Mechanisms for Generation of Near-Fault Ground Motion Pulses for Dip-Slip Faulting

NATALIA POIATA,1 HIROE MIYAKE,2 and KAZUKI KOKETSU2

Abstract—We analyzed the seismological aspects of the near-

fault ground motion pulses and studied the main characteristics of

the rupture configuration that contribute to the pulse generation for

dip-slip faulting events by performing forward simulations in

broadband and low-frequency ranges for different rupture scenarios

of the 2009 L’Aquila, Italy (Mw 6.3) earthquake. The rupture

scenarios were based on the broadband source model determined

by Poiata et al. (Geophys J Int 191:224–242, 2012). Our analyses

demonstrated that ground motion pulses affect spectral character-

istics of the observed ground motions at longer periods, generating

significantly larger seismic demands on the structures than ordinary

records. The results of the rupture scenario simulations revealed the

rupture directivity effect, the radial rupture propagation toward the

site, and the focusing effect as the main mechanisms of the near-

fault ground motion pulse generation. The predominance of one of

these mechanisms depends on the location of the site relative to the

causative fault plane. The analysis also provides the main candidate

mechanisms for the worst-case rupture scenarios of pulse genera-

tion for the city of L’Aquila and, more generally, the hanging-wall

sites located above the area of large slip (strong motion generation

area).

Key words: Ground motion pulses, dip-slip faulting, direc-

tivity effect, focusing effect, 2009 L’Aquila earthquake, worst-case

rupture scenario.

1. Introduction

Seismic recordings of past earthquakes indicate

that near-fault ground motions could be significantly

different from those observed at larger distances from

seismic faults. The presence of strong coherent long-

period pulses recorded at some stations, correspond-

ing to a specific geometry of the fault-station

configuration, was reported as a distinctive charac-

teristic of the near-fault ground motions (e.g., Aki

1968; Somerville et al. 1997; Koketsu and Miyake

2008). The rapid development of strong motion net-

works allowed to capture long-period ground motion

pulses near seismic faults during large damaging

events like the 1999 Kocaeli (Mw 7.6) earthquake and

the 1999 Chi–Chi (Mw 7.6) earthquake, as well as

during smaller events such as the 1994 Northridge

(Mw 6.7), 1995 Kobe (Mw 6.9), and 2003 Bam (Mw

6.6) earthquakes. Most of these earthquakes resulted

in substantial material damage and loss of human

lives. The effect that near-fault ground motion pulses

can have on the engineered structures was first

revealed during the 1994 Northridge earthquake. It

was recognized that the considerable damage that

was observed as the result of this earthquake was

caused by large pulse-like ground motions recorded

in the near-source area (e.g., Heaton et al. 1995;

Strasser and Bommer 2009). The building codes that

existed during that period did not consider any cur-

rently known near-source effects. An example

illustrating the importance of including the near-

source effects into the building practice of critical

facilities was given by the experience of the Kashi-

wazaki-Kariwa nuclear power plant during the 2007

Chuetsu-oki (Mw 6.6) earthquake in Japan. This

reverse faulting event was the world’s first major

earthquake that occurred on a causative fault

extending beneath a nuclear power plant (Miyake

et al. 2010). The ground motions recorded by the

instruments inside the plant and characterized by

three seismic pulses were significantly stronger than

those that were anticipated at the time of its design.

The detailed analyses of the earthquake’s source

process by Miyake et al. (2010) determined that the

observed ground motion pulses corresponded to the

combined effect of the distribution of asperities (ar-

eas of large slip), rupture propagation, and the

S-wave radiation pattern. Fortunately, the plant

1 National Institute for Earth Physics, 12 Calugareni, 077125

Magurele, Ilfov, Romania. E-mail: natalia@infp.ro2 Earthquake Research Institute, University of Tokyo, 1-1-1

Yayoi, Bunkyo-ku, Tokyo 113-0032, Japan.

Pure Appl. Geophys.

� 2017 Springer International Publishing

DOI 10.1007/s00024-017-1540-z Pure and Applied Geophysics

structures suffered only minor damage due to the

ground motion.

Long-period ground motion pulses recorded in

near-source regions are mainly attributed to the rup-

ture directivity effect. This phenomenon can be

observed in both strike-slip and dip-slip faulting (e.g.,

Koketsu et al. 2016). The rupture directivity effect is

observed when the rupture front propagates over the

earthquake source fault at high speed, typically

slightly less than the shear wave (S-wave) velocity of

the media. In this case, if a site is located in the

direction of rupture, most seismic energy of the wave

front will arrive in a single pulse of ground motion

(Fig. 1). The conditions contributing to the forward

directivity effect and the characteristics of the rupture

directivity pulses were identified and summarized by

Somerville et al. (1997) based on the near-fault

records of the 1992 Landers earthquake. The theo-

retical aspects of the physical mechanism describing

their generation were addressed in numerous studies

(e.g., Boore and Joyner 1978, 1989; Heaton 1990;

Joyner 1991; Miyatake 1998). It was also pointed out

that the forward directivity effect generates pulses

with dominant periods of C0.6 s, strongly affecting

the spectral content of the ground motions (e.g.,

Somerville et al. 1997; Somerville 2003; Koketsu and

Miyake 2008), and placing significant demands on

the structures near the earthquake source fault.

Moreover, Somerville (2003) indicated that the per-

iod of the pulse increases with the magnitude of the

event. Thus, even smaller earthquakes are able to

generate ground motions that could result in

increased seismic demand and have damaging effect

on structures in the near-fault areas.

The rupture directivity effect as described by

Somerville et al. (1997) explains well the fault-nor-

mal components observed during past strike-slip

faulting events (e.g., 1992 Landers). In case of dip-

slip faulting, however, both the near-fault ground

motion pulses and the hanging-wall effect (Abra-

hamson and Somerville 1996) are observed at some

stations. Inspection of 3-D S-wave radiation patterns

for dip-slip faulting (Fig. 2), as well as the near-fault

strong ground motion records from the stations on the

hanging-wall sites above the causative fault of the

2007 Chuetsu-oki (reverse faulting; Miyake et al.

2010), 2009 L’Aquila (normal faulting) and 2015

Gorkha (low-angle dip-slip faulting, Koketsu et al.

Forward directivity

Backward directivity

Direction of rupturepropagation

Wavefronts of S-wave

Site ASite B

Figure 1Snapshot of the S-wave fronts, illustrating the rupture directivity effect. Site A is located in the direction of the rupture propagation, while Site

B is located in the direction opposite to that of the rupture propagation. The propagating rupture is represented by a finite fault comprising four

point sources (white stars) placed at a distance R from each other; the rupture is assumed to start from the leftmost point and propagate toward

the right (black arrow) at a velocity Vr = 0.9Vs. The theoretically calculated synthetic waveforms from each of the four sources (black traces)

and the resulting total velocity waveforms (red traces) are shown for Sites A and B (marked as blue triangles)

N. Poiata et al. Pure Appl. Geophys.

Koketsu et al. 2016) earthquakes pointed toward a

more complex mechanism of pulse generation for

such events.

