ArticlePDF Available

Double-ramp on the Main Himalayan Thrust revealed by broadband waveform modeling of the 2015 Gorkha earthquake sequence

Authors:

Abstract and Figures

The 2015Mw7.8 Gorkha earthquake sequence that unzipped the lower edge of the Main Himalayan Thrust (MHT) in central Nepal provides an exceptional opportunity to understand the fault geometry in this region. However, the limited number of focal mechanisms and the poor horizontal locations and depths of earthquakes in the global catalog impede us from clearly imaging the ruptured MHT. In this study, we generalized the Amplitude Amplification Factor (AAF) method to teleseismic distance that allows us to model the teleseismic P-waves up to 1.5 Hz. We used well-constrained medium-sized earthquakes to establish AAF corrections for teleseismic stations that were later used to invert the high-frequency waveforms of other nearby events. This new approach enables us to invert the focal mechanisms of some early aftershocks, which is challenging by using other long-period methods. With this method, we obtained 12 focal mechanisms more than that in the GCMT catalog. We also modeled the high-frequency teleseismic P-waves and the surface reflection phases (pP and sP) to precisely constrain the depths of the earthquakes. Our results indicate that the uncertainty of the depth estimation is as small as 1–2 km. Finally, we refined the horizontal locations of these aftershocks using carefully hand-picked arrivals. The refined aftershock mechanisms and locations delineate a clear double-ramp geometry of the MHT, with an almost flat décollement sandwiched in between. The flat (dip ∼7 degrees) portion of the MHT is consistent with the coseismic rupture of the mainshock, which has a well-constrained slip distribution. The fault morphology suggests that the ramps, both along the up-dip and down-dip directions, play a significant role in stopping the rupture of the 2015 Gorkha earthquake. Our method can be applied to general subduction zone earthquakes and fault geometry studies.
Content may be subject to copyright.
Earth and Planetary Science Letters 473 (2017) 83–93
Contents lists available at ScienceDirect
Earth and Planetary Science Letters
www.elsevier.com/locate/epsl
Double-ramp on the Main Himalayan Thrust revealed by broadband
waveform modeling of the 2015 Gorkha earthquake sequence
Xin Wang a, Shengji Wei a,b,, Wenbo Wu c
aEarth Observatory of Singapore, Nanyang Tec hno lo gi ca l University, Singapore
bAsian School of the Environment, Nanyang Tech nol og ic al University, Singapore
cDepartment of Geosciences, Princeton University, United States
a r t i c l e i n f o a b s t r a c t
Article history:
Received 15 March 2017
Received in revised form 16 May 2017
Accepted 22 May 2017
Editor: P. Shearer
Keywords:
2015 Gorkha earthquake
waveform modeling
focal mechanism
Main Himalaya Thrust
fault geometry
The 2015 Mw7.8 Gorkha earthquake sequence that unzipped the lower edge of the Main Himalayan
Thrust (MHT) in central Nepal provides an exceptional opportunity to understand the fault geometry
in this region. However, the limited number of focal mechanisms and the poor horizontal locations
and depths of earthquakes in the global catalog impede us from clearly imaging the ruptured MHT.
In this study, we generalized the Amplitude Amplification Factor (AAF) method to teleseismic distance
that allows us to model the teleseismic P-waves up to 1. 5 Hz. We used well-constrained medium-
sized earthquakes to establish AAF corrections for teleseismic stations that were later used to invert
the high-frequency waveforms of other nearby events. This new approach enables us to invert the focal
mechanisms of some early aftershocks, which is challenging by using other long-period methods. With
this method, we obtained 12 focal mechanisms more than that in the GCMT catalog. We also modeled the
high-frequency teleseismic P-waves and the surface reflection phases (pP and sP) to precisely constrain
the depths of the earthquakes. Our results indicate that the uncertainty of the depth estimation is as
small as 1–2 km. Finally, we refined the horizontal locations of these aftershocks using carefully hand-
picked arrivals. The refined aftershock mechanisms and locations delineate a clear double-ramp geometry
of the MHT, with an almost flat décollement sandwiched in between. The flat (dip 7 degrees) portion
of the MHT is consistent with the coseismic rupture of the mainshock, which has a well-constrained
slip distribution. The fault morphology suggests that the ramps, both along the up-dip and down-dip
directions, play a significant role in stopping the rupture of the 2015 Gorkha earthquake. Our method
can be applied to general subduction zone earthquakes and fault geometry studies.
©2017 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND
license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
1. Introduction
Precise fault geometry is the key to understanding fault seg-
mentation, which plays a crucial role in the initiation, propagation
and termination of earthquakes, as well as in orogenic processes
(Cattin and Avouac, 2000; Wesnousky, 2006; Avouac, 2007; Elliott
et al., 2016; Hubbard et al., 2016; Qiu et al., 2016). The Main Hi-
malayan Thrust (MHT) that has hosted a series of damaging earth-
quakes, and is the location of the highest mountain range in the
world, has become the testing ground for fault geometry research
(Fig. 1). Previous studies have investigated the geometry of the
MHT in central Nepal through receiver functions (Schulte-Pelkum
et al., 2005; Nabelek et al., 2009; Duputel et al., 2016), struc-
*Corresponding author at: Earth Observatory of Singapore, Nanyang Technologi-
cal University, 50 Nanyang Avenue , N2-01a-14, 639798, Singapore.
E-mail address: shjwei@ntu.edu.sg (S. Wei).
ture geology (Pearson and DeCelles, 2005; Avouac, 2007; Hubbard
et al., 2016), electronic and magnetic surveys (Lemonnier et al.,
1999), seismicity (Pandey et al., 1995) and geodetic data (Elliott
et al., 2016). Most of these studies share a common feature: the
MHT is almost flat beneath the Lesser Himalaya, as a décolle-
ment that connects the Main Boundary Thrust (MBT) and the Main
Frontal Thrust (MFT) at the surface, extends southwards in the
Sub-Himalaya and steepens northwards in the Higher Himalaya,
usually through a ramp (Avouac, 2007). However, there are large
variations between the dimensions, depths and dips of these fault
geometries, e.g. the size of the ramp and whether or not existence
of other ramps beneath the Kathmandu Valley (Elliott et al., 2016;
Hubbard et al., 2016).
