1. Martyn Unsworth and Volkan Tuncer Weerachai Siripunvaraporn
Exploration for unconformity uranium deposits with audiomagnetotellurics
Martyn Unsworth and Volkan Tuncer
University of Alberta, Canada
Weerachai Siripunvaraporn
Mahidol University, Bangkok, Thailand
Jim Craven
Natural Resources Canada, Ottawa, Canada

2. Outline 1. Introduction 2. AMT field techniques
3. MacArthur River AMT dataset – data processing
4. MacArthur River AMT dataset – model verification
5. Other studies
6. Conclusions

3. Introduction After Ruzicka
1000
Crystalline
rocks
100
(Wm)
Sedimentary
rocks 10 1
Brines
Graphite
After Ruzicka
Why use audiomagnetotellurics (AMT) for uranium exploration?
Graphitic conductors are strong targets. Can also resolve structures above the unconformity
Logistically simple – no TX loops, small receiver
Good depth of penetration
Plane wave signal allow full 3-D inversion with modest computation

4. Audiomagnetotellurics (AMT)
2. AMT field techniques
Audiomagnetotellurics (AMT)
f = signal frequency
Depth of penetration
d = 500 * sqrt (r/f)
Measure resistivity of Earth
= Zxy / 2pmf
Zxy = Ex / Hy 2

5. 1980 2000 2. AMT field techniques Phoenix Geophysics V5-2000
• 24-bit A-to-D
• Low induction coil noise
• GPS time synchronized
• large data storage capacities
• low power consumption
• lower cost
Phoenix Geophysics V5-2000
Metronix AMT system

6. TE-mode Current flow along strike Ex and Hy Rho data Phase data
True model
100 10 1
10000
0 km
5 km
TE-mode
Current flow along strike
Ex and Hy
Rho data
1000 Hz
0.001 Hz
1 Hz
Phase data
1000 Hz
0.001 Hz
1 Hz
Rho fit
1000 Hz
0.001 Hz
1 Hz
Phase fit
Inversion model
0 km
5 km Wm 10
100
1000
Non-linear conjugate gradient inversion
10 % noise in rho and phase

7. TM-mode Current flow across strike Ey and Hx Rho data Phase data
True model
100 10 1
10000
0 km
5 km
TM-mode
Current flow across strike
Ey and Hx
1000 Hz
0.001 Hz
1 Hz
Rho data
1000 Hz
0.001 Hz
1 Hz
Phase data
Rho fit
1000 Hz
0.001 Hz
1 Hz
Phase fit
Inversion model
0 km
5 km Wm 10
100
1000
Non-linear conjugate gradient inversion
10 % noise in rho and phase

8. TE-mode tipper Current flow along strike10
0 km
5 km
TE-mode tipper
Current flow along strike
generates a vertical magnetic field
100 1
10000
True model
1000 Hz
0.001 Hz
1 Hz
Real data
Quad data
Real fit
1000 Hz
0.001 Hz
1 Hz
Quad fit
Inversion model
0 km
5 km Wm
1000
100 10
0.2
0.0
-0.2
Tipper
Non-linear conjugate gradient inversion
0.01 noise in Tyz (tipper)
Cannot determine absolute resistivity
Good horizontal resolution

9. Combined inversions TE, TM and tipper (Tyz) 0 km 5 km True model 0 kmWm 10
100
1000
TE+TM+Tyz
Non-linear conjugate gradient inversion
(Rodi and Mackie, Geophysics, 2000)
10 % noise in rho and phase
0.01 in Tyz

10. 3. MacArthur River AMT dataset – data processing
EXTECH IV was a cooperation between the Canadian government, industry and universities
tested a range of geophysical and geological techniques above a known deposit
Full tensor AMT data and vertical magnetic field recorded at all sites

11. Electric fields distorted
3. MacArthur River AMT dataset – data processing
Dimensionality - tensor decomposition
Forward problem
Measured electric fields = regional electric fields + distortion
Undistorted
electric fields
Tensor decomposition
Regional electric fields = measured electric fields - distortion
assumes a 2-D regional structure with local 3-D distortion
assumes no EM induction occurs in the distorter
computes strike angle and distortion (twist and shear angles)
r.m.s. misfit gives a measure of how well the above
assumptions are satisfied at each MT station
static shift still unknown
Electric fields distorted
by shallow structure

