SPE 38736 Applying NMR Total and Effective Porosity To Formation Evaluation
SPE 38736 Applying NMR Total and Effective Porosity To Formation Evaluation
SPE 38736 Applying NMR Total and Effective Porosity To Formation Evaluation
"litho/porosity crossplots" being derived by semi-empirical Historically, relaxation components with relaxation times
and semi-deterministic methods are strongly lithology de- T2 > 3 ms have been considered a measure of effective poros-
pendent. Optimal circumstances (i.e., clean, shale free, forma- ity, i.e., the pore space not associated with clay minerals.6
tions saturated with fresh water, good quality logs, good Thus, in liquid-filled pores, the number of hydrogen atoms
knowledge of mineral composition) are prerequisites to deduce "seen" by MRIL maps directly into porosity, since the hydro-
accurate formation porosity values. gen density is very close to that under calibration conditions.2
The following example shows that under certain, never- The newly introduced C/TP technique allows MRIL to
theless common circumstances, the conventional approach can quantify hydrogen with nuclear relaxation times as fast as 0.5
yield an incorrect porosity value. The true formation porosity ms and as slow as several seconds.6 The NMR signal inte-
shall be calculated by a combination of neutron porosity and grated over the complete relaxation spectrum is designated as
gamma-gamma density log. The density log shows a bulk den- MSIG (named after the sum-symbol Sigma).
sity of ρb = 2.32 g/cm3. The log analyst assumes a matrix den- In gas-bearing (non-flushed) formations, MSIG undercalls
sity ρma = 2.65 g/cm3 and fluid density ρfl = ρm = 1.00 g/cm3. true porosity proportional to gas saturation and inversely pro-
In a clean, water saturated formation the neutron log would portional to gas hydrogen index. Additional processing and
read the correct porosity value ΦN = 0.2. Using equation A-3 interpretation techniques are required to restore the porosity
(refer to Appendix A), an averaged neutron-density porosity reading.1
The permeability of the Wilcox reservoir units is deter- This example clearly demonstrates that conventionally de-
mined primarily by grain size, varying from sands size to clay rived porosity can severely suffer from its inherent lithology
size (Wentworth scale), though other detrital minerals as well dependence. Not being affected by such limitations, the lithol-
as authigenic changes may also influence the formations flow ogy independent MRIL total porosity provides more reliable
qualities. values than the conventional logs.
Log based formation evaluation of the Wilcox depends on
correctly assessing the water content of these units, a particu- South Texas: Fig. 2 provides a detailed look at logs recorded
larly difficult task since they tend to have high water content in another well logged in the Wilcox formation of South
even when free of water production. The correct assessment Texas. This well is located in the Zapala Delta of the La Salle
of porosity, needed for formation factor, and the clay mineral Delta system where the deposition tends to be shelf edge or
content, needed to account for equivalent water resistivity ef- slope region facies.8 The mineralogy in this well is similar to
fects, are critical to this process. the previous Wilcox example with similar formation evalua-
In track 1 of Fig. 1 is the uranium adjusted gamma ray tion issues.
CGR, caliper CAL, and the bulk densities from log RHOB and Shown in track 1 are uranium adjusted gamma ray CGR,
core analysis CROB that uses the inversion of core porosity caliper CAL, bulk density RHOB together with core derived
and matrix density. Track 2 shows density porosity PDSS, bulk density values CROB from five drilled sidewall samples.
neutron porosity PNSS and a neutron-density crossplot based Track 2 presents density porosity PDSS, neutron porosity
porosity PND, and core porosity PCOR. Track 3 compares PNSS and the neutron-density crossplot based porosity PND;
MRIL total porosity MSIG, with neutron-density porosity while track 3 compares MRIL total porosity MSIG and MRIL
PND and core porosity PCOR. effective porosity MPHI, with neutron-density porosity PND.
Generally speaking, the conventionally derived porosity The caliper log CAL indicates a moderately rugose bore-
(e.g., PDSS, PNSS, and PND) exhibit higher values than hole with enlargements generally less than 1 in. The disagree-
MPHI and PCOR. The average difference between PND and ment between the log measured bulk density RHOB and the
MPHI is about 3-5 p.u. core derived CROB suggests a log error.
