SPE 99833

Reservoir Modeling for Mature Fields—Impact of Work Flow and Upscaling on

Fluid-Flow Response
W.S. Meddaugh, SPE, Chevron Energy Technology Co.

Copyright 2006, Society of Petroleum Engineers

2 times the data driven range value also had little effect on
This paper was prepared for presentation at the SPE Europec/EAGE Annual Conference and reservoir response.
Exhibition held in Vienna, Austria, 12–15 June 2006.
An important issue surrounds the impact of up-scaling on
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
fluid flow response. Vertical up-scaling by factors commonly
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to used for full field simulation models has little impact on fluid
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at flow response based on studies of the New Mexico carbonate
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
reservoir and the shallow water clastic reservoir in Venezuela.
for commercial purposes without the written consent of the Society of Petroleum Engineers is However, areal up-scaling of models generated using a very
prohibited. Permission to reproduce in print is restricted to an abstract of not more than
300 words; illustrations may not be copied. The abstract must contain conspicuous fine 50 foot areal grid significantly alters the fluid flow
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836 U.S.A., fax 01-972-952-9435.
characteristics and warrants additional study.

Abstract Introduction
Scoping studies using data from three mature fields suggest This paper presents the results of several small studies done
that simple workflows that use only essential stratigraphic and over the past six years or so that provide insight into how to
facies constraints are as good in capturing overall reservoir efficiently build static models for mature fields that preserve
fluid flow response as complex, highly constrained workflows those geological features (e.g. heterogeneity) critical to fluid
that use detailed stratigraphic and facies constraints. Thus, flow. Numerous recent papers have addressed issues critical
considerable time and cost saving may be realized during to mature field characterization, static and dynamic modeling,
initial model building and updating if simple, but appropriate, and management.1-5 A variety of papers have been published
workflows are used. that address specific aspects of how best to capture critical
The reservoirs studied include a Permian-age carbonate geological heterogeneities in earth models prior to and during
reservoir in New Mexico, an Upper Miocene deepwater clastic upscaling for dynamic simulation.6-14 The primary focus of
reservoir in California, and an Eocene-age shallow water this paper is to compare the fluid flow response of dynamic
clastic reservoir in Venezuela. Two dimensional cross section models derived from static models generated using stochastic
models of the deepwater clastic reservoir showed that workflows that utilize differing amounts of geological
cumulative production and water breakthrough times were complexity (constraints) using data from a carbonate and two
essentially the same for models using the two major clastic reservoirs. Although not the primary focus of this
stratigraphic picks as for models constrained by 12 detailed paper, results from a study of vertical and areal upscaling of a
stratigraphic picks. Three dimensional streamline simulation carbonate reservoir are also presented.
was used to demonstrate that adding facies and rock type
constraints had little impact on recovery factors for a Carbonate Reservoir Study
carbonate reservoir scoping project area consisting of 25, 5- A scoping study was done in order to assess the effect of
spot waterflood patterns. Likewise, a very complex workflow incorporating varying amounts of geological detail or
for the shallow water clastic data set from Venezuela constraints using a data-rich portion of the Eunice Monument
constrained by eight facies and 16 detailed stratigraphic picks South Unit (EMSU) reservoir is located in Lea County, New
yielded the same reservoir response as a simple, two facies, Mexico about 25 miles south of the city of Hobbs (Fig. 1).
and four major stratigraphic picks constrained workflow. The field was discovered in 1929 and has produced about 15%
These studies suggest that for reservoirs with moderate to high of the estimated 1000 MMSTB OOIP from over 250 wells as
net to gross (>30-40%) or with small differences in the of early 2001, when this study was completed.
porosity vs. permeability trends of facies/rock types that
simple geological modeling workflows are adequate for Geological Setting
subsequent fluid flow simulation. Models generated using the The productive interval at EMSU is the Gaudalupian-age
shallow water clastic data sets and evaluated using three (Permian) Grayburg Formation. Within the study area, the
dimensional streamline simulation showed that varying the reservoir averages about 250 feet thick. The reservoir depth
semivariogram range parameters by factors between 0.25 and (top of Grayburg) is about 3900 feet. The average reservoir
porosity is about 8% with an average reservoir permeability of
2 SPE 99833

about 1-2 md. A structure map for the EMSU reservoir is specific transform equations developed from the available core
given in Fig. 2. The geology and stratigraphy of the EMSU data. Fig. 9 gives an example areal and vertical cross section
reservoir is discussed by Lindsay15 and Garber et al.16 from one of the “simple” workflow porosity realizations.
Meddaugh17, 18 presented summaries of the static and dynamic The second workflow investigated included the
modeling work described in detail below. depositional facies (shoal vs. lagoon) as an additional model
In the Permian Basin region of New Mexico and Texas constraint. Depositional facies maps for each of the four
the Grayburg Formation is a prolific hydrocarbon producing stratigraphic intervals were generated using the cored data
sequence (Fig. 3). The Grayburg reservoirs are located along from wells in and immediately surrounding the study area.
the east and northwest margins of the Central Basin Platform These maps along with facies logs for the wells with core data
and along the southern margin of the Northwest Shelf (Fig. 1). were used as constraints to distribute facies regions in the
The Grayburg Formation consists of shallowing-upwards reservoir models using sequential indicator simulation (SIS).
parasequences of interbedded carbonates and minor Porosity was then added using SGS constrained by layer and
siliciclastics. At EMSU the carbonates are completely facies region. Permeability was next added using facies
dolomitized as is the carbonate matrix of the generally very specific transforms derived from the core data. Overall, there
thin siliciclastic intervals. The EMSU carbonates were is little difference between the average porosity in the shoal
deposited on a shallow carbonate shelf that deepened to the facies and the lagoon facies (0.101 vs. 0.094, respectfully) but
west towards the Delaware Basin. a significant difference in the geometric average permeability
As shown on the structure map in Fig. 2 and the of the shoal facies and the lagoon facies (5.4 vs. 1.2 md,
schematic facies map given in Fig. 4, the EMSU reservoir respectively) as shown in Fig. 10. Note that the study area is
consists mainly of shoal and lagoon depositional facies. about 80% shoal facies by volume and that this accounts for
Dolostones deposited in the higher energy shoal complexes the higher average porosity and permeability of the study area
are mainly ooid grainstones and mud-poor packstones. In the compared to the EMSU reservoir overall. Fig. 11 provides
updip low energy back-shoal or lagoonal facies setting, the example areal and vertical cross section from one of the
dolostones consist mainly of mud-rich packstones and “facies-based” workflow realizations.
wackestones; mudstones are relatively rare. The shoal and The third workflow investigated adds lithology or rock
lagoonal depositional environments interfinger with one type as a constraint. This workflow is referred to as the
another as shown in the schematic cross section given in “lithology-based” workflow. Seven lithology or rock types
Fig. 5. were distinguished based on the available core descriptions:
Individual shallowing-upwards parasequences thicken (1) siliciclastic-rich carbonate (sandstone); (2) mudstone; (3)
downdip into the shoal complex. Updip the individual wackestone; (4) mud-rich packstone; (5) mud-poor packstone;
shallowing-upwards parasequences become thinner and more (6) grainstone; and, (7) rudstone. Table 3 summarizes the
numerous. Sets of parasequences stack to form six porosity and permeability by lithology-type. The lagoonal
stratigraphic units as shown in the cross sections shown in depositional facies are dominated by mud-rich rocks while the
Fig. 6. The zonal markers are dolomitic sandstones that are shoal depositional facies is dominated by mud-poor packstone
relatively easy to correlate across the field. The well sorted and grainstone (Table 4). Using the depositional facies
and very fine grained sand was transported by aeolian distributions generated as the first step of the “facies-based”
processes during times of low sea level and reworked into workflow as a starting point, the next step of the “lithology-
shallow marine blanket sands at times of high sea level. The based” workflow was to distribute the lithology-types by
siliciclastics have generally low permeability and tend to act multi-binary SIS constrained by depositional facies. Porosity
as field wide vertical barriers and baffles to fluid flow. and permeability were next distributed by SGS and transform
equation specific to each lithology type and stratigraphic layer.
Static Model Workflow Description Fig. 12 provides example areal and vertical cross section from
Three workflows with increasing amount of geological detail one of the “lithology-based” workflow realizations.
(constraints) were used in the EMSU carbonate reservoir The porosity histogram is “conserved” during the
scoping study. All workflows used a structural and stochastic modeling process. Only the distribution and
stratigraphic framework constructed from the top of Grayburg connectivity of the porosity is affected by the model
structure map given in Fig. 2 and the following sequence workflow. Table 5 gives the average porosity and average
stratigraphic horizons: GB (top of Grayburg Formation), Z2, permeability of the EMSU reservoir models generated by the
Z2A, Z3, and Z4. All of the stochastic models were generated three workflows. Note that conservation of the porosity
using a 50 foot areal cell size and vertical grid definition given histogram produces models with essentially the same OOIP.
in Table 1. The models consisted of 9,374,400 cells (186 x
126 x 400 layers). Note that the models were generated for Model Up-scaling
only a portion of the EMSU reservoir (Fig. 7). Fifteen Prior to dynamic evaluation, all of the realizations were up-
realizations were generated using each of the three workflows. scaled using a proprietary flux-based software package. The
In the “simple” workflow the model porosity was static models were vertically up-scaled from 400 layers to 18
distributed within each of the four stratigraphic intervals by layers and from a 50 foot areal cell size to a 100 foot areal cell
sequential Gaussian simulation (SGS) constrained by the layer size. The up-scaled models had a total of 105,462 cells (93 x
appropriate well data and semivariograms as summarized in 63 x 18 layers).
Table 2 and Fig. 8. Fifteen porosity model realizations were
generated and permeability added to each model using layer
SPE 99833 3

