F O C U S

Henri-Georges Doll: 1902-1991

Henri-Georges Doll, for decades Schlumberger’s technical A page from Henri

guiding light, died after a long illness near Paris, France Doll’s notebooks, fea-
turing the develop-
on July 25, 1991 shortly before his 90th birthday. He is
ment of the laterolog
buried, as he wished, next to his mentor Conrad Schlum- technique. This per-
berger in a small cemetery at the Schlumberger family mitted logging forma-
estate in Normandy, France. tion resistivity with
less borehole effect
In a career spanning more than 40 years, he invented,
and much improved
designed, built and field tested geophysical instruments, vertical resolution.
developed interpretation methods, created research and
engineering centers, and guided young scientists and
engineers who could be trusted to carry on. He planted
fertile seeds outside geophysics as well. During World
War II he formed a company for developing a detector for
metallic land mines. This venture later produced automat-
ic guidance and telemetry systems, industrial instrumen-
tation, photomultiplier tubes and sealed-tube neutron
generators. Even after retirement from Schlumberger, his
restless technical spirit couldn’t desist. For more than a
decade, Doll worked at his own expense developing med-
ical instruments for the in-situ measurement of blood-flow
rate. In all these efforts he was guided by faith that practi-
cal technical success rests on scientific research and
engineering of the highest quality.
Henri-Georges Doll possessed a gentlemanly bearing
that came as from another age:
“The first thing that impresses one about Henri-Georges
Doll is his natural elegance... Surprise him in the field
dressed in knickers, or at his work table in a tweed jacket,
and what will strike you first is the ease with which he fits
into his surroundings. One cannot imagine that a poten-

4 Oilfield Review
January 1992 5
 tiometer would fail him, or that the lead would dare to
break at the tip of his pencil....He seems to be one of
those men who could walk across the Gobi Desert without
getting dust on his shoes; the passage of years would only
accentuate his neatness and complicity with matter.”
So wrote his former wife, Annette Gruner Schlumberger,
in her 1977 book The Schlumberger Adventure. In Doll,
this complicity with nature joined forces with a roving
mind, long-range vision, tenacity and mathematical ability
to produce an engineer and inventor par excellence.
His work, chronicled by authorship of more than 70
patents and over 30 publications, drew the highest esteem
from industry and government. During his lifetime, Doll
received the Lucas Gold Medal of the American Institute of
Mining, Metallurgical and Petroleum Engineers, the first
Gold Medal Award of the Society of Professional Well Log
Analysts, a Certificate of Appreciation from the US govern-
ment and elevation by the French government to “Officier
de la Légion d’Honneur.”
Henri-Georges Doll was born in Paris, France, August
11, 1902, to a Swiss father and French mother. During
World War I, while his father was serving as an officer in
the French Army, young Henri attended a lycée in Lyons,
where he and his three brothers lived with their mother,
who was working in a hospital. A brilliant student, he was
a natural candidate for the Ecole Polytechnique in Paris,
from which he later graduated. He then went on to higher
studies at the Ecole des Mines, also in Paris.
While still a student there, the 22-year-old engineer
married Annette Schlumberger, Conrad’s daughter, and in
late 1925 joined Conrad and Marcel Schlumberger’s small
electrical prospecting group part-time. The following year,
6 Oilfield Review
after graduation, he participated in electrical surface sur-
veys of the Alsace plain. By 1927 he was to design and
test equipment for Conrad’s newly conceived carrotage
électrique, or “electric coring.” Using this equipment, he
and two assistants a short time later ran the now legendary,
hand-plotted first electric log in Péchelbronn, France.
A creative flowering filled the following decade. He
started serious mathematical analysis of electrical logging
and conceived a differential magnetometer. He was first to
recognize the origins of small voltages called spontaneous
potentials that appeared on the measurement electrodes
of electrical logging tools even when no survey current
was emitted—and to note that these differentiated shales
from permeable conglomerates. He brought to fruition
borehole measurements of temperature, dip and inclina-
tion, the last of these employing the first sonde containing
complex downhole instruments—not merely wires and
electrodes. He even developed a safe and simple method
for such a mundane but ubiquitous operating problem as
locating cable leaks.
Shortly after the outbreak of World War II in Europe, Doll
was dispatched to Houston, Texas, USA to establish
Schlumberger’s first research and development activity
outside France. When the United States entered the war,
he offered his services to the US Army through a small
nonprofit company that he had founded, Electro-Mechani-
cal Research. Here Doll led the development of the jeep-
borne mine detector for which he later received a Certifi-
cate of Appreciation from the US government.
After the war, Doll spearheaded the separation of
research and engineering in Schlumberger by founding in
1948 what was then the Schlumberger Well Surveying Cor-
poration’s research laboratory in Ridgefield, Connecticut,
USA. Upon his retirement in 1967, the board of directors
named it the Schlumberger-Doll Research Center in his
honor. No matter what his other responsibilities, the urge
for technical creation came to the fore.
At Ridgefield, Doll not only led electrical tool develop-
ment, but expanded the staff to include nuclear, sonic,
mathematics/computer and log interpretation groups,
among others. Although he was never as technically com-
fortable with nuclear and sonic logging as with electrical,
his support and encouragement sustained these groups
through their early, lean years. This foresight informed all
his thinking. Nearly 40 years ago, Doll predicted the even-
tual demise of the log analyst’s slide rule and colored pen-
cil, and their replacement by truck-borne computers doing
on-line interpretation.
Roughly coincident with Doll’s tenure as Schlumberger’s
manager of technique in the decade following World War
II, a second creative flowering took place. Two papers
“The SP Log: Theoretical Analysis and Principles of Inter-
pretation” and “The SP Log in Shaly Sands,” now clas-
sics, described the current and voltage distributions of the
spontaneous potential in and around the borehole. These
clarified formerly puzzling aspects of the subject. Perhaps
the most widely referenced of all papers on induction log-
ging is his 1949 “Introduction to Induction Logging and
Application to Logging of Wells Drilled with Oil-Base
Mud.” These were followed by important publications on
the microlog, the laterolog and the microlaterolog.
Louis Allaud and Maurice Martin, in their book Schlum-
berger, The History of a Technique, give Doll full credit for
the basic theory of resistivity determination in a perme-
able formation, which took into account all the parame-
ters—borehole, depth of invasion and shoulders. The thin-
bed problem challenged him also. From the beginning of
logging, bed resolution had been a preoccupation of the
Schlumberger brothers and Doll. As early as 1927, Conrad
proposed a sonde employing long guard electrodes to force
the survey current into a beam that would penetrate later-
ally into the formation, even in the presence of salty mud.
Many years later he conceived a “point electrode” system
aimed at the same goal. But it wasn’t until Doll actually
designed a workable system using the then barely ade-
quate technology that the first laterolog was run in 1949.
Taking a cue from the success of the microlog, one of
his earlier inventions that located permeable zones, he
extended the laterolog concept to a pad tool. The micro-
laterolog was born, giving for the first time a reasonable
estimate of invaded-zone resistivity (Rxo) and residual oil
saturation. Then in 1958, Doll came up with the proximity
log, which gave an even better determination of Rxo, one
Acknowledgements and Further Reading
Schlumberger AG: The Schlumberger Adventure. New
York, USA: ARCO Publishing, Inc. 1982.
Allaud L and Martin MH: Schlumberger: The History of a
Technique. New York, USA: John Wiley and Sons, Inc.,
1977.
January 1992
less affected by mudcake. He had already in 1946 successful-
ly field tested a resistivity dipmeter sonde. This essentially
employed three rudimentary laterologs on pads spaced
around the borehole, with a common survey current electrode
several feet below.
There is little doubt, however, that Doll’s crowning techni-
cal achievement was inventing the induction log. The impetus
for this tool was the felicitous joining of a problem, the increas-
ing use of nonconductive oil-base mud, and a technique,
electromagnetic induction, which Doll had already used suc-
cessfully in mine detection. Schlumberger colleagues strong-
ly opposed his proposal, citing problems posed by very small
signal strength, huge direct mutual coupling interference and
lack of adequate supporting technology. Nevertheless, Doll
persisted in this complicated and difficult task, leading a small
team that often developed order-of-magnitude improvements
in technology as needed. History validated his vision, perse-
verance and faith in the eventual success of well grounded
research. Induction logging became one of the most widely
used logging methods in the world. It not only solved the oil-
base mud problem, but also surmounted the obstacle present-
ed to electrode methods by high-resistivity invaded zones.
Throughout his professional life, Doll was self-confident but
reserved in manner, soft-spoken but precise and authorita-
tive. Above all, he was a kind and considerate man. His lega-
cy to geophysics and hydrocarbon exploration is a thriving
technology and a clear image of the fruit that can grow from
seeds of vision, knowledge, perseverance and integrity.
Doll is survived by his first wife, Annette Gruner Schlum-
berger, his daughters, Mrs. Frank Davidson, Mrs. Jean Lebel,
and Mrs. Arnaud de Vitry, 10 grandchildren and 23 great-
grandchildren.—JT
7
 Trends in Reservoir Management
Twenty-five years ago, oil companies could
hope for continuing discoveries of giant
fields and highly profitable exploitation
once oil was found. Today, the tables have
turned. More than three-quarters of current
additions to the world’s oil reserves comes
from better management of existing reser-
voirs. Less than one quarter comes from dis-
covery of new oil. Profitability in today’s
harsher economic climate depends on
increasing recovery from producing fields. If
recovery is being optimized better now than
in years past, at least four factors could
claim credit:
First is the oil companies’ clear vision that
a cross-disciplinary approach must prevail. If
producing oil requires the expertise of geol-
ogists, geophysicists, petrophysicists, reser-
voir engineers and numerous other special-
ists, it is best that the experts work together
rather than separately. It sounds simple, but
entrenched management structures can
make this difficult to put into practice.1
Second is dramatic innovation in interwell
measurements, notably three-dimensional
(3D) seismics, and perhaps, in the future,
interwell seismic tomography. Developed
initially to aid exploration and improve
understanding of complex structure, 3D
seismics is being used increasingly to
Prepared with assistance from:
Peter Briggs, Manager of Reservoir Technology
BP Exploration
Uxbridge, England
Tony Corrigan, Partner
Corrigan Associates
Ditchling, England
8 Oilfield Review
As oil becomes harder to find and known reserves must be more
phase than the industry has ever experienced.
enhance reservoir description.2 Still lacking
the resolution to systematically pick out fine
sedimentological structure, it has neverthe-
less revolutionized the mapping of faults
and illuminated well-to-well correlation in
complex environments where logs and
cores offer poor or negligible correlation.
Three-dimensional seismic techniques have
also aided fluid monitoring in certain
enhanced oil recovery (EOR) schemes, but
have yet to be proved capable of monitoring
conventional water- and gasfloods.3
The third factor is increasing computer
power, which has made possible the disci-
plines of geostatistics and reservoir simula-
tion. We may tire of the computer industry’s
dramatic predictions of undreamed-of com-
puter power, but must remember that so far
the predictions have come true. Computer
power has allowed geologists and reservoir
engineers to create a range of probabilistic
models that fill space between wells where
measured data are lacking. The models are
used as input to fluid-flow simulations,
another activity that benefits directly from
advances in computer technology.
The fourth factor is horizontal well tech-
nology, which changes all the rules about
producing from both simple and complex
structures. The industry has learned how to
Michael Fetkovich, Sr. Principal Reservoir Engineer
Phillips Petroleum Company
Bartlesville Oklahoma, USA
Michel Gouilloud, Executive Vice-President
Schlumberger Ltd
Paris, France
drill these wells and is mastering how to
complete and stimulate them. Less clear is
how horizontal wells can best benefit flood-
ing strategies and enhanced oil recovery
(EOR) schemes.4
Given the increasing complexity of devel-
oping many of today’s discoveries, particu-
larly those in deep formations and offshore,
these innovations are giving reservoir man-
agement the edge it needs to make exploita-
tion worthwhile. The basic principles of
reservoir management, though, remain
unchanged. Every reservoir progresses
through the same phases in its producing
life and similar decisions must be made at
any given phase (next page ).
The most critical occur during the earliest
phases—discovery and appraisal—when the
least data are available. With scanty geo-
logic information, with the exploration seis-
mic survey, and with cores, logs and tests
from a handful of wells, oil companies have
to rapidly assess reservoir size and pro-
ducibility, decide well locations and drive
mechanism, and then commission surface
facilities that may cost billions.
As the field is drilled—the development
phase—an abundance of data becomes
available, allowing more detailed under-
standing of production. Simulation that was
Tien-when Lo, Project Scientist
Exploration and Production Technology Department
Texaco Inc.
Houston, Texas, USA
Björn Paulsson
Senior Research Geophysicist
Chevron Oil Field Research Company
La Habra, California, USA
carefully exploited, reservoir management is entering a tougher, more challenging
Production
Discovery
Appraisal
Development
Plateau
previously based on a simple reservoir
model because of limited data can now
grow more sophisticated. As the field enters
the so-called plateau phase, when produc-
tion reaches planned levels, there is also an
accumulation of production data that simu-
lation can history match. The net result is
more reliable prediction of future produc-
tion. This helps prepare the way for the
decline phase, when the oil company must
decide how to extract the most from its
dying asset. Issues range from whether to
infill drill—and if so, where—to modifying
surface facilities, and finally perhaps to
planning EOR (see “A Niche for Enhanced
Oil Recovery in the 1990s,” page 55).
Nansen Saleri, Manager of Reservoir Engineering
Chevron Exploration and Production Services, Co.
Houston, Texas, USA
John Warrender, Staff Geologist
Reservoir Development
Conoco (U.K.) Limited, Aberdeen, Scotland
January 1992
nPhases in a
field’s exploitation.
Time
Decline
More or less paralleling this saga is a
gradual narrowing of focus for the data
gatherers—a continuing quest for finer
detail. At every scale, their primary quest is
to better understand reservoir heterogeneity.
Heterogeneity governs not only connectiv-
ity—the degree to which the permeable
zones are interconnected and connected to
wells—but also horizontal and vertical
sweep efficiency and residual oil saturation
in swept zones. All are critical factors that
determine recovery. The ongoing challenge
facing operators is learning enough about
the heterogeneity that may influence the
next phase of the reservoir’s life (next page).
Koenraad Weber, Professor of Production Geology
Delft Technical University, The Netherlands
and Consultant in Reservoir Geology
Shell Internationale Petroleum Maatschappij BV,
The Hague, The Netherlands
At the appraisal phase, the concerns are
twofold: determining reservoir boundaries,
including those of the driving mechanism,
and determining large-scale heterogeneities
that complicate the reservoir’s internal struc-
ture. This means identifying the main depo-
sitional units and the boundaries between
them, and identifying faults and fractures.
The extremities of the reservoir and its
faults are exclusively provided by seismic
data, although gross reservoir size and dis-
tances to certain boundaries can be esti-
mated from well tests.5 A system of sealing or
partially sealing faults can completely change
the way a field should be produced. Know-
In this article, Array-Sonic and Formation MicroScan-
ner are marks of Schlumberger.
1. Thakur GC: “Reservoir Management: A Synergistic
Approach,“ paper SPE 20138, presented at the 1990
Permian Basin Oil and Gas Recovery Conference,
Midland, Texas, USA, March 8-9, 1990.
2. Robertson JD: “Reservoir Management Using 3D Seis-
mic Data,“ Journal of Petroleum Technology 41 (July
1989): 663-667.
3. Greaves RJ and Fulp TJ: “Three-Dimensional Seismic
Monitoring of an Enhanced Oil Recovery Process,”
Geophysics 52 (September 1987): 1175-1187.
4. Horizontal well technology is described in several
articles in the July 1990 issue of Oilfield Review.
5. Hansen T, Kingston J, Kjellesvik S, Lane G, l’Anson K,
Naylor R and Walker C: “3D Seismic Surveys,” Oil-
field Review 1, no. 3 (October 1989): 54-61.
9
 Reservoir heterogeneity Recovery Data
Sweep Reservoir Production Well logging Rock
efficiency pressure data / tests samples
Scale

Sidewall core
swept zones

Tracer tests
Production
distribution
3D seismic

distribution

Pulse tests

Production

Production
interaction
Horizontal

Rock/fluid

Horizontal
continuity
Reservoir

Outcrop/
Standard

reservoir
cuttings
Special
ROS in
Vertical

Vertical

history

Cores
tests

ROS
logs
Sealing fault

Semi-sealing fault
(>1000 ft)

Nonsealing fault
Giga

Fracturing
-tight

-open

Boundaries
genetic units
(100-1000 ft)
Mega

Permeability
zonation within
genetic units

Baffles within
genetic units
Macro
(in.-ft)

Lamination
cross-bedding

Microscopic
heterogeneity
(microns)
Micro

Textural types

Mineralogy

Strong effect Moderate effect

nReservoir heterogeneities versus scale and the data required to characterize them. At the beginning of a field’s life, attention
focuses on large-scale structure. As a field matures, attention shifts to finer detail. (From Weber and van Geuns, reference 22.)

10 Oilfield Review
ing the fault system, then determining if
faults are sealing may be the most important
information an operator should acquire as it
contemplates development. This is illus-
trated by yearly estimates of recoverable
reserves required of operators by the UK
government. These estimates show trends
that correlate with the field’s tectonic history
and its faulting system in particular.6
In tectonically quiescent areas with little
faulting, operators tend to increase reserves
estimates (below). Many of these North Sea
fields were deposited as submarine fans,
and the sands have a more or less uniform
morphology. They have limited faulting and
produce with waterdrive that leads to more
efficient drainage resulting in a continual
upgrading of reserves. In tectonically active
areas, reserves estimates tend to decline.
Operators found these fields to be more
faulted than expected, and in some cases
saddled with high permeability contrasts.
The industry’s solution is 3D seismics, lots
of it. Take Shell Expro’s Cormorant field that
lies in the highly faulted Brent province in
the northern North Sea in the UK sector.
The southern part of the field was devel-
oped in the 1970s before 3D seismics
became commercial, and production started
in 1980. Reserves estimates dropped steadily
during the buildup to production as reser-
voir engineers gradually perceived the
faulted nature of their asset. By 1981, when
Block IV in the north of the field was drilled,
reservoir managers had the benefit of a reli-
able 3D survey (next page, top ). This reaf-
firmed the complex faulting and enabled
reservoir managers to ensure that injector-
producer well pairs were at least in the
same unit.7
Another 3D seismic survey was commis-
sioned in 1984 to provide yet more detail,
the previous survey having suggested that
the average distance between faults was
about the same as or even less than the sur-
vey line spacing. The more closely spaced
survey showed the same big picture, but
enough detail was different to merit fine-

350 nNorth Sea opera- 6. Corrigan T: “Factors Controlling Successful Reserve

tors’ changing esti- Prediction: A Cautionary Tale from the UK North
300 mates of their Sea,“ presented at the 2nd Conference on Reservoir
250
fields’ recoverable Management in Field Development and Production,
Forties reserves—the spots Norwegian Petroleum Society, Stavanger, Norway,
200 indicate beginning November 14-15, 1988.
of production. In 7. Gaarenstroom L: “The Value of 3D Seismic in Field
tectonically quiet, Development,” paper SPE 13049, presented at the
relatively non- 59th SPE Annual Technical Conference and Exhibi-
faulted fields (red), tion, Houston, Texas, USA, September 16-19, 1984.
140
estimates tend to Ruijtenberg PA, Buchanan R and Marke P: “Three-
go up because Dimensional Data Improve Reservoir Mapping,”
Piper Journal of Petroleum Technology 42 (January 1990):
original forecasts
120 22-61.
were conservative.
Grant I, Marshall JD, Dietvorst P and Hordijk:
Estimated reserves, million tonnes

In highly faulted
“Improved Reservoir Management by Integrated
fields (greens), esti- Study: Cormorant Field, Block 1,” paper SPE 20891,
100
mates go down as presented at Europec 90, The Hague, The Nether-
operators come to lands, October 22-24, 1990.
Magnus
grips with reservoir
complexity. These
80 trends show how
important it is for
Dunlin N W Hutton operators to under-
stand complex
60
structure as early
Claymore Field as possible in the
life of a field. For
Thistle typical North Sea
40
Hutton crudes, 1 tonne is
equivalent to 7.53
Maureen barrels. (From Corri-
20 gan, reference 6.)

Tartan
0
75

77

79

81

83

85

87

89
19

19

19

19

19

19

19

19

January 1992 11
 N

Platform

Oil 0 2 km
1974-1975 1981 1983-1984 1989 Water
2D Seismic 3D Seismic 3D Seismic Void block
Fault

N.Cormorant
nAn increasingly
Norway platform complex view of
Shetland Cormorant the fault system in
Field Block IV of Shell
Isles
Expro’s North Sea
Cormorant field.
East Shetland Two 3D seismic
Basin surveys, in 1981
and 1984 respec-
Orkney tively, progres-
Isles sively clarified the
picture. By 1989,
data from newly
North Sea drilled wells added
UK Block IV more detail. The
UMC
0 100 miles most recent simu-
Aberdeen lation (far right)
used a curvilinear
grid to follow the
faulting. (From
tuning the locations of wells to be drilled.
Shell Expro’s estimates of the Cormorant
field recoverables have rebounded and now
remain relatively constant (below ).
The newcomer in detecting large-scale
structure is well-to-well tomography, a tech-
nique that measures the acoustic signal
transmitted from a source located in one
well to a receiver located in a neighboring
well.8 Measurements are made for multiple
combinations of receiver and source depths,
creating a large data set from which com-
puter processing recreates an acoustic
velocity map of the interwell terrain. The
30
Estimated reserves, million tonnes

Cormorant Ruijtenberg et al, ref-

Southern part
Alpha platform erence 7.)

Block IV
0 2 km
20

8. “CAT-Scanning the Subsurface,” Oilfield Review 2,
no. 2 (April 1990): 4-6.
9. Paulsson BNP, Fairborn JW, Cogley AL, Howlett DL,
Melton DR and Livingston N: “McKittrick Cross-Well
Seismology Project: Part I. Data Acquisition and
Tomographic Imaging,” Expanded Abstracts, 60th
Annual International Meeting and Exposition, Society
of Exploration Geophysicists, San Francisco, Califor-
nia, USA, September 23-27, 1990: 26-29.
Lo T-w, Inderwiesen PL, Howlett DL, Melton DR, Liv-
ingston DN, Paulsson BNP and Fairborn JW: “McKit-
trick Cross-Well Seismology Project: Part II. Tomo-
graphic Processing and Interpretation,” Expanded
Abstracts, 60th Annual International Meeting and
Exposition, Society of Exploration Geophysicists, San
Francisco, California, USA, September 23-27, 1990:
30-33.
12
10. Lasseter T, Karakas M and Schweitzer J: “Interpreting
an RFT*-Measured Pulse Test with a Three-Dimen-
10
sional Simulator,” SPE Formation Evaluation 3
(March 1988): 139-146.
11. Slentz LW: “Geochemistry of Reservoir Fluids as a
Unique Approach to Optimum Reservoir Manage-
ment,” paper SPE 9582, presented at the SPE Middle
East Oil Technical Conference, Manama, Bahrain, 0
March 9-12, 1981. 75 77 79 81 83 85 87
Gibbons K: “Use of Variations in Strontium Isotope
Ratios for Mapping Barriers: an Example from the nEstimates of recoverable reserves in the
Troll Field, Norwegian Continental Shelf,” presented Cormorant field. Initially, only the south-
at the 6th European Symposium on Improved Oil ern part of the field was developed. Its
Recovery, Stavanger, Norway, May 21-23, 1991. complex structure caused estimates of
12. Lachance DP and Rezk AS: “Resolution of Fault reserves to drop. Cormorant Block IV was
Block Communication Leading to an Optimal Plan developed three years later and benefited
of Depletion, Abu Gharadig Gas Field, Egypt,” paper from 3D seismic surveys mapping its com-
SPE 17993, presented at the SPE Middle East Oil plex structure. Reserves increased after
Technical Conference and Exhibition, Manama,
the addition of the new field and
Bahrain, March 11-14, 1989.
remained steady with time because the
13. Weber KJ, Mandl G, Pilaar WF, Lehner F and Pre-
operator better understood its structure.
cious RG: “The Role of Faults in Hydrocarbon
Migration and Trapping in Nigerian Growth Fault
Structures,” paper OTC 3356, presented at the 10th
Annual Offshore Technology Conference, Houston,
Texas, USA, May 8-11, 1978.
Oilfield Review
technique was conceived in the 1960s, but
implementation is still hampered by techni-
cal difficulties, notably the engineering of a
sufficiently powerful downhole source. In nTwo abutting
tomograms
current experiments, acoustic frequency lies revealing possi-
in the hundreds of hertz (Hz), between the ble fault structure
low frequencies of surface and borehole between wells in
seismics (10 to 100 Hz) and the high fre- the shallow
quencies of acoustic logging (20,000 Hz). McKittrick oil
field in California,
This is designed to provide both formation USA. Tomograms
penetration and spatial resolution. Most are obtained by
experiments so far have been in shallow for- transmitting
mations. Texaco Inc. and Chevron Oil Field acoustic signals
between the wells
Research recently performed a tomographic (see inset below)—
survey to clarify reservoir structure in the the technique is
shallow McKittrick oil field in California, experimental.
USA ( right ). 9 The image created by two The data were
abutting tomograms maintains a resolution obtained jointly
by Texaco and
of 40 ft [12 m], reveals details of overthrust Chevron; the
faulting and provides a substantially tomogram was
improved picture of the field’s heterogene- Well 1 Well 2 processed and
ity. However, much work remains to ensure interpreted by
Texaco Inc.
the reliability of tomographic processing. (From Lo et al, refer-
Locating faults is just half the story, S R ence 9, courtesy of
though. Also crucial is determining whether Texaco Inc. and
there is fluid communication across a fault Chevron Oil Field
R Research.)
when sand abuts sand. Several methods S
help clarify the sealing question. In interwell
testing, a flow disturbance is created in a
R
well on one side of the fault and monitored S
using pressure gauges in a well on the other
side, either at surface or downhole. Lack of
reaction to the disturbance may indicate
sealing. 10 In tracer tests, radioactive or
chemical tracers injected in one well are
monitored in neighboring producers. No
nFault sealing
through clay
tracer appearing in the producer may indi- Shale smearing, a phe-
cate sealing. Chemical analyses of forma- nomenon in which
tion water and hydrocarbons on both sides clay smears plasti-
of a fault may also give insight into connec- cally in a growing
fault creating a
tivity across it, the analyses matching if the barrier to fluid
fault is nonsealing.11 Material balance cal- flow. The likeli-
culations that indicate the volume of the hood of a sealing
connected reservoir offer another diagnos- fault due to clay
smearing can be
tic—the reservoir will appear larger if a fault estimated from the
is nonsealing.12 In a reservoir that has pro- thickness of
duced for some time, a common method is nearby clay zones
to make pressure measurements using an and the throw of
RFT (Repeat Formation Tester) tool in wells Sand the fault. (From
Weber et al, refer-
either side of the fault. Differing pressure ence 13.)
declines across the fault indicate sealing.
A nondirect method is based on the
hypothesis that faults seal because clay beds
Clay
cut by fault displacement smear plastically
into the fault, filling it and preventing com-
munication (right ). Clay smearing has been
reproduced in laboratory experiments and
also observed in situ down mines.13 Clay
smearing is more likely to seal a fault if
there is plenty of clay available—this can be
determined from logs. But it is less likely the
more the fault has been displaced—dis-

