Core Analysis — what the rock will tell you, and what it won't
Core is the only direct measurement of the reservoir rock anyone ever gets. Everything else — logs, seismic, well tests — is inference calibrated against it. That authority is deserved, and it is also the problem: a core measurement carries the weight of ground truth even when the plug it came from was never representative, or when the laboratory changed the rock before measuring it. This guide is about both halves — the methods, and what each one quietly assumes.
Routine core analysis (RCAL) measures the static properties: porosity, permeability, grain density, fluid saturations. Special core analysis (SCAL) measures how two or three fluids share the pore space: capillary pressure, relative permeability, wettability, electrical properties. RCAL populates the static model. SCAL drives everything dynamic — sweep, recovery factor, saturation from logs, the shape of the transition zone.
The bias that runs through everything[1][2]
Before any method, one structural fact. Core data is not a random sample of the reservoir, and the departure from randomness is systematic rather than accidental.
- You core the reservoir, not the rock. Coring points are chosen for the interval expected to produce. Shales, tight streaks and the flanks are cored far less often, so the population is enriched in good rock before anything is measured.
- Recovery is selective. Friable and unconsolidated intervals are the ones that wash out and are lost in the barrel. What arrives at surface is the more competent fraction of what was cut.
- Plug selection compounds it. A plug must survive drilling out of the slab. Rubble and highly fractured intervals do not yield plugs, so they are absent from the poro-perm cloud.
- Poor plugs fail in the laboratory. Samples that disintegrate during cleaning or fail during a two-week SCAL test are dropped from the dataset, usually without a note.
All four filters point the same way. The consequence is not that core data is wrong, but that it describes the better part of the reservoir, and the arithmetic mean of a plug dataset is therefore an optimistic estimate of the interval. This is the same problem as PVT sample representativity, in a different medium: the measurement is excellent, and the question is what it is a measurement of. Reconciling core against log-derived properties over the full interval — including the parts nobody plugged — is what turns a plug dataset into a reservoir description.
From the bit to the laboratory
By the time a plug reaches a cell, several things have already happened to it, and each one limits what can still be measured.
This is the single highest-value planning decision in the whole sequence, and it has to be made months before any laboratory sees the rock. Restored-state SCAL needs preserved plugs and a crude oil sample from the same reservoir for ageing. Saturation measurement needs a sponge or pressure barrel and a tracer in the mud. None of these can be arranged retrospectively. A core cut without a SCAL plan will support RCAL and a wettability measurement that reports what the solvent did.
Routine core analysis[1][2]
Four measurements, and the recognised procedures for all of them are set out in API RP 40, which remains the reference the laboratories work to even though API formally withdrew it. McPhee, Reed and Zubizarreta was written in part to fill that gap, and covers SCAL and geomechanics in the detail RP 40 never reached.[1][2]
| Property | Method | What it assumes |
|---|---|---|
| Porosity | Helium expansion (Boyle's law) for grain volume, calliper or mercury immersion for bulk volume | Helium reaches all connected pores and adsorbs negligibly. Isolated porosity is not measured — a real gap in vuggy carbonates |
| Permeability | Steady-state or pulse-decay gas flow, corrected for slip | Darcy flow, no turbulence, and a slip correction that has actually been applied |
| Grain density | Grain volume from helium, dry weight | Sample fully cleaned and dried. A useful mineralogy cross-check, and one of the quickest ways to catch a mislabelled sample |
| Saturations | Dean–Stark distillation, or retort | That the fluids in the plug are the reservoir's. Usually false for conventional core — see above |
The Klinkenberg correction[3]
Gas does not obey the no-slip condition at a pore wall the way a liquid does. Molecules retain velocity at the surface, so gas flows more easily than Darcy's law predicts, and the measured gas permeability is always higher than the equivalent liquid permeability. The effect scales with the ratio of molecular mean free path to pore-throat size, so it grows as mean pressure falls and as the rock gets tighter.
where:
kg = measured gas permeability
k∞ = equivalent liquid (Klinkenberg-corrected) permeability
p̄ = mean pore pressure during the measurement
b = slip factor, characteristic of the gas and the pore structure
Plot kg against 1/p̄: the intercept at 1/p̄ = 0 is k∞
Ambient conditions are not reservoir conditions
Routine measurements are made at low confining stress. In the reservoir the rock carries net overburden, the frame compacts, and pore throats narrow. Both porosity and permeability therefore read high at ambient conditions.