Three Dimensional (3D) Streamline Model Description Geological Setting

3D streamline simulation models were used to evaluate the In the LL-652 CFB reservoir, the studied stratigraphic interval
fluid flow characteristics of the EMSU reservoir models (Fig. 17) is part of the large scale transgressive-regressive
generated by the three workflows. The 3D streamline model stratigraphic cycle that characterizes the Misoa Formation20,21.
represents the mathematical equivalent of a laboratory core This large scale cycle is punctuated by smaller-scale
flood. The 3D streamline model consisted of 20; five-spot sequences which represent higher frequency regressive-
waterflood patterns with production wells and injection wells transgressive cycles within the succession. Within the smaller-
spaced about 1400 feet apart. All field parameters selected for scale sequences, successive lowstand (LST), transgressive
the 3D streamline models, including well spacing, were based (TST) and highstand (HST) systems tracts can be recognized
on then current field practice. Producers were given minimum and correlated throughout the entire LL-652 area. This
flowing bottomhole pressures of 500 psia. Injectors were sequence hierarchy is the basis for the sequence stratigraphic
given maximum bottomhole injection pressures of 3500 psia, subdivision of the C-4-X.01 reservoir21 (Fig. 18).
to simulate approximate field operating conditions and to stay The LL-652 CFB C-4-X.01 reservoir interval is made up
below the 4000 psia fracture pressure. The wells were given of transgressive-regressive (T-R) sequences that were
pressure rather than rate constraints to allow the permeability deposited in a shallow, tidally dominated marine embayment
distributions to drive the 3D streamline simulation results. with a central estuarine mud flat. High quality reservoir
Fluid PVT property and rock relative permeability data were sandstones were deposited mainly during transgressive phases
obtained from the then most recent EMSU history matched that are typically estuarine and show an overall fining upward
finite-difference flow simulation model.18 Fig. 13 shows the sequence. The high quality sandstones include bay-head delta
variation of water saturation (Sw) with time over the eight and the dominant estuarine sands. These upwardly fining,
year simulation run. non-marine to marine sediments show both a landward
stepping depositional pattern that was confined to valleys and
EMSU Study Results and Discussion fluvial/estuarine deposition. The typical facies found in this
Fig. 14 shows the cumulative oil production vs. time plots phase are estuary distributary channels, associated tidal and
for the 15 realizations derived from the three workflows. non-marine muddy flats, and tidal sand flats. As the facies
Table 6 and Fig. 15 show the 3D streamline results in terms of migrated landward, the sandier facies were overlain by the
recovery factor (RF). Note that there is little difference muddier offshore facies. The regressive phases are composed
between the workflow results suggesting that for the EMSU of coarsening upward sands deposited in a tidally influenced
reservoir the additional geological constraints of depositional prograding bay-deltaic sequence. The characteristic deposits
facies and lithology-types do not alter the field wide fluid flow are sand shoals, bars, and flats22.
characteristics as compared to the “simple” workflow. No Petrophysical characterization indicates that depositional
effort was made in this study to assess individual well facies exert the primary control over reservoir properties.
performance as this scoping study was done to assess the Reservoir flow units are associated with the amalgamated
amount of detail needed to make field-scale decisions (e.g. distributary channels and tidal sand bars that define the LST of
pattern vs. peripheral water flood) rather than for detailed each sequence. Permeability baffles and barriers occur within
reservoir management at the single well scale. One should the tidally-influenced TST and HST deposits21.
expect that the different workflows, indeed the different
realizations from a given workflow, may give model results Static Model Workflow Description
that vary considerably at the single well scale. Four workflows with increasing amount of geological detail
(constraints) were used in the LL-652 CFB clastic reservoir
Shallow Water Clastic Reservoir Study scoping study. All of the stochastic models were generated
Another scoping study was done to investigate the effect of using a 35 m areal cell size and vertical grid definition given
geologic detail (constraints) on fluid flow results. Unlike the in Table 7. The models consisted of 6,148,800 cells (126 x
EMSU study, which was largely academic, the scoping study 122 x 400 layers). The same structural and stratigraphic
undertaken using data from the LL-652 reservoir (Lagunillas framework was used for all four workflows except that the
Field) in the Maracaibo basin in Venezuela was done to most complex workflow (referred to as the “complex facies”
support a business decision to use a particular workflow. Fig. workflow) used additional sequence stratigraphic detail.
16 provides a location map for the LL-652 reservoir. The Fifteen realizations were generated using each of the four
study uses data from a small area of the central fault block workflows described below.
(CFB) portion of the LL-652 reservoir as highlighted on the The “simple” workflow used for the LL-652 CFB study is
structure map for the Eocene C-4-X.01 member of the Misoa analogous to the “simple” workflow used for the EMSU study.
Formation is also given in Fig. 16. Cumulative oil production Porosity was distributed within each of the four stratigraphic
up to the time of this study (2002) was about 84 MMSTB intervals by SGS constrained by the layer appropriate well
(about 13% of the estimated OOIP). The productive portions data and semivariograms as summarized in Table 8 and Fig.
of the studied interval have an average porosity of 0.14 and 19. Fifteen porosity model realizations were generated and
average permeability near 40 md. Non-reservoir sections have permeability added to each model using layer specific
an average porosity of about 0.04 and average permeability of transform equations developed from the available core data.
about 0.3 md. The net/gross of the C-4-X.01 reservoir is The second LL-652 CFB workflow tested is referred to as
about 30%. The studied reservoir interval is about 300 m the “RnR-based” workflow. The RnR-based workflow is
thick. analogous to the “facies-based” workflow used in the EMSU
4 SPE 99833