January 1992 13
placement can be estimated from seismic
data and in some cases from logs in neigh-
boring wells. Both factors can be worked
into a quantitative prediction of fault com-
munication and integrated into reservoir
simulation, a feat recently performed for the
Cormorant Block IV field.14
A yet more difficult task awaits the opera-
tor when evaluating a reservoir’s natural
fracture system. That fractures influence
gross field behavior is demonstrated in a
recent report surveying 80 fields from pro-
ducing areas worldwide and covering every
production method from waterdrive to gas-
drive to surfactant flooding.15 The survey
shows that preferential flooding direction
closely parallels the direction of maximum
horizontal stress, the same direction a natu-
ral fracture system takes (right ). The survey
rules out the possibility that induced frac-
tures may be causing flooding directional-
ity—the correlation improves with greater
spacing between wells, unlikely unless the
fractures are natural.
nConfirmation of natural fractures from cores and logs, in a Texas carbonate. Fractures are visible on the borehole televiewer and
Formation MicroScanner logs (black streaks) at 2234 and 2238 ft. They also appear as two peaks on the Stoneley fracture width log
which is derived from the Array-Sonic log (far right). (From Adams et al, reference 18.)
14
Preferential Detecting a natural fracture system and
flooding estimating its directionality is therefore
direction mandatory, and needed as early as possible
N
in the development of a field. The question
is how? Apart from regional stress studies,
little can be done before wells are drilled
and logs and cores become available. Con-
W E ventional 3D seismics fails to see natural
fractures. There is hope, however, that shear
General seismic surveys, both from the surface and
channels the borehole, may help. Shear waves vibrate
trends transversely to the wave direction and can
be split into two parts by fractures. Shear
S
Range of major surface seismic experiments made by
horizontal stress Amoco Production Co. have demonstrated
this splitting as have shear borehole seismic
surveys in the Paris basin, France and Silo
nCorrelation between preferential flood- field, Wyoming, USA.16 Shear seismics is in
ing orientation and direction of maximum
horizontal stress determined from well- its infancy, though, and one stumbling block
bore breakouts, for 80 fields worldwide may prevent the technique ever maturing: it
with a variety of flooding mechanisms. is practically impossible to generate shear
This result reinforces the importance of wave energy in marine sediments because
understanding a reservoir’s natural frac-
ture system and its role in determining the ocean separating source from formation
waterflood efficiency. (From Heffer and Lean, does not support shear propagation. All
reference 15.) experiments so far have been on land.
Oilfield Review
 Concrete proof of natural fractures finally
emerges when wells are drilled and frac-
tures are recognized in cores and on logs
(previous page, bottom). Cores and logs also
herald the beginning of serious depositional
study. With the help of a standard petro-
physical interpretation of conventional logs,
use of dipmeter logs17 and more recently
Formation MicroScanner logs18 can reveal
important clues about the depositional pro-
cess (right ). Skillfully interpreted, all these
logs together with core data allow a picture
of the reservoir’s internal structure to gradu-
ally emerge. Particularly crucial are perme-
ability barriers between units and high-per-
meability streaks that allow production to
bypass large regions of the reservoir. Barriers
may be suggested by depositional analysis,
but must be corroborated with pressure
measurements and well testing.
Of increasing use in clarifying deposi-
tional environment is, once again, the 3D
seismic survey. Sophisticated processing on
the amplitude data that reveals dip and
azimuth, shaliness, and even net pay is

14. Bentley MR and Barry JJ: “Representation of Fault

Sealing in a Reservoir Simulation: Cormorant Block
IV, UK North Sea,” paper SPE 22667, presented at
the 66th SPE Annual Technical Conference and
Exhibition, Dallas, Texas, USA, October 6-9, 1991.
15. Heffer KJ and Lean JC: “Earth Stress Orientation—a
Control on, and Guide to, Flooding Directionality in
a Majority of Reservoirs,” presented at the 3rd Inter-
national Reservoir Characterization Technical Con-
ference, Tulsa, Oklahoma, USA, November 3-5,
1991.
16. For a review of shear seismics and its application to
identifying fractures:
“Formation Anisotropy: Reckoning with its Effects,”
Oilfield Review 2, no. 1 (January 1990): 16-23.
For surface shear seismics:
Lynn HB and Thomsen LA: “Reflection Shear-Wave
Data Along the Principal Axes of Azimuthal
Anisotropy,” Expanded Abstracts, 56th Annual Inter-
national Meeting and Exhibition, Society of Explo-
ration Geophysicists, Houston, Texas, USA (1986):
473-476.
For borehole shear seismics:
Crampin S, Lynn HB and Booth DC: “Shear-Wave
VSP’s: A Powerful New Tool for Fracture and Reser-
voir Description,” Journal of Petroleum Technology
41 (March 1989): 282-288. nDepositional interpretation from well data facilitated by the integration of petrophysi-
17. Gilreath JA: “Strategies for Dipmeter Interpretation cal logs, Formation MicroScanner logs and computer-aided interpretations. In con-
(Part 1),” The Technical Review 35, no. 3 (July structing this composite display, a geologist has reviewed the Formation MicroScanner
1987): 28-41. log—left track in “Sedimentology/Structure”—on an interactive workstation and sym-
Adams J, Bourke L and Frisinger R: “Strategies for bolically coded an interpretation that appears to the right. The “Petrophysics” and
Dipmeter Interpretation (Part 2),” The Technical “Rock Classification” tracks are computer-interpreted from conventional logs. (From Dar-
Review 35, no. 4 (October 1987): 20-31.
ling et al, reference 18.)
18. Adams J, Bourke L and Buck S: “Integrating Forma-
tion MicroScanner Images and Cores,” Oilfield
Review 2, no. 1 (January 1990): 52-65.
Darling H, Patten D, Young RA and Schwarze L:
“Single-Well Data Integration,” Oilfield Review 3,
no. 3 (July 1991): 29-35.

January 1992 15
proving that minute changes in the seismic
signal may reflect real geologic events and
have an important story to tell.19 A time
slice from a survey in the Matagorda area of
the Gulf of Mexico reveals a meandering
stream channel (next page ).20 Amplitude
maps from the top and base of another
channel sand from one of Shell’s offshore
fields in Sarawak, Malaysia, obtained by
automatic tracking on the 3D seismic ampli-
tude data of the relevant horizons, display a
distributary channel and crevasse splay.21
Both examples are from less than 1000 m
[3200 ft], shallow enough to preserve the
high seismic frequencies required to see this
kind of detail. Obtaining the same resolu-
tion at greater depths that currently attenu-
ate high frequencies will depend on techni-
cal improvements to the acquisition
process.
The general problem facing the entire
reservoir management team, as eyes focus
on smaller-scale heterogeneity, is the blank
area between wells. 3D seismic data may
reliably outline the overall reservoir archi-
tecture and the fault system with an occa-
sional clue to deposition, and well data may
reveal the detailed depositional environ-
ment close to wells, but the huge space
between wells is mostly unknown. True, in
depositionally simple areas, features on logs
may correlate from one well to the next, and
the reservoir structure emerges quite readily.
But in complex areas, correlation may be
sparse or nonexistent. The only recourse is
for the geologist to dig into a repertoire of
typical sizes, shapes and juxtapositions for
depositional units and fill in the blank space
with as plausible a picture as possible.
For clastic reservoirs, this art has been for-
malized by Weber and van Geuns of Shell
(below ).22 Reservoir architecture is divided
into three types—layer cakes, jigsaw puzzles
and labyrinths. Layer cakes describe reser-
voirs deposited by a single depositional
mechanism that show excellent correlation
between wells. In jigsaw puzzle reservoirs,
different sand bodies fit together without
major gaps but with occasional intervening
low-permeability zones. Labyrinths represent
more or less random arrangements of sands,
usually discontinuous.
The first step is to recognize which type of
architecture is present. The next step—
which becomes harder with more complex
architecture—is to draw on knowledge of
how particular sand bodies are shaped and
distributed and then create plausible scenar-
ios for the reservoir geometry. This has
spawned two disciplines. One involves
patient hours in the field studying outcrops
and compiling statistics on the geometry
and occurrence of depositional units and
their heterogeneity, a labor often subcon-
tracted by operators to university depart-
ments. The second takes place in front of a
terminal connected to a powerful computer.
Here, a geostatistician uses the outcrop
statistics to help create probabilistic models
of the reservoir (for detail on outcrop studies
and reservoir model building, see “Reservoir
Characterization Using Expert Knowledge,
Data and Statistics,” page 25 ).23
A probabilistic reservoir model first hon-
ors known, or deterministic, data from wells
and then fills the empty space with sand
bodies both shaped and placed in space
randomly. The result is called a realization,
a term coined by statisticians to describe
one outcome of a random process. Many
realizations are required to judge variations
in architecture caused by the random ele-
ment of the model building. In simple layer
cake architectures, most of the model may
be deterministic with little or no probabilis-
tic content. Labyrinth architectures, on the
other hand, have little deterministic content
and are mostly probabilistic.
Probabilistic model building comes in a
variety of guises, and models can be con-
structed at all scales. Some of the finest scale
clastic models are being built by Shell using
their proprietary software package,

600 m
Layer cake

Deterministic

Jigsaw puzzle

Labyrinth
Probabilistic

50 m

nClassification of large-scale sand geometries into layer cakes, jigsaw puzzles and labyrinths. Layer
cakes are deposited in one environment, have excellent well-to-well correlations and can be modeled
deterministically. Jigsaw puzzles represent sand bodies that fit together without major gaps, but with
occasional low permeability barriers. Correlation may be difficult and modeling must have a probabilis-
tic content. Labyrinths represent discontinuous sands with poor well-to-well correlation. Modeling
requires a strong probabilistic content. (From Weber and van Geuns, reference 22.)

16 Oilfield Review
Base sand
nTwo examples of
channel sands
identified with 3D
seismic data:
Left: amplitude
maps of two hori-
zons picked from a
3D seismic data set
obtained in one of
Shell’s offshore
fields in Sarawak,
Malaysia. The hori-
zons show the top
and bottom of a
distributary chan-
nel sand. The top
horizon clearly
Top sand shows a crevasse
splay, later con-
firmed by logs.
(From Rijks and Jauf-
fred, reference 21,
courtesy of Shell Inter-
nationale Petroleum Fault
Maatschappij B.V.
and Geophysics: The
Leading Edge of
Exploration.)
Right: a time slice
from a survey con-
ducted by ARCO in Channel
the Matagorda
area of the US Gulf
Coast reveals a
meandering
stream channel,
probably gas filled.
The meander of the
stream suggests a
channel width of
Dis 15 ft [4.5 m], while
trib seismic resolution
uta
ry was estimated at
Ch only 40 ft [12 m].
an
ne The better-than-
l
expected accuracy
may be caused by
Crevasse interference or by
Splay the presence of
gas. (From Riese and
Winkleman, reference
20, courtesy of ARCO
Oil and Gas Company
0 1 2km and the American
Association of
Petroleum Geologists.)

19. Sonnelund L, Barkved O and Hagenes O: “Reservoir
Characterization by Seismic Classification Maps,”
paper SPE 20544, presented at the 65th SPE Annual
Technical Conference and Exhibition, New Orleans,
Louisiana, USA, September 23-26, 1990.
20. Riese WC and Winkleman BE: “Shallow Overlooked
Channels, Offshore Gulf of Mexico: Application of
3D Seismic Analysis to Stratigraphic Interpretation,”
in Bally AW (ed): Atlas of Seismic Stratigraphy, vol.
3. Tulsa, Oklahoma, USA: American Association of
Petroleum Geologists (1989): 59-65.
21. Rijks EJH and Jauffred JCEM: “Attribute Extraction:
an Important Application in any Detailed 3D Inter-
pretation Study,” Geophysics: The Leading Edge of
Exploration (September 1991): 11-19.
22. Weber KJ: “How Heterogeneity Affects Oil Recov-
ery,” in Lake LW and Carroll HB Jr. (eds): Reservoir
Characterization. Orlando, Florida, USA: Academic
Press Inc. (1986): 487-544..
Weber KJ and van Geuns LC: “Framework for Con-
structing Clastic Reservoir Simulation Models,” Jour-
nal of Petroleum Technology 42 (October 1990):
1248-1253, 1296-1297.
23. Haldorsen HH and Damsleth E: “Stochastic Model-
ing,” Journal of Petroleum Technology 42 (April
1990): 404-412.

January 1992 17
 Early in life of reservoir
Water Sand, all types
Channel-fill sand
Oil Sheet sand
Coarsening-upward
sand
Gas Sand, all types
Coal
Foreground scale 10 ft
100 m
MONARCH, which divides space into vox-
els, the three-dimensional version of a pixel,
each measuring about 50 m × 50 m × 0.6 m
thick [164 ft × 164 ft × 2 ft]—the exact
dimensions can vary.24 The model honors
well data from logs, down to the 0.6-m verti-
cal resolution, uses structural information
provided by seismic and dipmeter data. It
ensures a given sand-shale ratio and distin-
guishes between at least three different sand
types—for example, in a fluvial environment
between channels, mouth bars, and crevasse
splays. Finally, it obeys the statistics on
depositional unit width, thickness and length
as derived from outcrop or other studies.
The result is a three-dimensional matrix of
millions of voxels. Although this must be
reduced in size—scaled up, in simulation
parlance—by almost two orders of magni-
tude for simulating fluid flow, the detailed
model provides insight into gross reservoir
connectivity by tracking flow paths through
the voxels. Applied to two reservoir units in
the Cano Limon field in Colombia, one unit
producing and the other overlying unit not
yet completed, a MONARCH model showed
that 24 existing wells tapping the producing
unit would also tap 60% of the overlying
unit if it was completed. This is invaluable
information for reservoir managers needing
to decide if extra wells are needed to tap the
overlying unit.
18
Ten years later
Programmed to track different fluids in the
sand bodies, MONARCH-type models can
also integrate production logs and pulsed
neutron logs that measure saturation behind
casing. This is a major advance. Currently,
the commonest data used to monitor reser-
voir performance are well pressures and
production, usually from individual wells
but sometimes from groups of wells. These
data track gross well performance at best,
masking the relative contributions of different
producing layers. MONARCH models can
track flood advances in detail and identify
unswept parts of the reservoir (above ).
Tracking fluids through voxels at one par-
ticular moment in time, though, is a far cry
from reservoir simulation which builds a
flow model of the reservoir from past pro-
duction data—the history matching
phase—and then predicts future
production.25 In flow simulation, the reser-
voir is divided into a large number of inter-
connecting homogeneous tanks called grid
blocks, and then multiphase fluid flow
through them is solved as a function of time.
The computation requires the largest com-
puters, and maximum model size is cur-
rently about 50,000 grid blocks, each block
being more than 200 times the volume of a
voxel. Most simulations use fewer and even
larger blocks, particularly during the
appraisal phase when data are scarce.
nTracking fluid fronts using MONARCH,
Shell’s probabilistic reservoir modeling
package. MONARCH divides space into
millions of voxels, small volume elements
measuring typically 50 m x 50 m x 0.6 m
thick. The models honor data measured
at wells (grey vertical columns), but fill
the rest of space probabilistically.
MONARCH models permit analysis of for-
mation connectivity and also of fluid dis-
tributions throughout the reservoir.
This example shows part of a North Sea
reservoir early in the reservoir’s life and
after ten years of production. The models
are viewed from the south and do not show
an overall 8° westerly dip. Given the dif-
fering vertical and horizontal scales, a true
visualization is obtained by tilting up the
right-hand edge of the figure by about 25°.
There are three types of sand, identifi-
able by color if oil saturated. All three
types are uniformly colored blue if water
saturated and green if gas saturated.
Shales are transparent. The rightward
progress of the waterfront is clearly evi-
dent. Ten years on, several pockets of oil-
saturated sand remain unswept. (From
Budding et al, reference 24, courtesy of Koninkli-
jke/Shell en Produktie Laboratorium.)
Oilfield Review
 Opinions on the value of simulation
vary—from being an unproductive chore,
required for unitization or by government
regulation, to being the key tool in reservoir
management. Nevertheless, most fields
undergo some form of simulation as soon as
the discovery well hits pay. The pessimist’s
view stems in part from unreasonable
expectations about what simulation can
bring, which in turn depends on when sim-
ulation is used in the life of the reservoir.26
During discovery and appraisal phases,
simple simulations with few grid blocks rep-
resenting little more than an extension of
classical reservoir engineering techniques
can assist major decisions on production
facilities and how the reservoir should be
produced. With little data to go on, such a
simulation cannot be expected to accurately
predict production. Rather, its use is to test
production sensitivity to different drive
mechanisms and to variations in large-scale
structure, such as external boundaries, strati-
fication and whether faults are sealing, par-
tially sealing or open.
As the field matures and more data
become available, the simulation grows
more complex with an increasing number
of grid blocks. Prediction becomes more
reliable and helps guide all manner of reser-
voir management decisions—about infill
drilling, perhaps the drilling of horizontal
wells, evolving surface facilities to cope
with changing production, modifying the
driving mechanism, or justifying workover
(see “Drilling for Maximum Recovery,” page
21). Yet, simulation remains essentially a
qualitative tool. Like weather forecasting, it
strives to predict a potentially chaotic situa-
tion from a huge number of initial condi-
tions—pressure everywhere in space for
weather forecasting and pressure and satu-
ration in every grid block for simulation.27
Both systems are what mathematicians
call overdetermined. In simulation, the
24. Budding MC, Flint SS, Paardekam AHM and
Dubrule ORF: “Three-Dimensional Reservoir Geo-
logical Modeling of the Cano Limon Field, Colom-
bia,” AAPG Bulletin 74, no. 5 (May 1990): 621.
Abstract only. Full text to appear in the AAPG Bulletin.
Budding MC, Paardekam AHM and van Rossem SJ:
“3-D Connectivity and Architecture in Sandstone
Reservoirs,” paper SPE 22342, to be presented at the
SPE International Meeting on Petroleum Engineer-
ing, Beijing, China, March 23-28, 1992.
25. For brief reviews of reservoir simulation:
Mattax CC and Dalton RL: “Reservoir Simulation,“
Journal of Petroleum Technology 42 (June 1990):
692-695.
Breitenbach EA: “Reservoir Simulation: State of the
Art,” Journal of Petroleum Technology 43 (Septem-
ber 1991): 1033-1036.
January 1992
number of model parameters that can be
adjusted to history match past production
data greatly exceeds the quantity of produc-
tion data being matched. This permits many
possible matches for the same data. To
make the history match, reservoir engineers
therefore adjust certain parameters and
ignore the rest. Traditional favorites for
adjustment are grid-block properties such as
permeability and relative permeability.
Much less popular, because they are far
more trouble to adjust but maybe just as sig-
nificant, are geologically dependent param-
eters such as the size and placement of the
grid blocks themselves.
Simulations do not necessarily have to
reproduce a reservoir in its full three-dimen-
sionality. Early in the life of a field, in partic-
Single Well
Sectional
Producer
Injector
nVarieties of reservoir simulations. Reservoir management may begin with a simple
simulation model and then evolve to more complex representations as more data
become available.
26. Thomas GW, “The Role of Reservoir Simulation in
Optimal Reservoir Management,” paper SPE 14129,
presented at the 1986 SPE International Meeting on
Petroleum Engineering, Beijing, China, March 17-
20, 1986.
Wood ARO and Young MS: “The Role of Reservoir
Simulation in the Development of Some Major
North Sea Fields,“ paper SPE 17613, presented at
the 1988 SPE International Meeting on Petroleum
Engineering held in Tianjin, China, November 1-4,
1988.
Wittmann M, Al-Rabah AK, Bansal PP, Breitenbach
EA, Hallenbeck LD, Meehan DN and Saleri NG:
“Exploring the Role of Reservoir Simulation,“ Oil-
field Review 2, no. 2 (April 1990): 18-30.
ular, much simpler simulations can yield
valuable insight into production (below ).
Single-well simulators, using a radial grid
system centered on the well, are invaluable
for studying coning, the economics of a par-
tial completion, or for interpreting a well
test. Sectional simulators that model flow in
a vertical plane are often used to study dis-
placement behavior between an injector
and a producer. Areal simulators that model
flow in a horizontal plane may be an opera-
tor’s first choice for analyzing production in
a single-layered homogeneous reservoir.
The results and analyses of any of these sim-
ulations can always serve as input to a more
sophisticated simulation.
Areal simulators were the first choice for
Saudi Aramco when planning water injec-
Areal
Full Field
27. Saleri NG and Toronyi RM: “Engineering Control in
Reservoir Simulation: Part I,” paper SPE 18305, pre-
sented at the 63rd SPE Annual Technical Conference
and Exhibition, Houston, Texas, USA, October 2-5,
1988.
Saleri NG, Toronyi RM and Snyder DE: “Data and
Data Hierarchy,” paper SPE 21369, presented at the
SPE Middle East Oil Show, Manama, Bahrain,
November 16-19, 1991.
19
 Reservoir I
1824 grid blocks 32,760 grid blocks

57

32 2520 grid blocks

13
60

60

42 42

Reservoirs I and II

1 I
Dense limestone
60
4 II
13
42 I

Dense limestone
60
4 II

42
Aquifer and Reservoir I

42,000 grid blocks

I
19
8

I
60

42

nEvolving simulation complexity for one of Saudi Aramco’s largest carbonate reservoirs. The development began with
areal simulations of Reservoir I, then a full-field simulation, next an investigation of the underlying Reservoir II, and finally
of the surrounding aquifer. The number of grid blocks starts at 1824 and finishes at 42,000.

20 Oilfield Review
tion in their largest, carbonate field (previ-
ous page ). 28 This first simulation was
designed in 1970, 20 years after parts of the
field started producing and used a 57 × 32
grid, giving 1824 grid blocks. It was
enlarged in 1978 to 60 × 42, giving 2520
grid blocks. These simulations were used to
plan individual well productions to sustain
an overall production rate, to forecast the
length of the production plateau and to esti-
mate final recovery.
By 1980, water encroachment had
become serious enough to consider giving
the areal simulation a third, vertical dimen-
sion—two horizontal dimensions rule out
the modeling of coning. Geologists identi-
fied 13 layers within three major producing
zones and superimposed them on the previ-
ous areal grid, giving a simulation compris-
ing 60 × 42 × 13 = 32,760 grid blocks. Con-
currently, reservoir engineers embarked on
a major program of reservoir-monitor log-
ging—flowmeters and pulsed neutron mea-
surements mainly—to provide sufficient
information to history match the field’s pro-
duction. By 1985, the simulator was history
matched and available for prediction stud-
ies. These have included determination of
the original oil/water contact, the problem
of varying depletion from different zones,
the possibility of also producing the reser-
voir’s gas cap and evaluation of gas lift. In
short, the 3D simulator has proved a key
tool for most reservoir management deci-
sions in the field.
About 500 ft [150 m] below this huge
reservoir lies a smaller reservoir of inferior
quality with about one-tenth the individual
well productivity. By the time the first well
in the smaller reservoir (Reservoir II) was
put on production in 1954—eight years
after the huge reservoir (Reservoir I) had
started producing—Reservoir II’s pressure
had dropped by about 900 psi. This indi-
cated communication between the two
reservoirs, which geologists attributed to a
fracture system permeating the intervening
500 feet of dense limestone. The urgent
question was how to contain the communi-
cation, since it could cause both oil and
water to migrate downward from Reservoir
I to Reservoir II as Reservoir II’s pressure
declined. Juggling the options required sim-
ulating both reservoirs simultaneously. The
first attempt divided Reservoir II into four
28. Al-Dawas SM and Krishnamoorthy AM: “Evolution
of Reservoir Simulation for a Large Carbonate Reser-
voir,” paper SPE 17938, presented at the SPE Middle
East Technical Conference and Exhibition, Manama,
Bahrain, March 11-14, 1989.
January 1992
layers, which were placed below a single
areal grid simulating Reservoir I. Later, the
four layers were combined with the full 13-
layer model previously established for
Reservoir I, giving a total of 37,800 grid
blocks. This is currently being used to
assess how to produce Reservoir II.
Yet more simulation has been required to
study the extensive aquifer that drives Reser-
voir I’s production and also that of other
nearby fields. The aquifer was modeled with
an areal grid in the middle of which Reser-
voir I occupies 19 × 8 grid blocks, much
coarser than the reservoir’s original 60 × 42
areal simulation. During history matching,
the aquifer and Reservoir I simulators were
run iteratively, with results from the finer
gridded reservoir simulation being scaled up
to the coarser grid blocks of the aquifer sim-
ulation. A final step in simulating this field
has been the use of locally refined gridding
that allows the integration of the fine 3D
reservoir model directly into the coarse,
areal aquifer model, obviating the iterative
process. This raises the number of grid
blocks to 42,000.
Saudi Aramco’s field was developed long
before the age of computers, so simulation
could be applied only years after production
commenced. In more recent fields, particu-
larly those in geologically complex and eco-
nomically harsh areas, simulation is often
used from the moment of discovery. The
Brent field in the North Sea, jointly owned
by Shell and Esso, provides an example.
The first simulation took place during the
appraisal phase using just 800 grid blocks.29
It successfully guided basic decisions on
field development and facilities planning.
The next simulation built during the devel-
opment phase used 11 layers and an areal
grid of 19 × 37, giving a total of 7733 grid
blocks.30 This proved adequate to confirm
the water injection program, investigate fur-
ther facilities requirements and estimate the
length of the plateau. As the field entered
decline phase, a third simulation was con-
ducted with an even finer grid to take
advantage of the abundant acquired data
and better track the water front. This time,
21 layers and an areal grid 30 × 54 pro-
duced 34,020 grid blocks. Reservoir engi-
neers used the simulation to pick new well
locations and optimize a gas injection pro-
gram.31 A comparison of production predic-
29. Johnson HA: “North Sea Reservoir Simulation: Prac-
tical Considerations,” paper SPE 19039, presented at
the SPE Middle East Technical Conference and Exhi-
bition, Manama, Bahrain, March 11-14, 1989.
30. Tollas JM and Sayers JR: “Brent Field 3-D Reservoir
Simulation,“ paper SPE 12160, presented at the 58th
SPE Annual Technical Conference and Exhibition,
San Francisco, California, USA, October 5-8, 1983.
Drilling for Maximum Recov-
ery
Whatever drive mechanism is chosen, ultimate
recovery depends on ensuring maximum connec-
tivity between wells and the producing formation.
This requires understanding the nature of the for-
mation heterogeneity and then guiding the drill
bit to reach as much connected oil as possible.
In reservoirs where faults subdivide the reser-
voir into isolated pockets, 3D seismics with its
ability to map large-scale structure offers the
best chance to ensure that each pocket gets
tapped by a well. An example is Chevron’s giant
Bay Marchand field, offshore Gulf of Mexico.1
This field, discovered in 1949, is located around
a huge salt dome and produces from highly
faulted sands. Oil production peaked at 75,000
barrels of oil per day (BOPD) in the late 1960s
and declined to just 18,000 BOPD by the mid
1980s. It looked like the end, until Chevron com-
missioned a 3D seismic survey of the entire field.
With over 60 platforms and 50 multiwell sur-
face structures scattered over the offshore site,
the logistics were daunting. Once completed,
however, the survey clarified the salt-sediment
interface, gave improved definition to the faulting
and provided a better understanding of the depo-
sitional environment. The result has been several
new producers in both mature and undeveloped
parts of the field, a realization that certain wells
were not worth maintaining, and improved water-
flood schemes. Field production now stands at
32,500 BOPD.
A different strategy is required in highly strati-
fied reservoirs where poor connectivity is caused
by discontinuous stringers, such as in Exxon’s
Fullerton field in Andrews County, Texas, USA.
Discovered in 1942, this field was rapidly
exploited with over 550 wells. By the 1960s, the
field was in decline and substantial water injec-
tion was required to maintain oil production at
around 10,000 BOPD. Although the producing
1. Abriel WL, Neale PS, Tissue JS and Wright RM: “Modern
Technology in an Old Area: Bay Marchand Revisited,”
Geophysics: The Leading Edge of Exploration (June
1991): 21-35.
21

