Development and Testing of New Automated Methods for the Capture of Quantitative Fracture Data from Outcrop Analogues* David W. Hunt1, Paul Gillsepie4, John Thurmond2, Giulio Casini2, and Erik Monsen3 Search and Discovery Article #40573 (2010) Posted July 30, 2010 *Adapted from oral presentation at AAPG Annual Convention and Exhibition, New Orleans, Louisiana, April 11-14, 2010. 1
Technology & New Energy, Statoil, Bergen, Norway (
[email protected]) Technology & New Energy, Statoil, Bergen, Norway 3 Schlumberger Stavanger Research, Schlumberger, Stavanger, Norway 4 Global Exploration Technology, Statoil, Stavanger, Norway 2
Abstract Natural fracture systems commonly act as an important control on the production of hydrocarbons from carbonate reservoirs. In the subsurface, significant enhancement of fracture models can be gained through the study of appropriate outcrop analogues (i.e. fracture orientation, spacing, height, connectivity etc.). However, a major bottleneck in the utilization of outcrop data is the time required for the collection of statistically meaningful fracture data from geological field work and/or remote sensing data. In order to eliminate this bottleneck we have successfully adapted 3D seismic technologies, originally developed for automated fault extraction, for the automated extraction of bedding and fracture data from a range of different digital remote sensing data types. The results derived from automatic analysis of the remote sensing data have been independently quality-controlled using traditional outcrop-based field studies on world-class carbonate exposures (USA, Europe, Middle East). The analyzed data include ground-based and airborne LIDAR-derived photorealistic models and orthorectified Quickbird satellite imagery combined with satellite-derived digital elevation models. Data are derived from surface surveys and also subsurface tunnel surveys. The examples have been chosen in order to 1) capture variability in terms of fractured carbonate reservoir types and structural setting and 2) to develop and prove the technology using a range of remote sensing data types and different data qualities. We contend that the research has led to development of a rapid and robust method that allows for the extraction of statistically representative fracture populations. The new technology frees the structural geologist from laborious digitizing work, and provides access to a plethora of relevant fracture data. The technology therefore allows the geologist to better focus on the interpretation and analysis of relevant outcrop analogue data in order to better parameterize the building of subsurface fracture models.
Copyright © AAPG. Serial rights given by author. For all other rights contact author directly.
Development and Testing of New Automated Methods for the Capture of Quantitative Fracture Data From Outcrop Analogues David Hunt, Paul Gillespie, John Thurmond, Giulio Casini (Statoil) Erik Monsen (Schlumberger)
1-
Objectives • Rationale • Photorealistic models, purpose specific LIDAR/Mesh - Vector
• Method Overview • Independent QC, results evaluation Image – Ant tracking
• Conclusions, directions
2-
Business Driver - Rationale • Middle East:- Fractured carbonate reservoirs intervals exposed only a few km from producing field/discoveries. • But:- Exceptional exposures, but are very steep, difficult to work on and therefore extract meaningful quantitative data (i.e. fracture stratigraphy).
N Fracture Density P10 (/m)
27
0
2
4
6
8
10
12
14
16
A1-1 A1-3 A2-2 B1 B2-2 C1 C2-2 C3-1 D1-1 D1-3
300m
174
10
D2-2 E1-1 E2-1 E2-3 E3-2 E3-4 E3-6 F1-2 F2-2
2
G1-1 G2-1
15
Kxx
120 m
G3-1 G3-3 H1-1
Mishrif, Middle East
16
H2 H4
3-
Presenter’s Notes: 1.
StatoilHydro geoscientists have had the opportunity to characterize fractured and diagemetically-modified carbonate reservoirs at on exceptional exposures that are located only a few kilometres from major discoveries and producing fields in the Middle east.
2.
A major challenge related to this work was the scale and steepness of the exposures, making it difficult to collect large quantoties of quantitative data. The images show fracture heterogeneity exposed on large bedding planes and also in steep section. The fractures clealy vary in orientation between different reseervoir layers.
3.
The fracture density also changes between reservoir layers. However, the density differences are based on 2D interpretation from photographs, and contain to diectional information.
4.
Fracture map (50 m squares) or Fracture map illustrates a grid of 50 m squares.
5.
The primary aim of the project, initiated in 2004, was to develop new techniques for the automated extraction of quantitative fracture data from outcrop analogues. The company is studying outcrops where the reservoir section is exposed only a few kilometres away form the prodicing fields, and show comparable fracturing in terms of both orientation and relationships to bedding/bed sets. Unfortunately despite superb exposures, access is limited and so the actuay capacity of the geologist to measure large numbers of stastically- represensentative quantitative fracture data is limited.
3
Business Driver - Rationale • Objective:- To develop rigorous automated techniques to rapidly and accurately extract quantitative fracture data from outcrop analogue data. • Initiative:- 2004 sought collaboration with Schlumberger to adapt ’ant tracking’ seismic technologies to work on digital outcrop datasets.