Earthquakes are direct evidence of active faults: precise earth-
quake location and mechanism can provide vital information to
infer the fault geometry of the MHT, where a series of large
earthquakes took place (Sapkota et al., 2013; Hayes et al., 2015;
Bollinger et al., 2016). However, the number of large earthquakes
http://dx.doi.org/10.1016/j.epsl.2017.05.032
0012-821X/©2017 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
84 X. Wang et al. / Earth and Planetary Science Letters 473 (2017) 83–93
Fig. 1. (a) Earthquakes from the USGS catalog 1970–2016 (https :/ /www.usgs .gov),
with events of M <5.0, 5.0 M <5.5and M 5.5before the 2015 Gorkha earth-
quake colored in gray, blue and red, respec tively. Earthquakes with M 5.0after
the 2015 Gorkha mainshock (including the mainshock) are colored in green. Main
Frontal Thrust (MFT), Main boundary Thrust (MBT) and Main Central Thrust (MCT)
are shown as black, pink and blue lines, respectively. (b) Earthquakes from the
GCMT catalog 1976–2016 (http :/ /www.globalcmt .org). Note the difference in siz-
able events before and after the 2015 Gorkha earthquake. (For interpretation of the
colors in this figure, the reader is referred to the we b version of this article.)
in Nepal that have modern broadband seismic records is very
limited: between the 1970 and 2015 Gorkha earthquake, only 8
events with magnitude larger than 5.5 occurred in the entire coun-
try (Fig. 1). The source parameters (location, mechanism, rupture
extents) for earthquakes that occurred before the era of modern
seismic instruments have been poorly constrained. The large uncer-
tainties of source parameters and lack of modern seismicity have
impeded us from inferring the structure of faulting from these his-
torical earthquakes.
The 2015 Gorkha earthquake sequence that unzipped the lower
edge of the MHT in central Nepal provides an exceptional op-
portunity to better understand the fault geometry in this region
(Fig. 2), as the number of sizable earthquakes (33 events with
M >5.0) in this sequence is comparable to the sum of events
that occurred from 1970 to 2015 (before the Gorkha earthquake)
(Fig. 1). However, only 9 of these 33 events—including the main-
shock and two aftershocks with normal focal mechanisms—have
Global Centroid Moment Tenso r (GCMT) solutions (Fig. 1). This is
because most of the aftershocks occurred immediately after the
mainshock and the Mw7.2aftershock, and therefore the long-
period seismic signals produced by the aftershocks were contam-
inated by the surface waves or the coda from the mainshock or
previous large event (Fig. 3). In addition, the depths are rela-
tively poorly constrained in the GCMT solutions—similar to other
long-period solutions for shallow events. Thus, it is difficult to
use the limited number of events with poor locations to delin-
eate the fault geometry of the MHT. Although the Gorkha seis-
micity has been relocated by various studies using either lo-
cal and/or teleseismic arrival time data (Adhikari et al., 2015;
Bai et al., 2016), the depths and horizontal locations of earthquakes
among these catalogs still vary significantly (Fig. 4 and Fig. S8). All
these factors obstruct our understanding of the geometry of the
MHT using seismicity.
To overcome these difficulties, we generalized the Amplitude
Amplification Factor (AAF) method to teleseismic distance that al-
lows us to determine the focal mechanisms of a portion of early af-
tershocks with high-frequency teleseismic P-waves, which resulted
in a dozen more solutions than those found in available catalogs.
We then modeled the high-frequency teleseismic depth phases to
precisely determine the depth of these events; we also further
relocated the horizontal position of these events using carefully
hand-picked P-wave arrival times. The refined earthquake catalog
illuminates a clear MHT that shows double ramps with the coseis-
mic slip of the mainshock sandwiched in between. In this paper,
we will describe the data and approaches used in greater detail,
followed by the results; we will then discuss the implications of
our findings.
2. Focal mechanism inversion for early aftershocks
Our approach in resolving the focal mechanism of more after-
shocks benefits from the usage of high-frequency (0.5 1.5 Hz)
teleseismic P-waves. At this frequency range, the signal-to-noise-
ratio (SNR) of P-waves is usually higher than that at longer periods,
particularly for some early aftershocks (Fig. 3), because the sig-
nals from the previous earthquakes are attenuated more and the
ambient noise level is usually lower. However, to use the wave-
form for inversion at this frequency range, we also cannot ignore
the site condition and structure complexity along the ray path.
To deal with this challenge, we took advantage of earthquakes
that have reliable long-period focal mechanism solutions to es-
tablish path calibration. We established this calibration by fixing
the focal mechanism to the long-period solution and predicting
the teleseismic P-waves at high-frequency ranges. We found that,
at 0.5 1.5 Hz, the shape of first 3 s of teleseismic P-waves can
still be well-fitted by the 1D synthetics with an amplitude ampli-
fication factor (AAF) applied to the synthetics to correct for the
imperfect green’s functions (Fig. S1, S5). A similar approach has
been successfully applied to the small earthquake focal mecha-
nism inversion, using regional waveform data in Southern Cali-
fornia (Tan and Helmberger, 2007). The key here is to find the
calibration events that have reliable long-period focal mechanism
solutions. To ensure the rupture complexity of calibration events
can be ignored at 0.5–1.5 Hz, ideally we need to select earth-
quakes with source durations less than about 0.6 s, so that they
can be considered as point sources. Longer source duration will
produce a roughly constant shift to all the AAFs. Our tests indi-
cate that the contribution to the standard deviation (see supple-
ment materials for more details) using a 1.5 s source time func-
tion is about 0.2 (Fig. S4), which is the threshold we used in this
study. It is also important to note that a stable and reliable long-
period focal mechanism solution is required for the calibration
events. Using these criteria, we found two aftershocks (2015/04/26
16:26 (UTC) Mw5.0 and 2015/05/16 13:34 (UTC) Mw5.2) in the
2015 Gorkha earthquake sequence that can be used as calibration
events.
