New Geodetic Constraints on Southern San Andreas Fault-Slip Rates, San Gorgonio Pass, California GEOSPHERE, V

Research Paper THEMED ISSUE: Seismotectonics of the San Andreas Fault System in the San Gorgonio Pass Region

GEOSPHERE New geodetic constraints on southern San Andreas fault-slip rates, San Gorgonio Pass, California GEOSPHERE, v. 17, no. 1 Katherine A. Guns1,†, Richard A. Bennett1, Joshua C. Spinler2, and Sally F. McGill3 1Department of Geosciences, University of Arizona, 1040 E. 4th Street, Tucson, Arizona 85721, USA https://doi.org/10.1130/GES02239.1 2Department of Earth Sciences, University of Arkansas at Little Rock, 2801 S. University, Little Rock, Arkansas 72204, USA 3Department of Geological Sciences, California State University San Bernardino, 5500 University Parkway, San Bernardino, California 92407, USA 13 figures; 7 tables; 1 set of supplemental files ABSTRACT Along the SSAF system, there are more than a dozen fault segments com- CORRESPONDENCE: [email protected] pactly arranged within a handful of tectonic provinces along the main San Assessing fault-slip rates in diffuse plate boundary systems such as the Andreas fault (e.g., the continental borderlands, the Eastern Transverse Ranges, CITATION: Guns, K.A., Bennett, R.A., Spinler, J.C., and McGill, S.F., 2021, New geodetic constraints on San Andreas fault in southern California is critical both to characterize seis- the Eastern California shear zone, the Los Angeles Basin, and the Western Trans- southern San Andreas fault-slip rates, San Gorgonio mic hazards and to understand how different fault strands work together to verse Ranges), which together accommodate ~52 mm/yr of plate boundary Pass, California: Geosphere, v. 17, no. 1, p. 39–68, accommodate plate boundary motion. In places such as San Gorgonio Pass, motion (Argus et al., 2010). Yet, these tectonic provinces each have their own https://doi.org/10.1130/GES02239.1. the geometric complexity of numerous fault strands interacting in a small area deformation styles, with varying fault types (right lateral, left lateral, reverse, or adds an extra obstacle to understanding the rupture potential and behavior oblique combinations), map view texture (multiple faults parallel to the plate Science Editor: Andrea Hampel Guest Associate Editor: Michele Cooke of each individual fault. To better understand partitioning of fault-slip rates boundary versus multiple faults striking east-west; straight and clearly defined in this region, we build a new set of elastic fault-block models that test 16 versus sinuous, disconnected traces), geologic level of maturity (e.g., recently Received 31 January 2020 different model fault geometries for the area. These models build on previ- formed in past <2 m.y. versus been accommodating slip since 10 Ma), and level Revision received 26 August 2020 ous studies by incorporating updated campaign GPS measurements from of active seismicity (Fig. 1). Determining which faults play the most important Accepted 20 October 2020 the San Bernardino Mountains and Eastern Transverse Ranges into a newly roles in actively accommodating plate boundary motion requires every tool in calculated GPS velocity field that has been removed of long- and short-term the active tectonics arsenal, including the application of historical geologic off- Published online 10 December 2020 postseismic displacements from 12 past large-magnitude earthquakes to set reconstructions (million-year time scale), tectonic-geomorphologic– based estimate model fault-slip rates. Using this postseismic-reduced GPS velocity geologic slip-rate studies and paleoseismic trenching (hundred-thousand-year to field produces a best-fitting model geometry that resolves the long-standing Holocene time scales), space-based geodetic techniques that allow decadal time- geologic-geodetic slip-rate discrepancy in the Eastern California shear zone scale measurement of the present-day motion of the crust, the pattern of crustal when off-fault deformation is taken into account, yielding a summed slip seismicity, and numerical modeling based on crustal deformation dynamics. rate of 7.2 ± 2.8 mm/yr. Our models indicate that two active strands of the One of the most complicated areas of fault interaction in southern California San Andreas system in San Gorgonio Pass are needed to produce sufficiently lies in and around San Gorgonio Pass, at the northern end of the Coachella low geodetic dextral slip rates to match geologic observations. Lastly, results Valley (Fig. 2). As the SSAF stretches north from the Salton Sea toward San Gor- suggest that postseismic deformation may have more of a role to play in gonio Pass, it splits from one main fault strand into three subparallel strands, affecting the loading of faults in southern California than previously thought. including, from southwest to northeast, the Garnet Hill, Banning-Coachella Valley, and Mission–Mill Creek faults. The 14–30 mm/yr of geologically and geodetically measured slip along the Coachella Valley segment of the SSAF ■■ INTRODUCTION near or to the south of this juncture (Becker et al., 2005; Meade and Hager, 2005; van der Woerd et al., 2006; Behr et al., 2010; Spinler et al., 2010; Blisniuk Complex plate boundary zones such as that of the southern San Andreas et al., 2013b; Lindsey and Fialko, 2013; Spinler et al., 2015) must somehow be fault (SSAF) system in southern California present an opportunity to investigate partitioned onto these three strands to the north, onto a possible emerging fault how recoverable elastic strain and nonrecoverable fault slip are distributed zone along the “Landers Mojave Earthquake Line” of Nur et al. (1993), or trans- within networks of interlocking and branching fault segments of different and ferred to other faults of the plate boundary zone through other mechanisms sometimes seemingly incompatible azimuths, lengths, ages, and maturities. such as block rotation (e.g., Carter et al., 1987; Powell, 1993). Understanding where this slip is being accommodated in the present day is vital to testing seismic hazard scenarios for this region, particularly in light of the evidence Katherine A. Guns https://orcid.org/0000-0002-2956-1536 This paper is published under the terms of the †Now at Institute of Geophysical and Planetary Physics, Scripps Institution of Oceanography, that the SSAF in the Coachella Valley is overdue for a major earthquake (Fumal CC‑BY-NC license. University of California, San Diego, 9500 Gilman Drive, La Jolla, California 92093, USA et al., 2002; Fialko, 2006; Field et al., 2015). However, the diffuse nature of the

GEOSPHERE | Volume 17 | Number 1 Guns et al. | New GPS-based block modeling accounting for postseismic deformation near San Gorgonio Pass Downloaded from http://pubs.geoscienceworld.org/gsa/geosphere/article-pdf/17/1/39/5219650/39.pdf 39 by guest on 27 September 2021 Research Paper

36˚N 2019, M7.1 Ridgecrest N Fault rlock Ga

Eastern California

100 km Shear Zone

S Block Modeling Boundary AF Mo Figure 1. Regional fault geometry and historical (≥Mw6.5) jav e S 1999, M7.1 Hector Mine earthquakes in southern California; earthquake surface 1971, M6.6 ec ruptures presented in red and focal mechanism solu- ti on tions plotted from catalog of Wang et al. (2009) and the U.S. Geological Survey Earthquake Catalog. Large- 1992, M7.3 Landers 1994, M6.7 magnitude earthquakes of the past 50 years are labeled Northridge with their location names. Earthquakes that make up the 34˚N Landers-Mojave Earthquake Line of Nur et al. (1993) are San shown with their years and names in black text (1979 S A Homestead; 1975 Galway; 1965 Calico; 1947 Manix; 2008 Gorgonio Palm F C Ludlow). The dashed white box presents the extent of Knot o Springs a our block model boundary in the context of the larger c h e region, and we highlight the location of the San Gorgo- Salton ll Continuous GPS Station a nio Pass knot in a solid white box. Campaign GPS Station Sea (CSU San Bernardino) 1987, M6.6 Campaign GPS Station (UA JOIGN network) 1979, M6.5 Faults (from USGS Quaternary Database) Rupture (from USGS 2010, M7.2 El Mayor Cucapah Historic Faults Database) 0 km 100 km −118˚W −116˚W

fault system here hampers our understanding of the role each part plays in there is the possibility of a more complex rupture involving several strands the earthquake cycle. simultaneously, even strands that are unmapped as yet (Yule and Sieh, 2003). An added wrinkle to the challenge that this complex fault geometry poses As the San Jacinto fault has evolved in close proximity to the San Andreas to understanding the distribution of slip rate is the question of how an indi- fault over the past ~1.5 m.y. (e.g., Morton and Matti, 1993; Kendrick et al., 2015; vidual earthquake rupture may or may not propagate along these closely Fattaruso et al., 2016), it has produced and reactivated many smaller connecting spaced faults in and around San Gorgonio Pass. The individual fault strands faults between itself and the San Andreas fault, particularly in the San Ber- together comprise a left-stepping restraining bend in the SSAF zone, which nardino Basin and eastern San Gabriel Mountains (Morton and Matti, 1993). represents a regional-scale structural knot in known fault geometry (e.g., Sykes In such intricate fault geometry, these smaller, less obvious (and potentially and Seeber, 1985; Yule and Sieh, 2003). Geometrical complexities such as this unmapped) faults could play a key role in a complex rupture, adding to the can serve as rupture barriers (e.g., Wesnousky, 2008; Lozos et al., 2015), but challenge of modeling seismic behavior here (Ross et al., 2017). It is critical to evidence from previous large-magnitude earthquakes along the San Andreas understand the overall fault geometry, slip histories, and interconnectedness of fault shows through-going ruptures propagating through large stepovers, such these related faults because these parameters have implications for the seismic as the 1857 Fort Tejon event rupturing through the “Tejon Knot” (Sykes and energy release and intensity of shaking during an earthquake rupture. These Seeber, 1985). While there could be a through-going rupture along one of the faults lie under and near some of the most populated counties in California, strands, as suggested by work by Douilly et al. (2020) and Castillo et al. (2019), including Riverside, San Bernardino, and Los Angeles counties, and better

characterization of their fault properties and rupture potential can help inform Helendale—South LockhartCamp Rock—Emerson Faults (from USGS A Lenwood—Lockhart efforts to mitigate the crippling effects that the next large-magnitude event Ludlow—CleghornQuaternary Lake database) Pisgah—Bullion is expected to have (Olsen et al., 2006; Jones et al., 2008; Field et al., 2015). Calico—Hidalgo (dashed where uncertain) One effective tool for evaluating how slip and elastic strain are partitioned

across a complex fault system is elastic fault-block modeling constrained by SAF — Mojave rontal F

h Johnson geodetic surface velocities. Many authors have employed elastic fault-block mod- t Valley r eling to better understand southern California crustal deformation (Bennett et N o