N Fracture Density P10 (/m)
27
0
2
4
6
8
10
12
14
16
A1-1 A1-3 A2-2 B1 B2-2 C1 C2-2 C3-1 D1-1 D1-3
300m
174
10
D2-2 E1-1 E2-1 E2-3 E3-2 E3-4 E3-6 F1-2 F2-2
2
G1-1 G2-1
15
Kxx
120 m
G3-1 G3-3 H1-1
Mishrif, Middle East
16
H2 H4
4-
Presenter’s Notes: 1.
StatoilHydro geoscientists have had the opportunity to characterize fractured and diagemetically-modified carbonate reservoirs at on exceptional exposures that are located only a few kilometres from major discoveries and producing fields in the Middle east.
2.
A major challenge related to this work was the scale and steepness of the exposures, making it difficult to collect large quantoties of quantitative data. The images show fracture heterogeneity exposed on large bedding planes and also in steep section. The fractures clealy vary in orientation between different reseervoir layers.
3.
The fracture density also changes between reservoir layers. However, the density differences are based on 2D interpretation from photographs, and contain to diectional information.
4.
Fracture map (50 m squares) or Fracture map illustrates a grid of 50 m squares.
5.
The primary aim of the project, initiated in 2004, was to develop new techniques for the automated extraction of quantitative fracture data from outcrop analogues. The company is studying outcrops where the reservoir section is exposed only a few kilometres away form the prodicing fields, and show comparable fracturing in terms of both orientation and relationships to bedding/bed sets. Unfortunately despite superb exposures, access is limited and so the actuay capacity of the geologist to measure large numbers of stastically- represensentative quantitative fracture data is limited.
4
LIDAR-Derived Photorealsitic Models • Laser scans outcrop at 10’s of millions of points • Adaptive intelligent mesh made from scan – reduces data size by c. 90% • Photo’s mapped onto this with cm-scale accuracy
5-
10 m
10 m
Data: Photorealistic Models • Same data, 3 models with different resolution Medium resolution
Low resolution
360 x 270 m Main stratigraphic surfaces
Stratigraphic detail, Karst mapping, facies High resolution
Fracture mapping
6-
Data: Photorealistic Models • Same data, 3 models with different resolution Medium resolution
360 x 270 m Main stratigraphic surfaces
Stratigraphic detail, Karst mapping, facies High resolution
Fracture mapping
7-
Data: Photorealistic Models • Same data, 3 models with different resolution Medium resolution
360 x 270 m Main stratigraphic surfaces
Stratigraphic detail, Karst mapping, facies High resolution
Fracture mapping
8-
Data: Photorealistic Models • Same data, 3 models with different resolution
360 x 270 m Main stratigraphic surfaces
Stratigraphic detail, Karst mapping, facies High resolution
Fracture mapping
9-
Data: Photorealistic Models • Same data, 3 models with different resolution Medium resolution
Low resolution
360 x 270 m Main stratigraphic surfaces
Stratigraphic detail, Karst mapping, facies High resolution
Fracture mapping
10 -
Data: Photorealistic Models • Same data, 3 models with different resolution
360 x 270 m Main stratigraphic surfaces
Stratigraphic detail, Karst mapping, facies High resolution
Fracture mapping
11 -
Data: Photorealistic Models • Same data, 3 models with different resolution Medium resolution
Low resolution
360 x 270 m Main stratigraphic surfaces
Stratigraphic detail, Karst mapping, facies
Fracture mapping
LIDAR – c 10 million points Each image (4500 x 3000 = 13 500 000) 12 -
7 images used
22.5 m
Image Resolution at 0.7-1.1 km (13.5 mpixels)
13 -
Image Resolution at 0.7-1.1 km (13.5 mpixels)
22.5 m
< 2 cm/pixel
14 -
LIDAR point cloud
Results: Fracture mapping Original 2D high resolution image
3D properties
World Image transform Images
Interpolation to image grid Normal vectors
Distance to scene
Edge enhancing filtering Ant Tracking Image World transform
High resolution
15 -
3D Interpretation
LIDAR intensity
LIDAR point cloud
Results: Fracture mapping 2D image curvature
3D properties
World Image transform Images
Interpolation to image grid Normal vectors
Distance to scene
Edge enhancing filtering Ant Tracking Image World transform
High resolution
16 -
3D Interpretation
LIDAR intensity
LIDAR point cloud
Results: Fracture Mapping 2D iterative Ant Tracking
3D properties
World Image transform Images
Interpolation to image grid Normal vectors
Distance to scene
LIDAR intensity
Edge enhancing filtering Ant Tracking - 1 Image World transform 3D Interpretation
High resolution
17 -
Note horizontal tracking shown
LIDAR point cloud
Results: Fracture Mapping 2D iterative Ant Tracking
3D properties
World Image transform Images
Interpolation to image grid Normal vectors
Distance to scene
LIDAR intensity
Edge enhancing filtering Ant Tracking - 2 Image World transform 3D Interpretation
High resolution
18 -
Note horizontal tracking shown
Results: Fracture Mapping 2DImage World Transform
LIDAR point cloud
3D properties
