Download presentation
Presentation is loading. Please wait.
1
Siyao Xu, Andre Jung Tapan Mukerji and Jef Caers
Annual Meeting 2013 Stanford Center for Reservoir Forecasting Statistical Similarity of Stacking Patterns: Linking Tank Experiments to Real Scale Siyao Xu, Andre Jung Tapan Mukerji and Jef Caers
2
Background - Statistics and Rules
Advantage & Challenge of Surface-based Model After Bertoncello et. al. 2011 Statistics & Rules Generate & Place Geometry Sequentially T Stack Surfaces for 3D Realization
3
Background - Tank Experiment
Source of Statistics and Rules (Xu et. al. 2012) T Sequential Measurement Deposition-Erosion Pattern Erosion Rules and Statistics Sedimentology Group, SAFL
4
… New Questions Given Field Data Tanks Which tank? What part?
Thickness Maps Seismic Sections Well Logs Saller et. al. 2008 … Tanks Sedimentology Group, SAFL Which tank? What part?
5
Problem – Quantify Heirarchy
at small scale 5 Lobes In Tanks Lobes at different scales Interpretation infeasible Automatic algorithm needed at large scale 2 Lobes
6
Problem – Statistical Similarity
Stacking Patterns as the Link From the Field Saller et. al. 2008 From the Tank Statistical Similarity Sedimentology Group, SAFL
7
Methodology - Preprocessing
Crop Lobes From Overhead Photo Series T Sedimentology Group, SAFL
8
Distances of 4 Parameters
Choose Scale Lobes to Point Statistics of Pattern Compare Patterns Lobe Hierarchy Geobody Proximity Distances of 4 Parameters Between Source Point Between Orientation Between Shape (Procrustes Analysis) Between Polygon (Haussdorf Distance)
9
𝑤 𝑆𝑜𝑢𝑟𝑐𝑒 =0.35; 𝑤 𝑂𝑟𝑖𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛 =0.2;
Choose Scale Lobes to Point Statistics of Pattern Compare Patterns Lobe Hierarchy Geobody Proximity Geological Distance between Lobes Weighted sum of 4 parameters d= 𝑤 𝑆𝑜𝑢𝑟𝑐𝑒 ∗𝑆𝑜𝑢𝑟𝑐𝑒𝐷𝑖𝑠𝑡+ 𝑤 𝑂𝑟𝑖𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛 ∗𝑂𝑟𝑖𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛𝐷𝑖𝑓𝑓+ 𝑤 𝑃𝑜𝑙𝑦𝑔𝑜𝑛 ∗𝑃𝑜𝑙𝑦𝑔𝑜𝑛𝐷𝑖𝑠𝑡+ 𝑤 𝑆ℎ𝑎𝑝𝑒 ∗𝑆ℎ𝑎𝑝𝑒𝐷𝑖𝑓𝑓 Empirical weights 𝑤 𝑆𝑜𝑢𝑟𝑐𝑒 =0.35; 𝑤 𝑂𝑟𝑖𝑒𝑛𝑡𝑎𝑡𝑖𝑜𝑛 =0.2; 𝑤 𝑃𝑜𝑙𝑦𝑔𝑜𝑛 =0.35; 𝑤 𝑆ℎ𝑎𝑝𝑒 =0.1;
10
Agglomerative Hierarchical Clustering
Choose Scale Lobes to Point Statistics of Pattern Compare Patterns Lobe Hierarchy Geobody Proximity Agglomerative Hierarchical Clustering Using Geological Distance Repeat till all points are in one group Group Closest Pairs
11
Acceptable NOT acceptable Acceptable
Choose Scale Lobes to Point Statistics of Pattern Compare Patterns Lobe Hierarchy Geobody Proximity Clustering With Temporal Constraints d13=d12+d So d13 > d12 & d13 > d23 Time Space t1 d12 d23 t2 t3 Acceptable NOT acceptable Acceptable
12
Reachability Plot dAB A B Lobe Hierarchy Choose Scale Lobes to Point
Statistics of Pattern Compare Patterns Lobe Hierarchy Geobody Proximity Reachability Plot dAB A B Hierarchical Clustering
13
Reachability Plot A valley is a cluster Choose Scale Lobes to Point
Statistics of Pattern Compare Patterns Lobe Hierarchy Geobody Proximity Reachability Plot A valley is a cluster
14
Scale Reachability Plot
Choose Scale Lobes to Point Statistics of Pattern Compare Patterns Lobe Hierarchy Geobody Proximity Reachability Plot Put a threshold on y-axis (Scale Of Interpretation) Scale
15
Scale Reachability Plot Each cut valley is a big lobe Choose Scale
Lobes to Point Statistics of Pattern Compare Patterns Lobe Hierarchy Geobody Proximity Reachability Plot Each cut valley is a big lobe Scale
16
Mark lobes by source points