al., 1996; Becker et al., 2005; McCaffrey, 2005; Meade and Hager, 2005; Spinler mong aSAF—San Bernardino ca u EurekaPinto Pk. Mtn. et al., 2010; Johnson, 2013; Liu et al., 2015; McGill et al., 2015; Evans et al., 2016; C Mill Creek Mission Crk Hearn, 2019), and each iteration of block modeling of this area has included more Sheep Hole 34˚N Banning precise and longer durations of geodetic data, as well as improved constraints San Jacinto Blue Cut SGPF Garnet Hill on geologic slip rates from mapping and age dating of Holocene and Late Pleis- Palm SAF—Coachella Valley tocene offsets. However, past modeling efforts for this region differ in the way Springs they chose to estimate long-term deformation from global positioning system Elsinore (GPS) coordinate time series observations. Transient deformation in the forms of N San Jacinto surface loading (e.g., Blewitt and Lavallée, 2002; Dong et al., 2002; Elósegui et al., 50 km 2003; Bennett, 2008; Bos et al., 2010) and earthquake co-seismic and postseismic displacements (e.g., Pollitz et al., 2000; Wei et al., 2011; Liu et al., 2015; Guns and Holocene Slip Rate B 7 6 Bennett, 2020) can affect final estimated crustal velocities for years to decades 6 Pleistocene Slip Rate 6 6 Continuous GPS depending on the process. Viscoelastic postseismic displacements in particu- 6 Campaign GPS site lar have been demonstrated to last decades to centuries after large-magnitude (JOIGN) Campaign GPS site earthquakes (Vergnolle et al., 2003; Freed et al., 2007; Hearn et al., 2013; Guns 17 (CSUSB) and Bennett, 2020), thereby affecting measurements of surface displacements 18 on a long-term scale. Some authors have mitigated the effects of postseismic 1214 11 14 displacements through data selection (selecting out the data most affected by 13 9 15 23 10 these signals, or estimating only the most recent postseismic transients [Shen et 4 34˚N al., 2011]), simulating viscoelastic motions within the block model (e.g., Johnson, 27 22 21 8 2013), or assuming postseismic motions are negligible (e.g., Bennett et al., 1996; 1,5 Becker et al., 2005; Meade and Hager, 2005; Spinler et al., 2010). These transient Palm 2,3 30 Springs motions must be properly reduced or removed from the geodetic time series in 25 order to reveal the component of the velocity field driven by relative plate motion, N 31 26 and consequently to estimate long-term fault-slip rates. 28 While geodetic data often form the basis for crustal deformation models in 50 km 50 km southern California, these models, which associate crustal strain-rate fields to −116˚W fault-slip rates, are highly dependent on the assumed fault geometry (Bürgmann and Thatcher, 2013). Often, multiple model geometries can fit the GPS velocities Figure 2. Data coverage and fault geometry within our chosen model boundary (dashed black line); campaign GPS stations from the California State University, San Bernardino equally well to within their uncertainties (d’Alessio et al., 2005; Spinler et al., (CSUSB) (yellow diamonds), and the Joshua Tree Integrative Geodetic Network (JOIGN) 2010; Bürgmann and Thatcher, 2013); therefore, we require a combination of (orange diamonds) allow for needed increased coverage in remote areas of the San Ber- field-based geologic and space-based geodetic data sets to address this chal- nardino Mountains and Joshua Tree National Park; Holocene and Pleistocene geologic lenge. Field-based slip-rate estimates aid in the interpretations of active fault slip rates (stars) are labeled by citation number (refer to Table 1). geometries, allowing us to weed out fault geometries that are less likely due to the lack of visible Quaternary surface offsets. In addition, geologic slip rates and measurements of overall fault displacement can serve as a priori constraints between geologic observations and estimated geodetic fault-slip rates have been on geodetically estimated slip rates (e.g., Johnson, 2013; Zeng and Shen, 2016), reported for the southern California region, including rate mismatches along the even though it is a precarious comparison to relate these data sets at time scales Garlock fault (e.g., McGill and Sieh, 1993; Hearn et al., 2013), the summed fault- that differ by orders of magnitude, e.g., thousands of years (of order 103 to 105 slip rates across the Eastern California shear zone (e.g., Oskin et al., 2008; Evans years) versus decades (101 years). Indeed, several long-standing discrepancies et al., 2016), and along the Mojave and San Bernardino strands of the SSAF (e.g.,

Meade and Hager, 2005; Chuang and Johnson, 2011). Previous researchers have bend in the fault trace began to develop, possibly related to the interference of attempted to resolve these discrepancies with modeling of the earthquake cycle, the left-lateral Pinto Mountain fault at the north end of the present-day Coach- which assumes repeating characteristic earthquakes in a layered viscoelastic ella Valley (Fig. 2) (e.g., Matti and Morton, 1993; Fattaruso et al., 2016). This bend, half space (Chuang and Johnson, 2011; Johnson, 2013), or by removing modeled located in the geographic region known as San Gorgonio Pass, decreased the

cycle-averaged viscoelastic perturbations of representative ≥Mw7.0 earthquakes mechanical efficiency of fault motion (Fattaruso et al., 2016) and introduced (Hearn et al., 2013). However, if long-term viscoelastic deformation could be the lasting effects to the fault geometries surrounding it. The San Jacinto fault root of these discrepancies, an alternative way to account for the majority of just to the west likely formed at this time, due to the restraining nature of the postseismic displacements is to quantify the displacement field earthquake by San Gorgonio Pass bend in the San Andreas fault zone, because it may have earthquake through time (Guns and Bennett, 2020). been easier to develop new faults than to force slip along the preexisting Here we present an updated elastic fault-block modeling approach wherein but unfavorably oriented San Andreas fault strands (Matti and Morton, 1993; we construct 16 different fault models to test and reevaluate which fault Morton and Matti, 1993; Bennett et al., 2004; Janecke et al., 2010; Blisniuk et geometries best fit the modern GPS velocity field, which best fit the geologic al., 2013a; Fattaruso et al., 2016). observations, and whether there is any overlap between the two sets of best- Today, the SSAF works together with the San Jacinto fault to accommo- fitting models. Our study improves on past work in this region in two main ways: date the majority of motion across the plate boundary at the latitude of Palm (1) we use a newly calculated, steady-state GPS velocity field that has been Springs and the Coachella Valley. Authors debate which of the two faults adjusted to remove viscoelastic postseismic transients associated with histor- accommodates the majority of slip, and there is evidence to support a variety ical and recent large-magnitude earthquakes (see Guns and Bennett, 2020, for of geologic and geodetic slip rates on the San Jacinto fault. Geologic slip- methodological details) to better approximate long-term slip rates; and (2) we rate estimates along different sections of the San Jacinto fault demonstrate incorporate comparisons to past and new geologic slip-rate observations using a range of rates from 8 to >23 mm/yr from measurements of offset surface a root mean square (RMS) score, but not as constraints in our model, to judge features and from paleoseismic trenching estimates (Rockwell et al., 1990; agreement of our 16 different model fault geometries relative to overall geologic Kendrick et al., 2002; Rockwell et al., 2006; Blisniuk et al., 2010; Janecke et slip-rate observations (Holocene and Pleistocene), Holocene-only geologic obser- al., 2010; McGill et al., 2012; Onderdonk et al., 2015), while different geodetic vations, and Pleistocene-only geologic observations. There have been noted databased estimates of fault slip indicate overlapping rates of 7–26 mm/yr “persistent” slip-rate discrepancies in southern California, namely within the (Bennett et al., 1996; Becker et al., 2005; McCaffrey, 2005; Meade and Hager, Mojave area of the Eastern California shear zone (ECSZ) (Liu et al., 2015; Evans 2005; Spinler et al., 2010; Lindsey and Fialko, 2013; Lindsey et al., 2014; Liu et al., 2016), and along the Mojave and San Bernardino sections of the SSAF et al., 2015; McGill et al., 2015; Spinler et al., 2015; Evans et al., 2016). The (Meade and Hager, 2005; McGill et al., 2015). Our new modeling sheds light on majority of recent estimates based on GPS data alone show that those along these particular discrepancies and suggests that with the reduction of the GPS the SSAF are higher than those along the San Jacinto fault, implying that it data by removal of ongoing long- and short-term postseismic transients in con- is still carrying most of the relative plate motion (Bennett et al., 1996; Becker junction with accounting for off-fault deformation, some of these discrepancies et al., 2005; McCaffrey, 2005; Meade and Hager, 2005; Spinler et al., 2010; Liu can be resolved, while others remain, indicating another possible mechanism et al., 2015; Evans et al., 2016), while some studies that include both GPS and may be at work such as time-variability of slip rate, larger-scale mode-switching InSAR data conclude that the slip rates are very nearly equal (Lindsey et al., behavior (e.g., Dolan et al., 2007), or even the possibility that lingering postseis- 2014) or that the San Jacinto slip rate is greater (Lundgren et al., 2009). How- mic deformation may function as a mechanism of short- or long-term elastic ever, despite it potentially being the main carrier of plate boundary motion in loading and/or unloading along faults. In addition, parameter estimate trade-offs this area, the behavior of the SSAF as it enters San Gorgonio Pass along the between different strands in the model geometry of San Gorgonio Pass itself Mission–Mill Creek (MMC), Banning, and Garnet Hill fault strands remains indicate a possible dependence of the San Gorgonio Pass thrust fault strand on enigmatic because of the apparent reduction in slip rate and fault activity the activity level of the controversial and debated Mission–Mill Creek fault strand. along the zone as it approaches the Pass (Fattaruso et al., 2014). We explore the results of all models and evaluate their individual capacity to fit Geologic slip-rate estimates for the MMC fault just south of the Indio Hills the GPS velocity field and known geologic observations in order to constrain a (Fig. 2; Table 1) reveal between 20 and 24 mm/yr (Blisniuk et al., 2013b; K. most likely scenario for motion in the San Gorgonio Pass region. Blisniuk, 2019, personal commun.) and 14–17 mm/yr (van der Woerd et al., 2006; Behr et al., 2010) of slip in the Holocene and Late Pleistocene at Pushwalla Canyon and Biskra Palms, respectively. Current slip-rate and hazard models ■■ GEOLOGIC SETTING OF SAN GORGONIO PASS indicate a progressive decrease in slip rate northward along the MMC fault segment (Field et al., 2015), following the evidence of previous studies show- Since its inception, the main trace of the SSAF has accommodated ~200 km ing little to no Quaternary offsets along the northern extension of the Mission of displacement (Powell, 1993). Circa 1.5 Ma, a major left-stepping restraining Creek fault where it meets the Mill Creek fault (Morton and Matti, 1993; Yule and

