Bulletin of Earthquake Engineering (2023) 21:4821–4849

https://doi.org/10.1007/s10518-023-01710-1

ORIGINAL ARTICLE

Fault displacement hazard estimation at lifeline–fault

crossings: A simplified approach for engineering applications

Vasileios E. Melissianos1 · Laurentiu Danciu2 · Dimitrios Vamvatsikos1 ·

Roberto Basili3

Received: 25 June 2022 / Accepted: 12 May 2023 / Published online: 5 July 2023
© The Author(s) 2023

Abstract
Lifelines, such as pipelines, roads, and tunnels, are critical infrastructure and when cross-
ing active tectonic faults, a reliable estimation of the fault displacement in case of an earth-
quake is required. The first and simplest approach is to use empirical fault scaling rela-
tions to compute the design fault displacement, but this may result in an unknown level of
safety. Thus, the probabilistic fault displacement hazard analysis (PFDHA) is the appropri-
ate tool to assess the fault displacement hazard within a performance-based framework.
Based upon an established PFDHA model, we present a simplified approach for engineer-
ing applications focusing on the lifeline–fault crossing along with appropriate simplifica-
tions and assumptions to extend its applicability to numerous faults. The aim is to provide
a structure-independent approach of PFDHA that can be used when a site-specific study
is not required, not possible (e.g., absence of recent sediments for dating past events), or
too cumbersome, e.g., for lifeline route selection. Additionally, an in-depth investigation
is presented on the key parameters, such as maximum earthquake magnitude, fault length,
recurrence rate of all earthquakes above a minimum magnitude, and lifeline-fault crossing
site, and how they affect the hazard level. This approach will be the basis for deriving haz-
ard-consistent expressions to approximate fault displacement for use within the Eurocodes.
The latter is intended to serve as a compromise between hazard-agnostic fault scaling rela-
tions and a comprehensive PFDHA, which requires detailed calculations and site-specific
seismological data.

Keywords Lifeline–fault crossing · Fault displacement hazard analysis · Fault scaling

relations · Seismological parameters · Uncertainties

* Vasileios E. Melissianos
[email protected]
1
School of Civil Engineering, National Technical University of Athens, 9 Iroon Polytechneiou str.,
Zografou Campus, 15780 Athens, Greece
2
Swiss Seismological Service, ETH Zurich, Sonneggstrasse 5, CH‑8092 Zurich, Switzerland
3
Istituto Nazionale di Geofisica e Vulcanologia, Via di Vigna Murata 605, 00143 Rome, Italy

13
Vol.:(0123456789)
4822 Bulletin of Earthquake Engineering (2023) 21:4821–4849

1 Introduction

Lifelines, such as fuel, water, and sewage pipelines, bridges, and tunnels, are critical infra-
structures for the operation and prosperity of communities. These structures extend over a
long distance and therefore multiple crossings of minor and major faults in seismic areas
might be inevitable (Rondoyanni et al. 2013). In such a case, the fault activation will force
the structure to develop excessive deformation with potentially devastating consequences,
as identified in previous earthquake events for buried pipelines (O’Rourke and Liu 2012;
Girgin and Krausmann 2016), above-ground pipelines (Honegger et al. 2004), and under-
ground tunnels (Anastasopoulos et al. 2008; Ghadimi Chermahini and Tahghighi 2019).
These facts render the implementation of the performance-based earthquake engineer-
ing framework (Cornell and Krawinkler 2000) necessary for the quantitative assessment
of lifeline safety against earthquake hazard. A risk reduction strategy for the design, con-
struction, operation, and maintenance of critical lifelines is, also, now dictated by pertinent
guidelines under the Sendai Framework for Disaster Risk Reduction 2015–2030 (United
Nations 2015).
The design fault displacement for lifelines founded on the ground surface or buried in
the ground is typically estimated via a set of empirical fault scaling relations (e.g., Wells
and Coppersmith 1994; Leonard 2014). This deterministic approach inherently leads to an
unknown level of conservatism and safety (Bommer 2002) because the fault displacement
is computed solely based on the fault dimensions and/or a characteristic earthquake magni-
tude. Even though this is a straightforward and easy-to-apply approach, the actual seismic-
ity of the fault, the location of the crossing site on the fault trace, and the rupture length,
among others, are disregarded. Given the importance of lifelines and in order to overcome
the aforementioned drawbacks, one could employ a case- and site-specific Probabilistic
Fault Displacement Hazard Analysis (PFDHA), which is the appropriate tool to quan-
tify the potential of coseismic surface fault displacement. A comparison between these
two alternative paths for estimating the design fault displacement is given by Melissianos
(2022) through the investigation of three characteristic examples of faults in Europe.
The necessary data for PFDHA have been reported by Coppersmith and Youngs (2000)
and its basis was established by Youngs et al. (2003) as the outcome of an extensive seis-
mic hazard investigation that was carried out for the planned nuclear repository at Yucca
Mountain, Nevada, USA (Stepp et al. 2001). Treiman (2010) presented a discussion from
the geological point-of-view on the problems arising from the estimation of future fault
displacement, focusing on rupture, deformation, and slip anticipation. Petersen et al.
(2011) extended the PFDHA for strike-slip fault mechanism, providing more detailed cal-
culations for off-fault displacements and focusing on mapping accuracy, fault complex-
ity, as well as the plan area of the structure (site of interest). Chen and Petersen (2011)
employed the approach of Petersen et al. (2011) to develop an empirical slip model for the
southern San Andreas fault in California, USA, to reduce the uncertainties confronted in
the design. Moss and Ross (2011) analyzed data collected from reverse fault ruptures and
provided appropriate fault displacement prediction equations and conditional probability of
slip functions. Recently, Chen and Petersen (2019) proposed an improvement regarding the
uncertainty of rupture location for PFDHA by increasing the correlation between the struc-
ture’s footprint, i.e., site of interest, and the fault displacement hazard.

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4823

Extensive research on the assessment of fault displacement hazards for engineering appli-
cations has been conducted mainly for the nuclear industry (Gürpinar et al. 2017; Interna-
tional Atomic Energy Agency 2010; Valentini et al. 2021). Contrarily, the pertinent studies for
lifeline-fault crossings are scarce. Indicatively, for the case of a buried pipeline–fault crossing,
Angell et al. (2003) presented an initial attempt to estimate the fault displacement hazard for
offshore pipelines using the newly established methodology of Youngs et al. (2003). Melis-
sianos et al. (2017a) provided a performance-based framework for evaluating the seismic risk
of buried pipelines at fault crossings. Then, Melissianos et al. (2017b) used this framework
to evaluate on a risk basis the effectiveness of alternative seismic countermeasures for buried
pipes subjected to faulting. At the same time, Cheng and Akkar (2017) also addressed the
probabilistic risk assessment of buried pipelines by estimating the fault displacement hazard
using Monte Carlo simulations.
The application of PFDHA requires detailed calculations and a large set of data, which
may not be available for every single fault crossing along the route of a lifeline. At the same
time, design engineers are typically unfamiliar with these calculations and data, and so they
tend to use the simpler empirical fault scaling relations. In that sense, our intension is to find
the middle ground between the simpler approach (i.e., deterministic via the scaling relations)
and the more “accurate” probabilistic approach (i.e., PFDHA). Thus, the aim is to formulate a
methodology for a hazard-consistent estimation of the fault displacement using limited data,
while being also simple enough to be applied by engineers. Such a methodology can be incor-
porated in a design code, such as Eurocode 8 (EN 1998) for the earthquake-resistant design of
structures. At this point, it is important emphasize that the current European and international
seismic design codes lack such provisions for estimating the design displacement for lifelines
crossing active tectonic faults.
Owing to the above, the present study provides the context towards this direction by pre-
senting an application of PFDHA for lifeline–fault crossings, with an emphasis on the caclu-
lation decisions and the assumptions made. This simplified approach for engineering appli-
cations is an extension of the previous work on the seismic performance-based assessment
of buried pipelines at fault crossings presented by Melissianos et al. (2017a). The methodol-
ogy presented is austere by modern seismic design recommendations and it is tailored for the
typical practitioner. Thus, the focus is only on the main properties of an active fault, namely,
dimensions (length, width, and area), mechanism, and seismicity, as they can be found in
available fault databases. Additional fault complexity such as secondary effects, fault segmen-
tation, and linkage are not considered in this analysis. Furthermore, a parametric investigation
is offered on the seismological parameters that contribute to the hazard level and shape the
hazard curve, followed by a discussion on their influence. The uncertainty associated with
the selection of alternative fault scaling relations that are required in the hazard calculations
is, also, discussed. The outcomes are applicable to thousands of faults given fault characteris-
tics present in a typical seismic source model used in Probabilistic Seismic Hazard Analysis
[PSHA (Cornell 1968)]. The proposed approach is generic and applicable at a regional scale,
however, for site-specific applications, in-depth characterization of the active faults via, for
example, geophysical prospecting or paleoseismological investigations, is required. As a final
remark, the proposed approach is structure-independent. Consequently, the obtained fault dis-
placement should be converted to imposed ground displacement on the lifeline by taking into
account the fault dip angle and the lifeline–fault crossing angle.

