1
Achieving Geologically Reasonable 3D Models Presentation by Eric de Kemp 3-D Imaging and Earth Modelling Geological Survey of Canada
Mike Hillier, Ernst Schetselaar, Gabriel Courrioux, Guillaume Caumon Florian Wellmann, Mark Lindsay, Gautier Laurent, Mark Jessell, Laurent Aillères
2
Introduction
What we are doing at GSC Implicit modelling Geologically Reasonable ? MapSim - Lets go back to 2D ! Spatial Agents (SABM) Conclusions
3-D Imaging and Earth Modelling Geological Survey of Canada
Geophysical and geological integration, visualization and modelling Ottawa Group: Don White – Seismology (CCS) Gilles Bellefleur – Seismology (Minerals) David Snyder – Teleseismic Ernst Schetselaar – 3D Integration Mike Hillier – Scientific computation Brian Roberts - Seismology (Hardware) Eric de Kemp - 3D Regional Modelling Jim Craven - Electromagnetics (MT) Maurice Lamontagne – Active Seismology Post Docs … PhD Msc + BSc Students
4
Sorting out the mess …
Standard approach to mapping !
5
Look at geometry and properties together !
6
Small things have big impact !
7
3D GRID MODELLING OF THE LALOR VMS DEPOSIT, SNOW LAKE MANITOBA, CANADA: BRIDGING THE GAP BETWEEN GEOLOGY AND GEOPHYSICS
Ernst Schetselaar, Gilles Bellefleur, Pejman Shamsipour
8
Building a Curvilinear-faulted Property Grid 6
1
2 5
4
3
9
Inputs to 3D modelling of geologic surfaces (steps 1-4) previous slide
Unit contacts from geologic map (Bailes et al, 1993)
NE
lithostratigraphic markers from drill logs (Bailes 2012)
lithofacies recoded from Hudbay drill logs
strike/dip s0 restored from bedding-core angles (Bailes, 2012)
SW 3D geologic surface model Lalor deposit
10
Curvilinear grid modelling meta-volcanic and volcaniclastic units Lalor deposit NE
SW
Nominal cell size: 20 x 20 x 5m
11
Seismic property modelling & simulation Density
Seismic response Flin Flon mine horizon, D. Melanson, 2014 using open source code of SOFI3D finite difference modelling Bohlen, 2002
Vp
Vs
Stochastic seismic property models Lalor deposit
12
Lessons learned…………. Incorporating geological expertise in the right way is vital to success (mineralogy, physical process, modelling) Systematic and generalized classification of rock types (lithofacies) required for data reconciliation and integration Big advantage of CEM: unprecedented opportunities for data reconciliation: natural driver for geophysicists, geochemists, geologists to join efforts in better understanding ore systems. Importance of reference holes - geophysical drill log data play an important role in upscaling physical rock properties (from sample – mine camp scale)
13
3D Modelling of the Purcell Anticlinorium “A project to make a 3D regional model – beyond the headframe…”
14
Purcell 3D
3D Model – Fault Network From Thomas 2013 Magnetic Modelling
C-S
F2 , L12
Map trace of fault
Line 11
Offset Moyie Dyke reflectors RMT SHORT COURSE 7: 3D INTEGRATED INTERPRETATION FOR GEOLOGIC MODELLING OF THE PURCELL-SULLIVAN SEDEX MINERAL SYSTEM 24 JANUARY 2015
Moyie thrust
SEDEX System & 3D Modelling • Ore system is embedded in current structural-stratigraphic architecture • Making a 3D model is essential to characterize the whole system • Needed new tools to deal with sparse regional data • The process is multi-scale, iterative and integrative • Need a process to do this or a ‘workflow’ to stay focused
SHORT COURSE 7: 3D INTEGRATED INTERPRETATION FOR GEOLOGIC MODELLING OF THE PURCELL-SULLIVAN SEDEX MINERAL SYSTEM 24 JANUARY 2015
Sullivan horizon (contact between Middle – Lower Aldridge Formation = LMC)
SHORT COURSE 7: 3D INTEGRATED INTERPRETATION FOR GEOLOGIC MODELLING OF THE PURCELL-SULLIVAN SEDEX MINERAL SYSTEM 24 JANUARY 2015
‘Barcode’ siltite markers provide unique 3D modelling constraints in hanging wall of the Sullivan horizon
Regional 3D Modelling Challenge, Purcell anticlinorium A) primary drill hole constraints
?
