Fault Seal: Of Means, Minima, Models, and Occam taking a shave

CONTENTS: Doomed! The analysis is doomed!; Fault seal relationship status: it’s complicated (or is it?); One of three; How much is that geomodel in the window (the one with the waggily tail)?; How simple is simple?; When is a seal a seal?; A little bit dippy; The max envelope? – what about the min envelope? Tickling the heterogeneous zones; Averages vs max and min; The best for exploration; Seal stratigraphy; Piecing the parabolae; Two basics right – Allan mapping; A torchlight to new ground, or to the right questions; Nature won’t always make it easy, but we don’t have to start hard.; Refs & useful further reading.; Glossary.

Doomed! The analysis is doomed!

Actually, I don't think the end is nigh quite yet. There are reasons to be cheerful, and fault seal is a constantly evolving topic. Like basin modelling, sometimes the number of variables involved are daunting, and at times it is tempting to throw the hands up in despair and say, it can never be done.  But can simplicity be distilled from the complexity, in a way that allows sensible predictions in an exploration context? And if we can’t always get the exact answers, can we do enough work to at least understand the key questions? Paetoro Consulting (that’s me) is actively working with Fault Seal Pty. Ltd. (Director: Titus Murray) and clients to see if that’s the case. What follows is my take -just my take - so far.

Fault seal relationship status: it’s complicated (or is it?)

Just about the only thing that’s agreed about fault seal is that it can be complicated. It would take a brave person to deny that, yet perhaps the right question to ask is not how complicated can it get, but rather – can it ever be simple?

There is lots of potential for a false complexity that arises from resolution problems in our data. Sometimes the issue is not real complexity, rather that imperfectly resolved data introduce irregularities of their own, and these are often used, unmodified, to try and fit simple models. That can’t work well. When we get a misfit to a simple solution, invoking additional complexities might not be the only solution. Perhaps the simplicity isn’t being resolved due to measurement, imaging, & interpretation errors. Checking for those, is a better first step than introducing complexity straight away.

There is much literature on the complexities than can affect fault seal. Cataclastic gouge. Diagenesis & hydrothermal cements. Stress orientation & dilatancy. Clay entrainment within fault zones. Near fault fracturing. Multi-strand distributed offset. Each of these can be important, sometimes. 

I’ve been fortunate over the years to meet many of the people involved in the fault-seal scene, and I have a lot of respect for them all. There has been a great wealth of ideas generated and progress made. The complexity of the issue however has been the common theme – which to my mind makes it all the more important to distil simplicity from it where we can.

One of three

One the best simple manifestations of fault seal is illustrated in Figure 1 from Yielding 2010. There are really only three basic scenarios of fault seal. In the first, juxtaposition puts a sealing lithology against a reservoir at the fault and seals that way. In the second, it is the fault rock itself which provides the seal to the reservoir, through any one of a number of mechanisms. In the third, a destructive process, the fault is reactivated, or fracture systems reopened, by new tectonic activity and associated stresses, allowing leakage.

Of these we can see intuitively that the first, juxtaposition seal, is the simplest. In this scenario the only things we need ask of the fault to understand the sealing potential are its offset and the stratigraphy on either side of it. These are things that are typically readily available from well data and seismic. The other two mechanisms though, immediately involve the fault rock itself, which is very rarely sampled directly, and which immediately involves all sorts of complexities in the analysis, many of which we have to infer. That makes them harder to get right. 

So, if juxtaposition is ever the dominant effect, it’s good news.  Lets be real, it’s almost inevitable that the fault rock is going to have important additional effects, maybe good ones too, but if we can do things to a good approximation from juxtaposition alone, then Hurrah! – that’s great news for our ability to predict things in exploration, since all we need in order to have a decent stab - is provided by offset wells and good (3D) seismic.

How much is that geomodel in the window (the one with the waggily tail)?

The ability to model some of these things can get ahead of itself. Tail wagging the dog and all that. There are complexities that it can be very important to consider – but the first job is surely always to see whether the simplest solutions can work first. Then if they don’t, that’s the time to think about which complexities to introduce. 

