Interseismic Strain Accumulation on Faults Beneath Los Angeles, California - Rollins_et_al-2018-Journal_of_Geophysical_Research__Solid

Interseismic Strain Accumulation on Faults Beneath

Los Angeles, California

Chris Rollins

1,2

, Jean-Philippe Avouac

, Walter Landry

, Donald F. Argus

and Sylvain Barbot

Division of Geological and Planetary Sciences, California Institute of Technology, Pasadena, CA, USA,

Now at Geophysical

Institute, University of Alaska Fairbanks, Fairbanks, AK, USA,

Walter Burke Institute for Theoretical Physics, California

Institute of Technology, Pasadena, CA, USA,

NASA Jet Propulsion Laboratory, California Institute of Technology, Pasadena,

CA, USA,

Earth Observatory of Singapore, Nanyang Technological University, Singapore

Abstract

Geodetic data show that the Los Angeles metropolitan area is undergoing 8

–

9 mm/year of

north-south tectonic shortening associated with the Big Bend of the San Andreas Fault. This shortening

has been linked to multiple damaging twentieth century thrust earthquakes as well as possible

≥

7.0

Holocene thrust events beneath central Los Angeles. To better characterize this seismic hazard, we assess

how this shortening is being accommodated by interseismic strain accumulation on subsurface faults,

incorporating detailed seismology- and geology-based models of fault geometry and the low-stiffness Los

Angeles sedimentary basin. We

fi

nd that strain accumulation on local strike-slip faults likely contributes no

more than 1

–

2 mm/year of the shortening. We formally invert the geodetic data for the pattern of

interseismic strain accumulation on the north dipping Sierra Madre, Puente Hills, and Compton thrust faults

and a master decollement. We explore the impact of the assumed material model, strain accumulation on

faults to the west and east, and other model assumptions. We infer that the three faults slip at 3

–

4 mm/year

over the long term and are currently partially or fully locked and accruing interseismic strain on their upper

sections. This locking implies an annual de

fi

cit of seismic moment, 1.6 + 1.3/

0.5 × 10

Nm/year in total,

which is presumably balanced over the long-term average by the moment released in earthquakes. The

depth distribution of moment de

fi

cit accumulation rate matches that of seismicity rates in Los Angeles to

fi

rst

order, in part, because the models incorporate the blind nature of the Puente Hills and Compton Faults.

1. Introduction

In California, the Paci

fi

c plate moves northwest at ~50 mm/year relative to the North American plate (e.g.,

Kreemer et al., 2014). The San Andreas Fault generally strikes subparallel to the relative plate motion direction

and accommodates the majority of the relative motion through right-lateral slip (e.g., Argus & Gordon, 2001;

Lisowski et al., 1991). North of Los Angeles, however, the San Andreas makes a leftward bend and is misa-

ligned by ~20° with the relative plate motion direction for ~200 km (Figure 1c), resulting in north-south short-

ening. In Los Angeles, Global Positioning System (GPS) data (Figure 1a) show ~6 mm/year of north-south

shortening between the Palos Verdes peninsula and the San Gabriel Mountains or 8

–

9 mm/year if Santa

Catalina, San Clemente, and San Nicolas Islands are included (e.g., Argus et al., 1999, 2005; Davis et al.,

1989; Feigl et al., 1993; Shen et al., 1996; Walls et al., 1998; Figures 1a and 2b). Geologic, seismologic, and

strain data indicate that this shortening is the principal strain in the Los Angeles region (Davis et al., 1989;

Hauksson, 1990; Li, 1996; Yang & Hauksson, 2013; Zoback et al., 1987). This shortening has been linked to

the damaging 1971

~ 6.7 San Fernando, 1987

~ 5.9 Whittier Narrows, and 1994

= 6.7

Northridge thrust earthquakes (Figure 1a; e.g., Argus et al., 1999; Dolan et al., 1995). Paleoseismologic studies

suggest that it may also have produced

≥

7.0 Holocene earthquakes on three extensive north dipping

thrust faults, the Sierra Madre, Puente Hills, and Compton faults (Leon et al., 2007, 2009; Rubin et al., 1998;

Figure 3). In this study, we characterize the seismic hazard associated with this shortening by using a re

fi

ned

GPS velocity

fi

eld and detailed 3-D representations of fault geometry and subsurface elastic structure to build

models of interseismic strain accumulation beneath Los Angeles. Following the preparation of various com-

ponents of the modeling effort and some insights from forward models of dip-slip and strike-slip faulting, we

formally invert the GPS data to estimate the long-term slip rates and the pattern of interseismic strain accu-

mulation on the Sierra Madre, Puente Hills, and Compton faults and a master decollement. We then vet these

ROLLINS ET AL.

