directional, but are more representative of the faultsystem as a whole than is the Cadna-owie pattern. Theapparent decreases in fault density above and below theCoorikiana horizon are due as much to the consistently higherfault throws on the Coorikiana, expressed as wider and longerfault heave separations, or fault polygons, (e.g. Fig. 4) as to thenumerous small faults which offset only the Coorikianahorizon.

152 J. WATTERSON ET AL.

Fig. 2. Two-way time structure contourmap for Coorikiana horizon, Lake Hope3D seismic survey (see Fig. 1 forlocation). Reds and yellows representhigher elevations (total elevation range is932–1276 ms two-way time). Maximumslopes on the domal structure are <1�.Solid lines a and b locate the in-lineand cross-line sections shown in Fig. 3.The white cross shows the position ofWell A.

Fig. 3. Seismic sections in in-line (a) and cross-line (b) directions (locations as in Fig. 2) with interpreted faults and horizons (Cadna-owie, Ca;Coorikiana, Co; Mackunda, M). Vertical exaggeration is c. 2.4.

ORIGIN OF A POLYGONAL FAULT SYSTEM 153

The Cadna-owie horizon map shows fewer fault-traceintersections than the overlying Coorikiana and Mackundahorizons. As the frequency of non-conjugate intersections willincrease with vertical distance from the most spatially orderedhorizon, the more complex patterns on the Coorikiana andMackunda horizon maps are interpreted as derivatives of thespatially ordered pattern on the Cadna-owie. Implicit in thisinterpretation of the horizon maps is the idea that the faultpattern either originated at or close to the Cadna-owie horizonor was defined by features at or near this depth.

A second order element of the map pattern, most evident onthe Mackunda horizon map, is a poorly defined annular ringdelimited by an increased density of more or less concentricallydisposed larger faults (Fig. 4) centred on a domed structuralhigh (Fig. 2). This ring decreases in prominence with increasingdepth and is barely perceptible on the Cadna-owie. Variationsin the prominence of the annular ring are a function of thevertical changes in fault densities and fault sizes (i.e. tracelengths and displacements), rather than an upward increase inamplitude of the structural high. Downslope of the annularconcentration of concentric faults the fault pattern is weaklyradial with faults striking at high angles to the horizon depthcontours. These non-isotropic elements of the fault distribu-tion and orientation are thought to be a consequence of thenon-orthogonal relationship between the inclined mechanicallayering associated with the dome, and the gravity field. Thestructural dips off the high which result in the anisotropy of the

Fig. 4. Fault maps of the (a) Mackunda, (b) Coorikiana and (c)Cadna-owie horizons. Fault-trace widths are proportional to throw.

Fig. 5. (a) Depth-converted fault profiles constructed using seismicsections oriented normal to the strike of each fault and passingthrough the maximum displacement. The single fault trace extendingbelow the Cadna-owie appears to be an upward continuation of acompactional fault in the pre-Cadna-owie sequence. (b) Aggregatethrow (filled circles) and heave (open circles) divided by line lengthfor all faults intersecting six horizons (Ca, Cadna-owie; TLD, toplow density layer; unnamed horizon; Co, Coorikiana; unnamedhorizon; M, Mackunda) on four cross-sections (two in-lines and twocross-lines) across the mapped area. Assumed seismic velocities arederived from Fig. 1c. Note that below the Coorikiana the heavevalues are greater than the throw values even though the fault dipsare >45�. This inconsistency reflects the difficulty of accuratedetermination of heave values from seismic sections so the throwsshould be taken as definitive.


fault pattern are in the range 0.3–0.9� (average 0.5�), about athird of the regional dip of c. 1.5� which resulted in the parallelfault array described by Higgs & McClay (1993).

An outstanding feature of all three fault maps (Fig. 4) is theabsence of a preferred direction of apparent extension, asis also the case in the polygonal fault system in thenear-horizontal Tertiary sequence of the Central North Sea(Cartwright 1994). The question of whether the faulting was aresponse to either radial extension from, or radial contractiontowards a number of fixed points represented by polygoncentres (Cartwright & Lonergan 1996), or to neither of thesewas investigated by measuring the true surface area of afaulted horizon. The aggregate area of the fault heave separa-tions is usually taken as a measure of the extension of a faultedarea. On this basis the Coorikiana horizon map area hasincreased by 5.28%, which, if extension was equal in alldirections, is equivalent to a radial extension of 2.6%. Recog-nizing the unlikely boundary conditions implied by regionalradial extension, Cartwright & Lonergan (1996) proposed thatthe extension implied by the bed-length plus heave separations,which they give as c. 10%, was matched by a lateral contrac-tion of the faulted layers due to volume loss, ascribed tocompaction and de-watering. To conclude from these data thatbeds have contracted radially by 10% requires that they arehorizontal, which they are not. The surface area of a non-horizontal bed is greater than that of the region which itoccupies. The true area of the Coorikiana horizon surface,exclusive of fault polygons, was measured as 98.22% of thetotal mapped area, the apparent 1.78% loss of horizon areabeing equivalent to a radial linear contraction of 0.89%. Theequivalent values for apparent linear contractions of theMackunda and Cadna-owie horizons are 1% and 0.241%respectively. The areal bed contraction implied by the data ofCartwright & Lonergan (1996) ranges from c. 12% to 41%. TheLake Hope measurements are relatively insensitive to velocityuncertainties but are well within the resolution limits of thehorizon maps, in respect of the definition of both polygonsand horizon dips (see text to Fig. 5b), even though the mapswere made carefully with such measurements in mind. Heavevalues based on fault polygon widths on horizon mapsgenerally are greater than the true values and the amount bywhich a polygon width is exaggerated reduces the horizonarea measured by either an equivalent or a greater amount,because horizons have their steepest dips adjacent to faults.For example, the heave polygon area of the Coorikianahorizon could be over-estimated by 30%, given the likelyerrors attached to definition of fault polygon widths. Theeffects of post-faulting compaction, which would havedecreased the area of a non-horizontal horizon, have notbeen incorporated in our calculations. The horizon areameasurements thus provide no basis for demonstratingeither lateral shrinkage or lateral extension of the faultedhorizon. Given the limited resolution of the data, a smallbut possibly critical amount of lateral extension cannot bediscounted.