The problem with all those complexities, is that there is no shortage to choose from, and if they are present, resolving their combined effect uniquely is non-trivial. So, if simple solutions can work, it provides considerable advantage. Going beyond the simple, it becomes extraordinarily hard to prove that the observed effect is actually due solely to the complexity we might be modelling. Non-uniqueness of solution. We don’t know what we don’t know. We might be able to explain observations by tweaking one variable, but then we might also be able to explain it by tweaking another one, or two together, or three and so on ad infinitum.

One of the greatest advantages of simplicity, is that if it can work as a premise, calibrated in one or two spots, then it easily lends itself to prediction in unsampled places, that can be tested.

How simple is simple?

What precisely do we mean when we talk about simple in the context of fault seal? Fundamentally, as the first figure illustrates, it is about juxtaposition. Whether flow units in the stratigraphy meet each other across a fault, or the flow units meet sealing units. In the most common context that translates to sands vs shales, though there are many others. Note though that a flow unit is not defined by our usual petrophysical definitions of reservoir or pay – “thief” flow units only need to flow over geological time, not the production time scales that usually drives our petrophysical flagging.

Two main things drive “juxtaposition” – it is the fault geometry (displacement profile) along its strike, and the stratigraphy at the fault (not near the fault, at the fault). By stratigraphy, we essentially mean a simple flow/seal unit stratigraphy. These two things in combination totally drive any simple juxtaposition model. We can celebrate that these two things are largely resolved by seismic and wells, and so can be extrapolated in a simple model for exploration purposes in undrilled areas. We are not at this point presuming that a simple model like this is going to work, just that it’s the most logical thing to check out first before resorting to anything more complicated.  

When is a seal a seal?

A mass of literature exists on what constitutes and forms a seal – but what we are really looking for here is to get empirical. Not to let pre-suppositions get in the way. To do that we need a few good HC occurrences at offset wells and to be able to see where and why HC saturations stop in the stratigraphic column.

 It’s not about having some pre-conceived notion of which petrophysical flag is going to make our seal work – it’s about looking at as many of the mud-log shows, the pressure data points, the salinities, the fluid inclusions, the well log suites, the petrophysical interpretations, and so on - as we can. This gives us a picture of where the hydrocarbon saturations and different reservoir waters stop, and what the rocks look like when they do. Armed with the criteria which most seem to matter in our individual sub-basin, this can be used to differentiate our simple seal & flow unit stratigraphy.

A little bit dippy

Seismic is an image. For our purposes, we are interested in how it can image a fault. There are fundamentally three things we would like it to tell us at any given fault-offset horizon – the location of the hanging wall (in x, y and time), likewise the location of the footwall, and thirdly, the dip of the fault. Any two of these things would in an ideal image define the other. Seismic though (try to contain your shock), does not deliver a perfect image.

What we usually end up with, for a given horizon, is an image of the horizon that stops short of the fault, both in the hanging wall and the footwall. This can be due to the fault zone itself, structural complexities within it, or it can be due to velocity contrasts and effects in the overburden above it, producing “fault shadows”. Typically, whatever seismic software we are using, it will mean that if we just use where our horizon’s grid “stops and starts” to define our fault polygon, the fault will have a shallower dip than reality, affecting not just juxtapositions but also estimates of GRV & in place hydrocarbons.

The max envelope? – what about the min envelope?

A lot of great work has been done over past years on shale entrainment in fault zones. SGR (shale gouge ratio) studies are just one aspect of that. 

A relationship between the maximum pressure difference plausible across a fault and the calculated SGR, such as that in Figure 2, is usually stipulated. One thing to note though, is that there are thousands of points below this maximum “envelope” – they don’t all fall on the line. That’s to say presence of a high SGR is no guarantor of the presence of a seal. 

The models appear to constrain the best that might be plausible, but not the actuality at any given place. For any given SGR, there appears to be a maximum pressure difference, but there is a distribution. OK - so this might be because the wells and associated pressures are not always sampling the maximum pressure difference at the crest, and in analyses of such data there is usually an effort to single out those at the crests of faults which do. This is attempted in Bretan’s 2003 paper and re-illustrated in many later papers from those studying SGR. Even these crestal points still have a wide distribution though; they don’t all fall on a line. SGR analysis seems to be telling us something, but maybe it's more about what can sometimes happen, not what always happens.