In this study, we use the observed near-field

ground motions of the 1992 Landers (Mw 7.3) strike-

slip and 2009 L’Aquila (Mw 6.3) normal faulting

earthquakes to demonstrate that the ground motion

pulses can significantly affect the shape of the

response spectra. This effect leads to an increase of

seismic demand on structures being potentially the

main cause of structural damage, such as those

observed during the analysed events. We also inves-

tigate the rupture configuration parameters that

contribute to the generation of the near-fault ground

motion pulses in dip-slip faulting events. This is

achieved by performing ground motion simulations

of the recordings at near-fault strong motion stations

in both broadband and low-frequency ranges using

the specific rupture scenarios based on the 2009

L’Aquila (Mw 6.3) earthquake source model. The

(a) (b)

(c)

δhanging wall

foot wall

X X

along strike rupture propagation

Site ASite B

Site A Site A

FN FN

Site A Site B Site A Site B

up-dip rupture propagation

(d) (e)

(f) (g)

up-dipalong strike

Figure 2a 3-D low-frequency and b high-frequency S-wave radiation patterns from a point source shear dislocation. The rectangles indicate the planes

of faulting and the arrows show the slip directions. c Schematic of a normal fault. The red star corresponds to the rupture starting point, the

up-dip and along-strike directions are shown by the full black arrows, and the slip directions by the open arrows. The blue triangles indicate

the location of the observation sites. Site A is located on the hanging-wall side and above the rupture plane, and Site B is located on the

footwall side of the fault. d Vertical section of the low-frequency radiation pattern of the up-dip rupture propagation for the normal fault (c).

The shaded area corresponds to the region of the maximum forward directivity effect, and the black arrow shows the orientation of the

directivity pulses observed in Site B. e Same as d, but for the high-frequency range. f Vertical section of the low-frequency radiation pattern of

the along-strike rupture propagation for the normal fault. c The slip direction is shown in circles next to the rupture starting point (red star).

g Same as f, but for the high-frequency range

Ground Motion Pulses for Dip-Slip Faulting

assumed scenarios are designed using the broadband

source model of Poiata et al. (2012). We further

analyzed the results of ground motion simulations

identifying the rupture configurations that are most

efficient in generating near-fault ground motion pul-

ses for the city of L’Aquila by comparing the

simulated and observed ground motions recorded at

the AQU strong motion station located inside the city.

1.1. Near-Fault Ground Motion Pulses

An earthquake results from shear dislocation

propagating along the fault. The 3-D low-frequency

radiation pattern of the S-wave generated by a point

shear dislocation on the horizontal plane is shown in

Fig. 2a (for a full mathematical description, see Aki

and Richards 1980). Moreover, high-quality wave-

form records from the stations at small and

intermediate distances from seismic sources indicated

that in the higher frequency range of 1–3 Hz, the

four-lobed radiation pattern of the S-wave is distorted

(e.g., Liu and Helmberger 1985; Takenaka et al.

2003; Takemura et al. 2009). The distortion results in

an isotopic distribution of S-wave amplitudes in all

directions for frequencies [3 Hz (Liu and Helm-

berger 1985). This phenomenon was attributed to the

scattering of the seismic waves by the heterogeneous

structure around the source region (‘‘heterogeneities

in the source-process’’; Liu and Helmberger 1985)

and along the propagation path of the seismic waves

(Takemura et al. 2009). Based on the results of these

studies and assuming some degree of simplification,

we represent the high-frequency radiation of S-waves

generated by a point shear dislocation as a sphere

(Fig. 2b). The limit between the low- and high-

frequency here is assumed at around 1–3 Hz, follow-

ing the results of Liu and Helmberger (1985).

Close to the earthquake fault, where the point-

source approximation is not valid, one has to consider

the fact that an earthquake is a shear dislocation that

propagates along a finite fault. Mathematical simpli-

fication of this is equivalent to a distribution (e.g.,

along a line) of the point sources over the fault, and

the propagation of the dislocation can then be

introduced by considering the appropriate time delays

(e.g., Kasahara 1981). This will result in an azimuth

dependence of the arrival times and summary

amplitudes of the waveforms observed at different

locations surrounding the ruptured fault, which is also

known as directivity (e.g., Aki and Richards 1980;

Lay and Wallace 1995). The effect of rupture

directivity on the S-wave point-source radiation

pattern symmetry can be described by the directivity

function (e.g., Aki and Richards 1980). The effects of

directivity on the S-wave radiation pattern in the low-

and high-frequency ranges of the seismic wave field

in the case of a dip-slip fault are shown in Fig. 2d, e.

The resulting elongation of the radiation pattern in

the direction of the rupture propagation corresponds

to the forward directivity effect discussed previously.

The forward rupture directivity effect at a given

site occurs under the following conditions (Somer-

ville et al. 1997): the rupture front must propagate

toward the site, the direction of the slip on the fault

must be aligned with the direction of the rupture, and

the rupture must propagate at a speed close to that of

the S-wave. These conditions are most easily fulfilled

in the case of strike-slip faulting, when the rupture

typically propagates horizontally along the strike, and

the slip on the fault is oriented along the direction of

the strike. Evidence for the directivity effects is seen

in the fault-normal components of the seismograms

recorded during the past strike-slip faulting events

(e.g., 1992 Landers earthquake; Fig. 3a) confirming

the above description. Following Somerville et al.

(1997), it is possible to show that the conditions for

the forward directivity effect are also satisfied in the

case of dip-slip faulting for both reverse and normal

faults (Fig. 2c). The alignments of the rupture

direction and of the slip direction along the dip of

the fault (Fig. 2d, e) generate a rupture directivity

effect at the stations located close to the surface trace

of the fault (Somerville et al. 1997; Somerville 2003).

Figure 2d, e illustrates this effect by showing low-

and high-frequency radiation patterns for an up-dip

rupture propagation in the case of a normal-faulting

event. The rupture directivity pulses in the case of

dip-slip faulting event are observed in the direction

normal to the dip of the fault and have components in

the vertical and horizontal directions normal to the

strike.