From the Coorikiana, throws on individual faults typicallydecrease rapidly downwards to zero, over a vertical intervalof c. 50–200 ms (c. 64–254 m). Interpretation of additionalhorizons adjacent to some faults has refined the profiles ofdecreasing throw from just below the Coorikiana down to theCadna-owie horizon (Fig. 5b). Resolved faults rarely extendbelow the Cadna-owie horizon. On depth-converted sections(e.g. Fig. 5b), down-dip throw gradients are in the range0.08–0.9 on fault surfaces with maximum throws of >25 m,

and are steepest just above the Cadna-owie horizon. Thesethrow gradients require footwall/hanging wall stratigraphicthickness ratios of up to c. 3.3 within mudstones immediatelyabove the Cadna-owie horizon, the origins of which areconsidered in a later section. The range of throw gradientsbelow the Coorikiana is similar to the range (0.2–0.9) observedby Higgs & McClay (1993) in a comparable intraformationalfault system with maximum fault throws of c. 60 m. Above theCoorikiana, up-dip throw gradients are generally lower andrange from 0.05 to 0.15.

Folds and horizon dipsCross-sections show a very characteristic association betweenfaults and folds. The cellular form of the fault-defined poly-gons means that conventional linear cross-sections (e.g. Fig. 3)provide a misleading impression of what, spatially, are highlyorganized structures. Instead, we use ‘dog-leg’ sections thattraverse between the centres of a succession of adjacentpolygons on the Cadna-owie horizon map and that formclosed loops (Fig. 6). Anticlinal folds occupy the mutualhanging walls of conjugate fault pairs with synclines occupyingmutual footwalls between adjacent pairs. The maximum am-plitudes of the folds is at the Coorikiana level and is up toc. 80 m, when fault offsets are discounted, similar to the faultmaximum throw values. The anticlines therefore correspond tothe boundaries of the polygonal cells defined by the fault traceson the Cadna-owie and the synclines to the centres of thepolygons. The wavelengths of the synclines are c. 5 timesgreater than those of the anticlines, so the three-dimensionalstructure is of isolated shallow dish-shaped depressionsseparated by a continuous network of relatively narrow ridges,with maximum relief of c. 80 m. For a typical polygonal cell ofdiameter 1.4 km, the anticlinal edges occupy c. 200 m and thecentral depression c. 1200 m. Most polygon diameters liewithin the range 800 m–2 km so the central depressions rangefrom c. 700 m to 1700 m in diameter (Fig. 7).

The folds begin abruptly at the Cadna-owie and maintainsimilar amplitudes, up to the Coorikiana from where theamplitudes decrease upwards. This vertical variation in foldamplitudes closely matches that of the vertical variation infault throws, both relatively and in absolute values (Fig. 6).Anticlinal hinges are most prominent at the level of theCoorikiana horizon even though folds of underlying horizonshave smaller or similar radii of curvature, because of thedownward decrease in the lengths of arc of the anticlines as thebounding faults converge downwards. The converse applies tothe synclines. Folds associated with intraformational faults inthe Tertiary mudstones of the southernmost North Sea show asimilar asymmetry, with anticlines locally developing intodiapiric structures (Henriet et al. 1991). The geometries offaults and folds are so complementary that the folds could aswell be described as reverse drag effects (see also Henriet et al.1991) but with rollover radii at least two orders of magnitudeless than would be expected on tectonic faults of similar size(Gibson et al. 1989).

The lower limit of folded layers corresponds to the top of alow velocity layer (Fig. 1c) and the overlying folded sequence isreferred to as the overburden sequence. The precise form of theanticlinal folds at the base of the overburden layer is uncertainbecause of both limited seismic resolution and complexitiesdue to the near convergence of conjugate-fault pairs. Theoriginal form of the anticlines at the base of the overburden


sequence is even more uncertain because the structure at thislevel has been modified by downfaulting of the overlyinggraben wedges, or mutual hangingwalls of conjugate-faultpairs. Collapse of graben wedges into the low density layer issupported by a number of observations: (i) the offset of theupper surface of the low density layer is c. 30% of themaximum throws/heaves in the overlying sequence (Fig. 5b),(ii) most of the maximum fault throws are less than twice thethickness of the low density layer, an amount which could beaccommodated by withdrawal of the low density layer beneaththe graben, with complementary uplift of immediately adjacentfootwalls into which displaced material was injected and (iii)horizon deformation and fault displacement patterns are con-sistent with thickened and upthrown footwalls and thinneddownthrown hanging walls (Fig. 7b). The asymmetry of foldsat shallower levels in the overburden sequence suggest thatthe original anticlinal ridges at the base of the overburdensequence were cuspate. Cuspate anticlinal folds of theboundary between gravitationally unstable layers are charac-