100
From 40-acre spacing
Continuity, %
From 20-acre spacing
From 10-acre spacing
0 reservoir. The inter-
0 1000 2000 3000
Distance between wells, ft
nPercentage of reservoir formation
hydraulically connecting two wells versus
distance between them, for Exxon’s Fuller-
ton, Texas field. The data were obtained
from wells drilled with 40-, 20-, and 10-
acre spacing. As distance between wells
increases, continuity worsens. And as
more data become available, previous
estimates on continuity appear overly
optimistic. (From Barber et al, reference 3.)
2. Stiles LH: “Optimizing Waterflood Recovery in a Mature
Waterflood, the Fullerton Clearfork Unit,” paper SPE
6198, presented at the 51st SPE Annual Technical Confer-
ence and Exhibition, New Orleans, Louisiana, USA, Octo-
ber 3-6, 1976.
3. Barber AH Jr, George CJ, Stiles LH and Thompson BB:
“Infill Drilling to Increase Reserves—Actual Experience in
Nine Fields in Texas, Oklahoma, and Illinois,” Journal of
Petroleum Technology 35 (August 1983): 1530-1538.
22
40-acre spacing Phase 1
Phase 2
- Producer - Injector
nTwo-phase sequence of infill drilling in Exxon’s Fullerton, Texas reservoir. In phase 1,
new producers were drilled in the center line of the previous three-line drive, and other
producers were converted to injection. In phase 2, another producer was drilled in each
80-acre rectangle. The infill program tapped reserves unconnected to previous wells
and prolonged the field’s life by at least ten years. (From Barber et al, reference 3.)
20 nIncrease in pro-
duction from infill
drilling in Exxon’s
Fullerton, Texas
Production, BOPD x1000
Other Infills ference production
is that lost from the
original wells to
10
Phase 1 the new wells.
(From Barber et al, ref-
Phase 2
erence 3.)
40-acre wells
Interference
0
1965 1975 1985 1995
4. McCoy TF, Fetkovich MJ, Needham RB and Reese DE: zone is several hundred feet thick, the pay is in
“Analysis of Kansas Hugoton Infill Drilling: Part I—Total carbonate stringers that correlate very poorly
Field Results,” paper SPE 20756, presented at the 65th
SPE Annual Technical Conference and Exhibition, New
between wells.2 Exxon found during infill drilling
Orleans, Louisiana, USA, September 23-26, 1990. that the correlation seemed to get worse the
Fetkovich MJ, Needham RB and McCoy TF, : “Analysis of closer you looked. A continuity measure was
Kansas Hugoton Infill Drilling: Part II—12-Year Perfor-
mance History of Five Replacement Wells,” paper SPE defined as the percentage of producing carbonate
20779, presented at the 65th SPE Annual Technical Con- that hydraulically connects between two given
ference and Exhibition, New Orleans, Louisiana, USA,
September 23-26, 1990.
wells—that is, the percentage that would theoret-
Fetkovich MJ, Ebbs DJ Jr and Voelker JJ: “Development ically get produced if one well was an injector
of a Multiwell, Multilayer Model to Evaluate Infill Drilling and the other a producer. Exxon then investigated
Potential in the Oklahoma Hugoton Field,” paper SPE
20778, presented at the 65th SPE Annual Technical Con- how this continuity depended on well spacing
ference and Exhibition, New Orleans, Louisiana, USA, (above, far left). Initially, the information came
September 23-26, 1990.
from the original wells, which were on a 40-acre
5. Freyss HP and Burgess K: “Overcoming Lateral Reservoir
Heterogeneities Via Horizontal Wells, ” presented at the spacing.
6th European Improved Oil Recovery Symposium, Sta- It came as no surprise that continuity worsened
vanger, Norway, May 21-23, 1991.
as well spacing increased. But as the field was
Oilfield Review
drilled with denser spacings, data showed that 500
2nd 3rd
earlier estimates of continuity were grossly opti-
mistic.3 The only means of increasing recovery 400

Production, BOPD x 1000

was to drill progressively denser well spacings
(previous page, top). This was carried out in two
phases. The first took well spacing to 20 acres,
the second to 10. When the second phase was
completed, the new wells accounted for 71% of
the field’s production, and each well added an
estimated 97,000 barrels to recoverable reserves
(previous page, middle).
However tempting it may look, though, infill
drilling is not a panacea. This is the view of
Fetkovich and colleagues at Phillips Petroleum
Co. with respect to the Hugoton gas field, the
largest in the lower 48 US states, spanning parts
of Kansas, Oklahoma and Texas. The Hugoton
comprises interspersed carbonate and layered
shaly units—continuity is excellent, the log sig-
nature being consistent from one well to the next.
When the state authorities authorized infill
drilling in the late 1980s, the Phillips team
proved through studying production histories and
conducting simulations that the reservoir was
essentially layered with no vertical crossflow.
They concluded that while infill drilling might
increase productivity, the new wells would add
nothing to the field’s reserves.4 Thanks to the for-
mation continuity, all the gas was already con-
nected to the existing wells.
Production from natural fractures, such as in
the Austin chalk formations in south Texas, poses
yet another type of heterogeneity that until
recently sorely tested operators’ attempts to
secure economic recovery. The Pearsall field
exploited by Oryx is typical. The oil is contained
in successive clusters of roughly parallel subver-
tical fractures. Each cluster acts as a separate oil
accumulation and lacks communication with its
neighbors. Produced through vertical wells, these
formations provide unsubstantial economic
return. But connect several together with a hori-
zontal well, and the economics improve immedi-
ately.5 Overcoming reservoir heterogeneity was a
key impetus behind the horizontal well revolution.
300
200
100
0
1977 1979 1981 1983 1985 1987 1989 1991 1993 1995
nProduction forecasts from second and third simulations for Shell and Esso’s North Sea
Brent field. The second simulation, made in 1982 during field development and
intended to investigate plateau-phase production, gave a reasonable forecast for five
years. The third simulation, made in 1988 during the decline phase, gives a more
accurate picture of the field’s later life and allows planning of remedial programs. (The
first simulation, not shown, was made at very coarse scale during the appraisal phase
of the field.) (From Johnson, reference 29)
tions from the second and third simulations these sands for a 3- by 2.5-kilometer [1.9-
shows how the finer simulation provides a by 1.6-mile] sector in the southern part of
more optimistic, and presumed truer, sce- the reservoir (next page, top). The sector
nario for the field’s decline phase (above ). was chosen because it had abundant well
What is the trend in simulation? Advances data—six oil producers, three water injec-
in computer power will undoubtedly spur tors and one gas injector. The probabilistic
simulations with more grid blocks, repre- model was then sandwiched between
senting finer reservoir detail. It is conceiv- deterministic models of the upper and
able, for example, that simulating grid lower parts of the Statfjord reservoir, and
blocks the size of the MONARCH system’s finally the whole thing was run through a
voxels may eventually be possible. If this simulator. The results were compared with
happens, reservoir management will a simulation of the Statfjord reservoir con-
become reliant on probabilistic modeling, structed entirely deterministically.
since no field technique, even the most Using the MONARCH package, three real-
experimental, appears capable of measuring izations of the channel sands were con-
to that degree of detail between wells—all structed using 172,000 voxels, each measur-
the simulations described in this article and ing 75 m × 25 m × 1.2 m thick [250 ft × 80 ft
almost all described in the literature are × 4 ft]. Each realization had to be reduced to
deterministic. The future may lie in con- about 10,000 grid blocks before sandwich-
structing a geologically probabilistic model ing and simulation. The comparisons
from the beginning and then firming it up showed that simulations using the proba-
deterministically as data become available. bilistic realizations predicted water break-
The latest simulation work in the Brent field through far more accurately than an entirely
represents one of the first applications of deterministic simulation. Credited with the
probabilistic modeling.32 improvement was MONARCH’s ability to
The Brent field actually comprises two create a more realistic picture of shale barri-
major reservoirs, the middle Jurassic Brent ers within the channel sands. These appear
reservoir and the lower Jurassic Statfjord to influence not only water breakthrough in
reservoir, that are separated by 250 m [820 the channels, but also breakthrough in the
ft] of shale. Some of the most complex overlying and underlying units that were
reservoir sediments and therefore most modeled deterministically.
troublesome during history matching are Simulation requires intense collaboration
channel sands in the middle part of the between geologists and reservoir engineers,
Statfjord unit. It was therefore decided to a cross-disciplinary challenge that operators
construct a detailed probabilistic model of face in virtually every area of reservoir man-

31. Tollas JM and McKinney A: “Brent Field 3-D Reser-
voir Simulation,” paper SPE 18306, presented at the
63rd SPE Annual Technical Conference and Exhibi-
tion, Houston, Texas, USA, October 2-5, 1988.
32. Keijzer JH and Kortekaas TFM: “Comparison of
Deterministic and Probabilistic Simulation Models
of Channel Sands in the Statfjord Reservoir, Brent
Field,” paper SPE 20947, presented at Europec 90,
The Hague, The Netherlands, October 22-24,1990.

January 1992 23
 Reservoir Geological 2nd 3rd Probabilistic agement. Other options and decisions in
Unit Simulation Simulation simulation reservoir development involve expertise
1 1 from a plethora of disciplines—drilling engi-
I 2 neering, environmental engineering, facili-
2 3 ties engineering to name a few. Integrating
4 them can be hard work.
3 5 Conoco (U.K.) Limited has identified sev-
II 6
4 7 eral factors that help.33 On a team of differ-
8 ent experts, there may have to be a boss,
Brent 9
but no single expertise should be allowed to
III 5 10 dominate. As far as management is willing
11 to devolve responsibility, the team should
12 set its own goals and audit its achievements.
IV 6 13 The team should be as independent as pos-
14 sible from the parent structure, and thus
have the freedom to react and make deci-
sions rapidly—in other words, act like
Shale
entrepreneurs (below, left ). Last, there must
be persistent cross-disciplinary communica-
tion and education, not only between
1 7
15 15 experts, but also a reaching out to financial
and business colleagues in the parent group.
Conoco put this theory into practice while
16
2 8 planning a highly deviated well in their geo-
17
Probabilistic logically complex North Sea Murchison
Statfjord
field. The well was intended to tap multiple
18 targets in the main field and south flank.
3.1 9
19 19 Geoscientists made a detailed interpretation
of the south flank so the planned well hit
10 20 20 the targets; drilling engineers decided if the
3.2
more than 80-degree well was drillable and
11 21 21
designed a casing program that could han-
dle the varying depletion pressures in the
nIncreasing complexity in modeling the producing layers in the Brent and Statfjord different units; and reservoir engineers
reservoirs of Shell and Esso’s North Sea Brent field. In the most recent simulation, sev- investigated whether the casing program
eral layers in the Statfjord reservoir were modeled probabilistically and sandwiched
between deterministic layers. (The first simulation, not shown, was made at very coarse could be simplified by shutting in existing
scale during the appraisal phase of the field.) producers or injectors to adjust these pres-
sure differentials. The design process was
Before nOrganizational iterative and had to involve partners and
changes that outside specialists. Yet, it was done speedily
Aberdeen London Conoco (U.K.) Ltd because all experts worked together and
implemented to
ensure a cross-dis- were in constant communication.
Exploration
ciplinary approach In tomorrow’s world, every aspect of
when planning a reservoir management will be developed
high-angle well in and managed through synergy—by integrat-
Engineering Drilling Development their North Sea
Murchison field. ing data, by coupling compatible method-
After restructuring, ologies and by building teams. Add better
Production Operations Reservoir Geology Geophysics the key disciplines measurements for seeing between wells and
(boxed)—drilling almost unimaginable increases in computer
engineers, geolo- power, and we may see hydrocarbon
After gists, reservoir
engineers and geo- exploitation becoming as much a science as
Aberdeen an art. —HE
physicists—all
worked in one team
in the same office. 33. King W and Warrender J: “The Murchison Field: A
(From King and War- Multidisciplinary Approach to Reservoir Manage-
Engineering Reservoir development
render, reference 33.) ment,” presented at Advances in Reservoir Technol-
ogy, Edinburgh, Scotland, February 21-22, 1991.

Drilling Reservoir Geology Geophysics

24 Oilfield Review
 Reservoir Characterization
Using Expert Knowledge, Data and Statistics
Prepared with assistance from:
Peter Corvi, Head of Effective Properties
Kes Heffer, Head of Reservoir Description
Peter King, Senior Reservoir Engineer,
Effective Properties
Stephen Tyson, Computing Consultant
Georges Verly, Senior Geostatistician,
Effective Properties
BP Research
Sunbury-on-Thames, England
Christine Ehlig-Economides, Reservoir
Dynamics Program Leader
Isabelle Le Nir, Geologist
Shuki Ronen, Geophysics Integration
Program Leader
Phil Schultz, Department Head
Etudes et Productions Schlumberger
Reservoir Characterization Department
Clamart, France
Patrick Corbett, Research Associate
Jonathan Lewis, Conoco Lecturer in Geology
Gillian Pickup, Research Associate
Phillip Ringrose, Senior Research Associate
Heriot-Watt University
Department of Petroleum Engineering
Edinburgh, Scotland
Dominique Guérillot, Sr. Research Engineer,
Reservoir Engineering Division
Lucien Montadert, Director of Exploration
and Production
Christian Ravenne, Principal Research Associate,
Geology and Chemistry Division
Institut Français du Pétrole
Rueil-Malmaison, France
Helge Haldorsen, Assistant Director, International
Field Developments
Norsk Hydro A.S.
Oslo, Norway
Thomas Hewett, Professor of Petroleum
Engineering
Stanford University
California, USA
January 1992
Accurately simulating field performance requires knowing properties
like porosity and permeability throughout the reservoir. Yet wells provid-
ing the majority of data may occupy only a billionth of the total reser-
voir volume. Transforming this paucity of data into a geologic model for
simulating fluid flow remains an industry priority.
The simplest representation of a reservoir is
a homogeneous tank with uniform proper-
ties throughout. Characterizing such a sim-
plified model requires merely enough data
to define its volume and basic petrophysical
properties. This information can then be
used to help forecast reservoir behavior and
compare possible production scenarios.
But efficient exploitation of reserves
requires a more sophisticated approach.
Reservoirs have been created by complex
sedimentary and diagenetic processes, and
modified by a history of tectonic change.
Rather than resembling simple tanks, reser-
voirs are heterogeneous structures at every
scale. The need for accurate characteriza-
tion is vital for oil company economic plan-
ning (next page, top ).
In the past, geologists have used expert
knowledge to interpolate between wells and
produce basic reservoir models. However,
during the last ten years, this effort has been
transformed by statistical modeling offering
insight into the effects of heterogeneity.
This article details how reservoir descrip-
tions are developed and is divided into
three stages (page 27 ):
•Defining the reservoir’s large-scale struc-
ture using deterministic data
•Defining the small-scale structure using
statistical techniques—geostatistics
•Rescaling the detailed geologic model to
be suitable input for a fluid-flow simulator.
Cray is a mark of Cray Research Inc. MicroVAX is a mark
of Digital Equipment Corp.
1. Weber KJ: “How Heterogeneity Affects Oil Recovery,”
in Lake LW and Carroll HB Jr (eds): Reservoir Charac-
terization. Orlando, Florida, USA: Academic Press
Inc. (1986): 487-544.
Defining the Large-Scale Structure
The first stage aims to construct as detailed a
geologic description of the reservoir as mea-
surements will allow. The available informa-
tion includes a structural interpretation from
seismic data, static data from cores and logs
and dynamic production data from testing.
To these, the geologist adds expert knowl-
edge on geologic structures, obtained from
field experience and outcrop studies (see
“Gathering Outcrop Data,” page 28).
The first step is to recognize depositional
units and then, if possible, correlate them
between wells. It helps to know the general
shape and order of deposition of the units,
and ideally the relationships between width,
thickness and length for each unit.1 Well-to-
well correlation is made using traditional
hard-copy maps and logs or, today, interac-
tive computer workstations. Workstations
allowing simultaneous visualization of seis-
mic, log and other data enable easier recog-
nition of subtle correlations.
In most cases, this large-scale characteri-
zation results in a conventional layer-cake
model—a stack of subhomogeneous layers.
There is no doubt that many reservoirs
should be characterized by more complex
arrangements of depositional units—such as
the “jigsaw” and “labyrinth” models (next
page, bottom ). 2 The ability to use these
more complex models routinely remains an
industry goal (see “Trends in Reservoir Man-
agement,” page 8 ).
2. Weber KJ and van Geuns LC: “Framework For Con-
structing Clastic Reservoir Simulation Models,” Jour-
nal of Petroleum Technology 42 (October 1990):
1248-1253, 1296-1297.
25
Hypothetical well data
Heterogeneity case A
Heterogeneity case B
Lithology
1 0.5
PV recovered
Defining the Small-Scale Structure Using
Geostatistical Modeling
Once the shape of the reservoir and its
large-scale structure have been described,
the next stage focuses on defining hetero-
geneity within each depositional unit. This
requires more detailed interwell measure-
ments than are currently available. Borehole
seismic data typically cannot span wells.
Although surface and cross-well seismic
data do reach everywhere, there is insuffi-
cient knowledge relating seismic data to
physical rock properties, and exploration
seismic data have insufficient resolution. In
addition, well testing data often lack direc-
tional information. Therefore nondetermin-
istic, or geostatistical, methods are required.3
Large-scale reservoir models typically
comprise a few thousand grid blocks, each
measuring about 100 m square and more
than 1-m thick [about 1000 ft2 by 3 ft] But
models that take into account small-scale
Predicted oil recovery
heterogeneity use smaller grid blocks—1
million or more are often needed. The sheer
quantity of input data required to fill so

2 0.4 Case B
Case A many blocks also favors geostatistics.
3 0.3 Geostatistical techniques designed to pro-
4 0.2 vide missing data can be classified in two
5 0.1 ways: by the allowable variation of the
Shale Barrier property at a given point and by the way
0
0 0.2 0.4 0.6 0.8 1.0 data are organized. There are two classes of
Pore volume (PV) injected property variation, continuous and discrete.
Continuous models are suited to properties
nHow heterogeneity affects productivity. In this example developed at BP Research, like permeability, porosity, residual satura-
Sunbury-on-Thames, England, two different models have been interpolated between a tion and seismic velocity that can take any
pair of hypothetical wells. The lithologies are identical: only their spatial correlation
has been altered. The graph shows how the pore volume recovered versus pore volume value. Discrete models are suited to geo-
injected varies for the two models. After one pore volume has been injected, the differ- logic properties, like lithology that can be
ence in recovery between the two models is about 0.1 of a pore volume. represented by one of a few possibilities.4
Data organization has developed along
two paths depending on how the models
are built up. One is grid based. All the
Barrier bar Tidal channel Crevasse splay properties are represented as numbers on a
Transgressive deposit Barrier bar
Barrier foot Distributary channel fill grid, which is then used for fluid-flow simu-
lation. The second is object based. Reser-
voir features such as shales or sands are
generated in space and a grid then superim-
posed on them.

Grid-Based Modeling
One of the key geostatistical tools used in
Layer cake Jigsaw Labyrinth
grid-based modeling is kriging—named after
a pioneer of rock-property estimation, D. G.
nDefining reservoir type. Current large-scale characterization tends to result in a sim- Krige, who made a series of empirical stud-
ple layer-cake model. However, efforts are being made to use more complex models ies in the South African goldfields.5 Kriging
like the jigsaw and labyrinth models shown. is designed to honor measured data and
produce spatially correlated heterogeneity.
The heart of the technique is a two-point
statistical function called a variogram that
describes the increasing difference (or
decreasing correlation) between sample val-
ues as separation between them increases.

26 Oilfield Review
 Core plugs Whole core Well logs Well testing
Kriging minimizes the error between the
interpolated value and the actual (but
unknown) value of the property.6
For a given property, a variogram is con-
structed from pairs of data generally mea-
sured in wells. In the oilfield, criticism of
kriging centers on the low density of well
data when compared to mining. To over- Borehole geophysics Outcrop studies Surface seismics
come this problem, variograms can alterna-
tively be constructed from data measured
on outcrops or in mature fields where better
sampling is available.
Although variograms are commonly used
to infer the spatial continuity of a single
variable, the same approach can be used to
study the cross-continuity of several differ- Geologist’s
ent variables—for example, the porosity at expert knowledge
one location can be compared to seismic
transit time. Once constructed, this cross-
correlation can be used in a multivariate
regression known as cokriging. In this way,
a fieldwide map of porosity can be com-
puted using not only porosity data, but also
the more abundant seismic data.7
Kriging and cokriging deal with quantita-
tive, or hard, data. A third method called Stage 1: Defining large-scale structure
soft kriging combines expert information,
also called soft data, with the quantitative
data. The expert data are encoded in the
form of inequalities or probability distribu-
tions. For example, in mapping a gas/oil
contact (GOC), if the contact is not reached
in a certain well, soft kriging uses the
inequality: GOC is greater than well total Stage 2: Defining small-scale structure
depth. At any given point where there is no
well, an expert may define the probability of
finding the GOC within a certain depth
interval. Soft kriging will use this probability.8
(continued on page 30 )

3. Haldorsen HH and Damsleth E: “Stochastic model-

ing,” Journal of Petroleum Technology 42 (April
1990): 404 -412. Haldorsen HH and Damsleth E:
Stage 3: Scaling up
“Challenges in Reservoir Characterization Research,”
presented at Advances in Reservoir Technology, Char-
acterization, Modelling & Management, organized by
the Petroleum Science and Technology Institute, Edin-
burgh, Scotland, February 21-22, 1991.
4. Journel AG and Alabert GF: “New Method For Reser-
voir Mapping,” Journal of Petroleum Technology 42 30,000 1 million
(February 1990): 212-218. grid blocks grid blocks
5. Krige DG: “A Statistical Analysis of Some of the Bore-
hole Values of the Orange Free State Goldfield,” Jour-
nal of the Chemical, Metallurgical and Mining Society
of South Africa 53 (1952): 47-70. Fluid-flow simulation
6. Journel AG: “Geostatistics for Reservoir Characteriza-
outflow face
inflow face

tion,” paper SPE 20750, presented at the 65th SPE

Annual Technical Conference and Exhibition, New
Orleans, Louisiana, USA, September 23-26, 1990.
7. Doyen PM: “Porosity From Seismic Data: A Geostatis-
tical Approach,” Geophysics 53 (October 1988):
1263-1275.
8. Kostov C and Journel AG: “Coding and Extrapolating
Expert Information for Reservoir Description,” in Lake
LW and Carroll HB (eds): Reservoir Characterization. nBuilding a reservoir model in three stages with measured data, expert knowledge
Orlando, Florida, USA: Academic Press Inc. (1986): and statistics.
249-264.