However, in case of dip-slip faulting events, the

fault plane normally forms an angle (d-dip angle)

with the horizontal plane (Fig. 2c). This geometrical

N. Poiata et al. Pure Appl. Geophys.

−150

0

150V

el. (

cm/s

)

0 5 10 15 20 25 30 35 40 45

Time (sec)

1992 Landers; station Joshua Tree; backward directivity

0

50

100

150

200

250

300

SV

(cm

/s),

(da

mpi

ng=

5%)

0 1 2 3 4 5 6 7 8 9 10

Period (sec)

observed

−150

0

150

Vel

. (cm

/s)

1992 Landers; station Lucerne; forward directivity

−150

0

150

Vel

. (cm

/s)

extracted pulse

−150

0

150

Vel

. (cm

/s)

0 5 10 15 20 25 30 35 40 45

Time (sec)

residual

0

50

100

150

200

250

300

SV

(cm

/s),

(da

mpi

ng=

5%)

0 1 2 3 4 5 6 7 8 9 10

Period (sec)

observed observedextracted pulseresidual

observed

(a) (b)

(c)

(d) (e)

13˚15'E 13˚30'E

42˚1

5'N

5 km

L’Aquila

Castelnuovo

Paganica

Onna

Mw 6.3

AQK

SP

SN

observedextracted pulseresidual

−50

−50

−50

0

50

Vel

. (cm

/s) 2009 L’Aquila; station AQK

0

Vel

. (cm

/s)

0

Vel

. (cm

/s)

0 5 10 15 20

Time (sec)

0

50

100

150

SV

(cm

/s),

(da

mpi

ng=

5%)

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Period (sec)

extracted pulse

residual

observed

epicenter

rupture propagation

Lucerne

Joshua Tree

backward directivityregion

forward directivityregion

20 sec

Figure 3a Map of the Landers region indicating the causative faults of the 1992 Landers earthquake (modified from Somerville et al. 1997). The red

star shows the epicenter location and the black arrow the direction of rupture propagation. The fault-normal ground velocities for the Lucerne

forward directivity and the Joshua Tree backward directivity stations are also shown. b Left fault-normal component of the velocity record

(black trace) from Lucerne station recorded during the 1992 Landers earthquake. The extracted pulse (red trace) and the residual (gray trace)

waveforms are shown below. Right velocity response time-series spectra. c Left fault-normal component of the velocity record from the

Joshua Tree station recorded during the 1992 Landers earthquake. Right velocity response spectra of the corresponding time series. d Map of

the L’Aquila area indicating the location of the fault plane (black rectangle) ruptured during the 2009 L’Aquila earthquake. The red star

shows the location of the epicenter; the blue triangle shows the location of the near-fault hanging-wall station AQK. e Left strike normal

component of the velocity record (black) from the AQK station, recorded during the 2009 L’Aquila earthquake. The extracted pulse (red

trace) and the residual (gray trace) waveforms are shown below. Right velocity response spectra of the corresponding time series

Ground Motion Pulses for Dip-Slip Faulting

configuration generates an area on the free surface

that is located on the hanging-wall side of the fault

and above the ruptured fault plane. Considering the

3-D aspect of the S-wave radiation pattern (Fig. 2a,

b), a rupture propagating along the strike of a dip-slip

fault could affect the waveforms recorded at a station

on the hanging-wall and above the ruptured fault

plane. Figure 2f, g shows the low- and high-fre-

quency radiation patterns for along-strike rupture

propagation in case of a normal-faulting event. This

implies that pulse generation in a dip-slip faulting

event could be driven by a mechanism more complex

than the directivity effect, as it is described above.

Indeed, the record of pulse-like velocity ground

motions from near-fault stations located on the

hanging-wall side above the ruptured fault plane of

the recent 2009 L’Aquila (Mw 6.3) normal-faulting

event supports this idea. These stations were not

located near the surface trace of the fault, where the

forward directivity effect would be maximum for a

dip-slip fault. Furthermore, Kagawa (2009) made a

similar observation based on the record from the

Kashiwazaki-Kariwa Nuclear Power Plant during the

2007 Chuetsu-oki (Mw 6.6) reverse faulting

earthquake.

2. Methodology

2.1. Extraction of Long-Period Pulses

from the Observed Ground Motions

To demonstrate that near-fault ground motion

pulses can produce increased seismic demands on

structures, we illustrated their possible effect on the

spectral characteristics of the ground motions (implic-

itly their contribution to the seismic hazard). We

calculated the velocity response spectra of the fault-

normal components recorded at the two stations near

the causative fault of the 1992 Landers (Mw 7.3)

earthquake. The relative locations of the stations and

the fault are presented in Fig. 3a. For comparison, we

show both the record of the Lucerne station located in

the direction of the rupture propagation, which

recorded the corresponding long-period pulse (Fig. 3a,

b), and the backward directivity station of Joshua Tree

(Fig. 3a, c). These records are representative of the

effect of rupture directivity from a strike-slip fault,

illustrating both the forward and the backward direc-

tivity, on the near-fault ground motions (Somerville

et al. 1997). We decomposed the velocity record from

the Lucerne station into pulse and residual components

by applying the wavelet decomposition method of

Baker (2007) and Baker and Shahi (2007). Then, the

corresponding velocity response spectra were calcu-

lated (Fig. 3b) for all three resultant ground motions

and compared against each other. Same analysis for a

dip-slip faulting event is performed using the ground

motion of the 2009 L’Aquila (Mw 6.3) normal-faulting

event recorded at the AQK hanging-wall station. The

resulted decomposed and original observed velocity

response spectra for the two selected earthquakes are

discussed in Sect. 3.1.

2.2. Forward Simulation of Ground Motion Pulses

for Normal-Faulting Event

The 2009 L’Aquila, Italy (Mw 6.3) earthquake

was a first well-recorded normal-faulting event that

provided an unprecedented number of strong motion

observations in the near-fault area. The hypocenter of

the earthquake was located 2 km west of the city of

L’Aquila (e.g., Fig. 3d) at 42.3476�N, 13.3800�E and

a depth of 9.5 km (INGV 2009). In spite of its

moderate magnitude, this event caused extensive

damage to the cities and villages in the Abruzzo

region. Most of the damage was concentrated in the

city of L’Aquila, located close to the hypocentre, and

the villages of Castelnuovo, Onna, and Paganica

located further southeast (Fig. 3d). Thus, the 2009

L’Aquila earthquake offered an important case study

for the effects of damaging moderate-magnitude

earthquakes occurring in densely populated areas.