teristic of a range of physical models (e.g. Vendeville 1987;Davison et al. 1996a; Alsop 1996) and of gravitational over-turn of both small scale (Anketell & Dzulynski 1968, fig. 4; seeFig. 8a) and large-scale natural analogues (e.g. Remmelts1995). The small scale analogue (Anketell & Dzulynski 1968) isparticularly instructive as it shows the changes in form anddecrease in amplitude of folds overlying the cusps, withincreasing height in the overburden layer. The similarity of theLake Hope structures and those of both the experimental andnatural examples point to a similar mechanism in which theboundary between the low velocity layer and the overburdenplays a major role.

Simple restoration of fault displacements on ‘dog-leg’sections (Fig. 6c & e) shows horizon geometries which,although they never existed, illustrate the folding componentof the deformation in the form of 1–2 km wavelength folds. Inaddition to reversal of fault displacements, restoration alsorequired local horizon rotations adjacent to faults in order tomaintain stratigraphic thicknesses (Fig. 6b–e).

Fig. 6. (a) Map showing locations ofdog-leg seismic sections with turningpoints numbered/lettered. (b) Dog-legseismic section with horizoninterpretations. Vertical lines correspondto turning points. (c) Horizonsinterpretated in (b) with restored faultthrows and local bed rotations; parts ofhorizons requiring rotation to maintainstratigraphic thickness are indicated bythin lines above or below the horizon(N.B. fault heaves and long wavelengthfolds are not restored). (d) and (e), as (b)and (c), for the second dog-leg loopshown in (a). Mean distance betweenturning points is 1.4 km, which is takenas the typical polygon size. Depths areshown in ms two-way time.


Low-velocity layerThe downward termination and convergence of conjugatefaults at or near the Cadna-owie horizon, and the occurrence ofthe highest dip-parallel throw gradients just above the Cadna-owie horizon (Fig. 5b) suggest a strong control of faulting bysome attribute of the stratigraphic sequence at or near thishorizon. The Cadna-owie horizon underlies a thin shale layer(c. 35 m) of anomalously low velocity (2420 m s�1, see Fig. 1b& c), which has been associated with drilling problems, e.g.bridging, attributed to over-pressuring (Alexander & Sansome1996). Basin subsidence curves (Kantsler et al. 1986; Tupper &

Burckhardt 1990; P. Boult pers. comm. 1997) indicate rapidsedimentation of the overlying mudstone-dominated sequence,which could have provided an effective seal above what isnow the low velocity layer, either preventing or reducing itsde-watering and compaction. The absence of syn-sedimentaryfaults within the faulted sequence suggests that >500 m ofsediment loaded the low density layer prior to the onset offaulting, giving rise to the widespread over-pressuring at thisstratigraphic level within the Eromanga Basin.

Another significant velocity change is evident at the levelof the Coorikiana. The underlying sequence, down to the

Fig. 7. (a) Coorikiana horizon dip map.Low dips, dark colours, occur in theshallow basins that occupy the centres ofpolygonal cells and high dips, blue-red,mark the polygonal cell boundaries asdefined by the anticlinal ridges. Faultpolygons are also shown (yellow). Thearea shown in (b) is indicated. (b)Isopach map of the interval between theCadna-owie and Coorikiana horizons forthe area indicated in (a) showing thepolygonal pattern defined by thicknessvariations. The thicker areas(purple/dark blue) generally correspondto highs adjacent to faults at theboundaries of polygonal cells. Individualpolygons have, ideally, a synclinal,dish-shaped geometry, as exemplified bythe polygon in the centre of this area.The footwall highs also represent thelimbs of anticlines that are bounded bycollapsed graben wedges containinganticlinal ridges, marked by thinnedisopachs (green/red) off the axis of theridges (light blue/dark blue) e.g. theanticlines to the north, east and west ofthe central polygon. Thicknesses are inmetres and change by about 105 m(200 m to 305 m) over the mapped area.


Cadna-owie, has a average velocity of 2680 m s�1 (range2170–3150) whereas that in the overlying interval, up to theMackunda, is significantly lower at 2435 m s�1 (range 2450–3250), similar to that in the low velocity layer associated withCadna-owie which is 400 m deeper. The present velocity dis-tribution is not necessarily the same as that prevailing at thetime the polygonal fault system and its associated structureswere formed. The low velocity layer could well have beenthicker, for example. The preservation of high fluid pressure(Alexander & Sansome 1996), the assumed cause of the lowvelocity, in a layer with a strongly faulted overburden appearsproblematical, though the mudstone dominated nature of thesequence may be a controlling factor.