January 1992 27
Gathering Outcrop Data
Every reservoir is unique. Yet recurring deposi-
tional conditions create families of reservoirs
that are broadly similar. If reservoir rock outcrops
at the surface, it presents a golden opportunity to
gather data which can be used to help character-
ize related subsurface formations.1
Outcrop studies allow sampling at scales that
match interwell spacing. Small-scale permeabil-
ity can also be measured and compared with
depositional characteristics. Then the statistical
and depositional information measured at the
outcrop can be incorporated into the stochastic
modeling of the subsurface analog.2
Generally, there are two kinds of outcrop
study.3 Bed or bed-set scale genetic models
(GEMs) concentrate on reservoir heterogeneity as
a function of the sedimentary process. To date,
most outcrop studies have been of this type. The
second type, a field analog model (FAM), looks at
heterogeneity caused not only by depositional
processes but also by diagenesis. FAMs present
two difficulties. First, it is hard, given the paucity
of reservoir data, to decide which diagenetic pro-
cesses are relevant. Second, finding outcrops
that have undergone specific diagenetic pro-
cesses can be difficult.
Although the study of outcrops is a traditional
skill learned by most geologists, in the past they
have concentrated on such goals as describing
the sedimentological environment and the
geometries of the major depositional units. For
reservoir characterization, a much more detailed
and quantitative picture of the heterogeneity
within depositional sequences is required.
Therefore, in addition to the geologist’s tradi-
tional mapping skills, new analytical techniques
have been devised. Analyzing photographs of the
outcrop helps clarify the lithofacies distribution.4
Clearly, more than one photograph is needed,
and this technique can succeed only if all the
photographs are at the same scale and taken at
the same distance from the rock. A helicopter-
mounted camera with a laser range finder is one
way of achieving this. Successive photographs
are taken with a 30 to 40% overlap and then com-
bined to construct a photomosaic of the whole
28
nPhotomosaic of the Tensleep Sandstone in the Big Horn basin, Wyoming, USA. This
Pennsylvanian-age sandstone is being studied because of its similarity to the Rotliegen-
des Sandstone, an important gas reservoir unit found in the Southern Basin of the North
Sea. Like most outcrops, it is irregularly shaped. Therefore, to obtain the mosaic with
constant scale and no distortion, overlapping photographs were taken from points in a
plane parallel to a major depositional surface on the outcrop. With a helicopter posi-
tioned in this plane, laser range finders were used to maintain a fixed distance from
the outcrop to ensure constant scale. This was double-checked by positioning scale
bars at regular intervals on the outcrop. The mosaic was prepared during studies by
Jon Lewis and Kjell Rosvoll at Imperial College, University of London, England and
Heriot-Watt University, Edinburgh, Scotland with the support of Conoco (UK) Ltd.
outcrop (above). The mosaic can be digitized to probe permeameters (next page, below).6 These
quantify lithofacies patterns. It can also be used evaluate permeability—from 0.5 millidarcies
to plan and execute small-scale petrophysical (md) to 15 darcies—by injecting nitrogen into a
measurements on the outcrop. prepared location on the outcrop. Thousands of
Gamma ray logging of outcrops is used to help nondestructive measurements can be made in
describe the composition of the lithofacies. A situ and the results compared to traditional labo-
standard truck-mounted sonde can be lowered ratory tests for quality control. This permits
down the face of an outcrop and gamma ray mea- examination of permeability changes on a very
surements continuously recorded. Although small-scale—on the order of millimeters.
washouts are known to affect borehole gamma Through such measurements, permeability has
ray logging, tests show that the sonde can be as been shown to vary dramatically within a deposi-
much as 0.6 m [2 ft] away from the rock before tional unit.7
the reading is compromised.5 A typical study, carried out by the University of
An alternative technique employs a lightweight, Texas, Bureau of Economic Geology, Austin,
portable gamma ray spectrometer to measure Texas, USA, centers on a Ferron Sandstone out-
either total radiation or individual radioelement crop in central Utah, USA. Ferron is a wave-mod-
concentrations. Total radiation measurements ified, deltaic sandstone with permeability distri-
are typically made at 0.65-m [2.1-ft] intervals. bution strongly related to lithofacies type and
Selected intervals may be logged in greater grain size. It is believed to be an analog for Gulf
detail if required. Because the tool measures an Coast deltaic reservoirs, which account for 64%
area of the rock face with a diameter of about 0.3 of Texas Gulf Coast gas production.8
m [1 ft], readings taken for thin beds will be First, the depositional framework of the out-
slightly influenced by adjacent strata. crop was established using color photomosaics to
One of the principal pitfalls in reservoir charac- map the distribution and interrelations of the
terization stems from the use of unrepresentative sandstone components. This involves determin-
permeability data measured from cores and core ing lithofacies architecture, delineating perme-
plugs taken from only a few wells. Densely sam- able zones and their continuity, identifying flow
pled outcrop permeability data are of key impor- barriers and baffles and establishing permeabil-
tance. These can be gathered by performing labo- ity trends.
ratory-based flow tests on samples gathered in On the basis of this framework, more than 4000
the field. However, a comparatively recent devel- minipermeameter readings were made, all on
opment has seen the introduction of portable
electronic minipermeameters—also known as
Oilfield Review
vertical exposures of rock to minimize the effects
of weathering. Two sampling strategies were
employed. First, more than 100 vertical strips
were mapped every 30 to 60 m [100 to 200 ft] on
three channel complexes. Permeability measure-
ments were taken at 0.3-m spacing on each strip.
Second, two sampling grids were constructed
to examine the small-scale spatial variability of
permeability within lithofacies. One was 12-m
[40-ft] square with 0.6-m subdivisions. The sec-
ond, inside the first, measured 1.8-m [6-ft]
square with 0.1-m [0.3 ft] subdivisions.
The study successfully identified patterns of
heterogeneity. Delta-front sandstones were found
to be vertically heterogeneous but laterally con-
tinuous over a typical well spacing. However, the
sand belts of the distribution system—composed
of amalgamated, lateral accretion point-bar
sandstones—were found to have both lateral and
vertical heterogeneity.
Gas supply R1
(nitrogen cylinder)
Four way valve
Flow units
Injection pressure
measurement system
January 1992
Variogram analysis confirmed that the hetero-
geneity of the permeability was structured in a
way that depended on the rock type sampled and
the sampling scale. For instance, at sampling
intervals of 0.6 m, the range over which perme-
ability shows correlation was 5.5 m [18 ft] for
tough-cross-bedded sandstone and 1.2 m [4 ft]
for contorted beds. Planar cross-bedded strata
exhibited a range of 3.6 m [12 ft] for a 0.6 m
sampling interval and 0.6 m for the 0.1 m inter- nGoing underground—outcrops are not
always on the surface. A three-dimen-
val. According to the University of Texas sional (3D) study of permeability varia-
researchers, this suggests that the permeability tion was undertaken in a subsurface sil-
ica sand mine in Morvern, Scotland. Some
structure may be fractal. 8000 permeability measurements have
This is just one example of work underway. been collected over a 1-km [0.6-mile]
square section of the mine. The work was
Over the past five years, there has been an explo-
carried out by Jon Lewis and Ben Lowden
sion of outcrop studies—a hundred or more liter- at Imperial College, University of London,
ature references (above, right). And many pro- England, and supported by Den Norske
Stats Oljeselskap A.S. (Statoil).
jects are afoot to bring these studies together and
create outcrop data bases. Key information
needed to condition geostatistical modeling will
become readily available.
nMinipermeame- 1. Weber KJ and van Geuns LC: “Framework for Construct-
R2 ter in detail. The ing Clastic Reservoir Simulation Models,” Journal of
probe is applied to Petroleum Technology (October 1990): 1248-1253, 1296-
the cleaned sur- 1297.
Four way valve face of the rock 2. North CP: “Inter-Well Geological Modelling of Continental
with enough pres- Sediments: Lessons From the Outcrop,” presented at
sure to ensure a Advances in Reservoir Technology, Characterisation,
good seal. Nitro- Modelling & Management, organized by the Petroleum
gen is injected into Science and Technology Institute, Edinburgh, Scotland,
the rock and the February 21-22, 1991.
pressure and flow 3. Lewis JJM: “A Methodology for the Development of Spa-
rate measured. tial Reservoir Parameter Databases at Outcrop,” pre-
sented at Minipermeametry in Reservoir Studies, orga-
These values,
nized by the Petroleum Science and Technology Institute,
along with the Edinburgh, Scotland, June 27, 1991.
area of the probe,
4. A lithofacies is a mappable subdivision of a stratigraphic
are then used to unit distinguished by its lithology.
Four way valve calculate perme-
5. Jordan DW, Slatt RM, D’Agostino A and Gillespie RH:
Gas injection probe ability. To measure “Outcrop Gamma Ray Logging: Truck-Mounted and
the range of per- Hand-Held Scintillometer Methods Are Useful for Explo-
meabilities found ration, Development, and Training Purposes,” paper SPE
Rock
at the outcrop, four 22747, presented at the 66th SPE Annual Technical Con-
different flow ele- ference and Exhibition, Dallas, Texas, USA, October 6-9,
ments are required. 1991.
Pressure
transducer (0-20psi) 6. Lewis JJM: “Outcrop-Derived Quantitative Models of Per-
meability Heterogeneity for Genetically Different Sand
Bodies,” paper SPE 18153, presented at the 63rd SPE
Annual Technical Conference and Exhibition, Houston,
Texas, USA, October 2-5, 1988.
Jensen JL and Corbett PWM: “A Stochastic Model for
Comparing Probe Permeameter and Core Plug Measure-
ments,” paper 3RC-24, presented at the 3rd International
Reservoir Characterization Technical Conference of the
National Institute for Petroleum and Energy Research and
the US Department of Energy, Tulsa, Oklahoma, USA,
November 3-5, 1991.
7. Kittridge MG, Lake LW, Lucia FJ and Fogg GE:
“Outcrop/Subsurface Comparisons of Heterogeneity in
the San Andres Formation,” SPE Formation Evaluation 5
(September 1990): 233-240.
8. Tyler N, Barton MD and Finley RJ: “Outcrop Characteriza-
tion of Flow Unit and Seal Properties and Geometries,
Ferron Sandstone, Utah,” paper SPE 22670, presented at
the 66th SPE Annual Technical Conference and Exhibition,
Dallas, Texas, USA, October 6-9, 1991.
 These kriging methods yield smooth inter- Measured data
polations, but do not describe small-scale
heterogeneity. In a process called stochastic Simulation 1
modeling that superimposes correlated

Value
Kriged
noise onto smooth interpolations, more real- interpolation
istic pictures emerge. A probability distribu- Actual data
tion determines how this noise is generated. Simulation 2
A number of pictures, or realizations, will
usually be created, each with different noise
sampled from the same distribution (top, Distance
right ). By analyzing many realizations, the nImitating reality’s variability. The kriging technique yields a
extent to which geologic uncertainty affects smooth interpolation, ignoring nature’s true small-scale het-
reservoir performance can be studied (mid- erogeneity. By adding noise to a kriged interpolation, stochas-
tic modeling produces more lifelike realizations.
dle, right ).
An example is sequential Gaussian simu-
Realization A
lation (SGS). The data are constructed grid 0 nTwo stochastic
block by grid block. At the first selected realizations of reser-
block, the smooth interpolated value is cal- 20 voir permeability
culated by kriging using the available mea- showing how differ-
40 ent noise added to
sured data. The interpolated value and its
the same kriged
variance, also calculated by kriging, define 60 interpolation gives
Depth, ft

a Gaussian distribution function from which quite different real-

80 izations. In both the
a noise value is randomly drawn and added Realization B
to the interpolated value. For the next point, 0
horizontal and ver-
tical directions,
selected at random, the process is repeated, realization A has
20
using as a base this newly derived data relatively high con-
point together with the measured data. As 40 tinuity whereas
the grid blocks are filled, all previously cal- realization B has
60 relatively low conti-
culated values contribute to computing the
nuity. Continuity is
next in the sequence. 80 gauged by measur-
The technique depends on the reservoir 0 200 400 500 600 800 1000 1200 1400 ing the size of groups
property being normally distributed, so a Distance, ft of contiguous blocks
property like permeability, which has a Above 100 md 1.0–0.1 md with permeability of
100 millidarcies
skewed distribution, requires transformation (md) or more.
to normality. The reverse process has to be 100–0 md Below 0.1 md
performed once all the grid blocks are filled.
10–1 md
An increasingly popular method for gen-
erating noise for stochastic modeling uses After Fogg GE, Lucia FJ and Senger RK: “Stochastic Simulation of Interwell-Scale Het-
the concept of fractals—a statistical tech- erogeneity for Improved Prediction of Sweep Efficiency in a Carbonate Reservoir,” in
Lake LW, Carroll HB Jr and Wesson TC (eds): Reservoir Characterization II. San Diego,
nique that produces remarkably realistic California, USA: Academic Press Inc. (1991): 355-381.
imitations of nature. Fractal objects exhibit
similar variations at all scales of observa-
tion. Every attempt to divide fractal objects
into smaller regions results in ever more
similarly-structured detail (right ).9 This sim-
plifies stochastic modeling. The variogram is nThe Sierpinski Gasket. This self-simi-
lar fractal structure was devised by the
defined from a single number—the fractal Polish mathematician W. Sierpinski
dimension, calculated from measured data about 90 years ago. It is formed by
in the reservoir or outcrops. And because dividing the largest triangle into
fractals are self-similar, the variance of the smaller triangles with sides half as
noise need be determined only at a single long as the original. In three of the
resulting four triangles the process is
scale. Fractal modeling has been used to repeated. The process is repeated at
predict the production performance of sev- progressively finer scales. At every
scale the same patterns can be found.
Random fractals generated by a simi-
lar process, but with added stochastic
variations, produce remarkably realis-
tic simulations of geologic variability.

30 Oilfield Review
eral large fields. Recent studies have also
applied fractal models to investigate the per-
formance of miscible gas floods.10
Discrete data require a special interpola-
tion technique called indicator kriging. This
can be combined with stochastic modeling
to form a process called sequential indicator
simulation (SIS). Take a model where either
sand or shale must be chosen—there is no
middle ground. At the same time, a random
element must be introduced to obtain multi-
ple realizations. Consider building a model
where sand is 1 and shale 0. Indicator krig-
ing will assign every grid block some value
between 0 and 1—giving an indication of
the likely lithology. For example, a value of
0.7 indicates a 70% chance of sand. In the
stochastic stage, this percentage is then used
to weight the random choice of sand or
shale for the grid block.11
Object-Based Modeling
In object-based modeling, a picture of the
reservoir model is built from generic
units—for example, sand bodies or shale
barriers, each unit having uniform internal
properties. These models are built by start-
ing with a uniform background matrix and
adding units with a contrasting property:
either a high permeability background to
which shale barriers are added, or a low
permeability background to which sand
bodies are added.
When starting with a high-permeability
background, the vertical distribution of
nStochastic shales in object-based modeling. The location of
each stochastic shale is random and independent of other
shales. The units are drawn randomly from a size distribution
function and placed at random locations until the required shale
density has been achieved.
January 1992
shales may be inferred from cores and logs,
particularly gamma ray logs. But unless the
well spacing is extremely dense, nothing is
revealed about the shales’ lateral dimen-
sions. Some shales correlate from well to
well, but most do not, and their lateral extent
must be generated statistically.
These stochastic shales are drawn at ran-
dom from a size distribution and placed
randomly until the precalculated shale den-
sity has been achieved. Shale density is esti-
mated from the cumulative feet of shale
measured in wells compared with the gross
pay, and then assumed to represent the
reservoir volume under study. To ensure that
the realizations honor the deterministic
data, shales observed in wells are placed in
the model first, their lateral extent still deter-
mined randomly. Subsequent, randomly
placed shales that intersect a well are
rejected (below ).12
To generate these models, the key statis-
tics specified by the geologist are shale
length and width, together with some guid-
ance as to their interdependence. Tradition-
ally, the geologist describes shales qualita-
tively as wide, extensive or lenticular—too
vague for building realistic stochastic mod-
els. Instead, ranges of length and width are
required, along with gross pay thickness and
average shale density. The depositional
environment has great bearing on these val-
ues and the primary sources of this informa-
tion are outcrop studies.13
9. Hewett TA: “Fractal Distributions of Reservoir Het-
erogeneity and Their Influence on Fluid Transport,”
paper SPE 15386, presented at the 61st SPE Annual
Technical Conference and Exhibition, New Orleans,
Louisiana, USA, October 5-8, 1986.
Hewett TA and Behrens RA: “Conditional Simulation
of Reservoir Heterogeneity With Fractals,” paper SPE
18326, presented at the 63rd SPE Annual Technical
Conference and Exhibition, Houston, Texas, USA,
October 2-5, 1988.
“Fractals and Rocks,” The Technical Review 36, no.
1 (January 1988): 32-36.
Crane SD and Tubman KM: “Reservoir Variability
and Modeling With Fractals,” paper SPE 20606, pre-
sented at the 65th SPE Annual Technical Conference
and Exhibition, New Orleans, Louisiana, USA,
September 23-26, 1990.
10. Perez G and Chopra AK: “Evaluation of Fractal Mod-
els to Describe Reservoir Heterogeneity and Perfor-
mance,” paper SPE 22694, presented at the 66th SPE
Annual Technical Conference and Exhibition, Dal-
las, Texas, USA, October 6-9, 1991.
Payne DV, Edwards KA and Emanuel AS: “Examples
of Reservoir Simulation Studies Utilizing Geostatisti-
cal Models of Reservoir Heterogeneity,” in Lake LW,
Carroll HB Jr and Wesson TC (eds): Reservoir Char-
acterization II. San Diego, California, USA: Aca-
demic Press Inc. (1991): 497-523.
11. Journel AG and Alabert FG: “Focusing on Spatial
Connectivity of Extreme-Valued Attributes: Stochas-
tic Indicator Models of Reservoir Heterogeneities,”
paper SPE 18324, presented at the 63rd SPE Annual
Technical Conference and Exhibition, Houston,
Texas, USA, October 2-5, 1988.
12. Haldorsen HH and Chang DM: “Notes on Stochastic
Shales; From Outcrop to Simulation Model,” in Lake
LW and Carroll HB Jr (eds): Reservoir Characteriza-
tion. Orlando, Florida, USA: Academic Press Inc.
(1986): 445-486.
13. Geehan GW, Lawton TF, Sakurai S, Klob H, Clifton
TR, Inman KF and Nitzberg KE: “Geologic Predic-
tion of Shale Continuity, Prudhoe Bay Field,” in Lake
LW and Carroll HB Jr (eds): Reservoir Characteriza-
tion. Orlando, Florida, USA: Academic Press Inc.
(1986): 63-82.
31

nBuilding an object-based model using SIRCH, a software pack-
age developed at BP Exploration. In this model, two different
types of fluvial channel belt have been generated in a three-
stage process. First, one type of channel belt is generated (yel-
low). Then a second set (purple) is added to the first. This second
type has associated overbank deposits (red)—which are thinner,
poorer quality sands that nevertheless improve connectivity.
Finally, the combined channel belts are faulted.
Each channel belt is created using about 50 properties. Some
are constants, some are picked at random from input distribu-
tions and some are calculated from other properties. Properties
that determine channel shape include width, thickness, reach
length and angle, azimuth and depth. Flow properties, like per-
meability and porosity, are assigned either by lithology or by
using a random sample. All well data are honored. Net-to-gross
ratio is used to control the quantity of sand generated.
32
In theory, a similar method could be
applied to generate sandstone bodies in a
low-permeability background. However, BP
Research, Sunbury-on-Thames, England, has
developed SIRCH—system for integrated
reservoir characterization—to generate real-
istic reservoir heterogeneities using a library
of depositional shapes. Each shape has a set
of attributes defined by sedimentology and
based on observations of depositional sys-
tems. These attributes determine the size
and position in the reservoir model of each
shape and its relations with other shapes.
For example, channel belts are con-
structed as continuous links of a specified
shape.14 As a belt is constructed, input val-
ues for the attributes are repeatedly sampled
from the library and the generated belts may
be conditioned to honor well data. The
model can also be faulted, structured to
account for regional dip and clipped back
using the mapped top and bottom of the
reservoir to give it a volume indicated by
structural maps (left ).
SIRCH is still stochastic, so several real-
izations are required to estimate reservoir
characteristics. The realizations can be used
to estimate the volume of sandstone within
a specified interval, reservoir connectivity
and hydrocarbons in place for a given
radius around a well. This process has
proved valuable in indicating sensitivity of
results to the input parameters. Influential
inputs can then be targeted for extra study.
Individually, all these techniques paint a
portion of the total picture. In most cases,
gaining a full view of the reservoir requires a
hybrid (see “The HERESIM Approach to
Reservoir Characterization,” next page ).
Object-based, discrete modeling may be
used to describe the large-scale hetero-
geneities in the reservoir—the sedimento-
logical units—while different continuous
models may describe the spatial variations
of properties within each unit. For example,
for fluvial environments, BP Research is
planning to use SIRCH to generate chan-
nels, SIS for the facies within them and SGS
for permeability within the facies.
(continued on page 36)
14. Channels are the depositing rivers; channel belts are
the resulting sandstone bodies.
Oilfield Review
The HERESIM Approach to Reservoir Characterization

An example of a software package that combines Geostatistical

traditional geologic analysis with geostatistical analysis Simulation of Simulation of Scaling up of
techniques is HERESIM, developed jointly by the • proportion lithofacies petrophysical petrophysical
curves data data
Institute Français du Pétrole, Rueil-Malmaison,
• variograms
France, and the Centre de Géostatistiques, Ecole
Nationale Supérieure des Mines de Paris, Foun-
tainebleau, France. This allows sedimentological Stratigraphic studies
and geostatistical structural analysis to be car- Fluid-flow modeling
Enhanced oil recovery
ried out on interactive workstations, followed by
studies
geostatistical conditional simulations using Reinterpretation of the results in
terms of sequential stratigraphy
small-scale grid blocks. Each simulation creates
an image of the reservoir geology, consistent
with well data. Petrophysical data, like perme- Connectivity studies
ability and porosity, are then mapped between Optimization of well spacing
wells using deterministic or stochastic Swept volume estimation
algorithms. Finally, the petrophysical information
nOrganization and use of HERESIM, a geostatistical reservoir
is scaled up for fluid-flow simulation (right).1 characterization package.
First, a sedimentological study is carried out to
subdivide the area under investigation, which can
▲▲▲▲
▲ ▲▲▲▲ ▲ ▲▲▲ ▲
▲▲▲ ▲ ▲Wells ▲ Variogram
include formations bordering the reservoir as 100%
well as the reservoir itself. The largest divisions
Sill
are depositional units that include sets of geneti-
Frequency

cally linked strata, bounded by major sedimen-

tary discontinuities or sudden variations in the

sedimentological environment. Distinction of
units is usually achieved by correlating features
between wells. The next step is to identify litho-
facies within each depositional unit, collect sta-
tistical information about them at the wells and
then generate stochastic realizations of lithofa-
cies between wells.
Statistical information obtained from well data,
seismic interpretation and geologic knowledge are:
• Proportion curves—the percentage occurrence
of each lithofacies in a depositional unit (right)
• Experimental variograms—to quantify the spa-
tial continuity of each lithofacies in the reser-
voir (far right).2
With this information and a form of indicator krig-
ing employing a Gaussian random function,
0
-1000 0 1000 2000
Relative distance
Lithofacies
1 2 3 4
nHorizontal proportion curve. Proportion
curves provide information on the relative
frequency of each lithofacies in a deposi-
tional unit. Here, four lithofacies have been
proportioned using data from 20 wells.
numerous realizations of lithofacies distribution
can be readily obtained for each unit. However,
the algorithm works only in rectangular formats,
so it is necessary to first geometrically transform
each unit into a rectangle. After stochastic real-
izations have been generated, the real shape of
the unit is restored. HERESIM offers two types of
transform, depending on whether the prevailing
environment since deposition has been domi-
nated by erosion or differential subsidence—
Value
Range Distance
nVariograms quantify the spatial continu-
ity of each lithofacies in the reservoir. The
sill equals the variance of data; while the
range indicates the separation beyond
which two points are uncorrelated.
1. Eschard R, Doligez B, Rahon D, Ravenne C and Leloch G:
“A New Approach for Reservoirs Description and Simula-
tion Using Geostatistical Methods,” presented at
Advances in Reservoir Technology, Characterization,
Modelling & Management, organized by the Petroleum
Science and Technology Institute, Edinburgh, Scotland,
February 21-22, 1991.
2. Matheron G, Beucher H, de Fouquet C, Galli A, Guérillot D
and Ravenne C: “Conditional Simulation of the Geometry
of Fluvio-Deltaic Reservoirs,” paper SPE 16753, pre-
sented at the 62nd SPE Annual Technical Conference and
Exhibition, Dallas, Texas, USA, September 27-30, 1987.

January 1992 33
Erosion transformation Subsidence transformation

Fill in eroded strata to create rectangle, then perform stochastic

realization.
Reverse effect of subsidence to create rectangle,then perform
stochastic realization.

Erode stochastic realization to recreate structure.

Subside stochastic realization to recreate structure.

nA step in HERESIM processing: creating rectangles to allow generation of stochastic realizations. Before the stochastic realization
can be generated, the depositional units have to be transformed into rectangles. HERESIM has two ways of doing this. One assumes
that the unit has been eroded and that all the correlation lines within the unit are parallel. To create a rectangle, these correlations
are extrapolated across the eroded sections of the unit. Then the rest of the structure can be stochastically generated and the result-
ing realization re-eroded to recreate the original structure. The alternative technique assumes that differential subsidence has
occurred, causing the correlation lines to diverge. To create a rectangle, the effects of the subsidence are reversed. Then the stochas-
tic realization is created and resubsided.

