Вы находитесь на странице: 1из 11

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/249552445

3D high-resolution digital models of outcrop


analogue study sites to constrain reservoir
model uncertainty: An example...
Article in Petroleum Geoscience October 2004
DOI: 10.1144/1354-079303-617

CITATIONS

READS

58

129

5 authors, including:
Jamie Keith Pringle

Anthony Robin Westerman

Keele University

none

100 PUBLICATIONS 676 CITATIONS

78 PUBLICATIONS 315 CITATIONS

SEE PROFILE

SEE PROFILE

Andy Gardiner
Heriot-Watt University
37 PUBLICATIONS 265 CITATIONS
SEE PROFILE

Some of the authors of this publication are also working on these related projects:
SHEER - Shale gas Exploration and Exploitation induced Risks - An EU Horizon 2020 Project View
project
Legacy seismic, mainly offshore west Cumbria. View project

All content following this page was uploaded by Jamie Keith Pringle on 12 October 2015.
The user has requested enhancement of the downloaded file. All in-text references underlined in blue
are linked to publications on ResearchGate, letting you access and read them immediately.

3D high-resolution digital models of outcrop analogue study sites to


constrain reservoir model uncertainty: an example from Alport Castles,
Derbyshire, UK
J. K. Pringle1, 2, A. R. Westerman1, J. D. Clark1, 3, N. J. Drinkwater4 and A. R. Gardiner1
1

Petroleum Engineering Institute, Heriot-Watt University, Edinburgh EH14 4AS, UK


(e-mail: jamie.pringle@pet.hw.ac. uk)
2
School of Earth Sciences & Geography, Keele University, Stas ST5 5BG, UK
3
Present address: ChevronTexaco, 6001 Bollinger Canyon Road, Building D-1180, San Ramon, CA94583, USA
4
ChevronTexaco Energy Technology Company, 4800 Fournace Place, Bellaire, TX 77401, USA
ABSTRACT: Advances in data capture and computer technology have made possible

the collection of 3D high-resolution surface and subsurface digital geological data


from outcrop analogues. This paper describes research to obtain the 3D distribution
and internal sedimentary architecture of turbidite channels and associated sediments
at a study site in the Peak District National Park, Derbyshire, UK. The 1D, 2D and
3D digital datasets included Total Station survey, terrestrial photogrammetry and
remote sensing, sedimentary logs and a Ground Penetrating Radar (GPR) dataset. A
grid of 2D GPR profiles was acquired behind a cliff outcrop; electromagnetic
reflection events correlated with cliff face sedimentary horizons logged by Vertical
Radar Profiling. All data were combined into a Digital Solid Model (DSM) dataset of
the site within reservoir modelling software.
The DSM was analysed to extract 3D architectural geometries for petroleum
reservoir models. A deterministic base model was created using all information,
along with a suite of heterogeneous turbidite reservoir models, using 1D, 2D or 3D
information. The model suite shows significant variation from the deterministic
model. Models built from 2D information underestimated connectivity and the
continuity of geobodies, but overestimated channel sinuosity. Advantages of using
3D digital outcrop analogue data for 3D reservoir models are detailed.
KEYWORDS: outcrop analogues, 3D digital model

INTRODUCTION
Deep-water clastic rocks are increasingly important sources of
petroleum production in most Atlantic basins, especially the
West African margin, North Sea, West of Shetland, Norwegian
shelf regions, eastern margin of South America and in the
Gulf of Mexico (Pettingill 1998; Cossey 2000; Lawrence &
Bosman-Smits 2000). With high quality 3D seismic and well
data, a detailed 3D stratigraphic framework of the reservoir can
be created (see Sikkema & Wojcikl 2000). However, important
sedimentary architectures such as intra-channel fill elements
may not be resolved. Fill properties may prove critical during
hydrocarbon production, especially in mature fields. A common practice is to use outcrop analogue geostatistical data to
infill the stratigraphic framework at sub-seismic resolution (e.g.
Chapin et al. 1994; Clark & Pickering 1996; Satur et al. 2000;
Drinkwater & Pickering 2001; Johnson et al. 2001; Lien et al.
2003). Well-exposed, 2D outcrop cliff faces are usually used to
obtain geostatistics (see Arnot et al. 1997), but limited outcrop
areas require significant data manipulation (e.g. Geehan &
Underwood 1993; Visser & Chessa 2000). A few, rare
exposures have multiple gullies dissecting the study area and
these allow 3D models of turbidite fans to be constructed (see
Petroleum Geoscience, Vol. 10 2004, pp. 343352

Batzle & Gardner 2000; Satur et al. 2000; Hodgetts et al. 2004).
However, such models are still limited by potential errors of
interpolation between outcrop faces. Furthermore, geostatistical measures may be skewed by outcrop shape and
orientation, rather than by original sedimentary features.
Parameters such as channel sinuosity, connectivity and
continuity of target sandbodies in 3D remain undefined,
although they control hydrocarbon sweep efficiency.
Some authors (e.g. Pickering & Corregidor 2000) have used
borehole information, drilled behind outcrop cliff faces, to
obtain pseudo-3D data, but this does not provide detailed, 3D
sedimentary architectural detail. Modern deep-sea fan environments can be studied in plan-view using side-scan SONAR
images, which give ideas of channel width, sinuosity and some
depositional processes (e.g. Kenyon et al. 1995). However, these
systems can still remain poorly understood. Rare 1D Deep Sea
Drilling Project (DSDP) drilling logs can provide subsurface
data (e.g. Pickering et al. 1986; Hiscott et al. 1997), but these are
limited in number and extent. Laboratory-based, flume tank
experiments have allowed process-based models to be investigated, chiefly by Kneller (1995), Kneller & McCaffrey (1999)
and Peakall et al. 2000, but these are dominantly empirical
1354-0793/04/$15.00  2004 EAGE/Geological Society of London

