Finding Goldilocks in a 3D world –how many models do we need to run?

Kate RobertsonStephan Thiel, Graham Heinson

@kate_rob31

Outline

• 2D and 3D inversion codes and the 2D to 3D transition

• Importance of understanding your data and your model

• Modelling tips

• Finding the best model

• Survey layout

What are we really modelling?

• Non-unique, under-parameterised problem

• How do we find the best model?

• Needs to be useful to the interpreter

‘All models are wrong, but some are useful’

George E.P. Box

Modelling codes – what are we working with?1

2D

• Occam (De-Groot-Hedlin & Constable, 1990)

• REBOCC (Siripunvaraporn & Egbert, 1999)

• Mare2DEM (Key, 2016)

• WinGLink (Rodi & Mackie, 2001)

• Geotools (CGG)

3D

• ModEM3DMT (Egbert & Kelbert 2012)

• WSINV3DMT (Siripunvaraporn & Egbert, 2009)***

• Mare3DEM (K. Key, in prep)

• Geotools (CGG)

• x3Di (Avdeev & Avdeeva 2009)1A far from exhaustive list, just what popped into my head

Open source codes

*** serial version is free, not parallel

Anisotropic codes

Invert for distortion

The transition from 2D to 3D

The transition from 2D to 3D

• Profile of MT data where bright colours show 3D data (|skew| > 5)

• Technically shouldn’t be modelled using a 2D code)

• Can model profiles in 3D now, or collect in arrays3D 3D

1D or 2D

6

Why test your models?

Exploring the search space (in a deterministic inversion scheme)

Bedrosian, P.A., et al (2018) Nature Geoscience, Janelle Simpson, pers. Comm 2018

Interpreting the models

As we get deeper the values

vary more and head toward

starting value

100

200

300

400

0

De

pth

(km

)

104

103

102

10

1

Resis

tivity

(Ωm

)

Interpreting the models

Water content (wt%)~0.001

wt%~5 wt%

Two models that both fit the data

can provide different water content

estimates:

10 Ωm, RMS 1.17, 5wt%

1000 Ωm, RMS 1.57, 0.001 wt%

10 Ωm starting

1000 Ωm starting

How do we know which is the best?

NOT from the global RMS…

11Bedrosian, P.A., et al (2018) Nature Geoscience,

How do we know which is the best?

• Global RMS NOT enough!

Mount St Helens example, Bedrosian et al 2018, supp. information12

Time considerationsSave time while modelling..

But ultimately we need faster 3D inversion software

13

• Not really in the scope of the talk, but can help to avoid many problems in modelling

• Orientation of the profile or array

• Site spacing and layout

• Depth penetration

• Survey aperture

• And more..

Survey Layout

Miensopaust 2018

Summary

• Can be difficult to determine the best model but need

to understand the range of models that can fit the data

• Understanding your code can save you expensive

computational time

• Much more to consider that has not been covered

here- distortion, anisotropy, error floors, rotating your

data etc…