Origin of polygonal fault system

Distinguishing features of fault systemIn addition to taking account of the low velocity over-pressured layer immediately underlying the faulted interval, amodel for the origin of the fault system and its associatedstructures should account for the following features:

(a) the absence of either marked regional horizontal extensionor contraction;

(b) the near isotropic fabrics of horizon fault maps;(c) the polygonal fault-trace pattern on the Cadna-owie

horizon;(d) the conjugate nature of many faults which define the

polygons on the Cadna-owie and younger horizons;(e) the maximum throws on faults occurring at or just below

the Coorikiana horizon, with sharp throw decreases down-ward from the maxima and more gradual decreasesupwards;

(f) the abrupt downward termination of the fault arraywith faults rarely continuing below the base of the over-pressured layer;

(g) constant dips of individual faults;(h) the fold geometries and the geometrical and temporal

relationships of folds and faults;(i) the less spatially ordered fault-trace patterns and the

multiplicity of abutting intersections on the youngerhorizons;

(j) the similarity between the maximum fault throws, foldamplitudes and the present thickness of the over-pressuredlayer;

(k) the diameters of the fault-defined polygonal cells on theCadna-owie horizon.

ModelSimply restored 2D dog-leg sections (Fig. 6c & e), togetherwith the Cadna-owie fault map (Fig. 4c) and the Coorikianadip map (Fig. 7), indicate a ‘non-faulted’ horizon geometry

Fig. 8. (a) Cross-section of diapiric ridges at the interface betweenan overturning low density, low viscosity layer (saturated clay richsediment) and its denser overburden (silt/sand sediment). Note thefolds in the overburden which are characterised by short wavelengthanticlines and longer wavelength synclines, both decreasing inamplitude upwards. Redrawn from Anketell & Dzulynski (1968,fig. 4). (b) Model cross-section geometry of concentric foldsoverlying a gravitationally unstable overturning layer (black) theupper boundary of which shows short wavelength cuspate anticlinesand longer wavelength synclines. The different wavelengths of theoverlying anticlines and synclines give rise to differential thinningresulting in incompatible strains between the synclinal and anticlinaldomains, seen as conjugate discontinuities, or faults, (heavy lines)converging towards the originally cuspate anticlinal folds on theupper boundary of the unstable layer. The misfit, or faultdisplacement, increases upwards to the level at which the foldamplitudes begin to decrease and eventually die out. In plan viewthe cuspate folds define the polygonal cell boundaries at the base ofthe overburden layer. (c) Plan view of the mature stage of theisothermal Rayleigh–Taylor overturn of two Newtonian fluids in a20 cm�20 cm physical model. Lines represent anticlinal ridges onthe upper surface of the lower, i.e. low density, fluid. The ridges, orspokes, intersect at hubs which are loci of diapiric rise of the lowdensity fluid. Redrawn from Talbot et al. (1991, fig. 4b).


dominated by a polygonal pattern of linear ridges which on theCadna-owie are the foci for downward converging conjugatefaults which parallel the ridges. The polygonal ridge pattern isfurther illustrated by isopachs of the interval between theCadna-owie and Coorikiana horizons for a small area(Fig. 7b). Thickness changes show that the boundaries ofpolygons are marked by footwall highs defining the limbs ofanticlines that are, in turn, bounded by the collapsed grabenwedges. The later collapse of the mutual hanging walls ofconjugate fault pairs has therefore created graben along theoriginal anticlinal ridges that preserve the form of theanticlines but have destroyed the ridges.

The polygonal ridge pattern is reminiscent of structuresassociated with salt diapirism. Alternatively, a non-uniformradial extension of the faulted sequence could arise fromuniform extension of the underlying and overlying layers, by aprocess comparable with boudinage in a system characterizedby an oblate (k=0) strain ellipsoid. The faulted layer would bepinned to the bounding layer or layers only at the polygoncentres, about each of which extension would be radial. Theannular ring of greater fault density with its surroundingsub-radial fault pattern (Fig. 2) could be an indication of radialflow off the structural high, but it is difficult to account for thedominant polygonal pattern, or an associated oblate strainellipsoid, by radial flow from a single centre unless theextended sequence was underlain by a decollement. What wesee as a conclusive argument against a critical role for thestructural high is that the Lake Hope structures appear to betypical of those elsewhere in the Eromanga Basin (Cartwright& Lonergan 1997) and further afield (Cartwright & Lonergan1996) that are not associated with structural highs. Geometri-cally, the Lake Hope structural pattern is so similar to thoseproduced by Raleigh–Taylor instabilities that we find it diffi-cult to avoid the conclusion that density inversion was thedominant factor in its formation. However, there remains thedynamical problem of how density instability alone couldgenerate structures in materials with finite strengths ratherthan purely viscous properties. This problem is the same asthat posed by many salt structures and we concentrate belowon the more tractable geometric and kinematic aspects of thestructures.

Most work concerning salt-related structures has concen-trated on more advanced stages of diapir development thanthat represented by the Lake Hope structures which did notevolve further than the equivalent of the salt anticline stage, i.e.the overburden sequence is conformable with the underlyinglow density active layer (Jackson & Talbot 1994), assumingthat the comparison is valid. Polygonal patterns strikinglysimilar to the Lake Hope structures, with polygon diameters15–30 km, are expressed topographically on the sea floor in thetabular salt-minibasin province of the Gulf Coast region(Diegel et al. 1995). These polygonal basins contain up to 6 kmof sediment overlying a salt layer several kilometres thick.

Features shared by the Lake Hope structures and manyphysical diapiric models are local ductile extension and thin-ning of the overburden overlying anticlinal ridges of the activeboundary. At Lake Hope the active boundary is that betweenthe top of the low velocity layer and its overburden and thelocalized extension of the overburden is partly accommodatedby conjugate normal faults which bound graben overlying theridges. Although it is counterintuitive for a graben to overlie aridge, it has long been recognized that graben can result fromlocal extension and thinning in the anticlinal domains thatoverly linear ridges (e.g. Hills 1963, p. 266). Faults bounding

such graben must therefore be contemporaneous with theirassociated anticlinal folds and with the local extension of thefolded layer or layers.