TABLE 1. GEOLOGIC SLIP RATES AND UNCERTAINTIES ALONG FAULTS NEAR SAN GORGONIO PASS Fault Study site Holocene slip rate Pleistocene slip rate Reference Citation (mm/yr) ±2σ (mm/yr) ±2σ number Coachella Valley Mission Creek fault Pushawalla Canyon 20 ± 2 20 ± 2 Blisniuk et al. (2013b) 1 Mission Creek fault Biskra Palms - 15.9 ± 3.4 van der Woerd et al. (2006) 2 Mission Creek fault Biskra Palms - 14–17 Behr et al. (2010) 3 Mission Creek fault Mission Creek Preserve - 20–30 Fosdick and Blisniuk (2018)* 4 Mission Creek fault Thousand Palms Oasis 4 ± 2 Fumal et al. (2002)† 5 Eastern California shear zone Calico-Hildalgo fault - 1.8 ± 0.3 Oskin et al. (2007) 6 Camp Rock–Emerson fault - ≤1.4 ± 0.6 Oskin et al. (2008) 6 Helendale–South Lockhart fault - ≤0.8 ± 0.3 Oskin et al. (2008) 6 Ludlow–Cleghorn Lake fault - ≤0.4 ± 0.2 Oskin et al. (2008) 6 Pisgah-Bullion fault - 1.0 ± 0.2 Oskin et al. (2008) 6 Calico-Hidalgo fault - 3.2 ± 0.4 Xie et al. (2018) 7 Eastern Transverse Ranges Blue Cut fault Hexie Mountains - 1.66 ± 0.44 Guns (2020) 8 Pinto Mountain fault Oasis of Mara 1.59–1.80 - Cadena (2013) 9 Pinto Mountain fault West Yucca Valley - 3.0 +0.6/−0.4 Gabriel (2017) 10 San Gabriel Mountains Sierra Madre–Cucamonga Day Canyon - 1.9 ± 0.4 (dip-slip) Lindvall and Rubin (2006) 11 San Bernardino Mountains and Cajon Pass San Andreas fault (SAF), San Bernardino section Cajon Pass 25 +6/−4 23 +5/−2 Weldon and Sieh (1985) 12 SAF, San Bernardino section Plunge Creek 6.3–18.5 7.0–15.7 McGill et al. (2013) 13 SAF, San Bernardino section Badger Canyon - 12.8 +5.3/−4.7 S. McGill, 2020, personal commun. 14 SAF, San Bernardino section Pitman Canyon - 14.5 +9.9/−6.2 S. McGill, 2020, personal commun. 14 SAF, San Bernardino section Wilson Creek 14.0–25.0 12.0–34.0 Harden and Matti (1989) 15 SAF, San Bernardino section Burro Flats 4.5 ± 2.5 Orozco (2004) 16 Mojave Section SAF, Mojave section Pallett Creek 35.6 ± 6.7 - Salyards et al. (1992)§ 17 SAF, Mojave section Wrightwood 20–40 - Weldon et al. (2002) 18 SAF, Mojave section Ritter Ranch/Key Slide 21.2 +4.2/−9.3 - Young et al. (2019)* 19 SAF, Mojave section Little Rock Creek 30 ± 10 Matmon et al. (2005)* 20 San Gorgonio Pass Banning fault Whitewater 2.3–5.9 - Gold et al. (2015) 21 San Gorgonio Pass fault Millard Canyon, Cabezon 5.7 +2.7/−1.5 (oblique) - Heermance and Yule (2017) 22 San Gorgonio Pass fault Millard Canyon, Cabezon ~0.64 to ~3.51 (strike-slip) Heermance and Yule (2017)# 22 Mill Creek fault 0 - Kendrick et al. (2015)* 23 Mission–Mill Creek fault 0 - Matti et al. (2019)* 24 Mission Creek fault Mission Creek Preserve - 20–30 Fosdick and Blisniuk (2018)* [4] Peninsula Ranges San Jacinto fault Anza 9.2 ± 2 12 +9/−5 Rockwell et al. (1990) 25 San Jacinto fault Thomas Mountain 12.4 +2.5/−2.0 13.4 +3.8/−2.5 Blisniuk et al. (2013a) 26 San Jacinto fault Quincy 12.8–18.3 - Onderdonk et al. (2015) 27 San Jacinto fault, Clark fault Rockhouse Canyon - 8.9 ± 2 Blisniuk et al. (2010) 28 San Jacinto fault, all strands Borrego Springs - ≥14.0 +7.1/−4.5 Janecke et al. (2010)* 29 Elsinore fault, Glen Ivy section Temescal Valley - 5.3–5.9 Millman and Rockwell (1986) 30 Elsinore fault, Widomar section Murietta 4.9 +1.0/−0.6 - Rockwell et al. (2000a) 31 Note: Citation numbers correspond to locations plotted in Figure 2B. *Not included in analysis due to location outside boundary or lack of point location, but shown for comparison. †Only potentially measuring one strand. §Not located within model boundary, but included in root mean square analysis. #This rate is a strike-slip rate calculated from the oblique “net slip” rate data reported by Heermance and Yule (2017) for surface Qt-3 in their supplemental materials.

Sieh, 2003; Kendrick et al., 2015; Matti et al., 2019; Yule et al., 2019). Therefore, areas prohibit permanent installations of recording instruments. Two of these the Banning, Garnet Hill, and San Gorgonio Pass faults are hypothesized to be areas are the San Bernardino Mountains adjacent to San Gorgonio Pass and the the main conduits of slip entering San Gorgonio Pass along the SSAF system. remote areas of Joshua Tree National Park in the Eastern Transverse Ranges to However, a recent study by Fosdick and Blisniuk (2018) reports compelling the southeast. In order to increase the spatial density of data coverage in these sedimentological evidence for the continued activity of the MMC strand into areas to better delineate how permanent and elastic strain are accommodated Quaternary time, suggesting a much higher than expected slip rate along this along the diffuse fault network, we incorporate campaign GPS measurements section, up to 20–30 mm/yr in the Pleistocene. Ongoing mapping analyses of from a total of 87 sites within the Joshua Tree Integrative Geodetic Network possible Holocene displacements suggest the possibility of a newer MMC (JOIGN) run by the University of Arizona and the San Bernardino Mountains strand replacing the older, inactive strand (Waco and Blisniuk, 2019). Nearer network run by California State University, San Bernardino (Fig. 2 and Fig. S11 to San Gorgonio Pass to the south of the MMC, Gold et al. (2015) completed in the Supplemental Material). The JOIGN network comprises 21 stations the first geologic slip-rate study along the Banning fault in the Coachella Valley, including 12 mast-mount, campaign-style monuments installed by the Uni- and demonstrated a low right-lateral slip rate in the range of 2–6 mm/yr, effec- versity of Arizona in 2005 and 11 historical National Geodetic Survey brass tively indicating that this strand carries only a small portion of plate motion markers that we occupy with constant height spike-mounts, which together at this latitude. Another study along the San Gorgonio thrust conducted by complement the 13 continuous GPS stations in the area, by providing increased Heermance and Yule (2017) presents a low oblique slip rate of ~6 +3/−2 mm/yr coverage in remote desert wilderness areas of Joshua Tree National Park. All across two strands of the fault and an argument for transfer of strain to some JOIGN stations have time series going back to 2005 or earlier and have been other portion of the plate boundary system. Moreover, the recent occurrence reoccupied for a minimum of 12 h to a maximum of 7 days at varying occu-

of three ≥Mw7.0 and three additional ≥Mw6.0 earthquakes over the past 28 pation frequencies (Fig. 3). All campaigns on the JOIGN network used Trimble

years in the ECSZ (1992 Mw6.1 Joshua Tree, Mw6.3 Big Bear, and Mw7.3 Landers, Zephyr Geodetic antennas and a variety of receiver types from the UNAVCO

the 1999 Mw7.1 Hector Mine, and the recent 2019 Mw6.4 and Mw7.1 Ridgecrest instrument pool, including Trimble 4000SSE/SSI, Topcon GB-1000, Trimble

earthquakes), not to mention the five ≥Mw5.0 events that occurred since 1947 5700/R7, and Trimble NetRS. GPS phase data were sampled continuously at along the same linear trend, lends support to the hypothesis of an emerging 15 s intervals nominally over their occupation observation periods. The San fault zone, dubbed the “Landers-Mojave Earthquake Line” of Nur et al. (1993). Bernardino Mountains network monuments are mostly historical survey mark- The prevalence of earthquakes along this zone raises the question of how ers and have varying measurement histories. Observations by CSUSB began elastic strain from the SSAF is being transferred and converted into perma- between 2002 and 2009 and continued through 2016 for most sites, but some nent strain in the form of fault slip during earthquakes in the ECSZ. In other sites include observations by other agencies that go back to the early 1990s. A JOIGN Campaign words: how might these two systems work in tandem to accommodate plate These monuments were reoccupied bi-annually to annually for a minimum of Network KEY2JT08 JT07 boundary deformation over geologic time scales? The fact that this “Land- eight hours to a maximum of five days, using tripods and spike-mounts and SGPK JT03 JT01 LITE WARR JT05 JT02 34˚N JT10 JT09 F726 PB21 ers-Mojave Earthquake Line” intersects the San Andreas system, just where either Ashtech ZXtreme or ZXII receivers with Ashtech choke-ring antennas JT06 JT04 PB15 SJPK 8252 BLOY JT12 JT11 JTRE DESO the SSAF trifurcates into the MMC, Banning, and Garnet Hill strands, suggests or Trimble 4000SSE/SSI or NetR9 receivers with Zephyr Geodetic antennas Palm 1114 Springs 1113 that it could potentially play a role in reducing the amount of slip accruing in (Fig. 3). All collected raw campaign data from the San Bernardino Mountains 50 km