The observations of the 2009 L’Aquila mainshock

and its aftershocks recorded at the Italian strong

motion network (RAN), operated by the Civil Pro-

tection Department, and at MedNet station AQU have

been integrated into the ITalian ACcelerometric

Archive (ITACA). The high-quality digital acceler-

ation records of the mainshock included the near-

fault stations (at epicentral distances of 1.7–5.0 km)

on the hanging wall of the ruptured fault (Fig. 4a)

that recorded horizontal pick ground acceleration

(PGA) values of 0.35–0.65 g. The velocity records of

N. Poiata et al. Pure Appl. Geophys.

13˚24'E 13˚36'E

42˚1

2'N

42˚1

8'N

42˚2

4'N

5 km

AQGAQV

AQKAQU

GSA

L’Aquila

Onna

Paganica

Castelnuovo

0.0

0.2

0.4

0.6

0.8

1.0Slip (m)

Mw 6.3

(a) (b)EGFM modelscenario

w =

1.3

(km

)

l = 2.0 (km)(c)

(d)

(e)

(f)

Mw 4.6

Vp

Vs

Velocity (km/s)

0

5

10

15

20

25

30

Dep

th(k

m)

AQU

0

5

10

15

20

25

30

Dep

th(k

m)

Vp

Vs

Velocity (km/s)

GSA

Figure 4a Map of the L’Aquila area presenting the relative location of the surface projection of the rupture plane for the 2009 L’Aquila earthquake

(after Poiata et al. 2012) and the near-fault hanging-wall (AQG, AQV, AQU, and AQK) and footwall (GSA) stations. The red star indicates

the location of the epicenter reported by INGV. The red contoured black star indicates the location of the EGF event according to INGV. The

thick red line shows the location of the observed coseismic surface ruptures modified after Boncio et al. (2010), and the yellow squares mark

the cities that suffered the largest damage. The gray rectangle corresponds to the SMGA estimated from the EGF analysis of Poiata et al.

(2012). The areas green and blue rectangles correspond to the two rupture scenarios of the linear up-dip and the along-strike rupture

propagations used in the forward simulation analysis of the near-fault pulses. b–d Plane view of the fault settings for the rupture scenarios

used for the forward simulations. The rupture scenarios assumed different rupture starting points marked by gray stars. The white contoured

black star corresponds to the rupture starting point of the SMGA model (from Poiata et al. 2012). e SMGA model of Poiata et al. (2012). f 1-D

velocity models for the AQU and GSA stations used for low-frequency waveform modeling

Ground Motion Pulses for Dip-Slip Faulting

these near-fault stations showed clear short-duration

pulses with peak ground velocity (PGV) exceeding

30 cm/s (e.g., Fig. 3e).

The high socio-economic impact of the 2009

L’Aquila earthquake and the large number of high-

quality observational data produced by this event

attracted great interest from the geophysical commu-

nity. The process of the mainshock was studied by a

number of researchers providing a low-frequency

(\1.0 Hz) source model through the inversion of

geodetic and/or strong motion datasets (e.g., Atzori

et al. 2009; Walters et al. 2009; Cirella et al. 2009;

Poiata et al. 2012). These studies provided robust

general features of the 2009 L’Aquila mainshock

source model: a rupture on a southwest-dipping fault

of about 24 km by 16 km in size characterized by a

single main asperity located 8–10 km southeast of the

hypocentre (Fig. 4a). According to the broadband

source model of Poiata et al. (2012) that reproduced

the ground motions in the frequency range of

0.2–10 Hz, the earthquake ruptured a strong motion

generation area (SMGA), coincident with the area of

large slip (asperity) estimated from the low-frequency

source inversion and located beneath the highly

damaged city of Onna (Fig. 4a).

We made use of the available dataset of strong

motion recordings as well as the results of the

broadband source model (Fig. 4a) of Poiata et al.

(2012) to study the conditions that contributed to the

generation of the near-fault ground motion pulses

recorded at the stations located on the hanging wall of

the ruptured fault. The analysis consisted of perform-

ing forward simulations for different rupture

scenarios. The main objectives of our analysis were

as follows: (1) to compare the contributions of the

along-strike and up-dip rupture propagation to pulse

generation; (2) to determine the rupture configuration

(direction of rupture propagation and fault-station

geometry) that contributes positively to the pulse

generation; and (3) to identify the worst-case earth-

quake rupture scenario for the city of L’Aquila.

The assumed rupture scenarios were based on the

SMGA model of Poiata et al. (2012) illustrated in

Fig. 4 and estimated using the empirical Green’s

function (EGF) method (Irikura 1986). We used this

model to perform the forward simulations of the

waveforms for the near-fault strong motion stations in

both the broadband (0.2–10 Hz) and the low-fre-

quency (0.05–0.5 Hz) ranges. This option was

selected to account for the difference in the low-

and high-frequency radiation patterns discussed in the

previous section. The calculations were done for both

stations located on the hanging-wall side of the

causative fault above the ruptured fault plane (AQ*

stations, Fig. 4a) and the hanging-wall stations

located further away, as well as for the footwall

station GSA. This selection of stations should provide

more details on the spatial variability of the near-fault

ground motion pulses (fault-station geometry). Here,

we present only the results of an analysis on the

example AQU hanging-wall station and on the GSA

footwall station.

The broadband synthetic waveforms (0.2–10 Hz)

were calculated using the EGF method of Irikura

(1986). The method allows estimating the ground

motions of a target earthquake in broad frequency

range by using records of nearby smaller events,

which are treated as empirical Green’s functions

incorporating the properties of propagation path and

local site effects. This allows to bypass the high-

frequency ([1 Hz) limitation of the deterministic

approaches given by the uncertainties of the velocity

structure and complexity of source description. Based

on similarity relationship between the target and

small earthquakes and the omega-squared model of

the source spectra, the EGF method requires two

parameters N and C representing, respectively, the

ratios of the fault dimensions and the stress drop

between the events. Once these parameters are

estimated, the fault plane of the target event is

divided into N 9 N subfaults of sizes equivalent to

the rupture area of the element event, and the

synthetic waveforms for the target event are calcu-

lated as superposition of the recordings of the

element event scaled by the factor C and accounting

for the difference in the source-time functions

between the two events. The study of Poiata et al.

(2012) used Mw 4.6 aftershock that occurred on 7

April 2009 (21:34:29 UT, Fig. 4a) as the empirical

Green’s function (EGF) event for estimating the

parameters N and C by the spectral ratio fitting

method of Miyake et al. (1999), and providing the

model of SMGA of the 2009 L’Aquila earthquake

using the EGF method of Irikura (1986) discussed

N. Poiata et al. Pure Appl. Geophys.

previously. They provided a broadband (0.2–10 Hz)

source model for the 2009 L’Aquila earthquake

represented by a SMGA of 41.6 km2 (8.0 km in

length by 5.2 km in width) with a rise time of 0.8 s

and composed of 4 9 4 subfaults corresponding to

the SMGA of small (EGF) event (Fig. 4a, e). The

rupture starting point was located at the northwestern

part of SMGA implying a predominant southeast and

up-dip directions of rupture propagation (Fig. 4a, e).