344

J. K. Pringle et al.

Fig. 1. (a) Stratigraphy of the


Derbyshire (Namurian) succession
with the study site location marked
(after Clark & Pickering 1996).
(b) Depositional environment
reconstruction for the Shale Grit
Formation (after Hampson 1997).

observations and subject to scale-model assumptions. Computer modelling studies (e.g. Boylan et al. 2002) provide
quantitative information in three dimensions, but depend upon
high-resolution, accurate input data that are rarely collected and
input in three dimensions.
Ever-improving computer hardware and software (e.g.
Dueholm & Olsen 1993) are being used increasingly with
survey (e.g. Xu et al. 2000), differential GPS (e.g. Bryant et al.
2000) or the new generation LIDAR (Light Detection And
Ranging) equipment (e.g. Ahlgrend & Holmund 2002; Bellian
et al. 2002) to obtain highly accurate, digital datasets of outcrop
study sites. These datasets can be used to create digital surface
models which can be interactively viewed from any desired
orientation to gain an appreciation of 3D distributions of
sedimentary architectures, for example (see Bryant et al. 2000;
Pringle et al. 2004). Whilst useful, these models usually cannot
be interrogated for data extraction and still do not comprise 3D
volumetric information.
High-resolution, shallow seismic geophysical data have
been obtained behind outcrop cliff faces, but usually with little
success (see Coleman et al. 2000). This is chiefly due to the
cemented nature of the sediments, providing little or no
acoustic impedance contrast between sedimentary intervals.
The relatively thin target zones are often still below nearsurface seismic resolution (Pringle 2003).
Ground Penetrating Radar (GPR) equipment has begun to
be used to acquire shallow subsurface, high-resolution datasets
(c. 0.25 m spaced beds to depths of 10 m+) on ancient outcrop
analogues (see Stephens 1994; Corbeanu et al. 2001; Szerbiak
et al. 2001; Pringle et al. 2003a). Whilst electromagnetic reflection amplitudes are controlled by changes in bulk electrical
properties, rather than by lithology boundaries (Davis & Annan
1989), Vertical Radar Profiles (VRPs) acquired down outcrop
cliff faces allow GPR reflection events to be correlated with
observed cliff-face sedimentary horizons (see Pringle et al.
2003b).
This paper describes datasets obtained from Carboniferous
(Namurian) turbidites of the Shale Grit Formation, exposed at
Alport Castles, Peak District National Park, (UK grid reference:
SK 143 914). The aim of the study was three-fold: to obtain
1D, 2D and 3D datasets (both surface and subsurface); to
integrate field data into a single digital model of the study site;
and to extract sedimentary architectures both to build a
deterministic reservoir model and test modelling sensitivities to
input data. GPR data provided the 3D near-surface, highresolution volumetric information.

ALPORT CASTLES STUDY SITE, PEAK DISTRICT


NATIONAL PARK, DERBYSHIRE, UK
Sedimentary background
The intra-cratonic Central Pennine Basin developed during
the Carboniferous, receiving its main clastic fill during the
Namurian (Walker 1966). In the Huddersfield sub-basin, the
stratigraphy of the Kinderscout area represents a prograding
basin fill, infilling pre-existing rift topography. Deposition
commenced with the deep-water, fine-grained Edale Shales,
which contain the Reticuloceras marine band biozone in upper
units. The successively overlying formations comprise, in
stratigraphical order: the Mam Tor Sandstones, Shale Grit,
Grindslow Shales and Kinderscout Grit (Fig. 1a). These formations form a prograding succession over 600 m thick (Walker
1966). The Namurian fill comprises a dominantly prograding
fan and delta system succession that fills the basin with small
regression episodes.
The Shale Grit and Grindslow Shales formations have been
interpreted as turbidite fan and slope sediments, respectively
(Collinson 1970), the lower part of the Shale Grit corresponding to medial submarine fans, the upper part to distributary
channels supplying sediment to these fans (McCabe 1978).
Hampson (1997) placed these formations in a sequence
stratigraphic framework, interpreting the Shale Grit as a lowstand systems tract (Fig. 1b). Within this framework, Hampson
(1997) interpreted the influx of sandstones into deep-water
environments to have been largely confined to periods of
fluvial incision on the delta plain and slope.
Study site characteristics
The Alport Castles exposure (Fig. 2a for location map), formed
by local faulting and landslips, consists of c. 400 m long, c. 50 m
high, sub-vertical cliff sections (Figs 2b, c), exposing
Carboniferous (Namurian) turbidite sheet-like sandstones and
small-scale channel fills (Walker 1966). Clark & Pickering
(1996) interpreted the sand-rich deposits as a series of stacked,
multi-storey, channel-fill sandstones associated with subsidiary
fine-grained, overbank deposits.
There have been several sedimentary facies classifications
since the original descriptions by Walker (1966). The most
recent study by Clark & Pickering (1996) subdivided the facies
into three types. Facies A consists of thin (25 cm thick)
parallel-bedded, sharp-based sandstone units and interbedded mudstone. Facies B consists of thick-bedded (0.63 m)
amalgamated sandstones with commonly erosive bases and
with numerous palaeocurrent indicators. Facies C is composed

3D digital outcrop analogue models

345

Fig. 2. (a) Location map of the study site at Alport Castles, Peak District National Park, Derbyshire, UK. (b), (c) Photomosaics of outcrop cliff
faces (see (a) for location). Sedimentary log sections are marked. (d) Close-up of intra-channel, mud-flake breccia; palaeocurrent data are shown
in (a). Team members (ringed) are for scale.