13
4824 Bulletin of Earthquake Engineering (2023) 21:4821–4849

2 Fault displacement hazard

Youngs et al. (2003) introduced two options for the PFDHA, namely the “earthquake
approach” that follows the PSHA methodology and the “displacement approach” that
requires paleoseismic data. Herein, the former is adopted because acquiring paleoseis-
mic data for every single fault crossing of a lifeline extending over hundreds of kilom-
eters is (1) practically not viable and (2) incompatible with seismic design codes. The
“principal faulting” is also adopted, neglecting any distributed faulting issues and dis-
placements “outside” the fault trace. Principal faulting stands for the slip along the main
plane of the earth’s crust discontinuity that leads to the release of energy during an
earthquake event. The seismic productivity of each fault is assumed to be constrained by
a rupture zone within the fault geometry. This assumption can be overly conservative for
a building located in this rupture zone. On the other hand, it is a fair approximation for
the total displacement imposed on a continuous lifeline that crosses the entirety of this
zone, encountering both primary and secondary faulting. Moreover, one could consider
the coseismic deformations persisting several kilometers around the fault trace, as well
as that the rake can vary along the rupture trace. These considerations can improve the
calculations, yet require site-specific data that is foreign to the seismic code and to most
engineers. The key parameters used (variables and abbreviations) throughout the study
are listed in Table 1.
The PFDHA yields the Mean Annual Frequency (MAF) of exceeding a predefined fault
displacement value 𝛿:
Mmax
⎡ ∞ ⎤
∫ ⎢∫
𝜆Δ (𝛿) = v fM (m)⎢ f (r�m)P(Δ > 𝛿�m, r)dr⎥dm (1)
⎥
Mmin ⎣0 ⎦

where 𝜆Δ (𝛿) is the MAF of fault displacement exceeding value 𝛿, v is the recurrence rate
of all earthquakes on fault above a minimum magnitude of engineering significance ( Mmin)
per year assuming a Poisson model, fM (m) is the probability density function of earth-
quake magnitude m being between a minimum Mmin and a maximum Mmax magnitude,
P(Δ > 𝛿|m, r) is the conditional probability that given an earthquake of magnitude m at a
distance r from the site of interest has occurred, the fault displacement will exceed value 𝛿,
and f (r|m) is the conditional probability density function for distance r from the site to an
earthquake of magnitude m occurring at the source under examination.
In the case of a lifeline–fault crossing, where the site of interest is located on the fault
trace (Fig. 1), the fault displacement hazard depends on the following:

• Earthquake (moment) magnitude ( M ) is the key factor describing the size of an earth-
quake, ranging between values Mmin and Mmax . Earthquake magnitude is the condition-
ing variable of the calculations.
• Rupture length ( RL ) is the segment of the fault that ruptures, acknowledging that dif-
ferent earthquakes may rupture fault segments of different lengths. Earthquake magni-
tude and rupture length are correlated (Wells and Coppersmith 1994; Leonard 2010).
Rupture length ranges within RLmin ≤ RL ≤ LF , where RLmin is the minimum rup-
ture length derived from Mmin via fault scaling relations and LF is the fault subsur-
face length that is typically reported by seismic hazard models and thus adopted herein.
It should be noted that the rupture length at depth can be up to 25% higher than the

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4825

Table 1  Key Parameters: variables and abbreviations

Variable* Explanation Abbreviation Explanation

ADS Average fault displacement at the AD Average Displacement

surface
ADD Average fault displacement at depth CCDF Complementary Cumulative Density
(subsurface) Function
b b-value of Gutenberg-Richter Law CPS Conditional Probability of Slip
LF Length of fault trace MAF Mean Annual Frequency
M/m Earthquake moment magnitude MD Maximum Displacement
RL Rupture length INT Interplate tectonic environment of fault
v Rate of all earthquakes on fault PFDHA Probabilistic Fault Displacement Haz-
above a minimum magnitude ard Analysis
x Distance to closest rupture-end PMF Probability Mass Function
XL Crossing point PSHA Probabilistic Seismic Hazard Analysis
Z Distance to closest fault-end SCR Stable Continental Region tectonic
environment of fault
Δ/𝛿 Surface fault displacement
𝜆 Mean annual frequency
𝜎 Standard deviation

*Lowercase letters denote the value of the variable

surface rupture length (Wells and Coppersmith 1994). The rupture length is correlated
with earthquake magnitude [see Eq. (12)].
• The position ( Pos) of the rupture length ( RL ) on the fault trace is reflecting the uncer-
tainty of the rupture location on the fault trace. Herein, by the term fault trace, we spe-
cifically refer to the projection of the (subsurface) top boundary of the fault plane onto
the surface. This is typically the one reported in seismological maps and it is also the
fault length employed in seismic source models. The actual manifestation of a surface
rupture may or may not fall on this trace, which is beyond the scope of our investiga-
tion.
• The location of the crossing point on the fault trace stands as the variable to account for
whether the lifeline is intercepted by the ruptured segment, which is represented by XL
that is the ratio of the distance (along the fault trace) of the lifeline–fault crossing point

Fig. 1  Lifeline–fault crossing plan view (definition of fault length LF , distance to closest fault-end Z , and
crossing point XL)

13
4826 Bulletin of Earthquake Engineering (2023) 21:4821–4849

( Z ) to the closest fault-end over the rupture length with 0 < XL = Z∕LF ≤ 0.50. Sym-
metry around XL = 0.50 is considered, resulting in a larger variability (Youngs et al.
2003).

Based on the “earthquake approach” for principal faulting, as developed by Youngs

et al. (2003), we focus on the fault displacement hazard calculations on the lifeline–fault
crossing site by accounting for the aforementioned four elements. Using the total probabil-
ity theorem and conditioning the hazard calculations primarily on the earthquake magni-
tude (Melissianos et al. 2017a), the MAF of exceeding fault displacement values is:
∑ ( ) ( )
𝜆Δ (𝛿) = v P Δ > 𝛿|mi PM mi (2)
i
( )
where P Δ > 𝛿|mi is the probability
( ) that the fault displacement Δ exceeds a defined value
𝛿 at the crossing site and PM mi is the probability of magnitude value ( mi ) falling within
the i-th bin, following an appropriate discretization of magnitude values between Mmin and
Mmax . Τhe lower bound Mmin represents the value below which no impact on the structure
is assumed (minimum magnitude of engineering significance), while the upper bound Mmax
is typically constrained by the fault dimensions. This probability of earthquake magnitude
M being lower than a specific value m is estimated using the Gutenberg-Richter Bounded
Recurrence Law (Gutenberg and Richter 1944):

PM mi = P M < mi + Δm||Mmin ≤ mi ≤ Mmax

( ) ( )

−P M > mi − Δm|Mmin ≤ mi ≤ Mmax

( )

where ∶
[ ( )] (3)
M < m||Mmin ≤ m ≤ Mmax =
( ) 1 − exp −𝛽 m − Mmin
PM [ ( )]
1 − exp −𝛽 Mmax − Mmin
with 𝛽 = b ln 10

The b-value of Eq. (3) defines the slope of the curve that provides the “expected” future
earthquake magnitudes. The effect of b-value is illustrated in Fig. 2 for a magnitude range
between 5.50 and 8.50 and b-value between 0.70 and 1.10 as given in the 2013 European
Seismic Hazard Model [ESHM13 (Woessner et al. 2015)]. The double truncation magni-
tude frequency distribution is the default assumption used in this study (Parsons and Geist
2009; de Santis et al. 2011; Page and Felzer 2015).

Fig. 2  Effect of b-value (slope)

of the Gutenberg-Richter Law
on the estimation of future earth-
quake magnitudes

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4827

( )
The probability term P Δ > 𝛿|mi in Eq. (2) is a complementary cumulative distribu-
tion function (CCDF) and is estimated via the expression:
( ) ∑∑∑ ( ) ( ) ( )
P Δ > 𝛿|mi = P Δ > 𝛿|mi , RLj , ADSt , Posj,k P RLj , ADSt |mi P Posj,k
(4)
j t k

where subscripts denote the integration over variable values: i for earthquake magnitude
( M ), j for rupture length ( RL ), t for average surface displacement ( ADS ), and k for RLj
positions on the fault trace ( Pos). Please note that variables RL and M , as well as, ADS and
M are correlated via empirical fault scaling relations. Also, RL and fault displacement are
correlated but the correlation of the residuals of ADD and RL conditioned on the magni-
tude is not provided in any pertinent
( study. )
The first right-hand term, P Δ > 𝛿|mi , RLj , ADSt , Posj,k , of Eq. (4) is a CCDF and is
estimated after Youngs et al. (2003) as:
( ) ( ) ( )
P Δ > 𝛿|mi , RLj , ADSt , Posj,k = P Δ > 𝛿|RLj , ADSt , Posj,k , Slip P Slip|mi (5)
( )
The term P Δ > 𝛿|RLj , ADSt , Posj,k , Slip of Eq. (5) is estimated as:
( )
P Δ > 𝛿|RLj , ADSt , Posj,k , Slip = 1 − F(y) (6)

where F(y) is the Conditional Probability of Exceedance depending on the fault mechanism.
F(y) is a cumulative distribution function (CDF), estimated via empirical data models like
the common approach for the ground motion prediction equations. Using historic data, fault
displacement values (𝛿) were normalized by the average surface displacement values ( ADS).
The distribution of the ratio 𝛿∕ADSt is a function of the crossing point XL. The fault displace-
ment is higher in the middle of the LF compared to the ends. In nature, the distribution of slips
along the rupture could be skewed toward one of the two fault-ends (Ward 1997). Which end is
favored is often unknown; therefore, the assumption of symmetry will tend to increase the vari-
ability (Moss and Ross 2011). The distribution of 𝛿∕ADSt is fitted with an appropriate model
depending on the fault mechanism. The average surface displacement for the normalization of
fault displacement data is represented by the variable y in the following Eqs. (7) and (8). The
gamma distribution proposed by Youngs et al. (2003) is used for normal and strike-slip faults:
y
u2

Γ u1 ∫
1 𝛿
F(y) = ( ) e−t zu1 −1 dz, y= >0
ADSt
0 (7)
with
( ) ( )
u1 = exp −0.193 + 1.628XLj,k and u2 = exp −0.009 + 1.476XLj,k

while the Weibull distribution proposed by Moss and Ross (2011) is used for reverse faults:
( )q1
− qy 𝛿
F(y) = 1 − e 2 ,y = >0
ADSt
with
( ) (8)
3 2
q1 = exp −31.8XLj,k + 21.5XLj,k − 3.32XLj,k + 0.431
( )
3 2
q2 = exp +17.2XLj,k − 12.8XLj,k + 3.99XLj,k + 0.380

