The description of a forthcoming specialist conference on basin modeling includes the text:
BPSM (Basin and Petroleum System Modeling) has become an indispensable tool in frontier basins to identify risk, reduce uncertainty, and identify new potential areas. This technology has become more important over time as a result of increased understanding of processes and the rapid development of computing power. Both the hardware and the software are evolving to quantify more complex processes. See here
One could only assess the veracity of the first part of this statement by carrying out a survey of companies to see how many use basin modeling as part of their evaluation process and how many consider it "indispensable". What I think can be said is that, if basin models are "identifying risk and reducing uncertainty", then that isn't showing up in exploration success rates. Industry surveys show that frontier basin success rates have not changed much over the last 20 years, remaining less than 10% for a commercial discovery. I would also dispute whether the ability to "quantify more complex processes" has made a difference - increasing the complexity of a model does not mean increased predictive power. In fact, often the reverse is true because of a greater tendency to fit the "noise" in the system rather than the "signal" (see Nate Silver's excellent book "The Signal and the Noise: The art and science of prediction")
Some of us think that the lack of improvement in the predictive power of basin models is because they only partly address the two things most affecting the chance of a prospect receiving charge: The kitchen yield in relation to the volume of the migration pathway to the trap and the interaction of trap closure height and seal capacity (it doesn't matter whether we are speaking of fault seal or top seal). This is a topic that will be taken up elsewhere but of note are the presentations and papers of Richard Bishop (e.g. Bishop et al., 2015) and two other entries in this blog on traps being filled/not filled and traps leaking and spilling at the same time (in relation to the latter, see also the paper by Sales et al. 1997).
I would like to highlight one other aspect of this discussion: the controls on the occurrence of oil or gas in a trap and our ability (or lack thereof !) to predict it. The intrinsic link between trap fill and phase has already been discussed by Bishop (2015) and Sales et al. (1997). However, in Sales et al. (1997) excess supply of both oil and gas is assumed as precursor to the discussion. Bishop (2015) considers that such excess is implied by the observation that nearly all traps are filled to their spill or leak point. I would argue that not only the total amounts of gas and oil are important here but their relative amounts, i.e, the gas to liquids ratio of the incoming fluid. This, together with the pressure and temperature of the trap and the mutual miscibility of the gas and oil (dependent on their compositions), determines whether the fill of any individual trap is single or dual phase.
Firstly, let's look at the relative masses of oil and gas expelled from the standard Pepper and Corvi (1995) source rock types (cumulative):
Some of us think that the lack of improvement in the predictive power of basin models is because they only partly address the two things most affecting the chance of a prospect receiving charge: The kitchen yield in relation to the volume of the migration pathway to the trap and the interaction of trap closure height and seal capacity (it doesn't matter whether we are speaking of fault seal or top seal). This is a topic that will be taken up elsewhere but of note are the presentations and papers of Richard Bishop (e.g. Bishop et al., 2015) and two other entries in this blog on traps being filled/not filled and traps leaking and spilling at the same time (in relation to the latter, see also the paper by Sales et al. 1997).
I would like to highlight one other aspect of this discussion: the controls on the occurrence of oil or gas in a trap and our ability (or lack thereof !) to predict it. The intrinsic link between trap fill and phase has already been discussed by Bishop (2015) and Sales et al. (1997). However, in Sales et al. (1997) excess supply of both oil and gas is assumed as precursor to the discussion. Bishop (2015) considers that such excess is implied by the observation that nearly all traps are filled to their spill or leak point. I would argue that not only the total amounts of gas and oil are important here but their relative amounts, i.e, the gas to liquids ratio of the incoming fluid. This, together with the pressure and temperature of the trap and the mutual miscibility of the gas and oil (dependent on their compositions), determines whether the fill of any individual trap is single or dual phase.