Choose Scale Lobes to Point Statistics of Pattern Compare Patterns Lobe Hierarchy Geobody Proximity Stacking Patterns Point Process Mark lobes by source points
17
ECDF of Proximity to Nearest Neighbor (NN)
Choose Scale Lobes to Point Statistics of Pattern Compare Patterns Lobe Hierarchy Geobody Proximity ECDF of Proximity to Nearest Neighbor (NN) Degree of local clustering NN Angle NN Polygon Distance NN Shape Similarity NN Source Distance Statistical Similarity
18
G Function & Attributes
Choose Scale Lobes to Point G Function & Attributes Compare Patterns Lobe Hierarchy Geobody Proximity 2 Sample Kolmogorov-Smirnov Test Inputs: 2 ECDFs Outputs: P-value Higher More similarity 0.067 Sample 1 Subset of 1 Sample 2 >>
19
Case Study - Patterns from Real Systems
Jegoua et. al. 2008 Amazon – 14 Lobe Complexes Borneo – 18 Lobes Saller et. al. 2008
20
Case Study – Patterns from Tanks
Tank B – 184 lobes Tank A – 424 lobes Sedimentology Group, SAFL
21
Case Study – Application of Method
Scale of Interpretation … 2. Hierarchical Clustering (Reachability Plot) 4. Compare 4 ECDFs Respectively 3. Choose a Scale 1. Measured Finest Scale
22
Case Study – Application of Method
5. Repeat for All Scales Scale of Interpretation …
23
Case Study – Application of Method
5. Repeat for All Scales Scale of Interpretation …
24
Case Study – Source CDF (Tanks vs. Borneo)
20 40 60 80 0.2 0.4 0.6 0.8 1 Scale of Interpretation (km) Similarity (P-Value) Tank A – More Significant Tank B - Less Significant Range of Interest For Surface-based Modeling X-axis: Calibrated to Real Scale Reservoir
25
Case Study – Angle CDF(Tanks vs. Borneo)
20 40 60 80 0.2 0.4 0.6 0.8 1 Scale of Interpretation (km) Similarity (P-Value) Tank B - Less Significant Tank A – More Significant Range of Interest
26
Case Study – Polygon Distance CDF(Tanks vs. Borneo)
20 80 0.2 0.4 0.6 0.8 1 Scale of Interpretation (km) Similarity (P-Value) 40 60 Tank A – More Significant Tank B - Less Significant Range of Interest
27
Case Study – Shape Similarity CDF(Tanks vs. Borneo)
20 80 0.2 0.4 0.6 0.8 1 Scale of Interpretation (km) Similarity (P-Value) 40 60 Tank A & Tank B – Approximate Significance
28
Case Study – 4 ECDFs (Tanks vs. Amazon)
Source Point Distance 0.2 0.4 0.6 0.8 1 100 200 Angle 0.2 0.4 0.6 0.8 1 100 200 Similarity (P-Value) Tank B - better Tank A – better 0.2 0.4 0.6 0.8 1 Polygon Distance 100 200 Shape Similarity 0.2 0.4 0.6 0.8 1 100 200 Tank A – better Tank B - better Scale of Interpretation (km)
29
Case Study - Summary Borneo Tank A more similar Amazon Neither
Different parameters Different scale of interest Scale of interest Assist surface-based modeling
30
Conclusion Statistical similarity
Relative proximity Data mining with geological constraints Hierarchy of geology Compare Stacking Patterns Most similar tank & scale of interest Assist Surface-based Modeling Patterns at scale of interests
31
Future Works Geological Side Engineering Side Geological validation
Calibrate method on different tanks Engineering Side Given a stacking pattern How to simulate?
32
Acknowledgement Professor Chris Paola
Sedimentology Group, San Anthony Falls Laboratory, University of Minnesota
33
Conclusion Statistical similarity
Relative spatial relationships Data mining with geological constraints Hierarchy of geology Compare Stacking Patterns Most similar tank & scale of interest Assisting Real Scale Modeling Patterns at scale of interests
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.