Modelling electromagnetic responses from seismic dataModelling electromagnetic responses from...

1
Modelling electromagnetic responses from seismic data Dieter Werthmüller [[email protected]], Anton Ziolkowski, and David Wright Introduction Good estimates of background resistivit- ies are often crucial in controlled-source elec- tromagnetic (CSEM) feasibility studies and in- versions. Seismic data and well logs are of- ten available prior to CSEM acquisition, but elastic waves and electromagnetic waves share no physical parameter. Contributions We present a methodology to estimate res- istivities from seismic velocities. We apply known methods, including rock physics, depth trends, structural in- formation, and uncertainty analysis. We show an example of the methodology with data from the North Sea Harding field. Rock Physics We use a Gassmann-based relation (f G ) for the transformation from P-wave velocity v to porosity φ, and the self-similar model (f s ) for the transformation from porosity φ to resistiv- ity ρ (e.g. Carcione et al., 2007): ρ = f s (ρ s f , m) , where φ = f G (K s ,K f ,G s ,% s ,% f , κ, v ) , m is the cementation exponent, K and G are bulk and shear moduli, % is density, κ is the Krief exponent, and subscripts s and f stand for solid and fluid fraction (see Fig. 5). EM-Line b-7 a-3 b-11 b-A01 b-8 6570000 6575000 412500 417500 4 0.0 0.1 0.2 0.3 0.4 Porosity (-) 1.5 2.5 3.5 4.5 Velocity (km/s) Gassmann 10 -1 10 0 10 1 Resistivity (Ωm) self-similar 5 The transform is done in three steps: 1) Calibrate transform (incl. depth trend, box below) with a well log nearby (b-8). 2) Apply to seismic velocity in area of in- terest (including uncertainty, box left). 3) Check transform with well log in area of interest (if available). 1 10 1 2 3 Depth (km) b-8 1 10 b-7 1 10 Resistivity (Ωm) b-11 1 10 b-A01 1 10 a-3 ρ s mode ±2σ 6 Start EM-Line End 1 2 3 Depth (km) Grid Sandstone Seabed Balder Formation Base Cretaceous Background resistivity model [Mode (Ωm)] 0.4 0.7 1.3 2.3 4.1 7 1D Modelling 1200 1300 1400 1500 CMP Position 0 0.2 0.4 0.6 Depth (km) Start model ρ m (Ωm) 1 2 3 4 5 9a 1200 1300 1400 1500 CMP Position 0 0.2 0.4 0.6 Depth (km) Final model ρ m (Ωm) 9b This resistivity model (box left) has two weak- nesses: 1) anisotropy (λ = p ρ v h , ρ m = ρ v ρ h ), 2) resistivities outside well control. CSEM impulse (IR) and step (SR) responses have different sensitivities to anisotropy (Fig. 10). Only if the anisotropy factor is cor- rect, inversion of IR and SR yield the same res- ult (Fig. 11). Short offset 1D inversions of measured CSEM data, with correct aniso- tropy factor, improve the background resistivity model in the shallow section, were we have no well control (Fig. 9); the resulting resistivities are in this case lower. 0.0 0.5 1.0 1.5 2.0 2.5 Time (s) 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 Amplitude (Ω/(m 2 s)) ×10 -10 ρ m = 1.0 ρ m = 2.0 ρ m = 3.0 λ = 1.0 λ = 1.5 λ = 2.0 10a 0.0 0.5 1.0 1.5 2.0 2.5 Time (s) 0 1 2 3 4 5 6 7 Amplitude (Ω/m 2 ) ×10 -11 10b 1.0 1.5 2.0 2.5 Anisotropy (-) 0 2 4 6 8 10 NRMSD (%) NRMSD IR NRMSD SR Model (SR - IR) 0 2 4 6 8 10 Δ Model IR-SR (Ωm) 11 Uncertainty Analysis 0.4 0.6 0.8 1.0 2 6 10 14 Prob. density (-) Parameter 0.4 0.6 0.8 1.0 Resistivity (Ωm) Model 0.4 0.6 0.8 1.0 2 6 10 14 Prob. density (-) Parameter & Model 1a 1b 1c 2 Data, rock physics model, and rock phys- ics parameter have errors. Chen and Dick- ens (2009) describe a methodology to account for the uncertainties related to rock physics parameters and the rock physics model it- self. They describe the rock physics model as gamma distribution in a Bayesian frame- work, with a defined error E , and the rock physics parameters as distributions, f (ρ|θ )= β α x α-1 Γ(α) exp (-βx) , where θ is a vector containing all model para- meter distributions, α =1/E 2 , and β =(α - 1)rp . Here, ρ rp is one realization of the rock physics model with a random set of model para- meters. We define the distribution of the velo- city from the data themselves (see Fig. 2). The distribution is defined as the difference between the measured log values and the val- ues of the smoothed log, v (z ) - v s (z ). This yields resistivity as a probability density function, instead of a deterministic resistivity value (Fig. 3c). 1 10 Resistivity (Ωm) 0.6 1.0 1.4 1.8 Depth (km) Grid Sandstone ρ s mode ±σ ±2σ 3a v f 2.0 3.0 4.0 v s Velocity (km/s) ρ f 1 10 ρ s Resistivity (Ωm) deterministic 0.6 0.8 1.0 1.2 Resistivity (Ωm) 1 3 5 Prob. density (-) 3b 3c Depth Trend Rock parameters are a function of, e.g., litho- logy and depth. We include the following dependences in our model: Depth: K s ,G s , κ, ρ s Temperature: ρ f Porosity : m Lithology: Grid Sandstones (delineated with seismic horizons) 1 10 Resistivity (Ωm) ρ ρ s ρ(φ[v ]) 2 3 4 Vel. (km/s) 0.6 1.0 1.4 1.8 Depth (km) v v s 10 30 (GPa) K s G s 1 10 (GPa) ρ f ρ s 2 3 (-) m κ Grid Sandstone 8 Acknowledgment We thank PGS for funding the research and the Harding partners, BP and Maersk, for per- mission to use the data. References Carcione, J. M., B. Ursin, and J. I. Nordskag, 2007, Cross-property relations between electrical conductivity and the seismic velocity of rocks: Geophysics, 72, E193–E204, doi: 10.1190/1.2762224. Chen, J., and T. A. Dickens, 2009, Effects of uncertainty in rock-physics models on reservoir parameter estimation using seismic amplitude variation with angle and controlled-source electromagnetics data: Geophysical Prospecting, 57, 61–74, doi: 10.1111/j.1365-2478.2008.00721.x. Werthmüller, D., A. Ziolkowski, and D. Wright, 2012, Background resistivity model from seismic velocities: SEG Technical Program Expanded Abstracts, 31, doi: 10.1190/segam2012-0696.1. Conclusions This method yields the range of back- ground resistivity models, consistent with the known seismic velocities. This model provides an additional data set, which can be used for integrated ana- lysis, or as a starting point for a detailed CSEM feasibility study or inversion. We will use this background resistivity model for 3D CSEM modelling for comparison with measured data.