Such analysis is also a case of remote sensing – using pressure and nearby wells to infer fault character. These facilitate calculation of an SGR and of a pressure difference, but there is no easy way (short of a new well) of actually sampling the fault zone to verify that this is indeed a cause and effect relationship, and that some other complexity is not responsible. There are other possibilities by which a fault rock can produce a pressure difference, many of which are related to fault offset too, so that larger pressure differences would again result on larger offset faults.  I wish we could send a drone through the subsurface to check it out, but alas…

Tickling the heterogeneous zones

While we may not be able to send a drone through the subsurface to check out a fault zone, the next best thing we have is surface outcrop analogues. These show us that shale entrainment into fault zones is a real process. However, it doesn’t take long looking at any one fault to conclude they are inherently complex and heterogeneous at outcrop scale. This is something Sosio de Rosa et al. (2018) and Murray et al. (2019) have investigated quantitatively.  While a lot of potentially sealing mechanisms happen along fault zones where (reservoir) flow units are juxtaposed, they rarely persist in an exactly predictable manner – at least to the point we can be sure such seal exists everywhere along the fault.  That is important. 

Averages vs max and min

SGR calculation algorithms are averages. Crudely speaking, the SGR is the average V-shale of the stratigraphic section to have slid past the footwall during that amount of offset. Nothing wrong with that per se, but it’s worth contemplating that it does implicitly carry the assumption that it is a process occurring homogeneously as a function of offset. 

 The observation that fault zones are frequently very heterogeneous on a metre scale, should therefore alert us to the fact that an approximation is in play – and one that may or may not always hold. Fault seal is driven explicitly by maxima and minima in the fault zone rock, not averages. It only takes one leak to bust the balloon. The average balloon skin integrity taken over the whole surface is worthy of note, and perhaps tells us what the balloon material might be capable of, but this average is not the primary control. It is the presence of those heterogeneities. The leak. The pin prick that makes the whole thing go bang. When that happens we don’t care what the average balloon skin thickness is.

I show a sample SGR calculation “Triangle diagram” in the figure above. It’s fairly straightforward to do in excel – a bit fiddly, but armed with the basic formula, a V-shale profile with depth, and some conditional formatting, it’s not hard to produce something like the above. It helps also to know that when viewed at a 10% zoom a big excel spreadsheet stops showing the text of a cell and just shows the colour fill. Like magic, with a bit of screen-dump jiggery-pokery – an SGR triangle falls out (once you’ve done it once, the excel template is more or less just a cut and paste job for the Vshale data next time, next well). These diagrams show the SGR value as a function of fault offset. At zero offset, the profile is just the V-shale depth profile driving the model, but at greater offsets, the averaging process kicks in. 

Diagrams like this illustrate how all the magnificent heterogeneity we get at left, from the well control, is smeared into one blurry average as offset continues. It is very obvious from this contrast how much of an averaging process is going on in SGR calculations. That’s not a criticism per se, but it does beg the question, from what we see of faults in outcrop, is this level of homogeneity falling out of the averaging process, what we really expect to see? 

I suspect what it is doing, is showing the best case possible, a perfect shale entrainment process. I also suspect that hardly ever exists. Sometimes it might get close, but how often, is a legitimate question. As with most things, the truth is likely to be somewhere between the worst-case scenario (i.e. juxtaposition driven, with no contributing SGR effect) and the best-case scenario, as given by SGR triangle diagrams.  Much of the effort at present is centred on discovering just where reality places us on that line between the best- and worst-case scenarios.

The best for exploration

Juxtaposition seal analysis lies at the other end of the scale to those methods quantifying fault rock qualities – it’s kind of the worst-case scenario, which may be another over-simplification. It assumes as a base case that the fault at flow-unit juxtapositions will leak.  There is no membrane seal. The beauty of this approach is, again, that it is simple. While the fault rock itself may be inherently heterogeneous, we don’t really care because we usually have a good definition from well and seismic of where large-scale sealing units are in a fault’s hanging wall, and it is the hanging wall lithology, not the fault rock, which provides the seal. That makes it eminently more predictable, if it can be shown to explain hydrocarbon occurrences on its own. That in turn, makes it a sensible first step to address if fault seal is considered important in an exploration context. 