B PT65 LUCS San Gorgonio Pass. Taken together, these recent events and studies indicate and the JOIGN network are archived and publicly available at UNAVCO, and MONA CSUSB GVWP LUNE MEEK RICU WMTN DEAD Campaign DEVL NWAL MEAD ONYXBRNS the need for a reevaluation of how slip is being partitioned within this com- station names and locations are presented in Table S1 (footnote 1). 6108 7211 MNA2 Network AWHD KELL SKYL U471 BOUR HIGH PITS 3SIS CHPA BRYN K526 DIVD FSCR 5297 6106 MILL ULMO INA5 HI65 plex fault zone, in order to both determine the distribution of plate boundary To process raw campaign data into final time-series products, we use the MOGO LS35CHER G011 INA3 34˚N SANO MATX LNGC CLSA G068 BRIN DVID 0358 NORC FATL G069 G021 0822 zone deformation and to better quantify the seismic hazard and possibility of GAMIT/GLOBK software suite (Herring et al., 2018) and follow standard con- RC67 0181 LACY LAST CABZ 0184 METB RYN6 Palm G076 Springs through-going rupture potential in San Gorgonio Pass. ventions to analyze the JOIGN and San Bernardino Mountains data sets in

Surveyed by G078 50 km UC Riverside conjunction with 24 local and regional continuous station data sets. These 24 −116˚W continuous stations serve as tie-in stations when we use GLOBK to combine ■■ METHODS our processed campaign data with the final SINEX solutions from the Plate Boundary Observatory (PBO) (archived and publicly available at UNAVCO 1 Supplemental Material. Tables of statistical details GPS Data [Herring et al., 2016]). This process incorporates the full set of available contin- regarding estimation of GPS velocities (including the uous GPS stations into our analysis. We include 337 total continuous stations map position and names of all campaign and continuous station sites); the full model assessments of GPS Collection and Processing spanning the breadth of southern California from just north of the Garlock fit; and figures that present details of all strike-slip fault to the southern border with Mexico, for all years between 2000 and 2018. and dip-slip fault-slip rates calculated within models. While southern California is known for its extensive continuous GPS data The final product is a collection of coordinate time-series data that record Please visit https://doi.org/10.1130/ GEOS.S .13123292 to access the supplemental material, and contact ed- coverage, a handful of areas still exist in which terrain is too rugged, satellite seasonal, interseismic, co-seismic, and postseismic site motions throughout [email protected] with any questions. visibility is too obscured by trees or other forms of vegetation, or wilderness the study region.

800

P511 LUCS 600

LAST 400 BRYN MLFP

AGMT East (mm) CHER WMTN ORMT 200 JT10 JT06 WIDC 0 JT05 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 Time (years)

LUCS P511 800 LAST MLFP 600 BRYN CHER AGMT WMTN 400 ORMT JT10 WIDC North (mm) JT06 200 JT05

0 2000 2002 2004 2006 2008 2010 2012 2014 2016 2018 Time (years) Figure 3. Selected examples of postseismic-reduced coordinate time series for continuous GPS stations (left) and campaign GPS stations (right) used in our block modeling analysis. Four-letter codes at right of each block are station names. Vertical gray dashed line delineates the time of the 2010 M7.2 El Mayor– Cucapah earthquake, and short blue dashed line in WIDC time series at left delineates a station antenna change offset. All time series presented in stable North America reference frame (NAM08) and red velocity models estimated using GAMIT/GLOBK utility TSFIT (Herring et al., 2018).

Accounting for Transient Viscoelastic Postseismic Displacements caused by large-magnitude earthquakes in southern California, with the goal of estimating long-term fault-slip rates unbiased by the secular components When crustal stress changes associated with large-magnitude earth- of transient postseismic surface motions. quakes perturb the viscoelastic lower crust and upper mantle, they excite To distinguish which earthquakes in the historical and recent earthquake time-dependent, viscoelastic, postseismic displacements measurable by GPS record could still be measurably contributing to GPS time series in southern instruments. Depending on the size and location of the event, these transient California, Guns and Bennett (2020) compute forward model displacements

postseismic displacements can last from years up to decades into the future. for 217 ≥Mw6.0 earthquakes that have occurred in the southwestern United If unaccounted for, postseismic displacements can bias estimates of surface States, Baja California, and Sonora Mexico since year 1800 (Wang et al., 2009). velocity derived from the trends of GPS coordinate time-series data. Ongo- To compute the viscoelastic component of deformation, we use the PSGRN/ ing viscoelastic effects from historical earthquakes occurring before the GPS PSCMP program (Wang et al., 2006) in combination with a laterally homo- observation period can still be contributing to the modern-day deformation geneous, layered Earth model developed by Broermann (2017). This layered field (Hearn et al., 2013). These ongoing viscoelastic effects can be difficult to Earth model consists of a 15-km-thick elastic upper crust, a 15-km-thick lower- isolate and account for due to the fact that the long-term components of post- crustal Maxwell rheology (viscosity = 1019.9 Pa s), followed by a 30-km-thick 19.6 seismic displacements are difficult or impossible to distinguish from secular lithospheric mantle Burger’s Body rheology (Maxwell η1 = 10 Pa s, Kelvin 18.6 trends after only a relatively short time following the earthquake, depending η2 = 10 Pa s), and a sublithospheric mantle Burger’s Body below a depth 18.8 1 7. 8 on whether it is horizontal or vertical deformation (Smith and Sandwell, 2004). of 60 km (Maxwell η1 = 10 Pa s, Kelvin η2 = 10 Pa s). The model adopts a Here we use the results of Guns and Bennett (2020), who applied a forward seismic velocity and density structure derived from CRUST 1.0 (Laske et al., modeling strategy to identify and reduce the ongoing short- and long-term 2013; see Guns and Bennett, 2020, for detailed structure). Broermann (2017) effects of viscoelastic postseismic deformation on modern GPS observations estimated this layered model to fit the wide, lithospherically diverse region

of southern Arizona, the Colorado Plateau, and the Basin and Range province, and the final estimated viscosity structure agrees well with many estimates of A Postseismic-reduced Velocity Field 20 ± 1 mm/yr viscosities in southern California (Guns and Bennett, 2020, and their table 1). In assessing these modeled postseismic displacements, we discovered 12 past events are likely still contributing measurable viscoelastic displacements to the deformation field in any year of our time series (2000–2018) within the uncertainties of sub-millimeter precision of continuous GPS data. These

earthquakes include, in time-order: the 1812 Mw7.5 Wrightwood, 1857 Mw7.9

Fort Tejon, 1872 Mw7.4 Lone Pine/Owens Valley, 1892 Mw7.3 Laguna Salada, 34˚N 1906 Mw7.8 San Francisco, 1918 Mw6.8 San Jacinto, 1952 Mw7.5 Kern County,

1992 Mw7.3 Landers, 1994 Mw6.7 Northridge, 1999 Mw7.1 Hector Mine, 2010 Mw7. 2

El Mayor–Cucapah, and the 2012 Mw7.0 Baja California events. Once identified, Palm Springs we subtracted the modeled co-seismic and postseismic displacements asso- N ciated with these earthquakes from our coordinate time series to produce a postseismic-reduced GPS data set (Figs. 3 and 4). To remove any further leftover 50 km postseismic displacements for recent, observed earthquakes and their postseismic displacements, we estimate parameters associated with logarithmic Postseismic Corrections−116˚W 5 mm/yr postseismic decay and remove any residual postseismic motions from our GPS time series. When we complete this logarithmic estimation step for the

well-observed 2010 Mw7.2 El Mayor–Cucapah earthquake, we calculate the variance reduction to be 60% using two different estimation programs, illustrating successful reduction of the majority of postseismic displacements for southern California using our modeling scheme. This ensures that we obtain a more accurate interseismic GPS velocity field for our block modeling approach. 34˚N Estimating Final Velocities Palm Springs We estimate our final velocities for our (1) postseismic-reduced (Fig. 4A), N (2) observed (unreduced), and (3) partially reduced GPS time series using a GAMIT/GLOBK utility called TSVIEW/TSFIT (Herring et al., 2016). These pro- 50 km grams estimate earthquake displacements, periodic seasonal motions, and secular trends using a weighted least-squares inversion in either a MATLAB- −116˚W based Graphical User Interface (TSVIEW) or a command line setting (TSFIT). Figure 4. Map plot of (A) final postseismic-reduced GPS velocity field, and (B) the post- We employ the RealSigma/First Order Gauss Markov Extrapolated option seismic earthquake corrections used for the summed effects of the 12 earthquakes (see to estimate noise within the time-series data by estimating an a posteriori Guns and Bennett, 2020, for details). Velocities from stations within dashed white model boundary line are used to constrain fault-slip rates in our model inversion. Velocity field scale factor to account for the noise processes by assuming a First-Order presented in stable North America reference frame (NAM08). Light-blue diamonds mark Gauss-Markov error process and analyzing post-fit residuals. All continuous locations of campaign GPS sites, while dark-blue diamonds mark continuous site locations. station velocities were estimated first using TSFIT, but in cases where specific station time series exhibited excessively noisy observations or unexplainable behavior, we choose to visually edit these time-series records by hand in unexplained modeled signals (e.g., RKMG). For estimating velocities from our TSVIEW, to estimate velocities from the good data rather than removing the campaign network sites, we use TSFIT to estimate the same known offsets station from our analysis. We edited 15 continuous station and five campaign and logarithmic postseismic terms that are applied to the continuous stations station time series by hand, the names and statistics of which are presented as well as secular trend, but we do not estimate any annual or semiannual in Table S2 (footnote 1). These stations were chosen because of the presence seasonal terms. We apply an outlier cutoff of 3-sigma for all campaign stations, of un-modeled cyclic noise (e.g., AVRY), short periods of high noise volume because some stations have distinct visible outliers. For stations that have out- (e.g., DYHS), un-modeled or unknown offsets (e.g., COSO), or because of liers that are not accommodated by this outlier cutoff, we estimate velocities by