The high reading of the conventional logs might be caused The close agreement of MSIG and the over-burden cor-
by significant clay mineral content in the formation, whereas rected core porosity PCOR in contrast to the disagreement
the model to derive porosity from neutron and density logs is between PCOR and PND supports the superiority of the NMR
based on a clean, in this case sandstone formation (refer to measurement of total porosity. This conclusion is further sup-
Appendix A). ported by comparison of MSIG and MPHI where there differ-
In particular the neutron porosity is very sensitive to an in- ence indicate a minor fraction of clay pore sizes to be present
crease in clay mineral volume as indicated by increased CGR in the interval thus reducing the possibility that clay minerals
values. This behavior is clearly expressed at 10286 – 10290 ft. may have influenced the core derived porosity.
There are other zones (e.g. 10264 – 10268 ft, 10307 – 10309 The MSIG-MPHI indication of clay bound water is also
ft, 10317 – 10319 ft) where PNSS appears influenced by in- consistent with the 6-9 p.u. spread in neutron-density porosity,
creased clay mineral content, although the overestimation of providing the involved clay is smectite, a distinct possibility in
porosity in these intervals is not as prominent as at 10286 – these formations.
10290 ft. The previously stated concern with conventional porosity
In contrast to the conventional logs, MRIL total porosity determination in mixed mineral formations is seen in this dis-
MSIG is in good agreement with core porosity PCOR. How- play. Clearly, mineralogy is affecting the conventional logs
ever, there are three zones where MSIG and PCOR exhibit making the benefit of a mineralogy independent total porosity
different characteristics: In zone 10244 – 10251 ft, core po- apparent.
rosity shows decreasing values from about 17 p.u. down to
approximately 8 p.u. whereas MSIG slightly increases from 15 North Sea: The next example (Fig. 3) shows data from an
to 18 p.u. In the same interval the conventional logs show the exploration well in the North Sea. The formations are basically
same characteristics as the core data. However, the absolute quartz based shaly sands from estuarine; beach, washover and
value e.g. for PND is still about 6 p.u. too high. tidal delta; upper shore face and beach barrier sand, and,
The discrepancy between MRIL data and core porosity can braided-delta front. The facies variation and the associated
be explained by the fact that MRIL's vertical resolution is ap- depositional energy changes results in a considerable variance
proximately 3 ft.2 The tool "sees" the high porosity formation in reservoir quality due to grain-size and clay mineral content
at 10251 ft correctly, but is not able to resolve the low porosity as well as lithic fragments, commonly associated with micas
section (10244 - 10249 ft). and other heavy minerals.
The same applies for depth intervals 10262 – 10277 ft and Track 1 presents the gamma ray GR, track 2 shows density
10288 – 10290 ft. In all three zones RHOB as well as CROB porosity PDSS, neutron porosity PNSS and the neutron-
show an increase in bulk density suggesting that the formation density crossplot based porosity PND as well as core porosity
is tighter with less pore volume, i.e., porosity.
4 G. R. COATES, S. MENGER, M. PRAMMER, D. MILLER, NUMAR SPE 38736
PCOR. Track 3 compares MRIL total porosity MSIG, with tron-density crossplot based porosity PND. Again, track 3
neutron-density porosity PND and core porosity PCOR. compares MRIL total porosity MSIG and MRIL effective po-
Similar to the previous example, the conventionally derived rosity MPHI, with neutron-density porosity PND.
porosity (e.g., PDSS, PNSS, and PND) exhibit significantly The caliper log (CAL) indicates a moderately rugose bore-
higher values than MSIG and PCOR. In most parts PND is hole with enlargements up to 1 inch. With the exceptions of
about 3 p.u. higher than MSIG. few zones to be discussed below MPHI shows very low values
Again the higher reading of the conventional logs can be (0-1 p.u.) over the entire interval (i.e., 5650 – 5760 ft). MSIG
explained by significant clay content. The clay effect is most indicates a fairly constant porosity of about 9 p.u. with varia-
prominent in the depth interval 12688 – 12710 ft, where core tions of ± 3 p.u., while the conventional logs exhibit unsteady
data and MRIL porosity show low porosity formation while readings between 6 and 27 p.u., i.e., for PND.