Seal stratigraphy

Our HC-occurrence well-calibrated seal-stratigraphy is the simple building block for juxtaposition seal analysis. There is of course complexity in sealing stratigraphies too. There may be thin thief sands or silts - with permeability within shales - that are not seismically resolvable. These might be juxtaposed across a fault in a position where we would map a shale. However, if they are that thin, there is a corollary that they are not likely to be very laterally continuous. That means even if they did leak across the fault, they might just form a small hanging wall stratigraphic trap and not “bust” the footwall trap - rather they would extend it a short way. There may be exceptions to that rule, but more laterally extensive thin units would be unusual and if present, correlatable on well data. 

Likewise, for seals, although quite thin shales can locally provide very effective capillary seal, there are implications for lateral continuity of a thin sealing unit too. That means thin ones aren’t likely to be extensive and hence effective as a prospect-scale seal. That in turn means we can fall back to seismic resolution scales of 10’s of m in conjunction with our well data, to constrain the location of the most effective seals within a sub-basin. All good news for regional analysis of juxtaposition seal in an exploration context. 

Piecing the Parabolae

Fault displacement profiles are also themselves subject to complexities of course, but in general lots of earthquake displacements over time on a simple planar fault do average to something that approximates a parabolic profile.  We can use that to QC fault geometries where there is a “crinkliness” of the displacement profile resulting from imaging and interpretation error. There are additional complications where faults bifurcate or transfer displacement at relay ramps – but in general its not too difficult to address this by breaking a fault into sensible segments. 

Two basics right – Allan mapping

The aim of the game then, is to get these two basics right, the sealing stratigraphy, and the fault geometries. Armed with those as good as we can get them, we progress to the bread and butter of juxtaposition seal analysis – Allan maps of flow unit juxtapositions across faults. These fall out naturally from our analysis. These can be used to help define fault blocks with crests and spill points and bounding faults and predict likely HC accumulations. These can be compared with the observed accumulations. Programs like FaultRisk (Fault Seal Pty Ltd.) are designed to input and QC these two basics, and output the Allan maps, incorporating uncertainty with distributions rather than just single deterministic values – though of course these fall out of the analysis too.  

In case the concept is new, these Allan maps are simply a picture of the fault plane itself, but with the flow and sealing units from both sides of the fault projected on to it. In the example below, produced in FaultRisk software, a) the greys are areas of no flow unit, i.e. seal, b) dark purply blue is where footwall flow units (reservoirs or thief zones) have no juxtapositions with flow units across the fault, c) the palest blue is where the hanging wall flow units have no juxtapositions across the fault plane, and d) the medium blue is where there is juxtaposition of hanging wall and footwall flow units. In programs like FaultRisk it is also straightforward in passing to calculate the SGR value in these areas of juxtaposition, and if we want to, to display this with an additional colour code in the juxtaposed areas. In the figure shown, yellow represents juxtapositions where the SGR value is low and wouldn’t help seal much, while the red areas are where this might be more of a possibility.

A torchlight to new ground, or to the right questions.

If there is a match of such HC column models to observed accumulations, maybe not exactly in column height – but in location of the major columns, then we are empowered to extrapolate to undrilled areas.  There may be too many variables to get the hole-in-one first time, in terms of exact column height predictions, but if there is less of a match, it can throw a light on precisely where the mismatches are and what could influence them, and hence focus attention on matters to be further checked. 

As we have also calculated the SGR involved with the observed seal stratigraphy and fault geometries as we go, and seen what SGR style analysis would predict at the same time, these can also be cross checked against actual HC occurrences. This can provide a sense of how wide or close to the mark theses estimates are too. We can therefore start to get a sense of where nature is placing us, on that line between a worst-case scenario of juxtaposition only and a best-case scenario of homogenously generated shale entrainment. That of course assumes these are the only two options, whereas in truth many other factors are at play, and not just fault seal related. Charge issues are also important to consider. Juxtaposition though, the simplest option, is the first and best stepping stone across the river of understanding.