7126

Journal of Geophysical Research: Solid Earth

RESEARCH ARTICLE

10.1029/2017JB015387

Key Points:

•

We build GPS-based models of strain

accumulation on faults beneath Los

Angeles that incorporate the effect of

the sedimentary basin

•

We infer that three major thrust faults

slip at 3-4 mm/year over the long

term and are partially or fully locked

on their upper sections

•

This locking corresponds to a total

seismic moment de

fi

cit rate of 1.6 +

1.3/-0.5 times 10

Nm/year, which is

presumably released in earthquakes

Supporting Information:

•

Supporting Information S1

Correspondence to:

C. Rollins,

john.c.rollins@gmail.com

Citation:

Rollins, C., Avouac, J.-P., Landry, W.,

Argus, D. F., & Barbot, S. (2018).

Interseismic strain accumulation on

faults beneath Los Angeles, California.

Journal of Geophysical Research: Solid

Earth

123

, 7126

–

7150. https://doi.org/

10.1029/2017JB015387

Received 22 DEC 2017

Accepted 9 JUL 2018

Accepted article online 16 JUL 2018

Published online 30 AUG 2018

Figure 1.

(a) Tectonics and shortening in the Los Angeles region. Dark blue arrows are shortening-related GPS velocities relative to the San Gabriel Mountains

(Argus et al., 2005). Contours are uniaxial strain rate (rate of change of

) in the N ~5° E direction estimated from the GPS using the method of Tape et al.

(2009). Background shading is the shear modulus at 100-m depth in the CVM*, a heterogeneous elastic model based on the Community Velocity Model (Süss &

Shaw, 2003; Shaw et al., 2015) that we create and use in this study (section 4). Black lines are upper edges of faults, dashed for blind faults. Epicenter

s of the

1971, 1987, and 1994 earthquakes are from Southern California Earthquake Data Center; focal mechanisms are from Heaton (1982) for 1971 and Global CMT

Catalog for 1987 and 1994. Pro

fi

le A-A

follows LARSE line 1 (Fuis et al., 2001) onshore and line M-M

of Sorlien et al. (2013) offshore. SGF = San Gabriel Fault;

SSF = Santa Susana Fault. VF = Verdugo Fault. SAF = San Andreas Fault. CuF = Cucamonga Fault. A-DF = Anacapa-Dume Fault. SMoF = Santa Monica Fault.

HF = Hollywood Fault. RF = Raymond Fault. UEPF = Upper Elysian Park Fault. ChF = Chino Fault. WF = Whittier Fault. N-IF = Newport-Inglewood Fault.

PVF = Palos Verdes Fault. (b) GPS velocities on islands. (c) Tectonic setting. Black lines and pairs of half-arrows, respectively, are major faults an

d their slip senses.

Black arrow is Paci

fi

c Plate velocity relative to North American plate from Kreemer et al. (2014). GF = Garlock Fault. SJF = San Jacinto Fault. EF = Elsinore Fault.

SB = Santa Barbara. LA = Los Angeles. SD = San Diego.

10.1029/2017JB015387

Journal of Geophysical Research: Solid Earth

ROLLINS ET AL.

7127

models of strain accumulation by comparing them with the depth distribution of seismicity in the

Los Angeles region, published geologic and paleoseismologic slip rates, and other constraints from

the literature.

2. Tectonics, Geology, and Geodetic Constraints

While the Sierra Madre Fault breaks the surface along the southern edge of the San Gabriel Mountains, the

Puente Hills and Compton are blind thrust faults, with top edges respectively at ~3- and ~5-km depths

directly beneath the Los Angeles metropolitan area (Figures 1a and 2a). Seismic re

fl

ection data,

Figure 2.

