Bulletin of the Seismological Society of America, Vol. 105, No. 4, pp. –, August 2015, doi: 10.

1785/0120140332

Regionally Adjustable Generic Ground-Motion Prediction Equation

Based on Equivalent Point-Source Simulations: Application
to Central and Eastern North America
by Emrah Yenier and Gail M. Atkinson

Abstract We develop a generic ground-motion prediction equation (GMPE) that

can be adjusted for use in any region by modifying a few key model parameters. The
basis of the GMPE is an equivalent point-source simulation model whose parameters
have been calibrated to empirical data in California, in such a way as to determine the
decoupled effects of basic source and attenuation parameters on ground-motion am-
plitudes. We formulate the generic GMPE as a function of magnitude, distance, stress
parameter, geometrical spreading rate, and anelastic attenuation. This provides a fully
adjustable predictive model, allowing users to calibrate its parameters using observed
motions in the target region. We also include an empirical calibration factor to account
for residual effects that are different from and/or missing in simulations compared to
observed motions in the target region. As an example, we show how the generic GMPE
can be adjusted for use in central and eastern North America (CENA), and calibrated
with the Next Generation Attenuation-East database. We provide median predictions
of ground motions in CENA for average horizontal-component peak ground motions
and 5% damped pseudospectral acceleration (periods up to T
10 s), for magnitudes
M 3–8 and distances up to 600 km.

Introduction
Reliable estimates of ground motions that may be pro-
duced by future earthquakes require robust modeling of the
earthquake source and attenuation attributes in the region of
interest. Ground-motion observations from past events pro-
vide a valuable empirical basis to develop ground-motion pre-
diction equations (GMPEs) that describe amplitudes as a
function of variables such as magnitude, distance, and site
condition. However, with the exception of well-monitored ac-
tive regions such as California and Japan, empirical ground-
motion data are generally sparse in the magnitude–distance
range of engineering interest (e.g., magnitudes M > 6:0 and
distances R < 50 km). Thus, there are insufficient data for de-
velopment of reliable GMPEs in many regions, with central
and eastern North America (CENA) being a classic example.
There are several alternative methods used for derivation
of GMPEs in data-poor regions. A widely used method is the
simulation-based approach, in which synthetic ground mo-
tions are generated over a wide magnitude and distance range,
and the GMPE is developed based on the simulated amplitude
data. The simulations are based on a seismological model of
the source, path, and site effects, with the parameters being
calibrated using the available empirical data for the region.
Simulations can be performed using a variety of techniques
ranging from simple stochastic point-source methods to more
sophisticated finite-source broadband simulations (e.g., Atkin-
son and Boore, 1995, 2006; Toro et al., 1997; Atkinson and
Silva, 2000; Somerville et al., 2001, 2009; Silva et al., 2002;
Frankel, 2009). Another common approach is the hybrid em-
pirical method (Campbell, 2002, 2003). This method calibra-
tes an empirically well-constrained GMPE in a data-rich host
region (e.g., western North America [WNA]) for use in a data-
poor target region (e.g., CENA) based on adjustment factors
obtained from response-spectral ratios of stochastic simula-
tions in the host and target regions (e.g., Campbell, 2002,
2003; Scherbaum et al., 2005; Pezeshk et al., 2011). A third
method is the referenced empirical approach introduced by
Atkinson (2008). It is similar to the hybrid empirical method
in concept but adjustment factors are determined empirically
using spectral ratios of observed motions in the target region to
predictions of an empirical GMPE in the host region (e.g., At-
kinson, 2008, 2010; Atkinson and Boore, 2011, Atkinson and
Motazedian, 2013; Hassani and Atkinson, 2015).
Both the hybrid empirical method and the referenced
empirical approach anchor their predictions to magnitude
scaling and saturation effects observed in data-rich regions,
assuming that these effects are transferable. Although the
magnitude scaling is assumed to be similar between regions,
no such assumption is made regarding the overall level of

BSSA Early Edition / 1

2 E. Yenier and G. M. Atkinson
ground-motion amplitudes. Differences in overall amplitude
level and distance scaling between regions are attributed to
regional differences in fundamental source and attenuation
parameters. The hybrid empirical method requires sound
knowledge of these parameters in both host and target
regions to determine host-to-target adjustment factors via
simulations reliably. This may restrict the applicability of the
method (Campbell, 2003). The referenced empirical ap-
proach resolves this issue by determining the adjustment
factors empirically, avoiding the need for assumptions of
the source and attenuation parameters for the host and target
regions. However, an important limitation of the referenced
empirical approach is that the available ground-motion data
in the target region may not sufficiently represent all impor-
tant regional characteristics (Atkinson, 2008).
In this study, we take advantage of key concepts from
both the hybrid empirical and referenced empirical approaches
to develop a robust simulation-based generic GMPE. The
generic GMPE can be adjusted for use in any region by modi-
fying a few key modeling parameters, and calibrated to match
the empirical data in the target region. The basic idea is that
we first develop a simulation-based GMPE using synthetic
motions generated based on parameters that were calibrated to
observed motions in California by Yenier and Atkinson
(2015). We parameterize this GMPE so as to isolate the effects
of the basic source and attenuation parameters on peak ground
motions and response spectra. The wealth of empirical data
available in California enables this parameterization to be
validated over a wide magnitude–distance range. By casting
the calibrated model in terms of a few basic seismological
parameters, we provide effective and transparent control over
the transferable factors between regions. The fundamental
seismological parameters that are used as predictive variables
in the generic GMPE include magnitude, distance, stress
parameter, geometrical spreading rate, and anelastic attenua-
tion coefficient. This provides an adjustable predictive model
with well-behaved scaling characteristics that is readily
calibrated with minimal regional data. In the generic GMPE,
we also consider an empirical calibration factor to account for
residual effects that are different from and/or missing in
simulations compared to empirical data. This closes any
remaining gap between simulated and observed motions.
By decoupling the effects of key model parameters in
the formulation, we have essentially created a “plug-and-
play” GMPE. One can easily produce a regional predictive
model by entering the associated source and attenuation
parameters into the generic GMPE, without the need of
performing multiple simulations and calculating model coef-
ficients, if their values are known through appropriate studies
for the target region. The parameterized form of the generic
GMPE also allows computation of the regional values of
source and attenuation parameters from the available empiri-
cal data, if such information is not known for the target
region. Finally, the formulation facilitates the efficient explo-
ration of the effects on the GMPE in terms of epistemic un-
certainty in key seismological parameters—stress drop and
BSSA Early Edition
attenuation. Importantly, sound overall scaling behavior of
the GMPE in terms of near-distance saturation effects and
magnitude-saturation effects is ensured by the calibration
of the model form to an extensive empirical data in California
obtained from the (Next Generation Attenuation-West 2)
NGA-West 2 database (Yenier and Atkinson, 2015).
We provide a recipe for the adjustment of the generic
GMPE to a specific region. As an example implementation
of the generic model, we use it to develop a GMPE for CENA
by adjusting the stress and anelastic attenuation, and cali-
brate the model using the NGA-East database. During the cal-
ibration exercise, we infer a magnitude- and depth-dependent
stress parameter model based on the values obtained from
study events, as well as a regional attenuation model. This
provides an example of how the generic GMPE model can
be used to determine stress and attenuation parameters from
a regional database. The developed GMPE provides median
predictions of ground motions in CENA for average horizon-
tal-component peak ground motions and 5% damped pseu-
dospectral acceleration (PSA; periods up to T
10 s), for
wide ranges of magnitude (M 3–8) and distance (< 600 km).
Functional Form of the Generic GMPE
A regionally adjustable generic prediction equation re-
quires a robust yet simple functional form that successfully
decouples the effects of fundamental source and attenuation
parameters on ground-motion amplitudes. We define the
generic GMPE as
ln Y
FE  FZ  Fγ  FS  C; 1
in which ln Y is the natural logarithm of a ground-motion
intensity measure. FE , FZ , Fγ , and FS represent functions for
earthquake source, geometrical spreading, anelastic attenua-
tion, and site effects, respectively. The term C is an empirical
calibration factor that accounts for the residual differences
between simulations and empirical data.
We determine the source and geometrical spreading
effects (FE and FZ ) using equivalent point-source simula-
tions with parameters calibrated to observed motions in
California. In the equivalent point-source method, seismic
waves are assumed to radiate from a virtual point source
located at an overall effective distance that mimics observed
close-distance saturation effects (e.g., Boore, 2009; Yenier
and Atkinson, 2014); this effective distance can be thought
of conceptually as including an “added depth” term that de-
pends on magnitude, such that motions from large ruptures
will essentially look like they come from farther away, on
average. The anelastic attenuation (Fγ ) is adjusted to opti-
mize observed frequency-dependent attenuation effects.
In this study, we provide predictions for the orientation-
independent horizontal component of peak ground acceler-
ation (PGA), peak ground velocity (PGV), and 5% damped
PSA, in which PGA and PSA are given in units of gravity
(g) and PGV is in centimeters per second.
Regionally Adjustable Generic GMPE Based on Equivalent Point-Source Simulations
The source function (FE ) describes the effects of mag-
nitude and stress parameter on ground-motion amplitudes as
FE
FM  FΔσ ; 2
in which FM represents the magnitude effect on ground-
motion amplitudes that would be observed at the source, if
there were no distance-saturation effects. It is defined for the
reference stress (Δσ), κ0 parameter, and site condition. We
choose Δσ
100 bar and κ0
0:025 s as the reference
modeling parameters based on the findings of Yenier and
Atkinson (2015) for California earthquakes. In equation (2),
FΔσ represents the stress adjustment factor that is needed
when Δσ is different than 100 bars.
The FM term is defined as a function of moment
magnitude (M), using a hinged-quadratic function:

