of 14
Special Section:
2019 Ridgecrest, California, Earthquake Sequence
Detailed 3D Fault Representations for the 2019
Ridgecrest, California, Earthquake Sequence
Andreas Plesch
*1
, John H. Shaw
1
, Zachary E. Ross
2
, and Egill Hauksson
2
ABSTRACT
We present new 3D source fault representations for the 2019 M 6.4 and M 7.1 Ridgecrest
earthquake sequence. These representations are based on relocated hypocenter catalogs
expanded by template matching and focal mechanisms for M 4 and larger events.
Following the approach of
Riesner
etal.
(2017)
, we generate reproducible 3D fault geom-
etries by integrating hypocenter, nodal plane, and surface rupture trace constraints. We
used the southwest
northeast-striking nodal plane of the 4 July 2019 M 6.4 event to con-
strain the initial representation of the southern Little Lake fault (SLLF), both in terms of
location and orientation. The eastern Little Lake fault (ELLF) was constrained by the 5 July
2019 M 7.1 hypocenter and nodal planes of M 4 and larger aftershocks aligned with the
main trend of the fault. The approach follows a defined workflow that assigns weights to a
variety of geometric constraints. These main constraints have a high weight relative to
that of individual hypocenters, ensuring that small aftershocks are applied as weaker con-
straints. The resulting fault planes can be considered averages of the hypocentral locations
respecting nodal plane orientations. For the final representation we added detailed, field-
mapped rupture traces as strong constraints. The resulting fault representations are gen-
erally smooth but nonplanar and dip steeply. The SLLF and ELLF intersect at nearly right
angles and cross on another. The ELLF representation is truncated at the Airport Lake fault
to the north and the Garlock fault to the south, consistent with the aftershock pattern. The
terminations of the SLLF representation are controlled by aftershock distribution. These
new 3D fault representations are available as triangulated surface representations, and
are being added to a Community Fault Model (CFM;
Plesch
etal.
, 2007
,
2019
;
Nicholson
etal.
, 2019
) for wider use and to derived products such as a CFM trace map and viewer
(
Su
etal.
, 2019
).
KEY POINTS
We present a 3D model of the source faults for the 2019
Ridgecrest, CA earthquake sequence.
We employ an objective method of defining faults using
hypocenter, focal mechanism, and geologic constraints.
Source faults consist of two main segments, the Southern
and Eastern Little Lake faults, and six large splays.
Supplemental Material
INTRODUCTION
Many of the fundamental aspects of earthquake science,
including event nucleation, dynamic rupture and wave propa-
gation, stress triggering, and other phenomena, are impacted
by the properties of fault zones, including its location and
geometry. Moreover, earthquake hazards assessments are
largely based on inferences about the location and magnitudes
of past and future earthquakes, which often involves defining
the activity and slip rates on faults using geologic, seismologic,
or geodetic observations. Among the most influential proper-
ties impacting earthquake phenomena and their associated
hazards are source fault location and geometry. As a result,
there have been many comprehensive efforts to map active
fault zones in earthquake-prone regions. In California, these
efforts began with the mapping of individual fault zones such
as the San Andreas (e.g.,
Lawson
et al.
, 1908
;
Allen, 1957
;
Dibblee, 1973
). These efforts expanded to comprehensive map-
ping and classification of active fault traces with regional and
national fault trace databases maintained by the California and
U.S. Geological Surveys, respectively (
Jennings and Bryant,
2010
; see
Data and Resources
). In recent decades, these maps
have been extended to develop 3D digital representations of
1. Department of Earth & Planetary Sciences, Harvard University, Cambridge,
Massachusetts, U.S.A.; 2. Seismological Laboratory, California Institute of
Technology, Pasadena, California, U.S.A.
*Corresponding author: andreas_plesch@harvard.edu
Cite this article as
Plesch, A., J. H. Shaw, Z. E. Ross, and E. Hauksson (2020).
Detailed 3D Fault Representations for the 2019 Ridgecrest, California, Earthquake
Sequence,
Bull. Seismol. Soc. Am.
110
, 1818
1831, doi:
10.1785/0120200053
© Seismological Society of America
1818
Bulletin of the Seismological Society of America
www.bssaonline.org Volume 110 Number 4 August 2020
Downloaded from https://pubs.geoscienceworld.org/ssa/bssa/article-pdf/110/4/1818/5114416/bssa-2020053.1.pdf
by California Inst of Technology user
on 27 August 2020
active fault systems in California (e.g.,
Seeber and Ambruster,
1995
;
Carena and Suppe, 2002
;
Shaw
et al.
, 2002
), and else-
where (
Plesch and Oncken, 1999
). The Southern California
Earthquake Center
s (SCEC
s) Community Fault Model
(CFM) represents the most comprehensive of these efforts,
including detailed 3D representations or more than 300 active
fault segments that are deemed capable of generating moder-
ate-to-large earthquakes (
Plesch
et al.