Although there are reported moment tensor solutions (e.g.,
GCMT, W-phase) for these calibration events and some other large
aftershocks, we still wanted to have our own solutions, in par-
ticular for depth, which reveal large variations among different
catalogs (Fig. 4). Here we used extended teleseismic P and SH
waves, which contain the depth phases (e.g. pP, sP, sS) that are
most sensitive to the focal depth to invert the focal mechanism
and depth. Many practices have demonstrated that teleseismic P
and SH waves are less affected by the 3D velocity structure and
can result in higher resolution in fault plane solutions, since most
of the ray paths lie in the relatively simple mantle (Zhan et al.,
2012). Given that we are using a 1D velocity model in the in-
version (as in most moment tenor inversion methods), and that
X. Wang et al. / Earth and Planetary Science Letters 473 (2017) 83–93 85
Fig. 2. (a) Map of source region of the 2015 Gorkha earthquake, showing the inferred coseismic slip distribution for the mainshock (red-to-yellow color coded with black
contours) and its largest aftershock on May 12th (cyan contours) from joint inversion of GPS, SAR/InSAR and seismic data (Wei et al., in preparation). Gray circles indicate
the locations of aftershocks from double difference relocation (Bai et al., 2016). Moment tensor solutions for aftershocks obtained by our inversion are shown as beach balls,
colored by focal depth and scaled to magnitude. Main Frontal Thrust (MFT), Main boundary Thrust (MBT) and Main Central Thrust (MCT) are shown as black, pink and blue
lines, respectively. Surface rupture of the 1934 event is shown as heavy red lines. Lower inset map shows the location of the 2015 Gorkha earthquake. The black line AA’
gives the location of the profile shown in Fig. 4. (b) Aftershock focal mechanisms with earthquake ID and magnitude are shown in the upper panel. The lower panel displays
the focal mechanisms obtained from different methods. AAF: High-frequency waveform inversion with AAF correction. NoAAF: High-frequency waveform inversion without
AAF correction. GCMT: Focal mechanisms from the GCMT catalog. LP_CAP: Mechanisms from Long-Period waveform Cut-And-Paste (CAP) inversion.
there is no high-resolution 3D velocity model available in this re-
gion, we preferred to use only teleseismic P and SH waves in
focal mechanism inversion, so as to minimize the impact of the
unknown 3D structure. We conducted the waveform inversion at
higher frequencies (e.g., 0.01 0.05 Hz) than that in GCMT (ap-
proximately 0.006 0.025 Hz) (Ekstrom et al., 2012) and W-phase
(typically 0.001 0.005 Hz) (Kanamori and Rivera, 2008)inver-
sions; this procedure allows us to better resolve the depth of
earthquakes because the sensitivity increases as a function of fre-
quency. We adopted the Cut-And-Paste (CAP) method (Zhao and
Helmberger, 1994; Zhu and Helmberger, 1996)to carry out grid
search for the best double-couple solution and centroid depth (see
more details about the method in the supplementary material).
Our results and corresponding uncertainties are shown in a boot-
strapping manner (Figs. 5 and 6). We also tested the source du-
ration against the depth, and always got consistent results (Figs. 5
and 6). Overall, we obtained long-period solutions for six events,
including the two calibration events, with results summarized in
Fig. 2(b). It turns out that our fault plane solutions are consistent
with the GCMT catalog, while we found more differences for the
depths.
We then fixed the focal mechanism of two calibration events
to predict the high-frequency (0.5–1.5 Hz) teleseismic P-waves. We
only focused on the beginning portion (first 3s) of P-waves, since
these are direct downgoing waves that do not include the depth
phases, which are more easily affected by the topography and shal-
low velocity structure. By comparing the 1D synthetics against the
data, we got the Amplitude Amplification Factors (AAFs) for the
two calibration events (see more details in the supplementary ma-
terial). The comparison between the AAFs derived from these two
events show consistent values (Fig. S2), although their horizon-
tal locations differ by 50 km, indicating that the AAFs are mainly
caused by the upper mantle attenuation and receiver site structure.
This is an important feature that is required to apply the AAFs to
other aftershocks in this earthquake sequence. To verify the AAFs,
we conducted crosschecking tests by inverting the focal mecha-
nism of the two calibration earthquakes using the AAFs derived
from the other events (Fig. S6). Our tests showed that the AAFs
could greatly improve the accuracy of focal mechanism inversion.
We also used the AAFs to derive the focal mechanism of an Mw4.6
aftershock, which occurred on 2015/08/23 09:02 (UTC), which was
well recorded by four local broadband stations (Fig. S7). Again, the
86 X. Wang et al. / Earth and Planetary Science Letters 473 (2017) 83–93
Fig. 3. (a) Aftershocks of the 2015 Gorkha earthquake sequence for the time window 12 hrs after the April 25th mainshock and the May 12th Mw7.2aftershock. Events that
have GCMT solutions are colored in blue. Events with focal mechanisms obtained by our study are colored in red. (b) The waveform record for the first hour after the Mw7.2
event at the teleseismic station COLA (distance 78, azimuth 21) at different frequency ranges. Some of early aftershocks’ theoretical P-wave arrivals are marked as
green dots, along with the origin time and magnitude (USGS catalog). Note the difference in signal-to-noise ratio at long-period waveform (5–100 s) and higher frequencies
(e.g. 0.5–1.5 Hz). (For interpretation of the colors in this figure, the reader is referred to the web version of this article.)
Tabl e 1
Source parameters of earthquakes obtained in this study.
Origin time Location Focal mechanism
Longitude
(E)
Latitude
(N)
Depth*
(km)
Strike
()
Dip
()
Rake
()
Magnitude
(Mw)
Long-period CAP inversion
2015/04/26 07:09 86.0147 27.7677 13.1 304 19 110 6.66
2015/04/26 16:26 85.8150 27.8070 13.0 304 20 122 5.02
2015/05/12 07:05 86.0970 27.7520 11.6 329 8 141 7.14
2015/05/12 07:36 86.1480 27.5840 10.0 270 32 68 5.92
2015/05/16 11:34 86.1510 27.5880 9.5 322 11 131 5.22
High-frequency AAF inversion
2015/04/25 06:56 85.7670 27.8050 12.1 336 21 149 4.85
2015/04/25 07:07 85.9220 27.7450 14.7 55 74 60 4.54
2015/04/25 07:47 85.6300 27.8840 14.9 334 22 161 4.58
2015/04/25 08:20 85.1150 27.7260 7.9 79 56 95 4.35
2015/04/25 08:29 84.6800 27.9470 9.4 249 16 75 4.47
2015/04/25 08:55 85.5160 27.6190 7.3 304 85 58 4.88
2015/04/25 12:44 84.6650 28.1050 7.4 13 28 180 4.73
2015/05/12 07:34 86.2640 27.6940 14.8 222 34 50 4.72
2015/05/12 08:13 85.8520 27.7640 11.5 300 15 120 4.60
2015/05/13 21:38 86.0720 27.6380 10.4 236 16 45 4.44
2015/05/15 01:42 84.7380 27.8620 10.0 258 44 102 4.45
*The depth is relative to the sea level.
teleseismic inversion with AAF corrections reveals a consistent fo-
cal mechanism as we derived from the local broadband waveform
inversion (Fig. S7), while result without AAF correction is much
more different.
