A i N li B h i f S il iAssessing Nonlinear Behavior of Soils in Seismic Site Response:

Statistical Analysis of KiK-net Strong-Motion DataSt o g ot o ata

Julie Régnier1, L.F. Bonilla2, E. Bertrand1 & J.F. Semblat2

1 CETE Méditerranée, laboratoire de Nice2 IFSTTAR, Marne-la-Vallée

Tokyo, the 25th of september 2013

Introduction

Non-linear soil behaviour: strong impact on site response1- Example with Tohoku recordings

Even for moderate events:2- Statistical analysis on all KiK-net sites2- Statistical analysis on all KiK-net sites

What are the effects on site response? pWhat are the pertinent parameters to predict such behavior?3- Statistical analysis on 54 KiK-net sites

E i ll i l i i it ?

Need to take this phenomenon into account

Especially in low seismicity areas? Numerical simulations. But, first we need site characterization/ expensive tests:

What is the depth until which this characterization is

2

prequired ?4- Inversion of non-linear site response

1- Example with Tohoku recordings

• Comparison of site responses (Fourier spectral ratios on borehole data)

Non-linear soil behaviour : strong impact on site response

Vs (km/s) Vs (km/s)

• Comparison of site responses (Fourier spectral ratios on borehole data)• Tohoku main event versus aftershocks at 4 KiK-net sites

0

20

40

0 1 2 3

th (

m)

30

40

50

60

e re

spon

se

IBRH11 Vs30= 242 m/s0 1 2 3

( )

30

40

50

60

e re

spon

se

IWTH21 Vs30= 521 m/s

60

80

100

Dep

t

0

10

20

Bor

ehol

e

0

10

20

Bor

ehol

e

0 60IBRH16 Vs30= 626 m/s

60MYGH04 Vs30= 850 m/s

20

40

60Dep

th (

m)

20

30

40

50or

ehol

e re

spon

se

20

30

40

50

oreh

ole

resp

onse

80

1005 10 15 20 25 30

0

10

Bo

Frequency (Hz)5 10 15 20 25 30

0

10

Bo

Frequency (Hz)

• Shift of the amplification peak toward low frequencies

• Decrease of the amplification at high frequencies 3

• Shift of the amplification peak toward low frequencies Increase amplification at low frequencies !!!

2- Statistical analysis on all KiK-net sites

Non-linear soil behaviour : strong impact on site response even for moderate events• Comparison of site responses (Fourier spectral ratios on borehole data)• For all events we calculated: PNLev (%) Percentage of non-linearity:

Frequency (Hz)

4

2- Statistical analysis on all KiK-net sites

Non-linear soil behaviour : strong impact on site response even for moderate events• Comparison of site responses (Fourier spectral ratios on borehole data)• For all events we calculated: PNLev (%) Percentage of non-linearity:

A1

Frequency (Hz)

5

2- Statistical analysis on all KiK-net sites

Non-linear soil behaviour : strong impact on site response even for moderate events• Comparison of site responses (Fourier spectral ratios on borehole data)• For all events we calculated: PNLev (%) Percentage of non-linearity:

A1

A2A1

A2PNLev=

Frequency (Hz)

6

2- Statistical analysis on all KiK-net sites

Probability PNLev > 10%: NL ≠ linear site response ?PGA > 75 cm/s2 at depth ≈ PGA > 150 cm/s2 at the surface

P ≥ 40% Whatever the site!!Probability of having large modification linear/non-lineary g gsite response is high even for moderate EQ

French seismic zonation(22 octobre 2010)

Very weak

Weak

Agr

70 cm/s2Weak

moderate

Medium

strong

70 cm/s2

110 cm/s2

160 cm/s2

300 cm/s2strong 300 cm/s2

http://www.planseisme.fr/Zonage-sismique-de-la-France.html

3- Statistical analysis on 54 KiK-net sites

Sites selection

• 54 sites

• 2 EQ PGA >50 cm/s2 PGA >50 cm/s(downhole)

• Variability Vs30 OK

8

3- Statistical analysis on 54 KiK-net sites

Per site : definition of fNL

Linear to non-linear site response ratioField et al (1997)

fNL frequency from which the ratio is >1

f>fNLBFSRlin > BFSRstrong

f<fNLBFSRlin < BFSRstrong

Fréquence (Hz)9

3- Statistical analysis on 54 KiK-net sites

Correlation fNL with soil and site response parameters

A

• f0 : Funadamental resoance frequency: H/V

Apred

q ysurface

• Apred : Maximal amplitude f th li itof the mean linear site-

response BFSR • F : Assoicated

fpred

• Fpred : Assoicated frequency to Apred f0

3- Statistical analysis on 54 KiK-net sites

fNL ≥ f fNL ≤ f

Correlation fNL with soil and site response parameters

fNL ≥ f0 fNL ≤ fpred

z)fN

L (H

z

fNL between f0 et fpred :0 pred HF more affected by NL soil behaviourNL takes mostly place in the soil layers linked to HF amplificationsShallow layers 11

3- Statistical analysis on 54 KiK-net sites

These observations suggest

• fNL between f0 et fpred : NL take mostly place in shallow soil layers?