, 2007
,
2016
;
Nicholson
et al.
, 2019
). Fault surfaces are defined in the CFM by a variety
of data constraints, including geologic fault traces, earthquake
hypocenters and focal mechanisms, well penetrations, and seis-
mic reflection and refraction studies. The SCEC CFM is widely
used in crustal deformation modeling, wave propagation sim-
ulations, earthquake simulators, and probabilistic seismic haz-
ards assessment (e.g., Uniform California Earthquake Rupture
Forecast, version 3 [v.3]). The CFM also directly contributes
to other property modeling efforts, such as the development
of 3D seismic-wavespeed models (e.g., Community Velocity
Models;
Magistrale
et al.
, 2000
;
Süss and Shaw, 2003
;
Shaw
et al.
, 2015
) that are used in many aspects of seismology,
including strong ground motion prediction.
Although the CFM has been shown to represent the signifi-
cant majority of earthquake sources in southern California
(
Evans
et al.
, 2020
), it is by no means complete. Moreover,
there is substantial uncertainty in the geometric representa-
tions of many fault zones given a lack of subsurface data.
The 2019 Ridgecrest, California, earthquake sequence (
Liu
et al.
, 2019
;
Ross, Idini,
et al.
, 2019
;
Chen
et al.
, 2020
;
Hudnut
et al.
, 2020
;
Ponti
et al.
, 2020
) is a clear example of the need to
expand and refine these models. The
M
6.4 and 7.1 Ridgecrest
earthquakes and their associated foreshocks and aftershocks
occurred on faults broadly within the regional Little Lake fault
zone associated with a series of highly segmented fault traces
(
Wills, 1988
) with evidence of previous activity (
Kozaci
et al.
,
2019
;
Thompson Jobe
et al.
, 2020
). However, neither of the
source faults, which we refer to as the eastern Little Lake fault
(ELLF) and southern Little Lake fault (SLLF), were represented
in the CFM.
This article describes our efforts to develop comprehensive,
3D representations of the Ridgecrest source faults. We employ
an objective, reproducible method (
Riesner
et al.
, 2017
)to
model these fault geometries based on weighting constraints
from mapped fault ruptures, earthquake hypocenters, and focal
mechanism. Parts of the Ridgecrest sequence terminated into,
or otherwise interacted with, other faults represented in the
CFM, including the Airport Lake and Garlock faults. Thus,
we sought to develop representations of the Ridgecrest faults
that were compatible with these other CFM representations.
We developed initial fault models in the days after the events
(
Plesch
et al.
, 2019
) and have subsequently refined these as
the field mapping and earthquake catalogs have evolved.
Here, we document these refinements to the Ridgecrest
source fault representations and describe in detail how the
improvements to the data constraints have impacted the latest
fault representations.
THE JULY 2019 RIDGECREST SEQUENCE
The 2019 earthquake sequence occurred in the area of the
northern Indian Wells Valley of eastern California, about
10 km northeast of the town of Ridgecrest (Fig.
1
). This area
is part of the eastern California shear zone (ECSZ), a tectonic
system of distributed right-lateral shear that accommodates a
component of relative Pacific and North American plate
motion. In the past few decades, the ECSZ has sourced two
major earthquakes, the 1992
M
7.3 Landers and 1999
M
7.1
Hector Mine events, with their epicenters located to the south
of the Ridgecrest earthquakes in the Mojave Desert. In the
decades prior to the 2019 Ridgecrest sequence, the Indian Wells
Valley has experienced numerous earthquakes that are typically
associated with swarms that extend over months (
Hauksson
et al.
,1995
). The largest events were the 1995
M
L
5.4 and
5.8 earthquakes. The 17 August
M
L
5.4 earthquake ruptured
anorth
south-trending fault and exhibited components of both
normal and right-lateral motion. The 20 September
M
L
5.8
event ruptured a north-northwest-striking fault and exhibited
right-lateral motion. Both events had thousands of aftershocks,
and although no direct surface ruptures of these source faults
were identified, both events produced triggered slip on the
Airport Lake fault, which lies to the north of the epicentral zones
(
Roquemore and Zellmer, 1986
;
Hauksson
et al.
,1995
).
The 2019 Ridgecrest sequence began with the 4 July
M
6.4
earthquake, which occurred about 10 km northeast of
Ridgecrest on a northeast
southwest-striking, left-lateral
strike-slip fault (Fig.
1
). The event had several moderate fore-
shocks (
Huang
et al.
, 2019
) and an extensive aftershock
sequence. The 5 July
M
7.1 earthquake occurred about 34 hr
later and 7 km to the west of the
M
6.4 earthquake. The
M
7.1
event ruptured a northwest
southeast-striking fault and exhib-
ited right-lateral strike-slip faulting (
Liu
et al.
, 2019
;
Ross,
Idini,
et al.
, 2019
;
Stewart
et al.
, 2019
;
Chen
et al.
, 2020
;
Hudnut
et al.