The two do not shift by the same amount, and that asymmetry is the point. Porosity loss is modest, because it is a volume ratio. Permeability loss is much larger, because flow depends on the narrowest throats and those are the first to close. More importantly, the correction is not uniform across the dataset: poorly consolidated and low-permeability samples lose proportionally more than clean, well-cemented ones. So the correction steepens the poro–perm trend rather than shifting it, and an uncorrected cloud is most optimistic for the rock that is hardest to produce.
Ask for measurements at net confining stress representative of the reservoir, and for the stress value used. If only ambient data exists, ask for a stress-correction subset — a dozen plugs spanning the permeability range, measured at both conditions — and derive the correction from those rather than applying a literature factor. A single correction factor applied across a wide permeability range is the wrong shape, for the reason above.
Cleaning — the step that decides what SCAL can measure[4][5]
RCAL needs a clean, dry sample. Cleaning — solvent extraction, Soxhlet, Dean–Stark — strips hydrocarbons and the polar compounds adsorbed on the mineral surfaces. Those adsorbed compounds are what made the reservoir mixed-wet or oil-wet in the first place. Remove them and the rock reverts to strongly water-wet.
That matters because a strongly water-wet sample gives a different answer for almost everything SCAL measures: a different residual oil saturation, different relative permeability curves, different capillary pressure hysteresis, and a different saturation exponent. Three sample states are therefore distinguished, and any SCAL number is meaningless without knowing which one produced it.
| State | How it is obtained | What it is good for |
|---|---|---|
| Native (preserved) | Plug preserved at the wellsite, never cleaned, never dried | The closest thing to reservoir wettability. Fragile, and compromised by any invasive mud |
| Cleaned | Solvent-extracted and dried | RCAL. And SCAL only where a strongly water-wet answer is genuinely what you want |
| Restored | Cleaned, resaturated with brine to irreducible water, then aged in reservoir crude at reservoir temperature | The practical route to reservoir-representative SCAL. Requires a live or dead crude from the same reservoir, and weeks of ageing |
Restored state is a reconstruction, not the original, and its quality depends on the ageing protocol — crude, temperature and duration. Ageing periods of several weeks are common and short ageing is a known way to end up with something still closer to water-wet than the reservoir. A SCAL report should state the ageing conditions. If it does not, the results are of unknown wettability, which for relative permeability means of unknown value.
Capillary pressure[6][7]
Capillary pressure sets the saturation distribution above the free water level, the irreducible water saturation, and the entry pressure a seal must exceed. It is the bridge between a static model and a saturation-height function.
| Method | Fluids | Speed | Trade-off |
|---|---|---|---|
| Mercury injection (MICP) | Mercury / vacuum | Hours | Fast, high pressure, excellent pore-throat resolution. Destroys the sample and uses fluids nothing like the reservoir's |
| Porous plate | Air–brine or oil–brine | Weeks to months | Direct equilibrium at each step, and the reference method. Slow, and limited by the plate's entry pressure |
| Centrifuge | Air–brine or oil–brine | Days | Faster than porous plate, but measures an average saturation along a saturation gradient and needs inversion — the Hassler–Brunner treatment[8] or a numerical equivalent |
Whichever method is used, the laboratory fluid pair is not the reservoir pair, so every curve must be converted before it means anything about the field.