hand within TSVIEW (see Table S2). Final campaign station velocities appear Pleistocene or Holocene observations, rather than long-term million-year offset together with our continuous station velocity estimates in Figure 4A. estimates) in order to have the closest possible comparison to geodetic rate, We estimate velocities for our three different GPS time-series data sets while making no assumptions regarding whether fault-slip rates may vary in order to better understand the differences in the resulting fault-slip rates over time scales of an order of 104–105 years. Although the geologic record estimated from each set. Using our newly calculated, postseismic-reduced shows that fault zones evolve on million-year time scales, there is currently GPS velocity field will allow block modeling to estimate long-term fault-slip no agreed-upon mechanism or geodynamic explanation for fault-slip rate rates because this velocity field has had both short-term (recent) and long- variations on time scales of less than an order of 100,000 years. As a result, term (historical) viscoelastic postseismic displacements removed. It should when discrepancies are observed between Holocene, Late Pleistocene, and be noted that while removing these short- and long-term transients brings us geodetic slip rates (e.g., Weldon et al., 2004), it can be difficult to assess much closer to a time-invariant velocity field, there could still be additional whether such differences are real and associated with a poorly understood postseismic deformation that is not accurately captured by the chosen model. deformation process, or if they represent epistemic errors in either the geo- Hence, we refer to this field as a postseismic-reduced field, rather than a detic or geologic methods used to estimate fault-slip rate. While making any postseismic-removed field. Inverting the unreduced, or currently observed comparison between modern-day geodetically estimated fault-slip rate and a GPS velocity field will enable us to capture how the current deformation field time-averaged, long-term geologic rate is potentially incompatible due to the (including all measured postseismic displacements) is mapped onto each of large difference in scale between the two observation data sets, it is making our fault geometries by estimating the modern-day slip rates. Our third velocity this comparison that can give us a window that may improve understanding field represents the type of data set that would normally be used in this type of of how fault-slip rate may have changed or remained constant through time. block modeling inversion: a partially reduced GPS velocity field that has only the short-term postseismic displacements from recent, observed earthquakes removed. This last velocity field will be free of the effects of recent earthquakes Block Modeling Using TDEFNODE

such as the 1999 Mw7.1 Hector Mine and the 2010 Mw 7.2 El Mayor–Cucapah earthquakes but will still contain ongoing viscoelastic postseismic deformation To better characterize how plate boundary motion is being transferred displacements from historical, large-magnitude earthquakes. through the network of faults in and around the San Gorgonio Pass area, we construct a suite of 16 different elastic fault-block models using the Fortran program TDEFNODE (McCaffrey, 2005; McCaffrey et al., 2007). This program Geologic Slip-Rate Data can use a variety of data inputs, in combination with a user-specified block geometry, to invert for rigid body rotations of fault-bounded rigid blocks, cal- Southern California has one of the most extensive records of geologic culate the elastic deformation created by locking along a fault interface, and fault-slip rates, slip histories, and earthquake recurrence intervals in the North thereby estimate the interseismic, or long-term, fault-slip rate between the America plate boundary (Scharer and Streig, 2019), which makes it an ideal rotating blocks. We use our postseismic-reduced GPS velocity field, our par- place for comparing geologic slip-rate measurements with estimates of geo- tially postseismic-reduced velocity field, and our unreduced observed GPS detic slip rate. Table 1 presents the slip-rate data either included in our analysis velocity field as data inputs into our inversions so that we may compare the of geodetic data fit or included in our discussion for comparison. In select- effect that reduction of postseismic deformation has on fault-slip rate estimates ing rates to use to assess how well our calculated geodetic rates agree with for each of our 16 different fault geometries. geologic rates, we choose those that have three main characteristics: (1) the geologic slip-rate measurement is in an accessible published paper, meeting abstract, or thesis or dissertation; (2) the slip rate has clearly defined location Inversion Parameters either through map figures, specific place names, or reevaluations of known locations; and (3) the rate has Holocene or Late Pleistocene age constraints Depending on what the user chooses to solve for in TDEFNODE, the pro- and clearly defined uncertainties. Slip-rate study locations presented in papers, gram can run a linear least-squares inversion, a simulated annealing process, abstracts or theses and/or dissertations were verified in Google Earth™. Well- or a grid search method, all minimizing the reduced χ2 misfit to the data inputs: 2 defined uncertainties are necessary in our calculation of geodetic fit to geologic N 2 1 (di ,obs di ,pred ) rates; therefore, we only use sources that describe their slip-rate calculations in reduced misfit = , (1) DOF 2 detail, approximating the probability distributions as either a zero mean Gauss- i =1 i ian with a reported standard deviation or uniform (boxcar) with a preferred where DOF is the degrees of freedom in the inversion problem (number of or “best” value within a specified range of permissible values. We choose to data observations minus the number of free parameters being estimated),

use only the most recent geologic time-based geologic slip rates (either Late N is the number of data observations, di,obs is the ith data observation, di,pred is

2 the ith model predicted observation, and σi is the ith variance, or uncertainty. as seismicity patterns, gravity lows, topography changes, and regional-scale In our model setup, we employ a linear least-squares inversion to estimate GPS velocity changes can also aid in testing block boundary fault locations the M number of Euler vectors describing crustal block motion using our (McCaffrey, 2005; Elliott et al., 2010; Spinler et al., 2010). We use a combination GPS velocity fields, which useN = 392 data observations for our postseismic- of known fault surface and subsurface geometries digitized by the Southern reduced velocity field. We do not include geologic data as constraints in any California Earthquake Center Community Fault Model (CFM) (Plesch et al., 2007; of our inversions. In our most complicated model construction, we estimate Nicholson et al., 2017), the Uniform California Earthquake Rupture Forecast 15 Euler vectors from 15 blocks, thereby estimating 45 individual parameters (UCERF3) report (Dawson and Weldon, 2013), and mapped fault traces or his-

(the Cartesian components of the Euler vectors, ωx, ωy, ωz for each of 15 blocks). torical rupture traces through the U.S. Geological Survey Quaternary Fault and The forward problem for our reduced GPS field case is shown as: Fold Database (U.S. Geological Survey and California Geological Survey, 2006) to construct our model fault geometries. We choose our most complex initial

dGPS = G m + e (2) block-model geometry to represent all major faults in the Peninsula Ranges, Euler vectors ( x , y , z ) San Gorgonio Pass, San Bernardino Mountains, Eastern Transverse Ranges, and Eastern California shear zone areas. This geometry is presented in Figure 5 d1 G1,1 G1,45 m1 e1 and includes 16 total blocks and 31 total fault segments. Table 2 presents our = + , (3) assigned locking depths and fault dips for all numbered faults in our models.

d392 G392,1 G392,45 m45 e392 The TDEFNODE program requires the specification of a hanging-wall block for each fault segment; therefore, to accommodate that specification, we assign all

where dGPS is the 392-element array containing the horizontal components of our postseismic-reduced GPS velocity field sampled at 196 observation points 34.8°N (i.e., with both north and east components), G is our mapping function matrix of Green’s Functions (unit response functions to motion along constructed 16 20 T WMOJ CMOJ EMOJ fault nodes at depth), mi,Euler vectors = (ωix, ωiy, ωiz) are the chosen model parame- 22 CLEG SBAR 23 MESV ters to estimate (block motions for each block in the model geometry), and e 28 17 15 18 represents measurement errors with variance matrix V = EeeT from our TSFIT 21 analyses (where E = the mathematical expectation operator). The diagonal SGAB27 4 SBMT 19 26 12 13 SGPA 14 elements of V are derived from the uncertainties provided as part of the GPS 31 7 11 MORO 24 30 9 JTNP coordinate time-series data product. The least-squares solution for the model 34°N 29 WWTR 6 10 parameters is given by: 3 8 5 1 1 Palm Springs 25 ˆ T 1 T 1 mLS = G V G G V dGPS, (4) ANZA ELSI PERA 2 CHIR

where mˆ LS is our array of model-predicted Euler vector parameter estimates for each block. This problem is overdetermined, meaning the number of observed 33.6°N N data, N, is greater than the number of model parameters, M, and the system 50 km 0 km 50 km is well posed in the sense that the matrix [GT V −1G]−1 exists without damping or regularization. For this case, there exists one set of model parameter esti- −117.8°W −116°W −115.2°W mates that minimizes Equation (1), providing the closest approximation to Figure 5. Block model geometry including block names and numbered faults for our most the observed velocity data values for each model geometry that we tested. complicated block geometry (SGP1). See Table 2 for numbered fault names and locking depths. Fault surface geometries were chosen through comparison with the mapped traces present in the U.S. Geological Survey Quaternary fault map and through the use of the Southern California Earthquake Center Community Fault Model (SCEC CFM) and the Model Construction Uniform California Earthquake Rupture Forecast (UCERF3), particularly for fault geometry at depth. Light-blue diamonds mark locations of campaign GPS sites, while dark-blue In defining our block geometries in TDEFNODE, we first begin by con- diamonds mark continuous site locations. ANZA—Anza, CHIR—Chiriaco, CLEG—Cleg- horn, CMOJ—Central Mojave, ELSI—Elsinore, EMOJ—East Mojave, JTNP—Joshua Tree structing the most complicated fault geometry and then alter it to generate National Park, MESV—Mesquite Valley, MORO—Morongo, PERA—Peninsula Range Area, our 15 other geometries. Most often, block geometry decisions are guided by SBAR—South Barstow, SBMT—San Bernardino Mountains, SGAB—San Gabriel, SGPA— known mapped surface traces of active faults, though lines of evidence such San Gorgonio Pass Area, WMOJ—West Mojave, WWTR—White Water.