Firstly, let's look at the relative masses of oil and gas expelled from the standard Pepper and Corvi (1995) source rock types (cumulative):
Although maturity is often thought of as the strongest control on the amounts of oil and gas expelled by a kitchen, source rock type is a stronger control within most of the maturity range. Furthermore, the amounts of oil and gas retained in the source rock vs. expelled has a major impact on expelled fluid gas to liquids ratio (GLR):
With the default P&C (1995) retained oil and gas amount settings (100 mg/g/TOC and 20 mg/g/TOC respectively) a marine clastic (B) kerogen expels a fluid with GLR of ~ 1100 scfs/bbl at 50% kerogen conversion and ~ 2200 scfs/bbls at full conversion. The GLR for fluvio-deltaic source rocks is very sensitive to the hydrogen index input chosen but for the standard kerogen is ~ 4400 scfs/bbl at 50% and 8800 scfs/bbl at full conversion (as a point of reference, the system-wide GLR for the Taranaki Basin of New Zealand is ~ 10,000 scfs/bbl).
Note that source rock type and expulsion/retention settings are INPUTS to basin models not OUTPUTS, so we can already see that basin modelling per se may not be good at predicting GLR
What happens when we put these fluids into a migration system (the culmination of which is our target trap) ? First lets look at how the mass/volume of oil vs. gas translates into phase and for this we need to use some standard bubble point and dew point curves: the ones in the diagram below are for UK North Sea oils and gases based on empirical observations (Glaso, 1980, England et al. 2002). There are many factors which affect the position and shape of these curves but that is a topic for another day and they are reasonable for our present purposes. The figure shows how the GLR at 50% kerogen conversion sits in relation to these curves for the P&C (1995) kerogen types:
The symbols here show the phase state of fluids in traps at different depths (assumes hydrostatic pressure) and the black bars highlight the intersection with the dew point/bubble point curves.
If we charge our system from a standard "B" type source (50%) conversion and all those fluids arrive in a trap, we can expect it to contain monophase oil if deeper than about 3100m and dual phase oil and gas if shallower than that. On the other hand, if our charge is from a standard "D/E" source all traps shallower than about 5100m would contain dual phase fluids. If we have a very gas prone type F (upper flood plain or paleozoic coals for example) we will hardly ever encounter anything other than gas. Similarly, if the source is a very oil prone lacustrine "C" or marine carbonate "A" (not shown in the figure) we will find mostly oil filled traps. Once again there are factors such as migration lag and in-trap alteration which will modify these conclusions in specific circumstances. However, their generality is borne out by the relative frequency of oil vs. gas discoveries in petroleum systems driven by one of the end-member source types. As examples one may cite the oil dominance in offshore Angola or the Bohai Basin of China (C type source) and the gas dominance on the outer Exmouth Plateau of Australia (F type source).
Now let's see how this plays out in a migration plumbing system. The diagram below shows a stylised series of three stacked reservoir/seal pairs with the top seal capacity varying both vertically and laterally for reservoirs 2 and 3 as shown. The actual values are not important here - it is the closure height to seal capacity ratio which matters - but the seal capacity does increase with depth as we might expect as the rocks compact. We are going to inject fluids with varying GLR into the base of the system (this whole exercise is done in Zetaware Trinity).
Here are the patterns of oil and gas obtained with varying input GLRs (nb: input GLR varies from chart to chart but is held constant during the migration fill process):
We can note several things from this:
1. At low input GLR gas does not displace oil up-dip: it can't do so if the system remains single phase
2. At very high input GLR we do not drop out an oil rim at any realistic depth. However for gas condensates with CGR of about 50 bbls/MMscf or higher oil rims do begin to drop out and may even lead to oil filled traps (the oil found here would be saturated with gas). Commercial oil pools can be (and are) found in dew point systems - although they may also sometimes be present as "nuisance" oil rims to commercial gas pools. The distribution of oil and gas in traps can be complex in all but the most oil or gas dominated systems
4. Discovering an oil or gas pool or even several does not necessarily define the system as "oil prone" or "gas prone". Compare the patterns of oil and gas occurrence for the 3000 scfs/bbl and 50 bbls/MMscf (= 20,000 scfs/bbl) input cases in the figure. This has not stopped some frontier basins with one or two oil or gas discoveries being labelled as "gassy" or "oily". In reality, a close look at the fluid properties and geochemistry is needed to make this call.