34 Oilfield Review
nStochastic realization generated using HERESIM.
although some forms of fracturing can also be
accommodated (previous page).
Having selected the most likely realizations of
lithofacies, geologists and reservoir engineers
assign porosity and permeability values.
HERESIM assumes that petrophysical data are
strongly related to lithofacies. For each lithofa-
cies, either constant data values can be assigned
or a Monte Carlo-type distribution employed.
The geologic model describes a reservoir at a
decametric scale (on average 50-m [165 ft] square
or smaller with a thickness of 0.5 m [1.5 ft]). As a
result, simulated grids are huge (millions of
cells), too big for fluid-flow simulation (above).
Scaling up in HERESIM is performed using a fast
algorithm that calculates absolute permeabili-
January 1992
ties.3 First, average permeability is estimated
from the root of the product of the arithmetic mean
and harmonic mean of the grid-block permeabili-
ties. The quality of this value can be judged by
computing the log of the difference between the
two means. The larger the difference, the less
satisfactory the simple average. Grid blocks that
fail this test are subjected to a more accurate and
time-consuming fluid-flow simulation.
3. Guérillot D, Rudkiewicz JL, Ravenne C, Renard G and
Galli A: “An Integrated Model for Computer Aided Reser-
voir Description: From Outcrop Study to Fluid Flow Simu-
lations,” presented at the 5th European Symposium on
Improved Oil Recovery, Budapest, Hungary, April 25-27,
1989. Reprinted in Revue de l’Institut Français du Pétrole
45, no. 1 (January-February 1990): 71-77.
35
Rescaling the Geologic Model For Fluid-
Flow Simulation
Geologic modeling can describe a reservoir
with millions of grid blocks. For the com-
puter, this mainly presents problems of stor-
age. But the next major operation is fluid-
flow simulation, which involves complex
numerical calculation—a monumentally
large task given a million grid blocks (see
“Simulating Fluid Flow,” page 38 ).
Today, fluid-flow simulators can cope
with up to about 50,000 grid blocks—the
nModeling the North Sea Rannoch forma-
tion. This reservoir sandstone is character-
ized by hummocky cross-stratified bed-
forms. The Department of Petroleum
Engineering at Heriot-Watt University
determined the flow performance of this
depositional structure using the two deter-
ministic simulations shown here. These
two images represent a slice of rock mea-
suring 2.4 m [7.9 ft] and 14.4 m [47.2 ft].
In the first (top), saturation is at initial
conditions. Low permeability, rippled
crests have the lowest oil saturations
(green). The second (bottom) shows satu-
ration distribution after 0.2 pore volumes
of water have been injected at a low
advance rate of 0.25 m/d. Saturation is
reduced most rapidly in the low perme-
ability, rippled crests (dark blue).
36
greater the number of grid blocks, the more
expensive the fluid-flow simulation is in
computation. So before the simulation is
run, the geologic model, with its high-reso-
lution petrophysical data must be enshrined
in a smaller number of larger grid blocks—a
process called scaling up. Reorganizing so
much data may introduce error, masking
heterogeneities in the geologic model that
directly affect simulated production.
The key to achieving scale up is deriving
a single set of reservoir properties—called
pseudos—for the larger blocks so that the
fluid-flow simulation after scale up is as
close as possible to what it would have
Recovery of original oil-in-place
been with the small-scale data.
Over the years, a variety of methods has
been developed to do this. These include
calculations based on simple averaging
procedures like the oft-quoted Kyte and
Berry method and more complex simula-
tions of parts of the reservoir using small-
scale grid blocks, measuring about 1 m
square and 0.5 m thick.15
The impact of small-scale effects on two-
phase flow performance is revealed in a
study of the North Sea Rannoch formation,
Recovery of original oil-in-place
carried out at Heriot-Watt University, Edin-
burgh, Scotland. Rannoch subfacies are
strongly laminated and rippled. The study
utilized newly acquired minipermeameter
data at a millimetric scale, much denser
than that conventionally obtained from
core analysis.
These data were combined with rock cap-
illary pressure curves to define pseudos for
each lamina subfacies. A geologic model for
hummocky cross-stratification combined the
subfacies at the bedform scale (left ).16 The
simulated flow performance of this model
was then compared with simpler flow simu-
lations calculated using arithmetic and har-
monic average permeabilities for horizontal
and vertical directions, respectively.
The flow behavior of the detailed model
was anisotropic, whereas the simpler
approach yielded more isotropic behavior
and significantly higher recoveries for verti-
cal displacement (right ). So, small-scale
heterogeneity affects reservoir producibility,
and in many cases, variations in capillary
pressure with rock type and small-scale
sedimentary structure cannot be ignored.
A predominant theme in scaling up is the
identification of a number of discrete scales
of heterogeneity. To preserve the effects of
heterogeneity at all scales, a series of scale-
up operations can be performed, each deal-
ing with heterogeneity larger than the previ-
ous operation. At each stage in the process,
the size of the scale up should be as large as
possible without introducing the next level
of heterogeneity. If a reservoir is divided into
depositional units, lithofacies, beds and
small-scale heterogeneity within a bed, the
scale-up process will have four stages. For
example, lithofacies become the basic build-
ing blocks at the depositional unit scale.17
Horizontal Flood
80%
Detailed geologic model
Arithmetic average and rock
curves
40%
0
0 50% 100%
Total pore volume injected
Vertical Flood
80%
Detailed geologic model
Harmonic average and
rock curves
40%
0
0 50% 100%
Total pore volume injected
nRannoch formation’s flooding character-
istics. Using densely-sampled permeabil-
ity data, capillary pressures measured
from appropriate rock types and knowl-
edge of the depositional structure to con-
struct a detailed geologic model, the
Department of Petroleum Engineering at
Heriot-Watt University simulated both hor-
izontal and vertical flooding. The results
using the detailed geologic model are sig-
nificantly different from those derived
using traditional techniques employing
average permeabilities and rock curves—
particularly for the vertical flood.
Oilfield Review
 Heterogeneity can be assessed by measur-
ing the average value of a reservoir property
within a growing volume. When the volume
is very small, the average value will fluctuate
because of microscopic heterogeneities. But
as the averaging volume increases, more of
these heterogeneities will be encompassed
and the average value will stabilize. Further
growth may result in capturing larger-scale
heterogeneity and the value will again fluc-
tuate (right ).18 Since this behavior varies
among properties, the size of the smallest
grid block should be selected so that all key
properties are stable. 19
No matter how many scales are used,
existing methods of calculating pseudos
have their drawbacks. The elementary
methods, using averaged properties, are
quick but not accurate enough. The simula-
tion methods are computationally expen-
sive. And if sensitivity studies are required
using a number of stochastic realizations,
the time and cost of scaling up can be pro-
hibitive. Recently, BP Research developed a
new algorithm—real-space renormaliza-
tion—that promises speedy yet accurate cal-
culation of pseudos.
The method starts with a group of eight
grid blocks (2×2×2) with permeabilities dis-
tributed to represent the original data. A new
relative permeability is calculated and
adopted for a new single block the size of
the eight original blocks—it becomes the
starting point for the next renormalization
when the process is repeated with seven
other similarly generated blocks. The new
block’s effective relative permeability is cal-
culated using a semianalytical technique that
depends on an analogy between Darcy’s law
and Ohm’s law—block permeabilities are
modeled by an equivalent resistor network.
Through a series of transformations, the
resistor network is reduced to a single resis-
15. Kyte JR and Berry DW: “New Pseudo Functions to
Control Numerical Dispersion,” Society of
Petroleum Engineers Journal 15 (August 1975): 269-
276.
Tompang R and Kelkar BG: “Prediction of Water-
flood Performance in Stratified Reservoirs,” paper
SPE 17289, presented at the SPE Permian Basin Oil
and Gas Recovery Conference, Midland, Texas,
USA, March 10-11, 1988.
January 1992
Normalized average value
Microscopic Macroscopic Megascopic
Volume
Fluid density ( ρ)
nHow sample volume affects
property averages. At small sam-
ple volumes, property averages
fluctuate because of microscopic
heterogeneity. As sample volume
0 is increased, heterogeneities will
average out and properties stabi-
the range of REVρ lize—this gives the representative
Porosity (φ) elementary volume (REV). This sta-
bility continues until larger-scale
heterogeneity begins to take
effect. Different properties have
different volumes where hetero-
geneity is stable. The ideal sample
volume should fall within ranges
of stability for all key properties.
0
the range of REVφ
Permeability (k )
0
the range of REVk
the range of REV for all
properties
16. Corbett PWM and Jensen JL: “An Application of 19. Haldorsen HH: “Simulator Parameter Assignment
Small Scale Permeability Measurements—Prediction and the Problem of Scale in Reservoir Engineering,”
of Flow Performance in a Rannoch Facies, Lower in Lake LW and Carroll HB Jr (eds): Reservoir Char-
Brent Group, North Sea,” presented at Miniperme- acterization. Orlando, Florida, USA: Academic Press
ametry in Reservoir Studies, organized by the Inc. (1986): 293-340.
Petroleum Science and Technology Institute, Edin- Kossack CA, Aasen JO and Opdal ST: “Scaling Up
burgh, Scotland, June 27, 1991. Heterogeneities With Pseudofunctions,” SPE Forma-
17. Lasseter TJ, Waggoner JR and Lake LW: “Reservoir tion Evaluation 5 (September 1990): 226-232.
Heterogeneities and Their Influence on Ultimate Norris RJ and Lewis JJM: “The Geological Modeling
Recovery,” in Lake LW and Carroll HB Jr (eds): of Effective Permeability in Complex Heterolithic
Reservoir Characterization. Orlando, Florida, USA: Facies,” paper SPE 22692, presented at the 66th SPE
Academic Press Inc. (1986): 545-559. Annual Technical Conference and Exhibition, Dal-
18. Any volume for which the property is stable is called las, Texas, USA, October 6-9, 1991.
a representative elementary volume (REV).
37
 Simulating Fluid Flow
Today, most simulators solve fluid-flow problems
by segmenting a portion of the reservoir into a
series of grid blocks—either two- or three-dimen-
sional. The fluid phases in each block are mod-
eled with finite difference equations similar to
the conventional volumetric material balance
equations. Darcy’s law is then used to describe
fluid flow between grid blocks.
Early efforts in reservoir simulation, in the late
1940s, began with simple material balance cal-
culations. Throughout the 1960s, black oil simu-
lations dominated the scene. These describe
nonvolatile oil as a single component containing
gas in solution as a second component. There are
three compressible, immiscible phases: gas,
water and oil. Recovery methods simulated
included pressure depletion and some forms of
pressure maintenance.1
The advent of enhanced recovery techniques
such as chemical flooding, steamflooding and in-
situ combustion, ushered in more sophisticated
simulators. Most describe the hydrocarbon in
terms of a number of components and solve ther-
modynamic equilibrium equations to determine
the distribution of components between the liquid
and vapor phases in the reservoir. 2
A common strategy for fluid-flow simulation is
to use history matching to adjust the geologic
38
model. Adjustments should be consistent with
measured data and improve upon postulated vari-
ations in reservoir properties that cannot be or
have not been measured. The results of the simu-
lation can be extremely sensitive to variations in
the reservoir description. Models for reservoir
management are usually routinely updated as
new production data or reservoir information are
acquired. This allows for reexamination of cur-
rent and future production scenarios. But history
matching always results in nonunique combina-
tions of variables. And it could be said that the only
fully simulated reservoir is a fully depleted one.
Simulator technology is by no means mature.
Some current developments center on improving
the ability to cope with different production sce-
narios, for example, naturally fractured forma-
tions and horizontal wells.
During production by gravity drainage in natu-
rally fractured reservoirs, oil drains from matrix
blocks into the fracture system. From there it can
pass either into other blocks below or along frac-
tures. A fractured reservoir is modeled using two
interacting continua, one for the matrix and one
for the fractures—the so-called dual porosity
concept. The matrix contains most of the oil and
the fractures most of the conductivity. So trans-
missibility between matrix blocks is usually
ignored. But to correctly model ultimate recovery,
the capillary contacts between matrix blocks
should be considered.
Work at Koninklijke/Shell Exploratie en Produk-
tie Laboratorium, Rijswijk, The Netherlands, has
shown that a vertical stack of oil-saturated matrix
blocks surrounded by gas does not drain indepen-
dently.3 Oil from one block passing into the frac-
ture system will be absorbed by the underlying
block, slowing production. Therefore, a flow sim-
ulation that ignores this effect will overestimate
initial production rates. To combat this, Shell has
devised a simulator that allows flow in all direc-
tions between neighboring blocks.
Fluid-flow simulation for horizontal wells is
currently being approached in at least two ways.
The productivity of a horizontal well can be
approximated, using numerical solutions of sim-
plified equations, and the approximation used in
a conventional reservoir simulator. Alternatively,
for single wells in which short-term performance
is being studied, fine-scale, black oil simulations
have been used. In this approach, pseudos can
be derived for a coarser, field-sized simulation.
1. Mattax CC and Dalton RL: “Reservoir Simulation,” Jour-
nal of Petroleum Technology 42 (June 1990): 692-695.
Breitenbach EA: “Reservoir Simulation: State of the Art,”
Journal of Petroleum Technology 43 (September 1991):
1033-1036.
2. Cheshire IM and Pollard RK: “Advanced Numerical Tech-
niques for Reservoir Simulation and Their Use on Vector
and Parallel Processors,” in Edwards SF and King PR
(eds): Mathematics in Oil Production. Oxford Science
Publications, Oxford, England: Clarendon Press (1988):
253-268.
3. Por GJ, Boerrigter P, Maas JG and de Vries A: “A Frac-
tured Reservoir Simulator Capable of Modeling Block-
Block Interaction,” paper SPE 19807, presented at the
64th SPE Annual Technical Conference and Exhibition,
San Antonio, Texas, USA, October 8-11, 1990.
Oilfield Review
tance, equivalent to the new block’s effective
relative permeability.
The effect of each stage of renormaliza-
tion is to produce larger blocks whose per-
meability approaches that of the whole.
Results of renormalization have been com-
pared with direct numerical simulation and
the maximum error was found to be only
7%.20 The main difference was in computa-
tional effort. Renormalization speeds are up
to two orders of magnitude faster than those
of direct numerical simulation. The largest
problem yet tackled totalled 540 million
grid blocks.
Renormalization has drawbacks: it is diffi-
cult to represent contorted flow paths with
small cells and the estimation of effective
permeability suffers. This is seen in very
shaly reservoirs, where there are high con-
trasts between neighboring permeabilities.
This problem may be resolved by another
method that short-circuits the need for scale
up and requires a different kind of reservoir nCutting simulation time using streamtubes, a network of conduits conveying fluid
characterization. This is the streamtube from injection to production wells. This field-scale model incorporates five solvent injec-
approach, championed by Chevron Oilfield tion wells (pink) and seven producers (yellow). Using streamtubes, simulating the injec-
Research Co., La Habra, California, USA, tion of 2.5 pore volumes of solvent took just 30 seconds on a MicroVAX 3200 worksta-
tion. Using conventional fluid-flow simulation for a coarse grid and a Cray XMP-48, the
among others.21 The plan view of the reser- calculation required 630 seconds. The two calculations predicted similar solvent break-
voir is simulated as a network of tubes that through. (After Hewett and Behrens, reference 21.)
carry the flow between injection and pro-
duction wells (right ). The technique takes
20. King PR: “The Use of Renormalization for Calculat-
into account well placement, areal hetero- may always be able to generate more com- ing Effective Permeability,” Transport in Porous
geneity and the relative flow rates of wells. plex geologic models than can be simu- Media 4 (July 1989): 37-58.
The key to the technique’s success is a pre- lated. Scaling up will therefore remain a key King PR, Muggeridge AH and Price WG: “Renor-
liminary flow simulation on two-dimen- element in reservoir characterization. malization Calculations of Immiscible Flow,” Sub-
mitted to Transport in Porous Media, August 1991.
sional vertical slices between wells. The The fluid-flow simulation bottleneck will 21. Emanuel AS, Alameda GK, Behrens, RA and Hewett
observed scaling behavior is then mapped have to be addressed, either by developing TA: “Reservoir Performance Prediction Methods
onto individual streamtubes and their contri- new, more rapid processing hardware like Based on Fractal Geostatistics,” SPE Reservoir Engi-
neering 4 (August 1989): 311-318.
butions summed to get a three-dimensional parallel processing 22 or by smarter algo-
Hewett TA and Behrens RA: “Scaling Laws in Reser-
prediction of reservoir performance. rithms, or probably both. Then, the capabil- voir Simulation and Their Use in Hybrid Finite Dif-
Because of the relatively greater arith- ity for cost-effective and repeated stochastic ference/Streamtube Approach to Simulating the the
Effects of Permeability Heterogeneity,” in Lake LW,
metic complexity of fluid-flow simulation simulation will be realized. The impact of Carroll HB Jr and Wesson TC (eds): Reservoir Char-
compared with model building, computers this capability on reservoir management, acterization II. San Diego, California, USA: Aca-
and ultimately on improved recovery, demic Press Inc. (1991): 402-441.
remains to be felt. —CF 22. Mayer DF: “Application of Reservoir Simulation
Models to a New Parallel Computing System,” paper
SPE 19121, presented at the SPE Petroleum Com-
puter Conference, San Antonio, Texas, USA, June
26-28, 1990.

January 1992 39
 $
RESERVOIR OPTIMIZATION
Taming the Geoscience Data Dragon
Cross-disciplinary uses of oilfield information are changing the face of petroleum data
management. Here is a look at exploration and production data management—where it
stands today and where it needs to go.
Steve Darden The center of attention for the exploration
John Gillespie and production geoscientist is the base
FINDER Graphics Systems, Inc. map—a map of geologic or geophysical fea-
Corte Madera, California, USA tures, such as structure, fluid type or bed-
ding tops. This map, built from cross sec-
LaRay Geist tions based on log and seismic data, is used
Geoffrey King to locate new prospects and plan develop-
BHP Petroleum (America) Inc. ment of existing ones.
Houston, Texas, USA To produce this map in early 1980s, the
E&P geoscientist spent 70 to 80% of the
Scott Guthery time locating, sorting and reprocessing data
Ken Landgren and making names and scales of data con-
Schlumberger Austin Systems Center sistent.1 At the end of the day, a map was
Austin, Texas, USA produced by batch processing on a main-
frame computer. Hours later, or the next
John Pohlman morning, the geoscientist draped the map
ExplorTech Computer Applications over a drafting table, beside previous maps.
Houston, Texas, USA To do the analysis, the scientist looked at
everything—compared maps, referred to
Samuel Pool seismic sections and logs that covered the
ARCO office wall, and hunted through production
Plano, Texas, USA and core reports. If the scientist could keep
track of the data, he or she could assemble
Dave Simonson a clear mental picture of the reservoir and
Consultant therefore generate the best possible map.
Oakland, California, USA Ultimately, the geoscientist would meet
with specialists from other oilfield disci-
Paul Tarantolo, Jr. plines and, through a series of meetings,
Exxon Production Research Company contribute to the integration of different map
Houston, Texas, USA versions. For example, the geologist’s map,
built from well log correlation, would be
Dan Turner integrated with the geophysicist’s map, built
Petrotechnical Open Software Corporation from seismic data.
Houston, Texas, USA To produce a map in the 1990s, the E&P
geoscientist sits before a workstation. Only
the remnants of clutter remain. The drafting
table is little used. Information from stacks
40
of paper and film have been loaded into the
computer. Porosity data are three keystrokes
away rather than buried in a pile ( next
page ). Seismic and log data pop up in sepa-
rate windows on the same screen. The
overnight batch job is replaced by an inter-
active map revision...let’s see what happens
if we assume the Dunlin formation here is
200 feet thick instead of 170 feet...The geo-
scientist no longer tracks data mentally; that
is done by the workstation. He or she can
devote more energy to data analysis and
more to data management. Now data are
shared across disciplines, and interdisci-
plinary integration is no longer a separate
step. It is contained in the interpretation.
The workstation is the most visible prod-
uct of a data revolution that, in just ten
years, has affected every level of the E&P
business.2 In many oil companies, central-
ized data processing is giving way, in whole
or part, to smaller units of “project data”
downloaded from mainframe computers to
workstations. The user’s demand for easier
communication between application pro-
grams is compelling software vendors and
oil companies to set aside rivalries and join
forces to write industry standards for data
formats, data exchange and computer
graphics. Data organization systems are
springing up that allow previously isolated
data to be accessed from one workstation.
Ten years ago, all this was mostly infeasible.
Oilfield Review
 nThe old world of
geoscience inter-
pretation meets the
new. Remnants of
clutter remain
around a demon-
stration worksta-
tion at FINDER
Graphics in Corte
Madera, Califor-
nia. Nearly all the
data surrounding
Arlene Fox and
Eric Erickson have
been loaded into
the data base
accessed by the
workstation. Note
on the far wall the
“correlation” estab-
lished between
logs, using pushpins
and rubberbands.
The same sort of
correlation has been
migrated into the
computer, using
graphic devices.

Gary Wagner photo

As a result of increasing computer power
and versatility, the exploration and produc-
tion data dragon has been caged but not
tamed (see “Evolution of Petroleum Data
Management, next page). In 1992, explo-
ration and production data management
stands at an uncertain and exciting thresh-
old, marked by rapid change in three areas:
computer hardware, computing standards
and database development.
Hardware
Reservoir data today are stored and ana-
lyzed on four kinds of hardware systems:
mainframe computers (including minicom-
puters and servers), workstations, a combi-
nation of mainframes and workstations, and
on small computers, usually workstations or
personal computers (PCs) dedicated to spe-
cific tasks, such as seismic or log analysis.
Many see this as an unstable transition
state, and predictions of the outcome vary.
Makers of mainframes note the petroleum
industry’s long-standing commitment to the
mainframe, citing its many advantages:
computing power, capacity for data security,
ease and reliability of data access (usually a
single interface for all data), and ease of
tracking data history. Workstation vendors
see their tool as the mouse—limited in
power, but flexible and inexpensive—that
(continued on page 44)

For help in preparation of this article, thanks to Jeffrey A.
Brown and J. William Bradford, GeoQuest Systems, Inc.,
Houston, Texas, USA; Ron Dietsch and Lisa Stennes,
Petroleum Information, Denver, Colorado, USA; Eric
Erickson and Arlene Fox, FINDER Graphics Systems,
Inc., Corte Madera, California, USA; Jerry House,
Petroleum Information, Houston, Texas, USA; Mel
Huszti, Gulf Canada Resources Limited, Calgary, Alberta,
Canada; Stephen Kenny, Digitech Information Services
Ltd., Calgary, Alberta, Canada; Pam Koscinski, Dwight’s
Energydata, Inc., Oklahoma City, Oklahoma, USA; Marc
Lador, Petroconsultants, Geneva, Switzerland; Gary
Meyers, ARCO Oil and Gas Company, Plano, Texas,
USA; Phil Freeman, Eric J. Milton, Dwight V. Smith and
Terry J. Sheehy, FINDER Graphics Systems, Inc., Hous- A list of acronyms used in this article appears on page 54.
ton, Texas, USA; Av Munger, Munger Oil Information 1. Citerne A and Yu K: “Knowledge-Based Well Data
Services, Los Angeles, California, USA; Bill Quinlivan, Management and Graphical Interface,” paper SPE
Schlumberger Austin Systems Center, Austin, Texas, USA; 17611, presented at the SPE International Meeting on
Ron Samuels, Dwight’s Energydata, Inc., Dallas, Texas, Petroleum Engineering, Tianjin, China, November 1-
USA; Bill Schork and Lisa Stewart, Schlumberger-Doll 4, 1988.
Research, Ridgefield, Connecticut, USA; Ron Uchida,
2. About 50 companies offer E&P hardware or software
FINDER Graphic Systems, Inc., Lakewood, Colorado,
systems for the exploration and production environ-
USA; Laramie M. Winczewski, Simon-Horizon Inc.,
ment. See Oil & Gas Journal 88, no. 11 (March 12,
Houston, Texas, USA.
1990): Special Supplement, 32-3–32-17.
The following marks and trademarks appear in this arti-
cle: Cray (Cray Research Inc.), FINDER (Schlumberger),
GeoShare (joint mark of Schlumberger and GeoQuest
Systems, Inc.), Intel (Intel Corp.), Macintosh (Apple Com-
puter, Inc.), MOTIF (Open Software Foundation, Inc.),
ORACLE (ORACLE Corporation), UNIX (AT&T Bell Labo-
ratories), VAXstation (Digital Equipment Corp.), VMS
(Digital Equipment Corp.) and X Window System (Mas-
sachusetts Institute of Technology).

January 1992 41
 Evolution of Petroleum Data Management

1920 1930 1940 1950 1960

•1919
• •1924
• •
•1928
•
•1934 •1945
• •
•1951
• •1955
• • •1956
Munger Oil Everett Petroleum Production Electronic Ferranti Mark I, Production Petroconsul-
Information De Golyer Information reports1 offered Numerator, first commer- information tants S.A.
Service first to discovers first Corp. (PI) (Texas Panhan- Integrator, Ana- cially manufac- available in starts monthly
offer drilling & oil field using started in dle) by company lyzer and Com- tured computer, digital form scouting report
completion single-fold Denver, that became puter (ENIAC) is installed at (Lockwood). service in Cuba;
reports (only in seismic data, Colorado, USA Dwight’s Energy- invented at Manchester expanded to
Southern Cali- Nash salt dome, as weekly drilling data. University of University (UK). world outside
fornia, USA). Brazoria County, report for Rock- Pennsylvania, North America
Texas, USA. ies only. USA, first fully in 1962.
•1953
functional elec-
First commer-
tronic calculator.
cial digital
computer (IBM).

1. Production reports typically include data on monthly and cumulative production, lease data, total depth, completion dates and
number of wells on the lease.
2. Scout reports, or “tickets,” typically include all drilling-related data from the time the well is permitted through to completion.
TRANSISTORS

VACUUM TUBES I

Paper and film

Core memory
Solid state m

Drum me
Magnetic ta
Punch cards and paper tape
Hard disk sto

Computing for Oil—A Short History of Data in the Petroleum Industry

Over the past 40 years, the life of the geoscientist
has been changed largely by advances in infor-
mation science—how data about the earth are
acquired, processed, stored, retrieved and ana-
lyzed. Before computers, prospects were found
mainly by surface reconnaissance and crude
seismic surveys. This required a tremendous
number of people, to both acquire and sort
through data. Armchair exploration using com-
mercial data services—scout tickets and other
well data—became possible in the 1930s, but not
practical until the 1950s, and was labor-intensive
until the 1970s.
The first use of computers in the petroleum
industry was far removed from exploration. In the
42
late 1950s, computers proved their mettle in
helping solve differential equations describing
the flow of oil through the reservoir. An early
application for exploration was analysis of a
small number of grids for 2D reservoir modeling.
Computational power jumped a level in the early
1960s, when discrete transistor technology arrived.
For reservoir modeling, this allowed adding a
third dimension and increasing the number of
grids, which improved resolution. But, more sig-
nificantly, for the first time earth properties with
nonlinear characteristics, like relative permeabil-
ity and capillary pressure, could be modeled.
Computers moved from novelty to commercial
popularity with the introduction of the IBM 360
series in 1964, and geoscientists began moving
data from bookshelves and file cabinets into com-
puters. They built well files that had mainly raw
log data, but also some interpreted data. To eval-
uate a prospect, they would take a seismic sec-
tion and correlate it by hand to the sonic log.
There was no automated correlation, and the
computer well files mimicked the paper well files
from which they descended. Later, computer fil-
ing grew more sophisticated and data bases
evolved within each discipline: geology, geo-
physics, drilling, petrophysics and reservoir
engineering. By the 1970s, it was standard to
manually tap information from these separate
computer data bases. Now, exploration didn’t
rely mainly on expensive field reconnaissance. It
ution of Petroleum Data Management

1960 1970 1980 1990

•
•1962
•
•1971
• •
•1972
•
•1976
•
•1978 •1987 •1989
• •• •1990
First digital Intel invents Scout reports Delivery of first National Pro- RISC technol- Development Scout informa-
scout2 reports microprocessor. available Cray-1 super duction System ogy becomes of International tion on CD
(PI). nationwide computer. (NPS), a com- commercial. Relational ROM (Dwight’s).
[USA] (PI). mon format for Information
First production data System
•1973 commercial and histories, (IRIS 21)
International satellite established by relational data
well and con- telecom- 13 major oil base for E&P
cession data munications companies. data
supplied on of log data. First on-line (Petroconsul-
computer tape system for con- tants).
(Petroconsul- veying produc-
tants). tion data world- Raster images
wide via phone of entire logs on
lines (Dwight’s). diskette, optical
disk or magneto-
optical disk (PI).
Early 1960s: Emergence of 2D Early 1970s: Intergraph first company to adapt interactive
multifold seismic data. computer display technology for oil industry application.

Mid- to late-1960s: Wireline logs and Late 1970s: Gulf Oil’s ISIS (Interactive Late 1980s: Interactive color graphics
seismic data on paper computer tape or Seismic Interpretation System), first on- capability available for ~ $20,000, mak-
cards. First capability to store and print screen integration of two disciplines: ing it commercial for the first time.
multiple copies of logs. seismics and wireline logs.