We then used the AAFs derived from the calibration events to
determine the focal mechanism of other early aftershocks, which
only have good SNR at high frequency (Fig. 3). By applying this
method, we obtained the focal mechanism of 12 events more than
that in the GCMT catalog; this is summarized in Fig. 2 and Ta-
ble 1. Note that our inversions with AAF corrections also reveal a
similar focal mechanism for those events having a GCMT solution,
including both the normal events and the shallow dip angle thrust
events. In the next, we will focus on using events with thrust focal
mechanism to delineate the geometry of the MHT.
3. Horizontal location and depth
To delineate the fault geometry with earthquakes, we need to
better resolve another key parameter, the location. As shown in
Fig. S8, different catalogs show large variations in earthquake hor-
izontal locations, which could be caused by the different inversion
methods and the data they used. While quantifying the discrep-
ancy between inversion methods is difficult, we are emphasizing
the data quality that is used in earthquake locating or relocating
procedures. We aligned the teleseismic P-waves (Fig. S9), based
on our carefully hand-picked arrivals for one of the aftershocks
(2015/05/12 07:36 (UTC) Mw6.0). Here the waveforms have been
filtered to 0.5–4 Hz, a frequency band that best shows the sig-
nal for this particular earthquake. It is clear that our hand picks
X. Wang et al. / Earth and Planetary Science Letters 473 (2017) 83–93 87
Fig. 4. Comparison of different catalogs in depth profiles. (a) The USGS catalog (http :/ /earthquake .usgs .gov). (b) The GCMT catalog (http :/ /www.globalcmt .org). (c) The catalog
from Adhikari et al. (2015). (d) The catalog from Bai et al. (2016). (e) Our refined catalog with focal mechanisms and error bars indicating the 95% confidence levels for
horizontal and vertical locations. For comparison, only those earthquakes studied here are highlighted in red in (a), (b), (c) and (d). (For interpretation of the colors in this
figure, the reader is referred to the web version of this article.)
are robust, with uncertainty less than about 0.3 s. Hence, we used
the relative relocation method to refine the horizontal location for
the events with AAF focal mechanisms (see more details in the
supplementary material). We selected an event (2015/04/26 07:09
(UTC) Mw6.7) as a reference, for which various catalogs show a
very small difference in the horizontal location (Fig. S9); we as-
sumed the averaged location between different catalogs is the best
location of this earthquake. To calibrate the path effect, we used
the residual between the theoretical predictions and handpicked
arrivals difference for two earthquakes observed at common sta-
tions. We then applied a grid search to find the best location
by minimizing the double differences (Fig. S9) (Waldhauser and
Ellsworth, 2000). Using a bootstrapping method, we also estimated
the uncertainty of the horizontal relocation. We discovered that
our relocated catalog is more consistent with that reported by Bai
et al. (2016), in which the data from a local temporal broadband
seismic network in China was used.
Traditionally, it is always difficult to precisely determine earth-
quake depth using P-wave arrival times, due to the trade-off be-
tween depth and origin time. The waveform modeling method we
applied to resolve earthquake depth does not suffer from such a
trade-off. Because the relative arrival times between direct P phase
and depth phases (e.g., pP, sP) which are key to constraining
earthquake depth—do not rely on accurate origin time (Fig. 7).
The resolution increases as the frequency increases, since depth
phases are easier to distinguish from the direct phases at higher
frequency. Here we performed high-frequency (0.5–1.5 Hz) wave-
form modeling to teleseismic P-waves to determine the depth of
the earthquakes we derived from AAF CAP inversions. As demon-
strated in Fig. S10, the depth phases of M5–5.5 events can be easily
identified, as they can be readily approximated as point sources
at this frequency band, and the surface-reflected phases (depth
phases) are well separated from the direct phases. During this pro-
cess, we fixed the focal mechanism of the earthquakes to those we
derived earlier and performed a grid search for earthquake depth;
thus, we obtained a best fitting depth for each station. Due to the
rapid increase in the number of global seismic stations, we have
hundreds of teleseismic P-waves and thus hundreds of estimations
of depth for each event. With a cut-off cross-correlation coefficient
of 70%, we can make a statistical analysis of the best depth and the
associated error (i.e., defined as 95% confidence ranges) (see more
details in the supplement material, Fig. S11). Observation and best-
fit synthetics at four representative stations are displayed in Fig. 7,
where both the direct phases and depth phases are easily identi-
fied and well-fitted by the synthetics. Here we arranged the events
according to their distance away from the MHT, and we can see
that the relative arrival time of depth phases are getting later as
the distance increases, indicating a deepening of the subducting
Indian plate.
88 X. Wang et al. / Earth and Planetary Science Letters 473 (2017) 83–93
Fig. 5. Source parameter inversion results for the calibration event 2015/04/26 16:26 (UTC) Mw5.0. The best solution and inversion parameters are shown in (a). (c) Displays
some of the waveform fits, with the black and red representing real data and synthetics, respectively. The station names are indicated at the upper right of each waveform
pair along with the epicenter distance (upper) and azimuth (lower) in degree on the left. The distribution of teleseismic stations is shown in (b) with time shifts and
cross-correlation coefficients (CCC) of the P-waves and SH-waves for all the stations used in this study. Lines connecting the source and receiver are colored by time shifts,
and stations are colored by CCC, which serves as an index for the quality of inversion. (d) Displays the waveform misfit as a function of depth, source duration and the best
focal mechanism. The best depth for this event is around 14 km. (a) Displays the best focal mechanism (blue lines) with the uncertainty estimated by a bootstrapping method
(gray lines). The corresponding distributions of the strike/dip/rake for the two fault plane solutions are shown in (e). The focal mechanism of this event is well constrained,
especially for the dip angle. (For interpretation of the colors in this figure, the reader is referred to the web version of this article.)
Apparently, the accuracy of depth is crucial to delineating the
fault geometry. To further verify the depth resolution, we con-
ducted a synthetic test in which we inverted the 3D synthetic data
at the same frequency ranges (0.5–1.5 Hz) as we used in the depth
phases modeling. A prominent feature in the region that might af-
fect the depth phases is the topography, which shows a strong gra-
dient across the Himalayan frontal area. Here we used a hybrid of
Spectral Element Method (SEM) (Komatitsch and Tromp, 1999) and
Direct Solution Method (DSM) (Cummins et al., 1994)to generate
3D synthetic data with numerical accuracy up to 2Hz (Fig. S12).