13
4828 Bulletin of Earthquake Engineering (2023) 21:4821–4849

Fig. 3  A graphical example of considering alternative rupture lengths at indicative non-overlapping posi-
tions (for illustration purposes) of the rupture length with respect to the crossing point. The changing vari-
ables in Eq. (5) are surface rupture length RLj and position Posj,k of the RLj on the fault trace: in Case A
j = 1 and k = 1, 2, … , 6, in Case B j = 2 and k = 1, 2, 3, and in Case C j = 3 and k = 1 [Note that the sub-
script t for integration over ADS in Eq. (5) is omitted because it is not related to fault crossing geometry]

where XLj,k = x∕RLj is the crossing point relevant to the position ( Posj,k ) of the RL (Fig. 1
and 3). It is noted that in case the crossing site is not intercepted by the RLj at position
Posj,k under examination, then by definition the crossing point XLj,k is not computed and
( ) ( )
consequently P Δ > 𝛿|RLj , ADSt , Posj,k , Slip = 0 → P Δ > 𝛿|mi = 0. A graphical
example of considering alternative RLs and positions of RL concerning the crossing point
is illustrated in Fig. 3.
At this point, it should be pointed out that Youngs et al. (2003) and Moss and Ross
(2011) provided the aforementioned data normalization also by the maximum fault dis-
placement (MD). Wells and Kulkarni (2014) reported that the MD is expected to occur
over a limited reach of the fault, and thus, adopting MD is a conservative approach. On the
other hand, the AD is more representative of the expected displacement along a larger part
of the fault because it is calculated using multiple measurements of displacement along the
rupture zone (Wells and Coppersmith 1994). Therefore, hereinafter only the AD is con-
sidered. It is noted that both AD and MD approaches were examined by Melissianos et al.
(2017a) within a logic tree formulation.

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4829

( )
The second right-hand term, P RLj , ADSt |mi , of Eq. (4) is the joint probability of RLj
and ADSt conditioned on the magnitude, being a probability mass function (PMF) due to
the discretization of the variables. Estimation of this term is based on fault scaling rela-
tions as per Sect. 3. If the corresponding relations are provided at depth rather than at the
surface, then a transformation is required. In the examined case, a simple linear transfor-
mation of ADDt = 1.32ADS(t is adopted, based ) on
( the findings)of Wells and Coppersmith
(1994). In turn, translating P RLj , ADSt |mi to P RLj , ADDt |mi become as simple as keep-
ing the individual probability masses constant and changing the AD coordinates that these
refer to. Given the simplicity of this change, in the following lines, only the derivation
concerning subsurface properties will be offered. Fault scaling relations allow the distribu-
tion of RL and ADD to be expressed as a joint lognormal function with positive correlation
f (RL, ADD|M) that deals with the fault rupture at the subsurface level and consequently
both quantities, namely rupture length and average displacement, refer to the subsurface
level. A sufficiently fine discretization of the variables into RLj and ADDt is adopted,
allowing the probability to be approximated via a single probability density function (PDF)
value at its center via the expression:
RLstep ADDstep
RLj + 2
ADDt + 2
( )
∫ ∫
Pi,j,t = f RL, ADD|mi dADDdRL
RLstep ADDstep
(9)
RLj − ADDt −
( )
2 2

≅f RLj , ADDt |mi ΔADDΔRL

Note that the sum of Pi,j,t over j and t for all RLj and ADDt should equal one and there-
fore a renormalization is performed for the derived PMF to account for any minor numeri-
cal issues:
� � � � Pi,j,t
P RLj , 1.32ADSt �mi = P RLj , ADDt �mi = ∑ ∑ (10)
j t Pi,j,t
( )
The function f RLj , ADDt |mi of Eq. (9) is estimated as a multivariate joint lognormal
PDF using the mean value and the covariance matrix of the corresponding normal. The
mean value [𝜇] is:
[ ]
𝜇log10RL
[𝜇] =
𝜇log10 ADD (11)

where 𝜇log10 RL and 𝜇log10 ADD are the mean log values of RL and ADD given an earthquake
magnitude mi and are computed using empirical fault scaling relations, such as those of
Wells and Coppersmith (1994), Leonard (2014), and Thingbaijam et al. (2017), in the gen-
eral form of:
RL ∼ g(M) (12)

ADD ∼ h(M) (13)

whereby the tilde operator ( ~) denotes that the functions g(⋅) and h(⋅) provide the distribu-
tion conditioned on their argument, for example, as defined by regression. Then, the covar-
iance matrix [𝜌] for Eq. (9) is estimated as:

13
4830 Bulletin of Earthquake Engineering (2023) 21:4821–4849

[ ]
2
𝜎log 𝜌log10 RL,log10 ADD 𝜎log10 RL 𝜎log10 ADD
10 RL
[𝜌] =
𝜌log10 RL,log10 ADD 𝜎log10 RL 𝜎log10 ADD 2
𝜎log (14)
10ADD

where 𝜎log10 RL and 𝜎log10 ADD are the log standard deviations derived from the regression that
produced the scaling relations of Eqs. (12) and (13), respectively, while 𝜌log10 RL,log10 ADD is
the correlation coefficient. ADD values based on RL are sampled using a fault scaling rela-
tion in the general form of:
ADD ∼ r(RL) (15)
Still, the correlation of the residuals of ADD and RL conditioned on the magnitude is not
provided in any pertinent study. A discussion on the effect of this correlation is offered in
Sect. 4. ( )
Finally, the third right-hand term, P Posj,k , of Eq. (4) is the probability of a fault seg-
ment of length RLj rupturing at the k-th position along the fault trace. Lacking any better
information, equal-length RLj are assumed to be located on all possible positions on the
trace with the same probability (Youngs et al. 2003). Therefore, this term is a PMF and is
estimated as:
( )
P Posj,k = 1∕Nj (16)

where Nj is the total number of potential positions for the RLj estimated as:
( )
LF − RLj
Nj = floor +1 (17)
RLmin

where the floor(⋅) function provides the largest integer lower than or equal to its argument
and the minimum rupture length of interest ( RLmin) is estimated using a pertinent fault
scaling relation and considering the minimum earthquake magnitude that is of significance
for the case at hand. ( )
The second term of Eq. (5) is the Conditional Probability of Slip (CPS), P Slip|mi , that
accounts for the probability of the rupture reaching the ground surface conditioned only
on earthquake magnitude. Petersen et al. (2011) reported that, for example, after the 1989
Loma Prieta, California ( M = 6.9) and the 2002 Nenana Mountain, Alaska ( M = 6.7)
earthquakes, no surface rupture was observed. The CPS is estimated after Wells and Cop-
persmith (1993) for normal (Youngs et al. 2003) and strike-slip (Petersen et al. 2011)
faults:
[ ] [ ]
P(Slip|m) = exp(−12.15 + 2.053m) ∕ 1 + exp(−12.15 + 2.053m) (18)

and after Moss and Ross (2011) for reverse faults:

[ ]
P(Slip|m) = 1∕ 1 + exp(7.30 − 1.03m) (19)

A comparison of Eqs. (18) and (19) is illustrated in Fig. 4 for earthquake magnitude values
ranging between 5.50 and 8.50. A significant difference is observed, because, for exam-
ple, to have over 80% probability [ P(Slip|m) > 0.80] for the rupture to reach the surface,
a magnitude M > 8.50 event is required in the case of a reverse fault, contrary to a magni-
tude M > 6.77 event in the case of normal or strike-slip fault.

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4831

Fig. 4  Empirical conditional

probability of slip (CPS), i.e.,
probability of the rupture reach-
ing the ground surface condi-
tioned on earthquake magnitude
( M ), with respect to the fault
mechanism (Youngs et al. 2003;
Moss and Ross 2011; Petersen
et al. 2011)

The fault displacement hazard curve at the crossing site is computed using Eq. (2) for
a range of input fault displacement values (𝛿). A discretization scheme is adopted for the
variables under consideration. At first, the earthquake magnitude range is discretized into
several bins from a minimum value ( Mmin) to a maximum one ( Mmax ) that is typically esti-
mated using a fault scaling relation of magnitude given fault length:
Mmax ∼ w(LF) (20)
Then, the implementation of Eq. (4) requires the discretization of LF to multiple RL .
Practically, a minimum RL is considered after Eq. (12) for Mmin , while all larger rupture
lengths are integer multiples of this minimum one. Then, every RL is considered at all
possible positions on the fault trace, which is represented by the variable Pos , keep-
ing track of those that intercept the lifeline (Fig. 3). Only the latter contribute to the
fault displacement hazard, contrary to what earthquake engineers are familiar with for
the ground shaking in PSHA, where the contribution of all nearby seismic sources is
accounted for at the site under investigation.
A case study to illustrate the outcome of the hazard calculations is provided sub-
sequently. The baseline parameters considered for the lifeline–fault crossing are an
interplate normal fault with length LF = 100km , recurrence rate v = 0.0066 years−1,
b = 1.00 , Mmin = 5.50 and Mmax = 7.57. The lifeline–fault crossing point is located at
a distance Z = 30km from the fault closest fault-end (crossing point XL = 0.30 ). The
estimated fault displacement hazard curve at the crossing site is presented in Fig. 5. It is
worth noting that contrary to a hazard curve for ground shaking obtained from PSHA,
which in the vicinity of a dominant fault with a rate of v would tend to saturate at v
for low enough intensities, the PFDHA curve will encounter a plateau at lower rates;
this is due to the CPS, as low-magnitude earthquakes will rarely lead to surface rupture
(Youngs et al. 2003; Moss and Ross 2011; Valentini et al. 2021). Finally, it is stated that
the recurrence rate v is in fact an external multiplier of the seismic hazard calculations
as per Eq. (2) and consequently the increase/decrease of v essentially leads to a propor-
tional upward/downward parallel “shift” of the hazard curve to higher/lower MAFs for
the same displacement.