Topographies of the partially restored horizons (Fig. 6c & e)are similar to that of an initially horizontal layer boundarydeformed by a Rayleigh–Taylor instability due to densityinversion (Talbot et al. 1991), with a spoke pattern of ridges onthe boundary between the contrasting materials (Talbot et al.1991), as shown in Fig. 8c. Three or more ridges converge tojoin at the hubs, ideally triple junctions, where elevations arehigher than on the ridges. The ridges are channels along whichmaterial flows radially towards the hubs. Flow within the ridgebounded polygonal cells is radially outward from the low atthe centre of each cell towards a spoke and then along thespoke towards the nearest hub. A crucial role for this type ofRayleigh–Taylor instability in the generation of a comparableintraformational type of fault system was proposed by Henrietet al. (1991), from whom we differ in identifying the principalunstable low-density layer as that which underlies the faultedand folded interval rather than the faulted interval itself;we cannot, however, discount the possibility that duringfaulting the sequence overlying the low density layer was alsoover-pressured to some degree.

The different wavelengths of the synclines and anticlines inthe overburden sequence resulted in strain incompatibilities, ormisfits, between anticlinal and synclinal domains, which areaccommodated by discontinuities manifested as faults (seeAppendix). This model of interdependence of folding andfaulting is consistent with the highly systematic relationshipsobserved and with the conclusion that faulting and foldingwere coeval.

Horizon restorations (Fig. 6c & e) are partial because onlythe fault displacements and some local horizon rotations havebeen restored, and not the broad wavelength (1–3 km) folds,hence the gaps representing fault heaves. Given that thesections are pinned at anticlinal culminations and synclinaldepressions, a dog-leg section joining the centres of severalpolygons will have the same length before and after deforma-tion unless there is also a component of regional extension.As the fold geometries require local bed-parallel stretching,line-lengths of horizons on cross-sections will not be con-served, but the distances between the centres of polygonal cellswill be unchanged. The part restorations of the dog-leg sec-tions show that folding and faulting were not simply sequen-tial, because the faults cannot be simply restored to providea realistic pre-faulting geometry. Furthermore, if the localextension is due to folding, when the folding ceased therewould be no local extension to be accommodated by faulting,i.e. no force to drive the faulting. The intimate geometricalrelationship between folds and faults requires a correspond-ingly close time relationship. Strain differences between anti-clinal and synclinal domains of concentric fold stacks (seeAppendix) require discontinuities between the two types ofdomain to maintain overall strain compatibility. If, as impliedby this interpretation, the faults represent geometrically neces-sary discontinuities rather than a change of failure mode fromfolding to faulting, then they are intrinsic to the foldingdeformation rather than the product of a separate event orstage of development. The fold and fault relationships shownin Fig. 8b illustrate some of the essential elements of thesystem, which include the systematic vertical variation inthrow on the faults. This variation is not consistent witheither substantial regional extension, or with gravity slidingor with lateral contraction due to dewatering, as none of


these processes would give rise to the observed displacementvariation with depth. Although Fig. 8b represents an internallyconsistent geometry of several aspects of the structures it bestrepresents the early stages of growth of the folds and faults ofthe study area. Specifically, Fig. 8b does not take account ofthe associated but later subsidence of the graben into the lowvelocity layer, probably excising this layer in the hanging wallsof the graben bounding faults and injecting the excisedmaterial into the footwalls of the faults where the increasedthickness contributed substantially to footwall uplift. Thetiming of this subsidence and its relationship with otherdisplacements will be described elsewhere.

Active, as opposed to passive and reactive, diapirismrequires an overburden thin enough to be broken but thickenough for discrete structures to form (Schultz-Ela et al. 1993).At Lake Hope the diapiric process was initiated but ceasedwhile still at an embryonic stage, possibly due to depletion ofthe low density layer, which is probably of zero thicknessbeneath the synclinal regions.

Current thinking on diapir initiation (Jackson 1995)emphasizes the solid properties of overburden sequences andthe consequent limited overburden thickness beneath whichgravitational overturn can initiate in the absence of regionalextension. Although physical models generating spokepatterns have employed viscous liquids for both the active andoverburden layers, the properties of the overburden, at least,are seen as neither wholly viscous nor wholly brittle but acomplex mixture of properties which vary with stress state andstrain rates (Poliakov et al. 1996). Nevertheless, a problemremains in how sufficient force could be generated by densityinversion alone to deform rocks with significant strength.

This model for the origin of the polygonal fault systemaccounts for many of its geometric and kinematic features(i–viii, above). Maximum fault throws would be expected to berelated in some way either to the original thickness of the basalincompetent low density layer (xii, above) or to the maximumrelief on this layer after deformation, which is estimated to beabout double its original thickness. The c. 35 m thickness ofthe low velocity layer, as measured in the single well for whichdata are available, is about a half of the maximum fault throw.This well is located on the flank of a ridge where the presentthickness of the low-velocity layer is likely to be close to itsoriginal thickness.