and the height above free water level:
h = Pc,res / ( Δρ g )
Indicative laboratory values:
mercury–vacuum σ ≈ 480 mN/m, θ ≈ 140°
air–brine σ ≈ 72 mN/m, θ ≈ 0°
oil–brine σ ≈ 30 mN/m, θ assumed
The conversion is where a large and silent error enters. Interfacial tension at reservoir conditions is not the laboratory value — it falls with pressure and temperature, and for a gas condensate near the dew point it collapses towards zero. The contact angle is almost never measured on the actual system and is usually assumed to be zero, which is the water-wet assumption again. Both assumptions push the converted curve the same way: they overstate capillary pressure.
Work the consequence through, because it is the opposite of what intuition suggests. An overstated Pc curve means every saturation appears to occur higher above the free water level, so the transition zone is stretched upward. At a fixed height, you therefore read a saturation further back along the drainage curve — a higher water saturation than the truth, and a correspondingly lower hydrocarbon volume. This is one of the few errors in the whole core-analysis chain that is conservative rather than optimistic, which is exactly why it is worth stating: the biases do not all point one way, and assuming they do is its own mistake.
Relative permeability[9][10]
Relative permeability determines fractional flow, and fractional flow determines when water arrives and how much oil is left behind. It is the most consequential measurement in SCAL and the most fragile.
Steady state versus unsteady state
| Steady state | Unsteady state (JBN) | |
|---|---|---|
| Procedure | Inject both phases at a fixed ratio, wait for pressure and production to stabilise, measure, then step to the next ratio | Inject one phase, record production and pressure drop against time, invert to relative permeability[9] |
| Duration | Weeks per sample | Days per sample |
| Gives | Direct kr at each stabilised saturation, across the full range | kr only over the saturation range swept after breakthrough |
| Main weakness | Cost, duration, and capillary end effect at low rates | Inherits every JBN assumption, and reports nothing between initial saturation and breakthrough |
The JBN inversion assumes one-dimensional, incompressible, viscous-dominated displacement with negligible capillary pressure. The last assumption is the one that fails, and it fails in a specific and recognisable way.
The capillary end effect
At the outlet face of a core the wetting phase must leave into an open space where capillary pressure is zero. The wetting phase therefore cannot exit until its saturation at the face has risen enough to satisfy that condition, and it accumulates there. In a water-wet core flooded with water, water banks against the outlet instead of producing. Less oil is swept from the body of the core than the injected volume implies, so oil that was never displaced is reported as residual — Sor comes out too high. The retained water also adds pressure drop that is capillary rather than viscous in origin, so the water relative permeability comes out too low. Both errors point toward a more pessimistic waterflood than the rock deserves.
Was the sample native, cleaned or restored, and if restored, what were the ageing conditions? Was a rate sensitivity run, and did the endpoints move? Were the fluids live or dead, and at what temperature? Was the interpretation an analytical inversion or a numerical history match including capillary pressure? A dataset that cannot answer these is not evidence about the reservoir — it is evidence about the laboratory.
Wettability — measured three ways, agreeing sometimes[5][11][12]
Wettability governs where each fluid sits in the pore, which sets residual saturation, the shape of the relative permeability curves, capillary pressure hysteresis and the saturation exponent. It is not a property that can be inferred from lithology, and it cannot be measured on a cleaned sample.