13
4832 Bulletin of Earthquake Engineering (2023) 21:4821–4849

Fig. 5  Illustrative example of a

fault displacement hazard curve
on the lifeline crossing site

3 Fault scaling relations

Fault scaling relations are developed using data from historic earthquakes and relate fault
characteristics and properties, namely earthquake magnitude, fault length, fault width, rup-
ture area, subsurface rupture length, average fault displacement, maximum fault displace-
ment. These regression equations are typically used within a deterministic seismic hazard
assessment framework to compute the fault displacement offset for the fault crossing at
hand. A comprehensive review of the available scaling relations can be found in Wang
(2018), while a new set of relations especially for strike-slip faults has been very recently
published by Anderson et al. (2021).
Hereinafter, the fault scaling relations of Wells and Coppersmith (1994), Leonard
(2014), and Thingbaijam et al. (2017), abbreviated as WC1994, L2014, and TMG2017,
respectively, are examined. In brief:

• Wells and Coppersmith (1994) systematically examined a database of 421 historical

earthquakes with a magnitude greater than 4.50, from which they selected 244 events
with known parameters of the earthquake rupture. The data were obtained from conti-
nental and intraplate tectonic environments, while subduction zones and oceanic slabs
were excluded. The authors employed the ordinary least-square regression method and
provided a set of independent fault scaling relations for normal, reverse, and strike-slip
fault mechanisms.
• Leonard (2010, 2012, 2014) proposed two scaling relations; one between rupture
width and rupture length, and the other between rupture displacement and rupture
area. By substituting these scaling relations in the definition of seismic moment, the
author developed a series of self-consistent scaling relations between seismic moment
(or moment magnitude) and different rupture characteristics by applying ordinary
least-square regression. The author has combined the data of normal and reverse fault
mechanisms to a single category, that of dip-slip. Additionally, a distinction was made
regarding the tectonic environment (TECT), namely Interplate (INT) and Stable Con-
tinental Region (SCR). It is noted that the Leonard (2014) relations are currently also
used in the development of the active faults sources as part of the 2020 update of the
European Seismic Hazard Model [ESHM20 (Danciu et al. 2021)].
• Thingbaijam et al. (2017) published a set of relations following the typical fault mecha-
nism characterization as normal, reverse, and strike-slip like (Wells and Coppersmith
1994) and differently from Leonard (2014) who analyzed normal and reverse together.
The database did not include SCR events but included subduction interface events in a

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4833

separate set of relations. The authors employed general orthogonal regression, which
allows for inverting the variables, and developed relations that are purely empirical, i.e.,
no prior scaling model was assumed to fit the relation coefficients, similarly to Wells
and Coppersmith (1994) and differently from Leonard (2010, 2012, 2014).

Fault scaling relations are used explicitly within the PFDHA calculations at the follow-
ing steps:

• Estimation of the minimum rupture length ( RLmin) under consideration via Eq. (12)
considering the minimum earthquake magnitude ( Mmin).
• Computation of the joint probability f (RL, ADD|M) [Eqs. (9) and (10)]: (1) where the
median values of rupture length ( RL ) and average subsurface fault displacement ( ADD)
given an earthquake magnitude are computed via Eqs. (12) and (13), respectively, and
(2) RL and ADD are sampled via Eq. (15).
• Computation of the maximum earthquake magnitude ( Mmax ), which is required for the
Gutenberg-Richter Law [Eq. (3)] via Eq. (20).

Thus, it is deemed necessary to examine the effect of selecting one over another set of
relations. The (subsurface) rupture length may be estimated as a function of earthquake
magnitude using the relations listed in Table 2 for Eq. (12). The comparison is presented
in Fig. 6 for normal, reverse, and strike-slip fault mechanisms within a range of earthquake

Table 2  Empirical fault scaling relations for Eq. (12): RL ∼ g(M)

Ref Median value ( RL in km) TECT Parameters a1 , 𝛽1 and standard deviation 𝜎 per fault mecha-
nism
Normal Reverse Strike-slip

WC1994 log10 (RL) = a1 + 𝛽1 M INT a1 = −1.88 a1 = −2.42 a1 = −2.57

𝛽1 = 0.50 𝛽1 = 0.58 𝛽1 = 0.62
𝜎 = 0.170 𝜎 = 0.160 𝜎 = 0.150
4.17 3.4 ≤ LF ≤ 40
{
L2014 log10 (RL) = (M − a1 )∕𝛽1 INT a1 = 4.24 a1 = 4.24
a1 =
5.23 LF > 40

1.667 3.4 ≤ LF ≤ 40
{
𝛽1 = 1.667 𝛽1 = 1.667
𝛽1 =
1.000 LF > 40

0.174 3.4 ≤ LF ≤ 40
{
𝜎 = 0.276 𝜎 = 0.276
𝜎=
0.390 LF > 40

4.25 1.6 ≤ LF ≤ 60
{
L2014 log10 (RL) = (M − a1 )∕𝛽1 SCR a1 = 4.32 a1 = 4.32
a1 =
5.43 LF > 60

1.667 1.6 ≤ LF ≤ 60
{
𝛽1 = 1.667 𝛽1 = 1.667
𝛽1 =
1.000 LF > 60

0.108 1.6 ≤ LF ≤ 60
{
𝜎 = 0.117 𝜎 = 0.117
𝜎=
0.185 LF > 60
TMG2017 log10 (RL) = a1 + 𝛽1 M INT a1 = −1.722 a1 = −2.693 a1 = −2.943
𝛽1 = 0.485 𝛽1 = 0.614 𝛽1 = 0.681
𝜎 = 0.128 𝜎 = 0.083 𝜎 = 0.151

13
4834 Bulletin of Earthquake Engineering (2023) 21:4821–4849

Fig. 6  Median estimate of rupture length ( RL ) given earthquake magnitude ( M ) using alternative empirical
fault scaling relations

magnitude between 5.50 and 8.50, where non-negligible variability is observed, especially
for lower magnitude earthquakes occurring at normal faults and for higher magnitudes at
strike-slip faults. As a remark, WC1994 and TMG2017 directly provide the standard devia-
tion 𝜎 of the output variable, while L2014 provides it for the intercept a1 of the relation
( )
M = a1 + 𝛽1 log10 (RL). With parameter 𝛽1 being a constant, this translates to 𝜎(M) = 𝜎 a1 .
In this application, the relation
( has)been(inverted
) to obtain Eq. (12) with the corresponding
standard deviation is 𝜎 log10 (RL) = 𝜎 a1 ∕𝛽1 . It is noted that Leonard (2014) intended
the equation (Y = a + 𝛽X ) to be invertible since it is not fitted in the classical sense. The
multiplicative constant 𝛽 is actually fixed, coming from a physical model, while optimiza-
tion is performed only on the intercept a. Thus, the linear part of the model is essentially
treated as deterministic, allowing the inversion to be mathematically correct.
The average subsurface fault displacement ( ADD) may be computed using the earth-
quake magnitude ( M ) [Eq. (13)] using the scaling relations listed in Table 3 for all fault
mechanisms and tectonic environments. Please note that the relation ADS ∼ h(M) provided
by Wells and Coppersmith (1994) for reverse fault mechanisms is not significant at a 95%
probability level, and thus, it is replaced by the relation provided by Moss and Ross (2011),
abbreviated as MR2011.
The obtained ADS values by Moss and Ross (2011), are also transformed to ADD
values. Moreover, regarding the standard deviations, WC1994 ( and ) MR2011 pro-
vide the relation ADS ∼ p(M) with standard deviation 𝜎 log10 (ADS) and thus the
required
( standard
) deviation
( for the subsurface
) ( average fault displacement
) ( is obtained) as
𝜎 log10 (ADD) = 𝜎 log10 (1.32ADS) = 𝜎 log10 (ADS) + log10 (1.32) = 𝜎 log10 (ADS) .
Then, L2014 provides the 𝜎 of parameter a2 for the relation M = a(2 +) 𝛽2 log10 (ADD).
With parameter 𝛽2 being a constant, this translates to 𝜎(M) = 𝜎 a2 . The relation
has
( been inverted
) to
( obtain
) Eq. (13) and thus the corresponding standard deviation is
𝜎 log10 (ADD) = 𝜎 a2 ∕𝛽2 . Finally, TMG2017 directly provide the output standard devi-
ation for the relation ADD ∼ h(M). The comparison of the different median (mean log)
estimates of ADD based on the earthquake magnitude are illustrated in Fig. 7 for normal,
reverse, and strike-slip fault mechanisms within a range of earthquake magnitude values
between 5.50 and 8.50. It is observed that roughly in all cases, the SCR case of Leonard
(2014) yields the higher average fault displacement values, while Thingbaijam et al. (2017)
expressions yield roughly lower average displacement values. The INT case of Leonard
(2014) is around “the middle” compared to the other sets of relations.