, 2020
;
Ponti
et al.
, 2020
). The earthquakes pro-
duced extensive ground rupture along planes consistent with
their preferred nodal planes and aftershock alignments
(
Kendrick
et al.
, 2019
). Notably, the two events are thought
to have occurred on interpenetrating, nearly orthogonal fault
systems (
Barnhart
et al.
, 2019
;
Ross, Idini,
et al.
, 2019
;
Hauksson and Jones, 2020
; Fig.
1
). Both the main events
occurred at about 8 km depth, with most aftershocks occurring
above 10 km depth and some extending to about 12 km depth
(
Ross, Idini,
et al.
, 2019
) although for most of the mainshock
rupture area, the aftershocks are generally shallower than
10 km. The aftershocks for the
M
7.1 event extended about
30 km south of the mainshock epicenter and generally termi-
nated at the Garlock fault. The northern termination of the
rupture is characterized by a diffuse zone of aftershocks
extending from about 8 to 25 km northwest of the mainshock
Volume 110 Number 4 August 2020 www.bssaonline.org
Bulletin of the Seismological Society of America
1819
Downloaded from https://pubs.geoscienceworld.org/ssa/bssa/article-pdf/110/4/1818/5114416/bssa-2020053.1.pdf
by California Inst of Technology user
on 27 August 2020
epicenter (Fig.
1
). This region is bounded to the west by the
Airport Lake fault zone.
Both the 4 and 5 July events occurred in a region where
there were a series of mapped faults thought to exhibit
Quaternary displacements (
Wills, 1988
). However, these
surface fault traces were gener-
ally short and discontinuous,
and did not belie the rather
planar, nearly continuous 15
and 45 km surface ruptures
for the
M
6.4 and 7.1 events,
respectively (
Kendrick
et al.
,
2019
). As noted, the earth-
quakes did not occur on a
fault represented in the SCEC
CFM. However, the
M
7.1
mainshock ruptured a fault
that is parallel to the Little
Lake fault, with the
M
6.4 event
potentially providing a struc-
tural linkage between the two
(Fig.
1
). Thus, for our analysis,
we will refer to the source
faults for the
M
6.4 and 7.1
events as the SLLF and ELLF,
respectively.
DATA AND METHODS
We base our new 3D source
fault representations of the
2019 Ridgecrest sequence on
four principal datasets: (1) field
mapping of rupture traces pro-
vided by
Kendrick
et al.
(2019)
;
(2) a relocated hypocenter cata-
log significantly expanded by
quake template matching
(QTM) methods (QTM catalog,
provided by Ross for this study,
through 25 July 2019); (3) a
relocated hypocenter catalog
that uses a regional set of sta-
tions to determine absolute
locations (
Hauksson
et al.
,
2020
); and (4) selected focal
mechanisms provided by the
Southern California Earthquake
Data
Center
(SCEDC;
Hauksson
et al.
,2020
). The
mapped trace of the rupture is
based on thousands of field
observations as well as on
analysis of remote sensing data.
Only field confirmed trace data were used for our modeling
effort. The first relocated hypocenter catalog contains more than
46,000 events, and was expanded from the Southern California
Seismic Network (SCSN) catalog by sensitive template match-
ing, allowing detection of events down to
M
0.3. The second
M 6.4
M 7.1
10 km
Indian Wells valley
Ridgecrest
ECSZ
Mojave
California
Pacific
SAF
North America
200 km
Airport Lake fault
Garlock fault
500
1000
1500
1000
1000
1000
420,000
430,000
440,000
450,000
460,000
470,000
3,990,000
3,980,000
3,970,000
3,960,000
3,950,000
3,930,000
Little Lake fault
3,940,000
Figure 1.
Map of the 2019 Ridgecrest sequence showing datasets used to constrain our 3D fault representations.
Dots show epicenter locations: red events occurred after 1 July 19; gray events before. Blue traces are fault ruptures
compiled by
Kendrick
et al.
(2019)
; light blue traces are a subset of these used as constraints for fault surface
interpolation. Black lines show fault traces from Qfault database (Quaternary fault and fold database [see
Data and
Resources
]). Blue
black polygons are Community Fault Model representations of Little Lake, Airport Lake, and
Garlock faults (
Plesch
et al.
, 2007
,
2016
;
Nicholson
et al.
, 2019
). Focal mechanisms (upper hemisphere) are from
the Southern California Earthquake Data Center (
Hauksson
et al.
,2020
); black
white circles are focal mechanism of
mainshocks. Gray outlines show the fault surfaces modeled in this study for reference. Circled numbers refer to
structural domains used in our analysis: 1, southern domain; 2, interaction domain; 3, central domain; 4, northern
domain. Contour interval is 100 m. Coordinates are for UTM
zone 11N, North American Datum 1927 (NAD 27). (Inset)
Location of event (large star), of the 1992 Landers and 1999
Hector Mine earthquakes (sma
ll stars), of the eastern
California shear zone (ECSZ) and of the San Andreas fault (S
AF). Red dots are relocated earthquakes from 1981 to 2019.