In the source region we used the SEM method to compute the 3D
wavefield for a velocity model that has the topography on top of
a layered background model; assuming the rest of the earth is a
1D model, we then propagated the wavefield to a teleseismic dis-
tance with the DSM method. One of the great advantages of such
a hybrid method is the reduction of computation time, as a global
SEM modeling to compute wavefields up to 2 Hz is too expensive
numerically. Here we assumed that a 1D background model is a
good approximation, since this region is mainly covered by hard
rock, and most of the earthquakes occurred at depths within the
uppermost crust—a thickness smaller than the resolution of any
current available 3D tomographic models. The complex subducted
slab structure at greater depth would similarly influence the depth
phases and direct phases, thus less impacting the relative timing
between them. To test the maximum possible impact of topogra-
phy, we placed the synthetic source at a location in the aftershock
zone that has the steepest topography (Fig. S12). The inversion of
the synthetic data shows that, despite the strong topography, its
maximum impact to the depth estimation is less than 1 km, which
is about the same as the error we estimated from the real data
(Fig. S13).
4. Results and discussion
With our refined earthquake location and focal mechanism, we
now have 16 events with a shallow dip angle thrust solution to
define the geometry of the MHT. Compared with the GCMT catalog,
which only has six thrust mechanism solutions, our results can
be used to better understand the fault zone structure. Our results
X. Wang et al. / Earth and Planetary Science Letters 473 (2017) 83–93 89
Fig. 6. Source parameter inversion results for the calibration event 2015/05/16 13:34 (UTC) Mw5.2. The rest of the figure’s caption is the same as in Fig. 5.
are shown in both map view (Fig. 2, Fig. 8d) and depth profiles
(Fig. 8a, b, c), with error bars indicating the 95% confidence levels
for horizontal and vertical locations. Most of the aftershocks are
distributed at the boundary of the coseismic slip of the mainshock,
particularly the down-dip region (Fig. 2), which is not unexpected,
as the stress changes from the mainshock concentrate on the edge
of the coseismic slip. It is also clear that many of these events,
including the Mw7.2 and Mw6.7aftershocks, are located to the
northeast of the coseismic slip, where the eastward and northward
slip asperities form a sharp corner. The northward slip asperity
is an interesting phenomenon in that it penetrates further to the
north of the aftershock belt in the down-dip region and there is
almost no aftershock surrounding the northern boundary of this
asperity. At the up-dip boundary of the coseismic slip, there are
five events with our focal mechanisms. Altogether, we have a good
spatial sampling of the aftershock zones, thanks to the additional
events with AAF focal mechanisms.
We projected all the shallow angle thrust events to the ver-
tical geological cross sections (Elliott et al., 2016; Hubbard et
al., 2016) and receiver function image (Duputel et al., 2016). As
shown in Fig. 8, our refined event locations are concentrated
at a depth range of 8–15 km, and to the first order consistent
with the proposed geometry of the MHT (thick black lines in
Fig. 8a–c), which is supported by various evidences that this is
the main plate boundary. In addition, our results distinctively de-
lineate a double-ramp feature with a near flat décollement sand-
wiched in between (yellow-dashed line in Fig. 8a–c). The flat por-
tion of this interface has a dip angle of 5–10, which is consis-
tent with the dip angle of long-period moment tensor solutions
(e.g. 7in GCMT) and that derived from geodetic data (Elliott
et al., 2016). The width (20 km) of this near horizontal in-
terface also agrees nicely with that of the eastern part of the
coseismic rupture area (slip >1m, Fig. 2), where most of the
down-dip aftershocks occurred. This suggests that the rupture of
the mainshock is primarily located on the flattened décollement.
The depth, which is relative to the free surface as we use depth
phases to determine the depth, of this portion is around 12 km,
in consistent with the result from geodetic and seismic inver-
sion for the mainshock coseismic slip (e.g., Elliott et al., 2016;
Wei et al., in preparation). The down-dip portion of the interface
steepens to a dip angle of 25and ranges from 13 km to 17 km
in depth and about 10 km horizontally. This ramp feature has been
commonly observed by previous geological and geophysical stud-
ies (e.g., Avouac, 2007; Duputel et al., 2016; Elliott et al., 2016;
Hubbard et al., 2016). At the shallower depth, two aftershocks
form another ramp that connects the up-dip end of the flat inter-
face. Compared with the deeper portion of the interface, this ramp
shows a similar dip angle (30), but is much smaller (a few km)
in dimension. This ramp (middle-ramp in Fig. 8b and d) agrees the
best with that reported by Hubbard et al. (2016)’s structural geol-
ogy study.
90 X. Wang et al. / Earth and Planetary Science Letters 473 (2017) 83–93
Fig. 7. (a) Ray paths of the direct P phase and the depth phases (pP and sP). At teleseismic distances, the ray paths for these phases are very close, except above the source
region. (b) Observations and best-fit synthetics at four representative stations, where both the direct phase and the depth phases are easily identified and well-fitted. The
corresponding focal mechanism and focal depth for each waveform pair are indicated on the right side, with the cross-correlation coefficients between data and synthetic
shown on the left. Here, we arranged the waveforms according to their distance away from the MHT. (c) Location of the four representative stations, with all avai lable stations
used in this study are shown as white triangles. All of the waveform fits for each earthquake can be found in Fig. S11 in Supplementary material.
The depth of our results are generally 2–3 km shallower than
those reported by Hubbard et al. (2016) and Duputel et al. (2016)
(Fig. 8b–d), while our results are more consistent with those re-
ported by Elliott et al. (2016) (Fig. 8a). Here we consider such
depth discrepancy to be a minor feature relative to the main fea-
ture of the fault geometry. Compared with the depth variation,
a more distinct difference between our result and these profiles
(Fig. 8a, b, c) is the turning point position of the down-dip ramp,
in which ours is about 10 km to the south of previous studies.
This discrepancy can be attributed to the lateral variation of the
geometry of the MHT (Fig. 8d), since most of the down-dip earth-
quakes we used are located in the eastern part of the coseismic
slip, while the geological and receiver function profiles are located
further to the west. Actually, the 3D fault geometry in Hubbard et
al. (2016) (Fig. 8d) does show that the down-dip ramp is located
about 10 km further to the south in the region where most of our
aftershocks are located. We also noted that the northern boundary
of the coseismic slip area in the eastern part is also about 10 km
to the south compared with the middle portion of the coseismic
slip. In general, our results show good agreement with the previ-
ous studies and provide independent constraints on the geometry
of the MHT.