Transcript of Modelling electromagnetic responses from seismic dataModelling electromagnetic responses from...

Page 1: Modelling electromagnetic responses from seismic dataModelling electromagnetic responses from seismic data Dieter Werthmüller [Dieter.Werthmuller@ed.ac.uk], ... 2 :5 3 :5 4 :5 Velocity

Modelling electromagnetic responses from seismic dataDieter Werthmüller [[email protected]], Anton Ziolkowski, and David Wright

IntroductionGood estimates of background resistivit-ies are often crucial in controlled-source elec-tromagnetic (CSEM) feasibility studies and in-versions. Seismic data and well logs are of-ten available prior to CSEM acquisition, butelastic waves and electromagnetic wavesshare no physical parameter.

Contributions• We present a methodology to estimate res-istivities from seismic velocities.

• We apply known methods, including rockphysics, depth trends, structural in-formation, and uncertainty analysis.

• We show an example of the methodologywith data from the North Sea Harding field.

Rock PhysicsWe use a Gassmann-based relation (fG) forthe transformation from P-wave velocity v toporosity φ, and the self-similar model (fs) forthe transformation from porosity φ to resistiv-ity ρ (e.g. Carcione et al., 2007):

ρ = fs(ρs, ρf,m, φ) , where

φ = fG(Ks,Kf, Gs, %s, %f, κ, v) ,

m is the cementation exponent, K and G arebulk and shear moduli, % is density, κ is theKrief exponent, and subscripts s and f standfor solid and fluid fraction (see Fig. 5).

EM-Line

b-7a-3

b-11

b-A01

b-86570000

6575000

412500

417500

4

0.0 0.1 0.2 0.3 0.4

Porosity (-)

1.5

2.5

3.5

4.5

Vel

oci

ty(k

m/s)

Gassmann

10−1

100

101

Res

isti

vit

y(Ω

m)

self-similar

5

The transform is done in three steps:1) Calibrate transform (incl. depth trend,box below) with a well log nearby (b-8).2) Apply to seismic velocity in area of in-terest (including uncertainty, box left).3) Check transform with well log in area ofinterest (if available).

1 10

1

2

3

Dep

th(k

m)

b-8

1 10

b-7

1 10

Resistivity (Ωm)

b-11

1 10

b-A01

1 10

a-3

ρs

mode

±2σ

6

Start – EM-Line – End

1

2

3

Dep

th(k

m)

Grid Sandstone

Seabed

Balder Formation

Base Cretaceous

Background resistivity model [Mode (Ωm)]

0.4

0.7

1.3

2.3

4.17

1D Modelling

1200 1300 1400 1500CMP Position

0

0.2

0.4

0.6Dep

th(k

m)

Start model ρm (Ωm)

1

2

3

4

59a

1200 1300 1400 1500CMP Position

0

0.2

0.4

0.6 Dep

th(k

m)

Final model ρm (Ωm)9b

This resistivity model (box left) has two weak-nesses:1) anisotropy (λ =

√ρv/ρh, ρm =

√ρvρh),

2) resistivities outside well control.CSEM impulse (IR) and step (SR) responseshave different sensitivities to anisotropy(Fig. 10). Only if the anisotropy factor is cor-

rect, inversion of IR and SR yield the same res-ult (Fig. 11). Short offset 1D inversions ofmeasured CSEM data, with correct aniso-tropy factor, improve the background resistivitymodel in the shallow section, were we have nowell control (Fig. 9); the resulting resistivitiesare in this case lower.

