PS

Carbonate Reservoir Modeling Using

Multiple-Point Statistics (MPS) / Facies Distribution Modeling (FDM)*
By
Marjorie Levy , Paul M. (Mitch) Harris1, and Sebastien Strebelle1
1

Search and Discovery Article #40293 (2008)

Posted August 1, 2008

*Adapted from poster presentation at AAPG Annual Convention, Houston Texas, April 9-12, 2006

1
Chevron Energy Technology Company, San Ramon, CA (levm@chevron.com; MitchHarris@chevron.com; stsb@chevron.com)

Abstract

We have explored the use of MPS/FDM modeling in a carbonate reservoir. We have modeled the platform top of an isolated platform
example and tested various scenarios for the distribution of grainstone facies.

The training image is a 3D conceptual model of the reservoir, containing information about facies dimensions and relationships among
facies. Five facies were considered: Bar crest = best reservoir quality due to sorting in highest energy setting; Bar flank = good
reservoir quality between bar crests, includes flanks of bars and intervening tidal channels; Island = localized areas where permeability
is enhanced by dissolution during meteoric diagenesis; Deeper platform = poorer reservoir quality in platform areas away from bars
and channels; and Background = tight intervals due to muddier facies or to porosity-plugging cementation.

The facies probability cube allows controlling the spatial distribution of the facies in the MPS model. First, facies depocenter maps
were generated for deeper platform, bar flank, bar crest and island. Then, the stratigraphy of the reservoir was modeled by digitizing a
vertical proportion curve reflecting the variations of facies proportions with depth. Three alternative vertical proportion curves were
created, representing respectively a gradual trend, cyclicity at the scale of composite sequences, and high cyclicity at the scale of
individual sequences. Corresponding alternative facies probability cubes were generated for these three cases.

Several scenarios were run: the gradual, cyclic and highly cyclic cases; both narrow and wide bar crests and bar flanks; and with
constant and variable azimuth. The wide bar crest/bar flank and very cyclic simulation produce results that qualitatively appear most
reasonable in both cross section and map views.
Carbonate Reservoir Modeling Using Multiple-Point Marge Levy, Paul (Mitch) Harris, and Sebastien Strebelle
Chevron Energy Technology Company, San Ramon, CA
Statistics (MPS) / Facies Distribution Modeling (FDM)
We have explored the use of MPS/FDM modeling in a carbonate Training Image Generation MPS (Multiple-Point Statistics) FDM (Facies Distribution Modeling)
reservoir. We have modeled the platform top of an isolated platform
example and tested various scenarios for the distribution of
1. Describe geometry of each facies:
grainstone facies. Define map view and cross-section shapes MPS is an innovative depositional facies
The training image is a 3D conceptual model of the reservoir, modeling technique, developed by Chevron in
Ellipse Lobe Semi-ellipse Sigmoid collaboration with Stanford University, that uses 3D
containing information about facies dimensions and relationships Vertical
among facies. Five facies were considered: Bar crest = best reservoir conceptual geological models as training images to Geologic
integrate geological information into reservoir models. Proportion Facies
quality due to sorting in highest energy setting; Bar flank = good Cross-section Section Depocenter
reservoir quality between bar crests, includes flanks of bars and Replacing the traditional variogram with a training
image allows MPS to capture complex spatial Trend Map
intervening tidal channels; Island = localized areas where
Define dimensions, orientation, sinuosity relationships between multiple facies, and to model
permeability is enhanced by dissolution during meteoric diagenesis;
Deeper platform = poorer reservoir quality in platform areas away non-linear shapes such as sinuous channels that
Orientation
from bars and channels; and Background = tight intervals due to conventional variogram-based modeling techniques
muddier facies or to porosity-plugging cementation. Width typically fail to reproduce. In addition, because MPS is Well Log
The facies probability cube allows controlling the spatial Length Sinuosity not an object-based, but still a pixel-based algorithm,
Facies
distribution of the facies in the MPS model. First, facies depocenter amplitude MPS models can be constrained by very large
Thickness and numbers of wells and 3D facies probability cubes
maps were generated for deeper platform, bar flank, bar crest and wave length
island. Then, the stratigraphy of the reservoir was modeled by derived from seismic data or from reservoir facies
digitizing a vertical proportion curve reflecting the variations of deposition interpretation (using FDM).
facies proportions with depth. Three alternative vertical proportion For the last three years, the MPS/FDM workflow
2. Specify relationships among facies: has been preferred to variogram-based and object-
curves were created, representing respectively a gradual trend,
cyclicity at the scale of composite sequences, and high cyclicity at Define facies Define vertical and/or based techniques to model important Chevron assets
the scale of individual sequences. Corresponding alternative facies erosion rules horizontal constraints in both shallow-water clastic reservoirs (Indonesia and
probability cubes were generated for these three cases. Angola), and deepwater reservoirs (Gulf of Mexico,
Several scenarios were run: the gradual, cyclic and highly North Sea, Angola, and Nigeria). In those projects, the
cyclic cases; both narrow and wide bar crests and bar flanks; and MPS/FDM workflow enabled the generation of
2 geologically realistic facies models, which significantly
with constant and variable azimuth. The wide bar crest/bar flank and 3D Facies Probability Cube
very cyclic simulation produce results that qualitatively appear most 1 improved reservoir OOIP and oil recovery uncertainty
reasonable in both cross section and map views. assessment.