study except that the facies distributed by SIS are the 3D streamline models, including well spacing, were based
petrophyically defined reservoir and non-reservoir “facies”. on then current field practice. Producers were given minimum
Table 9 shows the difference in porosity and permeability flowing bottomhole pressures of 3200 psia and injectors were
between the reservoir and non-reservoir facies. The given maximum bottomhole injection pressures of 3700 psia
petrophysical definition of the reservoir and non-reservoir to approximate field operating conditions. Initial model
“facies” is based on porosity and shale volume cutoffs. The conditions were: bubble point pressure = 3165 psia, initial
reservoir “facies” correspond to the bay-head delta and the reservoir pressure = 3510 psia at a datum depth of 7300 feet
estuarine sand facies in the depositional interpretation and an initial oil water contact at 7900 feet subsea. The wells
discussed above. Porosity was added by SGS and the were given pressure rather than rate constraints to allow the
permeability distributed by a collocated cokriging with SGS permeability distributions to drive the 3D streamline
algorithm using the porosity as secondary data. The porosity simulation results. Fluid PVT property and rock relative
distribution was constrained by the reservoir and non-reservoir permeability data were obtained from the then most recent LL-
“facies” and by stratigraphic interval. 652 CFB history matched finite-difference flow simulation
The third workflow examined using the LL-652 CFB data model23,24. Fig. 21 shows the variation of water saturation
set is the “lithology-based” workflow; analogous to the (Sw) with time over the eight year simulation run.
lithology-based workflow used for the EMSU study. The
seven lithology types identified in cored intervals as well as by LL-652 CFB Study Results and Discussion
log signature in the other wells in the study area are listed in Fig. 22 shows the cumulative oil production vs. pore volume
Table 10 along with the depositional facies corresponding to injected plots for the 15 realizations derived from the three
the lithology type. Table 11 summarizes the porosity and workflows. Table 14 gives the 3D streamline results in terms
permeability data for the seven lithology types. Multi-binary of recovery factor (RF). Note that there is little difference
SIS was used to distribute the seven lithology types between the “RnR-based”, “lithology-based”, and “complex
constrained by stratigraphic layer. Porosity was distributed by lithology-based” workflows. The “simple” workflow does
SGS constrained by lithology type and stratigraphic layer. give slightly optimistic results compared to the other
Permeability was distributed by a collocated cokriging with workflow results on a pore volume injected basis. This is
SGS algorithm using the porosity as secondary data. summarized schematically in Fig. 23. While the LL-652 CFB
The fourth workflow is referred to as the “complex study results show a difference when facies are used, the
stratigraphy” workflow. This workflow is analogous to the addition of facies constraints beyond two (e.g. reservoir and
“lithology-based” workflow described above except that rather non-reservoir) appears to add little value. Fig. 23 clearly
than using just the four major stratigraphic intervals as was the shows that the fluid flow characteristics at the field level of the
case for the “lithology-based” workflow, the “complex most complex model and the relatively simple “RnR-based”
stratigraphy” workflow uses the 16 detailed stratigraphic model are nearly the same. As was the case for the EMSU
intervals. Porosity and permeability average values for the study, no effort was made to assess the effect of workflow
detailed stratigraphic intervals are given in Table 12. complexity on individual well performance.
A comparison of the porosity models generated by the four
workflows tested using the LL-652 CFB data are shown in Deep Water Clastic (Turbidite) Study
Fig. 20. Table 13 provides a comparison of the porosity and A short study using data from several of the Elk Hills,
permeability averages of the models generated by the four California reservoirs was undertaken to determine if detailed
workflows. As expected, all of workflows return models with sequence stratigraphic interpretation that was partially
essentially the same average porosity due to conservation of completed needed to be finished prior to full field static and
the porosity histogram during modeling and the same dynamic modeling. The Elk Hills oil field is located in Kern
permeability (due to strong linkage with porosity). All models County, California about 20 miles from Bakersfield and
have essentially the same OOIP. consists of several reservoir structures as shown in Fig. 24.
For this study, models were developed for the Northwest
Model Up-scaling Stevens (NWS), 31S Mainbody B, and the 26R reservoirs. As
Prior to dynamic evaluation, all of the realizations were up- all of the Elk Hills reservoir studies yielded similar
scaled using a proprietary flux-based software package. The conclusions only the details of the NWS reservoir study are
static models were vertically up-scaled from 400 layers to 20 discussed in detail below.
layers and from a 35 m areal cell size to a 70 m areal cell size.
The up-scaled models had a total of 76,860 cells (63 x 61 x 20 Geological Setting
layers). The Mohnian-age (Upper Miocene) Stevens sands make up
the producing section at the NWS reservoir. The Stevens
Three Dimensional (3D) Streamline Model Description sands are part of the Antelope Shale Member (or Elk Hills
3D streamline simulation models were used to evaluate the Shale Member) of the Monterey Formation. First recognized
fluid flow characteristics of the LL-652 CFB reservoir models as turbidites by Sullwold25, the Stevens Oil Zone encompasses
generated by the three workflows. The 3D streamline model up to 4000 feet of turbidite sediments deposited in a deep
represents the mathematical equivalent of a laboratory core water environment.26, 27 The NWS reservoir is a sequence of
flood. The 3D streamline model consisted of 25; five-spot turbidite sands (Fig. 25). These sands are considered gravity
waterflood patterns with production wells and injection wells induced deep-water flows from feeder channels to the
spaced about 410 m apart. All field parameters selected for southwest and southeast. The A1 and A2 sands probably
SPE 99833 5