World Image transform Images
Interpolation to image grid Normal vectors
Distance to scene
LIDAR intensity
Edge enhancing filtering Ant Tracking Image World transform 3D Interpretation
10 m
• Note:- intersection of fracture/bedding with surface topography retained as a confidence factor of the planes strike and dip
19 -
LIDAR point cloud
Results: Fracture mapping
Images
Ant-tracks –image analysis
Interpolation to image grid Normal vectors
Fracture no
Dip
Dip direction
: 137 :
: 64 :
: 192 :
Ant Tracking
3D Interpretation
137
81
100
71
91
6
4 8 9
116
107
23
26
96
30 85
36 73
35
111
72
130
77 121
37
41
39
86
53 52
61
59
88
64
89
57
42 87
46
55
50
90
56
131
145 98
123
65
132
68
122 67 69
20 -
40
38
54
112 133
63
70
134
141
80
124
79 135
120
29
45 49
144
113
13676
84
44
58
62
78
143 97
51
60
66
31
32
129
33
110
34
43 48
24
119
27
19
22
83
25
109
95
47
108 118
21
106 28
82 128
20
104 105 75
16
93 103
15
117
11
139
17 18
74 94
142
138
102
12
14
10
126
7
127 140
99
101
114
2
5
115
13
LIDAR intensity
Image World transform
1 3
Distance to scene
Edge enhancing filtering
High resolution 92
3D properties
World Image transform
125
Field Data vs Ant Track Data
21 -
4.75 m
Ant Tracking: Quality Control - Mesh
Model 3 - Fracture
Model 1 - Stratigraphy
22 -
Mechanical Stratigraphy • Final LIDAR study phase:- develop and qc fracture-bedding relationships ’mechanical stratigraphy’ • 2007 field study (blind from Schlumberger) vs fully automated result – minor discrepencies Field data
23 -
Automated result
LIDAR point cloud
Results: Fracture mapping Y-component of normal vector field
3D properties
World Image transform Images
Interpolation to image grid Normal vectors
Distance to scene
Edge enhancing filtering Ant Tracking Image World transform
LIDAR - Medium/low resolution
24 -
3D Interpretation
LIDAR intensity
LIDAR point cloud
Results: Fracture mapping Y-component of normal vector field Vector analysis
3D properties
World Image transform Images
Interpolation to image grid Normal vectors
Distance to scene
LIDAR intensity
Edge enhancing filtering Ant Tracking Image World transform 3D Interpretation
Centre points
Best fit planes
25 -
Accuracy from mesh determined by photorealistic model inputs
Field Data vs Mesh Analysis - 1
Mesh analysis 2 clusters Superimposition on data
26 -
Field Data vs Mesh Analysis - 2
Mesh analysis – main cluster Difffers from field data
27 -
Field Data vs Mesh Analysis - 2
Mesh analysis – main cluster Difffers from field data
28 -
Field Data vs Mesh Analysis - 3
• Why the contrast between field data and vector analysis – are the vector data unreilable? • No – these sets exist in specific parts of the section 29 -
Field vs Automated Data Distribution
30 -
Field vs Automated Data Distribution
31 -
Field vs Automated Data Distribution
32 -
Field vs Automated Data Distribution
33 -
Results, Conclusions •
Different orientation, scales of data obtained from image analysis and vector analysis of LIDAR/mesh data – complimentary LIDAR/Mesh - Vector
Image – Ant tracking
34 -
Results, Conclusions •
Different orientation, scales of data obtained from image analysis and vector analysis of LIDAR/mesh data – complimentary
•
Limitations and strengths of the technology identified
LIDAR/Mesh - Vector
− Image quality, illumination, weatherning − Mesh quality, cliff topology
35 -
Image – Ant tracking
Results, Conclusions •
Different orientation, scales of data obtained from image analysis and vector analysis of LIDAR/mesh data – complimentary
•
Limitations and strengths of the technology identified
LIDAR/Mesh - Vector
− Image quality, illumination, weatherning − Mesh quality, cliff topology •
36 -
Blind testing & independent QC of study area
Image – Ant tracking
Conclusions
37 -
•
Technique is complimentary to, not a replacement for field geology
•
There is a bias in both field and automated datasets
•
The automated results force you to reevaluate field collected data and to develop better fracture models
Study Areas, Data Types 100’s m
1.
38 -
Guadalupe Mountains, USA - Ground Based LIDAR, photorealitsic model
km
2. Somerset, England, Famous Bench Ground-based LIDAR, Aerial + ground-based photorealistic models
Multi km
-
3. Mishrif, Middle East Ground basd LIDAR, 8 m tunnels through anticline, Integrated with orthorectified Quickbird data
Study Areas, Data Types High 100’s contrast m
1.
39 -
Guadalupe Mountains, USA - Ground Based LIDAR, photorealitsic model
High km contrast
2. Somerset, England, Famous Bench Ground-based LIDAR, Aerial + ground-based photorealistic models
High +low Multi contrast km
-
3. Mishrif, Middle East Ground basd LIDAR, 8 m tunnels through anticline, Integrated with orthorectified Quickbird data
Thank you Development and Testing of New Automated Methods for the capture of quantitative fracture data from outcrop analogues David William Hunt Project Manager, Carbonate Reservoirs & Traps
[email protected], tel: +47 97985842 www.statoil.com
40 -