eled in the response spectral domain rather than the
e0  e1 M − Mh   e2 M − Mh 2 M ≤ Mh
FM
; 3
e0  e3 M − Mh  M > Mh
in which the hinge magnitude Mh and model coefficients
e0 − e3 , are period dependent. This mimics the functional
form of magnitude scaling used by Boore, Stewart, et al.
(2014; hereafter referred as BSSA14) in their NGA-West 2
empirical GMPE. The stress adjustment term is defined as
FΔσ
eΔσ lnΔσ=100; 4
in which eΔσ describes the rate of the ground-motion scaling
with Δσ. Equation (4) describes the relationship between
stress parameter and response spectral amplitudes, facilitat-
ing the determination of Δσ from PSA data in the target
region.
We model the geometrical spreading effects based on the
equivalent point-source method, in which seismic waves are
assumed to radiate from a virtual point source placed at an
overall effective distance from the site, such that the empiri-
cally observed saturation effects are successfully reproduced.
The effective distance (R) is given as
q
R
D2rup  h2 ; 5
in which Drup is the closest distance from the site to the fault-
rupture surface and h is a pseudodepth term that accounts for
close-distance saturation effects. The h term is a magnitude-
dependent model-fitting parameter that expresses finite-fault
effects; it cannot be interpreted as hypocentral depth. The
pseudodepth is generally defined as a function of magnitude
to account for the extension of distance-saturation effects to
larger distances with increasing magnitude. In this study, we
define the pseudodepth as
h
10−0:4050:235M : 6
This equation was derived based on pseudodepths determined
from modeling of well-recorded M > 4:0 earthquakes around
BSSA Early Edition
3
the world (e.g., California, Italy, Japan, New Zealand, Taiwan,
and Turkey) by a trial-and-error procedure that achieved an
overall agreement with empirical data while preventing over-
saturation of predicted amplitudes for large magnitudes (Ye-
nier and Atkinson, 2015). By adopting this expression for the
pseudodepth, we are essentially assuming that finite-fault ef-
fects in all regions will act in a similar way in terms of their
effects on near-distance saturation of motions.
We define the geometrical spreading function (FZ ) as
FZ
lnZ  b3  b4 M lnR=Rref ; 7
in which Z represents the geometrical attenuation of Fourier
amplitudes, whereas the multiplicative component
b3  b4 M lnR=Rref  accounts for the change in the appar-
ent attenuation that occurs when ground motions are mod-
Fourier domain. The coefficients b3 and b4 are period depen-
ref is the reference effective distance, given
dent, and pR
as Rref
1  h2 .
In ground-motion modeling, Z is often modeled in its
simplest form as 1=R. A more rigorous way of modeling
regional attenuation is to consider Z as a piecewise continu-
ous function that describes the distance-dependent attributes
of geometrical spreading, to account for the complex attenu-
ation effects that result from direct-wave spreading at close
distances (which may differ from 1=R due to crustal layer-
ing), and multiple reflections and refractions at larger distan-
ces. Babaie Mahani and Atkinson (2012) evaluated the
ability of various functional forms to describe the geometri-
cal attenuation in North America, and concluded that a bi-
linear model provides a good balance between simplicity
and ability to capture the key attenuation attributes over a
broad distance range. In this study, we define Z using a
hinged bilinear model that provides for a transition from
direct-wave spreading to surface-wave spreading of reflected
and refracted waves:

b
R1 R ≤ Rt
Z
; 8
Rbt 1 R=Rt b2 R > Rt
in which Rt represents the transition distance, and b1 and b2
are the geometrical attenuation rates of Fourier amplitudes at
R ≤ Rt and R > Rt , respectively. In the generic GMPE, we
fix the transition distance at Rt
50 km based on the find-
ings of Yenier and Atkinson (2014).
The geometrical spreading rate at close distances is often
assumed to be b1
−1:0 (i.e., 1=R), based on the homo-
geneous whole space approximation. However, theoretical
waveform simulations suggest faster spreading rates, about
b1 ≈ −1:3, for typical layered earth models (Ojo and Mereu
1986; Burger et al., 1987; Ou and Hemann, 1990; Somerville
et al., 1990; Chapman and Godbee, 2012; Chapman 2013).
Empirical modeling of ground motions in various regions, in-
cluding WNA, CENA, and Australia also support this finding
(Atkinson, 2004; Allen et al., 2007; Babaie Mahani and At-
kinson, 2012; Yenier and Atkinson, 2014, 2015). Therefore,
4 E. Yenier and G. M. Atkinson
we define the geometrical spreading rate at R ≤ 50 km as
b1
−1:3 in the generic model. The geometrical spreading
rate at R > 50 km is fixed at the widely used value of
b2
−0:5, which is consistent with attenuation of surface
waves in a half-space (Ou and Herrmann 1990; Atkinson
2012).
Equation (7) effectively decouples the geometrical
spreading of Fourier amplitudes (Z) and the change in appar-
ent attenuation that occurs when ground motions are con-
volved by the response spectra transfer function. This
decoupling is a key element of our formulation, and is critical
to the development of a “plug-and-play” GMPE, in which a
new GMPE for response spectra can be defined for a specified
attenuation rate (i.e., in the Fourier domain), without the
need to reperform simulations. Although the shape and rates
of geometrical attenuation are fixed at their generic values in
our model, Z can be modified to have different shape and
attenuation rates (e.g., a trilinear model) if there is compel-
ling evidence supporting such a change. In such a case, the
preferred model as given in equation (8) can be replaced with
an alternative geometrical spreading model that is compatible
with the decay of the Fourier amplitudes in the target region.
The anelastic attenuation function (Fγ ) is given as
Fγ
γDrup ; 9
in which Drup is the closest distance to the rupture and γ is a
period-dependent anelastic attenuation coefficient that is em-
pirically determined from regional ground-motion data.
We develop the generic GMPE for a reference site con-
dition, for which the travel-time weighted average shear-
wave velocity over the top 30 m is V S30
760 m=s (referred
to as National Earthquake Hazards Reduction Program
[NEHRP], 2000 B/C site condition). In this study, we adopt
the site-effect model of BSSA14 (which was originally de-
rived by Seyhan and Stewart, 2014) to predict ground mo-
tions at different site conditions, as a function of V S30 :
FS
Flin  Fnl ; 10
in which Flin represents the linear site effects, and Fnl
represents the nonlinear site effects. The linear site response
is defined as

a set of self-consistent modeling parameters that were well
c lnV S30 =760 V S30 ≤ V c
Flin
; 11
c lnV c =760 V S30 > V c
in which c describes the V S30 scaling and V c is the limiting
velocity beyond which ground motions no longer scale with
V S30 . The nonlinear site response is given as
 