Given the number of relevant variables and the uncertaintiesin their values, no confidence could be placed in calculations ofthe conditions which would give rise to polygonal structures ofthe size range observed (xiii, above). Other natural systemsshow a wide range of polygon sizes or ridge spacings, i.e. <1 mto tens of kilometres. Salt related polygons range in diameterfrom 15–30 km in the Gulf Coast where the salt layer is severalkilometres thick (Diegel et al. 1995), 20–30 km in the NorthSea where a 1 km thick salt layer is overlain by 1 km ofoverburden (Davison et al. 1996b), and 7.5 km in the SouthernNorth Sea where polygons formed with a 1 km salt layeroverlain by <400 m overburden (Remmelts 1995). Polygons inthe Central North Sea which are unrelated to salt movementhave diameters in the range 0.5–1 km (Cartwright & Lonergan1996). Mud diapirs in the Gulf of Oman (Collier & White1990) are associated with probable linear mud ridges spaced at5–6 km intervals. In the Lake Hope region the typical ratio offold wavelength to the thickness of the low density layer is c. 35(wavelength�1400 m, low density layer thickness�40 m), butvaries from c. 20 to >60 according to polygon size. Compar-able ratios for flame structures and load casts appear to range

from <1 to c. 10, and the ratio in the experiments reported byTalbot et al. (1991) is c. 5. The range of these ratios for all sizesof gravity overturn structures is probably less than two ordersof magnitude, i.e. much less than the range of their absolutesizes.

A crucial question, to which we can provide no definitiveanswer, is whether gravitational instability alone initiated anddrove the formation of the structures or whether there was alsoa component of regional extensional.

DiscussionThe formation of the Lake Hope polygonal fault systemevidently did not lead to complete de-watering and compactionof the over-pressured layer because this layer remains over-pressured. It is possible that over-pressuring affected a thickerlayer prior to faulting, but as the faults converge downwardstowards the present top of the layer it is probable that this wasalso the top of the most over-pressured layer when gravi-tational overturn began. The mudstones between the over-pressured layer and the Coorikiana may have been de-wateredand hydraulically pressured prior to faulting. Previous authors(Henriet et al. 1991; Cartwright 1994) have identified high porepressure as an important element in the formation of this typeof fault system. A hydrofracturing mechanism cannot, how-ever, be primarily responsible for the formation of polygonaland related fault systems, as envisaged by those authors,because hydrofracturing gives rise only to Mode I fracturesand is not directly relevant to the formation of shear fractures,or faults. Overpressure can, however, promote faulting atlow values of shear stress by reducing the effective normalstresses.

Many outstanding questions remain regarding the geometryof the Lake Hope fault system. For example, a significantnumber of faults terminate downwards within the Coorikiana–Top Mackunda interval and above the Mackunda, but it isunclear whether these represent a separate tier of polygonalfaults, as reported from the North Sea (Cartwright 1994).Faults terminating downwards in the Coorikiana–TopMackunda interval do not do so at a well-defined level aswould be expected at the base of an upper tier.

The widespread occurrence of polygonal fault systems,noted by Cartwright (1994), reflects their characteristic settingin widely distributed and laterally extensive, post-rift, marine,mudstone sequences at shallow depths. These characteristicsprovide conditions favouring the development of overpres-sured low density layers which we regard as the crucialrequirement for the development of non-salt-related polygonalsystems. It is conceivable that the function of these layers is inproviding a decollement surface rather than, or in addition to,a density inversion. The role of density inversion and gravi-tational overturn was proposed by Henriet et al. (1991) but theassociated de-watering of the faulted sequence which theyenvisaged is not an essential part of the process. A densityinversion origin for North Sea polygonal fault systems issupported by the common occurrence of mud diapirism withinTertiary sequences which also contain polygonal fault systems(Fault Analysis Group, unpublished data).

In sequences with high proportions of mudstone, such asthose that typically host polygonal fault systems, faults wouldbe expected to be non-conductive to both across-fault andalong-fault fluid flow. It is not possible to determine from theseismic data whether or not the Lake Hope faults are sharp


discontinuities or are ductile shear zones. Although greatlyattenuated in shear zones, layers may still be continuous fromfootwall to hanging wall. Shear zones are thought to be morelikely than discontinuous faults at both Lake Hope and insimilar structures elsewhere (Henriet et al. 1991). A character-istic of the Tertiary sequence in the North Sea, in whichpolygonal fault systems are ubiquitous, is that hydrocarbonsoften are restricted to the topmost sand lenses in sequenceswith very low vertical matrix permeabilities (N. Telnaes pers.comm. 1997). Continuity of sand layers across the fault zonescould account for leakage of hydrocarbons along faults ofthese polygonal systems. Whether or not this explanation iscorrect, the hydraulic properties of intraformational faultsystems in mudstone dominated sequences is, as pointed out byHenriet et al. (1991), a question which concerns many aspectsof fluid migration. These aspects include the escape of hydro-carbons from source rocks, leakage of cap rocks, and a varietyof near-surface engineering problems.

The simplified model of graben formation which we describeis complicated by the later collapse of the graben into theunderlying ridges with excision of the low velocity layer andassociated uplift of adjacent footwalls. The kinematics of thissubsidence and its effects on fault geometries will be describedelsewhere.