MICROCIRCUITRY

INTEGRATED CIRCUITS

Video

Solid state memory

Mass storage unit (IBM, 1976)
Drum memory
Magnetic tape

Hard disk storage

Floppy disk storage
Optical storage

could also be performed in the computer room
with purchased data.
These first discipline-specific data bases were
hierarchical and so only relationships between
data sets anticipated by the author of the hierar-
chy could be considered. The production data
base, for example, might have been organized by
lease block only, not by hydrocarbon type, pro-
duction volume or completion date. If you wanted
to map productivity from the Wilcox formation,
you would have to manually identify the Wilcox in
each well in the petrophysics data base, then
compare that manually with each well in the pro-
duction data base—and the two data bases might
have been in different buildings, on separate
computers and using different software.
These data bases were part of the established
isolation of disciplines. By and large, the produc-
tion engineer never talked with the exploration
engineer. They worked for the same company, but
belonged to separate organizations, with sepa-
rate cost control points, separate budgeting
issues and often different management points
that converged only at the level of chairman.
Today, this separation has largely disappeared,
due in part to a force from outside the industry.
In the 1970s in California’s silicon valley, the
first papers were published on relational data
base technology, and by the early 1980s it was
becoming the rage in the computer industry.1
Although searches on relational data bases were
initially slower than on hierarchical ones, the
idea of searching by “relations” fostered a new
way of thinking about database design—you don’t
need to anticipate all the relations between data
when designing the data base. Through the 1980s,
this technology trickled into the oil industry, which
had become thoroughly computerized. Relational
technology made possible data integration across
disciplines—core and log porosities, for instance,
were combined in the same data base. Slowly,
1. The seminal paper that started relational technology was
by Edgar Codd of IBM’s research laboratory in San Jose,
California, USA: Codd EF: “A Relational Model of Data for
Large Shared Data Banks,” Communications of the ACM
13, no. 6 (June 1970): 377-387. An update on this work
is Codd EF: The Relational Model for Database Manage-
ment. Addison-Wesley Publishing Company: Reading,
Massachusetts, USA, 1990.
43
replaced the brontosaurus mainframe—
powerful, but cumbersome and expensive.
Users are taking several roads. A handful
emphasize mainframes, some are going
completely to workstations, and the major-
ity are trying to mix workstations and main-
frames. The consensus seems to be that
mainframes will not disappear, but their role
may change from data analysis to data
archiving and control. The main exceptions
are reservoir simulation and seismic data
processing, which benefit from being per-
formed with conventional microcomputers
linked in parallel or with supercomputers.
Urgently needed is a reliable link between
mainframes and workstations, since most oil
companies have both technologies and
need a way to get the most from each.
Many companies have assembled in-house
links that operate with varying degrees of
efficiency and reliability (right ). Often, how-
ever, mainframes and workstations cannot
communicate easily, if at all.
Much work remains before the two tech-
nologies can be married in a standard, non-
proprietary way. Meanwhile, a leading effort
for a proprietary linkage is a project orga-
nized by International Business Machines
(IBM) and consisting of IBM and a consor-
tium of oil companies. The project is an
attempt to develop a large IBM mainframe-
based relational data management system
with a mechanism for workstations to
upload and download data (see “Database
Development,” page 49, for discussion on
data bases).
The data base is unique in that it will be
several times larger than any existing oilfield
data base, able to hold information from
several fields. Because this system is the first
Boeing 747 in an industry that has so far
holes appeared in the walls separating types of
data and separating E&P organizations.
This incremental change was given a boost,
ending disciplinary isolation, by the oil price col-
lapse of 1986. Suddenly, fewer resources were
available to solve a growing number of E&P prob-
lems. Management consequently had to revise
business strategies, and one practice that
emerged was the interdisciplinary team. Today,
the idea of the geophysical team solving a prob-
lem on its own, independent of the geologist and
reservoir engineer, is almost nonexistent. But
before the interdisciplinary teams can work effi-
ciently, a final challenge remains—finding a way
to join data from different disciplines.
Mainframe
Log
Mapping
Analysis
Seismic
Interpretation
Application Application/data
programs interface
Data files/data base
nTransfer of data by links between pairs
of computers. The advantages of this
approach are that it does not hinder cre-
ation of new application programs; it iso-
lates the impact of change; and it allows
tight integration of programs within a dis-
cipline. The disadvantages are that it
does not provide integrated access to
data, requires transfer software for each
link pair, and makes tracking of data his-
tory and consistency difficult. The names
of the programs are taken as examples;
they could be any type. (From Guthery et al,
reference 3.)
Again, the roots of the solution came from out-
side the industry. From the ashes of the price col-
lapse rose a few companies offering relational
database technology for the petroleum industry,
among them, Ingres Corporation (purchased by
Ask Computer Systems Inc.), Oracle Systems
Corp. and Software AG. These data bases
enabled geoscientists to relate data that previ-
ously had not been recognized as related, and it
allowed users to define the relationship between
the data. For example, it allowed log values to
constrain a seismic inversion.
Along with the introduction of this new capabil-
ity, in the late 1980s, was the declining cost of
storage media, which increased the amount of
on-line information. And the more information at
evolved only to the Piper Cub, IBM faces
several new challenges. A technical chal-
lenge, given the size of the data base and
nature of its contents, will be a design that
permits establishing meaningful relations
between data. How the mainframe will
communicate with workstations is also
uncertain. Many oil companies and work-
station vendors want IBM to design the sys-
tem to be compatible with their own sys-
tems. This would simplify, for example,
networking, checking of data in and out of
the mainframe and data representation.
The shift to workstations also affects peo-
ple doing E&P work. As oil companies try to
maximize their resources, managers wel-
come workstations for their economic as
well as technical benefits. Buying and main-
taining workstations is significantly less
expensive than mainframe computers. Addi-
tionally, workstations let geoscientists take
greater responsibility for data organization
and quality, formerly the work of data man-
agers or geotechnicians.
The Battle for Computing Standards
Linguists estimate that when Christopher
Columbus landed in the New World 500
years ago, Native Americans spoke 1000
separate, mutually unintelligible languages.
Today’s E&P geoscientist, casting around for
a data system, may feel like Columbus (see
“Misconceptions about Data Management,”
next page and “Matching Needs and Capa-
bilities,” page 47 ). Among the numerous
standards, only a few are recognized con-
ventions—and sometimes within these are
several incompatible versions.
The current generation of exploration soft-
ware consists mostly of workstation- and
PC-based, stand-alone applications and
the geoscientist’s fingertips, the more obvious
the need for a way to organize it and make it
accessible between different hardware and soft-
ware configurations.
This brings us to the present, and the efforts of
standards groups to define an open systems
architecture that is acceptable to both vendors
and clients (see pages 46-48 of main text). The
goal now is to make all forms of data available to
any geoscientist working on any workstation,
seamlessly and effortlessly.
For further reading:
Augarten S: Bit by Bit: An Illustrated History of Computers.
New York, New York, USA: Ticknor & Fields, 1984.
Computer Basics: Understanding Computers. Time-Life
Books, Alexandria, Virginia, USA: 1989.
their proprietary data bases.3 While these
tools are powerful within their own realm,
they can neither talk to each other nor eas-
ily share data. Routine analysis usually
requires several software applications, often
from different vendors, and the user must
manually reformat the output from one
application before loading it into the next.
To make matters worse, applications often
reside on different computer systems, so
data must be moved from system to system,
and in many cases must be rekeyed when
data conversion is not performed automati-
cally. Even if two systems can be linked,
data sharing is often slow because the
underlying proprietary data base is struc-
tured to run only its specific application as
fast as possible.
This lack of communication emerged
because software developers seek speed
through proprietary solutions. Furthermore,
until recently, both oil companies and ven-
dors tended not to share solutions because
of the perceived risk of revealing proprietary
algorithms.4 The direct effect is a higher cost
of software engineering—each software
vendor must design user interfaces, graphics
standards and database architecture. And
this is often at the expense of analytical
capability. The indirect effects for the user
are less efficient use of software programs,
higher training and support costs and time-
consuming data management.
Geoscience computing entered the 1990s
with a lack of standards—such as for user
interface, data exchange and graphics. For-
tunately, by the mid-1980s, the computer
industry at large had recognized the futility
of a proliferation of standards, and that with-
out a cooperative effort, everyone would
lose their competitive advantage. Standards
groups sprang up for operating systems5,
quarter-inch tape manufacturers and
3. Guthery S, Landgren K and Waagbo K: “Integrated
Data Access for Geoscience Interpretation Systems,”
GEOBYTE 5, no. 5 (October/November 1990): 38-41.
4. Schwager RE: “Petroleum Computing in the 1990s:
The Case for Industry Standards,” GEOBYTE 6, no. 1
(February/March 1991): 9-13.
5. An operating system is a software family that lets
information flow through the computer as fast as pos-
sible but in an orderly fashion. It also serves as a trans-
lator or interpreter for the user. A user doesn’t talk to a
computer, but to an operating system, which com-
mands the hardware to fulfill the user’s request. Tasks
of an operating system include:
•Controlling movement of data into and out of com-
puter memory
•Supervising operation of the central processing unit
•Sending data to peripheral devices, such as monitors
and printers
•Keeping data and programs uniquely identified.
In oilfield computers, common operating systems are
UNIX, Disk Operating System (DOS), and Virtual
Memory System (VMS).
Misconceptions about Data Management
• Buying a data base or data management system
is enough. Buying a system is the beginning,
not the end, of problem solving. Once a system
is acquired, there are other issues to confront:
How are data converted and loaded? Who will
perform quality checks before data are loaded?
What will be required to ensure that the new
system interfaces with existing systems, or
other systems that might be acquired in the near
future? What kind of training is available? What
are the maintenance agreements?
Loading a large data base is a major task. In
the early 1980s, Exxon Production Research
built a seismic interpretation data base. Locat-
ing, converting and loading data for a relatively
small project—100 seismic lines—took
months. During database construction, a geo-
scientist working by hand could finish an inter-
pretation before another geoscientist awaiting
completion of the data base. But after the data
base was complete, mapping took half as long.
• All types of information can be dealt with the
same way. Certain systems may improve pro-
ductivity only for certain tasks. Most information
systems are good at only one or two tasks. One
oil company bought a mapping system that was
designed for cultural data used by city planners
and highway engineers. They found, however,
that well data were so dense that data “over-
posting” produced an uninterpretable image on
the screen. The vendor was not responsible for
the problem—its system was working as
intended. The oil company had to donate to the
vendor millions of dollars in software engineer-
ing to solve the problem.
• If data are in the data base, they must be right.
Many geoscientists can tolerate working with
paper copies of data that contain errors. They
perform quality assurance on the fly, or compen-
sate in their interpretation. But once data
become digitized from paper to an electronic
file, the perception is that the data must be cor-
rect—“otherwise, somebody wouldn’t have put
it in the data base.” It is valuable to recognize
that data are data, no matter the form. A valu-
able component of the data base is a record of
who loaded and checked the data.
Checking of data requires not only attention to
detail but a wide-ranging imagination. Exam-
ples of errors found in vendor data include
February having 30 days, an offshore platform
with the Kelly bushing at 2000 feet, and a 1000-
foot well with TD at 5000 feet. In the future, much
of checking for plausibility will be automatic.
• Just having a geographic information system
(GIS) or data base will increase productivity.
The learning curve for workstations can be on
the order of months to years. One seismic work-
station vendor’s own survey found that less than
1% of the users of its system ranked themselves
as “experienced,” even after eight years. One
reason is that understanding the nuances of
workstation operation takes time. Another rea-
son is that most users are occasional users,
who tend to learn the machine well enough for
their job but not to the point of mastery. The
learning curve is shrinking, as new machines
become easier to use, but they still require
some computer background. An average user of
the FINDER system, for example, can get the
basics in about one week—given a background
in UNIX and SQL and some experience with
ORACLE data bases.
A second part of this misconception is that the
only advantage of a GIS system is the ability to
work faster. True, the geoscientist can spend
less time looking for data and therefore produce
an interpretation faster. But an often overlooked
advantage is that a more thorough interpretation
can be performed in the same amount of time as
when done on a mainframe.
• Information management is somebody else’s
job. Anyone using a computer that is not a
stand-alone system is managing shared infor-
mation, and information management is part of
his or her job. This attitude helps prevent devel-
opment of an adversarial relationship between
people using an information system and people
managing it. Every explorationist is responsible
for the quality of data that go into the interpreta-
tion. Data loading and quality assurance may be
subcontracted, but the explorationist is respon-
sible for correctly placing the wells, scaling the
logs and picking the formation tops. It is in the
45
 best interest of the company if the explorationist’s
needs and understanding of the data are clearly
communicated to people managing the data base.
• A data base is an application. Workstation appli-
cation products, such as for log interpretation or
mapping, have an underlying internal data base
that is largely invisible to the user. Consequently,
the concept of a data base is often not appreci-
ated. The data base itself is not an application
but a reservoir of information accessed through
an application.
• A central (or a local) data base is preferred.
The conflict in the choice of central vs. local is
between the need to freely make changes in the
data and the need to prevent changes that
destroy the accuracy of data. Central data bases
have higher security, but can inhibit interactive
interpretation. One solution to these conflicting
needs is the use of a database management
system that facilitates the automatic creation of
local data bases, devoted to the study area. This
allows the geoscientist to manipulate a sub-
group of data, without altering the integrity of
the central data base.
• Graphics capabilities are of foremost impor-
tance. Good graphics capability in a workstation
is only one factor in its overall performance.
Often overlooked in judging the day-to-day prac-
ticality of a workstation is ease of data load-
ing—not something vendors rush to demon-
strate on the convention floor. Nevertheless,
finding and loading data take a large chunk of
the geoscientist’s time, and the easier the load-
ing, the sooner data analysis can begin.
• Speed is of the essence. Speed is important in
an application, but less so for a data base.
Database software and hardware can be tuned
to a specific application, increasing database
speed 20 to 200 times over that of an off-the-
shelf product. Also, workstation hardware is
likely to be upgraded before workstation soft-
ware, so that in two years, there is a good
chance the user will be running the same soft-
ware—or an updated version—on hardware that
is twice as fast.
46
40 41
Geopolitical
Data
Well Data Seismics
Logs Subsurface Surveys
Cores Model Cross
Cuttings sections
Production
Data
60
50
Barrels of Oil/Day
40
30
20
10
0
82 83 84 85 86 87 88 89 90
Year
nA central part of an E&P data model is the subsurface model, which is used to guide
the search for, and development of, petroleum reserves. The subsurface model is devel-
oped with information from the four surrounding disciplines. (Adapted from Berkhout et al,
reference 6.)
database programs. By the time the problem Flaws in data models often become
of multiple standards surfaced in petroleum apparent when a type of data does not fit
computing, the broader computer industry’s the model or when relations between data
efforts could be used as a guide. cannot be accommodated by the structure
As geoscientists mix interpretation from of the model—for example, horizontal wells
the four major disciplines—geology, geo- may require new ways of computing log
physics, petrophysics and reservoir engi- data, such as varying the measured value
neering—a foremost need is for a standard with horizontal distance rather than vertical
exploration and production data model: a depth. A data model may have nothing to
system for organizing and defining geo- do with computers. Computers provide only
science information and interrelationships a representation of that model, such as a
between data to model the reservoir in a relational or hierarchical database structure.
way that fulfills a need (above).6 Because there are so many ways to define
A typical question in assembling a data oilfield data and relations between them, oil-
model is the definition of a so-called major field data models are complex, attempting to
element. For example, is (A) a seismic line cover as many phenomena and data permu-
part of a seismic survey or is (B) a seismic tations as possible. Through the 1980s, the
survey part of a seismic line? A computer move toward integrating data across disci-
system using a model that organizes data as plines made it clear that a cooperative effort
(A) cannot easily exchange information with involving vendors and users was necessary
a computer system using a model that orga- to unify the field, or at least reduce the num-
nizes data as (B). ber of geoscience data models.
Oilfield Review
 The IBM effort at linking mainframes and
workstations depends ultimately on devel-
oping a satisfactory data model. Two, more
sweeping efforts have been initiated by the
Petrotechnical Open Software Corporation
(POSC), based in Houston, Texas, USA, and
the Public Petroleum Data Model (PPDM),
based in Calgary, Alberta, Canada. Both
groups are not-for-profit organizations.
The mission of POSC is to “...define,
develop and deliver, through an open pro-
cess, an industry standard, open systems
software integration platform for petroleum
upstream technical computing applica-
tions.” The PPDM seeks to “create an indus-
try standard means of assembling, remem-
bering and communicating data related to
the petroleum industry.”
POSC, founded in October 1990, is a cor-
poration funded by 35 companies from
Europe and North America representing the
E&P industry, oilfield services, software and
computer manufacturers and government
research establishments. POSC acts as a kind
of United Nations for geodata standards,
soliciting computer-based technology from
across the industry to be adopted in whole
or part in a POSC “software integration plat-
form.” To date, 20 companies have submit-
ted 34 models for consideration by POSC.
The POSC platform constitutes an E&P
data model, a user interface with a “com-
mon-look-and-feel,” software offerings and
a set of specifications and reference codes
that will allow vendor products to be com-
pared with selected industry standards.7 Its
first release—a model for data exchange
and user interface—is scheduled for the
third quarter of 1992. The group’s $6.5 mil-
lion budget funds a full-time staff of ulti-
mately 44, and is supported by member
companies expected to number 40.
Although the group currently operates solely
from Houston, its board of directors has
authorized the opening of a second office in
the UK.
The success of POSC is linked to the
emergence of the so-called open systems
environment: software and hardware design
that permits any application to run on any
hardware.8 Although “open systems” is on
the lips of many software engineers, the oil
industry has reason for some resistance
against it—a large investment in proprietary
software and hardware. POSC sees part of
its job as showing how these existing sys-
tems can be adapted to talk with each other
and with new systems.
POSC cites several advantages to the
open systems environment. It is in line with
January 1992
Matching Needs and Capabilities
Questions to Ask When Selecting an E&P Information System
A chief cause of dissatisfaction with information systems—from data bases to integrated interpretation
packages—is that the buyer had neither the time nor expertise to fully evaluate the capabilities and limi-
tations of the system. Answering several questions while shopping can help the buyer clarify his needs
and find the product that best satisfies them.
• What is going to be the present and future use of
the system? Who are the users?
• What do I want to store and why do I want to
store it that way?
• Does the data base have facilities that permit
easy transfer of data between different data
bases—between the corporate, basin and pro-
ject data bases?
• Can the data base exchange information with
software applications we use?
• Can the data base connect with any Structured
Query Language (SQL) data base or product, such
as spread sheets or geoscience applications?
• What is the size of the business unit I will be
working within—all the wells in the Adriatic?
The Niger delta? The Eastern Hemisphere?
• What is the form of data to be used—a monthly
report? A weekly report?
• How is information accessed? Are privileges
limited or unlimited? Do users need simultane-
ous access to the same data? Will I need a sys-
tems manager?
• Where will information be stored—centrally,
locally or regionally? Will I need one big disk
drive?
the oil company trend of buying and modi-
fying off-the-shelf software, rather than
developing it in-house, which has become
too expensive. It also prevents “box
lock”—when the time comes to retire a
computer, the replacement won’t necessar-
ily tie the user to a single hardware supplier.
POSC expects that these advantages will
outweigh one drawback that may be associ-
ated with open-system architecture, com-
promised speed. Proponents of open sys-
tems maintain that this problem will be
minimized by hardware advances—an often
quoted figure is doubling of computational
speed every 12 to 18 months.
The PPDM group, founded in 1989, is a
volunteer organization currently of 20 pay-
• Will workstations be networked together? If so,
is network software available and how well does
it perform?
• What platform(s) will the vendor support, and
when?
• What data input formats does the vendor sup-
port? Are they compatible with those that I have
or am likely to get?
• How will I load the various types of data that I
expect to put on the system? How does the sys-
tem handle paper data?
• Do I need satellite or phone links?
• Will it interface with existing internal systems
(for lease administration, production account-
ing, asset accounting, etc) or systems we plan
to acquire? On what standards is the system
based?
• What is the upgrade path for this system? In two
years will I feel stuck with this system, or will it
be able to grow to meet future needs?
• Can the data base be easily customized to fit the
needs of different operating groups?
• How thorough and helpful is the vendor’s user
support?
6. For an example of an E&P data model, see Berkhout
AJ, Smeets G and Ritsema I: “A Strategy for Develop-
ing a Standard E&P Data Model,” Geophysics: The
Leading Edge of Exploration (September 1991): 33-37.
7. Turner D: “Petrotechnical Open Software Corp.—A
Way Forward,” GEOBYTE 5, no. 5 (October/Novem-
ber 1990): 36-37.
8. Wilson DC and King JW: “An Integrated Geoscience
and Reservoir Engineering Software System,” Advances
in Reservoir Technology: Characterization, Modeling
& Management, Royal College of Physicians, Edin-
burgh, UK, February 21-22, 1991.
47
ing subscribers comprising oil companies
and vendors of applications and hardware.9
Membership is $900, which is used by the
group for printing and other operational
expenses. There are no paid employees.
The PPDM is a standardized, nonpropri-
etary database format that serves as a frame-
work on which to build petroleum com-
puter application programs. Like POSC’s
model, the PPDM is released in stages. The
first database format, called a schema, was
released in April 1990. Subsequent versions
are revised, based on user feedback and
released periodically. The model currently
covers exploration and production data,
including production and core analysis,
openhole pressure and drillstem test data.
Plans call for pipeline, reserve and eco-
nomic data, lease information (expiration
dates, price paid) and seismic and log data.
Since its first release, the PPDM has
become the de facto standard in Calgary.
Only a small percentage of purchasers of
the model are from outside Canada. In
hopes of expanding the model’s appeal,
parts of it have been submitted to POSC.
The proliferation of workstations has stim-
ulated demand for other kinds of computer
standards for petroleum applications. Here
are highlights of a few.10
•User interface. This is perhaps most in
need of a standard. The only standards
that exist today govern the structure of
what things look like on the screen. An
important standard for the oil industry in
this respect is OSF/MOTIF, which gives a
common look to the screen. Each window
has a similar configuration and border; it
is moved around the screen and closed in
the same way; it has buttons that the user
clicks on to operate certain functions. Pro-
grammers have called for an application
style guide, defining standards for screen
layout, menu bars, dialogue boxes and
data input fields. Standards are also
needed to allow the user to select a pro-
ject and automatically start required
applications and access data bases.11
48
•Data conversion. The first generation of
workstations were cumbersome because
data often had to be converted to a format
acceptable to the program, which was
often specific to the workstation. This
extra step is removed in the emerging
generation of workstations, which are
capable, for example, of reading standard
log tape formats.
A subset of the data conversion prob-
lem is data compression, particularly of
seismic data. Some precision is lost in
going from 32-bit to 8-bit seismic data. In
some workstations, compression changes
seismic amplitudes, resulting in anomalies
in amplitude mapping.
•Data definition. Trouble is in the making
when a user tries to integrate data from
two systems that do not use coincidental
definitions of fundamental concepts. For
example, is bottomhole temperature static
or circulating? Is porosity log- or core-
derived? An effort to address this long-
standing problem is the Petroleum Indus-
try Data Dictionary (PIDD). The PIDD has
been assembled over the past few years
by the American Association of Petroleum
Geologists (AAPG) working with repre-
sentatives of 17 companies and 2 United
States government agencies. Although the
effort is entirely US-based, the dictionary
is intended to standardize industry terms
for what things are called, how they are
named in the data base. The 2448-term
glossary is being serialized in the
petroleum computing magazine of the
AAPG, GEOBYTE, and is available for
$10 on diskette.12 The next step for the
PIDD is creation of a standard number of
bytes— physical computer characters—
for data referenced by the glossary terms.
•Data sharing. As oil companies search for
the best workstation, they often end up
with four, five or more mutually exclusive
platforms. Proprietary programs are writ-
ten to link these resources with each other
or with the mainframe. These links are
often highly specific and expensive to
develop and maintain (page 44). An alter-
native approach to connecting applica-
tions has been developed, called the
GeoShare data exchange standard. It is
not a pair-wise link, but a bus to which
any number of applications can be con-
nected. In order for an application to
make a connection through the GeoShare
standard, the application developer must
write a link between the application‘s pro-
prietary data base and the GeoShare stan-
dard. The standard is based on the Ameri-
can Petroleum Institute (API) standard for
data exchange known as Recommended
Practice 66 (RP 66) or Digital Log Inter-
change Standard (DLIS).13
The first release of the GeoShare standard,
in spring of 1991, is a program and blueprint
for a two-way bridge connecting applica-
tions with each other, with a data base or
both, regardless of platform (next page, top).
The specification is being written by Geo-
Quest Systems, Inc. and Schlumberger.
The GeoShare system is analogous to the
clipboard in the Macintosh environment. To
move text from one program to another, you
can move it temporarily into a clipboard.
The GeoShare standard largely defines the
format of data that can go in the clipboard.
But it is more complex. The Mac simply has
to define each letter in each word. In the
GeoShare standard, however, the sender
and receiver have to agree not only on the
definition of a letter or number, but on the
definition of a fault, horizon or a log. It
therefore conducts not only data but a dic-
tionary of words to define the data in a sin-
gular way. Later, the GeoShare standard is
expected to include graphics interface stan-
dards that will permit shipping images
between different workstations.
Several major oil companies believe they
can utilize the GeoShare standard to
achieve data integration approximating the
“tight” integration obtained with a common
data base (see “Some Data Management
Strategies,” page 51). By configuring their
systems as shown on the next page, they
can move study data from the common data
base to the selected workstation as needed
and therefore keep only one master copy
with the most current interpretation. Upon
completion of analysis on the workstation,
the results are then moved from the work-
station into the common data base. The
sooner interpreted data are moved back into
the common data base, the less chance for
confusion over the location of the correct
version of the interpretation.
Oilfield Review
 Data Integration nData integration
today, before indus-
try-wide adoption
of a common data
Geologic
Modeling Log Analysis model. Applications
can talk with each
other or a central,
shared data base
Exchan through the
e Data ge GeoShare data
Lease, Contract, ar St Seismic exchange standard.
h Interpretation

an
Administration

oS
This is a form of

da
Ge
loose integration.

rd
Note that applica-
Shared tions can talk with
Data Base each other without
having to pass
through the shared
data base, but at
Production, Seismic the end of an inter-
Accounting Modeling pretation cycle, the
interpretation can
Tight be downloaded to
Integration the shared data
with base. In this exam-
Data Base ple, mapping and
cross section appli-
Mapping Cross Sections cations bypass the
GeoShare link and
are tightly inte-
grated with the
Read and write application half-links data base.