the conventional logs, and PNSS in particular, erroneously There are several zones (5666-5668 ft, 5684 – 5688 ft,
suggest higher formation porosity of about 15 p.u.. Other 5702 – 5703 ft, 5718 – 5721 ft, 5738– 5739 ft and 5744 –
depth intervals (e.g., 12725 – 12728 ft, and 12752 – 12757 ft) 5748 ft) where MSIG, MPHI and PND closely match. In these
reveal similar problems if not as prominent. intervals, GR shows low readings relative to the rest of the
The only difference is found at 12630 – 12654 ft. Here gamma ray log.
neutron-density porosity PND seems to track the core porosity These observations suggest that MSIG gives the correct
more accurately while MSIG underestimates PCOR signifi- formation porosity. Only in relative clean (i.e., clay-free) zones
cantly. A detailed look at that section reveals, that PND is marked by the low GR readings, PND seems to indicate cor-
again strongly influenced by significant clay volume. Thus rect formation porosity. In those (clean) intervals, MPHI
PND suggests higher porosity (approximately 18 p.u.) than (which does not include the clay-bound porosity) approaches
found in the core data (13.5 p.u.). PND misses the 13.5 p.u. MSIG. This strongly suggests that those intervals contain only
core porosity point at 12646 ft. At this depth MSIG reads the little amount of clay.
lowest porosity with about 4 p.u. Higher up, i.e., 12642 – Only at 5684 – 5688 ft, the match between all three poros-
12638 ft, MSIG recovers. However, it still does not read the ity logs (i.e., MSIG, MPHI and PND) is almost perfect.
correct porosity of about 18 p.u. Whereas MSIG and PND correspond in all clean intervals,
As pointed out above MRIL's vertical resolution of about 3 MPHI exhibits the same characteristic as MSIG, but reads
ft might be the reason for this untypical behavior: The NMR lower values. An explanation could the 2 – 3 ft vertical resolu-
measurement correctly detects a low porosity zone at about tion of the MRIL tool already mentioned above. MPHI is not
12464 ft. But, as MRIL vertically integrates over about 2-3 ft., able to fully resolve thin beds with thickness of 1-2 ft.
it can not recover (i.e., in terms of depth) fast enough to accu- The conventional logs, in particular neutron porosity, are
rately map the high porosity part of the formation. severely distorted (i.e., they read too high) in layers with sig-
Since MRIL's results are lithology independent and its nificant clay contents as indicated by high GR and low MPHI
characteristics are well known, it should be possible to develop readings. While at places PND overestimates the porosity as
methods to account for and to correct such behavior. much as 18 p.u., MSIG correctly determines the true formation
porosity.
North East Utah: The example shown in Fig. 4 comes from These observations illustrate that, in contrast to conven-
the Douglas Creek member of the Green River formation in tional logs, MRIL measurements provide more reliable and
North East Utah. These formations are Eoscene deposits from accurate results where
Uinta mountain detritals that were deposited in an inland lake • borehole conditions are sub-optimal (rugosity),
environment. Reservoir salinity fluctuates from fresh to salty • formation porosity is relatively constant,
depending upon damming of the lake inlets. • mineral contents of the formation changes significantly.
The reservoir sands are dominated by quartzose deposits
with some organic materials present such as plant fragments.
Two depositional types of reservoirs are recognized, a fluvial-
delatic and a lacustrine to offshore bar form. The reservoirs
are stratigraphic traps related to sand quality variations associ- Conclusions
ated with grain size, and authegenic effects including cement- This study has found
ing and clay formation. • MRIL total porosity to closely match drilled core plug
Clay minerals are primarily illite or illite-smectite though porosity,
some authegenic kaolinite is also found, while cementing ma- • MRIL total porosity to be superior to conventionally de-
terials are carbonate dominated and often iron rich though rived porosity (i.e., density porosity, neutron porosity) in
some redeposited quartz cement is also found. mixed mineral formations and under sub-optimal borehole
In track 1 gamma ray, caliper and bulk density logs are conditions,
shown, GR, CAL and RHOB respectively. Track 2 presents • MRIL porosity to be an invaluable log for formation
density porosity PDSS, neutron porosity PNSS and the neu- evaluation.