Our results of the ramp geometry provide a lower bound es-
timate of ramp dimension, which can be used in earthquake dy-
namic modeling and seismic coupling study. Qiu et al. (2016)
simulated earthquake circles on a geological based fault geometry
(Hubbard et al., 2016) that has a double-ramp feature, and showed
that these barriers (ramps) are playing an important role in modu-
lating the earthquakes’ occurrence, particularly the big events that
rupture through the shallower ramp and reach the surface—similar
to the 1934 Mw8.2 earthquake located just to the east of the 2015
Gorkha earthquake. Such modeling studies usually assume specific
fault geometry and ignore the uncertainty of fault geometry. Our
estimate of ramp dimension can be used as a lower bound for
this portion of the fault geometry and narrow down the model
space in estimating the impact of fault geometry to earthquake
dynamic modeling. The complexity of fault geometry should also
be considered in the modeling of seismic coupling, especially for
the Himalaya frontal area where geodetic studies have shown that
the region is almost fully locked (Stevens and Avouac , 2015). Using
different fault geometry will change the coupling status and slip
partitioning on the fault, at least at the local scale. It is clear that
a more realistic fault geometry is required for more accurate cou-
pling modeling and thus can improve the seismic hazard estimate
in the region.
Our result indicates that the Mw7.2aftershock is not located
on the interface delineated by the smaller events (Fig. 8), suggest-
ing the event may have occurred on a secondary fault rather than
on the main plate interface (the MHT). The centroid depth of the
Mw7.2aftershock was estimated to be 12–14 km, which is sup-
ported by both seismic waveform modeling (Fig. 9a,b,c) and the
geodetic inversion (Wei et al., in preparation). The horizontal lo-
cation of the earthquake is well determined by the InSAR data
and agrees with our relocation result. Here we assume our relo-
cated epicenter is very close to the centroid location, since there
is no clear evidence of rupture directivity and the slip distribu-
tion of the Mw7.2aftershock is very compact (Fig. 2). The depth
of the Mw7.2 earthquake is about 5 km shallower than the other
smaller aftershocks on the down-dip ramp. The depth sensitivity
test (Fig. 9e) shows that the waveform fits at the depth of 18 km
is much worse than that at the preferred depth (14 km, relative
to surface). Additionally, the dip angle of shallow dipping fault
plane solutions does not fit with the ramp, which is only 10,
about 15smaller than that defined by the seismicity of smaller
aftershocks (25). The error of the dip angle estimated by a boot-
X. Wang et al. / Earth and Planetary Science Letters 473 (2017) 83–93 91
Fig. 8. (a) Geological cross-section from Elliott et al. (2016) superimposed with our focal mechanisms (red beach balls). Error bars indicate the 95% confidence levels for
horizontal and vertical locations. The heavy yellow dotted line shows the double ramps geometr y of the MHT (a flat décollement bounded on all sides by more steeply
dipping ramps) delineated by the seismicity in our study. (b) Geological cross-section from Hubbard et al. (2016) superimposed with our focal mechanisms. (c) S-to-P
receiver function stacking image from Duputel et al. (2016) superimposed with our focal mechanism. The thick lines in (a)/(b)/(c) mark the preferred geometry of the MHT
in each research paper. (d) Comparison between our refined earthquake catalog and Hubbard et al. (2016)’s 3D model. White contour lines show the inferred depth of the
MHT from Hubbard et al. (2016)’s 3D model, with intervals of 5 km. The blue and red dashed lines show the profile locations in Fig. 8a and b, respectively. The A-A’ profile
is used to project our refined earthquake location. Detailed explanations of the geological and receiver function cross-section can be found in the original papers.
strapping analysis of the long-period teleseismic body waves is
3(Fig. 9d), much smaller than the difference of dipping angles.
The shallow dip angle also is confirmed by previous research, and
ranges from 8to 11(Lay et al., 2016). These distinct and ro-
bust features all suggest that the Mw7.2aftershock has ruptured
a shallower unidentified fault above the MHT.
5. Conclusion
In summary, we have applied our broadband waveform mod-
eling methods to the 2015 Gorkha earthquake sequence to refine
the earthquake focal mechanism and location, which revealed a
distinct feature of double ramps on the MHT with flat décolle-
92 X. Wang et al. / Earth and Planetary Science Letters 473 (2017) 83–93
Fig. 9. Focal mechanism inversion results for the 2015/05/12 07:05 (UTC) Mw7.2 earthquake. The distribution of teleseismic stations and the selected veloci ty waveform
fits (black is data, red is synthetic) are shown in (b). (c) Displays the waveform misfits as a function of depth which shows the best depth for this event is around 14 km
(relative to the free surface). (a) Shows the best focal mechanism with uncertainty estimated by a bootstrapping method. The corresponding distributions of strike/dip/rake
for the two fault plane solutions are shown in (d). The arrows indicate the corresponding 95% confidence limits. (e) Shows the sensitivity tests for depth with a source time
function derived from finite fault inversion (Wei et al., in preparation), where the black is data and the red and blue are synthetics generated at 14 km and 18 km depth,
respectively. (For interpretation of the colors in this figure, the reader is referred to the web version of this article.)
ment sandwiched in between. Here we have provided a means
to study worldwide major earthquake sequences using teleseismic
data. This is particularly important for further global study, as most
of earthquakes take place offshore. Our approach can be general-
ized to other subduction zones and active seismic regions to refine
earthquake location and focal mechanism to better delineate the
fault geometry and improve understanding the tectonic setting in
the region as well as the physics of earthquakes.
X. Wang et al. / Earth and Planetary Science Letters 473 (2017) 83–93 93
Acknowledgements
This research was supported by the Earth Observatory of Sin-
gapore (EOS) startup grant M4430240.B50.706022. Seismic wave-
form data at teleseismic distance for this study were down-
loaded through the Incorporated Research Institutions for Seis-
mology (IRIS) website. Wave fo rm data at regional distance were
accessed from EOS technical office. We would like to thank techni-
cal office for installing and servicing the EOS-Nepal seismic array.