B) primary drill hole + secondary strike/dip constraints
SHORT COURSE 7: 3D INTEGRATED INTERPRETATION FOR GEOLOGIC MODELLING OF THE PURCELL-SULLIVAN SEDEX MINERAL SYSTEM 24 JANUARY 2015
20
Implicit surface modelling using Radial Basis Functions (RBF’s) s(x)
λiφi(x)
λ1φ1(x)
λ2φ2(x) Source: department of Information Technology, Uppsala University N
N
i =1
i =1
s(x) = ∑ λ i φ | x − x i | ≡ ∑ λ i φ i (x) φ : radial basis function [0, ∞] φ(r ) = r r = | x − xi | φ(r ) = r 3 φ(r ) = r 2 ln( r ) φ(r ) = e
− ( εr ) 2
Carr et al., 2000 source: Leapfrog solutions
21
Implicit approach at GSC
HILLIER, M., SCHETSELAAR, E. M. DE KEMP E. A. AND PERRON, G., 2014, MATHEMATICAL GEOLOGY, V46, P. 931-953.
22
Generalized RBF interpolation traditional RBF
gradient constraint
tangent / fold axis constraint
HILLIER, M., SCHETSELAAR, E. M. DE KEMP E. A. AND PERRON, G., 2014, MATHEMATICAL GEOLOGY, V46, P. 931-953.
23
Implicit surface modelling
on-contact points f(xi,yi,zi) = 0
i = {1,.,.,n}
off-contact points f(xi,yi,zi) > 0 f(xi,yi,zi) < 0 i={ n+1,.,.,N}
d0
Surface is extracted by tracing zero equipotential from a 3D scalar signed distance function
25
Including Inequality Points Without Inequality pts
With Inequality pts
off-contact points f(xi,yi,zi) > 0 f(xi,yi,zi) < 0 i={ n+1,.,.,N}
27
Radial Basis - Selection R3 – 3rd order radial
Non geologic at regional scale
MQ - Multiquadradic
Stratigraphic geometry
28
Radial Basis - Selection
Notice critical threshold for Shape parameter on MQ will send the RBF into R3 behavior, see overturned surface far away from data is not supported
29
Smoothness MQ Slack 50
MQ Slack 200
30
Arbitrary topology is the deal !
TURK and O’BRIEN, 2002
http://www-graphics.stanford.edu
31
Reasonable Models ?
32
Reasonable Models ? How do we measure “reasonableness” ? When there is poor – sparse data How do we extend our models ?
?
33
Geological Reasonable Models Reasonable models fit our knowledge & data Where do we start… what specific knowledge is essential ? A bit of knowledge with the right data can have a big impact ! How do we get knowledge embedded into our models ? Geology models must be compatible with geological processes and the expected patterns through earth history.
34
Metamorphic terrains We need to go there …. Mineral wealth, volumetrically significant
Folded gneiss. Teton Range, Wyoming. Courtesy of Marli Bryant Miller, Eugene, Oregon,
[email protected] Downloaded from http://marlimillerphoto.com/contact.html
35
Complex Geology Support modelling full geologic history T=151 T=0
T=23
S0
S1 T=345
S2 T=100
F3
S3 T=150
F2
Scott 2011 35
36
Implicit models – geologically reasonable ?
37
Implicit models – geologically reasonable ?
38
Implicit models – geologically reasonable ?