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4835

Table 3  Empirical fault scaling relations for Eq. (13): ADD ∼ h(M)
Ref Median value ( ADD in m) TECT Parameters a2 , 𝛽2 and standard deviation 𝜎
per fault mechanism
Normal Reverse Strike-slip

WC1994 log10 (ADS) = a2 + 𝛽2 M ADD = 1.32ADS INT a2 = −4.45 – a2 = −6.32

𝛽2 = 0.63 𝛽2 = 0.90
𝜎 = 0.330 𝜎 = 0.280
MR2011 log10 (ADS) = a2 + 𝛽2 M ADD = 1.32ADS INT – a2 = −2.2192 –
𝛽2 = 0.3244
𝜎 = 0.170
L2014 log10 (ADD) = (M − a2 )∕𝛽2 INT a2 = 6.84 a2 = 6.84 a2 = 6.85
𝛽2 = 2.00 𝛽2 = 2.00 𝛽2 = 2.00
𝜎 = 0.303 𝜎 = 0.303 𝜎 = 0.260
L2014 log10 (ADD) = (M − a2 )∕𝛽2 SCR a2 = 6.46 a2 = 6.46 a2 = 6.47
𝛽2 = 2.00 𝛽2 = 2.00 𝛽2 = 2.00
𝜎 = 0.100 𝜎 = 0.100 𝜎 = 0.050
TMG2017 log10 (ADD) = a2 + 𝛽2 M INT a2 = −4.967 a2 = −3.156 a2 = −4.032
𝛽2 = 0.693 𝛽2 = 0.451 𝛽2 = 0.558
𝜎 = 0.195 𝜎 = 0.149 𝜎 = 0.227

Fig. 7  Median estimate of average subsurface displacement ( ADD) given earthquake magnitude ( M ) using
alternative empirical fault scaling relations

The required sampling of ADD given rupture length RL is carried out using the scaling
relations listed in Table 4 after Eq. (15). The pertinent comparison is depicted in Fig. 8,
indicating significant variability in the results. Moreover, regarding the standard deviations,
L2014 and TMG2017 provide the 𝜎 for the relation ADD ∼ r(RL). Conversely, WC1994
provide the relation log10 (ADS) = a3 + 𝛽3 log10 (SRL), where SRL is the surface rupture
length. In such case, both sides of the relation should be transformed to the subsurface
level. Based on the mode of the distribution of ratios of average subsurface displacement
to average surface displacement calculated by Wells and Coppersmith (1994), the transfor-
mation ADD = 1.32ADS is used for the average displacement. Then, the transformation
SRL = 0.75RL is used for the rupture length, as Wells and Coppersmith (1994) found that,

13
4836 Bulletin of Earthquake Engineering (2023) 21:4821–4849

Table 4  Empirical fault scaling relations for Eq. (15): ADD ∼ r(RL)
Ref Median value ( ADD in TECT Parameters a3 , 𝛽3 and standard deviation 𝜎 per fault mecha-
m and RL in km) nism
Normal Reverse Strike-slip

WC1994 log10 (ADD) INT a3 = −2.024 a3 = −0.518 a3 = −1.709

= a3 + 𝛽3 log10 (RL)* 𝛽3 = 1.24 𝛽3 = 0.31 𝛽3 = 1.04
𝜎 = 0.370 𝜎 = 0.400 𝜎 = 0.320
−3.844 3.4 ≤ LF ≤ 40
{
L2014 log10 (ADD) = INT a3 = −3.799 a3 = −3.799
a3 =
a3 + 𝛽3 log10 (1000RL) 0.833 LF > 40
−2.310 3.4 ≤ LF ≤ 40
{
𝛽3 = 0.833 𝛽3 = 0.833
𝛽3 =
0.500 LF > 40

0.450 3.4 ≤ LF ≤ 40
{
𝜎 = 0.530 𝜎 = 0.530
𝜎=
0.455 LF > 40

−3.615 1.6 ≤ LF ≤ 60
{
L2014 log10 (ADD) = SCR a3 = −3.572 a3 = −3.572
a3 =
a3 + 𝛽3 log10 (1000RL) 0.8300 LF > 60

−2.022 1.6 ≤ LF ≤ 60
{
𝛽3 = 0.833 𝛽3 = 0.833
𝛽3 =
0.500 LF > 60

0.190 1.60 ≤ LF ≤ 60
{
𝜎 = 0.200 𝜎 = 0.200
𝜎=
0.190 LF > 60
TMG2017 log10 (ADD) = INT a3 = −2.302 a3 = −1.456 a3 = −1.473
a3 + 𝛽3 log10 (RL) 𝛽3 = 1.302 𝛽4 = 0.975 𝛽3 = 0.789
𝜎 = 0.252 𝜎 = 0.132 𝜎 = 0.276
*Coefficients and standard deviation after Eqs. (21) and (22), respectively

Fig. 8  Median estimate of average subsurface displacement ( ADD) given rupture length ( RL ) using alterna-
tive empirical fault scaling relations

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4837

on average, the surface rupture length equals 75% of the subsurface rupture. Owing to the
above, the median log10 (ADD) value may be estimated as:
log10 (ADS) = a3 + 𝛽3 log10 (SRL) →
→ log10 (ADD∕1.32) = a3 + 𝛽3 log10 (0.75RL) →
(21)
→ log10 (ADD) = a∗3 + 𝛽3 log10 (RL)
with a∗3 = − log10 (0.76) + a3 + 𝛽3 log10 (0.75)

The corresponding standard deviation can be estimated as:

( ) ( ) ( )
𝜎 log10 (ADD) = 𝜎 log10 (1.32ADS) = 𝜎 log10 (ADS) + log10 (1.32) →
( ) ( ) (22)
→ 𝜎 log10 (ADD) = 𝜎 log10 (ADS)

as the variance of a random variable plus a constant is equal to the variance of the former.
The maximum earthquake magnitude ( Mmax ) considered in the Gutenberg-Richter Law
[Eq. (3)] is computed using the scaling relations listed in Table 5 after Eq. (20).

Table 5  Empirical fault scaling relations for Eq. (20): Mmax ∼ w(LF)
Ref Median value ( LF in km) TECT Parameters a4 , 𝛽4 per fault mechanism
Normal Reverse Strike-slip

WC1994 Mmax = a4 + 𝛽4 log10 (LF) INT a4 = 4.34 a4 = 4.49 a4 = 4.33

𝛽4 = 1.54 𝛽4 = 1.49 𝛽4 = 1.49
4.17 3.4 ≤ LF ≤ 40
{
L2014 Mmax = a4 + 𝛽4 log10 (LF) INT a4 = 4.24 a4 = 4.24
a4 =
5.23 LF > 40

1.667 3.4 ≤ LF ≤ 40
{
𝛽4 = 1.667 𝛽4 = 1.667
𝛽4 =
1.000 LF > 40

4.25 1.6 ≤ LF ≤ 60
{
L2014 Mmax = a4 + 𝛽4 log10 (LF) SCR a4 = 4.32 a4 = 4.32
a4 =
5.43 LF > 60

1.667 1.6 ≤ LF ≤ 60
{
𝛽4 = 1.667 𝛽4 = 1.667
𝛽4 =
1.000 LF > 60
[ ]
TMG2017 Mmax = log10 (LF) − a4 ∕𝛽4 INT a4 = −1.722 a4 = −2.693 a4 = −2.943
𝛽4 = 0.485 𝛽4 = 0.614 𝛽4 = 0.681

13
4838 Bulletin of Earthquake Engineering (2023) 21:4821–4849

Fig. 9  Fault displacement hazard curves on lifeline crossing site for normal, reverse, and strike-slip fault
mechanism using alternative fault scaling relations

At this point, it is worth recalling that alternative fault scaling relations are examined to
capture the pertinent uncertainty on selecting alternative sets. To quantify this effect, the
baseline parameters of the case study of Sect. 2 are adopted. The obtained fault displace-
ment hazard curves are illustrated in Fig. 9 for all fault mechanisms and considering alter-
native sets of empirical fault scaling relations. It is observed that adopting the set of Thing-
baijam et al. (2017) leads to lower MAF values, while adopting the Leonard (2014) fault
scaling relations results in higher MAF values, especially for the interplate tectonic envi-
ronment. The hazard curve when the Wells and Coppersmith (1994) relations are adopted
are roughly in the middle of the other cases. In any case, the differences in terms of MAF
values increase as the fault displacement increases. Concluding, the variability of the haz-
ard curves highlights the fact that the selection of a set of fault scaling relations should be
treated as an important epistemic uncertainty.

4 Sensitivity analysis

The proposed approach of PFDHA is intended for engineering applications, thus being appli-
cable to a large number of essentially different faults at a regional level. Engineers must there-
fore become familiar with the effects of various parameters and aspects of the hazard calcu-
lations. To that effect, the impact of the main model parameters, namely b-value (Sect. 4.1.1)
and maximum earthquake magnitude (Sect. 4.1.2), which shape the hazard curve is first investi-
gated. Then, the effects of fault length (Sect. 4.1.3), fault mechanism (Sect. 4.1.4), and location
of the crossing site (Sect. 4.1.5) on the hazard level are studied. Finally, the critical aspects of
the calculation process, namely the correlation of RL and ADD (Sect. 4.2.1), the conditional
probability of the rupture reaching the surface (Sect. 4.2.2), and the crossing point (Sect. 4.2.3)
are discussed to provide insight into their significance in comparison to ground shaking haz-
ard calculations that engineers are more familiar with. The sensitivity analysis presented here
is not intended to assess the impact of all potential sources of uncertainty, but rather to demon-
strate the effect of different parameters. When a case-specific detailed fault displacement hazard
analysis is required, additional parameters would be considered by the experienced seismologist
conducting the analysis. It is noted that the sensitivity analysis is also structure-independent and
consequently there is no judgement on the relevant importance of each parameter on the lifeline
structural response. Ultimately, this is a designated objective of a case- and site-specific study.
The sensitivity analysis was conducted for the lifeline–fault crossing case study shown
in Sect. 2. The baseline parameters are summarized in Table 6 for the sake of clarity.