came from a channel in the 24Z area immediately to the and cumulative production through 5 years. Although the
northwest of the 29R Structure. The A3 to A6 sands are results were expected to be similar given the essentially
thought to have been deposited from a channel near the conformable nature of the stratigraphic layering, the identical
southeast plunge of the 29R structure, the same source channel nature of results was surprising. Clearly, the level of sequence
for the 26R reservoir sands. stratigraphic detail used to constrain the geostatistical models
The NWS reservoir is a stratigraphic trap where turbidite of effective porosity and permeability used as input to
sands pinchout updip to the southeast (Fig. 26). The sands are dynamic modeling has no significant effect on fluid flow for
typically arcosic and texturally immature. This suggests the the NWS reservoir. Similar results and conclusions were
sediments were deposited within a fairly short distance of the obtained for the Elk Hills 26R and 31S Mainbody B
sediment source provenance, probably the Salinian Block reservoirs.28
along the west side of the San Andreas Fault. A review of
whole core data shows that individual turbidite sands are thin, Effect of Semivariogram Range on Reservoir Fluid
typically 4-5 feet thick. In some places there is no shale Flow Response
between the sands and in some places shales can range up to As part of the study of the LL-652 CFB reservoir a quick
30-40 feet thick. Sequence stratigraphic markers were picked study of the effect of semivariogram parameters on fluid flow
in the more laterally continuous shales. response was completed. Models were generated using the
previously discussed “RnR-based” workflow with the
Reservoir Static Model Workflow Description semivariogram range parameters used by the SIS algorithm to
Five realizations were generated using models constrained by distribute the reservoir and non-reservoir “facies” were
varying amounts of stratigraphic control. The “two marker” arbitrarily changed as shown in Table 18. The resulting
models used only the top of the A1 and A3 sands as models were evaluated by 3D streamline simulation and the
constraints. The “major marker” models used the tops of the results summarized in Table 19. Note that over the
A1, A2, and A3 sands while the “all markers” models used all investigated range of the semivariogram parameters there was
eight of the available sequence stratigraphic markers. Table no significant impact on recovery factor.
15 provides a summary of the well log porosity by A similar assessment of the semivariogram parameter on
stratigraphic interval. The other Elk Hills study data sets reservoir fluid flow characteristics was done as part of a larger
showed more contrast in the average porosity values of the study of the Third Upper and Third Lower Sand in the Greater
stratigraphic layers than was the case at NWS. Table 16 and Burgan field in Kuwait showed that significant changes in the
Fig. 27 summarize the geostatistical constraints. Following range parameter yielded little, if any, change in the reservoir
construction of the appropriate stratigraphic framework, flow response.29,30
porosity was distributed using layer appropriate constraints via
SGS. Permeability was added by cloud transform using all Effect of Upscaling on Fluid Flow Response
available NWS core data for calibration. Fig. 28 shows cross During the course of the EMSU reservoir study a series of
sections through each of the model “types”. Note that the use dynamic models were generated using increasing amounts of
of all available sequence stratigraphic control gives models vertical and areal upscaling.
with more visually apparent lateral continuity. The models
used in this study were developed for a small portion of the Vertical Upscaling Effect on Fluid Flow Response
western part of the NWS reservoir shown previously in Fig. For the EMSU reservoir, the initial 50 foot areal cell size with
26. The static model grid was 50 x 50 x 96 layers (240,000 400 one foot (nominal) vertical layers was upscaled to a series
total cells). The areal cell size was 40 feet and the average of models (all with 100 foot areal cell size) with 167, 83, 48,
vertical layer was 4.5 feet thick. The grid definition used 18, and 9 vertical layers as shown in Fig. 32. The upscaling
enabled 2D cross sections to be extracted for finite difference was done using proprietary flux-based software. Fig. 33 plots
simulation without upscaling. cumulative water injected vs. time for one static model
evaluated at all of the selected levels of vertical upscaling.
2D Cross Section Finite Difference Model Description Note that Fig. 33 shows small differences between the results
To examine the effects of correlation detail on flow for the selected levels of vertical upscaling. The differences
simulation, 2D models were generated by cutting a north- shown in Fig. 33, however, are clearly not a direct function of
south and an east-west cross section from the reservoir the amount vertical upscaling as the range of fluid flow
property model sets. Fig. 29 summarizes the simple down-dip response is bracketed at the low by the model with the most
water injector, up-dip producer flow simulation models that vertical upscaling (9 layers) and the high end by the model
were run using a proprietary finite difference simulator in with the next most vertical upscaling (18 layers). Fig. 34
black oil mode. The down-dip injector was constrained to a shows the variation in fluid flow response for each of the 15
constant bottomhole pressure of 5180 psia. The up-dip realizations upscaled to 83, 18, and 9 vertical layers. Table 20
producer operated at a 250 bbls/day constant liquid rate. shows the results in terms of recovery factors (RF). Note that
Reservoir fluid properties are summarized in Fig. 30. there is essentially no effect of vertical upscaling on the
results.
NWS Reservoir Study Results and Discussion
Fig. 31 shows the change in Sw as a function of time for one Areal Upscaling Effect on Fluid Flow Response
of the simulation runs for each model case. Table 17 A series of EMSU reservoir models were also generated using
summarizes the results in terms of water breakthrough times differing amounts of areal upscaling. In these models the
6 SPE 99833

