On the estimation of earthquake recurrence along faults and the
crucial role of intermediate magnitudes in seismic hazard assessment
O. Scotti, C. Clement and F. Bonilla
Institut de Radioprotection et Sûreté Nucléaire, Bureau d'Evaluation du Risque Sismique BP 17 - 92262 FONTENAY-AUX-ROSES – FRANCE
Let us consider, for exemple, the
Durance fault system in SE France
P
Seismological and structural data:
•
•
•
•
1999-2004
A
length: 60 km
segments: 10 km long
seismogenic depth :5 to 10 km
slip motion: strike-slip to reverse
B
P
• slip rate: 0.07 (GPS) to 0.2 mm/yr
(geological markers)
• one M>6 paleo-earthquake
• four M~5 historical events (~every
100 years )
1000 years
IRSN/BERSSIN
permanent GPS
Système de failles de la Durance
NW
A
Pli de Manosque
1
0 km
Bassin de Valensole
Mio-Pliocène
Mioc ène inf.
Oligocène
1
B
Crétacé
1
0 km
1
P
2
2
• few 2<M<3 instrumental events in
the last 5 years
Seismic stations
SE
3
3
Jurassique
4
4
Trias
5
Socle varisque
6
5
6
~25000 years
old M>6 event
From Ghafiri, 1995
Adapted from Benedicto 1996
On the difficulty of choosing the appropiate earthquake recurrence model
truncated exponential (GR), characteristic (YC) and maximum magnitude (CE)
Let us assume that activity rates, l, can be deduced from slip rates using :
probzility density fonctions
- Maximum magnitude
Mmax = 4.33 + 0.90 log10(S)

Mmax
- Seismic moment rate
M0  μ Si vi  λi   Mmin f(m)M0(m)dm
- Seismic moment
Mo(m)=10(cM+d)
where v is the slip rate, S is the surface of the fault, m the rigidity modulus.
We then apply three models available in the literature to compute the probability density
functions for magnitude (as defined by Stewart et al., 2001).
The seismologiacl and structural data allows to formulate a variety of hypotheses on
the geometry and slip rates. Here we show the resulting return period estimates of
M>=5 events for a 10 km long segment. In spite of the important uncertainty in the
seismological data, the key uncertainty is the choice of the most adequate
earthquake recurrence model. The GR model, which is rarely used in practice, leads to
Moment rate (Nm/yr)
slip rate (mm/yr)
fault depth (km)
the highest return period estimates in this approach.Which is the most adequate
recurrence model that best describes Durance-type faults? GR, YC or CE?
On the difficulty of estimating activity rates: which is more reliable, catalogue data or slip rates?
Let us now use instead the historical record, which suggests M~5 events recur every 100 years along a segment of the fault. Assuming
such events can occur anywhere along the fault and that they represent the maximum magnitude events, then their return period on a
single fault segment can lower to 500 years. But which is the physical reason to limit Mmax at M~5? Fault maturity? Strain rate?
Which seismic hazard level and scenarios for such faults?
Assuming a 10 km long and 10 km deep vertical fault
segment and a slip rate of 0.07 mm/yr, seismic hazard is
calculated for a site located at a distance of 10 km, using
the activity rates and recurrence models discussed above.
Annual occurrence
1E-2
GR
CE
YC
Mmax=5.0
1E-3
1E-4
1E-5
1E-6
Attenuation relationship
Abrahamson and Silva (1997)
1E-7
Reverse faulting ; 3
0
s
50
100
150
200
Peak Ground Acceleration (cm/sec2)
Hazard levels and hazard
scenarios vary strongly due to
uncertainty in both recurrence
models and activity rate
estimates of intermediate
magnitude events
Deagregation of hazard for a 10-5 annual probability level
GR
CE
YC
Conclusions Using the Durance fault segment as an example we showed
-that it is the uncertainty in earthquake recurrence models that has the
greatest impact on return period estimates compared to seismicl data.
- that the large contribution to the hazard of low probability ground
motions allows lower magnitude bins to contribute down to very low
probability levels.
We thus emphasize the crucial role of intermediate magnitude
earthquakes in hazard assessment as well as the need to find
measurable field parameters that may be injected in numerical models
to help discriminate among the different seismicity models for faults in
moderate seismicity regions.
Benedicto A. ; 1996 : Modèles tectono-sédimentaires de bassins en extension et style structural
de la marge passive du Golfe du Lion (partie nord), sud-est de la France. PhD thesis, Univ.