| Method | What is measured | Scale | Limitation |
|---|---|---|---|
| Amott / Amott–Harvey[11] | Ratio of spontaneous to total displacement, for both water and oil | −1 oil-wet to +1 water-wet | Insensitive near neutral wettability, where spontaneous imbibition of both fluids is small |
| USBM[12] | Logarithm of the ratio of areas under the two forced-displacement capillary pressure curves | Positive water-wet, negative oil-wet | Needs a full centrifuge capillary pressure pair; gives a single index, no information on fractional wetting |
| Contact angle | Angle on a polished mineral surface | Degrees | Measured on a flat crystal, not on reservoir rock with its roughness and mineral mixture. Indicative only |
Amott and USBM are complementary rather than redundant: Amott resolves the strongly wetted ends poorly near neutral, and USBM resolves neutral wettability better but says nothing about fractional wetting. Running both is normal practice, and disagreement between them is itself information — commonly a sign of mixed wettability, where different pore surfaces have different preferences. Anderson's six-part literature survey remains the standard entry point to the subject.[5]
Electrical properties — where core meets the log[13]
Archie's relations convert resistivity into saturation, and the three parameters in them come from core measurements on the same rock.
Resistivity index: I = Rt / Ro = Sw−n
so that Sw = I−1/n
m = cementation exponent, from F against φ on brine-saturated plugs
n = saturation exponent, from I against Sw during desaturation
The cementation exponent is robust: it is measured on fully brine-saturated plugs, and cleaning does not affect it. The saturation exponent is not, and this is the most consequential link between a laboratory decision and a booked volume.
Measuring n requires desaturating the plug, which requires a
wettability state. On a cleaned, strongly water-wet plug, brine remains connected in surface
films throughout desaturation, conduction paths persist, and n comes
out near 2. In a mixed-wet or oil-wet reservoir, oil occupies some grain surfaces, the brine
films break, and the true n is higher — values well above 2 are
documented in oil-wet systems.[14]
Since Sw = I−1/n, a larger exponent gives a larger water saturation for the same measured resistivity. Take a resistivity index of 10: with n = 2, Sw = 0.32; with n = 4, Sw = 0.56. Using a cleaned-core n of 2 where the reservoir truly behaves like 4 therefore understates water saturation and overstates hydrocarbon in place — by a wide margin, across every well the log evaluation touches. This is one of the mechanisms behind pay that logs and cores disagree about; see low-resistivity pay for the family of problems it belongs to.
Scaling from the plug to the reservoir[6]
A plug is a few cubic centimetres. A simulation cell is millions of times larger. Two devices carry laboratory measurements across that gap, and both are approximations that should be stated as such.
The Leverett J-function collapses capillary pressure curves from samples of differing permeability and porosity onto a single dimensionless curve, on the argument that pore-throat size scales as the square root of permeability over porosity.[6]
Curves from one rock type should collapse onto one J-curve.
Curves that will not collapse are telling you there is more than one rock type.
That failure mode is the useful part. A J-function that will not collapse is not a bad correlation; it is evidence that the samples belong to different pore systems, which is where rock typing begins. Forcing a single J-function across a carbonate with both intergranular and vuggy porosity produces a smooth curve that describes neither.
Relative permeability curves are normalised to endpoint saturations before averaging, so that curves from samples with different Swirr and Sor can be combined without smearing the endpoints. Averaging the raw curves instead is a common and damaging shortcut: it produces an average curve whose endpoints belong to no sample, and it softens the very features that control breakthrough timing.
QC checklist before core data enters a model
- Depth. Core depth shifted to log depth, with the shift documented and the method stated — core gamma against the log is the usual basis.
- Representativity. Cored interval compared against the full reservoir interval. What was not cored, and how the un-cored fraction is being characterised.
- Recovery. Percentage recovered per core, and which intervals were lost. Systematic loss of a facies is a bias, not a gap.
- Stress conditions. Ambient or net confining stress, and the stress used. A stress correction derived from this core, not from literature.
- Klinkenberg. k∞ reported rather than kg, and the mean pressure stated.
- Sample state for every SCAL test: native, cleaned or restored — and for restored, the ageing crude, temperature and duration.
- Capillary pressure conversion. Which σ and θ were used, whether the interfacial tension was at reservoir conditions, and whether the contact angle was measured or assumed.
- Relative permeability interpretation. Analytical inversion or numerical history match; whether capillary pressure was included; whether rate sensitivity was tested.