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4839

Table 6  Baseline parameters of Parameter Value

the lifeline–fault crossing case
study
Fault mechanism Normal
Tectonic environment Interplate
Fault recurrence rate v = 0.0066years−1
Gutenberg-Richter b-value b = 1.00
Fault length LF = 100km
Minimum earthquake magnitude Mmin = 5.50
Maximum earthquake magnitude Mmax = 7.57
Crossing point XL = 0.30

Throughout the sensitivity analysis, selected parameters are changed to demonstrate

their effect on the hazard calculations. Due to the fact that a single fault is being exam-
ined, the slip rate must be constant. Therefore, a constant value of the fault slip rate
along strike of S = 0.5mm∕year is adopted, which falls within the range of the most
frequently encountered values in the seismically active faults of EFSM20 (Basili et al.
2022). Noteworthily, at 0.5 mm/year an active fault would have accumulated ~ 6 m
total offset over the Holocene (~ 12,000 years), or ~ 60 m over the Late Pleistocene
(125,000 years), thereby creating geomorphic features that could be possibly recognized
in the field. The recurrence rate has to be rederived after any adjustment to the parame-
ters in the sensitivity analysis to ensure consistency with the constant slip rate along the
fault, thus offering a common basis for comparing the different changes in parameters.
The recurrence rate v is obtained after Youngs and Coppersmith (1986):
( ( ))
bvM0 u exp −𝛽 Mmax − Mmin
𝜇AS = [ ( ( ))] (23)
(c − b) 1 − exp −𝛽 Mmax − Mmin

where 𝜇 = 3 ⋅ 1011 dyne∕cm2 is the fault rigidity after Bilek and Lay (1999), A = LF ⋅ W
is the fault area with the fault width W = 20km assumed to be constant, S is the fault slip
rate, b is the value of the Gutenberg-Richter Law, parameter 𝛽 is defined as 𝛽 = bln10,
M0 u is the maximum moment corresponding to the maximum magnitude Mmax of the fault.
The maximum moment is assumed to be deterministic and estimated as M0 u = 10d+cMmax
with parameters c = 1.5 and d = 16.1 (Hanks and Kanamori 1979). The first part of the
sensitivity analysis (Sect. 4.1) deals with the input parameters, namely the b-value of the
Gutenberg-Richter law, the maximum earthquake magnitude, the fault length, the fault
mechanism, and the location of the crossing site, while in the second part (Sect. 4.2) criti-
cal aspects of the hazard calculations are discussed.

4.1 Input parameters

4.1.1 Slope of the Gutenberg‑Richter Law (the b‑value)

The effect of the b-value to the fault displacement hazard estimates is illustrated in
Fig. 10, where it can be observed that for low fault displacement values the b-value has
a relatively low effect, while for higher displacement values, the effect tends to be insig-
nificant. The increase of the b-value increases the ratio of low versus high magnitude
events (Fig. 2).

13
4840 Bulletin of Earthquake Engineering (2023) 21:4821–4849

Fig. 10  Fault displacement hazard curves on lifeline crossing site considering different b-values of the
Gutenberg-Richter Law for normal, reverse, and strike-slip fault mechanisms

Table 7  Empirical fault scaling relation Mmax ∼ w(LF) [Eq. 20] and Table 5] of Leonard (2014): S(a4 ) is
the one standard deviation range of a4, i.e. the mean plus/minus one standard deviation
TECT S(a4 )[one standard deviation range of a4] per fault mechanism
Normal Reverse Strike-slip

3.87 − 4.45 3.4 ≤ LF ≤ 40

( ) ( ) {
INT S a4 = 3.81 − 4.73 S a4 = 3.81 − 4.73 ( )
S a4 =
4.84 − 5.62 LF > 40

4.07 − 4.43 1.6 ≤ LF ≤ 60

( ) ( ) {
SCR S a4 = 4.12 − 4.51 S a4 = 4.12 − 4.51 ( )
S a4 =
5.25 − 5.62 LF > 60

4.1.2 Maximum earthquake magnitude

Alternative values of maximum earthquake magnitude ( Mmax ) are typically considered

in any seismic hazard study. The Mmax is estimated by empirical fault geometry-mag-
nitude scaling equations. The median value, a higher or a lower value can be adopted,
using the range of parameter a4 of Table 5 after Leonard (2014) as per Table 7. The cor-
responding hazard curves are illustrated in Fig. 11, indicating a “parallel” displacement
of the hazard curves to lower MAF values for lower values of Mmax in particular for
low fault displacement values. A minor effect of the maximum earthquake magnitude is
observed in the hazard level for higher values of fault displacement. Due to the constant
slip rate, the released energy is constant for all cases. Thus, decreasing Mmax , the over-
all rate v is increased as it concerns smaller events [Eq. (23)]. Therefore, the epistemic
uncertainty stemming from the calculation of Mmax via empirical fault scaling relations
should be considered.

4.1.3 Fault length

The fault length is a parameter directly related to other important fault parameters such
as fault width, fault area, and maximum earthquake magnitude. For example, strike-
slip faults may extend up to hundreds of kilometers in length, while being shallow in
terms of width, contrary to reverse faults. Earthquake magnitude has also been found to

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4841

Fig. 11  Fault displacement hazard curves on lifeline crossing site considering different maximum earth-
quake magnitude for normal, reverse, and strike-slip fault mechanisms, where the low Mmax is computed
using the lower a4 value and the high Mmax using the higher a4 value of Table 7

be a linear function of the log10 LF . Herein, the LF under examination is ranging from
50 to 150 km and the effect of fault length on the displacement hazard curve is shown
in Fig. 12 for all fault mechanisms. Generally, increased fault lengths lead to higher
MAFs. This mild trend appears more clearly in Fig. 13 for the normal fault, where small
MAF increments for given fault displacements are observed even when the fault length
doubles or triples.

Fig. 12  Fault displacement hazard curves on lifeline crossing site for varying LF

Fig. 13  Mean annual frequency

of exceeding characteristic Δ val-
ues on lifeline crossing site of a
normal fault crossing considering
different LF

13
4842 Bulletin of Earthquake Engineering (2023) 21:4821–4849

Fig. 14  Fault displacement

hazard curves on lifeline cross-
ing site with respect to the fault
mechanism

4.1.4 Fault mechanism

A comparison of the hazard curves for different fault mechanisms, namely normal, reverse,
and strike-slip is depicted in Fig. 14. A notable lower hazard level is observed for reverse
mechanism compared to normal and strike-slip, which is mainly attributed to the condi-
tional probability of slip [see Eqs. (18) and (19)] that mainly differentiates this mechanism
from the normal and strike-slip, as explicitly outlined in Fig. 4.

4.1.5 Location of the crossing site

The crossing site location on the fault trace, as thoroughly discussed in previous sec-
tions, is critical to the fault displacement hazard level on the crossing site. Moss and
Ross (2011) and Youngs et al. (2003) have normalized fault displacement data with the
fault average surface displacement ( ADS ) to estimate the conditional probability that
the offset at a specific point will exceed a predefined value. The distribution of ratio
𝛿∕ADS is expressed as a function of the crossing point XL with function f (𝛿∕ADS|XL)
being symmetric about a maximum value at x∕RL = 0.50 . Accordingly, as shown in
Fig. 15 and regardless of the fault mechanism, the hazard curves for alternative crossing
points ( XL ) are parallel and the highest MAF values are obtained for XL = 0.50 because
in this case the crossing site is located at the middle of the fault trace length. The

Fig. 15  Fault displacement hazard curves on lifeline crossing site considering different XL (distance of the
crossing site to the closest fault-end) for normal, reverse, and strike-slip fault mechanisms

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4843

practical outcome for consideration within a route selection process of a lifeline (Seel
et al. 2014; Hamid-Mosaku et al. 2020) is that the engineer should expect higher fault
displacements if the crossing site is decided to be closer to the middle of the mapped
fault trace length.

4.1.6 Discussion on input parameters

To offer an overview of the sensitivity analysis, focusing on the input parameter to the
hazard calculations, the effects of b-value, maximum earthquake magnitude ( Mmax ), fault
length ( LF ), and crossing point ( XL ) are examined in terms of the obtained MAF for
predefined values of fault displacement with respect to the baseline case (see Table 6).
Figure 16 shows the variation of MAF with the baseline case being reported in the verti-
cal line.
The primary observation is that the location of the crossing site on the fault trace affects
significantly the resulting MAFs, regardless of the fault displacement value, as concluded
in Sect. 4.1.5. This is of primary importance for engineers during the route selection

Fig. 16  Variation of MAF with respect to the baseline case (see Table 6) for predefined values of fault dis-
placement regarding the input parameters b-value, Mmax , LF , and XL . Along the vertical axis, values of the
parameter investigated in each chart are increased

13
4844 Bulletin of Earthquake Engineering (2023) 21:4821–4849

procedure, and requires some fault-specific knowledge to be accurately accounted for.

Then, regarding the two main seismological input parameters, namely the b-value and
Mmax , the influence of the latter on the MAF is significantly higher, highlighting the need
for appropriate treatment within a logic tree formulation to account for its epistemic uncer-
tainty. Finally, as observed in both Figs. 12 and 13, the higher the fault displacement value
examined, the lower the effect of LF is.

4.2 Aspects of hazard calculations

4.2.1 Correlation of RL and ADD

( )
The term P RLj , ADDt |mi in Eq. (10) is introduced to account for the joint probability
of RL and ADD conditioned the magnitude. In particular, the estimation of the covari-
ance matrix [𝜌] in Eq. (14) requires the correlation coefficient 𝜌log10 RL,log10 ADD of the resid-
uals of RL and ADD conditioned on the earthquake magnitude. Unfortunately, this value
is not provided in pertinent studies (e.g., Leonard 2014; Thingbaijam et al. 2017; Wells
and Coppersmith 1994). Therefore, the effect of this factor is examined in terms of the
resulting hazard curve by considering no correlation ( 𝜌log10 RL,log10 ADD = 0 ), partial posi-
tive and negative correlation (± 50%) ( 𝜌log10 RL,log10 ADD = ± 0.50 ), and full negative and
positive (± 100%) correlation ( 𝜌log10 RL,log10 ADD = ±1.00 ). At the same time, though, the
value of this correlation coefficient cannot be accurately determined without recourse to
the data employed for fitting each model. As the sensitivity analysis presented in Fig. 17
shows, the increase in the correlation leads to higher MAF values. Therefore, in the
absence of any data, the 0% correlation option is adopted. This choice can be seen as
reasonable in light of a conservative design approach; it can be empirically suggested
that a negative correlation may be more appropriate, as a certain amount of energy (or
given M ) may be released over a large length with a small slip or over a shorter length
with a larger slip.