Another approach to data sharing is the A Data System Comparison

OpenWorks program, being developed by Public Library Oilfield Data System
Landmark Graphics Corp. There are several
Data base Books, magazines, Seismic, log, production
differences between OpenWorks and the newspapers, microforms and geopolitical data, etc.
GeoShare standard. Whereas the GeoShare
standard can be viewed basically as a DB Manager Librarian DB Manager or committee
bridge, OpenWorks is a link plus an attempt Quality control Staff evaluates condition of Geotechnicians, users,
to establish a unified user environment, a holdings, arranging for repair of data base administrators,
damaged items; correctly or all three. Data are checked
database management structure and inter- reshelves holdings for plausibility, correct scaling
process communications. Although Open- of logs, placement of wells,
Works has been made available to third- formation top picks, etc.
party software developers, its architecture Database Dewey decimal system Relational tables
optimizes Landmark applications. management Card/computer catalog Structured Query Language
system Holdings checkout Central processing unit
Database Development Hard copy device Photocopying machine, Printer, plotter
A data base is a collection of information printer
and a database management system is an
organizing system for storing, accessing and The first computer data bases were simply ple and networked systems, supplied by a
revising information that reflects the needs electronic means of storing what was in a vendor, capable of understanding the orga-
of the user (see “A Data System Compari- filing cabinet. The main advantage was that nization of different data bases on different
son,” above, right ). The key idea is that the the computer could locate data faster. Like machines and keeping a master list of every-
database organization satisfies a basic need, the first generation of application software, thing stored.14 (See “Bringing Data manage-
otherwise it’s just an unsorted pile of infor- oil companies mainly generated them in- ment into the 90s: The BHP Case Study,”
mation. Structure of the data base must also house. Database technology has evolved to page 52.)
accommodate the data model. the point where the state of the art is multi-

9. Rhynes PJ: “PRISM: Petroleum and Resource Indus-
try Information Storage and Management,” GEOBYTE
5, no. 5 (October/November 1990): 31-35.
The PPDM was founded by Applied Terravision
Systems, Digitech Information Services Ltd., FINDER
Graphics Systems, Inc. and Gulf Canada Resources
Limited.
10. Nation L: “Data Standards Work Progresses,” AAPG
Explorer (August 1991): 9.
11. Schwager, reference 4.
12. “Petroleum Industry Data Dictionary—Part 1,”
GEOBYTE 6, no. 3 (February/March 1991): 17-32.
The rest of the dictionary will be published in three
subsequent issues. For diskettes of the dictionary,
contact Pam Koscinski, Dwight’s Energydata, Inc.,
P.O. Box 270295, Oklahoma City, OK 73137.
Phone: (1) 405-948-7008; fax: (1) 405-948-6053.
13. Although the name DLIS suggests that the format is
good only for logs, it is in fact appropriate for
exchange of any kind of scientific data.
14. Gillespie JG: “Database Strategy for an Exploration
Workstation,” presented at the ASEG/SEG Interna-
tional Geophysical Conference and Exhibition, Ade-
laide, Australia, February 17, 1988.

January 1992 49
 There are two main types of data bases, nComparison of hierarchical and relational data base structures for the same informa-
tion used in a drilling operation. To find wells intersecting a gas zone completed by
classified by the structure they use to orga- XYZ company in November, 1990, the programming interfaces would be as follows.
nize information: hierarchical and relational (Adapted from Date CJ: An Introduction to Database Systems. Reading, Massachusetts, USA: Addi-
(right ). A rapidly evolving variation on these son-Wesley Publishing Co., 1977.)
two themes, called object-oriented method-
ology, is just becoming commercial. 15 Hierarchical Data Base
Today, it operates mainly as an overlay on
top of a relational foundation, but increas-
ingly it may function as a separate approach Field FIELD_NAME FIELD_TYPE REGION well_ptr nxt_field
to which relational, hierarchical or flat-file
structures are subsets.
A hierarchical data base organizes data
Well WELL_ID OPERATOR COMPL_DATE zone_ptr nxt_well
like the branches of a tree. To reach a cer-
tain leaf, one must always ascend the trunk
and follow the same set of branches. Or in
computer vernacular, a hierarchical system Porous_zone ZONE_NAME TOP BOT POROSITY FLUID nxt_zone
is a list from which an item is selected. This
item refers to another list, and another list
and so on until the end point is reached. For a hierarchical data base:
This system is well suited to data that are next_field:
always used in a fixed structure. Information get first/next FIELD
contained in well log headers, which is if not found then goto exit
always presented in the same order and for- next_well:
mat, is often represented hierarchically. The get first/next WELL where WELL.OPERATOR =“XYZ” and
advantage of this structure is its speed for a WELL.COMPL_DATE = “NOV 1990”
particular item. A disadvantage is absence if not found then goto next_field
of a shortcut to the end point. In a strict get first/next POROUS_ZONE where
hierarchical system, one must always POROUS_ZONE.FLUID = “GAS”
ascend and descend the same path. if found then print WELL.WELL_ID
A relational data base is more like a series goto next_WELL
of file drawers than a tree. Data items are exit:
loaded one at a time into lists and sublists,
in a way that allows each item to be located
independently. For example, a list of wells
might be made up of one actual list of Relational Data Base
records, but indexed in several ways: by
operator, by hydrocarbon type and by com-
pletion date. The relation between items is Field FIELD_NAME FIELD_TYPE REGION
built with the query, which tells the com-
puter how to order data coming out. Most
commercial data bases are relational, with Well WELL_ID OPERATOR COMPL_DATE FIELD_NAME
some sections arranged hierarchically and
some in other formats. Two advantages of
relational technology are the simplicity of its
internal structure, and the ease with which Porous_zone ZONE_NAME TOP BOT POROSITY FLUID WELL_ID
this structure can be modified, with little or
no effect on supporting software.
For a relational data base, a query interface might be: “Print WELL_IDs for wells inter-
Relational database technology became secting gas zones completed by XYZ Company in November 1990.”
practical in mid-1970s, when IBM intro-
duced Structured Query Language (SQL), select unique Well.well_ID from Well, Porous_Zone where
which has become the de facto industry
standard.16 This language enabled commu- Well.OPERATOR = “XYZ” and
nication between data bases or between Well.COMPL_DATE = “Nov 1990” and
application and data bases that adhere to Porous_Zone.FLUID = “GAS” and
the SQL standard. Well.WELL_ID = Porous_Zone.WELL_ID.

50 Oilfield Review
 Relational database technology still leads
the way into the 1990s, but another
approach, called object-oriented technol-
ogy, is drawing attention. This is considered
by some to be a new way to access data,
but by others to be a layer that fits on top of
a relational structure, allowing a new way
to group data.
The definition of an object-oriented data
base is still uncertain, but it is generally
agreed that the basic unit in an object-ori-
ented approach is an “object,” which is
managed as a single entity. An example is a
seismic section. In a conventional relational
data base, to call up a seismic section on
the screen, the user locates and requests the
various traces before the section can appear.
In an object-oriented system, the user asks
for the section by name or location, and the
computer locates the data and applications
needed to draw the section. Another exam-
ple of an object would be all the well logs
from a traverse across a reservoir, named,
for instance, A-A’. By simply marking A-A’
on the screen, the user can call up all the
logs without knowing anything about the
underpinnings of data in the data base. The
logs are treated as objects. Some object-ori-
ented products are coming out now, but the
technology isn’t expected to become signifi-
cant until the mid-1990s.
Data bases have recently been challenged
by growth in the amount and complexity of
data and demand for new connections
between data types. Two notable develop-
ments to meet these challenges are graphics
interaction and mixed-media storage.
In the last three years, advances in graph-
ics interactions have enabled computers to
emulate the way geoscientists think: in
maps. About 50 companies today are mak-
ing and selling mapping systems under the
generic name geographic information sys-
tem (GIS). Most of these first-generation sys-
tems are designed to handle 2D representa-
tions of the earth’s surface for use, for
example, in city planning and forestry pro-
jects. These GISs are limited, however, in
that they cannot gracefully satisfy the
demands of the E&P industry, which needs
to apply GIS principles to model the subsur-
face in three dimensions. These systems also
have no understanding of the E&P data
needs. Users have to write software to con-
nect data descriptions of leases, seismic sur-
veys, well logs and so on to the GIS’s recog-
nition of lines, points and polygons.
January 1992
The new generation GISs can model the
subsurface in 3D and have a complete com-
prehension of E&P entities. They understand
the data model used to store data in the data
base, and can translate this into lines, points
and polygons that allow a map to be built to
represent the subsurface data. Old GISs can-
not make this translation between the data
model and map. Users of the new genera-
tion can create searches of the data base
from a map. The new GIS shows on the map
all wells that satisfy a query—such as gas
wells the XYZ company completed in April
1990. By clicking on the well, the user can
bring up text annotations, such as who
logged the well, what logs are available,
total depth and bottomhole temperature.
The latest advance in database technol-
ogy—although still in commercial infancy—
is mixed-media capability, which has devel-
oped in response to the demand for different
kinds of data. Until recently, data bases
were limited mainly to digitized data, such
as seismic maps, log curves and tabular
data. With mixed-media storage, any kind of
information can be accessed—text, tables,
raster images of downhole data, such as
logs and core photos, as well as video and
audio information. Interpretations can be
stored as well as raw data.
With mixed media loaded into the data
base, the geoscientist can draw a polygon
around a set of wells and type a query,
“show core information.” A text track
appears with the list of wells cored, or cored
wells appear in a new color or a new sym-
bol. Then the user can query again: “show
Pierre formation,” and up comes a core
image, with thin section data and the core
analyst’s interpretation and notes—paleon-
tology, sedimentary structures, grain
description, and so on.
Advances have also taken place in rela-
tional database capabilities of workstations.
Some newer data management systems,
such as used in the FINDER system, have
the capability to withdraw two kinds of data
in one swipe: bulk data (sometimes called
vector data) and parametric data. Bulk data
are large chunks of data, composed of many
smaller data items that are usually retrieved
in a specific order—for example, a sequence
of values such as a porosity curve or a seis-
mic trace. Parametric data are single pieces
of data, such as those included in a log
header: mud density, bottomhole tempera-
ture, field name, and so on. In most data
bases, bulk and parametric data are stored
separately and retrieved separately—the
user must ask separately for a log curve and
header. The FINDER system, because it
stores some bulk data as if they were single
data items, allows the user to access both
types of data with one command. Storing
some data in bulk, such as well depths, per-
mits faster access because the computer
searches one already ordered entity, rather
than searching each depth and having to
impose order on it.
Some Data Management Strategies
Where data management stands today
might be described as “loose” integration
between platforms, and where it wants to go
as “tight” integration. While the distinction
may be mostly blurred, there are some clear
differences. The main difference is in the
number of occurrences of a given data item.
In a tightly integrated system, there is one
data store; in a loosely integrated system
there may be many. In general, today there
is often tight integration of application pro-
grams within an existing platform of a single
vendor. The long-term desire is for tight inte-
gration between platforms while, in the
short term, there is a need for loose integra-
tion between platforms.
To illustrate the difference between tight
and loose integration, consider two seismic
interpretation systems. One is used for seis-
mic structural interpretation, the other for
seismic inversion. It may be possible to
move a copy of the seismic section from the
interpretation to the inversion program, per-
form an inversion, then move it back to the
interpretation program to interpret an inverted
(continued on page 54)
15. Brewer KE and Pritchard RJ: “Integrated Technical
Computing of the 1990’s,” paper SPE 20359, pre-
sented at the 5th SPE Petroleum Computer Confer-
ence, Denver, Colorado, USA, June 25-28, 1990.
Rumbaugh JE, Blaha MR, Premerlani WJ, Eddy F and
Lorensen W: Object-Oriented Modeling and Design.
New York, New York, USA: Prentice-Hall, 1991.
Fong E, Kent W, Moore K and Thompson C: 1991,
X3/SPARC/DBSSG/OODBTG Final Report. National
Institute of Standards and Technology Draft, Septem-
ber 17, 1991. (This is a review draft of a study by the
object-oriented database technology group. This
landmark report states that object-oriented technol-
ogy and methodology is ready to use now by gov-
ernment agencies. For copies, contact Elizabeth
Fong, NIST, Building 225, Room A226, Gaithers-
berg, Maryland 20899, USA. Phone: (1) 301-975-
3250.)
16. Although IBM invented the SQL principle, its first
commercial implementation was in 1979 by Oracle
Corporation.
51
Bringing Data Management into the 90s: The BHP Case Study

1980s
In 1987, the Australian energy company, Broken
Seismic Geological
Hill Proprietary (BHP), recognized the need to Interpretation Interpretation
replace its Mapping and Geophysical Interpreta-
tion Computer (MAGIC), an in-house system for
Digitizing
exploration and production file management. Horizon Picks Well Completion Log Well Data Base
Reports Data Base
MAGIC was a suite of applications programs and
a data store that was mainly a repository for seis-
mic data. Its main limitation was that it operated MAGIC Data Base Log Displays,
only in batch mode, not interactively. Cross Sections
To assess its needs, BHP surveyed users
throughout the company on requirements for soft-
Base Maps Drafting
ware applications, data processing and data
management. A significant finding was that users
wanted closer interaction with their data. They
complained of a lack of control of data interpreta-
Hand Computer
Contoured Contoured
tion and processing functions, and wanted data to Maps Maps
be more accessible and more easily modified. Manual tasks

At the same time, the company recognized the Computerized tasks

industry trend toward workstations and user Drafting

access from the desktop, which fit well with the
user request. BHP also had the desire to store,
1990s
catalog and make available E&P data of all types,
not just seismic data. Geological Interpretation
Seismic Interpretation
Once BHP identified the company-wide (Cogniseis Development DLPS
(Landmark Seisworks
requirements, it looked for systems to meet these Interpretation Package
Module)
and Landmark Stratworks Module)
needs. The field was quickly narrowed to the Vor-
text system from Aangstrom Precision Corp. in
Michigan, the Exploration and Production Office
FINDER Exploration Data Base
System/Geoscience Information System
(EPOS/GSIS) from the TNO Institute of Applied
Geoscience in The Netherlands, and the FINDER Computer Log Displays,
data management system from FINDER Graphics Contoured Maps Cross Sections

Drafting Mapping
(Minor Alterations) (Zycor)

nBHP’s exploration technical systems overview.

52 Oilfield Review
Systems Inc. in California. The FINDER system
was found to have the greatest potential to satisfy
BHP’s requirements, and initial acquisition was
made in 1989. Since that purchase, BHP has
installed the system in offices in Melbourne, Aus-
tralia; Houston, Texas, USA; London, England
and Calgary, Canada.
Currently, BHP stores data in the FINDER sys-
tem as basin-wide data bases. In the Gulf of Mex-
ico, for example, the data base has about 7800
seismic lines representing 141,700 line miles
[228,000 km]; 202,600 well locations with 91,300
scout tickets, and cultural, lease and other types
of data that allow easy generation of base maps.
In the Melbourne office, the Timor Sea data set
comprises 5400 seismic lines representing about
123,000 line kilometers [76,400 miles], 200
wells, cultural, permit and interpretative data.
The hardware environments at BHP’s two main
sites are based on two different distributed com-
puting models. In Houston, the FINDER system
runs on seven stand-alone VAXstations with
access via a network from 19 X Window System
terminals, which are for graphics display. In Mel-
bourne, there is a central file server with seven
dataless VAXstations and X Window System ter-
minals attached to the server. Because of the
experience gained in these test configurations,
future implementations will probably be based on
a central server for each business unit with data-
less workstations and X Window System terminals.
The working environment is flexible. In some
cases, subsets of data are extracted as project
data bases within the FINDER system. At other
times, users who want regional context maps
work directly with the entire basin’s data set.
Within this framework, the FINDER system pro-
vides much of the routine functionality that BHP
had with its MAGIC system, plus the ability to
January 1992
interact with data. For capabilities not part of the are valid. The explorationist usually wants the
FINDER system, such as seismic interpretation or operator’s at the time of drilling, because well
geologic modeling, BHP has purchased packages logs will use this name, whereas the production
from other vendors, including Zycor Inc., Land- engineer usually wants to know the name of the
mark Graphics Corp. and Sierra Geophysics, Inc. operator of the producing well. Each well contains
BHP has also written routines to work with the dozens of data fields with similar possibilities for
FINDER system, such as a program that digitizes ambiguity. With about 47,000 wells in just the off-
horizon picks marked on paper displays of seis- shore portion of BHP’s Gulf of Mexico data base,
mic sections. the size of this problem can be daunting.
Through the FINDER system, BHP intends to Even when each data field is well defined, data
integrate different types of data and data from often have to be manually reorganized to fit the
many sources, and, most importantly, make format specifications of existing data loaders. If
those data available to the user via workstations this task is too large or cumbersome, custom
and X Window System terminals. It is hoped that data loaders must be written to convert the ven-
this flexibility will permit integration of geology dor data format into the database format. These
and geophysics in ways either not possible in the loaders are written either by FINDER Graphics or
past, or too difficult and, hence, impractical. BHP. Throughout the process, BHP draws on the
There have been several keys to BHP’s success expertise of its data administrators, who have
with the FINDER system. Early in the process, BHP learned from experience which data from which
acquired the FINDER Software Development vendors or areas need special attention.
Facility (source code for modifying the FINDER BHP’s long-term goal is to simplify the every-
system) and assigned dedicated in-house support day work flow, allowing professionals to be more
staff to the FINDER system. This has allowed BHP productive and more creative, by reducing the
to work closely with the FINDER development time spent searching, accumulating, requesting
staff and has given BHP the ability to add special- and moving data.
ized features and functionality to the system.
Each business unit at BHP Houston has a data
administrator whose sole responsibilities are
loading data, performing quality assurance, and
correcting and updating the FINDER data bases.
The data administrators work with the explo-
rationists on quality assurance, but the ultimate
responsibility for the integrity of data lies with
the explorationists.
Data loading and verification is an essential
function that BHP finds to be the most problem-
atic. Moving data from old data bases or mag-
netic tape into the FINDER system often requires
reformatting. Even new data can require consid-
erable manipulation because the organization
and definition of data may vary. For example,
commercial scout tickets from various vendors
often don’t agree on the name of the well opera-
tor. Vendors may furnish the name of the operator
at the time of drilling or during production. Both
53
section. Since there are now two copies of
the section, the two systems are said to be
loosely integrated. In a tightly integrated sys-
tem, processing and interpretation systems
would obtain data from a common data
store (possibly distributed across multiple
computing sites or machines), with a com-
mon file format, and would not require refor-
matting of data. Data handling is faster and
the border between computers and applica-
tions becomes invisible to the user.
There are good reasons why tight integra-
tion will be slow in coming. Tight integration
of applications written for independent sys-
tems requires modification of existing appli-
cations to permit them to use the common
data store. It also requires agreement on a
universal model for data exchange—one of
the long-term projects of POSC.
Loose integration between domains
requires a mechanism for moving data from
one to the other. The advantage of this
approach is that it does not require rewriting
programs. The GeoShare data exchange
standard, which uses RP66/DLIS as the
common data format, is one effort in this
direction, requiring only the writing of a link
to the GeoShare standard. Additionally, oil
companies are interested in prolonging the
value of their investment in proprietary sys-
tems. Therefore, the immediate efforts, such
Schlumberger’s geoscience data bus imple-
mentation using the GeoShare standard, are
for loose integration of independent sys-
tems, while a hopeful eye is kept on POSC’s
development of a universal system permit-
ting tight integration.
The shift toward workstations also calls
into question the means of data security and
control, which are often deeply rooted in
organization culture. Where should data be
kept—centralized, decentralized or in
regional data centers? Who controls data
quality? How can consistent quality of all
54
data be assured? Who will determine what
goes back in the data base and how is this
determined? Today, just as no two oil com-
panies have the same computer systems, no
two companies approach these challenges
the same way.
A leading debate in data management is
coping with simultaneous alteration of the
same data set. Imagine two geoscientists
working concurrently on different aspects
of the same project, using some of the same
well data. They have downloaded data
from the master data base, revised it, cor-
rected it and made interpretations. When
interpreted data are ready for storage on the
master data base, how is it decided which
version goes in?
There are various solutions. In BP USA,
automatic data revision is anathema. When
someone changes data in the master data
base, the computer produces a report, stat-
ing there are two versions of the same thing,
and listing the changes. It asks if those ver-
sions should be merged or supplanted. This
determination is made by the database man-
ager. The manager cannot be expert in
petrophysics, sedimentology, geophysics
and petrology in every province, but he or
she is still responsible for the integrity of the
master data base. The manager therefore sits
down with the geoscientists and, donning
the hats of diplomat, defender of data and
geoscientist, negotiates which version of
what to keep. BP finds that the successful
data manager must be high enough in the
organization to wield recognized authority,
yet close enough to the field to know the
area in question.
ARCO is installing a data management
system in its Plano, Texas, USA, facility and
expects to use a slightly different solution.
The data management plan calls for down-
loading project data from the mainframe to
the FINDER system. When an interpretation
is complete, it is screened at three levels
before being installed in the corporate data
base. First, the project manager approves all
formation top picks and performs a general
quality check. The appropriate operations
center then checks data pertaining to its
area, such as well location and geopolitical
data. Finally, with recommendations of the
previous two checkers attached, the inter-
pretation reaches the computing services
group in Plano, which then performs quality
checks as if it were vendor data, and installs
the interpretation in the corporate data base.
If the computing group questions the data, it
negotiates a solution with the project man-
ager, the operations group, or both.
Managing simultaneous alteration is one
challenge in this early stage of technology
transfer to workstations. The major chal-
lenge, in using workstations to reduce find-
ing and development costs, is development
and acceptance of a standard E&P data
model. Once this hurdle is cleared, only
minor obstacles in the path to seamless inte-
gration of geoscience data remain. —JMK
Acronyms
AAPG: American Association of
Petroleum Geologists
API: American Petroleum Institute
CD ROM: Compact disk read-only memory
CPU: Computer processing unit
DB: data base
E&P: Exploration and production
GIS: Geographic information system
OSF: Open Software Foundation
PC: Personal computer
PIDD: Petroleum Industry Data Dictionary
POSC: Petrotechnical Open
Software Corporation
PPDM: Public Petroleum Data Model
RISC: Reduced Instruction Set Computer
SQL: Structured Query Language
Oilfield Review
 A Niche for Enhanced Oil Recovery in the 1990s

Aging fields and dwindling prospects for finding new, large reserves are turning attention to improving

recovery from known oil fields. What is the status of enhanced oil recovery (EOR) and what role might it

potentially play in the next few years?

Larry W. Lake Traditional primary and secondary produc-
University of Texas tion methods typically recover one third of
Austin, Texas, USA oil in place, leaving two thirds behind. The
reasons for this are not difficult to under-
Raymond L. Schmidt stand. During the life of a well, there is
Chevron Oil Field Research always a point at which the cost of produc-
La Habra, California, USA ing an additional barrel of oil is higher than
the price the market will pay for that barrel.
Paul B. Venuto Production then halts. Under normal cir-
Mobil Research and Development Corp. cumstances, the well is abandoned, with
Princeton, New Jersey, USA 70% of the oil left in the ground.
Except for brief periods in which EOR was
economical, or perceived to be so, there
were good economic reasons not to nurse
every drop of oil from a well. Oil was easy
to find and another giant field was just
around the corner—the cost of a newly
found barrel of oil was far less than the cost
of an EOR incremental barrel. This situation
has begun to change, especially in North
America. Reserves in the aging oil fields of
the US and Canada are declining faster than
new oil is being added by discoveries. In
the US, for example, 70% of the approxi-
mately 500 billion barrels of oil discovered
were found during the earliest 20% of
drilling. About 130 billion barrels have been
produced to date and up to another 170 bil-
lion barrels are considered a long-term tar-
get for advanced EOR technology. The situ-
ation is similar in Canada. Given the
declining reserves and the low probability
of locating significant new fields, producers
sought additional oil in old reservoirs, mak-
ing North America a proving ground for
EOR techniques. Today, it is estimated that
North America produces more than half the
world’s EOR production (below ).
Research and development on many
fronts indicate that the risks of EOR are

Estimated Annual Wordwide EOR Produced Oil, B/D (x1000)

Country Thermal Miscible Chemical EOR Total %
USA 454 191 11.9 656.9 42
Canada 8 127 17.2 152.2 10
Europe 14 3 — 17.0 1
Venezuela 108 11 — 119.0 7
Other S. American 2 NA NA 17.0 1
USSR 20 90 50.0 160.0 10
Other (estimated ) 171* 280** 1.5 452.5 29
Total 777 702 80.6 1574.6 100

*Mainly Duri field (Indonesia) **Mainly Hassi-Messaoud (Algeria) and Intisar (Libya)

nEstimated annual worldwide production of oil by EOR.

(USSR data from Simandoux P and Valentin E: “Improved Recovery, Strategic Option or Not?” pre-
sented at the Offshore North Seas Conference, Stavanger, Norway, August 28-31, 1990; other data
from Oil & Gas Journal 88, no. 17 (April 23, 1990): 62-67.)