Sac2000, Taup and GMT were used for basic data processing and
figure development. We are grateful to Paul Tapponnier and Judith
Hubbard for useful discussions. We thank two anonymous review-
ers for their constructive reviews.
Appendix A. Supplementary material
Supplementary material related to this article can be found on-
line at http://dx.doi.org/10.1016/j.epsl.2017.05.032.
References
Adhikari, L.B., Gautam, P., Koirala, B.P., Bhattarai, M., Kandel, T., Gupta, R.M., Timsina,
C., Maharjan, N., Maharjan, K., Dahal, T., Hoste-Colomer, R., Cano, Y., Dandine, M.,
Guilhem, A., Merrer, S., Roudil, P., Bollinger, L., 2015. The aftershock sequence of
the 2015 April 25 Gorkha-Nepal earthquake. Geophys. J. Int. 203, 2119–2124.
Avouac, J.-P., 2007. Mountain building: from earthquakes to geological deformation.
Dynamic processes in extensional and compressional settings. In: Treatise on
Geophysics, vol. 6, pp. 377–439.
Bai, L., Liu, H.B., Ritsema, J., Mori, J., Zhang, T.Z., Ishikawa, Y., Li, G.H., 2016. Faulting
structure above the Main Himalayan Thrust as shown by relocated aftershocks
of the 2015 M(w)7.8 Gorkha, Nepal, earthquake. Geophys. Res. Lett. 43, 637–642.
Bollinger, L., Tapponnier, P., Sapkota, S.N., Klinger, Y., 2016. Slip deficit in central
Nepal: omen for a repeat of the 1344 AD earthquake? Earth Planets Space 68,
12.
Cattin, R., Avouac, J.P., 2000. Modeling mountain building and the seismic cycle in
the Himalaya of Nepal. J. Geophys. Res., Solid Earth 105, 13389–13407.
Cummins, P.R., Geller, R.J., Ta ke uchi, N., 1994. DSM complete synthetic seismograms
P–SV, spherically symmetrical, case. Geophys. Res. Lett. 21, 1663–1666.
Duputel, Z., Vergne, J., Rivera, L., Wittlinger, G., Farra , V., Hetenyi, G., 2016. The 2015
Gorkha earthquake: a large event illuminating the Main Himalayan Thrust fault.
Geophys. Res. Lett. 43, 2517–2525.
Ekstrom, G., Nettles, M., Dziewonski, A.M., 2012. The global CMT project 2004–2010:
centroid-moment tensors for 13,017 earthquakes. Phys. Earth Planet. Inter. 200,
1–9.
Elliott, J.R., Jolivet, R., Gonzalez, P.J ., Avouac, J.P., Hollingsworth, J., Searle, M.P.,
Stevens, V.L., 2016. Himalayan megathrust geometry and relation to topogra-
phy revealed by the Gorkha earthquake. Nat. Geosci. 9, 174–180.
Hayes, G.P., Briggs, R.W., Barnhart, W.D. , Ye ck , W.L., McNamara, D.E., Wald , D.J.,
Nealy, J.L., Benz, H.M., Gold, R.D., Jaiswal, K.S., Marano, K., Earle, P.S ., Hearne,
M.G., Smoczyk, G.M., Wald, L.A., Samsonov, S.V., 2015. Rapid characterization of
the 2015 M-w 7.8 Gorkha, Nepal, earthquake sequence and its seismotectonic
context. Seismol. Res. Lett. 86, 1557–1567.
Hubbard, J., Almeida, R., Foster, A., Sapkota, S.N., Burgi, P., Tapponnier, P., 2016.
Structu ral segmentation controlled the 2015 M-w 7. 8 Gorkha earthquake rup-
ture in Nepal. Geology 44, 639–642.
Kanamori, H., Rivera, L., 2008. Source inversion of W phase: speeding up seismic
tsunami warning. Geophys. J. Int. 175, 222–238.
Komatitsch , D., Tromp, J., 1999. Introduction to the spectral element method for
three-dimensional seismic wave propagation. Geophys. J. Int. 139, 806–822.
Lay, T., Ye , L., Koper, K.D., Kanamori, H., 2016. Assessment of teleseismically-
determined source parameters for the April 25, 2015 M W 7.9 Gorkha, Nepal
earthquake and the May 12, 2015 M W 7.2 aftershock. Tectonophysics 2016.
https://doi.org/10.1016/j.tecto.2016.05.023.
Lemonnier, C., Marquis, G., Perrier, F., Avouac, J.P., Chitrakar, G., Kafle, B., Sapkota, S.,
Gautam, U., Tiwari, D., Bano, M., 1999. Electrical structure of the Himalaya of
Central Nepal: high conductivity around the mid-crustal ramp along the MHT.
Geophys. Res. Lett. 26, 3261–3264.
Nabelek , J., Hetenyi, G., Vergne, J., Sapkota, S., Kafle, B., Jiang, M., Su, H.P., Chen, J.,
Huang, B.S., Hi, C.T., 2009. Underplating in the Himalaya–Tibet Collision Zone
revealed by the Hi-CLIMB experime nt. Science 325, 1371–1374.
Pandey, M.R., Tanduk ar, R.P., Avouac, J.P., Lave, J., Massot, J.P., 1995. Interseismic
strain accumulation on the Himalayan crustal Ramp (Nepal). Geophys. Res.
Lett. 22, 751–754.
Pearson, O.N., DeCelles, P. G . , 2005. Structural geology and regional tectonic signifi-
cance of the Ramgarh thrust, Himalayan fold-thrust belt of Nepal. Tectonics 24,
26.
Qiu, Q., Hill, E.M., Barbot, S., Hubbard, J., Feng, W.P., Lindsey, E.O., Feng, L.J., Dai,
K.R., Samsonov, S.V., Tapponni er, P. , 2016. The mechanism of partial rupture of
a locked megathrust: the role of fault morphology. Geology 44, 875–878.
Sapkota, S.N., Bollinger, L., Klinger, Y., Tapponnier, P., Gaudemer, Y., Tiwari, D., 2013.
Primary surface ruptures of the great Himalayan earthquakes in 1934 and 1255.
Nat. Geosci. 6, 71–76.
Schulte-Pelkum, V., Monsalve, G., Sheehan, A., Pandey, M.R., Sapkota, S., Bilham,
R., Wu, F., 2005. Imaging the Indian subcontinent beneath the Himalaya. Na-
ture 435, 1222–1225.
Stevens, V.L., Avouac, J.P., 2015. Interseismic coupling on the main Himalayan thrust.