The corresponding source parameters for small and

target events are summarized in Table 1. We use

these estimates and the same EGF event to perform

the forward simulations for the rupture scenarios

shown in Fig. 4b–d and compare them with the

SMGA model (Fig. 4e) of Poiata et al. (2012) as well

the observed records. For each of the rupture

scenarios, we assumed that the elementary subfault

corresponds to the fault size of the small event, and

the final source is composed by the superposition of

multiple elementary subfaults distributed according

to the designed geometry. The broadband synthetics

are then estimated by applying the EGF method

(Irikura 1986) for each of the scenarios. The assumed

source rupture scenarios were designed to allow

testing the contribution of the pure along-strike

(Fig. 4b) vs pure up-dip (Fig. 4c) rupture propagation

for the SMGA of the EGF model, the effect of the

different assumptions for the locations of the rupture

starting point (Fig. 4d) and the location of the rupture

area relative to the city of L’Aquila (Fig. 5a). For the

latter parameter, the AQU station was selected as

representative since no records were available from

the historical center of the city of L’Aquila (the most

damaged area of the city). The details of the assumed

rupture scenarios are summarized in Fig. 4.

The low-frequency (0.05–0.5 Hz) synthetic wave-

forms were determined using the extended reflectivity

method of Kohketsu (1985). The method allows a

deterministic calculation of the near-field seismo-

grams due to a finite fault (propagating rupture) in a

given 1-D velocity model. The finite fault is assumed

to be represented by a number of subfaults, each

corresponding to a point source with a given seismic

moment, mechanism, and rise time (s). The rupture

propagation is represented by a spherical rupture

front expanding at a speed Vr. For the calculation of

low-frequency synthetics, we used the same scenario

source models (based on SMGA of Poiata et al. 2012)

as for the broadband case (Fig. 4). We assumed that

each of the rupture fault’s subfaults represents the

EGF event (the Mw 4.6 aftershock on 7 April 2009,

21:34:29) with the size equivalent to that estimated

by Poiata et al. (2012) using the EGF method

(Table 1). The rise time for the ramp-type source-

time function and the seismic moment for each

subfault were estimated through low-frequency for-

ward modeling of the Mw 4.6 aftershock for the near-

fault hanging-wall stations shown in Fig. 4a assum-

ing that the source size of this event corresponds to

the subfault size. The results provided a seismic

moment M0 = 2.3 9 1015 Nm and a rise time

s = 1.0 s for the ramp-type source-time function.

The rupture propagation velocity was set to 2.9 km/s

(*0.85 VS), which is the same as that assumed for

the broadband modeling and derived from the EGF

analysis of Poiata et al. (2012). This value of lies in

the upper range of the commonly inferred earthquake

rupture velocities of 0.65–0.85 VS (e.g., Madariaga

1976). However, it is in good agreement with the near

shear wave speed of the rupture front propagation

Table 1

Source parameters of the EGF (small) event and the SMGA model used for the assumed rupture scenarios setup

C N EGF event

7 April 2009 21:34:29

SMGA (Poiata et al. 2012)

6 April 2009 01:32:40

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

l (km) w (km) s (s) L (km) W (km) S (km)2 T (s)

3.5 4 2.0 1.3 0.2 8.0 5.2 41.6 0.8 2.9 3.5 Broadband model

NA NA 1.0 NA

M0 (Nm) Low-frequency model

2.3 9 1015

Ground Motion Pulses for Dip-Slip Faulting

13˚24'E 13˚36'E

42˚1

2'N

42˚1

8'N

42˚2

4'N

5 km

L’AquilaPaganica

Castelnuovo

0.0

0.2

0.4

0.6

0.8

1.0Slip(m)

AQU

Acc. (cm/s/s)

−1000

0

1000

AQU station, f = 0.2 - 10 Hz

0 5 10 15Time (sec)

Vel. (cm/s)

−80

0

80

0 5 10 15Time (sec)

Displ. (cm)

−20

0

20

0 5 10 15Time (sec)

Acc. (cm/s/s)

−1000

0

1000

0 5 10 15Time (sec)

Vel. (cm/s)

−80

0

80

0 5 10 15Time (sec)

Displ. (cm)

−20

0

20

0 5 10 15Time (sec)

EW

−0.35

0.00

0.35

0 5 10 15 20 25 30Time (sec)

NS

−0.35

0.00

0.35

0 5 10 15 20 25 30Time (sec)

UD

−0.35

0.00

0.35

0 5 10 15 20 25 30Time (sec)

AQU station, f = 0.05 - 0.5 Hz

syn. Case 1

syn. Case3

syn. Case 1

syn. Case 3

Case1

Case2

(a)

(c)

Case 2.

Case 3.

syn. Case 2

syn. Case 2

GSA

(d)

−0.07

0.00

0.07

0 5 10 15 20 25 30Time (sec)

−0.07

0.00

0.07

0 5 10 15 20 25 30Time (sec)

−0.07

0.00

0.07

0 5 10 15 20 25 30Time (sec)

EW NS UDGSA station, f = 0.05 - 0.5 Hz

Acc. (cm/s/s)

GSA station, f = 0.2 - 10 Hz

0 5 10 15Time (sec)

Vel. (cm/s)

0 5 10 15Time (sec)

Displ. (cm)

0 5 10 15Time (sec)

Acc. (cm/s/s)

0 5 10 15Time (sec)

Vel. (cm/s)

0 5 10 15Time (sec)

Displ. (cm)

0 5 10 15Time (sec)

−150

0

150

−10

0

10

−2.5

0.0

2.5

−150

0

150

GS

A

−10

0

10

−2.5

0.0

2.5

syn. Case 2

syn. Case 1

syn. Case 2

syn. Case 1

EW NS

EW NS

Case 1.

Case 2.

Case 3.

(b)

AQU

AQU / GSA

Onna

N. Poiata et al. Pure Appl. Geophys.

required for generation of the rupture directivity

effect (e.g., Somerville et al. 1997). This is appro-

priate for the purpose of the study. The station-

specific 1-D velocity models used in calculations of

low-frequency synthetics were extracted from Poiata

et al. (2012). All the finite fault parameters are

summarized in Table 1 and the 1-D velocity models

are presented in Fig. 4.

3. Results

3.1. Effect of Ground Motion Pulses on the Response

Spectra

We investigated the response spectra of the fault-

normal component and the extracted long-period

ground motion pulse recorded during the 1992

Landers (Mw 7.3) earthquake at Lucerne station

located in the direction of the rupture propagation. By

comparing the velocity response spectra of the

original record and the extracted pulse (Fig. 3), we

observed that the peaks at longer periods ([4 s)

originate from the pulse, whereas the response

spectrum of the residual waveform is roughly com-

parable to that of the station situated in the backward

direction of the rupture propagation (Fig. 3c). Similar

results were provided by the analysis of the strike

normal (SN) component of the velocity record from

the AQK hanging-wall station (Fig. 3d) recorded

during the 2009 L’Aquila, Italy (Mw 6.3) normal-

faulting event (Fig. 3e). In this case, we observed

significantly larger values of the *1.8 s-period

response spectra for the 2009 L’Aquila earthquake.