(a) Cross sections of faults, structure, north-south contraction, and seismicity along pro

fi

le A-A

. Red lines are fault surfaces as meshed here (Figure 3),

dashed where uncertain (Shaw & Suppe, 1996; Shaw & Shearer, 1999; Fuis et al., 2012). Geometries of basin, basement, and mantle are from Shaw et al. (201

5);

geometry of base of Fernando Formation (boundary between beige and tan units of the basin) is interpolated from Sorlien et al. (2013; offshore), Wrigh

t (1991;

coastline to Whittier Fault), and Yeats (2004; Whittier Fault to Sierra Madre Fault); topography is from Fuis et al. (2012). (b) Projections of Argus e

t al. (2005) GPS

velocities (relative to San Gabriel Mountains) onto the direction N 5° E and 1

uncertainties. Note that stations on Palos Verdes are plotted left of the coastline as

the offshore section of pro

fi

le A-A

passes alongside Palos Verdes (Figure 1a). (c) Seismotectonic features. Distribution of shear modulus is from the CVM*, the

heterogeneous elastic model used in this study (section 4). Translucent white circles are relocated 1981

–

2016

≥

2 earthquakes whose epicenters lie within the

mesh area of the three thrust faults and decollement (Hauksson et al., 2012 and updated). PVF = Palos Verdes Fault; N-IF = Newport-Inglewood Fault; WF =

Whittier

Fault.

10.1029/2017JB015387

Journal of Geophysical Research: Solid Earth

ROLLINS ET AL.

7128

earthquake hypocenters, and geologic constraints suggest that the Sierra Madre and Puente Hills faults root

into a decollement beneath the San Gabriel Mountains (Fuis et al., 2001; Hadley & Kanamori, 1978; Meigs

et al., 2003; Myers et al., 2003; Ryberg & Fuis, 1998). This decollement may continue southward (Davis

et al., 1989; Humphreys & Hager, 1990; Wright, 1991) and connect to the north dipping Lower Elysian Park

ramp (Cooke & Marshall, 2006; Marshall et al., 2009; Sorlien et al., 2013). The latter likely extends upward to

10- to 11-km depth beneath central Los Angeles and may connect to the Compton Fault in a ramp-

fl

at-ramp

geometry (Shaw & Suppe, 1996), which would make the three faults a connected fold-and-thrust belt

(e.g., Davis et al., 1989; Figure 2a).

Although these three north dipping thrust faults are well oriented to accommodate north-south shortening

and evidently have produced

≥

7.0 Holocene earthquakes, geologic and paleoseismologic studies pro-

vide mixed evidence about how much of the total shortening they accommodate. On the one hand, the

San Gabriel Mountains are likely underlain by a crustal root whose dimensions imply that north dipping

thrust faults collectively accommodate ~7 mm/year of shortening and thickening (Godfrey et al., 2002). A

1 mm/year exhumation rate inferred in the San Gabriel Mountains (Blythe et al., 2000) also suggests a total

shortening rate of ~5.5 mm/year under considerations of isostasy (Donnellan et al., 2001; Meigs et al.,

2003). Meigs et al. (2003) also inferred a geologic offset rate of 3.6

–

5.7 mm/year on the Sierra Madre Fault.

However, most other geologic and paleoseismologic estimates of slip rates on the three faults, including

other estimates on the Sierra Madre, are around ~1 mm/year (Bergen et al., 2017; Shaw et al., 2002 ; Shaw

& Suppe, 1996; Tucker & Dolan, 2001; Walls et al., 1998) or ~2 mm/year at the maximum (Leon et al., 2007,

2009). Thus, although Argus et al. (2005)

fi

t the geodetic shortening to a model of a north dipping edge dis-

location accumulating strain at 9 ± 2 mm/year beneath northern Los Angeles and Meade and Hager (2005a)

fi

t the shortening to interseismic strain accumulation on two north dipping faults slipping at ~4 mm/year

each, these geologic and paleoseismologic slip rates suggest that deformation is more distributed. Walls

et al. (1998) postulated that more than half of the shortening could be accommodated by escape tectonics

on local strike-slip faults such as the right-lateral Palos Verdes, Newport-Inglewood, and Whittier faults and

the left-lateral Raymond-Hollywood-Santa Monica fault system (Figure 1a); Argus et al. (1999), however,

found that this model overpredicts relative east-west surface velocities in Los Angeles. Nevertheless,

Marshall et al. (2009) and Daout et al. (2016)

fi

t the geodetically inferred shortening to models of strain accu-

mulation on thrust and strike-slip faults with slip rates closer to the geologic rates. Although the former

model predicts a somewhat more gradual contractional gradient than that seen in GPS velocities, and the lat-

ter model

fi

ts only the northern portion of the shortening projected into the San Andreas-perpendicular

direction (N ~25° E) rather than the north-south direction, these studies suggest that the geodetic shortening

and the geologic slip rates may be reconcilable within more complex models of strain accumulation.