SPE 38736 APPLYING NMR TOTAL AND EFFECTIVE POROSITY TO FORMATION EVALUATION 5
14. Akkurt, R., Moore, M.A., Freeman, J.J.: "Impact of NMR in the The formation is irradiated with fast neutrons generated
Development of a Deepwater Turbidite Field", paper SS pre- either by a chemical source or an electrical source (like used in
sented at the 38th Annual Logging Symposium of the Society of pulsed neutron tools). On their way from the source to the
Professional Well Log Analysts, Houston, June 15-18, 1997.
detector the neutrons are slowed down by elastic and inelastic
15. Kleinberg, R.L., Vinegar, H.L.: "NMR Properties of Reservoir
Fluids"", The Log Analyst Vol. 37, No.6, 20 – 32, 1996.
scattering. Neutron tools record either backscattered neutrons
directly or capture gamma radiation (i.e., gamma photons that
are generated when neutrons are captured by nuclei).
Appendix A — Physical Principles of Well Logs to
Due to the similarity in mass of neutron and hydrogen nu-
determine Formation Porosity.
cleus, the neutrons interact most effectively with hydrogen.
Hence the neutron porosity log is sensitive to hydrogen in the
Gamma-Gamma Density Log
formation i.e., free fluid water and chemically bound hydro-
Gamma-gamma density logging is based on the attenuation of
gen.
gamma rays in the formation. Density tools determine the at-
Neutron tool count rates are calibrated in terms of porosity
tenuation of gamma rays originating from a gamma source of
for different lithologies (sandstone, dolomite and limestone)
known strength. The attenuation is caused by Compton scat-
assuming that neutron scattering occurs mainly in the free fluid
tering i.e., interaction between gamma ray photons (0.1 to 10
in pores while scattering with hydrogen chemically bound in
MeV) and electrons of the outer shell of atoms. The amount of
the formation can be ignored.
scattered gamma photons and, in turn, the attenuation is di-
Since it is essential that the neutrons experience scattering
rectly related to the number electrons per volume unit NEl.
in the formation only and not in the borehole fluid (mud), most
This number can be calculated by
modern tools bring both source and detector in close contact
with the formation. However, neutron porosity logs are very
NEl = NA (Z/A) ρb (A-1)
sensitive to borehole irregularities in particular rugosity.
with NA: Avogadro’s number,
Sonic Log
Z: atomic number, number of protons in nucleus,
Sonic tools excite the formation by emitting acoustic waves of
A: mass number, sum of the number of protons and
typically several kHz signal frequency and measure the time
neutrons in nucleus,
the signal needs for a given distance. This interval travel time
ρb: bulk density.
∆t or so-called slowness depends on elastic properties of the
rock matrix, the properties of the rock fluid and the rock tex-
Since the ratio Z/A is almost constant for most materials of
ture, i.e. rock porosity. Wyllie et al.9 suggested that in com-
geologic interest except hydrogen, NEl and in turn the gamma
pacted formations ∆t can be expressed as the sum of the travel
ray attenuation is directly proportional to the bulk density ρb.
time ∆tma in the matrix fraction and the travel time ∆tfl in the
Thus ρb can easily be calculated from the measured gamma ray
liquid fraction
attenuation.
Knowing or assuming the matrix and fluid densities, ρma
∆t = (1-Φ)∆tma + Φ ∆tfl (A-4)
and ρfl respective, the bulk density can be written as
Solving this equation for porosity yields
ρb = (1-Φ)ρma + Φ ρfl (A-2)
Φ = (∆t - ∆tma) / (∆tfl - ∆tma) (A-5)
where φ is the rock porosity. Solving for porosity yields Thus it is possible to determine the porosity if the matrix and
fluid slownesses are known.