Yet more complexity !
t=3 t=2 t=1
Complex intrusive emplacement histories
Relative Ages
Dykes, sills, plutons… Kimberlites, synvolcanic intrusions, ore pulses Topologic & volume changes Occurres at all scales Before, during & after deformation ! Topology change
Noranda Camp
39
40
What makes a good geology map ?
What makes a good geology map ?
What makes a good geology map ?
What makes a good geology map ?
What makes a good geology map ?
45
What makes a good geology map ?
Familiar geological forms and patterns … Nice colours ! Has geological meaning Embedded geological knowledge Consistent geologic topology ! Geologic relationships are key
46
Regional Bedrock Map Simulation 2D maps and cross sections for 3D GEOMAPPING
47
MapSim Overview Geologic map simulation to support bedrock interpretation and compilation Produces geological features in ArcGIS from selected observations, previous mapping and geologic knowledge Outputs 2D maps and cross-sections from 3D input data and 3D computation engine GRBF
48
Geologic relationships
Folded gneiss. Teton Range, Wyoming. Courtesy of Marli Bryant Miller, Eugene, Oregon,
[email protected] Downloaded from http://marlimillerphoto.com/contact.html
49
Topological encoding of spatial relationships to support geological Modelling in a 3-D GIS environment Ernst Schetselaar and Eric de Kemp
261 ! 28 25 22
27 24 21
26 23 20
Egenhofer, M. and Franzosa, R., 1991. Point-set topological relations. International Journal of Geographical Information Systems 5 (2): 161-174. Egenhofer, M., Sharma J. and Mark D., 1993. A critical comparison of the 4-intersection and 9intersection models for spatial relations: formal analysis. In R. McMaster and M. Armstrong (eds), Autocarto 11, Minneapolis, MN, pp. 1-11. Zlatanova, S., 2000, On 3D topological relationships. In Proceedings of the 11th International Workshop on Database and Expert System Applications (DEXA 2000), 6-8 September, Greenwich, London, UK, pp. 913-919.
261 !
50
Encoding of Geologic History “geologic map-model an extension of relationships”
Volumetric Relation
BURNS, K.L., 1975, ANALYSIS OF GEOLOGICAL EVENTS, MATHEMATICAL GEOLOGY, VOL. 7, NO.4, P. 295 - 320
51
Geological Evolution Schema (GES)
Michel Perrin, Mathieu Poudret, Nicolas Guiard, Sebastien Schneider, 2013, Chapter 6: Geological Surface Assemblage, In: Shared Earth Modeling, Knowledge drivern solutions for building and managing subsurface 3D geological models, p. 115-139
Map / Model = Spatial extension of relationships Depositional top S0 ~1m
S2
53
Mapping relationships
+
Event SpaceTransform
i.e. UVT, N1F2, F2I1, I1F3
Estimation-Simulation
=
54
Legend – Encoding of Geologic History
55
Map – Extended set of relationships
56
Map – Extended set of relationships
57
Map – Encoding of Geologic History
58
MapSim 3D Voronoi
Binary Points
+
Simulation Stage
=
59
60
Agent Surface Development Agent interogator: sniffers (E1,E2,E3 = plunge, facing, high frequency direction) Agent constructor: primitives (points, normals, curves, surfaces, tetrahedra) Can be used to do estimation, interpolation of geologic features Has potential but needs to be embedded with geologic rules
61
Spatial Agents (SABM)
62
Conclusions Demand Geologically Reasonable models ! Use all the DATA – structural data proxy for spatial continuity estimates (local plunge vectors…. F1,F2,L01,L02, S1,S2,S3….. ) Make more then a single model supporting range of alterative scenarios (ie. changing geologic history) Develop a geologic event history algebra for interpretive-constrained environments (knowledge & Data) Be open to exploring all available approaches from other domains (protein folding, urban design, animation, mechanics).
62
63
Many thanks ! Coordination is needed… resources limited !
Success is possible !