number of layers was fixed at 18 and the areal cell size significant affect the fluid flow characteristics of the reservoir.
increased from 50 feet (initial model, no areal upscaling) to The areal upscaling results, however, do challenge the
100, 200, and 300 feet as illustrated in Fig. 36. The upscaling prevailing paradigm of using flow models with large areal cell
was done using proprietary flux-based software. Fig. 37 sizes (e.g. 300 feet or more) or only 3-5 cells between wells
shows the variation in fluid flow response for each of the 15 for full field reservoir simulation studies.
realizations upscaled to 100, 200, and 300 foot areal cell sizes.
Tables 21-22 and Fig. 38 show the results in terms of RF. Acknowledgments
Note that there is a significant and systematic variation in RF The interpretations and conclusions presented in this paper are
as a function of areal cell size. The RF decreases significantly the author’s alone. However, several individuals did
as the areal cell size is increased. Additional studies are significantly assist the author during the course of the various
underway on mature reservoirs with different depositional projects summarized in this paper including Raymond A.
environments to further investigate the effect of vertical and Garber (EMSU geology and facies maps), Scott Williams
areal upscaling on large, essentially “full field”, reservoir (assistance with 3D streamline modeling), Jose Moros (LL-
models. 652 CFB geology) and Mauricio Villegas (assistance with the
Elk Hills 2D cross section simulations).
Conclusions
Based on the EMSU and LL-652 CFB studies the effect of Conversion Factors
adding increasing geological detail (constraints) on mature
field fluid flow response may be limited. The benefits of 1 foot = 0.3048 meters
obtaining additional data may not be warranted depending on 1 meter = 3.2808 feet
the scope and schedule of a particular project as well as the 1 mile = 1.6093 km
business decision that the model will be used to support. Fig.
39 summarizes this conclusion in qualitative terms. Clearly, References
there is a level of detail in reservoir models that needs to be 1. Xiaoguang, L., Xuewen, Z, Sincock, B., Handono, B.W., and
included and additional studies are needed to determine what Mayanullah, F.: “Integrated Approach for Improving
detail is truly essential.6,7,9,12 For example, the net/gross ratio Development of a Mature Field”, SPE 92895 presented at the
of the reservoirs studied is relatively high, on the order of 0.3 SPE Asia Pacific Oil & gas Conference and Exhibition, Jakarta,
Indonesia, 5-7 April 2005.
or higher. Perhaps, as has been suggested by the other studies, 2. Babadagali, T.: “Mature Field Development – A Review,” SPE
that reservoirs with relatively high net/gross may be modeled 93884 presented at the SPE Europec/EAGE Annual Conference,
more simply than reservoirs with lower net/gross. Reservoirs Madrid, Spain 13-16 June 2005.
with high net/gross are largely connected and given the data 3. Guimaraes, M.S., Schiozer, D.J., and Maschio, C.: Use of
density associated with large, mature reservoirs such as those Streamlines and Quality Map in the Optimization of Production
described in this report, relatively simple workflows with a Strategy of Mature Oil Fileds,” SPE 94747 presented at the SPE
small number of facies (two or three) seem to be sufficient to Latin America and Caribbean Conference, Rio de Janeiro,
model key elements of the fluid flow characteristics of these Brazil, 20-23 June 2005.
reservoirs, consistent with observations by Deutsch.31 4. Moulds, T.P., Trusell, P., and Haseldonckx, S.A.: “Magnus
Field: Reservoir Management in a Mature Field Combining
No effort was made in this study to assess individual well Waterflood, EOR, and New Area Developments,” SPE 96292
performance as these studies were done to assess the amount presented at the SPE Offshore Euroope Meeting, Aberdeen,
of detail needed to make field-scale decisions (e.g. pattern vs. Scotland, U.K., 6-9 September 2005.
peripheral water flood) rather than for detailed reservoir 5. Bischoff, R. and Bejaoui, R.: “Integrated Meodeling of the
management at the single well scale. One should expect that Mature Ashtart Field, Tunisia,” SPE 94007 presented at the SPE
the different workflows, indeed the different realizations from Europec/EAGE Annual Conference, Madrid, Spain 13-16 June
a given workflow, may give fluid flow model results that vary 2005.
considerably at individual wells. Geological and stratigraphic 6. Saleri, N.G.: “Reengineering Simulation: A Bottom-Line
complexity may be critical for detailed reservoir management Approach to Managing Complexity and Complexification,” SPE
36696 presented at the SPE Annual Technical Conference and
at the well or pattern level. Exhibition, Denver, Colorado, U.S.A., 6-9 October 1996.
Additional studies on mature fields representing other 7. Castellini, A., Chawathe, A., Larue, D., Landa, J.L., Jian, F.X.,
depositional settings are needed. Likewise, as facies-based and Toldi, J.L.: “What is Relavant to Flow? A Comparitive
modeling approaches become more commonly used, a critical Study Using a Shallow Marine Reservoir,” SPE 79669 presented
examination of the number of facies that are truly needed to at the SPE Reservoir Simulation Symposium, Houston, Texas,
capture a reservoir’s fluid flow characteristics should be done. U.S.A., 3-5 February 2003.
The number of needed facies may turn out to be a function of 8. Apaydin, O.G., Iwere, F.O., Luneau, B., and Ma, Y.: “Critical
the variability of the porosity to permeability relationship. As Parameters in Static and Dynamic Modeling of Tight Fluvial
shown in Fig. 39, using many facies may be justified only if Sandstones,” SPE 95910 presented at the SPE Annual Technical
Conference and Exhibition, Dallas, Texas, U.S.A., 9-12 October
there is a significant difference in the porosity vs. permeability 2005.
trend lines rather than the absolute value of porosity. 9. Kjonsvik, D., Doyle, J., Jacobsen, T., and Jones, A.: “The
The results of the upscaling study using only the EMSU Effects of Sedimentary Heterogeneities on Production from a
data set suggest that vertical upscaling, at least to the degree Shallow Marine Reservoir – What Really Matters,” SPE 28445,
routinely done for large, full field studies (i.e. minimum of 2-3 presented Europec, London, U.K., 25-27 October 1994.
flow model layers per geological “flow” unit) does not
SPE 99833 7