12. 3. MacArthur River AMT dataset – data processing
Dimensionality - tensor decomposition
used multi-site, multi-frequency algorithm of Gary McNeice and Alan Jones
plot best fitting geoelectric strike direction as map and rose diagram
r.m.s. misfit shows if assumptions are valid (should be in range 0.5 – 1.5 )
inherent ambiguity of 90 degrees in strike direction

13. ? 3. MacArthur River AMT dataset – data processing
Dimensionality – induction vectors ?
Projection of the real component of the vertical magnetic field
In the Parkinson convention, these vectors point at conductors.
Direction reverses above the conductor (as in VLF)
More sensitive than apparent resistivity data to structures to the side of AMT station

14. 3. MacArthur River AMT dataset – data processing
Apparent resistivity and phase curves on Line 224
Above conductor
Away from conductor
data rotated to strike direction defined by tensor decomposition
frequency is a proxy for depth
AMT dead band has weak signals
224

15. 3. MacArthur River AMT dataset – data processing
Pseudosection displays – Line 224
1-D analysis not appropriate since major lateral changes
Note the sign reversal in the tipper (Tzy)
Need to convert frequency to true depth
224

16. 3. MacArthur River AMT dataset – data processing
2-D inversion – Line 224
Inverted with NLCG6 algorithm developed by Randy Mackie
Inverse MT problem is inherently non-unique
Overcome this issue by imposing extra conditions on solution
(e.g. smooth model, discontinuity at known location etc). Note
that smoothing broadens the basement conductor
Full imaging requires both modes and tipper

17. 3. MacArthur River AMT dataset – data processing
2-D inversion - fit to data
Error floor used to give uniform fit
Note consistent apparent resistivity and phase
224

18. 3. MacArthur River AMT dataset – data processing
3-D inversion
Line 304
Line 254
Line 224
0 km
1 km
2 km
Mackie 2D
TE+TM+Tzy
0 km
1 km
2 km
Mackie 3D
TE+TM+Tzy
0 km
1 km
2 km
Siripunvaraporn 3D
TE+TM
304
224
254
Inverse MT problem is inherently non-unique
3-D inversion much more computationally demanding than 2-D

19. 3. MacArthur River AMT dataset – data processing
3-D inversion
Mackie 2D
inversion
Mackie 2D
inversion
Siripunvaraporn 3D
inversion

20. 4. MacArthur River AMT dataset – model verification
Comparison with well logs

21. 4. MacArthur River AMT dataset – model verification
Tests to justify a 2-D interpretation
Measured data at 10 Hz
Computed response of 3-D model
3-D effects in induction vectors not due to termination of conductors

22. 4. MacArthur River AMT dataset – model verification
Tests to justify a 2-D interpretation
Measured data at 10 Hz
Computed response of 3-D model
Rose diagram can hide 3-D behaviour
Large r.m.s. misfit values can be diagnostic of 3-D effects

23. 4. MacArthur River AMT dataset – model verification
Resolution from 2-D synthetic inversions 2 1
depth (km) 2 1
depth (km) 2 1
distance (km) 3 2 1
distance (km) 3
Resistivity values in ohm-m
10% noise added to synthetic AMT data
Invert TE, TM and Tzy data

24. 5. Other studies
GEOTEM channel 12
Vertical magnetic field
AMT study in Athabasca Basin by Leppin and Goldak (2006)
Inversion of TE tipper. Apparent resistivity data at every 4th station
Previous applications in USSR (Olex Ingerov, personal communication, 2006)

25. NW SE 6.Conclusions Future research
Depth and dip of the basement conductor
can be reliably mapped to 2 km
Vertical magnetic fields very useful
2D inversions validated by 3D inversions
(likely not true for all deposits)
Features above the unconformity may
be artifacts of the inversion – beware!
Future research
Evaluate other AMT datasets
Integrate various EM methods
Sharp bound inversions
More objective comparison of the 3D codes

26. Acknowledgements
Grants from the Natural Sciences and Engineering Research Council
of Canada (NSERC) and Alberta Ingenuity Fund to Martyn Unsworth
are gratefully acknowledged
AMT data collection was made possible by the financial support of
Cameco, Cogema, Geosystem and the Geological Survey of Canada
Charlie Jefferson (GSC) is thanked for his enthusiasm and initiative
during the EXTECH-IV project
The Geosystem field crew are thanked for the high quality of the AMT data
Alan Jones and Gary McNeice are thanked for the use of their tensor
decomposition code (STRIKE)
We thank Randy Mackie for use of his 2D inversion and for the
3D AMT inversion of the EXTECH-IV dataset