Fig. 17  Fault displacement

hazard curves on lifeline crossing
site by considering alternative
correlations for the residuals of
RL and ADD given the mag-
nitude (correlation coefficient
𝜌log10 RL,log10 ADD) for the term
( )
P RLj , ADDt |mi of Eq. (10)

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4845

Fig. 18  Fault displacement

hazard curves on lifeline
crossing site with consider-
ing [ P(Slip|m) < 1] or not
[ P(Slip|m) = 1] the probability of
the rupture reaching the surface

4.2.2 Conditional probability of the rupture reaching the surface

The Conditional Probability of Slip (CPS) is calculated using the empirical relation of
Eqs. (18) and (19) for each fault mechanism. These relations are logistic regression mod-
els produced by the analysis of worldwide data (Wells and Coppersmith 1993; Moss and
Ross 2011) to consider whether the rupture reaches the surface. The effect of the CPS
on the fault displacement hazard curve is depicted in Fig. 18, where the hazard curve is
obtained with and without considering the probability of the rupture reaching the surface.
As expected, considering the CPS yields lower MAF values for the entire range of fault
displacement values. It should be reminded that the CPS is an integral part of PFDHA
and stands as a substantial difference from typical PSHA calculations because in the lat-
ter all nearby seismic sources are considered to contribute to the hazard at the site under
examination.

4.2.3 Effect of crossing point

The effect of considering the crossing point is examined because contrary to what is antici-
pated in a PSHA where all potential ruptures contribute to the hazard integration at a given
site, this is not the case for lifeline crossings. This effect is illustrated in Fig. 19 where the
hazard curves are plotted with and without considering the location of the crossing point rela-
tive to each RL, in other words, by considering intercepting ruptures versus all ruptures. In
the first case, if the lifeline is not intercepted by the ruptured segment, then 𝜆Δ (𝛿) = 0 for that
particular RL. In the second case, the hazard on the crossing site is computed assuming all
RLs contain the crossing site regardless of the actual location of the RL on the fault trace, i.e.,
in any case 𝜆Δ (𝛿) > 0.

Fig. 19  Fault displacement

hazard curves on lifeline crossing
site by taking into account or not
the crossing point

13
4846 Bulletin of Earthquake Engineering (2023) 21:4821–4849

5 Conclusions

Safeguarding the integrity and ensuring the seismic resilience of lifelines that cross
active tectonic faults requires the reliable estimation of the fault displacement. To
achieve this, one could employ an existing dataset of mapped faults and a set of empiri-
cal fault scaling relations. Even though it is a relatively easy computation, the resulting
displacement comes with an unknown level of safety. On the other hand, the Probabil-
istic Fault Displacement Hazard Analysis (PFDHA) could be used to achieve a balance
between safety and cost-effectiveness while adhering to the performance-based engi-
neering framework. However, this requires an extensive effort in terms of input data and
calculations, as well as significant experience. In order to find a middle solution and
towards formulating a methodology for computing the fault displacement that can be
incorporated in a design code, we present in this study a structure-independent simplified
approach of PFDHA for engineering applications. Based on the established PFDHA of
Youngs et al. (2003) we focus on the lifeline–fault crossing and we introduce appropri-
ate simplifications and assumptions, allowing our approach to be applicable to different
faults at a regional level. In any case, the approach is not intended to replace a case- and
site-specific PFDHA accounting for all relevant uncertainties and specialized data if such
an analysis is required by the lifeline owner/operator and/or the regulatory authorities.
We use the presented approach in a next study to derive a set of simplified hazard-con-
sistent and code-compatible expressions, suitable for use in the lifeline route selection,
for the earthquake-resistant preliminary design of lifelines, and in cases where carrying
out a detailed PFDHA may not be possible, for example, when there are no recent sedi-
ments to date past earthquakes.
At first, the calculation steps are presented in detail, based on the work of Melissianos
et al. (2017a) and explanations are provided on the hazard calculation aspects. A thorough
discussion is then offered on the available alternative empirical fault scaling relations that are
required at certain critical steps of the hazard computations and how they affect the results.
Moreover, a sensitivity analysis on the influence of input parameters, such as the Gutenberg-
Richter b-value, maximum earthquake magnitude, fault length, and the location of the cross-
ing site provides engineers with insight into the effect of parameters and aspects on the haz-
ard level calculations. Generally, higher mean annual frequency values of exceeding low fault
displacement values were estimated for short faults, while vice versa higher mean annual
frequencies were obtained for high fault displacement values in case of longer faults. The
location of the crossing site on the fault trace was identified as a critical parameter because
the closer to the middle of the fault the site is, the higher the displacement hazard. The Mmax
(maximum earthquake magnitude) is strongly correlated with the fault dimensions, and thus
its effect on the fault displacement hazard curve is rather minimum. Additionally, the empiri-
cal fault-magnitude scaling equations that are incorporated in the hazard calculations were
evaluated, indicating a strong variability in results due to the different background of each
set of relations. Finally, the b-value plays a moderate role in terms that the higher the value,
the lower the hazard because small magnitudes have a low impact on the fault displacement
hazard.
In conclusion, the study offers a comprehensive exercise on how to calculate the fault
displacement hazard for lifelines at fault crossings, also considering the uncertainties asso-
ciated with seismological parameters. It provides practitioners with insights by demonstrat-
ing the impact of key parameters on both the hazard level and the shape of the hazard

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4847

curve. In addition, it serves as a background document for future reference, detailing the
effects of the simplifications and assumptions made, which are pertinent and relevant for
the engineering community.
Acknowledgements The authors acknowledge the comments made by the three anonymous reviewers who
helped us improve the overall quality of this manuscript.

Author contributions VE Melissianos: Formal analysis and investigation, Conceptualization, Methodology,

Writing—original draft preparation, Writing—review and editing. L Danciu: Conceptualization, Method-
ology, Writing—review and editing. D Vamvatsikos: Conceptualization, Methodology, Supervision, Writ-
ing—review and editing. R Basili: Methodology, Writing—review and editing.

Funding Open access funding provided by HEAL-Link Greece. The current work has been undertaken
as part of the Horizon 2020 Seismology and Earthquake Engineering Research Infrastructure Alliance in
Europe (SERA, Grant Agreement No. 730900). The first and the third author have received partial fund-
ing from the European Union’s Horizon 2020 research and innovation programme “METIS-Seismic Risk
Assessment for Nuclear Safety” under Grant Agreement No. 945121. Also, the financial support provided
by the Hellenic Foundation for Research and Innovation (H.F.R.I.) under the “2nd Call for H.F.R.I. Research
Projects to support Faculty Members & Researchers”, Project "TwinCity—Climate-Aware Risk and Resil-
ience Assessment of Urban Areas under Multiple Environmental Stressors via Multi-Tiered Digital City
Twinning ", (Number: 2515) is gratefully acknowledged.

Data availability The data generated during the current study are available from the corresponding author
upon reasonable request.

Declarations
Conflict of interest The authors declared no potential conflicts of interest with respect to the research, author-
ship, and/or publication of this article.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License,
which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as
you give appropriate credit to the original author(s) and the source, provide a link to the Creative Com-
mons licence, and indicate if changes were made. The images or other third party material in this article
are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the
material. If material is not included in the article’s Creative Commons licence and your intended use is
not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission
directly from the copyright holder. To view a copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​
ses/​by/4.​0/.

References
Anastasopoulos I, Gerolymos N, Drosos V et al (2008) Behaviour of deep immersed tunnel under combined
normal fault rupture deformation and subsequent seismic shaking. Bull Earthq Eng 6:213–239. https://​
doi.​org/​10.​1007/​s10518-​007-​9055-0
Anderson JG, Biasi GP, Angster S, Wesnousky SG (2021) Improved scaling relationships for seismic
moment and average slip of strike-slip earthquakes incorporating fault-slip rate, fault width, and stress
drop. Bull Seismol Soc Am 111(5):2379–2392. https://​doi.​org/​10.​1785/​01202​10113
Angell MM, Hanson K, Swan FH et al (2003) Probabilistic Fault Displacement Hazard Assessment for
Flowlines and Export Pipelines, Mad Dog and Atlantis field developments, Deepwater gulf of Mex-
ico. In: Offshore Technology Conference. OTC, Houston, TX, USA, p OTC 15402. https://​doi.​org/​10.​
4043/​15402-​MS
Basili R, Danciu L, Beauval C et al (2022) European Fault-Source Model 2020 (EFSM20): online data on
fault geometry and activity parameters. Rome. https://​doi.​org/​10.​13127/​efsm20
Bilek SL, Lay T (1999) Rigidity variations with depth along interplate megathrust faults in subduction
zones. Nature 400:443–446. https://​doi.​org/​10.​1038/​22739