PGA r  f 3
Fnl
f 1  f2 ln ; 12
f3
in which f2 represents the degree of nonlinearity as a func-
tion of V S30 :
BSSA Early Edition
f2
f4 expff5 minV S30 ; 760 − 360g
− expff5 760 − 360g: 13
In equations (11)–(13), parameters c, V c , f1 , f3 , f 4 , and f5
are model coefficients given in BSSA14 and PGA r is the
median peak horizontal acceleration predicted for the refer-
ence condition (V S30
760 m=s).
Determination of Model Coefficients
We calculate model coefficients of the magnitude effect
(FM ), geometrical spreading function (FZ ), and stress adjust-
ment factor (FΔσ ) from amplitude data generated from
ground-motion simulations. The simulations are based on
the equivalent point-source stochastic method with modeling
parameters calibrated to observed motions in California as
described by Yenier and Atkinson (2015); model parameters
are given in Table 1. To summarize briefly, we use the
additive double-corner-frequency source model of Boore,
Di Alessandro, et al. (2014) with a spectral-sag parameter (ε)
suggested by Yenier and Atkinson (2015). In simulations, the
geometrical decay of Fourier amplitudes (Z) is defined
in terms of effective distance, as given in equation (8)
(b1
−1:3, b2
−0:5, and Rt
50 km). We use the pseu-
dodepth model given in equation (6) to account for near-
distance saturation effects. We constrained the pseudodepth
function to avoid oversaturation of predicted amplitudes at
large magnitudes (Fig. 1). The simulations do not include
anelastic attenuation, because we will determine these ef-
fects empirically from regional ground-motion data. We
simulate ground motions for NEHRP B/C site condition
(V S30
760 m=s), using the generic crustal amplification
factors given by Atkinson and Boore (2006). We assume that
the near-surface high-frequency attenuation parameter is
κ0
0:025 s for this site class. Yenier and Atkinson (2015)
showed equivalent point-source simulations with these
modeling parameters (but also including regional anelastic
attenuation effects and a magnitude- and depth-dependent
stress parameter) can reproduce average observed spectral
amplitudes of earthquakes in California, within 25% error
band, for magnitudes up to M 7.5 and distances less than
400 km. The generic GMPE is defined for a base-case, using
calibrated to observed motions in California. Any inadequa-
cies or misfits between the simulations and empirical data
will map into unresolved residuals, which will be taken into
account through the calibration factor C.
We perform time-domain equivalent point-source sto-
chastic simulations using the widely cited Stochastic-Method
SIMulation (SMSIM) software (Boore, 2003, 2005), for mag-
nitudes from M 3–8 (with increments of 0.1 M units) and dis-
tances from 1 to 400 km (with increments of 0.1 log10 units),
for a fixed stress parameter of Δσ
100 bar. We generate
100 synthetic ground motions for each combination of M and
Drup . For each simulated time series, we calculate PGA, PGV,
Regionally Adjustable Generic GMPE Based on Equivalent Point-Source Simulations
Table 1
Parameter Values Used in Stochastic Equivalent Point-Source
Simulations (from Yenier and Atkinson, 2015)
Parameter Value
Shear-wave velocity β
3:7 km=s
Density ρ
2:8 g=cm3
Source model Generalized additive double-
corner-frequency source model
of Boore, Di Alessandro, et al.
(2014)
Spectral sag ε
min1; 101:2−0:3M 
Effective distance R
D2rup  h2 0:5
Pseudodepth h
10−0:4050:235M
Geometrical attenuation R−1:3 for R ≤ 50 km, and
50−1:3 R=50−0:5 for
R > 50 km
Anelastic attenuation Not considered in simulations
(determined empirically)
Site amplification (National Table 4 of Atkinson and Boore
Earthquake Hazards (2006) Frequency-amplification
Reduction Program pairs delimited by semicolons:
[NEHRP] B/C) 0.0001 Hz–1; 0.1 Hz–1.07;
0.24 Hz–1.15;
0.45 Hz–1.24; 0.79 Hz–1.39;
1.38 Hz–1.67; 1.93 Hz–1.88;
2.85 Hz–2.08; 4.03 Hz–2.2;
6.34 Hz–2.31; 12.5 Hz–2.41;
21.2 Hz–2.45; 33.4 Hz–2.47;
82 Hz–2.50
Kappa factor κ 0
0:025 s
Source duration 0:5=fa  0:5=fb , in which fa and
f b are the corner frequencies
Path duration* Table 1 of Boore and Thompson
(2014) Rupture distance–path
duration pairs delimited by
semicolons: 0 km–0 s; 7 km–
2.4 s; 45 km–8.4 s;
125 km–10.9 s; 175 km–17.4 s;
270 km–34.2 s. Path duration
increases with distance at a rate
of 0:156 s=km after the last
nodal point
Simulation calibration factor for Csim
3:16
California†
*In simulations, the nodal rupture distances are converted to effective
distance based on equation (6) at each magnitude level.
†
Factor applied to simulations for matching simulations to observed response
spectra in California with zero bias. (Refer to Yenier and Atkinson, 2015, for
more information regarding the Csim parameter).
and PSA at 31 periods from 0.01 to 10 s, then take the geo-
metric mean for each parameter over the 100 simulations.
The coefficients of the magnitude-scaling term FM are
computed from the regression of simulations obtained at
Drup
1 km (Y sim;1 km ). FM represents the magnitude scaling
of ground motions that would be observed at the source if
there were no saturation effects. Therefore, we need to remove
the saturation effects that we imposed in the simulations at
1 km to extract the unsaturated magnitude effects FM :
p
ln Y sim;1 km
FM − 1:3 ln 1  h2 ; 14
BSSA Early Edition
5
p
in which −1:3 ln 1  h2  accounts for the saturation effects
imposed in the simulations (i.e., FZ at Drup
1 km). We use
a grid search to determine the hinge magnitude (Mh ), where
we determine the coefficients e0 − e3 by regression of the
amplitudes at 1 km, for each trial value of Mh . We select the
best-fitting Mh and the associated coefficients (e0 − e3 ) based
on minimizing the residuals of the simulated amplitudes with
respect to the model equation. Figure 1 compares ground
motions simulated at Drup
1 km and the fitted model (equa-
tion 14) as a function of magnitude, for peak ground motions,
and response spectra. In Figure 1, the fitted functional form
captures the magnitude scaling and saturation effects implied
by simulations very well.
We determined the model coefficients of the geometrical
spreading function from regression of simulated amplitudes
at variable distances, after removing the magnitude effects
(i.e., ln Y sim − FM ). We use the form:
ln Y sim − FM
lnZ  b3  b4 M lnR=Rref ; 15
In this regression, we constrain the Z to the decay shape used
in the simulations (i.e., b1
−1:3, b2
−0:5, and
Rt
50 km). This forces the differences between the decay
rates of Fourier and response spectral amplitudes to map into
b3  b4 M lnR=Rref . In Figure 2, we compare the generic
model (i.e., FM  FZ ) against simulations to assess the per-
formance of the fitted FZ model. This shows that the generic
GMPE is in good agreement with the behavior of the simu-
lated amplitudes. The values of model coefficients for FM
and FZ are listed in Table 2. This specifies the generic GMPE
for California for the reference stress parameter (100 bars)
and the reference site condition (NEHRP B/C), but without
anelastic attenuation or overall amplitude calibration factor.
These factors can be determined empirically, as described
further later.
We generate another set of simulations to calculate the
stress adjustment factor. In this second set, we simulate
ground motions for the same magnitude range (M 3–8) but
for a fixed distance Drup
1 km and variable stress param-
eters (10 bar ≤ Δσ ≤ 1000 bar). Similar to the first set, 100
synthetic motions are generated for each combination of M,
Drup , and Δσ, and the geometric mean of the peak motions
and response spectra are calculated.
The stress adjustment factor FΔσ models the expected
change in amplitudes when Δσ is different than 100 bars. We
determine FΔσ using simulations obtained at Drup
1 km, as
FΔσ
lnY sim;1 km M;Δσ− lnY sim;1 km M;100 bar; 16
in which Y sim;1 km M; Δσ is the ground motion simulated at
Drup
1 km for a given magnitude and stress, and
Y sim;1 km M; 100 bar represents the ground motion simulated
at Drup
1 km for the same magnitude, but for the reference
stress (Δσ
100 bar). Figure 3 shows the required stress ad-
justment factors as a function of Δσ, for various magnitudes and
periods. This factor has an increasing trend with the stress, in
6 E. Yenier and G. M. Atkinson

Figure 1. Ground motions simulated at Drup
1 km (circles), and the fitted model (lines) as a function of magnitude.

which FΔσ
0 at Δσ
100 bar, by definition. The slope of
FΔσ , which is defined by coefficient eΔσ in equation (4), rep-
resents the strength of the ground-motion scaling with the stress
parameter. The steeper is the slope, the larger is the influence
of stress on ground motions. In Figure 3, Δσ has significant
influence at short periods (T < 0:2 s), regardless of magni-
tude. However, its effects weaken with increasing period, par-
ticularly for small-to-moderate magnitude events (M < 6). For
large magnitudes, the Δσ effects extend to longer periods due
to the shifting of the two corner frequencies with magnitude.
We regress the values of eΔσ (calculated for each mag-
nitude and period from the values of FΔσ using equation 4) to
the functional form:

the reference site. We perform an additional set of simulations
s0  s1 M  s2 M2  s3 M3  s4 M4 Δσ ≤ 100 bar
eΔσ
;
s5  s6 M  s7 M2  s8 M3  s9 M4 Δσ > 100 bar
17
in which s0 − s9 are period-dependent model coefficients. We
use two polynomials, because we require a different shape for
the eΔσ values for Δσ ≤ 100 bar and Δσ > 100 bar; we
constrain the regressions to attain FΔσ
0 at Δσ
100 bar.
BSSA Early Edition
Figure 4 shows how the values of eΔσ vary with magnitude and
period. The net effect of the stress parameter is complicated
because of interactions between scaling of the high-frequency
source amplitudes, shifting of the two corner frequencies, and
changes in spectral sag between the corner frequencies. In ad-
dition, the stress parameter affects the source duration, which
in turn influences the response spectral amplitudes. Coupling
of all these factors in the response spectrum domain requires a
high-order polynomial to satisfactorily model Δσ scaling over
a wide period range. The values of model coefficients for the
stress adjustment factor are listed in Table 3.
We have defined the source and attenuation functions for
NEHRP B/C site condition, assuming that κ 0
0:025 s for
to investigate the effect of κ0 on predicted amplitudes, for the
NEHRP B/C site condition. In this set, we simulate ground
motions for variable κ 0 from 0.01 to 0.04 s and the reference
stress (Δσ
100 bar). The selected range of κ 0 parameter
(i.e., 0.01–0.04 s) covers values obtained in different studies
for NEHRP B/C site condition (e.g., Youssef et al., 2014).
Simulations are performed for the same magnitude (M 3–8)
and distance (up to 400 km) ranges. One hundred synthetic
Regionally Adjustable Generic GMPE Based on Equivalent Point-Source Simulations 7

Figure 2. Simulations (symbols) in comparison to predictions of the generic model (lines), for magnitudes M 3–8 (Δσ
100 bar,
V S30
760 m=s). No anelastic attenuation is included in either simulations or the generic ground-motion prediction equation (GMPE) be-
cause this effect is determined empirically. The color version of this figure is available only in the electronic edition.

motions are generated for each combination of M, Drup , and
κ0 , and the geometric mean of the peak motions and response
spectra are calculated. We calculate the κ 0 effect based on the
ratio of simulations obtained from variable κ 0 values to those
obtained for the reference κ 0 , for each M − Drup − κ 0 combi-
nation. We found that the κ0 parameter primarily influences
ground motions at T < 0:2 s and M < 4. Its effects are gen-
erally less than 25% for T > 0:2 s and M > 4, regardless of
the distance. Therefore, we neglect the κ0 effects in the generic
GMPE to achieve model simplicity. We note that any neglected
κ0 effects that may exist will map into the regional calibration
factor C (i.e., if the κ 0 parameter for NEHRP B/C sites in
the target region is different than the presumed value of
κ0
0:025 s).
Adjustment of the Generic GMPE
for a Target Region
Adjustment of the generic model to a specific region in-
cludes any required modifications to the source and attenua-
tion parameters, as well as determination of an empirical
calibration factor that accounts for residual effects that are
missing and/or different in the simulations compared to the
observed motions. We assume that the magnitude (FM ) and
saturation (h) effects determined from simulations are trans-
ferable to other regions, but that the stress parameter may vary
regionally. The generic GMPE can be adjusted for this effect
when the regional value of stress parameter is plugged into
FΔσ ; in the example application, we show how regional data
can be used to develop a model for the regional stress param-
eter. The required modifications for regional attenuation are
made by means of Z and γ, which are also determined from
regional data as shown in the following section. We recom-
mend keeping the presumed Z model (geometric spreading)
as it is defined in the generic model, unless there is compelling
evidence for its modification. The anelastic attenuation coef-
ficient γ is determined using empirical data at regional distan-
ces for the region of interest; such data can be obtained from
weak-motion studies. The calibration factor C is calculated
following other regional adjustment steps, through the