0.0 0.5 1.0 1.5 2.0 2.5

Time (s)

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Am

pli

tud

e(Ω/(m

2s)

)

×10−10

ρm = 1.0

ρm = 2.0

ρm = 3.0

λ = 1.0

λ = 1.5

λ = 2.0

10a

0.0 0.5 1.0 1.5 2.0 2.5

Time (s)

0

1

2

3

4

5

6

7

Am

plitu

de

(Ω/m

2)

×10−11

10b

1.0 1.5 2.0 2.5

Anisotropy (−)

0

2

4

6

8

10

NR

MSD

(%)

NRMSD IR

NRMSD SR

Model (SR - IR)

0

2

4

6

8

10

∆M

odel

IR-S

R(Ω

m)

11

Uncertainty Analysis

0.4 0.6 0.8 1.0

2

6

10

14

Pro

b.

den

sity

(-) Parameter

0.4 0.6 0.8 1.0

Resistivity (Ωm)

Model

0.4 0.6 0.8 1.0

2

6

10

14

Pro

b.

den

sity

(-)Parameter & Model

1a 1b 1c

-1 0 1∆v : v(z)−vs (z)

pdf

data

0

1

2

3

4

5

Pro

b.

den

sity

(-)

2 3 4Vel. (km/s)

0.6

1.0

1.4

1.8D

ep

th (

km

)

v

vs

2

Data, rock physics model, and rock phys-ics parameter have errors. Chen and Dick-ens (2009) describe a methodology to accountfor the uncertainties related to rock physicsparameters and the rock physics model it-self. They describe the rock physics model asgamma distribution in a Bayesian frame-work, with a defined error E, and the rockphysics parameters as distributions,

f(ρ|θ) =βαxα−1

Γ(α)exp (−βx) ,

where θ is a vector containing all model para-meter distributions, α = 1/E2, and β = (α −1)/ρrp. Here, ρrp is one realization of the rockphysics model with a random set of model para-meters.

We define the distribution of the velo-city from the data themselves (see Fig. 2).The distribution is defined as the differencebetween the measured log values and the val-ues of the smoothed log, v(z) − vs(z). Thisyields resistivity as a probability densityfunction, instead of a deterministic resistivityvalue (Fig. 3c).

1 10

Resistivity (Ωm)

0.6

1.0

1.4

1.8

Dep

th(k

m)

Grid Sandstone

ρs

mode

±σ±2σ3a

vf 2.0 3.0 4.0 vs

Velocity (km/s)

ρf

1

10

ρs

Res

isti

vit

y(Ω

m) deterministic

0.6 0.8 1.0 1.2

Resistivity (Ωm)

1

3

5

Pro

b.

den

sity

(-)

3b

3c

Depth TrendRock parameters are a function of, e.g., litho-logy and depth. We include the followingdependences in our model:• Depth: Ks, Gs, κ, ρs

• Temperature: ρf• Porosity: m• Lithology: Grid Sandstones(delineated with seismic horizons)

1 10

Resistivity (Ωm)

ρ

ρs

ρ(φ[v])

2 3 4

Vel. (km/s)

0.6

1.0

1.4

1.8

Dep

th(k

m)

v

vs

10 30

(GPa)

Ks

Gs

1 10

(GPa)

ρf

ρs

2 3

(-)

m

κ

Grid Sandstone

8

AcknowledgmentWe thank PGS for funding the research andthe Harding partners, BP and Maersk, for per-mission to use the data.

ReferencesCarcione, J. M., B. Ursin, and J. I. Nordskag, 2007, Cross-property

relations between electrical conductivity and the seismic velocityof rocks: Geophysics, 72, E193–E204, doi: 10.1190/1.2762224.

Chen, J., and T. A. Dickens, 2009, Effects of uncertainty inrock-physics models on reservoir parameter estimation usingseismic amplitude variation with angle and controlled-sourceelectromagnetics data: Geophysical Prospecting, 57, 61–74,doi: 10.1111/j.1365-2478.2008.00721.x.

Werthmüller, D., A. Ziolkowski, and D. Wright, 2012, Backgroundresistivity model from seismic velocities: SEG Technical ProgramExpanded Abstracts, 31, doi: 10.1190/segam2012-0696.1.

Conclusions• This method yields the range of back-ground resistivity models, consistentwith the known seismic velocities.• This model provides an additional dataset, which can be used for integrated ana-lysis, or as a starting point for a detailedCSEM feasibility study or inversion.• We will use this background resistivity model

for 3D CSEM modelling for comparison withmeasured data.