1 of 37

Experimental Design for Seismic Mass Movement Monitoring

Dominik Strutz [1] , Tjeerd Kiers [2], Cedric Schmelzbach [2] ,

Hansruedi Maurer [2], Andrew Curtis [1]

[1]: University of Edinburgh, contact: dominik.strutz@ed.ac.uk

[2]: ETH Zurich

2 of 37

2

Cuolm da Vi slope instability

3 of 37

3

model prior information

experimental design

4 of 37

4

experimental design

Design 1

5 of 37

5

experimental design

Design 1

6 of 37

6

experimental design

Design 1

7 of 37

7

experimental design

Design 1

8 of 37

8

experimental design

Design 1

9 of 37

9

experimental design

Design 1

10 of 37

10

experimental design

Design 1

11 of 37

11

experimental design

Design 1

12 of 37

12

experimental design

Design 1

13 of 37

13

experimental design

Design 1

14 of 37

14

experimental design

Design 1

15 of 37

15

experimental design

Design 1

16 of 37

16

Design 1

Design 2

experimental design

quality criterion

17 of 37

data prior information

17

currently hidden

18 of 37

data prior information

18

What can we learn from this

seismic velocity model

noise statistics

heuristic attenuation

currently hidden

19 of 37

19

data prior information

currently hidden

20 of 37

experimental design input

  • model prior

20

21 of 37

experimental design input

21

  • model prior
  • velocity model

​

22 of 37

22

  • model prior
  • velocity model
  • velocity model uncertainty
  • noise statistics
  • (heuristic) attenuation
  • receiver angular sensitivity

​

​

experimental design input

23 of 37

23

  • model prior
  • velocity model
  • velocity model uncertainty
  • noise statistics
  • (heuristic) attenuation
  • receiver angular sensitivity
  • EIG algorithm

experimental design input

NMC Method

Nested loop Monte Carlo Estimator

 

Laplace Method

Assumes Bayesian posterior to be Gaussian

Used to develop framework that allows to implement

heuristic attenuation, and receiver angular sensitivity

24 of 37

24

  • model prior
  • velocity model
  • velocity model uncertainty
  • noise statistics
  • (heuristic) attenuation
  • receiver angular sensitivity
  • EIG algorithm
  • optimisation algorithm

experimental design input

 

25 of 37

25

benchmark

26 of 37

26

benchmark conclusions

27 of 37

design recommendations for DAS

27

28 of 37

design recommendations for DAS

28

29 of 37

design recommendations for DAS

29

30 of 37

design recommendations for DAS

30

31 of 37

design recommendations for DAS

31

32 of 37

design recommendations for DAS

32

33 of 37

design recommendations for DAS

34 of 37

design recommendations for DAS

35 of 37

design recommendations for DAS

36 of 37

Conclusion

  • Densely instrumented real-world scenario allows to benchmark experimental design algorithms and choices
  • Incorporating receiver angular sensitivity allows to give design recommendations for DAS cable layouts

​

​

36

Outlook

  • Run (synthetic) benchmarks on full landslide area to see effect of model prior and attenuation better
  • Summarise results and give recommendations for future deployments

​

37 of 37

Accessibility

37

Visit my website (dominik-strutz.github.io) for further updates and contact information

github.com/dominik-strutz/GeoBED

Implementations of all algorithms presented in this talk.

Will be extended and made more user friendly in the future.

A paper discussing experimental design for seismic monitoring is in preparation