January 1992 55
 24
22
20
Oil production, billions of barells/yr
18
16
14
12
10
8
6 wells through the rock pores, sweeping the
4 is practical for light to medium crudes. Over
2 oil production
0 Some wells remain economical with a
76
nA slow, but steady increase in percent
of world oil produced by EOR. Percent of
oil produced by EOR has more than dou-
bled since 1982, when EOR oil accounted
for 0.9% of worldwide production. In 1990,
EOR accounted for 450 million—about
2%—of the 22 billion barrels of oil pro-
duced. (From Oil & Gas Journal data.)
being reduced and the potential for EOR
profitability increased (above ). Computer-
ized characterization of the reservoir, which
quantifies the physical characteristics and
dynamic behavior of a field, is becoming
one of the most important tools for
improved oil recovery (see “Reservoir Char-
acterization Using Expert Knowledge, Data
and Statistics,” page 25 ). Success of oil
recovery depends on applying the energy of
injected fluids in the right place, in the right
amount and at the right time—a strategy
that a well-constructed reservoir simulator
can help develop.
EOR is an imprecise term that historically
has been used to describe the third step (ter-
tiary recovery) in oil and gas production.
The term “improved oil recovery” (IOR) has
come into use to describe all recovery
methods other than natural (primary) pro-
duction, reserving the designation EOR for
those processes beyond simple waterflood
and gasflood—basically, recovery by injec-
tion of anything not originally in the reser-
voir (next page, top ). The three major EOR
methods are thermal (application of heat),
miscible (mixing of oil with a solvent) and
chemical (flooding with chemicals).
Primary recovery, in long accepted prac-
tice, is defined as production by natural
56
reservoir pressure, or pumping, until deple-
tion. Until the early 1940s, economics dic-
tated when a well was to be plugged and
abandoned, usually after recovery of 10 to
Worldwide 25% of original oil in place (OOIP).
oil production
Secondary recovery methods are gener-
ally used to repressure the reservoir and
drive out some of the remaining oil.
Because water is usually readily available
and inexpensive, the oldest secondary
recovery method is waterflooding, pumping
water through injection wells into the reser-
voir. The water is forced from injection
oil ahead of it toward production wells. This
Worldwide EOR
time, the percentage of water in produced
fluids—the water cut—steadily increases.
78 80 82 84 86 88 90
Year
water cut as high as 99%. But at some
point, the cost of removing and disposing of
water exceeds the income from oil produc-
tion, and secondary recovery is then halted.
Extensive waterflooding, which began in
the 1940s, within a few decades became
the established method for secondary oil
recovery, usually recovering about another
15% of OOIP. On average, about one-third
of OOIP is recovered, leaving two-thirds, or
twice as much oil as is produced, in the
ground after secondary recovery.1
Another recognized secondary recovery
technique is injection of a hydrocarbon-
based gas into an existing gas cap or
directly into the oil itself. Gas may be
injected over a considerable period of
time—up to a year—while producing wells
are shut in, until reservoir pressure is
restored and production resumed. Another
method is injection of gas to sustain pres-
sure during production. Gas injection
requires a nearby source of inexpensive gas
in sufficient volume.
While waterflooding is effective in nearly
all reservoirs, no single EOR technique is a
cure-all. Most reservoirs are complex, as are
most EOR processes. Efficient reservoir
management treats EOR as a high-cost,
high-risk but critical component of a com-
prehensive plan that spans primary recovery
through abandonment.2
Once preliminary reservoir information
has been assembled and used to select EOR
options, engineering project design usually
follows several steps.
•Laboratory studies test the proposed EOR
processes in corefloods with samples of
reservoir rock and fluids. These small,
one-dimensional flow tests in relatively
homogeneous media do not always suc-
cessfully scale up to reservoir dimensions.
But if the process fails in the laboratory, it
will more than likely fail in the field.
•Fluid-flow simulations, based on a geo-
logic reservoir model, can start with
assessment of primary and secondary
recovery, matching the production history
to determine residual oil and waterflood
recovery. Then EOR process-variable sen-
sitivities can be calculated, followed by
predictions of EOR recovery, incremental
production rate and payout economics.
Reservoir geologic models are always
constrained by sparse data, simplified
concepts of reservoir structure and
dynamics, inadequate data for history
matching and increasing computational
uncertainty as calculations are extrapo-
lated into the future. Consequently, pre-
dictions that cover years of EOR perfor-
mance may be seriously in error. In
addition, small-scale heterogeneities,
which are difficult to define, are critical to
the success of EOR.
•Usually, a pilot test of the proposed EOR
process is carried out to investigate a
novel technique or to confirm expected
performance before an expensive, full-
scale implementation. Ideally, the pilot
test is performed in an area that is geolog-
ically similar to the field and large enough
to be statistically representative of overall
heterogeneity. Monitoring and data acqui-
sition throughout pilot testing provide
information needed to plan a full-scale
commercial operation.
•For commercial operations, important
considerations are secure sources of water
and other injectants, storage and trans-
portation facilities (like pipelines), surface
processing, separation, recycling and
upgrading facilities, and environmental
and safety requirements.3
The same principles of EOR engineering
may not apply to offshore oil fields. Because
offshore wells tend to be highly deviated or
extended reach, the distance between them
is often greater than between onshore wells.
This extends the time between EOR initia-
tion and meaningful results and flattens the
recovery response. These effects complicate
process control and limit the number of
EOR techniques that may be applicable.
Greater spacing between wells also
increases the likelihood of undetectable het-
erogeneities between wells, impairing simu-
lations of well behavior. Because the num-
ber of wells that can be drilled from a
platform is fixed, infill drilling, often an
important strategy for both secondary recov-
ery and EOR, may not be possible. High
costs and extended time before EOR pro-
duction begins mean that offshore EOR pro-
jects must be planned and started early
Oilfield Review
 Oil Recovery Mechanisms enough so that production increases incre-
mentally before primary and secondary pro-
Conventional Primary duction begins to decline. Otherwise,
Oil Recovery marginal costs may be too high to sustain
profitability. However, this option must be
balanced against other risks: insufficient
Natural flow Artificial lift reservoir description at early stages of field
Secondary production and lack of time to acquire pilot
test results to evaluate the EOR process.4
Various mechanisms thwart recovery of
Waterflood Pressure much of OOIP after secondary recovery.5
Tertiary maintenance Reservoir geologic heterogeneities may
cause a large volume of mobile oil to be
bypassed and remain within a field. This is a
result of poor sweep efficiency when
EOR
Chemical Thermal Miscible Other: microbial, electrical, injected displacement water moves prefer-
chemical leaching, entially through higher permeability zones
mechanical (vibrating, toward the production well. Even in regions
horizontal drilling) that have been swept by large quantities of
water, residual, immobile oil is held in the
pore spaces by capillary forces.
Surfactant Polymer Caustic CO2 Miscible Inert gas Many techniques have been tried in the
solvent laboratory and field in hopes of recovering
this additional oil. All employ one or more
of three basic mechanisms for improving on
Steam stimulation Steam or In-situ waterdrive alone:
or cyclic steam hot water combustion •Increase the mobility of the displacement
injection
medium by increasing the viscosity of the
water, decreasing the viscosity of the oil,
or both.
Foam •Extract the oil with a solvent.
displacement •Reduce the interfacial tension between the
oil and water.
nOil recovery mechanisms. The three major EOR processes—thermal,
(Adapted from Venuto PB, reference 2 and Donaldson EC et al, reference 5.) miscible and chemical—are each subdi-
vided into several categories. Among the
60 nCost-perfor- three, thermal processes dominate, having
mance compari- the greatest certainty of success and poten-
son of major EOR tial application in about 70% of enhanced
50 methods. (Adapted
oil recovery worldwide. Thermal methods
Incremental oil cost, $/barrel

from Simandoux P,
Champlon D and also give the highest recoveries at the lowest
40 Velentin E: “Manag- costs (left ).
Surfactant
ing the Cost of The term miscible means the mixing of
Enhanced Oil Recov-
two fluids—for instance, oil and a solvent
30 ery,” Revue de L’Insti-
Thermal
tut Français du Pét- such as carbon dioxide [CO2]—into a sin-
role 45, no. 1 gle-phase fluid. It may also apply to a conti-
20
CO2 injection (January-February nuity between the oil and injected gas, due
1990): 131-139.) to a multiphase transition zone between the
Polymer
two. Use of miscible gasdrive has grown
10 rapidly in recent years, and today the
Waterflooding method accounts for about 18% of EOR
0
applications worldwide. It has been suc-
0 10 20 30 40 50 60 70 80 cessful at depths greater than 2000 ft [610
m] for CO2 and greater than 3000 ft [915 m]
Total recovery, (% OOIP)
for other gases.
1. Smith RV: “Enhanced Oil Recovery Update,” 4. Simandoux P and Valentin E: “Improved Recovery, a EOR chemical processes, such as surfac-
Petroleum Engineer International 60, 61, four-part Strategic Option or Not?” Presented at the Offshore tant (detergent) flooding, have tantalized
series (November 1988 to March 1989). North Seas Conference, Stavanger, Norway, August
28-31, 1990. the industry with promises of significantly
2. Venuto PB: “Tailoring EOR Processes to Geologic Envi-
ronments,” World Oil 209 (November 1989): 61-68. 5. Smith RV, reference 1. improved recovery. As yet, cost and techni-
3. Venuto PB, reference 2. Donaldson EC, Chilingarian GV and Yen TF (eds): cal problems have precluded them from
Schmidt RL: “Thermal Enhanced Oil Recovery—Cur- Enhanced Oil Recovery, II—Processes and mainstream application. Waiting in the
rent Status and Future Needs,” Chemical Engineering Operations, Developments in Petroleum Science.
Progress (January 1990): 47-59. Amsterdam, The Netherlands: Elsevier Science Pub-
lishers, 1989.
January 1992 Schmidt RL, reference 3. 57
 Oil viscosity, centipoise, at reservoir conditions wings are processes like microbial EOR
EOR method
0.1 1.0 10 100 1000 10,000 100,000 1,000,000 (MEOR) and some novel and exotic propos-
als; these await confirmation by lab and
Hydrocarbon field experimentation and evaluation before
miscible
taking their place as accepted practice.
Nitrogen and Each EOR process is suited to a particular
flue gas
type of reservoir. Because unexpected or
CO2 miscible unknown reservoir characteristics cause
most EOR failures, EOR begins with thor-
Surfactant/ ough geologic study. Technical rule-of-
polymer
thumb screening criteria are available to aid
Polymer preliminary evaluation of a reservoir’s suit-
ability for EOR (left ). After these criteria are
Alkaline applied to a prospect, stringent economic
analysis follows, generally through repeated
Fireflood reservoir simulations6 (see “Trends in Reser-
voir Management,” page 8 ).
Steamdrive
Thermal
Thermal methods are the main means of
Permeability, md recovering heavy oils, those with gravity
EOR method
10 100 1000 10,000 less than 20° API, representing viscosities of
200 to 2000 centipoise (cp). Such heavy oils
Hydrocarbon Not critical if uniform
miscible generally don’t respond significantly to pri-
mary production or waterflooding so initial
Nitrogen and Not critical if uniform
flue gas oil saturation is typically high at the start of
a thermal recovery project. The principle of
CO2 miscible High enough for good injection rates thermal recovery is simple: increasing the
oil’s temperature dramatically reduces its
Surfactant/
polymer viscosity, improving the mobility ratio (next
page, left ). The two primary methods of
Polymer heating reservoir oil are injection of fluid
heated at the surface or production of heat
Alkaline directly within the reservoir by burning
some of the oil in place.
Fireflood Although the idea of heating reservoirs
dates back more than 100 years, large-scale
Steamdrive steamdrive projects began in heavy oil fields
in the US in the early 1950s and were fol-
lowed shortly by projects in The Nether-
Depth, ft lands and Venezuela. A relative of steam-
EOR method
2000 4000 6000 8000 10,000 drive is cyclic steam injection, also called
Hydrocarbon steam soak or “huff and puff.” It was
miscible Deep enough for required pressure discovered accidentally in 1960 during a
Nitrogen and Venezuelan recovery project. Cyclic steam
flue gas Deep enough for required pressure recovery uses a single well for both injec-
tion and production. Steam is injected into a
CO2 miscible Deep enough for optimum pressure well for several days or weeks, then the well
Surfactant/ is shut in for several days to a month or
polymer Limited by temperature more, the soak period. After this, the well is
produced for up to six months, after which
Polymer Limited by temperature the process is repeated. The steam heats the
rock and fluids surrounding the wellbore
Alkaline and also provides some drive pressure; by
the time production resumes, the steam has
Fireflood Deep enough for required pressure condensed and oil and water are produced.

Steamdrive Normal range 6. Taber JJ and Martin FD: “Technical Screening Guides
for the Enhanced Recovery of Oil,” paper SPE 12069,
presented at the 58th SPE Annual Technical Confer-
Crudes up to 20° API ence and Exhibition, San Francisco, California, USA,
October 5-8, 1983.
Lighter crudes, 30° API Venuto PB, reference 2.
7. Taber JJ and Martin FD, reference 6.
Good Possible Fair Difficult Not feasible Smith RV, reference 1.
Donaldson et al, reference 5.
nSelection of EOR techniques by oil viscosity, permeability and depth. (Adapted from Venuto PB, reference 2.
Taber JJ and Martin FD, reference 6.)
Schmidt RL, reference 3.
The advantage of “huff and puff” is the rela-
tively short “huff” time so that the well is
producing most of the time. The disadvan-
tage is that only the reservoir near the well-
bore is stimulated. This method therefore
often has to be followed by continuous
steam injection to drive oil toward a sepa-
rate production well. Reservoir pressure,
dictated by depth, imposes a limit on steam-
flooding: higher pressures require more fuel
to generate higher temperatures on the sur-
face to produce saturated steam needed for
efficient steamflood. The higher temperature
also leads to greater heat losses.
The recovery mechanism of steamflood-
ing is complex. Besides viscosity reduction,
the second most important recovery mecha-
nism is steam distillation of lighter compo-
be insulated, as well as downhole tubing
below 1500 ft [460 m]. Thermal expansion
may damage downhole equipment and
cause cement failure. Steam is very reactive,
causing pipe corrosion and scaling, miner-
alogical dissolution or reprecipitation, clay
swelling and changes in permeabilities.
Steam is more mobile than oil, overriding
the oil and channeling through thief zones;
no satisfactory methods have been found to
improve sweep efficiency. Steam is usually
generated by burning natural gas, but if
lease crude is used—typically at about one
barrel of crude for three or four barrels of oil
recovered—the air pollution byproduct may
be costly to control.
The other significant thermal recovery
process, in-situ combustion or fireflooding,
attempts to recover oil by igniting a portion
of the in-place crude by injecting air or oxy-
gen or by chemical or electrical means
(below ). Fireflooding appears attractive: it is
thermally more efficient than steam, has no
depth restriction and is well suited to rela-
tively thin (less than 25 ft [8 m]) reservoir
sands. But in practice it is not so simple. Cap-
ital costs are high, the process is extremely
complicated and difficult to predict or con-
trol, and operational problems from the high
temperatures include cement failures, sand-
ing/erosion, corrosion at both injection and
production wells because of oxygen and
moisture, and high gas production rate.
Despite some successes and more than 30
years of laboratory and field trials, in-situ
combustion applications have not increased.7

10,000

10°
AP
I
Viscosity, centipoise

100

20°
AP
I
10
30°
AP
I

2
100°F 200°F 300°F
38°C 93°C 149°C
Crude oil temperature

nThe empirical relationship of crude oil

viscosity and temperature depends on oil
composition. (Adapted from Lake LW, refer-
ence 10.)

nents of the oil which form a solvent bank

ahead of the steam. Other factors are ther-
mal expansion, solution gasdrive and misci-
ble and emulsion drive. Steamflooding has
been tested in light oil reservoirs where the
viscosity mechanism plays a minor role and
distillation to solvents predominates. Some
field tests of light oil steamfloods have
shown virtually 100% recovery of oil from
zones reached by the steam.
All is not roses with steamflooding, as
numerous technical problems may com-
bine to make a recovery project uneco-
nomical, inefficient or even dangerous.
Heat losses in surface equipment, in the
wellbore, in rocks around the reservoir, in
connate water and in the gas cap may
defeat the process. Surface equipment must
nChanges in a 3D seismic section for “preburn,” “mid-
burn” and “postburn.” The bright spot—an increase in
envelope amplitude, marked by arrowheads—appeared
at midburn time and expanded by postburn time. Below
the bright spot is a dim spot, caused by a decrease in
envelope amplitude. The bright spot is caused by
increased gas saturation along the top of the reservoir
boundary. The dim spot shows good correlation with the
burn volume in distribution and direction, as deter-
mined by core analysis. From studies like these, seismic
data can be used to map an estimated burn thickness.
(Courtesy of the Society of Exploration Geophysicists. From
Greaves RJ and Fulp TJ: “Three-dimensional Seismic Monitoring
of an Enhanced Oil Recovery Process,” Geophysics 52, no. 9
(September 1987): 1175-1187.)
59
Miscible
The fastest growing EOR process, miscible
flooding, uses a solvent that mixes fully with
residual oil to overcome capillary forces and
increase oil mobility. Displacement effi-
ciency nears 100% where the solvent con-
tacts the oil and miscibility occurs. The
numerous successful solvents include liqui-
fied petroleum gas (LPG), nitrogen, CO2, flue
gas (mainly nitrogen and CO2) and alcohol.
A sufficient supply of a particular solvent
and the value of injected hydrocarbons,
like natural gas, have considerable impact
on the choice and economics of miscible
flood projects. In Canada, for example, the
abundance of natural gas makes that the
choice, while the US has huge reserves of
CO2 in the western states that have been
tapped for EOR in the Permian Basin fields
of west Texas.
Miscible displacement EOR can be subdi-
vided into three significant processes: misci-
ble slug, enriched gas and high-pressure
lean gas, including CO2. For each of these
processes, there is a range of pressures, or
depths, temperatures and oil gravities neces-
sary to achieve and maintain miscibility.
Solvents have much lower viscosity than oil,
so reservoir stratification—vertical and hori-
zontal permeability contrast—strongly
affects sweep efficiency. Early breakthrough,
bypassing substantial amounts of oil, and
viscous fingering problems have plagued
many field projects involving the four misci-
ble displacement processes:
•In the miscible slug process, a slug of liq-
uid hydrocarbons, about half the reservoir
pore volume (PV), is injected and mixes
with the oil on contact. This is followed
by water or chase gas to push the slug
through the reservoir.
•For the enriched, or condensing, gas pro-
cess, 10 to 20% PV of natural gas,
enriched with intermediate molecular
weight hydrocarbons, is injected, fol-
lowed by lean gas and, in some cases,
water. At the proper reservoir pressure,
the C2–C6 components transfer from the
enriched gas to the oil, forming a miscible
solvent bank.
8. Taber JJ and Martin FD, reference 6.
Smith RV, reference 1.
Donaldson et al, reference 5.
Stosur G, Singer M, Luhning R and Yurk W:
“Enhanced Oil Recovery in North America: Status
and Prospects,” Energy Sources 12, no. 4 (1990):
429-437.
9. Taber JJ and Martin FD, reference 6.
Smith RV, reference 1.
10. Taber JJ and Martin FD, reference 6.
Lake LW: Enhanced Oil Recovery. Englewood Cliffs,
New Jersey, USA: Prentice-Hall, Inc., 1989.
60
Injector
one
an k
le z
CO2
Oil b
cib
Mis
a
CO2
b
CO2
c
•Similarly, in vaporizing gasdrive, lean gas
is injected at high pressure, 3000 to 6000
pounds per square inch gauge, and C2–C6
components from the oil are vaporized
and mutually exchanged between the oil
and gas during multiple contacts, eventu-
ally forming a miscible slug.
Carbon dioxide is a special case of high-
pressure miscible recovery. This gas is
highly soluble in crude oil, swelling the oil
and reducing its viscosity, while simultane-
ously extracting lighter hydrocarbons by
vaporization. The displacing gas front,
enriched by vaporized hydrocarbons
through multiple contacts, forms a miscible
slug as long as minimum miscibility pres-
sure (MMP) is maintained. Since CO2 can
extract heavier components, it is miscible
with crude oils having fewer C2–C6 compo-
nents. Carbon dioxide has a lower MMP
than natural gas, nitrogen or flue gas, and
therefore can be applied in shallower (lower
pressure) wells.
A major problem with miscible gasflood
EOR is the adverse mobility ratio caused by
the low viscosity of the typical injectant gas
compared to oil, perhaps by one or two
11. Bondor PL: “Dilute Surfactant Flooding for North Sea
Applications—Technical and Economics Considera-
tions,” Proceedings of the 6th European Symposium
on Improved Oil Recovery, Stavanger, Norway, May
21-23, 1991, volume 2: 749-758.
12. Taber JJ and Martin FD, reference 6.
Smith RV, reference 1.
Venuto PB, reference 2.
Lake LW, reference 10.
13. Donaldson et al, reference 5.
Simkin EM and Surguchev ML: “Advanced Vibro-
seismic Technique for Water Flooded Reservoir
Stimulation, Mechanism and Field Tests Results,”
Proceedings of the 6th European Symposium on
Improved Oil Recovery, Stavanger, Norway, May
21-23, 1991, volume 1: 233-241.
Producer
nPhase behavior
and flow dynamics
in miscible flooding
Residual oil (a) for ideal perfor-
mance, (b) during
the influence of
fluid density and
(c) in the setting of
viscosity contrasts,
which produce fin-
gering of the EOR
gas into oil.
Residual oil (Adapted from Venuto
PB, reference 2.)
Residual oil
orders of magnitude. The result is an unsta-
ble front between the gas and oil which
allows viscous fingers to form and propa-
gate through the displaced fluid, leaving
much of the hydrocarbon uncontacted
(above). Today, the primary means of attack-
ing this problem is the water-alternating-gas
(WAG) technique. In this process, water-
flood and gasflood are alternated, with the
design parameters being timing and the
ratio of water to gas. WAG claims the
virtues of decreasing the mobility of the gas,
maintaining pressure and saving operating
costs by substituting inexpensive water for
relatively expensive gas. Ideally, the gas pro-
vides miscibility and the water improves
sweep efficiency. However, one study of
WAG in 15 CO 2 flood projects showed
lower recoveries than for cases using a sin-
gle injection of CO2 followed by a water-
flood. Gravity segregation between water
and gas is thought to compromise the effec-
tiveness of the WAG process.8
Performance of many EOR techniques
suffers from differences in mobility between
the EOR product and the oil it is supposed
to recover. One possible solution to this is
foam, which is dispersed gas bubbles in a
liquid. Foam can reduce reservoir gas
phase permeability to less than 1% of its
original value. Typical surfactant-based
foams can last indefinitely. Such foams
have been used for mobility control misci-
ble gas injection and steamfloods with
mixed results. Problems include abnormally
high injector-to-producer pressure differen-
tials required for propagation, rapid
changes in foam stability and quality as it
migrates away from the injection well and
foam breakdown in small pores.
Chemical
Chemicals used in EOR include polymers,
surfactants and alkalis. All are mixed with
water and, occasionally, other chemicals
before injection. Broadly speaking, targets
for chemical recovery are crudes in the
range between the heavy oils recovered by
thermal processes and light oils recovered
by miscible gas injection.
Polymer flooding is simply an accompani-
ment to waterflooding. It is the most com-
monly used chemical enhancement process
since it is easy to apply and requires rela-
tively small investment. Although polymer
flooding increases recovery by a modest
amount, on the order of 5%, it can yield
solid profits under the right circumstances.
Adding high molecular weight polymers
increases the viscosity of water and, with
some polymers, reduces the aqueous phase
permeability without changing the relative
permeability to oil. This can greatly improve
waterflood volumetric sweep efficiency.
Polymer concentrations are 100 to 1000
parts per million (ppm) and treatment may
require injection of 15 to 25% PV over sev-
eral years followed by a typical waterflood.9
Cross linking, or gelling, polymers in situ
with metallic ions can augment perfor-
mance in sweep profile control—helping to
plug high conductivity zones or minor frac-
tures that degrade sweep efficiency. The
main drawbacks to the use of polymers are
their high cost and the low injection rate
caused by high viscosity (which impacts
economic rate of return), degradation at
higher temperatures, intolerance to high
salinity, polymer deterioration from shear
stress imparted by pumping, flow through
tubulars and perforations, and long-term
instability in the reservoir environment.
Surfactant flooding, also known as deter-
gent, micellar-polymer or microemulsion
flooding, uses low concentrations of surfac-
tants in water to reduce the interfacial ten-
sion between oil and water. Although the
idea has been around since the 1920s, seri-
ous research and field trials did not start
until 20 years ago, and uniformly profitable
field performance is still elusive. Of all EOR
processes, surfactant floods may be riskiest,
involving the most difficult design decisions,
requiring large capital investments, and
being strongly affected by reservoir hetero-
geneities. For these reasons, and as a result
of marginal field performance, interest in
surfactant flooding has declined.
A surfactant flood must be designed for a
specific crude oil in a specific reservoir tak-
ing into account such factors as salinity,
temperature, pressure and clay content.10
Generally, multiple slugs are used. Surfac-
January 1992
Injector
Polymer-
Flood- thickened
water fresh
water
nSurfactant-polymer flooding, showing the surfactant slug on its way toward the pro-
ducer, pushed by polymer-thickened fresh water and floodwater from the injector well.
(Adapted from Donaldson EC et al, reference 5.)
tant performance is optimal over a narrow
salinity range and is subject to adsorption
and retention through ionic exchange with
reservoir rocks. To avoid these problems,
the first slug may be a preflush water solu-
tion. However, it is often ineffective. The
second slug, perhaps 10 to 30% PV, con-
tains the surface active agent (5 to 10% by
volume), hydrocarbons, electrolyte and
cosolvent, usually alcohol. This is followed
by a slug of polymer-thickened water for
mobility control and, finally, typical water-
flood (above ).
Koninklijke-Shell Exploratie en Produktie
Laboratorium in Rijswijk, The Netherlands,
developed a surfactant capable of being
injected in low concentrations (less than 1%
by volume) in seawater in an offshore envi-
ronment during late secondary recovery
waterfloods.11 Shell’s analysis of the eco-
nomics of implementing this technique in a
moderately sized North Sea field—100 mil-
lion stock-tank barrels of OOIP—is reveal-
ing. Shell found that high front-end capital
costs would include building a chemical
plant to manufacture the product and addi-
tional dedicated floating facilities to handle
injection and production. The surfactant
cost alone was $18 per barrel of incremen-
tal oil recovered, undiscounted at an esti-
mated 9 kilograms (kg) [20 lb] of product
injected for each barrel of oil. Projections
showed that front-end investment of $750
million would be incurred in front-end capi-
tal costs and chemicals injected prior to the
production of significant amounts of incre-
mental oil. The analyzed technical cost
came to $90 per barrel assuming a 15% dis-
count rate (cost of capital).
The caustic or alkali flooding process
relies on a chemical reaction between the
caustic and organic acids in the crude oil to
produce in-situ surfactants that lower inter-
facial tension between water and oil. Other
mechanisms that may enhance recovery are
changing rock from oil-wet to water-wet,
which lowers interfacial tension, and emul-
Producer
Surfactant
slug
Oil-water Residual oil and
bank resident brine
sification, which lowers viscosity. Alkalies
such as sodium or potassium hydroxide are
used. Caustics can react strongly with min-
erals in the connate water and with the
reservoir rocks to the detriment of the pro-
cess. This complex process is poorly under-
stood; it has had some technical successes
but no great financial successes.12
Besides the mainstream techniques
discussed, many interesting ideas have
been proposed and tried. Today, most of
these are not economical or have some
technical flaw:
•Microbial EOR injects bacteria and nutri-
ents into the reservoir where the bacteria
multiply and biochemically manufacture
polymers and surfactants; this technique is
still unproved, although some successes
have been reported.
•Thermal enhancements include downhole
steam generation, mixing gas and solvents
with steam, downhole radio frequency
(induction) heating, nuclear steam genera-
tion and jet leaching by high-pressure hot
water with additives.
•Other ideas: injection of direct current for
electro-osmotic effect, chemical alteration
of micellar configuration of in-situ
petroleum, earthquake simulation using
high-powered surface vibrators (successful
in the USSR13).
EOR technology does not seem to be on
the threshold of any dramatic technological
breakthroughs. Instead progress will proba-
bly come through gradual evolution, stimu-
lated by growing motivation to recover
more oil from known fields. The major con-
tribution to EOR is likely to be from the con-
stantly improving art of reservoir characteri-
zation for predicting EOR response and
from horizontal drilling. —SM
61