vertical faults the near-vertical dip of 89°. For all other faults, we assign patch TABE 2. MODE FAT OCKING DEPTHS AND DIPS dips based on values within the SCEC CFM5 and the UCERF3 fault geometry Fault name Fault number ocking depth Dip catalogs (Plesch et al., 2007; Dawson and Weldon, 2013). in model assigned assigned By specifying coupling on the defined fault geometry, we can produce km elastic deformation in the blocks, which is then estimated at the surface by the San Jacinto 1 1 80–89 TDEFNODE code using an elastic half-space dislocation model (Okada, 1985, San Andreas fault, Coachella alley 2 10 89 1992). Locking is chosen by defining the coupling fraction term,f , to be either 0 San Andreas fault, Banning–Garnet Hill– 3 20 21–86 San Gorgonio Pass or 1, where f = 0 is freely slipping (no coupling) and f = 1 is completely locked San Andreas fault, San Bernardino 1 86–89 (McCaffrey et al., 2007). A common method for applying locking is to assign f San Andreas fault, Mission Creek ,6 1 81–89 values of 1 to all fault nodes down to the locking depth, d, and to then assign Galena Peak 30 1 89 values of 0 to all nodes below d to simulate the freely slipping nature of a dis- Mill Creek 7 1 81–86 location below the locking depth (Savage and Burford, 1973; McCaffrey, 2005) Burnt Mountain Eureka Peak 8,9 1 78 (Fig. S2 [footnote 1]). We implement this scheme of keeping the faults locked (f Blue Cut 10 10 89 Pinto Mountain 11,12,13,1 10 89 = 1) to the assigned locking depth at each fault, and letting the fault freely slip est North Frontal 1 10 6–89 (f = 0) below that locking depth. All faults have nodes placed at 5 km down-dip Helendale–South ockhart 16,17 10 89 intervals (down to 25 km), ensuring that they connect together properly in the East North Frontal 18 10 3 subsurface, and they are locked (f = 1) to their assigned locking depth. In order Johnson alley 19 10 89 to incorporate more realistic details into our fault geometries, and to lower our North Emerson 20 10 89 overall misfits to the GPS velocity field data, we incorporate variable locking Pisgah-Bullion 21 10 89 Calico-Hidalgo 22 10 89 depths along different sections of the fault geometries, which are presented udlow–Cleghorn ake 23 10 89 with assigned fault names and numbers in Table 2. These locking depths are Sheep Hole 2, 2 10 89 selected based on the depth of located microseismicity in the subsurface fault Cucamonga 26 1 8 zone, observed from work by Hauksson (2000). We also implemented slightly San Jacinto, Caon Pass 27 20 8 different locking depths as represented through work by Richards-Dinger and San Andreas fault, Moave 28 1 89 Shearer (2000) but found that using the microseismicity depths of Hauksson Elsinore 29 1 68–88 (2000) produced lower χ2 misfits to the GPS data. Assigned based on depth of seismicity from Hauksson 2000. In order to test various plausible hypotheses of fault activity in this com- Assigned based on dips from the SCEC Community Fault Model CFM and CFM and CERF3 values. plex zone of deformation, we chose to construct 16 different models that each consist of as little as five total blocks to the full 16 blocks (Fig. 6). Each block model tests a different hypothesis of fault activity and possible block motion in this network and is named by decreasing model complexity, with our most We test three end-member scenarios including (1) a San Jacinto + Banning– complex model being SGP1 (Table 3). To create our different models, we start San Gorgonio Thrust + Mission–Mill Creek system–only model, testing the with our most complex model (Fig. 5), and assign groups of adjacent blocks viability of total motion accommodated by the “main” San Andreas system to the same pole of rotation effectively forcing them to move together. When (SGP15); (2) a San Jacinto + Eastern Transverse Ranges and Eastern California multiple blocks are assigned the same pole of rotation, they rotate together shear zone block rotation–only model testing the viability of possible micro as one body (with zero motion on the faults between them) thereby acting as block rotation transferring motion and strain away from the main San Andreas one larger block. Each model run produces its own estimate of Euler poles, system to the Eastern California shear zone (SGP11); and a (3) San Jacinto so each time a model geometry includes “grouped” blocks as one block, a + Burnt Mountain–Eureka Peak (Landers Mojave Earthquake Line) scenario that new Euler pole is estimated for that “grouped” block. Therefore, in Figure 6, tests the possibility of an emerging fault zone taking the brunt of plate motion when multiple blocks have all been colored the same color, it means they have (SGP16). The rest of the models test variations of fault-geometry hypotheses been forced to move together as one block. Other approaches to grouping that envision possibilities of strain and plate motion being accommodated blocks, including total regularization variation (e.g., Evans et al., 2015; Evans across different sections of the diffuse fault network. Each of these geome- et al., 2016) and cluster analysis (e.g., Thatcher et al., 2016), rely on geodetic tries was selected and included in order to test the activity levels of certain data as a primary input to define block boundaries. We chose to define our faults (e.g., SGP9 tests whether GPS data are better fit with an inactive Mill block boundaries in a way that is primarily informed by geologic information Creek fault section, while SGP7 tests whether the Blue Cut fault needs to be in the San Gorgonio Pass Area; however, future work could benefit from a active to fit GPS data best, etc.). See Table 3 for our entire list of hypothesis more algorithmic approach. descriptions.

SGP1 N = 48 SGP2 N = 45 SGP3 N = 42 SGP4 N = 42

34ºN 34˚N 34˚ 34˚

50 km 0 km 50 km 50 km 0 km 50 km 50 km 0 km 50 km 50 km 0 km 50 km SGP5 N = 42 SGP6SGP6 N = 39 SGP7 N = 39 SGP8SGP8 N = 39

34ºN 34˚ 34˚ 34˚

50 km 0 km 50 km 50 km 0 km 50 km 50 km 0 km 50 km 50 km 0 km 50 km SGP9 N = 39 SGP10 N = 36 SGP11 N = 30 SGP12 N = 24

34ºN 34˚ 34˚N 34˚

50 km 0 km 50 km 50 km 0 km 50 km 50 km 0 km 50 km 50 km 0 km 50 km SGP13 N = 21 SGP14 N = 18 SGP15 N = 18 SGP16 N = 15

34ºN 34˚ 34˚N 34˚N

50 km 50 km 50 km 0 km 50 km 0 km 50 km 50 km 0 km 50 km 0 km 50 km −116˚W −116˚W −116˚W −116˚W

Figure 6. Illustration of all tested block model geometry hypotheses. See Table 3 for hypothesis descriptions. See Figure 5 in conjunction with Table 2 for block and fault names.

TABE 3. BOCK MODE FAT GEOMETRY HYPOTHESES Model Number of Number of model Geometry description name blocks parameters SGP1 16 8 Testing the possibility of an active Galena Peak fault lying between the Mission–Mill Creek fault and the San Gorgonio Pass fault SGP2 1 Testing the possibility of all potentially active faults presently accommodating plate motion together SGP 1 2 Testing the possibility of simultaneous transfer of plate motion through San Gorgonio Pass fault and through block rotation in the Eastern Transverse Ranges no Burnt Mountain fault SGP 1 2 Testing the possibility of simultaneous transfer of motion through all faults but the Blue Cut fault SGP3 1 2 Testing the possibility of all potentially active faults presently accommodating plate motion together, ecept for the removal of the estern North Frontal fault SGP9 13 39 Testing the possibility of an active San Gorgonio Pass Fault and Mission Creek Fault no Mill Creek or Burnt Mountain fault motion SGP8 13 39 Testing the possibility of an active San Gorgonio Pass fault with no Mission–Mill Creek motion SGP7 13 39 Testing the possibility of an active San Gorgonio Pass fault and Mission–Mill Creek system, with the westernmost section of the Pinto Mountain fault and the Burnt Mountain fault inactive SGP6 13 39 Testing the possibility of an active San Gorgonio Pass fault, active Mission–Mill Creek system, with no activity on the Blue Cut or Burnt Mountain faults in the Eastern Transverse Ranges SGP10 12 36 Testing the possibility of an active San Gorgonio Pass fault only no Mission–Mill Creek system working with a anders-Moave Earthuake ine one SGP11 10 30 Testing the end-member scenario of near total block rotation in the Eastern Transverse Ranges transferring slip to the Eastern California shear one SGP12 8 2 Testing the possibility of an active San Gorgonio Pass fault and Mission–Mill Creek system working together with the anders Moave Earthuake ine to accommodate plate motion SGP13 7 21 Testing the possibility of all faults that have Holocene displacements or historical earthuake ruptures being active SGP1 6 18 Testing the end-member scenario of ust San Andreas motion with an active San Gorgonio Pass fault, Mission–Mill Creek system SGP1 6 18 Testing the possibility of the combination of only the San Jacinto, the San Gorgonio Pass fault, and the anders-Moave Earthuake ine one to accommodate all plate motion SGP16 1 Testing the end-member scenario in which the anders-Moave Earthuake ine one works with the westernmost faults to accommodate plate motion Note: See Figure 6 for block model geometry maps.

Goodness-of-Fit Scores of freedom, N–M (data–parameters). For the majority of models tested, the χ2 One of the caveats of elastic fault-block modeling is that one can often end misfit values decrease as the number of model blocks increases (see Fig. 7). up with multiple models that fit the data with equally low reduced χ2 misfit To calculate this probability value, one must compare only two models at a scores. However, using reduced χ2 misfit (defined in Equation 1) to delineate time—one model with higher misfit and lower number of parameters, and which models in a suite best fit the input data is useful in weeding out model the second model with lower misfit and higher number of parameters. Then geometries that have very poor fits to the observed input data. Here, we attack one can directly assess whether the added parameters actually were neces- this problem by purposefully not choosing one overall best-fitting model, but sary to decrease misfit, or whether that decrease is due to chance. We use a rather a handful of best models that point toward best fits to either the GPS MATLAB-based program written by Anderson and Conder (2011) to calculate data or to observed geologic slip-rate data. In order to determine what best the F-test probability values, which takes the number of data, number of free fits our observed (GPS and/or geologic) data, we use three different scores: parameters for each of the two models, and the unreduced χ2 misfit for each (1) reduced χ2 misfit (defined earlier), (2) F-test probabilities between low χ2 of the models as inputs. misfit model pairs, and (3) RMS scores to geologic observations for overall For our RMS misfit scores to geologic observations, we devise a calcula- geologic rates, Holocene-only rates, and Late Pleistocene rates. tion that compares the residuals between model-produced geodetic fault-slip We assume that, in general, models that have increased complexity and rates and the geologic slip rates from the literature. In order to assess whether number of parameters tend to fit the data better with lowerχ 2 scores. This geodetic fault-slip rates might match more recent geologic rates or those result is not guaranteed, however, because it is possible to create a model that farther in the past, we calculate a Holocene-rate–only RMS score and a Late contains many parameters but in, for example, an unrealistic fault configu- Pleistocene–rate–only RMS score, in addition to an overall geologic-rate RMS ration, which would not produce as good a fit as a smaller but more realistic score. Our overall geologic rate RMS score is calculated as: fault configuration might. Nevertheless, we may use an F-test for the discrim- 1 N 2 2 2 1 (geologic slip ratei − geodetic slip ratei ) ination between models of varying complexity (i.e., differing M) and χ misfit, Geologic MS score = 2 2 N i =1 geologic slip rate uncertainty ( 1σ) + geodetic slip rate uncertainty (1 σ) by calculating a probability (p-value) that a lower misfit (improved model fit ( i 1 i ) N 2 2 1 (geologic slip ratei − geodetic slip ratei ) to observations) is justified given the corresponding reductionGeologic in the MSdegrees score = , (5) N 2 2 i =1 (geologic slip rate uncertaintyi ( 1σ) + geodetic slip rate uncertaintyi (1 σ) )