Next let's see what happens when we have a more realistic charge scenario, with the input GLR increasing as maturity of the source increases:
There is naturally more gas in the migration pathway as the source matures. However, notice that even at maturities above 1.3% Ro (the conventional "top gas window") it is possible to find oil. We might, for example, drill the middle trap, find that it contains gas or oil+gas and then deepen the well to find oil. Again, the decision about what to do should hinge on what the fluid property and geochemistry data for the first discovered fluid tell us about the petroleum system. The highest proportion of oil containing traps occur when the source is low mature but this also means fewer trap overall have received charge. If only oil is commercial in our area of interest, we trade off reduced phase risk against an increased risk of finding nothing at all.
This raises the question of charge sufficiency: Bishop (2015) observes that charge is not the limiting factor for trap fill even in systems apparently charged by lean source rocks. I suspect that the source rock quality and yield has been underestimated in many of these "lean" source rock cases because the true source - often deep in the kitchen - has never been drilled. This would explain why some of the data of Sluijk and Nederlof (1984) represent instances where more hydrocarbons were found in traps than were generated in the corresponding kitchen.
Studies such as those of Sluijk and Nederlof (1984), Biteau et al. (2010) and others cited by Bishop (2015) suggest that the supply of HCs to a trap may commonly be 1 - 2 orders of magnitude higher than the amount needed to fill it. However, we cannot conclude from this that charge sufficiency for an individual trap is never a problem: The next figure shows the distribution of oil and gas in our artifical migration pathway for scenarios in which the kitchen expels 28, 57 and 115 mmbboe/km2. For the purposes of this example we assume no migration losses other than those required to fill each trap in the pathway. In reality, some hydrocarbons will also be lost in reaching the critical saturation threshold in the rocks around the source interval itself and in sub-seismic waste zones
In the low yield case many traps, including those we would be most likely to drill, never receive charge. Although basins without sufficient charge may be rare, for every basin there must be a point at which hydrocarbons run out - equivalent to the maximum migration "range". It might be further from the kitchen than we expect but it must exist. This should be thought of not as a sharp line (even though it is sometimes drawn that way on play chance maps) but rather as a zone of increased probability of drilling a dry hole (nb: this means drilling into an empty trap, not a partly filled trap, since this is statistically unlikely - see http://petroleumsystem.
It is also interesting to consider the impact of phase separation and where it occurs along the pathway: If vertical migration happens early in the sequence phase separation also occurs earlier and the volumetric expansion of gas with reducing pressure means that the same mass of hydrocarbons equates to a much larger volume. This in turn means a greater lateral migration range compared to situations in which most migration happens in deeper carrier beds.
Thus, source rock UEP can be thought of as a kind of "master variable" which controls not only the chance of finding a hydrocarbon filled trap but also - for mixed oil and gas systems particularly - the phase of hydrocarbons found in that trap. Furthermore, and again as discussed in other posts in this blog, UEP is a major control on charge timing. Hence, we see that many things we traditionally expect a basin model to tell us - the chance of a trap receiving charge, the timing of charge relative to trap formation, the phase state of the trapped HCs - are highly dependent on the inputs we choose for the source rock.
We can see also that the pattern of migration depends on top and fault seal capacity of each intermediate trap along the migration pathway: whether it leaks at the crest, leaks through fault juxtaposition or through a non-sealing fault plane. Have we any realistic chance of estimating this for a whole, three dimensional migration pathway (four dimensional if you also expect fault seal capacity to change over time )? In his recent paper, Bishop (2015) discusses the inherent difficulty in determining whether a single trap is fill to the leak point, whether this is set by top or fault seal or by stratigraphic pinchout. I would add to this the observation that there have been many cases where the extent of compartmentalisation of discovered fields has been badly misread, even after extensive appraisal drilling. If we have trouble working out the plumbing of discovered and multiply drilled fields, what chance have we got of doing it for a whole migration pathway, especially since much of it will have, at best, coverage by 2D seismic ?
We can, I think, deal with this issue in several ways. Firstly, we can run multiple scenarios sampling the input space probabalistically or deterministically (or a combination as suggested by Bishop 2015). Secondly we can use our knowledge of compartmentalisation of discovered fields: There are several extant schemes or algorithms relating reservoir continuity to geological characteristics such as structural type, depositional environment, fault throw vs. net to gross, propensity for shale gauge etc. Nature is fractal so the same logic should apply to migration pathways: perhaps we can use such schemes to assign at least a relative efficiency to a migration pathway. Demaison and Huizinga (1994) referred to this with their low and high "impedance" systems but we can probably address the issue in a more detailed manner now, especially when we have 3D seismic attributes over some or all of the pathway. I do not believe we can do it deterministically in basin models because we cannot provide such models with inputs of sufficient detail to define specific migration pathways.