This interval also corresponds to the period range of

the pulse recorded at the AQK station (Fig. 3b).

These examples illustrate that near-fault ground

motion pulses can generate significantly larger

demands on structures than ordinary records. This

also underlines that the design criteria established

without considering such long-period near-fault

pulses will inevitably result in underestimation of

the seismic demands.

3.2. Conditions for the Generation of Near-Fault

Ground Motion Pulses: The 2009 L’Aquila

Earthquake

The results of the forward simulations for differ-

ent rupture scenarios indicated the following: (1) both

up-dip and along-strike rupture propagations con-

tribute to the generation of near-fault ground motion

pulses, although the up-dip rupture propagation

contribution is more significant, and (2) in a more

general case (Fig. 4d), for all stations (hanging wall

or footwall), the generation of pulses is conditioned

by rupture propagation toward the site. More partic-

ularly, the results of forward simulations for the pure

along-strike and pure up-dip rupture propagation

scenarios (Fig. 4b, c) show that for the near-fault

stations (both hanging wall and footwall) not located

above the ruptured fault plane (e.g., GSA station

Fig. 5d), both along-strike and up-dip rupture prop-

agations contribute to the generation of pulses.

Furthermore, scenarios testing the different rupture

initiation points for the SMGA (Fig. 4d) indicate that

the main condition for pulse generation at these

stations is that the direction of rupture propagation

should be toward the site. The analysis also points out

that in case of a more complex rupture configuration,

it is not possible to identify a predominant direction

of rupture propagation that will be more efficient in

generating the ground motion pulses at near-fault

stations not located above the ruptured fault plane.

Figure 5d illustrates this conclusion for the example

of the GSA footwall station.

Next, we discuss in more detail the results of the

forward simulations for the rupture scenarios, assum-

ing different rupture starting points (Fig. 5b, c) and

different locations of the SMGA for near-fault

stations positioned on the hanging wall above the

bFigure 5

a Map showing the relative location of the AQU and GSA stations

(blue triangles) and the rupture scenario corresponding to the

predominant along-strike rupture propagation (Cases 1 and 2;

rupture starting points are shown by the gray and green stars,

respectively) and the focusing effect (Case 3; rupture starting point

is shown by the brown star). b Vertical view of the rupture

scenarios. The stars correspond to the location of the rupture

starting points. The blue triangle shows the station location. c,

d Synthetic broadband and low-frequency waveforms for the

rupture scenarios of Case 1 (gray traces), Case 2 (green traces),

and Case 3 (brown traces) of the AQU and GSA stations,

respectively

Ground Motion Pulses for Dip-Slip Faulting

ruptured fault plane (AQ* stations from the Aterno

Valley).

3.2.1 Near-Fault Stations Located on the Hanging

Wall above the Ruptured Fault

Figure 5 shows examples of rupture scenarios assum-

ing the rupture of the entire SMGA and different

rupture starting points (Fig. 4d). These scenarios

result in different directions of rupture propagation

relative to the AQ* stations. We select two extreme

cases of these scenarios, noting them Case 1 and Case

2 (Fig. 5a, b). In Case 1, rupture propagates toward

the AQ* stations, and the fraction of the ruptured area

of the fault between the hypocenter and the stations is

maximum; while in Case 2, rupture propagates away

from the stations, and the fraction of the ruptured area

of the fault between the hypocenter and the stations is

minimum. The results of the forward simulations for

the broadband and low-frequency ranges (Fig. 5c)

show that near-fault ground motion pulses are

efficiently generated for the scenarios implying the

propagation of rupture toward the stations. We also

confirm that the amplitude of the pulses, for both the

broadband and the low-frequency components at

station AQU, is maximum when the rupture starting

point is located in the southeast end of the fault (Case

1, Fig. 5c). This scenario includes the along-strike

(predominant) and the up-dip propagations of rupture

in the direction of the station. We will refer to this

rupture scenario as ‘‘rupture propagation toward the

site’’, in order to distinguish it from the rupture

directivity effect.

The rupture scenarios assuming different loca-

tions of the SMGA relative to the AQ* stations are

examined to determine the worst-case scenario of

pulse generation for the city of L’Aquila and compare

it to the SMGA model (Fig. 4e) based on the EGF

simulation of Poiata et al. (2012), as well as observed

ground motions. Here, we focus on the synthetic

waveforms at the AQU station. We identify two

possible locations of the rupture area and the rupture

starting point relative to the AQU station that result in

pulse generation for both the broadband and low-

frequency synthetic waveforms. These rupture sce-

narios are summarized in Fig. 6, showing their

location relative to the AQU station. The first

scenario (Case 1) corresponds to the rupture propa-

gation toward the station (Fig. 5a–c, Case 1), whereas

the second scenario (Figs. 5a–c, 6, Case 3) represents

the case when the station is located above the

ruptured fault plane and the distance to the rupture

starting point is minimum (Fig. 5b, right). This case

corresponds to the ‘‘focusing effect’’ discussed by

Kagawa (2009). The effect is caused by the simul-

taneous arrival of the rupture front from different

points of the fault, which are at the same distance (on

the same isochrones) from the rupture starting point

(Fig. 7b). The synthetic broadband and low-fre-

quency waveforms for Cases 1 and 3 are presented

in Figs. 5c and 6a. Rupture propagation toward the

site (Case 1) and the focusing effect (Case 3) generate

ground motion pulses significantly exceeding the

observations as well as the simulation results for the

SMGA based on the EGF model of Poiata et al.

(2012) (Fig. 6a). It can be noticed from Fig. 6a that

the SMGA model of Poiata et al. (2012) reproduces

well the ground motion pulses recorded at the AQU

station during the 2009 L’Aquila earthquake. The

comparison of the velocity response spectra for the

observed and synthetic waveforms (Fig. 6c, d) indi-

cates that in the period 2.0–3.5 s, in both cases, the

spectral ordinates of these observed ground motions

are significantly below the level provided by the two

scenarios of Cases 1 and 3. Moreover, in Case 3, the

spectral values of synthetics are significantly higher

than the observations for the period range of

1.0–1.3 s (Fig. 6c), corresponding to the resonance

period of typical residential buildings in the area.

These results point out the important role that the

direction of rupture propagation and the location of

the SMGA relative to the hanging-wall sites play in

generation of the near-fault ground motion pulses in

dip-slip faulting events.