Rules-Based Deepwater Sheet

Shallow Marine Deepwater Channel /
Estuarine System System Training Image
System Sheet Complex

Training Image
Library:

Fluvial System Prograding Shallow Rudist Grainstone / Deepwater Channel Training Image
Marine/Tidal System Reef Margin with Internal Facies Architecture
 Carbonate Reservoir Modeling Example Using MPS / FDM
Build
4b. Single Facies Training Images
Sgrid
3. Model Region Carbonate platform Lens/upper half ellipse
Build
Training Image and Layering
based on
Stratigraphic model
Platform
Build facies
Probability Cube
Region
Using FDM

Build
Azimuth Field
Outline of MPS Fine-scale MPS sgrid. There are
sgrid; Areal cell 45 layers, with an average cell 800 x 150 x 6m
Island
MPS
Facies Simulation dimensions 50m thickness of ~2 meters. Oriented 45
x 50m Bar crest

1. Carbonate Workflow QC & Recycle Ellipse/upper half ellipse Lens/upper sigmoid Bar flank

Deeper
platform
2. Study Rationale
4. Build Training Images Bkgrd
400 x 150 x 3m 800 x 150 x 6m
4a. Rationale of Training Images Oriented 135 Oriented 45
Focus on isolated carbonate
platform example Platform top has widespread grainstones. We suspect that some type of bar
and channel (flank) morphology is likely.
Use modern analogs for Island = localized areas where permeability is 4c. Combined Facies Training Images
conceptual facies models enhanced by dissolution during meteoric
diagenesis

Test application of current

MPS/FDM tools to carbonate Bar crest = best reservoir quality due to
sorting in highest energy setting
modeling
Bar flank = good reservoir quality
Develop new tools specific between bar crests, includes flanks of
for carbonates bars and intervening tidal channels

Deeper platform = poorer reservoir

quality in platform areas away from
Model only platform and test bars and channels
varied scenarios for
distribution of grainstone Background = tight intervals due to
muddier facies or to porosity-plugging
facies cementation Landsat image
Berry Islands, Great Bahama Bank Training image using the bar Training image using the bar
crest narrow and bar flank crest wide and bar flank wide
narrow facies files facies files
 5. Generate Facies Probability Cube 7b. MPS Simulation in Map View
6. Add Well Data Conditioning and other
Constant Azimuth Field Variable Azimuth Field
5a. Facies Modeling Constraints
Depocenter
Maps 6a. Control from Cored Wells
Bar Crest MPS simulation with
Facies control provided from a subset of wide bar crest and wide
Bar Flank platform wells with considerable core bar flank facies

Deeper
Island
Platform MPS simulation with
Background narrow bar crest and
Deeper Platform narrow bar flank facies
Bar Flank
Bar Crest
Island
5b. Facies Proportion Curves
Gradual Cyclic
7c. Wide Barcrest/Barflank, High Cyclicity Model
High Cyclicity
Lithofacies described in each core were regrouped into the MPS/FDM
facies
MPS/FDM facies indices were assigned to each cored well and used
for conditioning of the model

Global Facies Proportions: Global Facies Proportions: Global Facies Proportions:

6b. Add Variable Azimuth
0: 13; 1: 40; 2: 19; 3: 25; 4: 3 0: 15; 1: 40; 2: 18; 3: 24; 4: 3 0: 20; 1: 40; 2: 16; 3: 21; 4: 3
Layer 17: Layer 18:
Higher energy facies; note how the Lower energy facies; note how background Layer 19:
5c. Alterntive Facies Probability Cubes central platform is dominated by facies impinges on deeper platform and Higher energy facies return
bar crest and bar flank facies high energy barcrest/barflank facies
Gradual
Cyclic High Cyclicity
Background Facies
Background Facies Background Facies 7. Run MPS Simulation
8. Model Summary
Matches facies regions of platform top which
7a. MPS Simulation in Cross Section were defined during reservoir characterization

High Cyclicity Model Matches stratigraphic variation defined during Facies Maps
Cyclic Model, 4 cycles
Gradual Model 14 cycles
No cycles
(= composite sequences) sequence stratigraphy studies with facies
Bkgrd Bkgrd Bkgrd (= sequences)
Layer 11 Layer 14 proportion curves
.8 Layer 11
Matches conceptual models for facies types
from modern analogs with training images
MPS
Generally matches facies identified in cored Simulation
Bar Crest Bar Crest Bar Crest wells
Layer 11 Layer 14 Layer 14
0