BSSA Early Edition

8 E. Yenier and G. M. Atkinson

Table 2
Model Coefficients for the Magnitude Scaling (FM ) and Geometrical Spreading (FZ ) Functions
T (s) Mh e0 e1 e2 e3 b3 b4
0 −1 −1 −1 −1
0.010 5.85 2:23 × 10 6:87 × 10 −1:36 × 10 7:64 × 10 −6:21 × 10 6:06 × 10−2
0.013 5.90 2:28 × 100 6:85 × 10−1 −1:29 × 10−1 7:62 × 10−1 −6:26 × 10−1 6:13 × 10−2
0.016 5.85 2:27 × 100 6:97 × 10−1 −1:23 × 10−1 7:59 × 10−1 −6:31 × 10−1 6:19 × 10−2
0.020 5.90 2:38 × 100 7:00 × 10−1 −1:07 × 10−1 7:49 × 10−1 −6:38 × 10−1 6:25 × 10−2
0.025 6.00 2:56 × 100 6:84 × 10−1 −9:42 × 10−2 7:41 × 10−1 −6:31 × 10−1 6:10 × 10−2
0.030 6.15 2:81 × 100 6:61 × 10−1 −9:09 × 10−2 7:39 × 10−1 −6:03 × 10−1 5:64 × 10−2
0.040 5.75 2:73 × 100 7:03 × 10−1 −1:09 × 10−1 7:38 × 10−1 −5:48 × 10−1 4:82 × 10−2
0.050 5.35 2:56 × 100 7:19 × 10−1 −1:64 × 10−1 7:54 × 10−1 −5:10 × 10−1 4:28 × 10−2
0.065 5.75 3:00 × 100 6:84 × 10−1 −1:55 × 10−1 7:55 × 10−1 −4:67 × 10−1 3:64 × 10−2
0.080 5.20 2:58 × 100 7:65 × 10−1 −2:43 × 10−1 7:87 × 10−1 −4:21 × 10−1 3:07 × 10−2
0.100 5.45 2:78 × 100 7:12 × 10−1 −2:62 × 10−1 7:94 × 10−1 −3:77 × 10−1 2:47 × 10−2
0.130 5.35 2:64 × 100 7:35 × 10−1 −3:32 × 10−1 8:12 × 10−1 −3:55 × 10−1 2:22 × 10−2
0.160 5.25 2:47 × 100 8:09 × 10−1 −3:87 × 10−1 8:41 × 10−1 −3:26 × 10−1 1:92 × 10−2
0.200 5.45 2:55 × 100 8:19 × 10−1 −3:86 × 10−1 8:43 × 10−1 −2:87 × 10−1 1:38 × 10−2
0.250 5.60 2:52 × 100 8:67 × 10−1 −3:77 × 10−1 8:78 × 10−1 −2:43 × 10−1 9:21 × 10−3
0.300 5.85 2:63 × 100 8:47 × 10−1 −3:63 × 10−1 8:76 × 10−1 −2:12 × 10−1 5:16 × 10−3
0.400 6.15 2:67 × 100 8:50 × 10−1 −3:47 × 10−1 8:97 × 10−1 −1:93 × 10−1 4:85 × 10−3
0.500 6.25 2:54 × 100 8:86 × 10−1 −3:49 × 10−1 9:18 × 10−1 −2:08 × 10−1 8:54 × 10−3
0.650 6.60 2:62 × 100 8:76 × 10−1 −3:16 × 10−1 9:25 × 10−1 −2:28 × 10−1 1:37 × 10−2
0.800 6.85 2:66 × 100 9:05 × 10−1 −2:89 × 10−1 8:94 × 10−1 −2:52 × 10−1 1:91 × 10−2
1.000 6.45 1:99 × 100 1:34 × 100 −2:46 × 10−1 9:83 × 10−1 −2:97 × 10−1 2:76 × 10−2
1.300 6.75 2:01 × 100 1:39 × 100 −2:06 × 10−1 1:00 × 100 −3:50 × 10−1 3:78 × 10−2
1.600 6.75 1:75 × 100 1:56 × 100 −1:68 × 10−1 1:05 × 100 −3:85 × 10−1 4:43 × 10−2
2.000 6.65 1:25 × 100 1:75 × 100 −1:32 × 10−1 1:19 × 100 −4:35 × 10−1 5:36 × 10−2
2.500 6.70 9:31 × 10−1 1:82 × 100 −1:09 × 10−1 1:29 × 100 −4:79 × 10−1 6:14 × 10−2
3.000 6.65 5:16 × 10−1 1:91 × 100 −8:98 × 10−2 1:42 × 100 −5:13 × 10−1 6:76 × 10−2
4.000 6.85 3:44 × 10−1 1:93 × 100 −7:47 × 10−2 1:51 × 100 −5:51 × 10−1 7:43 × 10−2
5.000 6.85 −7:92 × 10−2 1:98 × 100 −6:21 × 10−2 1:59 × 100 −5:80 × 10−1 7:90 × 10−2
6.500 7.15 −6:67 × 10−3 1:97 × 100 −5:45 × 10−2 1:63 × 100 −5:96 × 10−1 8:12 × 10−2
8.000 7.50 2:56 × 10−1 1:94 × 100 −5:23 × 10−2 1:59 × 100 −6:09 × 10−1 8:30 × 10−2
10.000 7.45 −2:76 × 10−1 1:97 × 100 −4:63 × 10−2 1:72 × 100 −6:20 × 10−1 8:42 × 10−2
PGA 5.85 2:22 × 100 6:86 × 10−1 −1:39 × 10−1 7:66 × 10−1 −6:19 × 10−1 6:03 × 10−2
PGV 5.90 5:96 × 100 1:03 × 100 −1:65 × 10−1 1:08 × 100 −5:79 × 10−1 5:74 × 10−2

analysis of residuals between observed motions in the target
region and the GMPE, after application of the derived regional
values of Δσ, Z, and γ.
An Example Application: Adjustment of the Generic
GMPE for CENA
As an example of the method, we adjust the generic
GMPE for CENA using ground motions obtained in the re-
gion. We use the database of PGA, PGV, and 5% damped
PSA from the NGA-East flatfile (see Data and Resources),
for CENA earthquakes of M ≥ 3:0 that were recorded by
at least three stations within 600 km. We consider both natu-
ral and potentially induced earthquakes (flagged in the NGA-
East flatfile). However, ground motions recorded in the Gulf
Coast region are excluded due to considerably different at-
tenuation attributes in this region (Electric Power Research
Institute [EPRI], 2004). (The generic GMPE could also be
used to develop a separate predictive model for the Gulf
Coast region using the available empirical data and regional
source and attenuation characteristics reported in the litera-
ture.) We use the average orientation-independent horizon-
tal-component ground motions calculated based on the
BSSA Early Edition
RotD50 measure (Boore, 2010), as provided in the NGA-East
flatfile; this is approximately equivalent to geometric mean
motions as provided in the simulations. Figure 5 shows a
map of the epicenters of the study events and Figure 6 is
a map of stations and their site condition. Figure 7 shows
the magnitude–distance distribution of the selected records.
In the analysis, we consider response spectra up to a
maximum usable period to reduce the impact of long-period
noise on the adjusted GMPE. For a given ground-motion
record, the maximum usable period T max is defined as
1
T max
; 18
max1:25flc ; fmin 
in which flc is the low-cut filter frequency of the record
reported in the NGA-East flatfile and fmin is the limiting
frequency below which spectral amplitudes are assumed to
be noise dominated. We describe fmin as
fmin
max0:1; 100:75−M=3 : 19
This equation is defined such that it provides an overall
agreement with the geometric mean of the factored filter
Regionally Adjustable Generic GMPE Based on Equivalent Point-Source Simulations 9

Figure 3. Stress adjustment factors (FΔσ ) determined from simulations (symbols). The color version of this figure is available only in the
electronic edition.

frequencies (i.e., 1:25flc ), as seen in Figure 8. For M < 6, the
f min model given for CENA is relatively less conservative
than that was used for California by Yenier and Atkinson
(2015) because ground motions attenuate more slowly in
CENA, providing usable signal to greater distances.
We correct the recorded ground-motion amplitudes
to the equivalent values for NEHRP B/C site conditions
(V S30
760 m=s) using the FS function adopted from
BSSA14. This function is based on the values of V S30 and
PGA r for each record, where the V S30 values are given in the
NGA-East flatfile, and we assume that PGA r can be reason-
ably estimated from BSSA14 as an approximation. We de-
liberately use BSSA14 rather than a CENA GMPE for this
purpose, as we do not wish the higher frequency content in
CENA to impose greater nonlinearity. We assess the appli-
cability of the adopted site-effect model for CENA based
on the model residuals (discussed later).
Regional Attenuation
Empirical studies suggest that the geometrical spreading
of Fourier amplitudes in CENA can be adequately described
BSSA Early Edition
as R−1:3 within 50 km and R−0:5 at farther distances (Babaie
Mahani and Atkinson, 2012; Atkinson and Boore, 2014). We
therefore use the generic bilinear Z model determined by
Yenier and Atkinson (2015) for California (b1
−1:3,
b2
−0:5, and Rt
50 km) without modification. The only
attenuation adjustment needed is for the regional anelastic
attenuation. As described in the methodology presented ear-
lier, we determine the regional anelastic attenuation (γ CENA )
from the empirical data using
ln Y B=C;ij − FM;i  FZ;ij 
Ei  γ CENA Drup;ij ; 20
in which Y B=C;ij represents the B/C-corrected motion for
event i and station j. FM;i and FZ;ij are the magnitude
and geometrical spreading functions evaluated for the known
magnitude and distance (Drup;ij ) of the record, respectively.
The Ei term is an event term, which provides the average
adjustment required to match observed amplitudes from
event i. Its value can be attributed to two main factors:
(1) the difference between the reference stress implicitly car-
ried by the FM function (100 bars) and its true value for the
ith event (modeled by FΔσ ) and (2) the overall difference
10
Figure 4. Stress-scaling coefficients (eΔσ ) obtained from simulations (symbols) and the fitted model (equation 17, lines). The color
version of this figure is available only in the electronic edition.
between synthetics and observed motions in CENA (modeled
by C). We calculate the regional anelastic attenuation coef-
ficient (γ CENA ) and event terms (Ei ) for each oscillator period
and ground-motion parameter; the values of the γ CENA term
are listed Table 4.
Regional Stress Parameter
The stress parameter is often determined by matching
the predicted and observed spectral amplitudes at short peri-
ods for the specified moment. However, this approach results
in a nonunique solution for Δσ due to the trade-off between
earthquake source and attenuation (Boore et al., 2010; Yenier
and Atkinson, 2014). Moreover, Δσ has little effect on the
response spectrum at long periods (Fig. 3), especially for
small-to-moderate events, restricting our ability to calibrate
the response spectral amplitudes at long periods. To ensure a
model calibration that is consistent over a wide period range,
we determine the stress parameter by matching the observed
spectral shape for the known moment (i.e., the corner fre-
quency), rather than spectral amplitudes. This breaks the
trade-off between source and attenuation parameters, trans-
BSSA Early Edition
E. Yenier and G. M. Atkinson
ferring the overall amplitude difference to the calibration fac-
tor C (Yenier and Atkinson, 2015). Following this technique,
we use a grid search to determine Δσ for each event sepa-
rately. We select the best-fitting Δσ based on the minimum
standard deviation of residuals between Ei and FΔσ , over a
wide period range (0:01 s ≤ T ≤ 10 s); by minimizing the
standard deviation of residuals, we are effectively finding
the best shape, rather than the best level. This shape-based
determination of stress can be made from response spectra
because of the way we have decoupled effects in our model
formulation, thus avoiding the need for separate studies in
the Fourier domain.
Figure 9 shows the shape-based Δσ values obtained
from CENA events as a function of focal depth (d). The mean
stress determined for evenly spaced focal depth bins shows
an increasing trend from Δσ ≈ 30 bar at d
2:5 km to Δσ ≈
250 bar at d
10 km; it remains relatively constant at
greater depths. Figure 10 shows the best-fitting Δσ as a func-
tion of magnitude. For M < 5, the stress parameter shows
large variability. Despite the large variation of Δσ values
at small magnitudes, the depth effect is clearly visible by
 Table 3
Model Coefficients for the Stress Adjustment Factor (FΔσ )
T (s) s0 s1 s2 s3 s4 s5 s6 s7 s8 s9