Φ = (ρma - ρb) / (ρma - ρfl) (A-3) Equation A-5 yields reliable results only in compacted
formations. For other formations like carbonates or uncom-
Since gamma density measurements have only a very shal- pacted formations other models10, 11 have been suggested
low depth of investigation typically smaller than the region which similarly to equation A-5 allow to determine rock po-
invaded by the borehole fluid, most of the time it is appropri- rosity from travel time measurements.
ate to ρfl = ρm with ρm density of the borehole fluid. However, Sonic logs can fairly reliably corrected for environmental
under certain circumstances e.g. in formations where the inva- effects i.e., mud velocity, borehole diameter. However, since
sion zone is very shallow, the derived formation porosity may the depth of investigation proportional to the wavelength of the
be invalid. acoustic signal, which in turn is dependent on the slowness of
the formation, the depth of investigation can only be assumed.
Neutron Porosity Log According to Bassiouni12 typical values vary between 0.75 ft
While gamma-gamma density logging is based on the attenua- for soft formations and 3.75 ft for hard formations. Hence the
tion of gamma rays, neutron porosity logging exploits the value of fluid slowness ∆tfl to be used in equation A-5 is sub-
scattering phenomena of neutrons. ject of uncertainty.
SPE 38736 APPLYING NMR TOTAL AND EFFECTIVE POROSITY TO FORMATION EVALUATION 7
Nuclear magnetic resonance (NMR) times applied between measurements.13 The echo-by-echo
NMR tools excite previously aligned hydrogen atom spins in difference between these signals highlights the presence of
the fluid or gas contained in the pore space to rotate about a long spin-lattice relaxation time T1 characteristic for light hy-
fixed magnetic field thus inducing an electric current recorded drocarbons.
by the tool.2 The recorded signal amplitude is proportional to In addition, hydrocarbon gases appear as fast decaying sig-
the number of hydrogen atoms within a tool specific measure- nals in the echo difference due to their rapid diffusion. Light
ment volume ("sensitive volume", e.g., about 750 cm3 for the oils diffuse much slower and cause a slowly decaying signal.
MRIL tool) and is independent of the rock formation. For this scheme to be quantifiable, the relaxation times of
In liquids, the number of hydrogen atoms NH per unit vol- oil and gas need to be predictable. The required parameters
ume (i.e., hydrogen density) is a weak function of pressure p can be extracted from the MRIL log itself by zone averaging14
and temperature T which tend to cancel each other between and can also be calculated from temperature and pressure of
surface and reservoir conditions. Hence, for fluids the hydro- the gas phase.1
gen density can be easily calculated
calibrations under all possible hole conditions regardless of After stacking 24 phase-alternated pairs and the customary
temperature and magnetic debris. vertical stacking, the fast echoes are virtually noise-free. In
Another total porosity mode is shown in Fig. B-1. Here, T1 this situation, phase correction is unnecessary and the magni-
contrast was considered irrelevant and instead of variable wait tude suffices for further processing. The T2 inversion algo-
times, short and long CPMG sequences are run in adjacent rithm employed senses the differences in signal-to-noise ratios
measurement volumes, interleaved with a partial-recovery, between standard, 1.2ms echoes, and fast echoes and adjusts
fast-echo burst sequence. The long and short CPMG’s share its internal stabilization parameter accordingly. A complete
the same echo spacing of 1.2ms and hence can be stacked to analysis of this algorithm can be found in the paper SPE-
improve the precision in the first few, very crucial echoes. 28368 (1994). The T2 inversion uses a base system of 0.5, 1, 2,
Typical echo counts are 400 for the long train and 10 for the 4, 8 and infinity (i.e., no decay) ms. In practice, the sensitivity
short one. As indicated in Fig. B-2, the first 1.2ms echo is dis- to fast components trails off at around 0.3ms. The amplitudes
carded because it contains some amount of redundant fast- associated with 8ms and infinity are discarded as remnants of
decaying signal. After phase and amplitude corrections, the incompletely recovered long-T2 components. The amplitudes
“standard” 1.2-ms echo train is converted into a T2 distribution for 0.5, 1, 2 and 4ms form the leftmost end of the T2 spectrum;
using 10 base components, logarithmically spaced between from there the standard distribution takes over and extends the
8ms and 2048ms. spectrum up to 2048ms.