Comparing Model Complexity and χ2 Misfit Values compares to known geologic data. These scores of fitness between geodetic 60 data and geologic data allow us to determine which models fit best for each Reduced Field data set and even both data sets. Observed Field 50 Partially Reduced Field ■■ RESULTS 40 We introduce results of TDEFNODE modeling for three sets of 16 differ-

Misfit Value ent model geometries: one set inverting postseismic-reduced GPS velocities 2 30 (Figs. S3 and S4 [footnote 1]), one set inverting the partially reduced GPS Lowest) Lowest) Lowest) rd

th velocity field (Figs. S5 and S6), and one set inverting the observed (unre- nd 2 (4 (3 (

(Lowest Misfit) duced) GPS velocity field (Figs. S7 and S8). Table 4 presents the comparison

Reduced χ 20 of calculated reduced χ2 misfit and geologic RMS score goodness-of-fit results SGP8 SGP5 SGP1 SGP2 for the postseismic-reduced and observed model sets (see Tables S3A–S3C

10 for full model parameters and GPS NRMS/WRMS fits for all three data sets). The lowest reduced χ2 misfit models are SGP2, SGP1, SGP8, and SGP5 from N = 39 N = 42 the postseismic-reduced GPS data set of models (see Fig. S9 for calculated 0 residual fields). While their counterparts in the observed GPS data set also 10 15 20 25 30 35 40 45 50 share low relative misfits to other model geometries in the observed data set, Number of Model Parameters (N) over all models, it appears that the full postseismic reduction in the GPS data 2 Figure 7. Plot of χ misfit against the number of model parameters for each model lowered misfits by up to 51% when compared to the observed GPS data set geometry using the fully reduced GPS velocity field (diamonds), the observed (Fig. 7; Table S4). Moreover, using the fully reduced data sets results in up to GPS velocity field (circles), and the partially reduced GPS velocity field (triangles). Models names are given based on model complexity, with SGP1 as the model a 45% decrease in misfit when compared to the partially reduced GPS data set. to the farthest right (most complicated model) and SGP16 as farthest left (least Models SGP2, SGP1, SGP8, and SGP5 are four of the eight most complicated complicated model). Our top four best-fitting models to GPS data are labeled. geometries (the highest number of blocks and faults), and due to this added complexity, we assess whether the larger number of parameters reduced misfit by chance, or whether those additional parameters are warranted in their where N is the number of geologic slip rates in the calculation (Holocene rates improved fits. For this, in Table 5, we present the results of our F-tests, which only N = 14, Late Pleistocene rates only N = 18, and overall geologic rates is N = demonstrate that the probability that these lower misfits are due to chance 27 because we give preference to Holocene rates at sites where there are both is less than 0.03% in all comparison cases, except SGP8 and SGP1, which

Pleistocene and Holocene rates [Table 1]), geologic slip ratei is the ith geologic suggests that the added blocks and fault boundaries in these more complex

slip rate in a given set, geodetic slip ratei is the ith geodetic model fault-slip rate models (particularly SGP2 and SGP5) are necessary to improve overall reduced that corresponds to the closest point location to the location of the associated χ2 misfit values. In addition, we run an F-test between our most complicated

geologic slip-rate observation, geologic slip rate uncertaintyi is the uncertainty model (SGP1) and one of our least complicated models (SGP15); this F-test on the geologic rate as determined by the authors of the particular study and produces a probability of 0%, which underscores the fact that more added

converted to 1σ for use here, and geodetic slip rate uncertaintyi is the 1σ uncer- parameters are warranted to decrease the overall misfit (Table 5). tainty calculated by the TDEFNODE model. These values are squared to form The models that have the lowest RMS scores to the overall geologic slip- variances, resulting in a measure of the average agreement (Equation 5), which rate data set are SGP4, SGP5, SGP2, and SGP1. This set of models is intriguing is unitless. If a geologic slip-rate investigation has reported both Holocene and given that three of these same models best fit the postseismic-reduced GPS Late Pleistocene rates (refer to Table 1 for specific authors), then we use the velocity data set. The postseismic reduction also decreases model RMS scores Holocene rate in the overall geologic RMS score calculation, and then use the for the overall geologic slip-rate data set category by again up to 19%, indi- specific Holocene and Late Pleistocene rates in each respective score calculation. cating that postseismic reduction brought geodetic fault-slip rates closer to For model geometries in which there is a geologic slip rate recorded along Pleistocene and overall geologic slip rates. In the case of Holocene slip rates, a fault, but that fault is not active in the model, we penalize the RMS score however, RMS scores actually increase between the observed GPS data set calculation by assigning a geodetic rate of zero to be compared with the and the postseismic-reduced data set. The lowest Holocene RMS score mod- measured nonzero geologic rate. This ensures that all geologic observations els for the postseismic-reduced data set are SGP7, SGP13, SGP6, and SGP5, are included and that the RMS score is meaningful to tell us how each model suggesting that very different model geometries actually fit Holocene slip-rate

TABE . BOCK MODEING GOODNESS OF FIT REST COMPARISON Model Postseismic-reduced GPS velocity field Observed GPS velocity field no reduction of postseismic deformation name χ Misfit Number of RMS residuals RMS residuals to RMS residuals to χ2 Misfit Number of RMS residuals RMS residuals to RMS residuals to alue model to geologic Holocene geologic Pleistocene value model to geologic Holocene geologic Pleistocene parameters slip rates slip rates geologic slip rates parameters slip rates slip rates geologic slip rates SGP2 .1 .02 3.2 SGP1 8 .2 3.97 6.09 8 .83 3.3 6.82 SGP8 8.06 39 6.00 .31 6.67 12.71 39 6.37 3.9 7.36 SGP 8.27 2 3.79 6.10 13.13 2 6.00 3.11 7.12 SGP9 8.6 39 .76 .0 6.2 13.82 39 6.3 .26 7.37 SGP 8.72 2 3.82 13.7 2 SGP3 8.7 2 .2 .7 6.19 12.9 2 6.18 3.3 7.1 SGP7 8.98 39 .73 6.9 13.9 39 6.02 2.89 7.21 SGP10 9.19 36 .97 .26 6.63 1.32 36 6.22 3.66 7.2 SGP6 9.0 39 .6 3.78 6.3 1.0 39 6.32 7.7 SGP12 12.38 2 8.88 8.91 7.87 20.9 2 9.69 8.7 9.32 SGP13 12.79 21 6.89 8.1 21.71 21 8.72 3.1 10.1 SGP1 13. 18 9.21 8.98 8.6 22.11 18 10.67 8.8 10.77 SGP11 16.86 30 6.28 .19 7.16 21.0 30 7.73 .0 9.0 SGP16 17.33 1 10.21 9.6 9.9 26.3 1 11.60 9. 12.11 SGP1 28.07 18 8.67 9.18 7.22 7.8 18 8.70 9.21 7.21

Notes: Table is sorted by postseismic-reduced GPS field χ2 misfit. Bolded terms are the two lowest values in each category. e present the χ2 misfit values for each GPS field and the normalied root mean suare RMS scores between our geodetically determined fault-slip rates and published available geologic slip rates for all geologic slip rates columns , 9, Holocene geologic slip rates only columns , 10, and Pleistocene geologic slip rates only columns 6, 11; lower RMS scores indicate geodetic rates that are closer to geologic rates within the sum of 1σ geologic and 1σ geodetic uncertainties. See tet for eplanation of calculation of the RMS residual terms, and χ2 misfit values.

data better than the best-fitting geometries to the GPS data. Lastly, for Late Pleistocene geologic slip-rate RMS scores, the lowest models are SGP2, SGP4, SGP1, and SGP5, and once again, the postseismic reduction decreases RMS TABE . F-TEST RESTS BETEEN MODES scores by up to 20%. Overall, the best-fitting model to postseismic-reduced Model name nreduced χ2 misfit value Number of Degrees GPS data and Pleistocene geologic slip rate data is SGP2, while the best-fitting reduced GPS field parameters of freedom model to Holocene-only geologic slip rates is SGP7, and the best-fitting model SGP2 2730.20 37 to overall geologic rates is SGP4 (Fig. 8). Model SGP6 is the only model that SGP1 276.7 8 3 is in the top four of every category, while SGP1 and SGP2 are in the top four SGP8 28.12 39 33 of every category except Holocene RMS scores. SGP 289. 2 30 SGP9 303.80 39 33 SGP1 1098. 18 37 ■■ F-test probability value DISCUSSION vs. SGP1 vs. SGP2 vs SGP6 SGP8 0.286000 0.0272000 - To evaluate the results of our block modeling, we examine two first-order SGP 0.009000 0.000120 - questions inherent in our investigation: (1) what can we learn by comparing our SGP9 0.000088 0.0000007 0.00033109 geodetic fault-slip rates to geological observations along the faults of San Gorgo- SGP1 0.0000000 0.0000000 0.00000000 nio Pass? And (2) how does the reduction of ongoing postseismic displacements Notes: An F-test probability indicates the etent to which the added affect estimated fault-slip rates across southern California? We explore what the parameters in the lowest misfit scenario are warranted, or whether the better fit is due to chance. A score closer to ero indicates the parameters are low misfit and low RMS score models tell us about active fault geometry and slip warranted, while a score closer to 1 would indicate the lower misfit would be distribution within the San Gorgonio Pass area and the possible implications of due to chance. the fits to geologic data for studies of earthquake recurrence and the determination of steady-state slip rates using geodetic and geologic data. In addition,