0.010 −2:05 × 100 1:88 × 100 −4:90 × 10−1 5:67 × 10−2 −2:43 × 10−3 −1:44 × 100 1:24 × 100 −2:89 × 10−1 3:09 × 10−2 −1:25 × 10−3
0.013 −1:92 × 100 1:80 × 100 −4:71 × 10−1 5:47 × 10−2 −2:36 × 10−3 −1:35 × 100 1:20 × 100 −2:80 × 10−1 3:01 × 10−2 −1:23 × 10−3
0.016 −1:71 × 100 1:66 × 100 −4:36 × 10−1 5:09 × 10−2 −2:20 × 10−3 −1:08 × 100 1:04 × 100 −2:47 × 10−1 2:69 × 10−2 −1:11 × 10−3
0.020 −1:16 × 100 1:27 × 100 −3:34 × 10−1 3:91 × 10−2 −1:70 × 10−3 −1:27 × 100 1:25 × 100 −3:17 × 10−1 3:62 × 10−2 −1:55 × 10−3
0.025 −1:54 × 100 1:59 × 100 −4:29 × 10−1 5:10 × 10−2 −2:24 × 10−3 −1:45 × 100 1:37 × 100 −3:37 × 10−1 3:73 × 10−2 −1:54 × 10−3
0.030 −1:06 × 100 1:20 × 100 −3:13 × 10−1 3:62 × 10−2 −1:55 × 10−3 −2:24 × 100 1:98 × 100 −5:09 × 10−1 5:78 × 10−2 −2:44 × 10−3
0.040 −8:57 × 10−1 1:04 × 100 −2:68 × 10−1 3:08 × 10−2 −1:33 × 10−3 −3:31 × 100 2:66 × 100 −6:68 × 10−1 7:42 × 10−2 −3:06 × 10−3
0.050 −9:63 × 10−1 9:83 × 10−1 −2:16 × 10−1 2:08 × 10−2 −7:42 × 10−4 −4:23 × 100 3:29 × 100 −8:32 × 10−1 9:30 × 10−2 −3:87 × 10−3
0.065 −2:23 × 100 1:95 × 100 −4:90 × 10−1 5:49 × 10−2 −2:29 × 10−3 −3:96 × 100 2:87 × 100 −6:67 × 10−1 6:88 × 10−2 −2:65 × 10−3
0.080 −3:68 × 100 2:96 × 100 −7:51 × 10−1 8:42 × 10−2 −3:51 × 10−3 −3:14 × 100 2:18 × 100 −4:67 × 10−1 4:47 × 10−2 −1:60 × 10−3
0.100 −4:05 × 100 3:10 × 100 −7:62 × 10−1 8:33 × 10−2 −3:39 × 10−3 −2:45 × 100 1:57 × 100 −2:89 × 10−1 2:30 × 10−2 −6:57 × 10−4
0.130 −4:17 × 100 3:09 × 100 −7:44 × 10−1 7:98 × 10−2 −3:21 × 10−3 −1:38 × 100 6:26 × 10−1 −1:16 × 10−2 −1:09 × 10−2 8:28 × 10−4
0.160 −3:96 × 100 2:82 × 100 −6:50 × 10−1 6:72 × 10−2 −2:61 × 10−3 −2:00 × 10−1 −3:37 × 10−1 2:57 × 10−1 −4:25 × 10−2 2:18 × 10−3
0.200 −2:71 × 100 1:73 × 100 −3:30 × 10−1 2:82 × 10−2 −9:06 × 10−4 8:20 × 10−1 −1:08 × 100 4:40 × 10−1 −6:10 × 10−2 2:85 × 10−3
0.250 −1:77 × 100 9:83 × 10−1 −1:31 × 10−1 6:00 × 10−3 −1:16 × 10−5 1:78 × 100 −1:77 × 100 6:07 × 10−1 −7:83 × 10−2 3:50 × 10−3
0.300 −3:18 × 10−1 −1:39 × 10−1 1:70 × 10−1 −2:85 × 10−2 1:42 × 10−3 2:25 × 100 −2:00 × 100 6:33 × 10−1 −7:70 × 10−2 3:27 × 10−3
0.400 2:02 × 100 −1:86 × 100 6:12 × 10−1 −7:67 × 10−2 3:34 × 10−3 2:42 × 100 −1:94 × 100 5:56 × 10−1 −6:17 × 10−2 2:39 × 10−3
0.500 3:96 × 100 −3:29 × 100 9:88 × 10−1 −1:20 × 10−1 5:14 × 10−3 8:56 × 10−1 −4:53 × 10−1 6:46 × 10−2 5:22 × 10−3 −8:30 × 10−4
0.650 3:65 × 100 −2:82 × 100 7:93 × 10−1 −8:93 × 10−2 3:55 × 10−3 −6:67 × 10−1 9:28 × 10−1 −3:71 × 10−1 6:18 × 10−2 −3:43 × 10−3
0.800 2:40 × 100 −1:65 × 100 4:09 × 10−1 −3:71 × 10−2 1:05 × 10−3 −2:12 × 100 2:15 × 100 −7:30 × 10−1 1:05 × 10−1 −5:29 × 10−3
1.000 1:07 × 100 −4:55 × 10−1 3:74 × 10−2 1:03 × 10−2 −1:08 × 10−3 −4:47 × 100 4:05 × 100 −1:27 × 100 1:71 × 10−1 −8:14 × 10−3
1.300 −2:51 × 100 2:52 × 100 −8:45 × 10−1 1:21 × 10−1 −6:02 × 10−3 −5:49 × 100 4:77 × 100 −1:44 × 100 1:85 × 10−1 −8:46 × 10−3
1.600 −5:26 × 100 4:74 × 100 −1:48 × 100 1:96 × 10−1 −9:28 × 10−3 −5:88 × 100 4:98 × 100 −1:46 × 100 1:83 × 10−1 −8:16 × 10−3
2.000 −6:64 × 100 5:77 × 100 −1:74 × 100 2:24 × 10−1 −1:03 × 10−2 −6:01 × 100 4:99 × 100 −1:43 × 100 1:75 × 10−1 −7:59 × 10−3

BSSA Early Edition

2.500 −8:08 × 100 6:84 × 100 −2:02 × 100 2:54 × 10−1 −1:14 × 10−2 −4:88 × 100 3:95 × 100 −1:09 × 100 1:26 × 10−1 −5:17 × 10−3
3.000 −7:98 × 100 6:64 × 100 −1:92 × 100 2:37 × 10−1 −1:04 × 10−2 −4:18 × 100 3:32 × 100 −8:86 × 10−1 9:89 × 10−2 −3:85 × 10−3
4.000 −7:12 × 100 5:78 × 100 −1:61 × 100 1:90 × 10−1 −7:98 × 10−3 −2:63 × 100 1:96 × 100 −4:62 × 10−1 4:24 × 10−2 −1:18 × 10−3
5.000 −6:39 × 100 5:08 × 100 −1:38 × 100 1:58 × 10−1 −6:36 × 10−3 −1:38 × 100 9:09 × 10−1 −1:42 × 10−1 1:32 × 10−3 7:11 × 10−4
6.500 −4:80 × 100 3:68 × 100 −9:37 × 10−1 9:76 × 10−2 −3:47 × 10−3 −3:93 × 10−1 9:83 × 10−2 9:53 × 10−2 −2:78 × 10−2 1:96 × 10−3
8.000 −3:42 × 100 2:51 × 100 −5:80 × 10−1 5:15 × 10−2 −1:34 × 10−3 −6:87 × 10−3 −1:89 × 10−1 1:69 × 10−1 −3:53 × 10−2 2:20 × 10−3
Regionally Adjustable Generic GMPE Based on Equivalent Point-Source Simulations

10.000 −2:19 × 100 1:51 × 100 −2:87 × 10−1 1:53 × 10−2 2:38 × 10−4 2:68 × 10−1 −3:86 × 10−1 2:17 × 10−1 −3:97 × 10−2 2:30 × 10−3
PGA −2:13 × 100 1:94 × 100 −5:04 × 10−1 5:82 × 10−2 −2:50 × 10−3 −1:44 × 100 1:24 × 100 −2:85 × 10−1 3:02 × 10−2 −1:22 × 10−3
PGV −2:25 × 100 1:95 × 100 −5:18 × 10−1 6:14 × 10−2 −2:73 × 10−3 −1:76 × 100 1:38 × 100 −3:26 × 10−1 3:50 × 10−2 −1:42 × 10−3
11
12 E. Yenier and G. M. Atkinson

Figure 5. Epicenters of study events in central and eastern North America (CENA). Circles show epicenter locations of naturally occur-
ring earthquakes and squares indicate events that have been flagged as potentially induced in the Next Generation Attenuation (NGA)-East
flatfile. Dashed line marks the Gulf Coast region. The color version of this figure is available only in the electronic edition.