Fig. 3. Fence diagram interpretation of cliff sections A & B at Alport Castles (modified from Clark & Pickering 1996). The sinuous cliff
(orientated almost perpendicular to palaeocurrent) meant that only an approximation of horizontal scale has been applied to the correlated
section. The top bed of channel element 12 is used as a datum.

of mud-flake breccias (Fig. 2d), consisting of angular shale


clasts in a coarse sand matrix, with clasts commonly aligned
parallel to bedding. Facies C is interpreted to relate to sub-

marine channel incision and migration, which is of particular


interest as the breccias could cause potential permeability
barriers to reservoir intervals.

346

J. K. Pringle et al.

Fig. 4. A 3D digital surface model of


the study site visualized in Petrel
modelling software. DEMs from digital
aerial and terrestrial photogrammetric
output have been imported and
gridded. Ortho-Rectified Images (again
from photogrammetric output) have
been draped onto their respective
surfaces. Sedimentary logs (represented
by pseudo-wells) and GPR data
acquisition area (box) are shown.
Interpreted base channel 3 in Figure 3
is shown as a horizon surface.

Fig. 5. Snapshot of the GPR dataset


without interpretation, showing GPR
profile locations and velocity analysis
sampling points (CMP & VRP). Note
the incised channel forms in the
dataset, consistent with palaeocurrents
obtained at the cliff-face (inset).

Data acquisition and processing


1D to 2D sedimentary data: sedimentary logs. Detailed sedimentary
logs of the cliff faces at Alport Castles were acquired, surveyed
(locations marked in Fig. 2) and combined with photomosaic interpretations. This allowed the separation of the 12
interpreted sheet and channel element end-members (Fig. 3).
The end-members are (a) sheet-like geometry elements (1, 2, 4,
5, 7) of massively bedded sandstones and (b) locally incised
channel elements (3, 6, 812). The interpreted channel elements
show a stepped development, eroding into underlying units.
Mud-flake breccias were recognized at the base of all the
channels. The breccias are interpreted as the erosive detritus of

either severe channel-bank incision up-current or channel


migration. Asymmetric, inclined mud-flake breccia horizons
were recognized in the channel fill (Fig. 2d) and these may
represent lateral migration.
Sedimentary logs were converted to pseudo-well logs,
with facies, interpreted elements and appropriate depth values
incorporated. The pseudo-wells were then imported into
Schlumbergers Petrel PC-based reservoir modelling software.
3D surface data: digital photogrammetry. Digital photogrammetry is
a standard technique in surveying, which produces both a 3D
surface (termed a Digital Elevation Model or DEM) and a
map-view corrected surface image (termed an Ortho-Rectified

3D digital outcrop analogue models

347

then applied to create a highly accurate (c. 0.15 m average error)


digital image-draped cliff surface model. This model was then
registered to UTM coordinates and integrated into the digital
aerial photogrammetry model within Petrel (Fig. 4). The resulting 3D Digital Outcrop Model (DOM) not only provides an
accurate surface framework within which the rest of the field
data can be integrated, but it is also in real-world coordinates
and can be interrogated repeatedly for surface stratigraphy,
architectural geometry, fracture distribution, etc. Figure 4
shows one cliff surface interpreted horizon.

Fig. 6. Common Mid-Point (CMP) data using 50 MHz dominant


frequency antennae. Constant velocity air and ground waves are
marked, with dipping weaker reflection (R16) events. A velocity of
0.07 m ns1 was calculated. The strongest reflectors were imaged
from Traces 68 (marked), indicating 3 m optimal antenna spacing
not 2 m, as suggested by the acquisition software.

Image or ORI), which can be draped onto the DEM. This


process is explained in more detail in Sgavetti, (1992) and
Pringle et al. (2004). Digital photogrammetry has the benefit
of correcting for photographic lens distortion and producing 3D image-draped surface models of study sites.
Although aerial photographs are usually used as input, terrestrially obtained ground photographs were also used in this
study.
A stereo-pair of scanned aerial photographs covering the
study site, together with Ground Control Points (GCPs)
obtained from local topographic maps of identifiable features
observable on the overlapping area of both photographs, were
used to create a digital surface model of the study site (Fig. 4).
However, the cliff outcrop is poorly represented in aerial
photogrammetry, since the vertical lines of sight are sub-parallel
to the cliff-face. Therefore, the site was visited over a one-day
period to obtain a cliff photograph stereo-pair and theodolite
survey data were used for control (see Pringle et al. 2004 for
detail). Digital terrestrial photogrammetric techniques were

3D subsurface data: Ground Penetrating Radar. After an initial phase


of investigation to decide the optimum dominant frequency
antennae, the site was then re-visited for a five-day study
period. A grid of survey lines in a northsouth and eastwest
orientation, set 50 m apart, was set out behind the cliff face
by measuring tapes, pegs and bearing information (Fig. 5a).
The lines were profiled subsequently by a GPR pulseEKKO
100 system, using 50 MHz dominant frequency antennae.
Parameters used were as follows: 0.5 m and 1 m trace sample
spacing in in-line and cross-line direction, respectively, 800 ps
sample interval, 64 stacks and a Time Window of 600 ns. 4001
GPR traces were obtained, covering a total distance of 2.4 km.
After analysis of Common Mid-Point (CMP) data (Fig. 6), a set
transmitter/receiver antenna spacing of 3 m was used to obtain
the 2D GPR profile dataset (Fig. 5b).
GPR profiles were then processed to optimize image data
quality (following the methodology in Pringle et al. 2003b). The
resulting GPR dataset was then converted from time to depth,
using the velocity analysis techniques of CMPs and VRPs, both
explained in detail in Pringle et al. (2003b). VRPs are important
to obtain, as these link definitively the observed sedimentary
horizons on the cliff face with GPR reflection events into first,
the VRP, and are correlated subsequently into the subsurface
GPR dataset.
Once processed, the GPR profiles were imported into the
reservoir modelling software. The inherent seismic interpretation module was used to interpret mappable horizons across
the dataset, these being predominantly the rockhead and the
top and bases of channelized features (Fig. 7). Interpreting
profiles in 3D allows tie-checks between in-lines and cross-

Fig. 7. (a) 2D, 50 MHz GPR profile


in-lines X100 and X5/b have been
processed, imported and interpreted in
Schlumberger Petrel software. (b) Main
horizons interpreted are the rockhead,
bases and tops of channelized features.