Finally, it must be said that the migration "cartoons" used in this post to illustrate concepts take no account of the lack of mixing in many hydrocarbon pools. Calculations of in-reservoir mixing times (see Smalley et al. 2004 for example) suggest that they are often longer than the typical filling time. This was supported by the observations of Stainforth (2004) who argued that compositional grading of petroleum pools is the norm rather than the exception. My own experience includes fields which are clearly unmixed, as reported for the Forties Field (England, 1990) but also some that show remarkable homogeneity over large inter-well distances. The latter cases can arise when a trap has access to charge from multiple directions so that a natural "averaging" process occurs: different migration paths have different lengths and volumes. If we apply the same logic to the intermediate traps along a migration pathway it follows that the migration lag effect on fluid properties and phase - though a fundamental aspect of the migration process - may not always be significant in practice.
With this post I hope to have made the point that the prediction of phase at the trap level is (a) fundamentally linked to the overall charge risk and therefore subject to similar uncertainties (b) inherently difficult in any mixed oil and gas charged petroleum system. I do not think this is a reason for pessimism or for not attempting to assign a phase risk to our prospects. Rather, given that it is hard to enough to find hydrocarbons in the first place - witness the low success rates in frontier basins - we should not worry about hydrocarbon phase at the trap level. If only one phase is likely to be economic we need to explore in basins where a dominance of that phase is likely, e.g. those likely to host very oil prone or very gas prone source rocks. Migration scenario testing can then help us home in on areas with the best chance of traps filled with the desired phase.
Once we are in a play or basin however, any hydrocarbon discovery is valuable, regardless of the phase: Examination of the fluids will tell us if we are in a fundamentally oil prone, gas prone or mixed system and guide our decision about what to do next - drill up-dip, down-dip, farm down or exit the play. Petroleum geochemistry has a major role to play here as compositional and isotope signatures exist for source type, relative maturity of expulsion, evaporative fractionation and secondary alteration by in-reservoir cracking, biodegradation and water-washing. All of these affect the GLR of trapped fluids.
All comments/criticisms etc. are welcome,
Rgds,
AM
It is also interesting to consider the impact of phase separation and where it occurs along the pathway: If vertical migration happens early in the sequence phase separation also occurs earlier and the volumetric expansion of gas with reducing pressure means that the same mass of hydrocarbons equates to a much larger volume. This in turn means a greater lateral migration range compared to situations in which most migration happens in deeper carrier beds.
Thus, source rock UEP can be thought of as a kind of "master variable" which controls not only the chance of finding a hydrocarbon filled trap but also - for mixed oil and gas systems particularly - the phase of hydrocarbons found in that trap. Furthermore, and again as discussed in other posts in this blog, UEP is a major control on charge timing. Hence, we see that many things we traditionally expect a basin model to tell us - the chance of a trap receiving charge, the timing of charge relative to trap formation, the phase state of the trapped HCs - are highly dependent on the inputs we choose for the source rock.
We can see also that the pattern of migration depends on top and fault seal capacity of each intermediate trap along the migration pathway: whether it leaks at the crest, leaks through fault juxtaposition or through a non-sealing fault plane. Have we any realistic chance of estimating this for a whole, three dimensional migration pathway (four dimensional if you also expect fault seal capacity to change over time )? In his recent paper, Bishop (2015) discusses the inherent difficulty in determining whether a single trap is fill to the leak point, whether this is set by top or fault seal or by stratigraphic pinchout. I would add to this the observation that there have been many cases where the extent of compartmentalisation of discovered fields has been badly misread, even after extensive appraisal drilling. If we have trouble working out the plumbing of discovered and multiply drilled fields, what chance have we got of doing it for a whole migration pathway, especially since much of it will have, at best, coverage by 2D seismic ?