Montpellier, 235 pp
Ghafiri A. (1995). Paléosismicité de failles actives en contexte de sismicité modérée: application
à l'évaluation de l'aléa sismique dans le Sud-Est de la France. Thèse de Doctorat, Univ. Paris XI,
337 pages
J. P. Stewart, S.J. Chiou J., D. Bray R.W., Graves, P. G. Somerville and N. Abrahamson (2001)
Ground Motion Evaluation Procedures for Performance-Based Design. PEER Report 2001/09.
On the estimation of earthquake recurrence along faults and the
key role of intermediate magnitudes in estimatic seismic hazard
O. Scotti, C. Clement and F. Bonilla
The 4th International Workshop on
Statistical Seismology
9-13 January 2006
Japan
nstitut de Radioprotection et Sûreté Nucléaire, Bureau d'Evaluation du Risque Sismique BP 17 - 92262 FONTENAY-AUX-ROSES – FRANCE
Consider, for example, the seismological and tectonic setting of the Durance fault system in SE France
Length
Depth
60 - 90km
Slip motion
5-10 km
reverse to strike-slip
Slip rate (mm/yr)
5 Segments
0.07 (GPS) to 0.2 (geological)
Seismicity
~10 km long (geophys./cartog.)
1999-2004
Location
1
M>6
9000-25000 yrs old
Segment 3
4
M~5
in the last 400 yrs
Segment 3
< 10
2<M<3
in 5 yrs
Mostly on
segments 2,3,5
IRSN/BERSSIN
1000 years
permanent GPS
Seismic stations
Choosing the appropriate earthquake recurrence model: GR,YC and/or CE?
1E-2
Input data: Geometry of segment 3 (M~6 vertical fault), slip rates (v) and a regional b-value
1E-3
M0  μ Si vi  λi   Mmin f(m)M0(m)dm
Mmax
Mo(m)=10(cM+d) (S = segment surface; m =rigidity modulus; i=segment 3; M=magnitude)
and Seismic moment:
the activity rate (l) can be estimated by distributing the moment rate following the GR, YC and CE recurrence models.
For acceptable depth and slip rates hypothesis, resulting return period estimates vary by 50 to 80% a given recurrence model,
Uncertainty in the choice of the recurrence model, however, reaches 100%.
Annual probability
Given Seismic moment rate:
GR
CE

YC
1E-4
1E-5
truncated exponential
(GR)
1E-6
characteristic (YC)
Which model is more compatible
100000
with the seismicity catalogue?
maximum magnitude (CE)
1E-7
5
10000
CE
GR
YC
5,5
6
Magnitude
6,5
7
Less than 10 events of 2<M<3 have occured in 5 yrs along segment 3, 2 and 5; 4 M~5 historical
events and one M~6 paleoearthquake on segment 3; segment 1 and 4 appear to be “inactive”.
Although the seismicity record may not be sufficiently long to capture the steady-state
1000
behaviour of the entire fault system, the seismic behaviour of segment 3 appears more
compatible with a GR-type recurrence model. Assuming the entire fault is active, should we
100
Moment rate (Nm/yr)
fault depth (km)
slip rate (mm/yr)
1,05E+14
2,09E+14
1,49E+14
5
10
5
0,07
2,99E+14
2,99E+14
10
5
0,1
5,98E+14
10
0,2
apply the GR model to the other segments as well? Is this behaviour typical of low strain-rate
faults? Alternatively, is the absence of seismicity evidence of their CE-character?
Are there parameters other than seismicity that can be considered as indicative of each recurrence model?
Impact of recurrence models and on seismic hazard estimations?
Conclusions
Example of hazard curves for segment 3. Hazard is calculated for a 10 km distant site
using the attenuation relationship of Abrahamson and Silva (1997), three recurrence

models with Max=6.5 and M 0 = 2,09e+14 Nm/yr.
1E-2
GR
CE
Annual exceedance
1E-3
YC
Uncertainty in recurrence models,
as well as in the estimation of “observed”
1E-4
activity rates combined with
1E-5
ground motion variability lead to
great differences in hazard estimates and
1E-6
in the scenarii that contribute the most to
s
1E-7
0
50
100
150
200
Peak Ground Acceleration (cm/sec2)
the hazard.
Using the Durance fault as an example we showed that :
1. Using catalogues to estimate return periods of earthquakes in regions of moderate–to-low
seismicity is delicate and subjected to the “hazard” of the catalogue window.
2. On the other hand, return periods of intermediate magnitude events estimated through slip
rates and fault geometries depend strongly on the choice of the recurrence model.
3. Segment 3 appears to be more GR-like, but what can be said of the other segments?
The GR model is the least commonly used recurrence model in seismic hazard assesment
studies.
