Patterns in Geophysical Data and Models

Jens Tronicke

Angewandte GeophysikInstitut für Geowissenschaften

Universität Potsdam

jens@geo.uni-potsdam.de

Near-surface geophysics

• Using geophysical tools to explore the shallow subsurface

• Scale of interest: Tens of centimeters up to tens of meters

• Common techniques:- Magnetics- Seismics (reflection and refraction)- Electrical resistivity and EM induction methods- Georadar (GPR)- Borehole and logging techniques

• Typical applications are from the fields of…

- Geology - Engineering geology and geotechnics- Sedimentology - Hydrogeology- Soil sciences - Environmental sciences- Archaeology - Agriculture- Civil engineering - …

⇒ Research topics are directed towards developing and improving geophysical techniques for specific applications

Patterns in Geophysical Data & Models

Data gathering(employing magnetics, seismics, electrics,

georadar, borehole techniques, …)

Processing(editing, analyzing, filtering, inversion, …In part specially developed techniques)

General flow of a geophysical experiment

Analyzing & enhancing specific features in the data space

Examples from seismic & georadar:

• Residual statics (cross-correlation)• Velocity analysis (coherency

analysis of CMP gathers)• NMO stacking • Coherency & dip filtering

• What is a pattern? - “Opposite of chaos” (Watanabe, 1985)- “A regular repetition of values in an image” (Sheriff, 2002)

Patterns in Geophysical Data & Models

Data gathering(employing magnetics, seismics, electrics,

georadar, borehole techniques, …)

Processing(editing, analyzing, filtering, inversion, …In part specially developed techniques)

General flow of a geophysical experiment

Interpretation(generating earth models based on one

or several geophysical techniques)

Analyzing, extracting patterns & features in the model space, e.g.:

• Visualizing specific structures (faults, sedimentological units, archaeological features, …)

• Integrating all available models & data (including non-geophysical information, e.g., core data, geological background, geo-technical or hydrological data, …)

• What is a pattern? - “Opposite of chaos” (Watanabe, 1985)- “A regular repetition of values in an image” (Sheriff, 2002)

Overview

Selected examples from geophysical data interpretation

• 3-D georadar data:

Attribute analysis to extract fault related features

• Multi-technique data sets:

Integrated earth models based on cluster analysis

Overview

Selected examples from geophysical data interpretation

• 3-D georadar data:

Attribute analysis to extract fault related features

• Multi-technique data sets:

Integrated earth models based on cluster analysis

(Coates, 2002)

Maleme Fault Zone, New Zealand

0

3

6

9

-30 -20 -10 0 10 20 30

y [m]

300

200

100

0

Tim

e [n

s]

Depth [m

]

NW SE

Central cut through migrated 3-D volume (x = 14 m)

• Goal of the study: Mapping shallow geometry of the fault zone and stratigraphic details within near surface sediments using 2-D and 3-D georadar

• 3-D data set (100 MHz antennas) covering an area of ~ 20 x 70 m

Attribute analysis (3-D data)

Migrated data cube

Instantaneous phase:

Local gradients: dx and dy

Slope:

φ

cos ( )φ

22yx ddS +=

Emphasizing continuity/discontinuity

Emphasizing local changes (including information on directions)

• Post-processing sequence to visualize near vertical features

Maleme Fault Zone: 3-D data

0

3

6

9

-30 -20 -10 0 10 20 30

y [m]

300

200

100

0Ti

me

[ns]

Depth

[m]

NW SE

Central cut through 3-D volume (x = 14 m)

• Reflection A • Reflection B

0

3

6

9

-30 -20 -10 0 10 20 30

y [m]

300

200

100

0Ti

me

[ns]

Depth

[m]

NW SE

195 ns

Maleme Fault Zone: 3-D data

Central cut through 3-D volume (x = 14 m)

Maleme Fault Zone: 3-D Daten

195 ns (~5.85 m)

NW SE

Maleme Fault Zone: 3-D Daten

NW SE

195 ns (~5.85 m)

Maleme Fault Zone: 3-D Daten

NW SE

195 ns (~5.85 m)

Maleme Fault Zone: 3-D Daten

NW SE

195 ns (~5.85 m)

Maleme Fault Zone: 3-D Daten

NW SE

195 ns (~5.85 m)

Main discontinuities picked in 3-D slope cube

8.07.57.06.56.05.55.04.5

Depth [m]

Maleme Fault Zone: 3-D data

Overview

Selected examples from geophysical data interpretation

• 3-D georadar data:

Attribute analysis to extract fault related features

• Multi-technique data sets:

Integrated earth models based on cluster analysis

• Hydraulic relevant parameters (e.g., porosity, permeability)

• Internal structures

Groundwater table

Depth to bedrock

• Borders of the aquifer

Unsaturated sediments

Saturated sediments

Bedrock

Geophysical tools for aquifer characterization

For example: contaminant site characterization

Crosshole georadar and seismic tomography

• Exploring the inter-borehole plane- Traveltime of the signals ⇒ velocity v- Amplitudes of the signals ⇒ attenuation α

• Scale of such experiments: several meters to tens of meters

• Resolution: ~ 1 m

• Sensitive to hydrological relevant parameters: water content, porosity, …

⇒ Often ideal complement to the more conventional hydrological field techniques

n sources

m receivers

Synthetic example

How to generate an integrated model from different geophysical models ?

Fuzzy c-means (FCM) cluster analysis

Parameter 1

Para

met

er 2

FCM cluster analysis⇒ mij : membership of data

point xj to cluster i

n-dimensional model space

Parameter 1

Para

met

er 2

Fuzzy c-means (FCM) cluster analysis

z [m

]x [m]

FCM cluster analysis⇒ mij : membership of data

point xj to cluster i

n-dimensional model space

Defuzzification⇒ “hard“ cluster separation

(Colour saturation pro-portional to membership)

Zoned multi-parameter model including information

regarding its reliability

Fuzzy c-means (FCM) cluster analysis

FCM cluster analysis⇒ mij : membership of data

point xj to cluster i

n-dimensional model space

Defuzzification⇒ “hard“ cluster separation

(Colour saturation pro-portional to membership)

Zoned multi-parameter model including information

regarding its reliability

Multi-prameter model:

: mean of p for cluster i

Spatial model of the petrophysical target

parameter

ip

Target parameter p(e.g. porosity-logs)

∑=

=c

iiijj pmp

1

z [m

]

x [m]

Synthetic example

Results FCM cluster analysis

• Colour saturation reflects reliability of the zonation

Reconstructed porosity distributions

Three-cluster solution

Resampled input model

Results cluster analysis

Velocity Attenuation Clustered model

Real data example: Boise Hydrogeophysical Research Site (BHRS)

Membership functionfor cluster 1

(“low porosity cluster”)

Fuzzy c-means(after defuzzification)

Analyzing reliability of the zonation

Fuzzy c-means(after defuzzification)

Petrophysical parameter reconstruction

Reconstructed porosity distribution

Conclusions

• Pattern & feature recognition and related methods are fundamental steps in the geophysical workflow ⇒ Data processing⇒ Interpretation

• Examples illustrated the potential to …- extract specific features in a data set- integrate different data sets in a quantitative fashion

Acknowledgements

• Deutsche Forschungsgemeinschaft (DFG)

• Schweizerischer Nationalfonds (SNF)

• P. Villamor (GNS, New Zealand)

• W. Barrash & M. Knoll (Boise State University)

• K. Holliger & H. Paasche (ETH Zürich)