4. Discussion and Conclusions

In this study, we demonstrated that long-period

near-fault ground motions pulses observed at near-

source regions can strongly affect the duration and

spectral characteristics of the observed ground

motions. Thus, the design criteria established without

considering such ground motion pulses will

N. Poiata et al. Pure Appl. Geophys.

EW componentAcc. (cm/s/s)

−1000

0

1000

0 5 10 15Time (sec)

Vel. (cm/s)

−80

0

80

0 5 10 15Time (sec)

Displ. (cm)

−20

0

20

0 5 10 15Time (sec)

NS componentAcc. (cm/s/s)

−1000

0

1000

0 5 10 15Time (sec)

Vel. (cm/s)

−80

0

80

0 5 10 15Time (sec)

Displ. (cm)

−20

0

20

0 5 10 15Time (sec)

obs.

syn. EGF

syn. Case 1

syn. Case 3

(a)

(c)

(d)

238.0 20.5 3.8

280.2 17.5 3.7

327.1 31.3 12.1

623.0 46.6 14.7

289.4 28.9 7.7

269.0 23.2 6.0

340.5 37.8 12.0

1143.0 78.8 13.9

EW obs.NS obs.EW syn.3NS syn.3

Case 3

0

100

200

300

400

SV

(cm

/s),

(da

mpi

ng=

5%)

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

Period (sec)

EW syn.EGFNS syn.EGF

Case1

0

100

200

300

400

SV

(cm

/s),

(da

mpi

ng=

5%)

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

Period (sec)

EW obs.NS obs.EW syn.1NS syn.1EW syn.EGFNS syn.EGF

13˚24'E 13˚36'E

42˚1

2'N

42˚1

8'N

42˚2

4'N

5 km

Onna

Paganica

Castelnuovo

0.0

0.2

0.4

0.6

0.8

1.0Slip(m)

AQUL’Aquila

Case1

Case 3

EGF

(b)

Figure 6a Comparison of the observed (black traces) acceleration, velocity, and displacement waveforms for the horizontal components of the

broadband record of the 2009 L’Aquila mainshock from the AQU station, and the synthetics calculated for the SMGA by the EGF simulation

of Poiata et al. (2012; red traces), the rupture scenario of the rupture propagation towards the AQU station (Case 1; gray traces), and the

rupture scenario corresponding to the case of the focusing effect (Case 3; brown traces). The numbers above the waveforms correspond to the

maximum amplitudes. b Map of the near-fault area of the 2009 L’Aquila earthquake showing the relative location of the AQU station (blue

triangle), the city of L’Aquila, and the rupture areas for the assumed scenario and EGF model (gray rectangles). The stars correspond to the

location of the rupture starting points. c 5% damped velocity response spectra comparing the observed horizontal components of the AQU

record (in black), and d the synthesis calculated for the two cases of the rupture scenarios and the EGF model (red traces)

Ground Motion Pulses for Dip-Slip Faulting

inevitably result in underestimation of seismic

demands in building structures.

The forward rupture directivity effect is generally

considered the main cause of the near-fault ground

motion pulses observed in cases of strike-slip and

dip-slip faulting. The conditions for the rupture

directivity effects summarized by Somerville et al.

(1997) provided a good explanation of the fault-

normal components observed during the past strike-

slip faulting events (e.g., 1992 Landers). In the case

of dip-slip faulting, however, both the near-fault

ground motion pulses and the hanging-wall effect

(Abrahamson and Somerville 1996) are observed at

some stations. The inspection of the 3-D S-wave

radiation patterns for dip-slip faulting, as well as the

near-fault strong ground motion records from the

stations on the hanging-wall sites above the causative

fault of the 2007 Chuetsu-oki (reverse faulting), the

2009 L’Aquila (normal faulting) and the 2015 Gor-

kha (low-angle dip-slip faulting) earthquakes point

toward a more complex mechanism of pulse gener-

ation for such type of events.

We also inspected the main aspects of the rupture

configuration that contribute to the generation of the

near-fault ground motion pulses for a dip-slip faulting

event by performing the forward simulations of the

different rupture scenarios for the 2009 L’Aquila,

Italy, (Mw 6.3) earthquake. We examined rupture

scenarios based on the broadband source model of the

2009 L’Aquila earthquake determined by Poiata et al.

(2012) and performed forward simulations of the

waveforms for the near-fault strong motion stations in

both broadband and low-frequency ranges. The

results of our analysis indicate that the generation

mechanism of the near-fault ground motion pulses in

the case of dip-slip faulting depend on both the rup-

ture configuration and the location of the site relative

to the fault plane. We confirmed that the rupture

directivity effect is predominant in stations located on

the footwall of the causative fault. In all hanging-wall

stations, pulse generation occurs as the result of the

radial rupture propagation toward the site (Fig. 7a).

However, in the stations located on the hanging wall

above the ruptured fault plane, where the hanging-

wall effect is observed (Abrahamson and Somerville

1996), the focusing effect (Kagawa 2009) could be an

alternative mechanism of pulse generation. Thus, our

results indicated that that the actual 2009 L’Aquila

earthquake did not constitute a worst-case scenario

for the city of L’Aquila regardless of its proximity to

the epicenter. We confirmed that, in the case of dip-

slip faulting earthquakes, along-strike propagation

toward the site and the focusing effect (Fig. 7) are the

main candidates for the worst-case rupture scenarios

of pulse generation for the hanging-wall sites located

above the area of large slip (strong motion generation

area; Fig. 6).

Acknowledgements

The authors would like to thank Luis Dalguer and

two anonymous reviewers for their help in improving

the manuscript. We would also like to express our

gratitude to the Italian Department of Civil Protection

and the Istituto Nazionale di Geofisica e

AQU

AQU

δ

hanging wallfoot wall

(a) (b)

Figure 7Schematics of a the radial rupture propagation toward the site and b the focusing effect

N. Poiata et al. Pure Appl. Geophys.

Vulcanologia of Italy for providing free access to

seismological data and the strong ground motion

records of the 2009 L’Aquila earthquake and its

aftershocks. The original near-field acceleration

waveforms used in the study were downloaded from

the website of the database of the Italian strong

motion records (http://itaca.mi.ingv.it) maintained by

the Working Group ITACA. Some figures were made

using the Generic Mapping Tools software (Wessel

and Smith 1995).

REFERENCES

Abrahamson, N. A., & Somerville, P. G. (1996). Effects of the

hanging wall and footwall on ground motions recorded during

the Northridge earthquake. Bulletin of the Seismological Society

of America, 86, S93–S99.

Aki, K. (1968). Seismic displacement near a fault. Journal of

Geophysical Research, 73, 5359–5376.

Aki, K., & Richards, P. G. (1980). Quantitative seismology (p.

932). San Francisco: W. H. Freeman.

Atzori, S., Hunstad, I., Chini, M., Salvi, S., Tolomei, C., Bignami,

C., et al. (2009). Finite fault inversion of DInSAR coseismic

displacement of the 2009 L’Aquila earthquake (central Italy).

