Constraining gravity and magnetic Constraining gravity and magnetic inversions for mineral explorationinversions for mineral exploration

Nick WilliamsNick Williams1&21&2, Doug Oldenburg, Doug Oldenburg11 & Peter Leli& Peter Lelièèvrevre11

11 University of British Columbia University of British Columbia –– Geophysical Inversion FacilityGeophysical Inversion Facility2 2 Geoscience AustraliaGeoscience Australia

Sponsors:Sponsors:

Both density models explain the gravity data equally well.Both density models explain the gravity data equally well.

Which is the most reliable prediction of the subsurface?Which is the most reliable prediction of the subsurface?

Inversion and constraintsInversion and constraints•• InversionInversion

–– Seeks a 3D physical property model that explains observed Seeks a 3D physical property model that explains observed geophysical datageophysical data

–– Calculation is nonCalculation is non--unique: many solutions are possibleunique: many solutions are possible

–– Finding a solution is easy. Finding a Finding a solution is easy. Finding a realisticrealistic solution is hardersolution is harder

•• ConstraintsConstraints–– Impose preferences or limitations on inversionImpose preferences or limitations on inversion

–– Ensure recovered model is consistent with geology Ensure recovered model is consistent with geology andand explains explains the geophysicsthe geophysics

–– Types:Types:•• Geological observations Geological observations •• Geological interpretationsGeological interpretations•• Mathematical constructs that approximate geologyMathematical constructs that approximate geology

UBCUBC––GIF inversion approachGIF inversion approach

•• Numerical optimisation problemNumerical optimisation problem

•• Seeks model with Seeks model with ““smallsmall”” & & ““smoothsmooth”” characteristicscharacteristics

•• SmallnessSmallness–– Deviations from a supplied reference model are smallDeviations from a supplied reference model are small

–– Add only as much detail as needed to fit the dataAdd only as much detail as needed to fit the data

•• SmoothnessSmoothness–– 1) Deviations from reference model are distributed smoothly, or1) Deviations from reference model are distributed smoothly, or

–– 2) Model itself is smooth2) Model itself is smooth

•• Constraints implemented using 4 main parametersConstraints implemented using 4 main parameters

UBCUBC––GIF constraint implementationGIF constraint implementation

•• 1) Reference model1) Reference model–– Model of expected physical properties based on existing Model of expected physical properties based on existing

geological knowledgegeological knowledge

UBCUBC––GIF constraint implementationGIF constraint implementation

•• 2) Smallness weights (2) Smallness weights (≥≥ 1)1)–– Reliability of the supplied reference modelReliability of the supplied reference model

UBCUBC––GIF constraint implementationGIF constraint implementation

•• 3) Property bounds, lower & upper3) Property bounds, lower & upper–– Range of properties allowed in each cellRange of properties allowed in each cell

UBCUBC––GIF constraint implementationGIF constraint implementation

•• 4) Smoothness weights (sharp < 1; smooth > 1)4) Smoothness weights (sharp < 1; smooth > 1)–– How smoothly or sharply properties change across each cell How smoothly or sharply properties change across each cell

face, in each direction (north, east, vertical)face, in each direction (north, east, vertical)

–– Based on geological structures/trendsBased on geological structures/trends

Recovering Desired GeometriesRecovering Desired Geometries•• A simple constraint is the overall geometry of the modelA simple constraint is the overall geometry of the model

•• Assigned using length scales in each direction:Assigned using length scales in each direction:

(default)

Aspect ratio 1:1 Aspect ratio 5:1 Aspect ratio 1:3

Lx = Ax · 2 · ΔxLy = Ay · 2 · ΔyLz = Az · 2 · Δz

Length scaleAspect ratio

Cell width

Recovering Desired GeometriesRecovering Desired Geometries

Inversion goalsInversion goals•• What do you want from the inversion modelling?What do you want from the inversion modelling?

•• Obtain a final 3D synthesis of all geological/geophysical data Obtain a final 3D synthesis of all geological/geophysical data and interpretations?and interpretations?–– Mature geological understandingMature geological understanding

–– Highly developed geological interpretations/modelsHighly developed geological interpretations/models

–– Option 1: Inversion as end productOption 1: Inversion as end product

•• Obtain an initial or updated 3D base map for future Obtain an initial or updated 3D base map for future interpretation and targeting?interpretation and targeting?–– Limited geological understandingLimited geological understanding

–– Minimal geological data/modelsMinimal geological data/models

–– Option 2: Inversion as base mapOption 2: Inversion as base map

Option 1Option 1: Inversion as an end product: Inversion as an end product•• Create full 3D interpretation of geological surfacesCreate full 3D interpretation of geological surfaces

–– Time, money, and data intensive, and difficult to updateTime, money, and data intensive, and difficult to update

•• Model is one interpretation Model is one interpretation inversion tests one hypothesisinversion tests one hypothesis

•• Identifies where interpretations are inconsistent with the dataIdentifies where interpretations are inconsistent with the data