Figure 6. Locations of recording stations and their National Earthquake Hazards Reduction Program (NEHRP) site classification: A,
V S30 > 1500 m=s; B, 760 m=s < V S30 ≤ 1500 m=s; C, 360 m=s < V S30 ≤ 760 m=s; D, 180 m=s < V S30 ≤ 360 m=s; and E, V S30 ≤ 180 m=s
(NEHRP, 2000). We excluded stations located in the Gulf Coast region (dashed line). The color version of this figure is available only in the
electronic edition.

BSSA Early Edition

Regionally Adjustable Generic GMPE Based on Equivalent Point-Source Simulations 13

Table 4
Anelastic Attenuation Coefficients for Central and
Eastern North America (CENA) and California-
Adjusted Ground-Motion Prediction Equations
(GMPEs)
T (s) γ CENA γ California
−3
0.010 −4:66 × 10 −9:82 × 10−3
0.013 −4:69 × 10−3 −9:83 × 10−3
0.016 −4:69 × 10−3 −9:83 × 10−3
0.020 −4:67 × 10−3 −9:82 × 10−3
0.025 −4:88 × 10−3 −9:88 × 10−3
0.030 −5:11 × 10−3 −1:01 × 10−2
0.040 −5:27 × 10−3 −1:08 × 10−2
0.050 −5:47 × 10−3 −1:13 × 10−2
0.065 −5:71 × 10−3 −1:19 × 10−2
0.080 −5:79 × 10−3 −1:24 × 10−2
0.100 −5:64 × 10−3 −1:25 × 10−2
0.130 −5:24 × 10−3 −1:22 × 10−2
0.160 −4:77 × 10−3 −1:17 × 10−2
0.200 −4:20 × 10−3 −1:09 × 10−2
Figure 7. Magnitude–distance distribution of the selected 0.250 −3:65 × 10−3 −1:02 × 10−2
ground motions in CENA. Ground motions recorded beyond 0.300 −3:12 × 10−3 −9:43 × 10−3
600 km are not considered. The color version of this figure is avail- 0.400 −2:44 × 10−3 −8:26 × 10−3
able only in the electronic edition. 0.500 −2:04 × 10−3 −7:36 × 10−3
0.650 −1:64 × 10−3 −6:45 × 10−3
0.800 −1:43 × 10−3 −5:85 × 10−3
1.000 −1:26 × 10−3 −5:13 × 10−3
1.300 −1:06 × 10−3 −4:35 × 10−3
1.600 −1:17 × 10−3 −3:90 × 10−3
2.000 −1:02 × 10−3 −3:36 × 10−3
2.500 −1:06 × 10−3 −3:01 × 10−3
3.000 −1:09 × 10−3 −2:72 × 10−3
4.000 −1:30 × 10−3 −2:12 × 10−3
5.000 −9:35 × 10−4 −1:70 × 10−3
6.500 −7:87 × 10−4 −1:31 × 10−3
8.000 −6:43 × 10−4 −1:06 × 10−3
10.000 −3:65 × 10−4 −8:49 × 10−4
PGA −4:67 × 10−3 −9:81 × 10−3
PGV −2:79 × 10−3 −6:31 × 10−3

made in Figures 9 and 10, we constrain the model to attain

Δσ
300 bar for M ≥ 5 and d ≥ 10 km. The mean value of
the stress parameter for earthquakes in CENA is expressed as

ln Δσ CENA
5:704  min0; 0:29d − 10

Figure 8. Minimum usable frequency (f min ) model considered  min0; 0:229M − 5: 21
for records in CENA (solid line). Squares indicate the geometric
mean of the factored low-cut filter frequencies (i.e., 1:25flc ) deter-
mined for evenly spaced magnitude bins. The error bars represent The estimates of equation (21) for different magnitudes and
one standard deviation about the mean values. The dashed line in- depths are shown in Figures 9 and 10. The mean residuals
dicates the fmin model used for California by Yenier and Atkinson between the observed and predicted Δσ values attain values
(2015). The dotted line shows the corner frequency of the Brune
around zero, as illustrated in Figure 11. Overall, the proposed
(1970) source model for Δσ
100 bar. The color version of this
figure is available only in the electronic edition. Δσ model provides a good agreement with the Δσ values
determined from CENA events based on the inferred spectral
the distinct separation of depth-clustered stresses. For M > 5, shape.
the stress parameter attains a value of Δσ ≈ 300 bar, on aver-
age; we note that this is about three times the average stress
Calibration Factor
for California events.
We regress the best-fitting Δσ values to develop a An overall calibration factor is needed to reconcile the
regional stress model for CENA. Based on the observations predictions with observed amplitudes in the target region,
BSSA Early Edition
14 E. Yenier and G. M. Atkinson

Figure 9. Best-fitting stress parameters (Δσ) determined for
CENA events as a function of focal depth (d). The Δσ values are
clustered into different magnitude bins as shown in the legend.
Hatched symbols indicate Δσ values obtained from the potentially
induced events. Diamonds represent the mean Δσ calculated for
evenly spaced focal depth bins over all magnitudes, and the error
bars show standard error about the mean stress. Lines indicate the
derived Δσ model (equation 21) evaluated for M 3 (solid) and M 5
(dashed). The color version of this figure is available only in the
electronic edition.
Figure 10. Best-fitting stress parameters (Δσ) determined
based on matching the observed response spectral shape for CENA
events, as a function of magnitude. The Δσ values are clustered into
different focal depth (d) bins as shown in the legend. Hatched sym-
bols show Δσ values obtained from the potentially induced events.
Lines indicate the derived Δσ model (equation 21) evaluated for
d
2:5 km (dotted), d
7:5 km (dashed), and d ≥ 10 km (solid).
The color version of this figure is available only in the electronic
edition.

accounting for effects missing and/or different in simulations
(e.g., discrepancies between the assumed and true values of
crustal properties, site amplification, κ 0 , and path duration).
We calculate the calibration factor based on the analysis of
residuals:
δij
ln Y B=C;ij
− FM;i  FΔσCENA;i  FZ;ij  γ CENA Drup;ij ; 22
in which δij represents the residual for the ground motion
obtained from event i at station j, for a given spectral period
or peak motion. FΔσCENA;i is the stress adjustment factor
evaluated for Δσ from equation (21) for the known magni-
tude and focal depth of event i. The last term accounts for the
regional anelastic attenuation determined earlier. P
Figure 12 shows the event residuals (δi
δij =ni , in
which ni is the number of records obtained from event i;
ni ≥ 3 at a given period) as a function of magnitude. δi generally
attains negative values and appears to be randomly distributed,
showing no distinct attributes for natural and potentially induced
events. In general, the mean δi values determined at evenly
spaced magnitude bins shows no magnitude-dependent trends.

Figure 11. Residuals between the best-fitting Δσ values obtained from CENA events and the estimates of the Δσ model (equation 21)
evaluated for the known magnitudes and focal depths of the study events. Residuals are clustered into different magnitude and depth bins in
graphs (a) and (b), respectively. The color version of this figure is available only in the electronic edition.

BSSA Early Edition

Regionally Adjustable Generic GMPE Based on Equivalent Point-Source Simulations 15

Figure 12. Average of residuals determined for each event that have at least three observations at a given period (δi , circles). Diamonds
show mean of δi values determined for evenly spaced magnitude bins, and error bars represent the standard error about the mean. Dashed
lines indicate the event-based calibration factors (Ce;CENA ) that is defined as the average of δi values over all magnitudes, at a given period.
The color version of this figure is available only in the electronic edition.

This suggests that the magnitude scaling of ground motions in

CENA is well captured by the FM function, at least for the
available data. Based on these observations, we calculate an
event-based calibration factor (Ce;CENA ) as the average of δi
values over all magnitudes, for each spectral period or peak
motion. The Ce;CENA term fluctuates with period between 0
and −0:5 (ln units) for periods T < 3 s and attains positive
values with an increasing trend at larger periods, as shown
in Figure 13. This increasing trend at long periods may be
due to the fact that stochastic simulations are inherently
limited in their ability to generate the coherent motions seen
at long periods, and do not adequately model surface-wave
effects. We describe Ce;CENA as
8
< −0:25  max0; 0:39 lnT=2 T ≤ 10 s
Ce;CENA
−0:25 for PGA :
:
−0:21 for PGV
23 Figure 13. Event-based calibration factor for CENA (equa-
tion 23, solid line). Circles indicate average Ce;CENA values deter-
We subtract the event-based δi term from the individual mined for all events at each period, and error bars represent the
residuals standard error about the mean.
P to calculate the average residual at each station (i.e.,
δj
δij − δi =nj , in which nj is the number of observa-
tions at station j; nj ≥ 3 at a given period). Figure 14 illus- value suggesting that the BSSA14 site-effect model is rea-
trates the variation of δj as a function of V S30 . In general, the sonable for this site class. However, ground motions on
mean δj determined for NEHRP C sites attain near-zero NEHRP B sites are underpredicted by ∼15% and ground

BSSA Early Edition

16 E. Yenier and G. M. Atkinson

Figure 14. Event-corrected average residuals for each station (δj , circles) as a function of V S30 . Mean of δj values for NEHRP site classes
are shown by squares (standard errors for the mean values are smaller than the symbols). The color version of this figure is available only in
the electronic edition.