348

J. K. Pringle et al.

Fig. 8. Screen view of the GPR


dataset, and main interpreted horizons
(base) and (top), respectively. Abseiled
sedimentary logs and terrestrial
photogrammetry datasets are also
visualized, with the larger-scale aerial
photogrammetry surface grid also
shown.
Table 1. Input parameters for bed facies and element type in the stochastic Schlumberger Petrel model of the study site at Alport Castles
Facies/Element

Facies
2
3

Name in well log

Element type

Body shape

Body size (m)

Modelling rules

Thickness

Width

Length

channel_sand
sheet_sand

Channel
Sheet

1/2 cylinder
Ellipsoid

12
12

3thick
1000

3width
1000

4
Element
1
2

thin_beds

Sheet

Ellipsoid

12

1000

1000

no eroding

base_sheet
diac_with_3

Sheet
Sheet

Ellipsoid
Ellipsoid

0.12
0.12

1000
1000

1000
1000

no eroding
no eroding

3
4

main_chan
amalg_cont_sheet

Channel
Sheet

1/2 cylinder
Ellipsoid

0.12
0.12

3thick
1000

3width
1000

unconf_sheet

Sheet

Ellipsoid

0.12

1000

1000

no eroding

6
7

thin_channel
diac_with_8

Channel
Sheet

1/2 cylinder
Ellipsoid

0.12
0.12

3thick
1000

3width
1000

no eroding
no eroding

8
9
10

small_chan
pinch_chan
up_thin_chan

Channel
Channel
Channel

1/2 cylinder
1/2 cylinder
1/2 cylinder

0.12
0.12
0.12

3thick
3thick
3thick

3width
3width
3width

erodes into 7
no eroding
erodes into 9

11
12

homog_chan
large_chan

Channel
Channel

1/2 cylinder
1/2 cylinder

0.12
0.12

3thick
3thick

3width
3width

no eroding
erodes into 11

13

back_shale

Sheet

Ellipsoid

0.12

1000

1000

lines, which are difficult to do accurately in two dimensions.


The channel base was interpreted to be the base of element 3,
with the top being interpreted as element 8 in Figure 3.
Palaeocurrent measurements taken at the cliff face were used
to check channel orientations. Figure 8 shows a view of the
GPR dataset, with abseiled sedimentary logs, digital aerial and
terrestrial photogrammetry dataset also shown. The interpreted
horizons were then used in the reservoir model-building
process.
From interpreting the GPR dataset, although the interpreted
cliff-face sheet-like elements are present, interpreted channel
horizons mapped in the top of the section within the GPR
dataset were relatively more incised than the channels exposed

erodes into 3
no eroding

erodes into 2
no eroding

no eroding

Sedimentary detail

% of
section

base sst, unconfined


homogeneous,
unconfined
lateral fill, thins upward

39
52

base sst, unconfined


unconfined, diachronous
with 3
lateral fill, thins upward
homogeneous,
unconfined
homogeneous,
unconfined
lateral fill, thins upward
homogeneous,
unconfined
lateral fill, thins upward
pinches out
continuous across
exposure
small-scale
large continuous
channel
thin bedded and shale
units

15
8

10

8
10
5
3
4
7
4
9
2
15
10

in the cliff face deeper in the stratigraphic section. These


observations are consistent with a regressive event within the
stratigraphic successions, as described previously.
DIGITAL SOLID MODELS (DSMS)
All field data, including digital aerial and terrestrial photogrammetric output (DEMs and ORIs), 1D sedimentary logs
converted to 1D pseudo-well logs and 3D GPR dataset and
interpreted horizons, were integrated within reservoir modelling software. A highly detailed 3D Digital Solid Model
(DSM) of the study site was created. This dataset is highly
accurate with dense three-dimensional data all within a
single digital model in one software package, which can then be

3D digital outcrop analogue models

349

Fig. 9. Reservoir models produced from (a) 1D and 2D data and (b) 3D data, with average facies plots (map view) of (c) 1D/2D and (d) 3D
models, produced by summing facies types in each grid cell pillar/the number of cells. Little change is seen in (c), with (d) identifying
channel-rich zones.

easily data mined, depending upon the information required.


Analysing a three-dimensional dataset removes any errors in
estimating geobody dimensions from 2D information and
corrects palaeocurrent direction for a 3D geobody a key
factor for reservoir modelling, as noted by Hodgetts et al.
(2004). Resulting reservoir models built using this information can be compared directly with the input DSM
dataset another advantage of keeping the dataset within
modelling software.
A bonus of creating such a dataset is that it can be
interrogated repeatedly for different research aims for
example, 3D sedimentary architectures at varying scales, smallscale fracture distribution studies, synthetic seismic modelling
comparisons or surface outcrop distributions over the study
site. Detailed mapping in 3D allows features to be identified
over larger scales, which may be missed using conventional
sedimentary mapping techniques. The DSM can also be utilized
as a virtual field trip, with modern visualization technologies,
avoiding field trip costs, reducing field time and pinpointing
areas for subsequent field investigation. A digital fly-through of
the Alport Castles DSM can be viewed in Pringle et al. (2001).