10. Cook, G., Chawathe, A., Larue, D., and Legarre, H.: 27. Zumberge, J.E, Russell, J.A., and Reid, S.A.: Charging of Elk
“Incorporating Sequence Stratigraphy in Reservoir Simulation: Hills Reservoirs as Determined by Oil Geochemistry,” AAPG
An Integrated Study of the Meren E-01/MR-05 Sands in the bull., v. 89, n. 10, 2005, p. 1347-1371.
Niger Delta,” SPE 51892 presented at the SPE Reservoir 28. Meddaugh, W. S.: “Geostatistical Modeling of the Elk Hills
Simulation Symposium, Houston, Texas, U.S.A., 14-17 26R, 31S, and Northwest Stevens Reservoirs, California,”
February 1999. AAPG Annual Meeting, Denver, Colorado, U.S.A., 2001.
11. Ezekwe, J.N., and Filler, S.L.: “Modeling Deepwater 29. Meddaugh, W. S., Al-Hamoud, J., and Kirby, R. H.:
Reservoirs,” SPE 95066 presented at the SPE Annual Technical “Geostatistical Modeling of the Third Upper and Third Lower
Conference and Exhibition, Dallas, Texas, U.S.A., 9-12 October Sands of the Burgan Formation, Greater Burgan Field, Kuwait,”
2005. AAPG International Meeting, Cairo, Egypt, 2002 (abstract).
12. Meddaugh, W.S. and Williams, S. D.: “Pragmatic 30. Meddaugh, W. S., Al-Humoud, J., Chakravarty, A., Kirby, R.,
Considerations for Reservoir Modeling Using Geostatistical Cook, G.R., and Wadowsky, J.: “Risked Reservoir Development
Techniques”, Houston, Texas, U.S.A., April, 2002 (abstract). Planning Using Geostatistical Modeling and 3D Streamline
13. Chawathe, A., and Taggart, I.: “Insights Into Upscaling Using Simulation, Greater Burgan Field, Kuwait,” GEO2000 Meeting,
3D Streamlines,” SPE REE 88846 (August 2004). Bahrain, 2000 (abstract).
14. Stern, D: “Practical Aspects of Scaleup of Simulation Models,” 31. Deutsch, C. V.: Geostatistical Reservoir Modeling, Oxford
SPE89032, SPE JPT (September, 2005). University Press, 2002.
15. Lindsay, R.F.:”Role of Sequence Stratigraphy in Reservoir
Characterization: An Example from the Grayburg Formation,
Permian Basin in Permian Basin Exploration and Production Table 1 – Grid Definition for Eunice Monument South Unit
Strategies,” p 19-26 in West Texas Geological Symposium, (EMSU) Reservoir Models
November 5-6, 1992. Stratigraphic Thickness (feet) Number of Layers
16. Garber, R. A., Meddaugh W. S., Villegas M., and Ring, J.: Interval
“Facies and Lithology-based Geostatistical Modeling of the GB-Z2 49 100
Eunice Monument South Unit (EMSU) Reservoir, New X2-Z2A 51 100
Mexico,” AAPG Annual Meeting, Denver, Colorado, U.S.A., Z2A-Z3 46 100
2001. Z3-Z4 43 100
17. Meddaugh, W. S., 2004: “Static and Dynamic Evaluation of
Stochastic Reservoir Model Workflows with Varying Amounts Table 2 – Average Porosity and Permeability Values and
of Geological Detail – Eunice Monument South Unit (EMSU), Semivariogram Range Parameters and Azimuths for
New Mexico,” AAPG Hedberg Conference on Carbonates, El Porosity Data by EMSU Model Stratigraphic Zone
Paso, Texas, 2004 (abstract). Interval Average Geometric Range Range Azimuth
Porosity and 1 2
18. Meddaugh, W. S., Garber, R. A., Villegas, M. and Ring, J.:
(fraction) Arithmetic (feet) (feet)
“Reservoir Characterization and Fluid Flow Modeling for a New Average
Mexico Waterflood”, AAPG/EAGE International Research Permeability
Conference on Carbonates, El Paso, Texas, 2000 (abstract). (md)
19. Meddaugh, W. S.: “Evaluation of Stochastic Earth Model GB-Z2 0.065 0.5 (10) 3250 2530 N50E
Workflows, Vertical Upscaling and Areal Upscaling Using Data X2-Z2A 0.088 1.3 (8) 3650 2775 N60E
from the Eunice Monument South Unit (New Mexico) and the Z2A-Z3 0.088 1.7 (15) 3640 2775 N60E
LL-652 Central Fault Block (Venezuela)”, p. 743-750 in Z3-Z4 0.081 1.1 (10) 3300 2740 N05E
Geostatistics Banff 2004, Leuangthong, O. and Deutsch, C.V. All semivariogram are exponential models with nugget = 0 and sill = 1.
(eds.), Springer, 2006.
20. Meddaugh, W.S., Moros, S. J., and Olivier, W.: “Stochastic Table 3 - Average Porosity and Permeability Values by
Reservoir Modeling and Workflow Evaluation, C3-C4 Interval, EMSU Depositional Facies
Central Fault Block (CFB), LL-652 Reservoir, Venezuela” Facies Average Porosity Geometric and
AAPG Annual Meeting, Dallas, Texas, U.S.A., April, 2004 (fraction) Arithmetic Average
(abstract). Permeability (md)
Shoal 0.101 5.4 (41)
21. Meddaugh, W.S. and Moros, S. J.: “Facies, Sequence
Lagoon 0.094 1.2 (18)
Stratigraphy and Flow Units in the Eocene C-4-X.01 Reservoir,
LL-652 Area (Lagunillas Field), Western Venezuela”, AAPG
Annual Meeting, Dallas, Texas, U.S.A., April, 2004 (abstract). Table 4 - Lithology Distribution by EMSU Facies
Facies SS MDST WKST mr- mp- GST RST
22. Levy, M. and Reed, A. A.: “LL-652 Field Reservoir PKST PKST
Characterization and Modeling,” unpublished internal company Shoal 0 0 0 0 20 70 10
report, 1999. Lagoon 42 5 13 40 0 0 0
23. Ring, J. N. and Garber, R. A.: “CFB Waterflood Pilot Reservoir Lithology Key: SS = Sandstone, MDST = Mudstone; WKST =
Simulation Study,” unpublished internal company report, 1999. Wackestone, mr-PKST = mud-rich Packstone, mp-PKST = mud-poor
24. Todd, W. and Meddaugh, W.S.: “LL652 Central Fault Block – Packstone, GST = Grainstone, RST = Rudstone
Full Field Reservoir Simulation Study – 2002”, unpublished
internal company report, 2002. Table 5 – Porosity and Permeability Average Values for
25. Sullwood, H. H.: “Turbidites in Oil Exploration” in Geometry EMSU Models Generated Using the Three Workflows
of Sandstone Bodies (Peterson, J. And Osmond, J. C., eds.), Workflow Porosity* Permeability*
AAPG Special Publication, 1961 (fraction) (md)
26. McJannet, G. S.: “General Overview of the Elk Hills Field,” Simple 0.078 (0.039) 21 (177)
SPE 35670 presented at the 66th Annual Western Meetuing, Facies 0.080 (0.038) 23 (136)
Anchorage, Alaska, U.S.A., 22-24 May 1996. Lithology 0.079 (0.035) 25 (188)
* Standard Deviation in Parenthesis
8 SPE 99833