Why? How to discriminate among the different seismicity models for Durance-type faults?
Which are the key parameters?
Deagregation of hazard for a given target level
It is important to provide answers to these questions because:
GR
CE
YC
3. Intermediate magnitude events contribute to the hazard down to very low probabilities due
to the strong contribution of high epsilons.
4. Scenarios resulting from GR-models are very different than those resulting from YC or CE
recurrence models.
Sismicité
instrumenta
le sur la
FMD
Benedicto Dcalae de
,
la base du
0.08Miocene
Durance 0.14
1996
depuis 20
(v)
(modified
ma
)
Dcalae
verticale
Hippolyte de la base Valavoi
&
du
re (NE
0.1
(v)
Dumont, chevauxhem Sisteron
ent de la
)
2000
nappe de
Dine (6ma)
Cushin et
0.05Dcalaes de Manosqu
al.,
0.22
cours deau
e
1997
(h)
Etude
omorpholo
ique,
modlisation Manosqu
Baroux,
0.11
e
de la
2000
(v)
dformation
fold
de
lanticlinal
de Manosque
ranche de
alosismicit
0.07Ghafiri,

Manosqu
0.13
1995
Valvranne,
e
(v)
sisme entre
Localisati
ons 3D
19992004
.
2. Potentiel
sismique de
la FMD
1. Paléosismologie
2. Dimension de la faille
a)
Segmentatio
n
b)
Géométrie
3D
c)
Enracinement
?
3. Vitesse de la faille
Représentation simplifiée du système de failles
FMD – support de modélisation - quantification
2. Potentiel
sismique de
la FMD
Exemples d’études permettant de quantifier la vitesse
de la faille
Auteur
1. Paléosismologie
2. Dimension de la faille
a)
Segmentation
b)
Géométrie
3D
c)
Enracinement
?
3. Vitesse de la faille
Marqueur/mét Lieu Vitess
hode
e mm/a
Benedicto, 1996
(modified)
Décalage de la base du
Miocene depuis 20 ma
Durance
0.080.14 (v)
Hippolyte & Dumont,
2000
Décalage verticale de la
base du chevauxhement
de la nappe de Digne
(6ma)
Valavoir
e (NE
Sisteron)
0.1 (v)
Cushing et al., 1997
Décalages de cours d’eau
Manosq
ue
0.050.22 (h)
Baroux, 2000
Etude géomorphologique,
modélisation de la
déformation de
l’anticlinal de Manosque
Manosq
ue fold
0.11 (v)
Ghafiri, 1995
Tranchée de
Paléosismicité à
Valvéranne, séisme entre
25 et 9 ka
Manosq
ue
0.070.13 (v)
Siame et al. 2001
Datation de terrasse en
utilisant les isotopes
cosmogéniques
Manosq
ue
0,02 –
0,11 (v)
GPS permanent depuis 4
ans
StMichell’O. Ginasser
vis
0,05 (h)
Nockey,2005
CONCLUSION
Sismicitrguli
re modre M ~ 5 5,5
Vitesse de glissement par diffrentes
approches de lordre de 1/10e mm/an
Faille segmente (10- 20 km) Magnitude >=
6,5 Conditionne par la profondeur
dinitiation de la rupture
Priode de retour des sismes majeurs de
lordre de 10000 ans. ! On ne connat
pas le comportement des failles lentes.
Si toute la faille casse, la magnitude peut
atteindre 7. Priode de retour plus longue
Let us consider, for example, the Durance fault system in SE France: Seismological and tectonic setting
Length
Seismogenic depth (km)
Slip motion
Slip rate (mm/yr)
Number of Segments
60 km
Geologic cross-section: 5 km ?
Geological:reverse
GPS: 0.07
5 geophysical lines
Seismicity: 5-10 km
Instrumental: s/s
Geological: 0.2
~10 km long
1999-2004
NW
Système de failles de la Durance
A
Pli de Manosque
1
0 km
1
B
Bassin de Valensole
Mio-Pliocène
Mioc ène inf.
Oligocène
Crétacé
A
4
Trias
Socle varisque
6
B
1
3
Jurassique
4
5
0 km
2
2
3
SE
1
5
6
Adapted from Benedicto 1996
PP
IRSN/BERSSIN
Seismicity
M>6: 1
permanent GPS
M~5: 4
From Ghafiri, 1995
Segment 3
based on paleo
Seismic stations
P
~25000 years
old M>6 event
Location
Segment 3
in last 400 yrs
1000 years
M<3: less than
10 in 5 yrs
All segments