motions at NEHRP D sites are overpredicted by ∼20%, on result in some overestimation of CENA motions for Drup <
average. The mean δj determined for NEHRP A sites is near 150 km because the presumed WNA path-duration model is
zero, except for short periods. For T < 0:1 s, ground motions implicitly carried via the FZ function to CENA.
at NEHRP A sites are underpredicted by ∼20%, on average. We consider an additional minor calibration term for
Based on the observations made in Figure 14, we deduce that regional differences in the path duration. We describe this
the BSSA14 site-effect model is reasonably applicable to the path-related calibration (Cp;CENA ) as
CENA data in the NGA-East database (at least when applied

in combination with the determined calibration constants), Δb3 lnR=150 R ≤ 150 km
Cp;CENA
; 24
within an acceptable error band (20%), on average. 0 R > 150 km
Finally, we correct the individual residuals for the event
and station terms (i.e., δ′ij
δij − δi − δj ) to assess the per- in which Δb3 represents a calibration for the geometrical at-
formance of the assumed geometrical spreading function. tenuation rate in response spectrum domain. We determine
Figure 15 compares the δ′ij values as a function of rupture the Δb3 term from the regression of δ′ij based on equa-
distance. The mean δ′ij determined for log-spaced distance tion (24) at each period and peak motion separately. Figure 16
bins attains near-zero values at Drup > 150 km, suggesting shows the variation of Δb3 coefficients as a function of
that γ CENA parameter can successfully represent the overall period. Its value could be determined only up to T
3 s
attenuation at far distances. However, the mean δ′ij deviates due to the limited data at Drup < 100 km for longer periods.
from the horizontal zero line and decreases with distance for We smooth Δb3 values as
Drup < 150 km, as shown in the figure. This discrepancy
might be attributable to the path-duration model. In the sim- Δb3
ulations, we used a path-duration model derived primarily 8 minf0:095;0:030max0;0:095lnT=0:065g T ≤10s
<
from observed motions in WNA. Boore and Thompson
0:030 for PGA :
(2015) recently reported that the path duration in eastern :
0:052 for PGV
North America (ENA) is much longer than that in WNA, par-
ticularly at distances less than 150 km. This difference could 25

BSSA Early Edition

Regionally Adjustable Generic GMPE Based on Equivalent Point-Source Simulations 17

Figure 15. Event- and site-corrected residuals (δ′ij ) as a function of distance, for ground motions obtained from natural and potentially
induced events. Squares show the mean δ′ij values determined for logarithmically spaced distance bins and error bars indicate the standard
error about the mean. Solid line represents the fitted path-related calibration model for CENA (Cp;CENA , equation 24). The color version of this
figure is available only in the electronic edition.

The total calibration needed for the CENA-adjusted

GMPE is the summation of Ce;CENA and Cp;CENA terms. This
closes the systematic gaps between simulation-based predic-
tions and observed motions in CENA. The resultant CENA-
adjusted prediction equation is given as

ln Y CENA
FM  FΔσCENA  FZ  γ CENA Drup  FS

 Ce;CENA  Cp;CENA : 26

For example, the median PGA that would be expected at

a NEHRP B/C site (V S30
760 m=s) located at Drup

10 km from an M 6 event (focal depth, d
10 km) in CENA
can be predicted by plugging the model coefficients
(Tables 2–4) into equation (26): PGA
exp2:33  0:71−
3:54 − 0:05  0 − 0:25 − 0:07
0:42g. Figure 17 illus- Figure 16. The Δb3 values determined from regression analy-
trates PSA values predicted from equation (26) for magnitudes sis (circles) and the Δb3 smoothed model for CENA (equation 25,
M 4–8 as a function of rupture distance, for NEHRP B/C site solid line).
condition (V S30
760 m=s). The B/C-corrected ground mo-
tions obtained from earthquakes in CENA are also shown in Discussion
the figure for two magnitude ranges: M 3.5–4.5 and M 4.5–
5.5. The CENA-adjusted GMPE is in good agreement with the To enable the comparison of expected motions in eastern
empirical data, where available, and provides seismologically and western North America, we repeat the calibration exer-
informed predictions of average ground motions for moderate- cise described above for the California events contained in
to-large magnitudes (M > 6). the NGA-West 2 database (Ancheta et al., 2014). Following
BSSA Early Edition
18 E. Yenier and G. M. Atkinson

Figure 17. Pseudospectral acceleration (PSA) predictions from the CENA-adjusted GMPE (equation 26) for magnitudes M 4–8 (focal
depth, d
10 km), for V S30
760 m=s (lines). Circles represent the B/C-corrected ground motions obtained from earthquakes in CENA for
two magnitude ranges: M 3.5–4.5 and M 4.5–5.5. The color version of this figure is available only in the electronic edition.

Yenier and Atkinson (2015), we assume that the geometrical ln Δσ California

spreading in California can be modeled as R−1:3 at distances
< 50 km and R−0:5 at farther distances (this does a better job
at matching attenuation trends than the traditional model 1=R
model at distances < 50 km, particularly for M < 5:5
events). Therefore, we use the generic bilinear Z model
(b1
−1:3, b2
−0:5, and Rt
50 km) without modifica-
tion. The only attenuation adjustment needed is for the
regional anelastic attenuation. Following the methodology
described above, we compute the regional anelastic attenu-
ation coefficient and overall calibration factor for California
based on the response spectra and peak motions compiled
from the NGA-West 2 flatfile (Ancheta et al., 2014). We
adopt the stress parameter model developed for California
earthquakes Δσ California by Yenier and Atkinson (2015):
BSSA Early Edition

2:18  minf0; max0:06; 0:3 − 0:04Md − 12g ln 10;
27
in which d is the focal depth in kilometers. The resultant
California-adjusted prediction equation is given as
ln Y California
FM  FΔσ;California  FZ  γ California Drup
 FS  CCalifornia ; 28
in which FΔσ;California is the stress adjustment factor evaluated
for the mean stress in California using equation (27). The
anelastic attenuation coefficients γ California are provided in
Table 4. We define the overall calibration factor for the
region as
Regionally Adjustable Generic GMPE Based on Equivalent Point-Source Simulations 19

Figure 18. Comparison of PSA predictions for CENA (equation 26, solid line) and California (equation 28, dashed line) for M 5 and M 8,
for focal depth d
10 km and NEHRP B/C site condition (V S30
760 m=s). The color version of this figure is available only in the
electronic edition.

CCalifornia
8 Conclusion
>
> max−0:25; − 0:25  0:36lnT=0:1 T ≤ 0:2 s
< We conclude that the generic GMPE approach provides a
max0; 0:39lnT=1:5 0:2 s < T ≤ 10 s