1820
Bulletin of the Seismological Society of America
www.bssaonline.org Volume 110 Number 4 August 2020
Downloaded from https://pubs.geoscienceworld.org/ssa/bssa/article-pdf/110/4/1818/5114416/bssa-2020053.1.pdf
by California Inst of Technology user
on 27 August 2020
relocated catalog contains about 33,000 events of the sequence,
extending up to 31 October 2019. Both catalogs clearly resolve
internal structure to the rupture, including multiple splays for
both the main ELLF and the SLLF. Our goal is to define the
main, continuous splays of these two faults and other large
(
>
5 km length) structures that are defined by the seismicity.
The earthquake sequence clearly illuminated other smaller fault
segments, as well as distributed zones of seismicity around the
main fault splays that we do not attempt to model. We incor-
porated all
M
4 and larger events listed by the SCEDC focal
mechanism catalog in our analysis. We assumed that the fault
plane is locally parallel to one of the nodal planes.
The method we use to interpolate and fit 3D surfaces to the
data is presented by
Riesner
et al.
(2017)
. This fault modeling
approach is based on constraining a scalar field in a meshed
box volume, the interpolation mesh, finding a best 3D fit to
the constraints, and then extracting a fault representation as
an isosurface from the scalar field (Fig.
2
;
Frank
et al.
, 2007
).
Three constraints are implemented in our workflow. The first
is used to fit a surface to 3D point locations (e.g., traces or
hypocentral locations), and the second one to locally fit to a
3D orientation based on nodal planes. A third constraint con-
trols overall roughness of the fault surface by limiting the gra-
dient of the scalar field to a constant between neighboring
regions. This constraint is necessary, in practice, to allow for
converging solutions. All constraints can be weighted. For a
given fault segment, we typically apply a weight to the collec-
tion of all hypocenters that is similar to the summed weight of
all points along the field-mapped surface trace. The orientation
constraint naturally has a larger impact on fault-plane regions
that are closer to the location of the constraint. Because the
focal mechanism catalog lists event locations that are not
always consistent with locations in relocated catalogs, we gen-
erally assign relatively lower weights to orientation constraints.
Figure
3
shows the effect of varying the weight of these orien-
tation constraints. For the gradient constraint, we use a con-
stant weight of 0.1 that effectively serves as a baseline for all
other weights. The final fitting is a discrete-smooth inter-
polation (DSI;
Mallet, 1992
) of the scalar field that iteratively
minimizes residuals to the constraints in a least-squares sense.
Computationally, the interpo-
lation is efficient and only takes
a few seconds on standard
desktop PC to converge.
The modeling workflow
consists of a series of defined
steps that can be iterated and
reproduced given changes to
input data and workflow
parameters such as the weight-
ing of constraints (Fig.
4
). The
first step involves collecting the
input data to model an identi-
fied fault. Input data typically
include the surface rupture
trace and a selection of hypo-
centers that show alignment
with the rupture trace and focal
mechanism nodal planes in the
Figure 2.
Constraint-based interpolation of fault representations.
(a) Constraints used to interpolate a scalar
f
in a volume.
u
i
are the
barycentric coordinates,
α
i
are the coordinates of nodes, and
N
,
d
are the
strike and dip vectors of orientation constraint Centered dots represent dot
products (b) resulting field interpolated on the tetrahedral mesh. The fault
surface corresponds to the constant value
Φ
of the variable
f
(adapted from
Riesner
et al.
, 2017
).
Figure 3.
Example of how lowering the weight of orientation constraints (red nodal planes) adjusts the resulting
modeled fault surface (white contours). (a) A weight of 0.5; (b) a weight of 0.1. Arrow highlights the region of the
fault surface that is affected most.
Volume 110 Number 4 August 2020 www.bssaonline.org
Bulletin of the Seismological Society of America
1821
Downloaded from https://pubs.geoscienceworld.org/ssa/bssa/article-pdf/110/4/1818/5114416/bssa-2020053.1.pdf
by California Inst of Technology user
on 27 August 2020
vicinity of these hypocenters. Not all modeled faults show sur-
face rupture. Rather, such faults are recognized from subplanar
alignments in the hypocentral distribution. This initial deter-
mination is typically accomplished by visual inspection of the
data. We found use of a fully 3D, geoscience-focused, and inte-
grative modeling environment (Emerson GOCAD/SKUA),
efficient for this purpose. Subsequently, this process of identi-
fying faults illuminated by seismicity can be refined by a stat-
istical analysis of the distances between hypocenters and a
modeled 3D fault surface.
After collecting all input data, the interpolation constraints
are defined. Location constraints are typically defined first.