- Wettability measured, by which method, and whether Amott and USBM agree.
- Saturation exponent and the wettability state it was measured in. If n came from cleaned core in a mixed-wet reservoir, say so where the volumes are reported.
- Rock typing. Whether J-functions collapse within each assigned rock type. If they do not, the rock typing is not finished.
- Reconciliation. Core porosity against log porosity, core permeability against well-test permeability. Two independent measurements of the same rock that disagree are information about scale, not an error to be averaged away.
Core is expensive, unrepeatable and irreplaceable. A programme is cut once, and the decisions that determine what it can ever answer — sponge or conventional, preserved or not, which SCAL tests, on which state — are all made before the bit turns. That is the argument for spending an afternoon on the plan rather than a year explaining the results.
References
- American Petroleum Institute. Recommended Practices for Core Analysis, API RP 40, 2nd edition, February 1998. Formally withdrawn by API but still the procedural reference commercial laboratories work to; remains publicly downloadable.
- McPhee, C., Reed, J. & Zubizarreta, I. Core Analysis: A Best Practice Guide. Elsevier (2015). Written explicitly to provide the updated recommended practices for SCAL and geomechanics that RP 40 does not cover, with emphasis on data quality control and diagnostic plots.
- Klinkenberg, L.J. (1941). The Permeability of Porous Media to Liquids and Gases. API Drilling and Production Practice, 200–213. Origin of the gas slippage correction.
- Amyx, J.W., Bass, D.M. & Whiting, R.L. Petroleum Reservoir Engineering: Physical Properties. McGraw-Hill (1960). Classical treatment of core measurement and capillary behaviour.
- Anderson, W.G. (1986–1987). Wettability Literature Survey, Parts 1–6. Journal of Petroleum Technology. Part 2, Wettability Measurement, JPT 38(11), 1246–1262, SPE-13933-PA; Part 6, The Effects of Wettability on Waterflooding, JPT 39, 1605–1622, SPE-16471-PA.
- Leverett, M.C. (1941). Capillary Behavior in Porous Solids. Petroleum Transactions of the AIME, 142, 152–169. Source of the J-function.
- Purcell, W.R. (1949). Capillary Pressures — Their Measurement Using Mercury and the Calculation of Permeability Therefrom. Transactions of the AIME, 186, 39–48.
- Hassler, G.L. & Brunner, E. (1945). Measurement of Capillary Pressures in Small Core Samples. Transactions of the AIME, 160, 114–123. Basis of the centrifuge capillary-pressure inversion.
- Johnson, E.F., Bossler, D.P. & Naumann, V.O. (1959). Calculation of Relative Permeability from Displacement Experiments. Transactions of the AIME, 216, 370–372. The JBN unsteady-state method. Some reference lists give the page range as 370–376.
- Craig, F.F., Jr. The Reservoir Engineering Aspects of Waterflooding. SPE Monograph Volume 3, Henry L. Doherty Series, Society of Petroleum Engineers. Standard treatment of relative permeability behaviour and its wettability dependence.
- Amott, E. (1959). Observations Relating to the Wettability of Porous Rock. Transactions of the AIME, 216, 156–162, doi:10.2118/1167-G. Origin of the spontaneous-to-total displacement ratio index.
- Donaldson, E.C., Thomas, R.D. & Lorenz, P.B. (1969). Wettability Determination and Its Effect on Recovery Efficiency. SPE Journal, 9(1), 13–20, SPE-2338-PA, doi:10.2118/2338-PA. The USBM wettability index.
- Archie, G.E. (1942). The Electrical Resistivity Log as an Aid in Determining Some Reservoir Characteristics. Transactions of the AIME, 146, 54–62.
- Morrow, N.R. (1990). Wettability and Its Effect on Oil Recovery. Journal of Petroleum Technology, 42(12), 1476–1484. On the saturation exponent and recovery consequences of non-water-wet states.