Table 11 – Porosity and Permeability Averages for the

Table 6 – Summary of Recovery Factors (RF) Obtained for Lithology Types Used in the LL-652 CFB Models
EMSU Models Generated Using the Three Workflows Lithology Porosity (fraction) Permeability (md)
Workflow Mean RF Std. RF Min. RF Max. Runs Cross Bedded 0.139 66.9
RF Dev. Sandstone
Simple 0.311 0.006 0.303 0.324 15 Tidal Delta 0.101 11.5
Facies 0.294 0.010 0.274 0.308 15 Sandstone
Lithology 0.315 0.008 0.302 0.338 15 Sheet Sandstone 0.092 12.8
RF calculated at end of 8 year runs. Std. Dev. = Standrad Deviation, Heterolithic Fill 0.113 24.9
Min. = Minimum, Max.= Maximum. After Meddaugh18. Channel Shale 0.103 22.0
Tidal Delta Shale 0.070 3.2
Table 7 – Grid Definition for LL-652 CFB Reservoir Models Transgressive Shale 0.057 3.6
Stratigraphic Thickness (feet) Number of Layers
Interval Table 12 – Porosity and Permeability Average Values
FS C4-6 95 100 for the Detailed Sequence Stratigraphic Zones in the
C4X 155 160
LL-652 CFB Reservoir
SB C4 45 90
Marker (Top of Porosity (fraction) Permeability (md)
SB C5-11 50 50 Interval)
FS_C4-6 0.093 7.8
Table 8 – Average Porosity and Permeability Values and SB 0.083 3.5
Semivariogram Range Parameters and Azimuths for FS_C4-5 0.104 19
Porosity Data by LL-652 CFB Stratigraphic Zone SB 0.085 4.2
Interval Average Arithmetic Range Range Azimuth FS_C4-4 0.112 38
Porosity Average 1 2 SB 0.084 20
(fraction) Permeability (feet) (feet) FS_C4-3 0.117 43
(md) SB 0.084 10
FS C4-6 0.091 8.5 1500 650 N60E FS_C4-2 0.125 68
C4X 0.109 38.9 900 650 N55E SB 0.092 13
SB C4 0.066 23.5 1350 700 N50E FS_C4-1 0.130 80
SB C5- 0.077 14.9 600 550 N45E SB_C4 0.056 15
11 FS_C5-11 0.091 68
All semivariogram are exponential models with nugget = 0 and sill = 1. SB 0.043 1.0
FS_C5-10 0.094 11
Table 9 – Reservoir Properties by Reservoir, Non-Reservoi SB = Sequence Boundary at base of stratigraphic interval. Based on
“Facies” in the LL-652 CFB Input Data work of Levy and Reed21
Interval Reservoir Non- Reservoir Non- RnR
Facies Reservoir Facies Reservoir Ratio Table 13 – Porosity and Permeability Average Values for
Porosity Facies Perm Facies LL-652 CFB Models Generated Using the Four Workflows
(fraction) Porosity (md) Perm
Workflow Porosity (fraction) Permeability (md)
(fraction) (md)
Simple 0.091 29.1
FS C4- 0.141 0.073 27 0.7 24/56
RnR 0.092 25.7
6
C4X 0.143 0.069 62 1.9 58/42 Lithology 0.091 25.9
Complex Lithology 0.088 24.0
SB C4 0.147 0.048 124 0.8 20/80
SB C5- 0.134 0.052 48 0.5 32/68
11 Table 14 – Summary of Recovery Factors (RF) Obtained for
*RnR Ratio = Reservoir Facies/Non-Reservoir Ratio LL-652 CFB Models Generated Using the Four Workflows
Workflow Mean RF Std. RF Min. RF Max. Runs
Table 10 – Lithology Types and Corresponding RF Dev.
Depositional Facies Included in the LL-652 CFB Models* Simple 0.453 0.007 0.442 0.473 15
Lithology Depositional (Facies) Setting RnR 0.456 0.002 0.452 0.461 15
Cross Bedded Sandstone Estuarine Channel Fill Lithology 0.440 0.009 0.418 0.450 15
Tidal Delta Sandstone Bayhead Delta Complex 0.454 0.007 0.436 0.464 15
Sheet Sandstone Delta Front Subtidal Lithology
Heterolithic Fill Estuarine Margin Fill RF calculated at end of 8 year runs. Std. Dev. = Standrad Deviation,
Channel Shale Bay Fill Min. = Minimum, Max.= Maximum. After Meddaugh18.
Tidal Delta Shale Bayhead delta Shale
Transgressive Shale Prodelta
Based on work of Levy and Reed21
SPE 99833 9

Table 19 - Recovery Factors of Cases Used to

Table 15 – Stratigraphic Definition and Average Porosity Investigate the Effect of Semivariogram Range on
Values for the Northwest Stevens (NWS) Models Fluid Flow Results for the LL-652 CFB Reservoir
Correlation Detail Stratigraphic Average Porosity Semivariogram Mean Runs
Case Interval (fraction) Case
Two Marker A1a-A3a 0.14 Base Case (Data 46.2 5
Major Marker A1a-A2a 0.14 Driven)
A2a-A3a 0.13 Range Down 25% 46.3 5
All Markers A1a-A1b 0.096 Range Down 100% 46.3 5
A1b-A1c 0.16 Range Up 25% 46.7 5
A1c-A1d 0.16
A1d-A1e 0.15
A1e-A2a 0.14 Table 20 – Effect of Vertical Upscaling on Recovery
A2a-A2b 0.13 Factor, EMSU Reservoir (“Simple” Workflow)
A2b-A2c 0.16 Number of Upscale Average Runs
A2c-A3a 0.099 Layers in Factor Recovery
Upscaled Factor
Model
Table 16 –Range Parameter for the Isotropic 400 25% 0.320 1
Semivariograms (Spherical Form) Used to Distribute 167 10% 0.321 1
Porosity in the NWS Reservoir Models 83 5% 0.320 15
Correlation Detail Stratigraphic XY Variogram 18 1% 0.311 15
Case Interval Range 9 0.1% 0.328 15
(Feet) Base model has 400 layers and 50’ areal cell size. All upscaled
Two Marker A1a-A3a 3400 models have a 100’ areal cell size. Upscale factor is ratio of number of
Major Marker A1a-A2a 2350 cells in upscaled model divided by number of cells in raw model. .
A2a-A3a 3360 Standard deviation for recovery factors is about 0.005. After
All Markers A1a-A1b 1950 Meddaugh18.
A1b-A1c 2600
A1c-A1d 3500 Table 21 – Effect of Areal Upscaling on Recovery
A1d-A1e 3500 Factor, EMSU Reservoir (“Simple” Workflow)
A1e-A2a 3420 Areal Cell Upscale Average Runs
A2a-A2b 3770 Size in Factor Recovery
A2b-A2c 5050 Upscaled Factor
A2c-A3a 3800 Model
(feet)
Table 17 - Summary of Flow Simulation Results for 50 4% 0.339 15
100 1% 0.311 15
the Three Levels of Correlation Detail, A1 and A2
200 0.25% 0.307 15
Sands, NWS Reservoir, Elk Hills, California. 300 0.06% 0.302 15
Correlation Water Range of Cumulative Range of Base model has 400 layers and 50’ areal cell size. All upscaled
Detail Case Break- Water Production Cumulative models have 18 layers. Upscale factor is ratio of number of cells in
Through Break- through 5 Production upscaled model divided by number of cells in raw model. Standard
(Days) Through Years through 5 deviation for recovery factors is about 0.005. After Meddaugh18.
(Days) (Mbbl) Years
(Mbbl)
Two 1091±66 1003-1230 388 ± 10.4 373-407 Table 22 – Effect of Areal Upscaling on Recovery
Marker Factor, EMSU Reservoir (“Lithology-based”
Major 1106±64 981-1212 387 ± 7.6 373-399 Workflow)
Marker Areal Cell Upscale Average Runs
All Marker 1063±98 842-1250 381 ± 11.6 365-404 Size in Factor Recovery
Upscaled Factor
Table 18 - Description of Cases Used to Investigate Model
(feet)
the Effect of Semivariogram Range on Fluid Flow
50 4% 0.349 15
Results for the LL-652 CFB Reservoir 100 1% 0.315 15
Semivariogram Range 1 Range 2 Azimuth 200 0.25% 0.306 15
Case 300 0.06% 0.296 15
Base Case (Data 2300 1400 N45E Base model has 400 layers and 50’ areal cell size. All upscaled
Driven) models have 18 layers. Upscale factor is ratio of number of cells in
Range Down 25% 1750 1050 N45E upscaled model divided by number of cells in raw model. . Standard
Range Down 1150 700 N45E deviation for recovery factors is about 0.005. After Meddaugh18.
100%
Range Up 25% 2900 1800 N45E
10 SPE 99833

Figure 1. Map showing the location of the Eunice Monument

South Unit (EMSU) reservoir in New Mexico (after Lindsay15).

Figure 3. Generlized stratigraphic column for the Permian Basin

in New Mexico and Texas, U.S.A.