However, the order in which these constraints are defined is
generally not significant. The vertices of the polylines that
make up the rupture trace often represent one location con-
straint. Connectivity between vertices is not considered. In
addition, for numerical stability, it is necessary to have at least
one vertex as a location constraint with a different scalar value
than the constant target value for the fault. Conceptually, this
off-fault location establishes an initial gradient that then can be
perturbed to fit the main constraints. This single vertex
constraint is defined at the mesh perimeter. However, its exact
location is not important for the final fault geometry in the
space where the fault is defined by other constraints. The next
distinct location constraints are the hypocentral locations
selected for the fault. These can be assigned a weight different
from that of the fault trace locations. The weight given is per
hypocenter. In principle, it would be possible to vary the
weights for each hypocenter or groups of hypocenters by con-
sidering, for example, moment release, location uncertainty, or
distance from an initial model fault. While we will explore this
possibility, we found that the parameter space is sufficiently
large to arrive at solutions that match observations by
assigning the same weight to all hypocenters. After the location
constraints are defined, the orientation constraints are pro-
vided using nodal planes of collected focal mechanisms. The
orientation constraint is given in terms of a gradient control
vector. This vector controls the direction of the local gradient
of the scalar field during the interpolation. We choose to use
only direction control with this constraint, without considering
the magnitude of the gradient or its polarity. This choice
addresses the potential mismatch of locations of focal mech-
anisms and of the hypocenter for the same event. As a practical
matter, we use the normal vectors of a disk representing the
nodal plane as the gradient control vector, at the vertices of
a polygon at the circumference of the disk. There are multiple
such normal vectors (eight in our representation) for each
nodal plane that we account for by lowering the weight per
vertex constraint proportionally. As the final constraint, a
global constant gradient constraint is defined, which is applied
throughout the interpolation volume and is not tied to a loca-
tion. This constraint is used during the interpolation to keep
the variation in gradient small between neighboring regions, in
effect controlling overall roughness.
After all constraints are set up, a DSI (
Mallet, 1992
) inter-
polation of the scalar field is performed, using a matrix formu-
lation (
Muron
et al.
, 2005
), with a fitting factor of 2, unlimited
numbers of iterations, and a convergence error of 10
7
. The
fitting factor is a trade-off between overall smoothness and
fit to given constraints. The fault geometry is implicitly defined
by the scalar field as an isosurface of the target value given
for the constraints (Fig.
2
, we use zero as the target value
for the isosurface). We use standard extraction of the isosurface
at all edges of the tetrahedra that constitute the interpolation
mesh to generate a triangulated surface of the modeled fault.
As an isosurface, the fault surface is defined for the complete
interpolation volume. Because the interpolation volume
encompasses an enlarged region surrounding the complete
earthquake sequence, defining the lateral and vertical limits
of a fault becomes a separate, next step. For the upper vertical
limit, we use topography for faults that have a rupture trace
and the upper limit of seismicity for faults, which did not rup-
ture the surface. The lower vertical limit is provided by the
regional base of seismicity surface, which is used to limit
Delete all constraints
Add location constraint for trace vertices
Add location constraint for hypocenters
Add gradient constraint for nodal planes
Add constant gradient constraint
Add location constraint for off fault location
DSI interpolation of scalar field
Extract isosurface from scalar field
Adjust weights
Weight: 1
Weight: 0.001
Record parameters and isosurface as fault
Weight: 100
Weight: 0.5
Weight: 0.1
Fit. factor: 2
Automated
Define interpolation constraints
Figure 4.
Steps of the workflow designed to be partially automated (gray
box) with input of typical weights. Automation does not include the
selection of the various constraints. The constraint selection criteria for each
fault are described in the main article. DSI, discrete-smooth interpolation.
1822
Bulletin of the Seismological Society of America
www.bssaonline.org Volume 110 Number 4 August 2020
Downloaded from https://pubs.geoscienceworld.org/ssa/bssa/article-pdf/110/4/1818/5114416/bssa-2020053.1.pdf
by California Inst of Technology user
on 27 August 2020
the depth of the CFM (
Nazareth and Hauksson, 2004
;
Plesch
et al.
, 2007
). The lateral fault limits are either given by other
faults, such as the Garlock and Airport Lake faults, or by the
extent of seismicity.
The workflow is currently implemented using the
StructureLab plugin (
Caumon
et al.
,2019
, Research for
Integrative Numerical Geology [RING] group at École
Nationale Supérieure de Géologie, France) for Emerson/
Paradigm SKUA/GOCAD and a small macro assembling the
necessary steps in a convenient manner, allowing for rapid iter-
ations. We use manual iterations to fine-tune weighting within
the ranges outlined earlier.
INITIAL, NEAR-REAL-TIME INTERPRETATION
In a rapid response to the earthquake sequence, we used the
workflow outlined earlier to generate a representation of the
3D source fault geometry (Fig.
5
;
Plesch
et al.
, 2019
). This
analysis incorporated a very limited dataset, consisting of
SCSN catalog locations (
Hutton
et al.
, 2010
) from the first four
days of the sequence and focal mechanisms of
M
4 and larger
events. Field mapping was not yet available.