:
>
> −0:25 for PGA calibrated model of predicted ground motions in CENA that
:
−0:15 for PGV agrees with average motions from the NGA-East database,
and is constrained by simulation-based scaling principles
29
that have been demonstrated to work in California over a
The adjusted California model does not require calibra- wide range of magnitudes and distances. We have provided
tion for the path effects (i.e., Cp;California
0) because calibrated median predictions of ground motions in CENA for
simulations that were used for the derivation of the generic average horizontal-component peak ground motions and
GMPE were generated based on the WNA-compatible dura- 5% damped response spectra (up to T
10 s), for magni-
tion model of Boore and Thompson (2014). Figures 18 and tudes M 3–8 and distances < 600 km.
19 illustrate the comparison of PSA predictions for CENA and The approach that we have taken, in casting our model
California as a function of distance and period, respectively. into a framework that is parameterized by the basic seismo-
The effects of differences in regional stress parameter and logical parameters of moment, stress, and attenuation, has
anelastic attenuation between California and CENA are both conceptual and practical advantages. The generic GMPE
apparent in these figures. effectively decouples the influence of basic source and
BSSA Early Edition
20
Figure 19. Comparison of predicted response spectra for CENA (equation 26, solid line) and California (equation 28, dashed line) at
Drup
10 km and Drup
100 km, for M 4–8. The response spectra are computed for focal depth d
10 km and NEHRP B/C site con-
dition (V S30
760 m=s). The color version of this figure is available only in the electronic edition.
attenuation parameters on ground-motion amplitudes. This al-
lows the determination of regional values of modeling param-
eters from observed response spectral data in the target region.
The generic GMPE proposed in this study is a self-adjusting
model, and is readily calibrated to the target region once the
regional parameter values are plugged into the model. Thus, it
does not require performing ground-motion simulations to
determine regional adjustment factors while ensuring that
the determined model will be robust and sound in its scaling
behavior. In addition, the generic values of source and attenu-
ation parameters that are implicitly carried into the generic
model are known. Therefore, the modeling parameters that re-
quire modification for regional use can easily be determined.
This provides effective and transparent control over the trans-
ferable factors between regions. Another advantage of the
generic GMPE is that we can easily create understandable
and documentable alternative GMPEs, by considering a range
of possible parameter values that might be reasonable for the
region (or a subset of the region). For each parameter set, we
may use the empirical data to derive a new calibration factor
for each frequency, such that the overall residuals are mini-
mized for the given model. Analysis of the residual trends
and their variability under the alternative models then provides
information on the limitations of the alternative parameter sets.
This allows modeling of the epistemic uncertainty in GMPEs,
which is an important consideration for probabilistic seismic-
hazard analysis applications.
Data and Resources
Electronic versions of Tables 2, 3, and 4 are available at
http://www.seismotoolbox.ca (last accessed May 2015). We
compiled ground motions for central and eastern North
America (CENA) earthquakes from the Next Generation At-
BSSA Early Edition
E. Yenier and G. M. Atkinson
tenuation (NGA)-East flatfile provided by Christine A. Gou-
let (written comm., 2014), and ground motions for California
earthquakes from the NGA-West 2 flatfile available at http://
peer.berkeley.edu/ngawest2/databases (last accessed July
2014). Ground-motion simulations were performed using
the Stochastic-Method SIMulation (SMSIM) version 3.8 soft-
ware that is available at http://www.daveboore.com/software_
online.html (last accessed October 2014). All graphics were
produced using CoPlot software (www.cohort.com; last ac-
cessed October 2014).
Acknowledgments
This study was supported by the Natural Sciences and Engineering
Research Council of Canada, by the Next Generation Attenuation-East
project, and by the U.S. Geological Survey National Earthquake Hazards
Reduction Program under Grant Number G13AP00067. We would express
our sincere thanks to Annemarie Baltay, John Douglas, Basil Margaris, Rob-
ert Mereu, Shahram Pezeshk, and Kristy Tiampo for their useful comments
that enabled us to improve the draft manuscript.
References
Allen, T., P. Cummins, T. Dhu, and J. Schneider (2007). Attenuation of
ground-motion spectral amplitudes in southeastern Australia, Bull.
Seismol. Soc. Am. 97, 1279–1292.
Ancheta, T., R. Darragh, J. Stewart, E. Seyhan, W. Silva, B. Chiou,
K. Wooddell, R. Graves, A. Kottke, D. Boore, et al. (2014). PEER
NGAWest2 database, Earthq. Spectra 30, 989–1006.
Atkinson, G. M. (2004). Empirical attenuation of ground-motion spectral
amplitudes in southeastern Canada and the northeastern United States,
Bull. Seismol. Soc. Am. 94, 1079–1095.
Atkinson, G. M. (2008). Ground motion prediction equations for eastern
North America from a referenced empirical approach: Implications
for epistemic uncertainty, Bull. Seismol. Soc. Am. 98, 1304–1318.
Atkinson, G. M. (2010). Ground-motion prediction equations for Hawaii from a
referenced empirical approach, Bull. Seismol. Soc. Am. 101, 1304–1318.
Atkinson, G. M. (2012). Evaluation of attenuation models for the northeastern
United States/southeastern Canada, Seismol. Res. Lett. 83, 166–178.
Regionally Adjustable Generic GMPE Based on Equivalent Point-Source Simulations
Atkinson, G. M., and D. M. Boore (1995). Ground-motion relations for
eastern North America, Bull. Seismol. Soc. Am. 85, 17–30.
Atkinson, G. M., and D. M. Boore (2006). Earthquake ground-motion pre-
diction equations for eastern North America, Bull. Seismol. Soc. Am.
96, 2181–2205.
Atkinson, G. M., and D. M. Boore (2011). Modification to existing ground-
motion prediction equations in light of new data, Bull. Seismol. Soc.
Am. 101, 1121–1135.
Atkinson, G. M., and D. M. Boore (2014). The attenuation of Fourier am-
plitudes for rock sites in eastern North America, Bull. Seismol. Soc.
Am. 104, 513–528.
Atkinson, G. M., and D. Motazedian (2013). Ground-motion amplitudes for
earthquakes in Puerto Rico, Bull. Seismol. Soc. Am. 103, 1846–1859.
Atkinson, G. M., and W. Silva (2000). Stochastic modeling of California
ground motions, Bull. Seismol. Soc. Am. 90, 255–274.
Babaie Mahani, A., and G. M. Atkinson (2012). Evaluation of functional forms
for the attenuation of small-to-moderate-earthquake response spectral
amplitudes in North America, Bull. Seismol. Soc. Am. 102, 2714–2726.
Boore, D. M. (2003). Simulation of ground motion using the stochastic
method, Pure Appl. Geophys. 160, 635–676.
Boore, D. M. (2005). SMSIM—Fortran programs for simulating ground mo-
tions from earthquakes: Version 2.3—A revision of OFR 96-80-A,
U.S. Geol. Surv. Open-File Rept. 00-509, 59 pp.
Boore, D. M. (2009). Comparing stochastic point-source and finite-source
ground motion simulations: SMSIM and EXSIM, Bull. Seismol. Soc.
Am. 99, 3202–3216.
Boore, D. M. (2010). Orientation-independent, non geometric-mean mea-
sures of seismic intensity from two horizontal components of motion,
Bull. Seismol. Soc. Am. 100, 1830–1835.
Boore, D. M., and E. M. Thompson (2014). Path durations for use in the
stochastic-method simulation of ground motions, Bull. Seismol.
Soc. Am. 104, 2541–2552.
Boore, D. M., and E. M. Thompson (2015). Revisions to some parameters
used in stochastic-method simulations of ground motion, Bull. Seis-
mol. Soc. Am. 105, 1029–1041.
Boore, D. M., K. W. Campbell, and G. M. Atkinson (2010). Determination
of stress parameters for eight well-recorded earthquakes in eastern
North America, Bull. Seismol. Soc. Am. 100, 1632–1645.
Boore, D. M., C. Di Alessandro, and N. A. Abrahamson (2014). A gener-
alization of the double-corner-frequency source spectral model and its
use in the SCEC BBP validation exercise, Bull. Seismol. Soc. Am. 104,
2387–2398.
Boore, D. M., J. P. Stewart, E. Seyhan, and G. M. Atkinson (2014). NGA-
West2 equations for predicting response spectral accelerations for shal-
low crustal earthquakes, Earthq. Spectra 30, 1057–1085.
Brune, J. N. (1970). Tectonic stress and the spectra of seismic shear waves
from earthquakes, J. Geophys. Res. 75, 4997–5009.
Burger, R. W., P. G. Somerville, J. S. Barker, R. B. Herrmann, and D. V.
Helmberger (1987). The effect of crustal structure on strong ground
motion attenuation relations in eastern North America, Bull. Seismol.
Soc. Am. 77, 420–439.
Campbell, K. W. (2002). Development of semi-empirical attenuation rela-
tionships for the CEUS, Report to U.S. Geological Survey, NEHRP
Award Number 01HQGR0011, 38 pp.
Campbell, K. W. (2003). Prediction of strong ground motion using the
hybrid empirical method and its use in the development of ground-
motion (attenuation) relations in eastern North America, Bull. Seismol.
Soc. Am. 93, 1012–1033.
Chapman, M. C. (2013). On the rupture process of the 23 August 2011
Virginia earthquake, Bull. Seismol. Soc. Am. 103, 613–628.
Chapman, M. C., and R. W. Godbee (2012). Modeling geometrical spread-
ing and the relative amplitudes of vertical and horizontal high-
frequency ground motions in eastern North America, Bull. Seismol.
Soc. Am. 102, 1957–1975.
Electric Power Research Institute (EPRI) (2004). CEUS ground motion
project final report, Technical Report 1009684, Electric Power
Research Institute, Dominion Energy, Glen Allen, Virginia, Entergy
BSSA Early Edition
21
Nuclear, Jackson, Mississippi, and Exelon Generation Company,
Kennett Square, Pennsylvania. Palo Alto, California.
Frankel, A. (2009). A constant stress-drop model for producing broadband
synthetic seismograms: Comparison with the Next Generation Attenu-
ation relations, Bull. Seismol. Soc. Am. 99, 664–681.
Hassani, B., and G. M. Atkinson (2015). Referenced empirical ground-mo-
tion model for eastern North America, Seismol. Res. Lett. 86, 477–491.
National Earthquake Hazards Reduction Program (NEHRP) (2000). NEHRP
Recommended Provisions for Seismic Regulations for New Buildings
and Other Structures (FEMA 368), Part 1: Provisions, Federal
Emergency Management Agency, Washington, D.C.
Ojo, S., and R. Mereu (1986). The effect of random velocity functions on the
travel times and amplitudes of seismic waves, Geophys. J. Roy. Astron.
Soc. 84, 607–618.
Ou, G.-B., and R. B. Herrmann (1990). A statistical model for ground
motion produced by earthquakes at local and regional distances, Bull.
Seismol. Soc. Am. 80, 1397–1417.
Pezeshk, S., A. Zandieh, and B. Tavakoli (2011). Hybrid empirical ground-
motion prediction equations for eastern North America using NGA
models and updated seismological parameters, Bull. Seismol. Soc.
Am. 101, 1859–1870.
Scherbaum, F., J. J. Bommer, H. Bungum, F. Cotton, and N. A. Abrahamson
(2005). Composite ground motion models and logic trees: Methodology,
sensitivities, and uncertainties, Bull. Seismol. Soc. Am. 95, 1575–1593.
Seyhan, E., and J. P. Stewart (2014). Semi-empirical nonlinear site ampli-
fication from NGAWest2 data and simulations, Earthq. Spectra. 30,
1241–1256.
Silva, W., N. Gregor, and R. Darragh (2002). Development of regional hard
rock attenuation relations for central and eastern North America,
Report to Pacific Engineering and Analysis, 57 pp.
Somerville, P., N. Collins, N. Abrahamson, R. Graves, and C. Saikia (2001).
Ground motion attenuation relations for the central and eastern United
States, Report to U.S. Geological Survey, NEHRP Award Number
99HQGR0098, 38 pp.
Somerville, P., R. Graves, N. Collins, S. G. Song, S. Ni, and P. Cummins
(2009). Source and ground motion models of Australian earthquakes,
Proc. of the 2009 Annual Conference of the Australian Earthquake
Engineering Society, Newcastle, Australia, December 2009.
Somerville, P., J. McLaren, C. Saikia, and D. Helmberger (1990). The 25
November 1988 Saguenay, Quebec earthquake: Source parameters
and the attenuation of strong ground motion, Bull. Seismol. Soc.
Am. 80, 1118–1143.
Toro, G., N. Abrahamson, and J. Schneider (1997). Model of strong ground
motion in eastern and central North America: Best estimates and
uncertainties, Seismol. Res. Lett. 68, 41–57.
Yenier, E., and G. M. Atkinson (2014). Equivalent point-source modeling of
moderate-to-large magnitude earthquakes and associated ground-
motion saturation effects, Bull. Seismol. Soc. Am. 104, 1458–1478.
Yenier, E., and G. M. Atkinson (2015). An equivalent point-source model for
stochastic simulation of earthquake ground motions in California,
Bull. Seismol. Soc. Am. 105, 1435–1455, doi: 10.1785/0120140254.
Youssef, M. A. H., A. R. Kottke, J. P. Stewart, K. W. Campbell, B. Kim, C.
Moss, S. Nikolaou, E. M. Rathje, and W. J. Silva (2014). Reference
rock site condition for central and eastern North America, Bull.
Seismol. Soc. Am. 104, 684–701.
Department of Earth Sciences
Western University
1151 Richmond Street
London, Ontario
Canada N6A 5B7
emrah.yenier@gmail.com
gmatkinson@aol.com
Manuscript received 3 November 2014;
Published Online 21 July 2015