• Measure of NL soil properties enough in the shallow soil layers to characterize the whole soil column NL response?

•Quantification of the depth where it occurs mostly?•Quantification of the depth where it occurs mostly?

• Use of inversion techniques: Inverting the site response to recover the equivalent Vs profile : Weak motion versus strong motionq p g

12

4- Inversion of linear/non-linear site responseApproach

Linear site response

Adjusting the elastic Vs(z)

InversionComparison Vs inversion linear / NLlinear / NL

finding the i l V ( )

Non-linear response

equivalent Vs(z) for strong

deformation

I i

13

Inversion

4- Inversion of linear/non-linear site response

Site selection criteria: - 1-D site configuration, - with weak and strong motion recordings

Comparison empirical and numerical linear

i

Comparison empirical linear and non-linear

site responsesea a d o easite responses

14

4- Inversion of linear/non-linear site response

Best solution from inversion : layer 3 deeper

Vs adjustement

Best solution from inversion : layer 3 deeper

in agreement with geological log

15

Inversion non-linear response

Parametrization & initial model

• Depth of soil layers fixed from linear inversion

• Vs intervals from linear inversion CV = 60 %Vs intervals from linear inversion CV 60 %

• Test on 3 EQs

EQ Mjma Depi Depth PGAsurf

Km Km cm/s2

1 9 202 24 372

2 7.1 28 10 128

3 6.3 24 10 1103 6.3 24 10 110

16

Inversion non-linear response

% f d f V d t l ti ti

EQ 1 EQ 2 EQ 3

PGA f 372 128 110

% of decrease of Vs compared to elastic properties

PGAsurf 372 128 110

Vs1 150 43 16

Vs2 150 43 16

m)

Vs3 31 13 20

Vs4 23 0.02 0.02

Z (

m

Vs5 - - -

• EQs 2 et 3 : layers 1, 2 et 3 affectedEQs 2 et 3 : layers 1, 2 et 3 affected

• EQ 1: layers 1, 2, 3 et 4 affected

17

Inversion non-linear response

% f d f V d t l ti ti

EQ 1 EQ 2 EQ 3

PGA f 372 128 110

% of decrease of Vs compared to elastic properties

m)

PGAsurf 372 128 110

Vs1 150 43 16

Vs2 150 43 16

Z (

m

Vs3 31 13 20

Vs4 23 0.02 0.02

Vs5 - - -

• EQs 2 et 3 : layers 1, 2 et 3 affectedEQs 2 et 3 : layers 1, 2 et 3 affected

• EQ 1: layers 1, 2, 3 et 4 affected

18

Superficials layers more affected during the 3 EqsDeeper layers affected during the strongest

Conclusions

Non-linear soil behaviour effects on site response:Decrease of amplification at high frequencyA lifi i i l f iAmplification increase at low frequencies

Probability of having large modification linear/non-linearProbability of having large modification linear/non-linearsite response is high even for moderate EQ (PGA>150 cm/s2)

Site-specific analysis showed that NL soil behaviour: Occurs on superficial soil layersDepth also depends on incident motion intensityDepth also depends on incident motion intensity

In accordance with previous studies:Regnier et al 2013: Statistical analysis on 54 KiK-net sitesAguirre and Irikura (1997): Vs reduction in the second layer between 5 m to 16 m depthPavlenko and Irikura (2006) Soil nonlinearity in the upper 15-25 m of the soil profilesDe Martin et al. (2010) Vs reduction occurred mostly around 35 m depth Assimaki and Li (2012), showed that Vs30 (superficial soil parameter) was a relevant

parameter to analyze non-linear site responses19

Thanks for your attention!!!Thanks for your attention!!!

Bonilla, L. F., Tsuda, K., Pulido, N., Regnier, J., & Laurendeau, A. (2011). Nonlinear site response evidence of K-net and KiK-net records from the Mw 9 Tohoku earthquake. Earth Planets Space, 58.

Régnier, J., Cadet, H., Bonilla, F.-L., Bertrand, E., Semblat, J.-F. (2013-a) Assessing nonlinear behavior of soils in seismic site response: Statistical analysis on KiK-net strong motion data. Bulletin of Seismological Society of America, accepted in february 2013, Volume 103, Issue 3.

Régnier J Bonilla F L Bertrand E Semblat J F Influence of the velocity profile at depth on the linear

20

Régnier, J., Bonilla, F.-L., Bertrand, E., Semblat, J.-F. Influence of the velocity profile at depth on the linear characterization of site effects: Tests on the KiK-net database, Bulletin of Seismological Society of America, submitted in january 2013-b