Our goal was to define the main rupture planes associated
with these earthquakes, and we did not make an attempt to
resolve subsidiary fault splays in more detail. As such, we iden-
tified two trends, the main southeast
northwest trend and the
orthogonal southwest
north-
east, and, assigned all hypocen-
ters and nodal planes in each
trend, as constraints in the
workflow. This rapid pro-
cedure resulted in two contigu-
ous 3D fault planes that proved
very useful for communicating
the overall structure to the
community and for placing it
into the context of the CFM
(
Plesch
et al.
, 2007
) and
Qfault databases (see
Data
and Resources
). These general-
ized fault planes also serve as a
reference to define the addi-
tional structure our more
detailed analyses described
later was able to resolve.
In this initial representa-
tion, the ELLF is fairly linear
and subvertical, extending
from the Garlock fault in the
south to the Airport Lake fault
in the north. The SLLF is also
fairly linear and subvertical,
and crosses and is orthogonal
to the ELLF.
DEVELOPING DETAILED FAULT REPRESENTATIONS
To develop a more detailed representation of the 2019
Ridgecrest source faults using the complete set of high-quality
hypocenter locations and detailed rupture trace maps, we
divide the rupture into a number of structural domains defined
by rupture trace and seismicity patterns along the ELLF and
SLLF (Fig.
1
). These include (1) a southern domain, where
two subparallel faults are mapped, consisting of an eastern
and western branch. We represent the eastern branch as part
of a continuous, main ELLF (south section) and the western
branch as a splay (west splay); (2) a complex zone of interac-
tion between the ELLF and SLLF. Faults in this region generally
intersect and interpenetrate one another. We subdivide the
SLLF into subparallel strands consisting of a shorter, northern
segment and a main, southern strand; (3) a central domain,
which includes the
M
7.1 mainshock epicentral zone. The
ELLF in this region consists of the main strand to the west,
and an eastern, more limited branch. The eastern branch is
apparent in the mapped field rupture traces and distinguished
by an intervening gap in seismicity; and (4) the northern
domain, where distributed seismicity including a large number
of
M
4 plus events define a number of subparallel faults that
trend obliquely with respect to the main ELLF trend, as well as
a number of cross faults. We represent this domain by
Figure 5.
Initial, simplified 3D source fault geometry developed in the weeks following the earthquake sequence
(
Plesch
et al.
, 2019
).
Volume 110 Number 4 August 2020 www.bssaonline.org
Bulletin of the Seismological Society of America
1823
Downloaded from https://pubs.geoscienceworld.org/ssa/bssa/article-pdf/110/4/1818/5114416/bssa-2020053.1.pdf
by California Inst of Technology user
on 27 August 2020
describing the four main seismicity lineaments as discrete
faults. In addition, there are a number of small cross faults
throughout all four domains of which we represent the six larg-
est: two in the southern domain, one in the central domain,
and three in the northern domain.
Having identified distinct structural domains from rupture
trace and seismicity patterns, we separated input data for each
of the 14 identifiable faults and applied the modeling workflow.
We first model the main strand of ELLF (Fig.
6d
) by selecting
the most continuous traces mapped in the field along the
(a)
m
m
m
m
M 7.1
M 6.4
M 7.1
Cross, south, south
South, main
South, west
m
Southern Little Lake fault
Cross, center
Cross, south, north
North
Main
North, e. boundary
North, east
North, center
North, west
Cross, north, south
1
Eastern Little Lake fault
North, main
M
Center, east
,
south
South, main
South, we
So
h
A
Sout
h,
A'
(b)
(d)
460,000
,
470,000
,
460,000
,
450,000
,
430,000
,
440,000
,
430,000
,
470,000
,
3,940,000
3,950,000
3,940,000
3,980,000
3,960,000
3,970,000
3,940,000
3,930,000
(c)
Cross, north, north
1,000
0
–1,000
–2,000
–3,000
–4,000
–5,000
–6,000
–7,000
–8,000
–9,000
–10,000
–11,000
–12,000
Z
(m)
–1,000
–2,000
–3,000
–4,000
–5,000
–6,000
–7,000
–8,000
–9,000
–10,000
–11,000
–12,000
Z
(m)
0
2,500
5,000
0
2,000
4,000
0
7,000
14,000
0
4,000
8,000
Figure 6.
Maps showing data used as constraints for the modeled faults:
(a) southern area of eastern Little Lake fault (ELLF); (b) central area with
intersection of southern Little Lake fault (SLLF) and ELLF; (c) northern area of
ELLF; and (d) main strand of ELLF. Blue dots represent hypocenters, circled in
(b,d) to distinguish constraints for eastern splay of the ELLF, central section;
red traces are field-mapped sections; yellow
blue circles are focal mech-
anisms (upper hemisphere projection), with the nodal plane that is the
closest to the fault trend used as constraint. Black circles in (a
c) and
colored circles in (d) are hypocenter not used as constraints. Color bars show
depth scale. Coordinates in UTM zone 11N, NAD 27 coordinates.