Although the term polygonal fault system is widely used, itis recognized (Cartwright 1994) that faults systems apparentlyformed by the same mechanism are not polygonal if the hostsequence has a significant regional dip at the time of faultingbut tend to form parallel to the regional dip with the faultsdipping upslope. The term intraformational fault system, asused by Henriet et al. (1991), may prove to be more appropri-ate as increased attention is paid to these common and verywidely distributed fault systems.

ConclusionsThe fault system terminates downwards at the approximatelevel of the Cadna-owie horizon, the fault-trace map of whichis truly polygonal. On the Cadna-owie the paired fault tracesdefine the boundaries of 0.8–2 km diameter, typically 1.4 km,polygonal cells the centres of which are shallow basins. Thefault traces are paired because this horizon is close to theconvergence level of conjugate faults dipping towards eachanother.

In the faulted mudstone dominated sequence at levels abovethe Cadna-owie the fault-trace pattern is spatially less orderedand high-angle fault intersections are common. In cross-section the mutual hanging walls of pairs of conjugate faultsare anticlines that overlie and parallel the polygonal cellboundaries while the centres of the cells are occupied by longerwavelength dish-shaped synclines. Faulting and folding werecoeval rather than sequential and the faults separate anticlinaland synclinal domains which have incompatible strains, due tothe differences in wavelengths of the anticlinal and synclinalconcentric folds. Given the resolution limits of the horizonmaps, minor lateral shrinkage or extension of the overburdenlayer can be neither demonstrated nor discounted, eitherlocally or regionally.

The synclines and associated conjugate faults overlie poly-gonal ridges that formed at the interface between the faultedsequence and an underlying 35 m thick high-velocity, low-density layer in which a high fluid pressure was maintained bythe top seal afforded by the faulted sequence. Burial of the

low-density layer by an overlying normally pressured sequencecreated a density inversion. The pattern of polygonalridges bounding dish-shaped depressions is interpreted as theproduct primarily of embryonic overturn of this gravitation-ally unstable system. This interpretation is satisfactorygeometrically and kinematically, but less so dynamically.

There are indications that, in similar mudstone sequencehosted polygonal fault systems elsewhere, faults have acted aschannels for vertical migration of hydrocarbons. Determi-nation of the hydraulic characteristics of this type of fault is ofpractical concern.

We thank Santos Ltd (Australia) for providing the 3D seismicdata on which this work was based. This work was funded by theDTI funded LINK Programme project ‘Faulted Reservoir Analy-sis System— Development of Reservoir Geological AnalysisTechniques’, the EU Hydrocarbon Reservoirs Programme project‘Secondary Migration of Petroleum through Caprock and Carriersequences’ (contract JOF3CT950014) and by the Fault AnalysisGroup. Thanks are due to our colleagues in the Fault AnalysisGroup for assistance at all stages of the work, to P. Boult forproviding and advising on well data, to C. Talbot for helpfulcomments on an early version of the manuscript and to B.Vendeville for a review which led to significant modification of ourconclusions.

Appendix: Incompatible strains in non-symmetricconcentric foldingThe overburden folds have a significant component of concentricgeometry, which is evident not only from their profiles in seismicsections but also by the non-vertical and opposed dipping boundariesbetween the anticlinal and synclinal domains. With similar foldgeometries these boundaries would be parallel and vertical.

Concentric folding of a sequence with fixed lateral boundariesrequires layer-parallel stretching and complementary thinning of thelayers, with no neutral surface. The extension of a layer is given by thedifference in length between an arc of a circle, representing the foldedstate, and the original length as represented by the horizontal chordjoining the two ends of the arc (Fig. A1). The thinning of a layer isapproximately inversely proportional to the stretch.

The stretch (arc length/chord length) varies with the angle sub-tended by the arc at the centre of the circle (Fig. A1). This angleincreases as the radius of curvature of the arc decreases. If the anglesubtended at the centre of the circle is 2�, the ratio of lengths ofarc/chord is given by

arc/chord=(�/360)�(2�/sin�)

The variation of layer thinning with change in the angle subtendedby the arc at the centre of the circle is shown in Fig. A2.

Fig. A1. The angle subtended at the centre of a circle by an arc ofthe circle, radius r, representing the folded state, is 2�. The chordjoining the two ends of the arc represents the pre-folding state.


At the base of the overburden the wavelengths of synclines andanticlines differ by a factor of c. 5 (see Fig. 6), i.e. the anticlines havemuch smaller radii of curvature than the synclines and are associatedwith correspondingly greater layer thinning. For the c. 400 m intervalfrom the Cadna-owie to the Coorikiana at which fault throw is amaximum, a difference in average thinning of 20% between anticlinaland synclinal domains would give rise to a vertical misfit of 80 m. Thisdifference would result from angles subtended by fold arcs at thecentres of the respective circles of approximately 40� and 70�, but thedata are insufficient for the real values to be determined. As the misfitbetween the anticlinal and synclinal domains is cumulative upwards,throws on the faults will increase upwards. Above the Coorikiana theradii of curvature of the synclines become smaller than those of theanticlines but both increase upwards, as the fold amplitudes decrease,so the misfit decreases upwards from the Coorikiana and faultsterminate upwards.

ReferencesA, E.M. & S, A. 1996. Lithostratigraphy and environments of

deposition. In: A, E.M. & H, J.E. (eds) The petroleumgeology of South Australia Volume 2, Eromanga Basin. South AustralianDepartment of Mines and Energy, Eastwood, South Australia, 49–86.