The fast-echo pulse burst consists of 50 CPMG’s with 10 For the log analyst, these data acquisition and signal proc-
echoes each. The first two CPMG’s are used to establish a essing issues are mostly transparent. The end product is a sin-
steady-state dynamic equilibrium condition corresponding to a gle T2 spectrum spanning the range from fractions of a milli-
wait time of 20ms. The first 20 echoes are not processed. The second to several seconds. With proper hydrogen index cor-
remaining 48 CPMG’s form 24 phase-alternated pairs, which rection, the area under this spectrum equals total porosity from
are co-added both in I and Q channels, resulting in 10 complex MRIL. The interpretation effort consists, as described in the
numbers. Between pairs, the resonance frequency is varied in a main body of the present paper, in partitioning total porosity
fashion to ensure an artifact-free first echo, which is crucial for into the clay-bound water volume, the effective porosity and
the accurate determination of very fast T2 components. the free fluid volume.
SPE 38736 APPLYING NMR TOTAL AND EFFECTIVE POROSITY TO FORMATION EVALUATION 9
GR PDSS
0 API 150 30 PU 0
CAL PNSS MSIG
6 IN 16 30 PU 0 30 PU 0
RHOB PND PND
DEPTH
2.15 G/C3 2.65 FEET 30 PU 0 30 PU 0
CROB PCOR PCOR
2.15 G/C3 2.65 30 PU 0 30 PU 0
10200
Fig. 1: Logs and core data over a section of the Wilcox formation from a South Texas Well.
10 G. R. COATES, S. MENGER, M. PRAMMER, D. MILLER, NUMAR SPE 38736
MPHI
30 PU 0
GR PDSS MSIG
0 API 150 30 PU 0 30 PU 0
CAL PNSS PND
DEPTH
6 IN 16 FEET 30 PU 0 30 PU 0
RHOB PND PCOR
2.15 G/C3 2.65 30 PU 0 30 PU 0
11150
Fig. 2: Logs and core data from a different South Texas Field. The interval shown is also from the Wilcox formation.
SPE 38736 APPLYING NMR TOTAL AND EFFECTIVE POROSITY TO FORMATION EVALUATION 11
PDSS
30 PU 0
GR PNSS MPHI
0 API 150 30 PU 0 30 PU 0
RHOB DEPTH PND PND
2.15 G/C3 2.65 FEET 30 PU 0 30 PU 0
CROB PCOR PCOR
2.15 G/C3 2.65 30 PU 0 30 PU 0
12700
12800
Fig. 3: Logs and core data from a North sea shaly sand sequence.
12 G. R. COATES, S. MENGER, M. PRAMMER, D. MILLER, NUMAR SPE 38736
GR PDSS MPHI
0 API 150 30 PU 0 30 PU 0
CAL PNSS MSIG
DEPTH
6 IN 16 FEET 30 PU 0 30 PU 0
RHOB PND PND
2.15 G/C3 2.65 30 PU 0 30 PU 0
5700
Fig. 4: Logs from a Green River formation interval in a well in North East Utah.
SPE 38736 APPLYING NMR TOTAL AND EFFECTIVE POROSITY TO FORMATION EVALUATION 13
vol.2
ignore
partial stacking stacking
T2 Distribution
Fig. A-2—Processing sequence for total porosity.
Data with 1.2ms echo spacing is amplitude and phase corrected and converted into a T2 distribution between 8ms and
2048ms. The first 1.2-ms echo is redundant and is therefore eliminated. The fast-echo data consists of 24 phase-
alternated, frequency-swept pairs which are averaged to yield ten, essentially noise-free echoes. They are transformed into
the (left) fast end of the T2 spectrum consisting of the principal components 0.5ms, 1ms, 2ms and 4ms.