1824
Bulletin of the Seismological Society of America
www.bssaonline.org Volume 110 Number 4 August 2020
Downloaded from https://pubs.geoscienceworld.org/ssa/bssa/article-pdf/110/4/1818/5114416/bssa-2020053.1.pdf
by California Inst of Technology user
on 27 August 2020
southeast
northwest trend. We then added smaller, more dis-
connected traces in the northern structural domain that cor-
responded with the main trend of the fault. The selected trace
supplies about 12,000 vertices as constraints input. We then
identified 19 focal mechanisms, including the
M
7.1 main-
shock, along the same trend. We extracted the nodal plane
from each focal mechanism following that trend. The selection
of hypocenters for this fault occurred in three phases of con-
tinued refinement. The first selection included about 33,000
events in a 6-kilometer-wide band along the trend, excluding
events in the southern branch previously identified. In two
subsequent phases, this selection was trimmed to about
14,000 events by excluding clusters of events that separated
from the main fault trend and thus used as input data for other
modeled fault splays. Most of these excluded events were
located in the diffuse seismicity of the northern domain.
The final set of constraints results in a 60-kilometer-long fault
surface that is generally steeply dipping but changes from a
moderate northeast dip in the southern (75°) and central
(55°) domains to a steep southwest dip (85°) in the northern
domain. Below about 6 km depth, the fault becomes nearly ver-
tical. With 14 out of the 19 nodal planes within 1.5 km from
the fault, the fault surface is locally curved, with mean local
curvatures reaching a maximum of 0.001. The smoothness
of the shallow fault geometry is controlled by the trace and
exhibits comparable amounts of mean curvature.
As previously described, we modeled two faults branching
from the main strand of the ELLF. In the southern domain, this
includes a western branch (Fig.
6a
) for which three segments of
rupture traces were selected corresponding to about 2000 ver-
tices. Along the trend of this trace, we selected about 1000
hypocenters of aligned events that are clearly separated from
the main trend of seismicity by a gap of about 2 km width. For
this splay, there is an increased number of events in the QTM
catalog that have a strong planar alignment. These hypocenters
show clearly that there is a single, steeply dipping fault plane at
seismogenic depths (Fig.
7
). We used this alignment from the
QTM catalog for selection of events in our regional catalog in a
box with the same dip as the alignment. No nodal planes were
available as input constraints for this fault segment. The result-
ing fault surface (western segment of the ELLF, south section)
is about 14 km long, closely follows the rupture trace, and is
moderately northeast dipping (75°) in the upper 2.5 km. At
depth, the surface turns subvertical. Thus, both the western
and main segments of the ELLF are essentially parallel.
The second, modeled branch of the main ELLF is situated in
the central domain, about 3 km to the east of the main fault
strand (Fig.
6b
). The selected trace consists of two main seg-
ments, defined by 1800 mapped vertices. The gap in hypocen-
ters that separates the branch from the main fault follows
moderately dipping plane, and is therefore not apparent in
map view. This planar orientation also connects hypocenters
best to the discontinuous surface trace. Therefore, we used this
orientation to select hypocenters for input constraints for mod-
eling the fault surface. There are 3400 events associated with
this fault segment. The continuity of seismicity along the trend
does not indicate lateral terminations or segmentation of this
branch. We therefore terminate this branch with the main
strand of the ELLF to the north and by the northern strand
Figure 7.
Cross section A
A
(see Fig.
6a
for trace location and projection
box) shows planar alignment of hypocenters in quake template matching
catalog (red) for western splay of the ELLF, southern section (gray line).
Hypocenters in regional catalog are black, including both open and solid
circles. Solid black events were used constraints for modeling the fault.
Volume 110 Number 4 August 2020 www.bssaonline.org
Bulletin of the Seismological Society of America
1825
Downloaded from https://pubs.geoscienceworld.org/ssa/bssa/article-pdf/110/4/1818/5114416/bssa-2020053.1.pdf
by California Inst of Technology user
on 27 August 2020
of the SLLF (described later) to the south. For orientation con-
straints, we used nodal planes of nine focal mechanisms that lie
along the trend. The resulting fault surface is about 13 km long,
dips about 75° in the upper 6 km, and is nearly vertical at
depth. We note that the modeled branch maintains separation
from the main strand at depth.
In the northern domain, we model four faults that are oblique
to the northern section of the ELLF: the western, central, eastern,
and eastern boundary faults (Fig.
6c
). These are primarily
defined by seismicity lineaments that all share a similar trend
and steep dip to the northeast. Because of this dip, the epicenters
are rather distributed in map view but can be readily identified
and distinguished in an inclined view down the dip direction.