Figure 2. Structure map of EMSU reservoir on Top of Grayburg

Fm. The Arrowhead Grayburg Unit (AGU) is to the south of Figure 4. Schematic depositional facies map for the Grayburg
EMSU. Fm. at the EMSU reservoir. The EMSU field limits are shown by
the outline. North is to top of figure. The EMSU field is about 5
miles wide.
SPE 99833 11

Figure 5. Schematic cross section (NW to SE) through the EMSU

reservoir (from Lindsay15).

Figure 9. Areal and west to east vertical cross section through

the EMSU “simple” workflow earth model (left) and upscaled
model (right). Note that porosity (red) is generally high to the
west (shoal depositional environment) in these sections.

Figure 10. Histograms

of porosity (left) and
permeability (right) in
Figure 6. NW to SE stratigraphic cross section through the the EMSU shoal
Grayburg Fm. at the EMSU reservoir. Datum = top of Grayburg (bottom) and lagoon
Fm. Length of cross section is about 1.8 miles. GR curve shown (top) depositional
of left; lithology curve on right (blue/purple = mud-rich rocks). facies. Note that the
porosity distribution
in both facies is
similar. However, the
permability in the
shoal is significantly
higher than in the
lagoon facies.

Figure 7. Map showing EMSU wells and location of project study

area. Bold symbols correspond to wells with core data. West to
east line in project area detail shows location of model cross
sections shown below (after Meddaugh18).

Figure 8. Example
directional semivariogram
for porosity in the EMSU
reservoir. Example is from
the GB-Z2 stratigraphic
interval. Figure 11. Areal and west to east vertical cross section through
the EMSU “facies” workflow earth model (left) and upscaled
model (right). Sections at top show the facies distribution (shoal
to the west; lagoon to the east). High porosity shown in red.
12 SPE 99833

Figure 12. Areal and vertical cross sections from the “lIthology-
based” EMSU models. Top shows the initial facies distribution.
Sections second from top show the lithology distribution. The
mud-rich rocks (poor reservoir quality) occur mainly to the east Figure 14. Plots showing cumulative oil production vs. time for
(right) in the lagoon facies. Third from top sections show the the three EMSU workflows: simple (top), facies-based (middle),
porosity distribution prior to upscaling (bottom). High porosity and lithology-based (bottom). Fifteen model realizations are
shown in red. shown for each workflow case. Note that there is little difference
in the cumulative oil production curves for models generated
using the different workflows.

Figure 15. Plot showing distribution of recovery factors from

each of the 15 realizations generated from the three workflows.

Figure 13. Time slices showing change in water saturaturation for

an EMSU 3D streamline model. The wells shown on the cross
section are all injectors.
SPE 99833 13

Figure 16. Location of the LL-652 reservoir in Lake Maraciabo,

Venezuela. Structure map is shown at right. Rectangle in
structure map shows the location of the study area in the LL-652
reservoir central fault block (CFB).

Figure 18. Summary chart showing sequence stratigraphic

interpretation for the LL-652 CFB freservoir interval studied.

Figure 17. Stratigraphic column for the LL-652 CFB reservoir

showing the stratigraphic unit studied (C-3-X / C-4-X). Log at
rights highlights the limited distribution (red) of high quality
reservoir in the studied interval.

Figure 19. Example directional semivariogram for porosity in the

LL-652 CFB reservoir. Example is from the FS_C4-6 to C4X
stratigraphic interval (top of study interval).
14 SPE 99833

Figure 22. Cumulative oil vs. pore volume injected plots for the
four LL-652 CFB workflows from 3D streamline simulations for 15
realizations of each workflow. Note that there is little difference
between the RnR-based, lithology-based, and complex
(stratigraphy) lithology-based results

Figure 20. Comparison of porosity distribution generated using

the four LL-652 CFB workflows investigated. Blue = low porosity;
Red = high porosity. High porosity shown in red.

Figure 23. Schematic figure derived from results show above.

Note that the simple LL-652 CFB workflow models give slightly
pessimistic results compared to the models derived from the
other three workflows investigated which essentially overlap.

Figure 21. Time slices showing change in water saturaturation for Figure 24. Location map showing the various Elk Hills, California
an LL-652 CFB 3D streamline model. The wells shown on the reservoirs included in study.
cross section are all injectors.
SPE 99833 15

Figure 25. Schematic west to east cross section for the Elk Hill
Northwest Stevens (NWS) reservoir showing updip pinchout of
sand layers.

Figure 28. South to north cross sections through porosity

models generated with varying levels of stratigraphic control.

Figure 26. Structure map (A3 Sand top) for NWS reservoir. Small
square in western portion of map shows location of the 3D earth
models constructed for fluid flow evaluation.

Figure 29. Schematic description of 2D finite difference model

used to evaluate the NWS reservoir models.

Figure 30. Summary of fluid properties used in the NWS reservoir

models (oil gravity = 29° API).

Figure 27. Summary of isotropic semivariogram modeled for

porosity for each of the various stratigraphic levels of detail
studied. All semivariogram forms are spherical. Note that the
semivariograms are “well defined” by the NWS reservoir well log
data.
16 SPE 99833

Figure 34. 3D
streamline
results for 15
realizations
generated
using the EMSU
data set and the
simple
workflow. Plots
at left
correspond to
upscaled
models with 83,
18, and 9 layers
(initial model
has 400 layers).

Figure 31. Cross sections through single realizations of each of

the stratigraphic-control cases investigated showing change in
water saturation through time. The injector is on the left and the
producer on the right. Note that there is little difference between
the three cases investigated.

Figure 35. Plot of recovery factors (after 8 years) calculated for

each of realizations upscaled from 400 vertical layers to 83, 18,
and 9 vertcal layers.

Figure 36. Areal

and vertical cross
sections from
models upscaled to
18 vertical layers
and 50 foot, 100
Figure 32. Cross sections through EMSU model generated via the foot, 200 foot, and
simple workflow and upscaled to various number of vertical 300 foot areal cell
layers as shown. The reservoir property shown is porosity. size. The intial
model generated
Figure 33. Water using the simple
injection vs. time workflow and the
plot for one EMSU adapt set
EMSU model used 400 layers
realization built and a 50 foot areal
using the simple cell size. The
workflow and reservoir property
upscaled to shown is porosity.
varing amounts.
SPE 99833 17

Qualitative Comparison of Workflows

Relative Cost or “Value”

High

Low

Geological
Modeling

Precision

Characteristics
Costs
Data Costs

Fluid Flow
High

Complexity
Model
Low

Figure 39. Summary “cartoon” showing the relationship inferred

from the studies reported in this paper between model complexity
(e.g. number of geological constraints), geological precision, and
fluid flow characteristics (after Meddaugh18).

Figure 37. 3D streamline results for 15 realizations generated

using the EMSU data set and the simple workflow. Plots above
correspond to upscaled models with 50 foot, 100 foot, 200 foot,
and 300 foot areal cell sizes and 18 layers. The initial model used
a 50 foot areal cell size and 400 layers.

Figure 38. Plot showing variation in recovery factor as a function

of areal cell size for EMSU models derived from the simple
workflow (Series 1) and the lithology-based workflow (Series 2).
Points plotted represent average obtained from 15 realizations
from the two workflows.