LHC June 2015 First high energy collisions
64
Extra Material … Yet more details …
65
Implicit surface modelling, including bedding strike/dip constraints
Hillier, Schetselaar, de Kemp 2013
66
Implicit modelling taking into account structural orientation data
Co-kriging potential field scalar function contact points
structural data
[Calgacno et al., 2008, geomodeller]
67
Finding the weights: 1 contact & 1 gradient
Hillier, Schetselaar, de Kemp 2013
68
Igneous and Ore Systems
? Ore Body
Noranda, Central Camp Intrusive breccias of the ca. 3.45 Ga Theespruit pluton into amphibolites from the Lower Onverwacht Group of the Barberton Greenstone Belt (Kisters and Anhaeusser 1995). Photograph (PhotoID260) courtesy of JF Moyen.
69
Purcell 3D GRBF
“3D Modelling beyond the head frame” Observation based 3D model
Purcell – 3D
Math – 3D Geometry Engine 2014 Hillier, Schetselaar , de Kemp and Perron MG
Structural knowledge 50 km
© Haakon Fossen 2010
GOCAD / SKUA
70
Feasibility GRBF
3D Observational Data Store
Math – 3D Geometry Engine 2014 Hillier, Schetselaar , de Kemp and Perron MG
Geo Knowledge
Mobile Mapping
3D Models
71
Agent Based Modelling (MAB)
Natural Emergent Features
Agent Based Modelling • What are Agents? – – – –
Entities within model space that probe their environment Communicates with other agents Used to determine an emergent phenomena Makes decisions based on information gathered from its environment – Given rule sets to follow Example Rule Set • • • • • •
Make a path of medium to strong curvature Stop if you come within 500 meters of a fault Don’t cross your own field lines Stay in zones of greater then 2.7 density Stop when local variance is too high Stop at 5 km from data
Des Bourdon Heureux dans le ruche …
73
Agent Based Modelling (ABM) Future Work
(just ideas !)
Consolidate agent tool box – interpolation, interogation, multi-scale metrics and communication library Extend to horizon and fault prototyping (local RBF) Add more types of data (geophysical, geochemical). “ Making decisions used to be easy “ – Agent 007
Incorporate knowledge database into Agent’s memory. e.g. ArcGIS (Relationships, Fold Shapes, Crustal Level …)
Agent adaptive learning techniques (Learn in high data regions, dynamic spatial characterization i.e. moving variograms)
74
Simulation of Complex Systems Examples – 3D Protein folding http://www.youtube.com/watch?v=1eSwDKZQpok&feature=related
Agents vs Classic Approach Classic • Strong spatial dependency of occurrences to give result (sparse
Agent Based • Explores model space and can make many or no solutions (sparse patterning)
global mean)
• Resolution of results more important then resolution of the data • Minimal learning from data • Global algorithms applied to most of the data
• Natural to multi-scaler complex systems, can preserve all data • Rich variety of Rules, Missions (Beliefs) and behaviors can be applied while interacting with data • Many algorithms including classic ones can be applied (Kriging, DSI, RBF, IDW etc. )
• Nested rule-set difficult to achieve • Harder to parallelize – Needs large matrix management • Less interpretive (more objective)
• Produces expressions that EMERGE from simple agents interactioncommunication (i.e. hexagons, honey) • More but smaller matrix which benefits from faster CPU • More interpretive (less objective)
Emergent 2D Boundary
Realization 01
Blue curve is the emergent pattern (i.e. the interface)
Agent Based Modelling (ABM) Structural Vector Field - Preliminary results…
Polydeformed calc-silicate turbidites Ugab Basin Namibia
Field data – S2a,S2b,S3, L12 …
78
Agent Based Modelling (ABM) Structural Vector Field - Preliminary results… Regional forland fold belt – Labrador, Quebec
79
Morphimetrics of folds Parameterization of 3D fold surfaces
Lisle, R. and Toimil, N. , Geology, Vol. 35, No. 6, June. 2007, pp. 519-522