The selected constraints include about 1000 events for the
western, about 1000 events for the central, about 300 events
for the eastern, and about 400 events for the eastern boundary
fault. There are only a few surface rupture traces along these
faults. The eastern and the eastern boundary fault each has a
shorter trace of 1
2 km length resulting in less than 20 mapped
vertices used as location constraints. Orientation constraints are
provided by nodal planes of five, four, and two focal mecha-
nisms for the western, central, and eastern faults, respectively.
The northern limit of all faults in the northern domain is the
limit of aligned seismicity. To the south, all faults extend in a
generalized fashion to the main strand, which we used to trun-
cate them. The resulting fault representations have increasing
lengths from 7 km for the western fault, to 20 km for the eastern
boundary fault. They are subparallel, dipping about 80°
85° to
the northeast, throughout the seismogenic crust.
In the northern domain, we additionally represent three
faults that trend at a high angle to the main trend, for example,
southwest
northeast (Fig.
6c
). We use the term
cross faults
to describe structures with this trend. These cross faults are
primarily defined by seismicity lineaments with closely corre-
sponding trends and steep dips to the southeast. The northern
and north-central cross faults of the ELLF, northern section,
are constrained by 200 events and by 700 events, respectively,
clustering along steeply southeast-dipping planes, at the north-
ernmost limit of the rupture. The southern cross fault is con-
strained by 800 events aligned along a steeply southeast-
dipping plane, as well. This plane projects upward into a small,
2-kilometer-long surface rupture trace with about 50 mapped
vertices, which provide location constraints. No focal mecha-
nisms were available as orientation constraints. The lateral
extents of these cross faults are defined largely by the extent
of aligned seismicity. This results in an en echelon, right-step-
ping arrangement of the northern and north-central cross
faults, with the northern cross fault interpenetrating the main
strand of the ELLF to the west, and north-central fault being
truncated by the eastern splay fault of the northern section of
the ELLF to the east. The resulting, combined length of the
fault representations is about 6 km, with a dip of about 85°
to the southeast. For the southern cross fault, the workflow
results in an about 6-kilometer-long representation, dipping
about 80° to the southeast in the upper 8 km, and 85°
90°
at depth. The southern cross fault also interpenetrates the main
strand of the ELLF to the west, and crosses the eastern splay
fault of the northern section of the ELLF to the east.
In the central domain, we represent one cross-fault (Fig.
6b
)
that is well defined by hypocentral alignment. Hypocenters there
align in a band a few hundred meters in width along a moder-
ately northwest-dipping plane. This band contains about 700
events that define location constraints. No surface rupture trace
or focal mechanisms were available as further constraints. The
lateral extent of this cross fault is defined by the extent of the
seismicity. To the north, the band of seismicity associated with
this cross fault can be clearly followed across the main strand of
the ELLF, and then truncates against the eastern branch of the
ELLF, central section. Because no rupture trace was mapped at
the surface, the cross fault is considered blind. This fault repre-
sentation is vertically limited to the depth range of selected
hypocenters. Application of the workflow results in a
7-kilometer-long fault representation, which dips 77° to the
northwest, consistently throughout the entire depth range.
The upper tip line is at about 2.5 km below sea level.
The southern domain exhibits two additional cross faults
that we represent
a northern (Fig.
6b
) and southern cross
fault (Fig.
6c
) of the ELLF, southern section. All are primarily
defined by hypocentral lineaments. The southern hypocentral
alignment dips to the northwest, in the same direction as most
of the cross faults in other regions of the rupture appear to.
There is some steepening of the dip with depth. The northern
hypocentral alignment, however, dips steeply to the southeast.
Selecting events for each dipping alignment generates about
400 hypocentral location constraints for both the northern
and southern cross fault. No surface rupture trace and no focal
mechanism were available as further constraints. Therefore,
the upper limit of the faults is defined by seismicity. Laterally,
the northern cross fault is limited by the extent of associated
seismicity that crosses the main strand of the ELLF. The
southern cross fault does not clearly cross the main strand
at its eastern limit and has sporadic seismicity extending for
about 6 km to its western limit. The resulting representations
for the northern and southern cross faults are 6 and 12 km
long, respectively, and dip about 85° to the southeast and
75° above 4 km to 85° at depth to the northwest, respectively.
After modeling the ELLF, we collected input data for the
southern, main strand and the northern strand of the SLLF
(Fig.
6b
). This included the initial
M
6.4 mainshock, which
is located in the area where the fault interacts with the ELLF.
A preliminary analysis using distance measurements of hypo-
centers to the initial, rapid response representation of the SLLF
as a single surface revealed a bimodal distance distribution
(Fig.
8
). This observation is best explained by the existence
of a northern, parallel strand. The main strand is defined
by a 13-kilometer-long, very continuous surface trace that
1826
Bulletin of the Seismological Society of America
www.bssaonline.org Volume 110 Number 4 August 2020
Downloaded from https://pubs.geoscienceworld.org/ssa/bssa/article-pdf/110/4/1818/5114416/bssa-2020053.1.pdf
by California Inst of Technology user
on 27 August 2020