We can, I think, deal with this issue in several ways. Firstly, we can run multiple scenarios sampling the input space probabalistically or deterministically (or a combination as suggested by Bishop 2015). Secondly we can use our knowledge of compartmentalisation of discovered fields: There are several extant schemes or algorithms relating reservoir continuity to geological characteristics such as structural type, depositional environment, fault throw vs. net to gross, propensity for shale gauge etc. Nature is fractal so the same logic should apply to migration pathways: perhaps we can use such schemes to assign at least a relative efficiency to a migration pathway. Demaison and Huizinga (1994) referred to this with their low and high "impedance" systems but we can probably address the issue in a more detailed manner now, especially when we have 3D seismic attributes over some or all of the pathway. I do not believe we can do it deterministically in basin models because we cannot provide such models with inputs of sufficient detail to define specific migration pathways.
Finally, it must be said that the migration "cartoons" used in this post to illustrate concepts take no account of the lack of mixing in many hydrocarbon pools. Calculations of in-reservoir mixing times (see Smalley et al. 2004 for example) suggest that they are often longer than the typical filling time. This was supported by the observations of Stainforth (2004) who argued that compositional grading of petroleum pools is the norm rather than the exception. My own experience includes fields which are clearly unmixed, as reported for the Forties Field (England, 1990) but also some that show remarkable homogeneity over large inter-well distances. The latter cases can arise when a trap has access to charge from multiple directions so that a natural "averaging" process occurs: different migration paths have different lengths and volumes. If we apply the same logic to the intermediate traps along a migration pathway it follows that the migration lag effect on fluid properties and phase - though a fundamental aspect of the migration process - may not always be significant in practice.
With this post I hope to have made the point that the prediction of phase at the trap level is (a) fundamentally linked to the overall charge risk and therefore subject to similar uncertainties (b) inherently difficult in any mixed oil and gas charged petroleum system. I do not think this is a reason for pessimism or for not attempting to assign a phase risk to our prospects. Rather, given that it is hard to enough to find hydrocarbons in the first place - witness the low success rates in frontier basins - we should not worry about hydrocarbon phase at the trap level. If only one phase is likely to be economic we need to explore in basins where a dominance of that phase is likely, e.g. those likely to host very oil prone or very gas prone source rocks. Migration scenario testing can then help us home in on areas with the best chance of traps filled with the desired phase.
Once we are in a play or basin however, any hydrocarbon discovery is valuable, regardless of the phase: Examination of the fluids will tell us if we are in a fundamentally oil prone, gas prone or mixed system and guide our decision about what to do next - drill up-dip, down-dip, farm down or exit the play. Petroleum geochemistry has a major role to play here as compositional and isotope signatures exist for source type, relative maturity of expulsion, evaporative fractionation and secondary alteration by in-reservoir cracking, biodegradation and water-washing. All of these affect the GLR of trapped fluids.
All comments/criticisms etc. are welcome,
Rgds,
AM
References:
Bishop R.S. (2015). Implications of source overcharge for prospect assessment. Interpretation, 3, 93-107, AAPG
Biteau et al. (2010). The why and wherefores of the SPI-PSY method for calculating the world hydrocarbon yet-to-find figures. EAGE First Break, 28, 53-64
Demaison G. and Huizinga B. (1991). Genetic classification of petroleum systems. AAPG. Bull., 75, 1626-1643
England W.A. (1990). The organic geochemistry of petroleum reservoirs. Org. Geochem., 16, 415-425
England W.A. (2002) Empirical correlations to predict gas/gas-condensate phase behaviour in sedimentary basins. Org. Geochem., 33, 665-673
Glaso O. (1980) Generalised pressure-volume-temperature correlations. SPE 8016, 785-795
Pepper A.S. and Corvi P.J. (1995) Simple kinetic models of petroleum formation: Part 1: oil and gas generation from kerogen. Marine and Petroleum Geology, 12, 291-319 (see also part II and III of this series of papers)
Sales J.K. (1997) Seal strength vs. trap closure - a fundamental control on the distribution of oil and gas. In: Seals, traps and the petroleum system, AAPG memoir 67, 57-83
Sluijk D. and Nederlof M.H. (1984). Worldwide geological experienceas as as systematic basis for prospect appraisal. In: Demaison G and Murris R.J. eds. Petroleum geochemistry and basin evaluation. AAPG Memoir, 35, 15-26.