13
4848 Bulletin of Earthquake Engineering (2023) 21:4821–4849

Bommer JJ (2002) Deterministic vs. probabilistic seismic hazard assessment: an exaggerated and obstruc-
tive dichotomy. J Earthquake Eng 6:43–73. https://​doi.​org/​10.​1080/​13632​46020​93504​32
Chen R, Petersen MD (2011) Probabilistic fault displacement hazards for the Southern San Andreas fault
using scenarios and empirical slips. Earthq Spectra 27(2):293–313. https://​doi.​org/​10.​1193/1.​35742​26
Chen R, Petersen MD (2019) Improved implementation of rupture location uncertainty in fault displacement
hazard assessment. Bull Seismol Soc Am 109(5):2132–2137. https://​doi.​org/​10.​1785/​01201​80305
Cheng Y, Akkar S (2017) Probabilistic permanent fault displacement hazard via Monte Carlo simulation
and its consideration for the probabilistic risk assessment of buried continuous steel pipelines. Earthq
Eng Struct Dyn 46:605–620. https://​doi.​org/​10.​1002/​eqe.​2805
Coppersmith KJ, Youngs RR (2000) Data needs for probabilistic fault displacement hazard analysis. J Geo-
dyn 29:329–343. https://​doi.​org/​10.​1016/​S0264-​3707(99)​00047-2
Cornell CA (1968) Engineering seismic risk analysis. Bull Seismol Soc Am 58(5):1583–1606. https://​doi.​
org/​10.​1785/​BSSA0​58005​1583
Cornell CA, Krawinkler H (2000) Progress and challenges in seismic performance assessment. PEER
Center News 3:1–4
Danciu L, Nandan S, Reyes C et al (2021) The 2020 update of the European seismic Hazard Model -
ESHM20: Model Overview. EFEHR Technical Report 001 v1.0.0. Zurich, Switzerland. https://​doi.​org/​
10.​12686/​a15
de Santis A, Cianchini G, Favali P et al (2011) The Gutenberg-Richter law and entropy of earthquakes:
two case studies in central Italy. Bull Seismol Soc Am 101(3):1386–1395. https://​doi.​org/​10.​1785/​
01200​90390
Ghadimi Chermahini A, Tahghighi H (2019) Numerical finite element analysis of underground tunnel
crossing an active reverse fault: a case study on the Sabzkouh segmental tunnel. Geomech Geoeng
14:155–166. https://​doi.​org/​10.​1080/​17486​025.​2019.​15733​23
Girgin S, Krausmann E (2016) Historical analysis of U.S. onshore hazardous liquid pipeline accidents
triggered by natural hazards. J Loss Prev Process Ind 40:578–590. https://​doi.​org/​10.​1016/j.​jlp.​
2016.​02.​008
Gürpinar A, Serva L, Livio F, Rizzo PC (2017) Earthquake-induced crustal deformation and con-
sequences for fault displacement hazard analysis of nuclear power plants. Nuclear Eng Design
311:69–85. https://​doi.​org/​10.​1016/j.​nucen​gdes.​2016.​11.​007
Gutenberg R, Richter CF (1944) Frequency of earthquakes in California. Bull Seismol Soc Am
34(4):185–188. https://​doi.​org/​10.​1785/​BSSA0​34004​0185
Hamid-Mosaku IA, Oguntade OF, Ifeanyi VI et al (2020) Evolving a comprehensive geomatics
multi-criteria evaluation index model for optimal pipeline route selection. Struct Infrastruct Eng
16:1382–1396. https://​doi.​org/​10.​1080/​15732​479.​2020.​17124​35
Hanks TC, Kanamori H (1979) A moment magnitude scale. J Geophys Res 84:2348–2350. https://​doi.​
org/​10.​1029/​JB084​iB05p​02348
Honegger DG, Nyman DJ, Johnson ER et al (2004) Trans-Alaska pipeline system performance in the 2002
Denali fault, Alaska, earthquake. Earthq Spectra 20(3):707–738. https://​doi.​org/​10.​1193/1.​17792​39
International Atomic Energy Agency (2010) Seismic hazards in site evaluation for nuclear installations
SSG-9. Austria, Vienna
Leonard M (2010) Earthquake fault scaling: self-consistent relating of rupture length, width, average
displacement, and moment release. Bull Seismol Soc Am 100(5A):1971–1988. https://​doi.​org/​10.​
1785/​01200​90189
Leonard M (2012) Erratum to Earthquake fault scaling: Self-consistent relating of rupture length, width,
average displacement, and moment release. Bull Seismol Soc Am 102(6):2797. https://​doi.​org/​10.​
1785/​01201​20249
Leonard M (2014) Self-consistent earthquake fault-scaling relations: update and extension to stable conti-
nental strike-slip faults. Bull Seismol Soc Am 104(6):2953–2965. https://​doi.​org/​10.​1785/​01201​40087
Melissianos VE (2022) Onshore buried steel fuel pipelines at fault crossings: a review of critical analy-
sis and design aspects. J Pipeline Syst Eng Pract 13(4):03122002. https://​doi.​org/​10.​1061/​(ASCE)​
PS.​1949-​1204.​00006​61
Melissianos VE, Vamvatsikos D, Gantes CJ (2017a) Performance assessment of buried pipelines at fault
crossings. Earthquake Spectra 33(1):201–218. https://​doi.​org/​10.​1193/​12201​5EQS1​87M
Melissianos VE, Vamvatsikos D, Gantes CJ (2017b) Performance-based assessment of protection meas-
ures for buried pipes at strike-slip fault crossings. Soil Dyn Earthq Eng 101:1–11. https://​doi.​org/​
10.​1016/j.​soild​yn.​2017.​07.​004
Moss RES, Ross ZE (2011) Probabilistic fault displacement hazard analysis for reverse faults. Bull Seis-
mol Soc Am 101(4):1542–1553. https://​doi.​org/​10.​1785/​01201​00248

13
Bulletin of Earthquake Engineering (2023) 21:4821–4849 4849

O’Rourke MJ, Liu JX (2012) Seismic design of buried and offshore pipelines. Monograph MCEER-
12-MN04. Multidisciplinary Center for Earthquake Engineering Research, Buffalo, NY, USA
Page MT, Felzer K (2015) Southern San Andreas fault seismicity is consistent with the Gutenberg-Rich-
ter magnitude–frequency distribution. Bull Seismol Soc Am 105(4):2070–2080. https://​doi.​org/​10.​
1785/​01201​40340
Parsons T, Geist EL (2009) Is there a basis for preferring characteristic earthquakes over a Gutenberg-
Richter distribution in probabilistic earthquake forecasting? Bull Seismol Soc Am 99(3):2012–
2019. https://​doi.​org/​10.​1785/​01200​80069
Petersen MD, Dawson TE, Chen R et al (2011) Fault displacement hazard for strike-slip faults. Bull
Seismol Soc Am 101(2):805–825. https://​doi.​org/​10.​1785/​01201​00035
Rondoyanni T, Lykoudi E, Triantafyllou A et al (2013) Active faults affecting linear engineering projects:
examples from Greece. Geotech Geol Eng 31:1151–1170. https://​doi.​org/​10.​1007/​s10706-​013-​9641-7
Seel K, Dragan M, Coulombe-Pontbriand M et al (2014) A Spatial Multi-Criteria Analysis Process to
Optimize and Better Defend the Pipeline Route Selection Process. In: Proceedings of the 10th
International Pipeline Conference. AmericanSociety of Mechanical Engineers, Calgary, Alberta,
Canada, p IPC201433221. https://​doi.​org/​10.​1115/​IPC20​14-​33221
Stepp JC, Wong I, Whitney JW et al (2001) Probabilistic seismic hazard analyses for ground motions
and fault displacement at Yucca Mountain, Nevada. Earthq Spectra 17(1):113–151. https://​doi.​org/​
10.​1193/1.​15861​69
Thingbaijam KKS, Mai PM, Goda K (2017) New empirical earthquake source-scaling laws. Bull Seismol
Soc Am 107(5):2225–2246. https://​doi.​org/​10.​1785/​01201​70017
Treiman JA (2010) Fault rupture and surface deformation: Defining the hazard. Environ Eng Geosci
106(1):19–30. https://​doi.​org/​10.​2113/​gseeg​eosci.​16.1.​19
United Nations (2015) Sendai framework for disaster risk reduction 2015–2030. Switzerland, Geneva
Valentini A, Fukushima Y, Contri P et al (2021) Probabilistic fault displacement hazard assessment
(PFDHA) for nuclear installations according to IAEA safety standards. Bull Seismol Soc Am
111(5):2661–2672 https://​doi.​org/​10.​1785/​01202​10083
Wang JH (2018) A review on scaling of earthquake faults. Terr Atmos Ocean Sci 29:589–610. https://​doi.​
org/​10.​3319/​TAO.​2018.​08.​19.​01
Ward SN (1997) Dogtails versus rainbows: Synthetic earthquake rupture models as an aid in interpreting
geological data. Bull Seismol Soc Am 87(6):1422–1441. https://​doi.​org/​10.​1785/​BSSA0​87006​1422
Wells DL, Kulkarni VS (2014) Probabilistic fault displacement hazard analysis - Sensitivity analyses and
recommended practices for developing design fault displacements. NCEE 2014 - 10th US National
Conference on Earthquake Engineering: Frontiers of Earthquake Engineering.
Wells DL, Coppersmith KJ (1993) Likelihood of surface rupture as a function of magnitude. Seismol Res
Lett 64(1):54. https://​doi.​org/​10.​1785/​gssrl.​64.1.​13
Wells DL, Coppersmith KJ (1994) New empirical relationships among magnitude, rupture length, rupture
width, rupture area, and surface displacements. Bull Seismol Soc Am 84(4):974–1002. https://​doi.​org/​
10.​1785/​BSSA0​84004​0974
Woessner J, Laurentiu D, Giardini D et al (2015) The 2013 European seismic hazard model: key compo-
nents and results. Bull Earthq Eng 13:3553–3596. https://​doi.​org/​10.​1007/​s10518-​015-​9795-1
Youngs RR, Coppersmith KJ (1986) Implications of fault slip rates and earthquake recurrence models to
probabilistic seismic hazard estimates. Int J Rock Mech Min Sci Geomech Abstr 23(4):125. https://​doi.​
org/​10.​1016/​0148-​9062(86)​90651-0
Youngs RR, Arabasz WJ, Anderson RE et al (2003) A methodology for probabilistic fault displacement haz-
ard analysis (PFDHA). Earthq Spectra 19(1):191–219. https://​doi.​org/​10.​1193/1.​15428​91

Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and
institutional affiliations.

13