These studies, however, reveal another discrepancy: in the Argus et al. (2005), Marshall et al. (2009), and

Daout et al. (2016) models, the best

fi

tting locking depths on the faults accruing strain

—

above which they

are inferred to be locked and accumulating interseismic strain as part of the cycle of stick-slip behavior

and below which they are inferred to be freely creeping

—

are only 6 ± 2, 8, and 3 km, respectively. If true,

the low-inferred strain accumulation below these depths should presumably be re

fl

ected in low seismicity

rates; however, most seismicity in Los Angeles in fact occurs below these depths (Figure 2c). This discrepancy

may somewhat result from the fact that these studies model the Earth as a homogeneous elastic half-space.

In reality, Los Angeles sits atop a deep sedimentary basin ringed by batholithic crustal rocks (e.g., Shaw et al.,

2015), a heterogeneity that may signi

fi

cantly affect the relationship between subsurface strain accumulation

and surface deformation. In particular, the Puente Hills and Compton faults mostly underlie the Los Angeles

basin (Figure 2a), and previous studies have shown that if a fault lies below a low-stiffness near-surface layer,

an analysis using a homogeneous elastic model will infer slip on the fault as being shallower than it actually is

(Arnadottir & Segall, 1991; Bernard et al., 1997; Cattin et al., 1999; Hager et al., 1999). The models we develop

in this study incorporate this consideration.

3. The Geodetic Shortening

Geodetic strain in the Los Angeles area can result from (1) local tectonic strain accumulation, (2) deformation

due to management of aquifers and oil

fi

elds, and (3) regional scale strain accumulation on the San Andreas

system. Although we are only interested in the

fi

rst term, the second and third must be characterized and

10.1029/2017JB015387

Journal of Geophysical Research: Solid Earth

ROLLINS ET AL.

7129

removed in order for a GPS data set to be of use in estimating the

fi

rst. Argus et al. (2005) prepared a

fi

eld of

horizontal GPS velocities in Los Angeles that accounts for both anthropogenic motion (as estimated from

Interferometric Synthetic Aperture Radar data) and strain accumulation on the San Andreas (as estimated

using a best

fi

t screw dislocation model). The residual velocities after subtracting both terms are expressed

in the reference frame of the San Gabriel Mountains de

fi

ned using a block model of long-term motion.

This velocity

fi

eld (Figure 1, dark blue arrows) is thus appropriate for estimating the pattern of local strain

accumulation, and we use it here. Although strain accumulation on the San Andreas system is the

dominant geodetic signal in southern California (e.g., Meade & Hager, 2005a), dislocation modeling shows

that the inferred shortening in Los Angeles is mostly independent of the model assumed for the

San Andreas (Figure S1 in the supporting information) and is therefore likely robust.

The methods we use to estimate interseismic strain accumulation on faults beneath Los Angeles (section 5)

require the simultaneous estimation of present-day strain accumulation on faults and of the long-term

motions across them. To reduce the number of free parameters in this approach, we assume that the

Figure 3.

Meshed geometries of the three main thrust faults beneath the Los Angeles basin (section 4), colored by depth, and 1981

–

2016

≥

2.5 earthquakes within

the mesh area from Hauksson et al. (2012 and updated), scaled by magnitude (white-

fi

lled circles). Gray-

fi

lled circles are 1981

–

2016

≥

4.5 earthquakes outside the

mesh area. Inferred paleoearthquakes are from Rubin et al. (1998) and Leon et al. (2007, 2009). SAF = San Andreas Fault.

10.1029/2017JB015387

Journal of Geophysical Research: Solid Earth

ROLLINS ET AL.

7130