Nature won’t always make it easy, but we don’t have to start hard.

Chances are of course, that nature won’t let it happen quite that simply. Always there are new questions that arise as much as answers. That however is progress.  Even if the analysis does not yield slam-dunk answers immediately, there is a satisfying confidence that the simplest solution – Occam’s razor, has been tested, and can form a useful basis for investigation of any other complexities that might be relevant. Back to basics is a well-worn cliché, and there is never any guarantee that the basics alone will work. What is clear though is that getting the basics done right first is never wasted and forms a useful basis for whatever comes next. In contrast, getting complicated too soon carries the danger of false trails and befuddlement, and having to go back to square zero after a lot of work. 

I want to stress that I have greatly enjoyed watching the fault seal discussions evolve over the years. I stand in awe at the work many stalwarts have done on so many aspects of fault seal with such determination and perseverance. It has taken us forward immensely. The processes involved are now increasingly well understood – though always room exists for surprises and improvement, and of course many questions remain. 

I believe the challenge now though is to wade into all those beautiful 3D datasets we get these days and try and assess the relative importance of each of these processes.  To do that, requires getting the basic, simple models right first. That means understanding how far along the road we can get with just juxtaposition seal. It might take us almost all of the way, it might take us only part of it, but if we don’t get this understanding right first, any discussion of other fault seal complexities is built on shaky foundations.

Better to do the simple models first, and pocket stepping stone one for perpetuity. If we’re happy our data has resolved things to an acceptable resolution and we understand the juxtapositions, but still have some mismatches to explain, then it’s the time to delve into all those other weird and wonderful fault seal complexities.

Just my view…

Paetoro Consulting UK Ltd (London) gratefully works with SHS Geo and Fault Seal Pty Ltd. (Sydney), to utilise the latter’s “FaultRisk” software (& wealth of experience) - to probabilistically model juxtaposition-seal, following on from well-calibrated seal-stratigraphy designation and QC’ed fault geometry construction. Further explanations of the techniques involved are discussed in Murray et al. (2019). The views expressed in this article are mine, and not necessarily those of my associates. It's also fair to say little of what I say is original - it's a loose repackaging of lots of other people's good ideas.

Refs & useful further reading:

Allan, 1989: Model for Hydrocarbon migration and entrapment within faulted structures, AAPG Bull. V. 73 p803-811.

Bretan et al., 2003: Using calibrated shale gouge ratio to estimate column heights, v. 87, p 397-413.

James et al, 2004: Fault Seal analysis using a stochastic multi-fault approach, v. 88, p 885-904.

Knipe et al 1998: Faulting, fault sealing, and fluid flow in hydrocarbon reservoirs: an introduction, Geol. Soc. Lond. Spec Pub v. 14,7 p vii-xxi.

Murray et al, 2019: Validation and Analysis procedures for Juxtaposition and Membrane fault Seals in Oil and Gas Exploration, Geol. Soc. Lond. (in-press).

Pei et al, 2015: A review of fault sealing behaviour and its evaluation in siliciclastic rocks, Earth Science Reviews, v. 150 p 121-138.

Sosio de Rosa et al 2018: Along-strike thickness variations of a fault in poorly lithified sediments, Miri (Malaysia), J. Struc. Geol. V 116, p. 189-206.

Yielding et al, 2010: Fault Seal calibration: a brief review. Geol. Soc Lond. Spec. Pub. V. 347 p 243-255.

Glossary

V-shale: The percentage or ratio of rock volume considered to be shale at a given point, typicaly determined petrophysically (and hence with associated uncertainties), and expressed as a profile with depth in a well. V-clay also sometimes used, in a similar fashion.

SGR: One of a number of quantitative algorithms for looking at shale entrainment within a fault zone, as a function of fault offset and V-shale or V-clay. Typically used to estimate the maximum pressure differences that can be sustained across a fault. The algorithm inplicitly assumes a uniformly/homogeneously occuring entrainment process.