Future field data can also be integrated into the current dataset,
the updated model providing ever-increasing quantitative information.
RESERVOIR MODELLING
Recent reservoir modelling practises quantify increasingly
the reservoir uncertainty by adopting either a probabilistic
approach based on a best-guess deterministic model incorporating all uncertainty, or the scenario approach, based on a
suite of conceptually different models (Floris & Peersmann
2002). High-resolution modelling presented in this paper,
therefore, has used both approaches to produce a suite of
different scenario models and a best-guess deterministic
model.
Reservoir model dimensions were the same for all studies,
the model size being 1.75 km  0.75 km  0.06 km thick over
the area of interest (box in Fig. 5 showing model area). The
resulting grid cells generated were relatively small for a reservoir
study; only c. 175 000 cells (typical static stochastic reservoir
models being c. 1020 million cells, Christie, pers. comm.), but

350

J. K. Pringle et al.

at a comparatively high resolution, 10x, 10y and 0.5z m


respectively. The grid was rotated by 55, parallel to the main
outcrop face, so that any geobody edge distortion effects were
minimized.
Suite of stochastic models
For initial stochastic reservoir models, the geostatistics derived
from the outcrop cliff-face interpretation (Fig. 3) were used
to subdivide the image volume into three facies types:
channel-sandstone (38%), sheet-sandstone (52%) and thinbedded units composed of 110 cm thick sandstone and shale
beds (10%). Pseudo-well logs, converted from the sedimentary
logs (upscaled to 0.5 m intervals in the vertical direction), acted
as control for initial models. Intra-channel fill was made more
heterogeneous by the addition of inclined, discontinuous shale
horizons. These horizons were noted within channels on the
outcrop face (Fig. 2d); 3D shale horizon dimensions were
obtained from a similar study on the Carboniferous Ross
Formation turbidite channels described in Pringle et al. 2003a.
The 3D shale horizon data from the Ross Formation study
were scaled to fit the Alport Castles turbidite channel dimensions, being of similar Upper Carboniferous age and deposited
in a medial turbidite environment. Comparisons of resulting
models with observed sedimentary horizons at the outcrop cliff
face showed, surprisingly, little similarity; therefore, the facies
were further subdivided into the appropriate interpreted
elements 113 (Fig. 3) to create more realistic models; the input
parameters are listed in Table 1.
Multiple models, or realizations, were run using input
parameters that followed standard practice (see Floris &
Peersmann 2002), in order to create a range of models from
which the reservoir modeller could pick the best-fit model
with the reservoir data. An example of a stochastic reservoir
model is shown in Figure 9a.
Analysis of facies distributions (Figs 9a, c) showed a
generally even unit distribution, honouring the input pseudowell log data. Connectivity analysis of the first 20 major
geobodies within stochastic models (Figs 9b, 10a) showed a
mean value of 4.5%, standard deviation of 10.6%, kurtosis
value of 5.1 and skewness of 2.5. Major connected geobodies
decrease exponentially in connectivity with reducing volumes
(Fig. 10a). Sinuosity analysis of geobodies (using a directional
trend analysis option available in Petrel) showed a bimodal
distribution (Fig. 10b), with an azimuth of 174, a standard
deviation of 20.7, kurtosis value of 0.5 and a 0.5 skewness
value. Variogram analyses were not undertaken for this study,
but could be generated easily and incorporated directly into full
reservoir-scale models within the Petrel software framework for
comparison, if required.
Interpretation of sinuosity analysis showed a significant
deviation from the mean palaeocurrent direction as well as
a flatter, positively skewed distribution. Results were due
primarily to the input palaeocurrent values obtained from the
outcrop sedimentological study and the distribution of geobody
elements.
Deterministic model
A deterministic model was created incorporating all the available field and remote sensing data (including the 3D GPR
interpreted horizons), which was the main aim of this study.
After a comparison period matching the resulting model with
input data, a high-resolution best-guess model was created.
GPR interpreted horizons were correlated first with observed
outcrop cliff face sedimentary horizons then used to subdivide

Fig. 10. Histogram statistics taken from models shown in Figure 9


for (a) 1D and 3D model connected volumes and (b) directional
trend (sinuosity) of geobodies.

the model volume into three zones. These zones were modelled
stochastically using the input parameters in Table 1, following
standard practice (see Floris & Peersmann 2002). The resulting
model is shown in Figure 9c.
Analysis of facies distributions (Figs 9b, d) showed an
uneven unit distribution, with negative skewness and a small
standard deviation observed. Connectivity analysis of the first
20 major geobodies within the deterministic model (Figs 9d,
10a) showed a mean value of 3.8%, standard deviation of 8.7%,
kurtosis value of 4.3 and a skewness value of 2.3. Major
connected geobodies also decrease exponentially in connectivity with reducing volumes (Fig. 10a) but, importantly, have a
maximum at major geobody 3. Sinuosity analysis of geobodies
(again using Petrels directional trend analysis option) showed a
bimodal distribution (Fig. 10b), with an azimuth of 170, a
standard deviation of 16.5, kurtosis value of 0.2 and 0.7 value of
skewness.
Interpretation of sinuosity analyses showed a smaller
deviation from the mean palaeocurrent as well as a normal,
positively skewed distribution. Results were due to input
palaeocurrent values obtained from the outcrop sedimentological study, GPR interpreted horizon constraints and the
distribution of geobody elements.
Model comparisons
Comparing facies unit distribution analyses between stochastic
models built from 1D and 2D outcrop information with the