Geophysical Research Letters, 36, L15305. doi:10.1029/

2009GL039293.

Baker, J. W. (2007). Quantitative classification of near-fault ground

motions using wavelet analysis. Bulletin of the Seismological

Society of America, 97, 1486–1501.

Baker, J.W. & Shahi, S. (2007). Quantitative classification of near-

fault ground motions using Wavelet analysis (supporting soft-

ware), http://www.stanford.edu/*bakerjw/pulse-classification.

html. Accessed Apr 2017.

Boncio, P., Pizzi, A., Brozzetti, F., Pomposo, G., Lavecchia, G., Di

Naccio, D., et al. (2010). Coseismic ground deformation of the 6

April 2009 L’Aquila earthquake (central Italy, Mw6.3). Geo-

physical Research Letters, 37, L06308. doi:10.1029/

2010GL042807.

Boore, D. M., & Joyner, W. B. (1978). The influence of rupture

incoherence on seismic directivity. Bulletin of the Seismological

Society of America, 68, 283–330.

Boore, D. M., & Joyner, W. B. (1989). The effect of directivity on

the stress parameter determined from ground motion observa-

tions. Bulletin of the Seismological Society of America, 79,

1984–1988.

Cirella, A., Piatanesi, A., Cocco, M., Tinti, E., Scognamiglio, L.,

Michelini, A., et al. (2009). Rupture history of the 2009 L’Aquila

(Italy) earthquake from non-linear joint inversion of strong

motion and GPS data. Geophysical Research Letters, 36,

L19304. doi:10.1029/2009GL039795.

Heaton, T. H. (1990). Evidence for and implications of self-healing

pulses of slip in earthquake rupture. Physics of the Earth and

Planetary Interiors, 64, 1–20.

Heaton, T. H., Hall, D. F., Wald, D. J., & Halling, M. W. (1995).

Response of high-rise and base-isolated buildings to a hypo-

thetical M 7.0 blind thrust earthquake. Science, 267, 206–211.

Irikura, K. (1986). Prediction of strong acceleration motions using

empirical Green’s function. In Proceedings of 7th Japan Earth-

quake Engineering symposium (pp. 151–156).

Istituto Nazionale di Geofisica e Vulcanologia (INGV) (2009), The

L’Aquila seismic sequence—April 2009, Italian Accelerometric

Archive (ITACA). Available at: http://itaca.mi.ingv.it/ItacaNet/.

Accessed Feb 2015.

Joyner, W. B. (1991). Directivity for nonuniform ruptures. Bulletin

of the Seismological Society of America, 81, 1391–1495.

Kagawa, T. (2009). Conditions of fault rupture and site location

that generate damaging pulse waves. In Proceedings of the 6th

International Conference on Urban Earthquake Engineering (pp.

53–58).

Kasahara, K. (1981). Earthquake mechanics, (p. 248), Cambridge:

Cambridge University Press.

Kohketsu, K. (1985). The extended reflectivity method for syn-

thetic near-field seismograms. Journal of Physics of the Earth,

33, 121–131.

Koketsu, K., & Miyake, H. (2008). A seismological overview of

long-period ground motion. Journal of Seismology, 12,

133–143.

Koketsu, K., Miyake, H., Guo, Y., Kobayashi, H., Masuda, T.,

Davuluri, S., et al. (2016). Widespread ground motion distribu-

tion caused by rupture directivity during the 2015 Gorkha, Nepal

earthquake. Scientific Reports, 6, 28536. doi:10.1038/srep28536.

Lay, T., & Wallace, T. C. (1995). Modern global seismology (p.

521). San Diego: Academic Press.

Liu, H., & Helmberger, D. V. (1985). The 23:19 aftershock of the

15 October 1979 Imperial valley earthquake: more evidence for

an asperity. Bulletin of the Seismological Society of America, 75,

689–708.

Madariaga, R. (1976). Dynamics of an expanding circular fault.

Bulletin of the Seismological Society of America, 66,

163–182.

Miyake, H., Iwata, T., & Irikura, K. (1999). Strong ground motion

simulation and source modeling of the Kagoshima-ken Hoku-

seibu earthquakes of March 26 (MJMA 6.5) and May 13 (MJMA

6.3), 1997, using empirical Green’s function method. Zisin, 51,

431–442. (in Japanese with English abstract).

Miyake, H., Koketsu, K., Hikima, K., Shinohara, M., & Kanazawa,

T. (2010). Source fault of the 2007 Chuetsu-oki, Japan, earth-

quake. Bulletin of the Seismological Society of America, 100,

384–391.

Miyatake, T. (1998). Generation mechanism of strong ground

motion pulse near the earthquake fault. Zisin, 2(51), 161–170. (in

Japanese with English abstract).

Poiata, N., Koketsu, K., Vuan, A., & Miyake, H. (2012). Low-

frequency and broadband source models for the 2009 L’Aquila,

Italy, earthquake. Geophysical Journal International, 191,

224–242.

Somerville, P. G. (2003). Magnitude scaling of the near fault

rupture directivity pulse. Physics of the Earth and Planetary

Interiors, 137, 201–212.

Somerville, P. G., Smith, N. F., Graves, R. W., & Abrahamson, N.

A. (1997). Modification of empirical strong ground motion

attenuation relations to include the amplitude and duration

effects of rupture directivity. Seismological Research Letters, 68,

199–222.

Strasser, F. O., & Bommer, J. J. (2009). Review: strong ground

motions—have we seen the worst? Bulletin of the Seismological

Society of America, 99, 2613–2637.

Ground Motion Pulses for Dip-Slip Faulting

Takemura, S., Furumura, T., & Saito, T. (2009). Distortion of the

apparent S-wave radiation pattern in the high-frequency wave-

field: Tottori-Ken Seibu, Japan, earthquake of 2000. Geophysical

Journal International, 178(2), 950–961.

Takenaka, H., Mamada, Y., & Futamure, H. (2003). Near-source

effect on radiation pattern of high-frequency S waves: strong

SH–SV mixing observed from aftershocks of the 1997 North-

western Kagoshima, Japan, earthquakes. Physics of the Earth and

Planetary Interiors, 137, 31–43.

Walters, R. J., Elliott, J. R., D’Agostino, N., England, P. C.,

Hunstad, I., Jackson, J. A., et al. (2009). The 2009 L’Aquila

earthquake (central Italy): a source mechanism and implications

for seismic hazard. Geophysical Research Letters, 36, L17312.

doi:10.1029/2009GL039337.

Wessel, P., & Smith, W. H. F. (1995). New version of the generic

mapping tools released, EOS. Transactions American Geophys-

ical Union, 76, 329.

(Received July 30, 2016, revised March 23, 2017, accepted March 24, 2017)

N. Poiata et al. Pure Appl. Geophys.