Smalley et al. (2004). Rates of reservoir fluid mixing: implications for interpretation of fluid data. In: Cubitt J.M., England W.A. and Larter S. (eds.) Understanding petroleum reservoirs: towards an integrated reservoir engineering and geochemical approach. Geol. Soc. Lon. Spec. Pub. 237, 99-113
Stainforth J.G. (2004). New insights into reservoir filling and mixing processes. In: Cubitt J.M., England W.A. and Larter S. (eds.) Understanding petroleum reservoirs: towards an integrated reservoir engineering and geochemical approach. Geol. Soc. Lon. Spec. Pub. 237, 115-132
See also the blog posts:
http://petroleumsystem.
Andrew, flow rates in porous media are mainly a function of permeability and viscosity, both of which vary by several orders of magnitude, so mixing time scales should naturally vary by several orders of magnitude. In addition, we are always underestimating the degree of reservoir compartmentalization. We cannot see faults with less than 10 meter throws on seismic. We look at a well log, or a core section, what do we see in terms of permeability variation? Perhaps these can partially explain the variation in degree of mixing?
ReplyDeleteYes, using the analytical approximations in Smalley et al. (2004) the rate of density driven fluid pool homogenisation is a linear function of the initial density difference and permeability and an inverse function of viscosity. So we are indeed going to see large variations in the time required for GOR of an oil column reach the gravity stable condition. This does not explain all the variations between fields I referred to however. We have done these calculations using the analytical equations and also numerically modelled mixing times using the Permedia simulator. Within reasonable ranges of values for perm and viscosity we still see widely varying degrees of mixing among fields and I think this must relate to different initial states (i.e. fluid composition being "averaged" on the way into the trap, not just within the trap and post-emplacement.
ReplyDeleteJust to add to that comment: Heterogeneity in permeability - especially if laterally continuous thereby creating tortuousity in particle movement pathways - has a huge impact on mixing times, much more than the absolute perm values. For example, in the Vincent Field the spread of solution gas through a heavy oil column is increased by 3 orders of magnitude if one puts thin silty stringers at regular intervals among the otherwise multi-Darcy sand tank. These stringers are thin and only need to be one order of magnitude lower perm than the matrix to have this impact.
ReplyDeleteSorry, I meant "the time required for spread of solution as through heavy oil is increased by 3 orders of magnitude.."
ReplyDeleteI suspect different people have different perceptions of what is reasonable range of permeability and diffusivity variations. Vertically I could imagine centimeter, or millimeter scale barriers (note this is well below log resolution) down to 1 micro Darcy in most reservoirs. We just don't know the input for any of these models.
ReplyDeleteIf viscosity has a range of 3 orders of magnitude, and permeability has a range of 3 orders of magnitude, the range of rates are 6 orders of magnitude. So it could take one million years, or one billion years.
As for flow simulations, model builders like to use average properties. That is fine for volume/storage calculations, but not for flow rates, which is limited by the end member values. So vertically the flow rate is limited by the least permeable barrier.
I suspect different people have different perceptions of what is reasonable range of permeability and diffusivity variations. Vertically I could imagine centimeter, or millimeter scale barriers (note this is well below log resolution) down to 1 micro Darcy in most reservoirs. We just don't know the input for any of these models.
ReplyDeleteIf viscosity has a range of 3 orders of magnitude, and permeability has a range of 3 orders of magnitude, the range of rates are 6 orders of magnitude. So it could take one million years, or one billion years.
As for flow simulations, model builders like to use average properties. That is fine for volume/storage calculations, but not for flow rates, which is limited by the end member values. So vertically the flow rate is limited by the least permeable barrier.
No doubt that is true, and that perception will be influenced by both experience and (as you have noted) by the scale of the observation. In the Vincent Field mixing model the lower perm (by only one order of mag) intervals were modelled as only 10 cm thick. It is almost independent of thickness as modelled as it is lateral continuity that is important, but in reality a thinner unit means more chance there are holes in it (less continuity).
ReplyDeleteNot to be pedantic but 6 orders of magnitude is the difference between a million and a TRILLION ! (unless you were referring to the English "Billion").
Cooking Oil Distributors
ReplyDeleteCastor Seed Oil Distributors