351

3D digital outcrop analogue models


deterministic model built from 3D information shows interesting variations (Fig. 10). The stochastic model facies distribution
averages showed an even distribution, compared to the uneven,
positively skewed facies distribution of the deterministic model.
Model geobody connectivity comparisons between the different
models show a similar exponentially decreasing trend with
geobody size, but there were significantly more and smaller
geobodies in the deterministic model that were more connected
compared to the fewer but larger geobodies modelled in the
stochastic models. Increased connectivities were seen in the
deterministic model compared with the stochastic models.
These variations are explained as connecting geobodies that
appear isolated in 2D but which are, in fact, connected in 3D;
especially as geological unit thicknesses are increased.
Analysis of the directional trend (palaeocurrent direction) of
geobodies within the models showed that whilst the azimuth
values were similar (174 and 171), the standard deviation of the
stochastic model geobody trends were significantly higher than
those geobodies from the deterministic model. Thus, models
created from outcrop faces had significantly larger palaeocurrent variations than the deterministic model geobodies, implying that overestimation of palaeocurrent variations is inherent
in reservoir models using 1D and/or 2D surface outcrop
information.
RESEARCH WORKFLOW
Having the final, integrated digital dataset within Petrel has
made analysis, data extraction and model building simple and
relatively efficient. However, a significant amount of effort was
required to obtain the various dataset types from the study site,
post-data acquisition processing (e.g. GPR dataset needing
to be processed to optimize subsurface images and convert
from two-way time to depth), integrating survey data with
photographs (digital photogrammetry in this case) and GPR
(using spatial positioning information to locate accurately GPR
profiles, for example), to finally create the complete Digital
Solid Model. Whilst each study is site specific, a generalized
workflow can be constructed (Fig. 11). Whilst complicated, this
workflow is still a simplification of the steps involved during
this study.
Similarly intensive studies are starting to be undertaken on
outcrops in other turbidite sub-environments see Hodgetts
et al. (2004) for another well-documented example. Potentially,
these further studies could be integrated to construct a fully
detailed, deterministic model of turbidite fans and associated
sediments (Hodgetts et al. 2004). The workflow described can,
of course, be applied in other depositional environments.
CONCLUSIONS
The key outcrop geological parameters for 3D flow modelling
require a combination of surface and subsurface observational
techniques. Adapting conventional surveying and using digital
photogrammetric techniques generated accurate 3D surface
models of study sites. Integration of accurate sedimentary
information derived from cliff-face sections and interpretations
of subsurface GPR datasets allow the 3D volumetric component to be completed. The resulting 3D digital dataset
can then be analysed, in this case, for three-dimensional
sedimentary architectures. Initial reservoir models using 3D
volumetric information show significant discrepancies from
those built using surface information. Significant increased
connectivity between geobodies is present within reservoir
models using 3D volumetric outcrop analogue data. Sinuosity
variations of geobodies are larger using 2D input data than

Fig. 11. Summary workflow used to create a DSM of the test site at
Alport Castles. The workflow is six-stage:
(a) aerial photogrammetry produces a DEM and ORI; (b) terrestrial
photogrammetry produces a DEM and ORI; (c) low-resolution
(aerial) and high-resolution (terrestrial) datasets are integrated to
form a Digital Outcrop Model (DOM); (d) sedimentary logs and
observed cliff sections are interpreted to form spatially positioned
pseudo-well logs used as input for stochastic reservoir modelling
studies; (e) VRPs used to calibrate GPR profiles with outcrop cliff
faces, then 2D profiles and interpreted horizons are integrated with
positional information to form a 3D GPR dataset; (f) all data are
integrated to create a Digital Solid Model (DSM). Geometries are
extracted from the DSM for reservoir modelling studies.

those using 3D input data. Significant research effort is needed


to build a digital database of 3D outcrop analogue studies
using the techniques described, but this is expected to become
less onerous as new surveying and software tools become
available.
Financial assistance was provided by the Genetic Units Project,
GEOTIPE Projects and the Dynamic Geoscience Laboratory (partfunded by Fairfield Industries) all based at Heriot-Watt University.
The authors would also like to thank Dom Tatum, Ben Morgan,
Matt Porter, Justyn Clark and the Peak District National Park
wardens for field assistance and logistical support. GPR equipment
and software support was provided by Alan Hobbs of the National
Environment Research Councils (NERC) Geophysical Equipment
Pool (GEP), based in Edinburgh, UK. Rob Messenger, Emma
Preston and Bill Verkaik of Schlumberger supplied Petrel software
technical support. The PC-based VISTA software, supported by
Mike Galbraith and Randy Kolesar, was supplied by SIS Ltd,
Calgary. Reviewers, Steve Cossey and Dave Ellis, and editor, Antony
Treverton Buller, are thanked for constructive comments.

REFERENCES
Ahlgren, S. & Holmlund, J. 2002. Outcrop scans give new view. American
Association of Petroleum Geologists Explorer, July, 2223.
Arnot, M.J., Lewis, J.M. & Good, T.R. 1997. Photogeological and imageanalysis techniques for collection of large-scale outcrop data. Journal of
Sedimentary Research, 67, 984987.
Batzle, M. & Gardner, M.H. 2000. Lithology and fluids: seismic models of the
Brushy Canyon Formation, West Texas. In: Bouma, A. & Stone, C. (eds)
Fine-grained turbidite systems. American Association of Petroleum Geologists