A, G.I. 1996. Physical modelling of fold and fracture geometries associatedwith salt diapirism. In: A, G.I., B, D.J. & D, I. (eds)Salt Tectonics. Geological Society, London, Special Publications, 100,227–241.

A, J.M. & D, S. 1968. Patterns of density controlled convo-lutions involving statistically homogeneous and heterogeneous layers.Rocznik Polskiego Towarzystwa Geologicznego Annales de la SocieteGeologique de Pologne, 38, 402–409.

C, J.A. 1994. Episodic basin-wide hydrofracturing of overpressuredEarly Cenozoic mudrock sequences in the North Sea Basin. Marine andPetroleum Geology, 11, 587–607.

—— & L, L. 1996. Volumetric contraction during the compaction ofmudrocks: a mechanism for the development of regional scale polygonalfault systems. Basin Research, 8, 183–193.

—— & —— 1997. Seismic expression of layer-bound fault systems of theEromanga and North Sea Basins. Exploration Geophysics, 28, 323–331.

C, J.S. & W, R.S. 1990. Mud diapirism within Indus fan sediments:Murray Ridge, Gulf of Oman. Geophysical Journal International, 101,345–353.

D, I., A, I. & B, D. 1996b. Salt tectonics: some aspects ofdeformation mechanics. In: A, G.I., B, D.J. & D, I.(eds) Salt Tectonics. Geological Society, London, Special Publications, 100,1–10.

——, B, D., A, G.I. & A-A, M.H. 1996a. Deformationand sedimentation around active Miocene salt diapirs on the TihamaPlain, northwest Yemen. In: A, G.I., B, D.J. & D, I.(eds) Salt Tectonics. Geological Society, London, Special Publications, 100,23–39.

D, F.A., K, J.F., S, D.C., S, R.C. & T, P.R. 1995.Cenozoic structural evolution and tectono-stratigraphic framework of thenorthern Gulf Coast continental margin. In: J, M.P.A., R,D.G. & S, S. (eds) Salt tectonics: A global perspective. AmericanAssociation of Petroleum Geologists, Memoirs, 65, 109–151.

G, J.R., W, J.J. & W, J. 1989. Modelling of bed contours andcross-sections adjacent to planar normal faults. Journal of StructuralGeology, 11, 317–328.

H, J.-P., D B, M. & V, M. 1991. Early fracturing ofPalaeogene clays, southernmost North Sea: relevance to mechanisms ofprimary hydrocarbon migration. In: S, A.M. (ed.) Generation,accumulation and production of Europe’s hydrocarbons. European Associ-ation of Petroleum Geoscientists, Special Publications, 1, 217–227.

H, E.S. 1963. Elements of Structural Geology. Jarrold & Sons, Norwich.

H, W.C. & MC, K.R. 1993. Analogue sandbox modelling of Mioceneextensional faulting in the Outer Moray Firth. In: W, G.D. & D,A. (eds) Tectonics and Seismic Sequence Stratigraphy. Geological Society,London, Special Publications, 71, 141–162.

J, M.P.A. 1995. Retrospective salt tectonics. In: J, M.P.A.,R, D.G. & S, S. (eds) Salt tectonics: A global perspective.American Association of Petroleum Geologists, Memoirs, 65, 1–28.

—— & T, C.J. 1994. Advances in salt tectonics. In: H, P.L. (ed.)Contintental deformation. Pergamon Press, Oxford, 159–179.

K, A.J., C, A.C. & Z, M. 1986. Organic maturation in theEromanga Basin. In: G, D., M, P.S. & P, G.M. (eds)The geology and hydrocarbon potential of the Eromanga Basin. GeologicalSociety of Australia, 12, 305–322.

O, A.C. & G, N.M. 1995. Lake Hope 3D—A case study. Explor-ation Geophysics, 26, 383–394.

P, A.N.B., P, Y.Y., D, E.C. & T, C.J. 1996.Salt diapirism with simultaneous brittle faulting and viscous flow. In:A, G.I., B, D.J. & D, I. (eds) Salt tectonics. GeologicalSociety, London, Special Publications, 100, 291–302.

R, G. 1995. Fault-related salt tectonics in the southern North Sea, theNetherlands. In: J, M.P.A., R, D.G. & S, S. (eds)Salt tectonics: A global perspective. American Association of PetroleumGeologists Memoirs, 65, 261–272.

S-E, D.D., J, M.P.A. & V, B.C. 1993. Mechanics ofactive salt diapirism. Tectonophysics, 228, 275–312.

T, C.J., R, P., S, H., K, H. & J, M.P.A.1991. Diapiric spoke patterns. Tectonophysics, 188, 187–201.

T, N.P. & B, D.M. 1990. Use of the methylphenanthrene indexto characterise expulsion of Cooper and Eromanga Basin oils. A.P.E.A.Journal, 373–385.

V, B. 1987. Champs de failles et tectonique en extension. Modelisationexperimentale. Memoires et Documents du Centre Armoricain d’EtudeStructurale des Socles, 15.

V, M. 1992. An integrated 3-D approach to clay tectonic deformation.Doctoral thesis, University of Gent.

Received 9 June 1998; revised typescript accepted 6 May 1999.Scientific editing by Richard England.

Fig. A2. Layer thinning as a function of 2�. Estimated maximumvalues of 2� for anticlines and synclines are approximately 70� and40� respectively.