Geophys. Res. Lett. 42, 5828–5837.
Tan, Y., Helmberger, D., 2007. A new method for determining small earthquake
source parameters using short-period P waves. Bull. Seismol. Soc. Am. 97,
1176–1195.
Wald hause r, F., Ellsworth, W.L., 2000. A double-difference earthquake location algo-
rithm: method and application to the northern Hayward fault, California. Bull.
Seismol. Soc. Am. 90, 1353–1368.
Wei, S., Chen, M., Wa ng, X., Graves, R., Lindsey, E., Wa ng, T., Karakas, C., Helmberger,
D., in preparation. Source model and ground shaking simulation of the 2015
Gorkha (Nepal) Earthquake sequence.
Wesnousky, S.G., 2006. Predicting the endpoints of earthquake ruptures. Nature 444,
358–360.
Zhan, Z.W., Helmberger, D., Simons, M., Kanamori, H., Wu, W.B., Cubas, N., Duputel,
Z., Chu, R.S., Tsai, V.C., Avouac, J.P., Hudnut, K.W., Ni, S.D., Hetland, E., Culaciati,
F.H.O., 2012. Anomalously steep dips of earthquakes in the 2011 Toho ku-Ok i
source region and possible explanations. Earth Planet. Sci. Lett. 353, 121–133.
Zhao, L.S., Helmberger, D.V., 1994. Source estimation from broad-band regional seis-
mograms. Bull. Seismol. Soc. Am. 84, 91–104.
Zhu, L.P., Helmberger, D.V., 1996. Advancement in source estimation techniques us-
ing broadband regional seismograms. Bull. Seismol. Soc. Am. 86, 1634–1641.

Supplementary resource (1)

... In the days following the main shock, dozens of the largest aftershocks were systematically studied at teleseismic distances (Letort et al. 2016;Wang et al. 2017), revealing some spatial heterogeneity of intermediate earthquakes realizations along the western, eastern and southern ends of the rupture (Fig. 2). This heterogeneity was attributed, and sometimes confirmed by the focal mechanisms (Fig. 2), to the presence of ramps and tear faults along the Main Himalayan Thrust fault and their spatial variations at depth Letort et al. 2016;Wang et al. 2017). ...
... In the days following the main shock, dozens of the largest aftershocks were systematically studied at teleseismic distances (Letort et al. 2016;Wang et al. 2017), revealing some spatial heterogeneity of intermediate earthquakes realizations along the western, eastern and southern ends of the rupture (Fig. 2). This heterogeneity was attributed, and sometimes confirmed by the focal mechanisms (Fig. 2), to the presence of ramps and tear faults along the Main Himalayan Thrust fault and their spatial variations at depth Letort et al. 2016;Wang et al. 2017). ...
... Bai et al. 2016;Kurashimo et al. 2019 ;Karplus et al. 2020). They confirmed that part of the seismicity that happened during the first months following the main shock is controlled by the geological structures at depth (Baillard et al. 2017;Hoste-Colomer et al. 2017;Wang et al. 2017 ;Bai et al. 2016Bai et al. , 2019Mendoza et al. 2019;Yamada et al. 2020). The small earthquakes associated with these structures therefore help reveal the morphology of the fault system at depth or the mechanisms at work (e.g. ...
Article
The Mw 7.9 April 25, 2015 Gorkha earthquake is the latest of a millenary-long series of large devastating Himalayan earthquakes. It is also the first time a large Himalayan earthquake and its aftershocks were recorded by a local network of seismic stations. In the five years following the mainshock, more than 31 000 aftershocks were located by this permanent network within the ruptured area, including 14 362 events with ML greater than 2.5, 7 events with ML > 6, including one large aftershock with Mw 7.2 on May 12, 2015. In 2020, five years after the mainshock, the seismicity rate along the ruptured fault segments was still about 5 times higher than the background seismicity before the Gorkha earthquake. Several bursts of earthquakes, sometimes organized in clusters, have been observed from a few days to several years after the mainshock. Some of these clusters were located at the same place as the clusters that happened during the decades of interseismic stress build-up that preceded the large earthquake. They also happened in the vicinity of the high frequency seismic bursts that occurred during the main shock. These heterogeneities contribute to a persistent segmentation of the seismicity along strike, possibly controlled by geological structural complexities of the Main Himalayan Thrust fault. We suggest that these pre-2015 clusters revealed the seismo-geological segmentation that influences both the coseismic rupture and the postseismic relaxation.
... This ramp corresponds to the thermally controlled brittle-ductile (kinematically locked-creeping) transition zone highlighted by a band of microseismicity, showing an area of high stress buildup; the midcrustal ramp also corresponds to local anomalies in erosion and thermochronometric ages at the surface, where it has been suggested to be associated with a midcrustal duplex below (Fig. 1B) (e.g., Bilham et al., 1997;Bollinger et al., 2006;Ader et al., 2012;Grandin et al., 2012;Avouac, 2015;Bilham, 2019;Ghoshal et al., 2020;Johnston et al., 2020). The M w 7.8 2015 CE Gorkha (Nepal) earthquake highlighted the seismic hazard in the Himalaya and revealed a more accurate picture of the geometry of the MHT (e.g., Elliott et al., 2016;Wang et al., 2017), enabling us to perform detailed calculations in this study. ...
... The midcrustal ramp in the central Himalaya dips at ∼26° and the upper décollement at ∼8° (Elliott et al., 2016;Wang et al., 2017), so the change in fault dip (φ) is ∼18°. Because the midcrustal ramp steps up from one décollement to another, (1) the ratio of transferred slip (i.e., the ratio of the updip slip to the downdip slip) for the midcrustal ramp can be defined as the slip ratio crossing the upper bend of the midcrustal ramp; and (2) the change in the fault dip and the cutoff angle are identical and the axial angle can be calculated by (after Suppe, 1983 (2) ...
... Geodetically, small-scale fault patches with lower coupling may not be identified due to the limited spatial resolution of measurements (e.g., Mukul et al., 2018); the choice of geodetic model also impacts the estimates of moment accumulation significantly (Fig. 3C). Additionally, the aftershocks of the 2015 Gorkha earthquake and microseismicity reveal previously unidentified faults, indicating that the more complicated structures and sporadic out-of-sequence thrusting cannot be disregarded (e.g., Whipple et al., 2016;Wang et al., 2017;Dey et al., 2019;Laporte et al., 2021). ...