352

J. K. Pringle et al.

Memoir 72 / Society of Economic Palaeontologists & Mineralogists


Special Publication, 68, 127142.
Bellian, J.A., Jennette, D.C., Kerans, C. & Gibeau, J. 2002. 3-D digital outcrop
data collection and analysis using eye-safe laser (LIDAR) technology.
Paper presented at the American Association of Petroleum Geologists
technical session, Quantitative Frontier of Stratigraphy and Sedimentology:
New Tools and Techniques for Outcrop Studies, 1013 March,
Houston, USA.
Boylan, A., Waltham, D., Bosence, D., Badenas, B. & Aurell, M. 2002. Digital
rocks: linking forward modeling to carbonate facies. Basin Research, 14,
401415.
Bryant, I., Carr, D., Cirilli, P., Drinkwater, N., McCormick, D., Tilke, P. &
Thurmond, J. 2000. Use of 3D digital analogues as templates in reservoir
modelling. Petroleum Geoscience, 6, 195201.
Chapin, M.A., Davies, S., Gibson, J.L. & Pettingill, H.S. 1994. Reservoir
architecture of turbidite sheet sandstones in laterally extensive outcrops,
Ross Formation, Western Ireland. In: Weimar, P., Bouma, A.H. & Perkins,
B.F. (eds) Gulf Coast Section Society of Economic Palaeontologists & Mineralogists
Foundation, 15th Annual Research Conference, Submarine Fans and Turbidite
Systems, December 47, 5368.
Clark, J.D. & Pickering, K.T. 1996. Submarine channels, processes and architecture.
Vallis Press.
Coleman, J.L., Browne, G.H. & King, P.R. et al. 2000. The inter-relationships
of scales of heterogeneity in subsurface, deep water E & P Projects
Lessons learned from the Mount Messenger Formation (Miocene),
Taranaki Basin, New Zealand. In: Weimar, P., Bouma, A.H. & Perkins,
B.F. (eds) Gulf Coast Section Society of Economic Palaeontologists & Mineralogists
Foundation, 20th Annual Research Conference Proceedings, Deep-Water Reservoirs of
the World, Houston, USA, December 36, 263283.
Collinson, J.D. 1970. Deep channels, massive beds and turbidity current
genesis in the central Pennine Basin. Proceedings of the Yorkshire Geological
Society, 37, 495520.
Corbeanu, R.M., Soegaard, K. & Szerbiak, R.B. et al. 2001. Detailed internal
architecture of a fluvial channel sandstone determined from outcrop, cores
and 3-D ground-penetrating radar: example from the middle Cretaceous
Ferron Sandstone, east-central Utah. American Association of Petroleum
Geologists Bulletin, 85, 15831608.
Cossey, S.P.J. 2000. Producibility of Deepwater Reservoirs in the North Sea
and Gulf of Mexico. GCSSEPM Foundation 20th Annual Research Conference,
Deep-Water Reservoirs of the World, Dec 36, 284292.
Davis, J.L. & Annan, A.P. 1989. Ground-penetrating radar for highresolution mapping of soil and rock stratigraphy. Geophysical Prospecting, 37,
531551.
Drinkwater, N.J. & Pickering, K.T. 2001. Architectural elements in a
high-continuity sand-prone turbidite system, late Precambrian Kongsfjord
Formation, northern Norway: application to hydrocarbon reservoir
characterisation. American Association of Petroleum Geologists Bulletin, 85,
17311757.
Dueholm, K.S. & Olsen, T. 1993. Reservoir analog studies using multimodel photogrammetry: a new tool for the petroleum industry. American
Association of Petroleum Geologists Bulletin, 77, 20232031.
Floris, F.J.T. & Peersmann, M.R.H.E. 2002. Integrated scenario and
probabilistic analysis for asset decision support. Petroleum Geoscience, 8, 16.
Geehan, G. & Underwood, J. 1993. The use of length distributions in
geological modelling. International Association of Sedimentologists Special
Publication, 15, 205212.
Hampson, G.J. 1997. A sequence stratigraphic model for deposition of the
Lower Kinderscout delta, an Upper Carboniferous turbidite-fronted delta.
Proceedings of the Yorkshire Geological Society, 51, 273296.
Hiscott, R.N., Pirmez, C. & Flood, R.D. 1997. Amazon submarine fan
drilling: a big step forward for deep-sea fan models. Geoscience Canada, 24
(1), 1324.
Hodgetts, D., Drinkwater, N.J., Hodgson, D., Kavanagh, J., Flint, S., Keogh,
K.J. & Howell, J. 2004. Three dimensional geological models from outcrop
data using digital data collection techniques: an example from the Tanqua
Karoo depocentre, South Africa. In: Curtis, A. & Wood, R. (eds) Geological
Prior Knowledge. Geological Society, London, Special Publications, 239,
5775.
Johnson, S.D., Flint, S., Hinds, D. & Wickens, H. 2001. Anatomy, geometry
and sequency stratigraphy of basin floor to slope turbidite systems, Tanqua
Karoo, South Africa. Sedimentology, 48, 9871023.
Kenyon, N.H., Amir, A. & Cramp, A. 1995. Geometry of the younger
sediment bodies of the Indus Fan. In: Pickering, K.T. & Hiscott, R.N. (eds)
Atlas of deep water environments: architectural style in turbidite systems, 2, 8993.
Kneller, B.C. 1995. Beyond the turbidite paradigm: physical models for
deposition of turbidites and their implications for reservoir prediction. In:
Hartley, A.J. & Prosser, D.J. (eds) Characterisation of Deep Marine Clastic
Systems. Geological Society, London, Special Publications, 94, 3149.