•• Requires Requires smooth modelsmooth model--differencedifference inversion: sharp boundaries inversion: sharp boundaries along contacts where model changesalong contacts where model changes

Referencemodel

Constrainedinversionresult

Joutel VHMS example sources: Fallara et al. (2003),Marquis et al. (2003), Oldenburg and Pratt (2007)

Option 2Option 2: Inversion as a base map: Inversion as a base map•• Supply raw geological observations directly as constraintsSupply raw geological observations directly as constraints

–– Minimal interpretationMinimal interpretation

•• Quick and relatively easy results, consistent with geologyQuick and relatively easy results, consistent with geology

•• May predict new features/targets May predict new features/targets –– best for explorationbest for exploration

•• Requires Requires smooth modelsmooth model inversion: smoothly extrapolate inversion: smoothly extrapolate properties away from observationsproperties away from observations

Types of inversion smoothnessTypes of inversion smoothness

Option 2:Limited

observations

Option 1:Full 3D model

workaround

Synthetic example, basic constraintsSynthetic example, basic constraints

Map + 2 drill holes

But we know more than thatBut we know more than that•• Extrapolate observations along strike to capture expected Extrapolate observations along strike to capture expected

continuity of geologycontinuity of geology

•• Based on observed or inferred structural orientationsBased on observed or inferred structural orientations

Reference propertyReference property–– no buffersno buffers

Reference propertyReference property–– with bufferswith buffers

Synthetic example, extrapolated constraintsSynthetic example, extrapolated constraints

Map + 2 drill holes + buffers

Synthetic example, basic constraintsSynthetic example, basic constraints

Map + 2 drill holes

A real exampleA real example•• Perseverance nickel sulphide deposit, Western AustraliaPerseverance nickel sulphide deposit, Western Australia

–– Archean graniteArchean granite--greenstone terranegreenstone terrane

–– Established mine seeking extensions to ore reservesEstablished mine seeking extensions to ore reserves

•• Available geological data:Available geological data:–– Outcrop geology map ESRI ArcView shapefileOutcrop geology map ESRI ArcView shapefile

–– Basement geology shapefile interpretationBasement geology shapefile interpretation

–– DistrictDistrict--wide drilling database:wide drilling database:•• Density measurementsDensity measurements•• Geology observationsGeology observations

–– Density measurements on 30 surface rocksDensity measurements on 30 surface rocks

–– Partial 3D geological model created from drilling and potential Partial 3D geological model created from drilling and potential field interpretationfield interpretation

Default inversion resultDefault inversion result

•• Explains the gravity data Explains the gravity data

•• Smooth, minimal structureSmooth, minimal structure

•• DoesnDoesn’’t match known geologyt match known geology

Gravity dataGravity data Recovered densitiesRecovered densities

500 m1 km

+ thickness of regolith (~70-100 m)+ denser basement rocks

• Reproduces variable-thickness regolith

• Enhances density contrasts• Refines targets• Still fits the gravity data

Add surface mappingAdd surface mapping

Geological constraints

Default result

Constrained result

500 m

500 m500 m

+ drilling+ basement map interpretation+ partial 3D geology model

• Reproduces variable-thickness regolith

• Still fits the gravity data

• Geologically realistic

Add all geological Add all geological knowledgeknowledge

Geological constraints

Default result

Constrained result

500 m

500 m500 m

SummarySummary•• Geological constraints ensure inversion is consistent with Geological constraints ensure inversion is consistent with

geophysics and geologygeophysics and geology–– Mandatory for most problemsMandatory for most problems

•• Geological knowledge can be translated directly into the Geological knowledge can be translated directly into the appropriate constraint parametersappropriate constraint parameters

•• Must clearly define the problem to be solvedMust clearly define the problem to be solved

•• Inversion as an end productInversion as an end product–– Substantial geological interpretation prior to inversionSubstantial geological interpretation prior to inversion

–– Provides a final synthesis of all informationProvides a final synthesis of all information

•• Inversion as a base mapInversion as a base map–– Use raw geological constraints to derive base map to plan data Use raw geological constraints to derive base map to plan data

acquisition, targeting, and interpretationacquisition, targeting, and interpretation

–– Continually refined throughout exploration programContinually refined throughout exploration program

AcknowledgmentsAcknowledgments•• Project sponsorsProject sponsors

–– BHP Billiton for data and permission to publishBHP Billiton for data and permission to publish

•• Ph.D. supervisorsPh.D. supervisors–– Doug Oldenburg & Dick TosdalDoug Oldenburg & Dick Tosdal

•• ColleaguesColleagues–– MDRUMDRU--GIF inversion project:GIF inversion project:

•• Dianne Mitchinson, Nigel PhillipsDianne Mitchinson, Nigel Phillips

–– GIF:GIF:•• Roman ShekhtmanRoman Shekhtman

–– Geoscience Australia:Geoscience Australia:•• Richard LaneRichard Lane