Kneller, B.C. & McCaffrey, W. 1999. Depositional effects of flow nonuniformity and stratification within turbidity currents approaching
a bounding slope: deflection, reflection, and facies variation. Journal of
Sedimentary Research, 69, 980991.
Lawrence, D.T. & Bosman-Smits, D.F. 2000. Exploring deepwater technical
challenges in the Gulf of Mexico. In: Weimer, P., Slatt, R. & Coleman, J.
(eds) Deep-Water Reservoirs of the World: Gulf Coast Section of SEPM 20th
Annual Research Conference, 473477.
Lien, T., Walker, R.G. & Martinsen, O.J. 2003. Turbidites in the Upper
Carboniferous Ross Formation, western Ireland: reconstruction of a
channel and spillover system. Sedimentology, 50, 113148.
McCabe, P.J. 1978. The Kinderscoutian Delta (Carboniferous) of northern
England; a slope influenced by density currents. In: Stanley, D.J. &
Kelling, G. (eds) Sedimentation in submarine canyons, fans and trenches. Dowden,
Hutchinson and Ross, Stroudsberg, 116126.
Peakall, J., McCaffrey, W.D. & Kneller, B.C. 2000. A process model for the
evolution of submarine fan channels: implications for sedimentary
architecture. In: Bouma, A.H. & Stone, C.G. (eds) Fine-grained Turbidite
systems. American Association of Petroleum Geologists Memoir 72/
Society of Sedimentary Geology Special Publication, 68, 7388.
Pettingill, H.S. 1998. World Turbidites 2, Lessons learned from 43 turbidite
giant fields. Oil & Gas Journal, 96, 9395.
Pickering, K.T. & Corregidor, J. 2000. 3D reservoir-scale study of Eocene
confined submarine fans, South-Central Spanish Pyrenees. In: Weimar, P.,
Bouma, A.H. & Perkins, B.F. (eds) Gulf Coast Section Society of Economic
Palaeontologists & Mineralogists Foundation, 20th Annual Conference of Deep-Water
Reservoirs of the World Proceedings, December 36 Houston, USA, 776781.
Pickering, K.T., Colman, J.M. & Cremer, M. et al. 1986. A high sinuosity,
laterally migrating submarine fan channel-levee-overbank: results from
DSDP Leg 96 on the Mississippi Fan, Gulf of Mexico. Marine and Petroleum
Geology, 3, 318.
Pringle, J.K. 2003. Integrated techniques for the acquisition and visualisation of 3-D
meso- to macro-scale sedimentary architectures of petroleum reservoir outcrop analogues.
PhD Thesis, Institute of Petroleum Engineering, Heriot-Watt University.
Pringle, J.K., Clark, J.D., Westerman, A.R., Stanbrook, D.A., Gardiner, A.R.
& Morgan, B.E.F. 2001. Virtual Outcrops: 3-D reservoir analogues. In:
Ailleres, L. & Rawling, T. (eds) Animations in Geology. Journal of the Virtual
Explorer, 3.
Pringle, J.K., Clark, J.D., Westerman, A.R. & Gardiner, A.R. 2003a. Using
GPR to image 3-D turbidite channel architecture in the Carboniferous
Ross Formation, County Clare, Western Ireland. In: Bristow, C.S. & Jol,
H. (eds) GPR in Sediments. Geological Society, London, Special
Publications, 211, 309320.
Pringle, J.K., Westerman, A.R., Clark, J.D., Guest, J.A., Ferguson, R.J. &
Gardiner, A.R. 2003b. The use of Vertical Radar Profiles (VRPs) in GPR
surveys of rock strata. In: Bristow, C.S. & Jol, H. (eds) GPR in Sediments.
Geological Society, London, Special Publications, 211, 221242.
Pringle, J.K., Westerman, A.R. & Gardiner, A.R. 2004. Virtual geological
outcrops: fieldwork and analysis made less exhaustive? Geology Today, 20 (2),
6469.
Satur, N., Hurst, A., Cronin, B.T., Kelling, G. & Grbz, K. 2000. Sand
body geometry in a sand-rich, deep-water clastic system, Miocene
Cingoz Formation of Southern Turkey. Marine and Petroleum Geology, 17,
239252.
Sgavetti, M. 1992. Criteria for stratigraphic correlation using aerial photographs: examples from the South-Central Pyrenees. American Association of
Petroleum Geologists Bulletin, 76, 708730.
Sikkema, W. & Wojcik, K.M. 2000. 3D visualisation of turbidite systems,
Lower Congo Basin, offshore Angola. In: Weimar, P., Bouma, A.H. &
Perkins, B.F. (eds) Deep-Water Reservoirs of the World, Gulf Coast Section Society
of Economic Palaeontologists & Mineralogists Foundation, 20th Annual Conference of
Deep-Water Reservoirs of the World Proceedings, December 36 Houston, USA,
928939.
Stephens, M. 1994. Architectural element analysis within the Kayenta
Formation (lower Jurassic) using ground-penetrating radar and sedimentological profiling, Southwestern Colorado. Sedimentary Geology, 90, 179211.
Szerbiak, R.B., McMechan, G.A., Corbeanu, R., Forster, C. & Snelgrove, S.H.
2001. 3-D characterization of a clastic reservoir analog: from 3-D GPR
data to a 3-D fluid permeability model. Geophysics, 66 (4), 10261037.
Visser, C.A. & Chessa, A.G. 2000. Estimation of length distributions from
outcrop datasets application to the Upper Permian Cutler Formation,
Utah. Petroleum Geoscience, 6, 2936.
Walker, R.G. 1966. Shale Grit and Grindslow Shales: transition from turbidite
to shallow water sediments in the Upper Carboniferous of Northern
England. Journal of Sedimentary Geology, 36, 90114.
Xu, X., Aiken, C.L. & Bhattacharya, J.P. et al. 2000. Creating virtual 3-D
outcrops. The Leading Edge, February, 197202.

Received 8 December 2003; revised typescript accepted 22 June 2004.

Вам также может понравиться