Lowest RMS Score to Holocene Rates SGP2 Lowest χ2 misfit, Lowest3.2 RMS Score0.4 to Pleistocene Rates SGP7 3.2 0.4 1.8 1.8 7.0 1.4 2.6 1.8 6.8 1.4 1.7 2.0 0.2 1.00.8 0.5 1.01.5 0.8 0.8 6.4 6.3 0.2 2.50.7 1.7 0.5 1.61.4 2.0 7.0 6.9 11.9 2.5 0.5 1.6 35.6 0.2 0.7 1.8 35.613.1 1.4 2.1 11.9 7.1 7.0 13.1 30.0 11.9 −3.9 0.2 2.5 30.0 −4.7 0.5 1.6 2.1 −0.2 0.7 1.8 13.1 −0.3 −4.2 1.3 −3.2 2.4 2.0 25.014.914.5 3.2 25.0 2.4 0.1 0.1 2.4 1.7 −0.2 14.5 4.5 1.5 1.9 12.8 6.5 −1.0 12.8 8.3 4.5 −0.9 6.3 0.1 2.4 8.2 4.6 3.0 12.4 0.2 −1.8 12.4 4.5 −1.8 9.7 1.6 −2.6 0.9 9.5 8.3 −3.6 0.1 19.56.4 −3.0−1.65.0 0.3 1.1 19.5 −3.0 4.5 −1.0 0.2 2.8 9.3 6.2 1.1 2.8 9.1 3.7 1.6 4.5 5.4 1.8 5.5 4.5 3.0 4.8 4.9 34˚N 15.6 34˚N 15.6 2.1 4.15.9 2.1 4.13.4 4.8 2.8 9.0 8.2 −1.66 −3.5 2.8 8.9 7.7 −1.66 15.2 −3.5 −3.5 −3.5 13.9 2.6 9.7 22.5 −0.9 2.7 9.6 22.5 1.7 2.6 15.4 2.6 9.3 15.9,15.514.8 9.5 Palm 15.9, Palm Springs 1.7 −0.9 5.6 14.9 5.6 2.6 9.6 Springs15.5 15.4 2.7 9.4 9.2,9.6 15.3 9.2,9.4 14.8 4.92.6 12.1 4.92.7 12.1 N 9.6 15.4 N 9.4 14.9 8.9 50 km 8.9 50 km −116˚W0 km 50 km −116˚W0 km 50 km

Lowest RMS Score to Overall Geologic Rates SGP4 3.2 0.4 1.8 Holocene Preferred or Median Geologic Slip Rates 6.8 1.4 1.4 1.9 1.0 0.8 6.3 0.2 Pleistocene Preferred or Median Geologic Slip Rates 1.3 1.6 1.9 6.8 1.3 35.611.7 0.2 1.6 1.9 11.7 6.9 30.0 −4.2 0.2 1.2 11.7 −0.5 1.5 1.9 −3.6 1.9 -25 -10 0 10 25 25.014.5 2.7 0.2 1.8 2.1 1.1 1.8 (-)Left lateral (+)Right lateral 12.8 1.7 1.7 −0.8 7.0 −2.9 mm/yr −2.3 −1.8 9.3 12.4 6.9 1.9 −4.5 −0.6 19.5 −3.0 5.2 −1.9 −0.4 3.0 9.0 1.2 4.5 4.3 4.7 5.6 15.6 34˚N 2.1 4.14.7 4.8 Figure 8. Our three best-fitting models across different categories of fitness, 2.9 8.7 9.8 −1.66 14.6 with their slip-rate estimates along all included fault strands; top left is 2.8 9.3 22.5 1.3 15.0 2 2.7 9.1 15.9, model SGP2, which has the lowest χ misfit to GPS data and the lowest Palm Springs 1.3 5.6 2.8 9.2 15.5 15.1 root mean square (RMS) score to the Late Pleistocene geologic slip rates. 9.2,9.2 15.0 At top right is model SGP7, which has the lowest RMS score to the Ho- 4.92.8 12.1 locene only geologic slip rates; at bottom left is model SGP4, which has N 9.2 15.1 8.9 the lowest RMS score to the Overall Geologic slip rates. 50 km −116˚W−116˚W0 km 50 km

we explore the extent to which our postseismic reductions, in conjunction with in this area; and (2) the geologic data are better fit by model geometries with accounting for off-fault deformation, can partially or fully resolve the known at least two fault strands. Geologic slip-rate observations within San Gorgo- geologic-geodetic slip-rate discrepancies in southern California. nio Pass consist of three main studies along the Banning–San Gorgonio Pass Thrust–San Bernardino fault strand (Orozco, 2004; Gold et al., 2015; Heermance and Yule, 2017) and three main regional slip-rate studies along the Mission– Comparing Geologic and Geodetic Rates within San Gorgonio Pass Mill Creek (MMC) (Morton and Matti, 1993; Kendrick et al., 2015; Fosdick and Blisniuk, 2018; see also Matti et al., 2019; Yule et al., 2019). Rates along the Strike-Slip Rates along the San Gorgonio Pass Thrust and Mission– Banning–San Gorgonio Pass Thrust–San Bernardino strand seem to be low Mill Creek Faults along the Banning fault to the southeast (2.3–5.9 mm/yr, but 3.9–4.9 mm/yr preferred, Gold et al., 2015), decreasing slightly to the northwest along the Two main observations come to light when we compare our estimated San Gorgonio Pass Thrust (approximate dextral motion of 0.6–3.5 mm/yr with geodetic slip rates with those produced from geologic investigations: (1) the an oblique motion of 4.2–8.4 mm/yr across two fault strands as measured GPS data overall seem to prefer a model with at least two SSAF fault strands by Heermance and Yule [2017]), and increasing farther toward the main San

Bernardino section as evidenced by a study by Orozco (2004). This study mea- our third lowest reduced χ2 misfit model, SGP8, has a geometry that does not sured 4–12 mm/yr of oblique slip and 2–9 mm/yr of dextral slip at Burro Flats include the MMC fault as active and therefore has geodetic rates of 0 mm/yr where the San Gorgonio Pass Thrust fault bends to meet the San Bernardino along its length (Fig. 9). This SGP8 model structure would agree well with strand of the SSAF. For the MMC fault strand, regional rate estimates differ geologic observations of Morton and Matti (1993) and Kendrick et al. (2015). significantly between as low as 0 mm/yr (inactive, no displacements over the On the other hand, our fourth lowest reduced χ2 misfit model, SGP5, exhibits late Quaternary period; Morton and Matti, 1993; Kendrick et al., 2015) and a slip rates along the MMC strand of 3.4–8.2 mm/yr, which implies a lot more maximum of 20–30 mm/yr within the Pleistocene (Fosdick and Blisniuk, 2018). slip being accommodated along this strand than the other two models. Yet, While all four of our lowest reduced χ2 misfit to GPS data models, SGP2, neither this model, nor any of our other models estimate MMC dextral slip SGP1, SGP8, and SGP5, yield geodetic rate estimates that are low enough to rates within the maximum range calculated by Fosdick and Blisniuk (2018). be within the limit of the dextral rate at Burro Flats (2–9 mm/yr), only model Of our four best-fitting models, only SGP1, SGP2, and SGP5 fault geome- SGP5 achieves low enough dextral rates to be within uncertainty of both the tries allow for the estimation of fault-slip rates along the San Gorgonio Pass 0.6–3.5 mm/yr or 2.3–5.9 mm/yr rate ranges at the sites to the southeast (Figs. 9 thrust that are low enough to just touch or to be within upper limits of geologic and 10). Our two lowest reduced χ2 misfit models, SGP2 and SGP1, have low rates within uncertainties. Model SGP8 produces rates along the San Gorgonio dextral rates estimated along the MMC fault (between 0–2.0 mm/yr), while Pass thrust of 6.5–7.1 mm/yr, which represents more dextral slip than current

Best Fitting Geodetic Models (lowest χ2 misfit) NW SE A. San Gorgonio Lowest χ2 misfit 40 Pass Knot SGP2 Best Pleistocene 30 Rate RMS score & third-best 20 Overall Geologic RMS score 10 Slip Rate (mm/yr)

50 km 0 km 50 km 0 −116˚ 50 100 150 200 Second Lowest B.40 SGP1 χ2 misfit Figure 9. Fault-slip rate comparison between model-calculated geodetic (plotted lines) and One of four best 30 observed geologic (plotted points and shaded RMS scores to background area) rates for the top three low- Overall Geologic 20 2 & Pleistocene est χ misfit models; slip rates plotted against distance along the plate boundary from the 10 Rates western edge of the model boundary (red line Slip Rate (mm/yr) 50 km 0 km 50 km on map plot): (A) SGP2, the lowest misfit model 0 −116˚W C.40 Third Lowest and best Pleistocene root mean square (RMS) SGP8 χ2 misfit score model; (B) SGP1, the second lowest misfit 30 model and one of the top four RMS scores to overall geologic and Pleistocene rates; (C) SGP8 20 the third lowest misfit.

Slip Rate (mm/yr) 50 km 0 0 km 50 km 50 100 150 200 −116˚W Distance from western model boundary along plate boundary (km) Holocene Geologic SAF Mojave - San Bernardino - SGP - Coachella Geodetic Rate (±2σ unc.) Slip Rate (±2σ unc. San Jacinto Geodetic Rate (±2σ unc.) or boxcar range) Pleistocene Geologic Mission Creek - Mill Creek Geodetic Rate (±2σ unc.) Burnt Mountain - Eureka Peak Geodetic Rate (±2σ unc.) Slip Rate (±2σ unc. Galena Peak Geodetic Rate (±2σ unc.) or boxcar range)

Best Fitting Models to Geologic Rate Observations NW SE A. San Gorgonio Best RMS 40 Pass Knot SGP4 scores to Overall Geologic 30 Rates & second- best RMS score to 20 Pleistocene Rates

Slip Rate (mm/yr) 10

50 km 0 km 50 km B. 0 Fourth Lowest 40 SGP5 χ2 misfit Figure 10. Fault-slip rate comparison between Top four best RMS 30 model-calculated geodetic (plotted lines) and scores to observed geologic (plotted points and shaded 20 Overall Geologic, Holocene, and background areas) rates for the models that best fit measured geologic rates overall and within 10 Pleistocene Rates Slip Rate (mm/yr) San Gorgonio Pass; slip rates plotted against

50 km 0 km 50 km 0 distance along the plate boundary from the C. western edge of the model boundary (red line 40 50 100 150 200 Lowest RMS score SGP7 to Holocene rates on map plot); (A) SGP4, which has the lowest 30 Closest to Overall Geologic root mean square (RMS) score Geologic Rates and is the second lowest RMS score to Pleis- 20 in SGP tocene rates; (B) SGP5, which has the fourth lowest χ2 misfit and is in the top four best-fitting RMS scores to Overall Geologic, Holocene, and

Slip Rate (mm/yr) 10 Pleistocene rates; (C) SGP7, which has the low- 50 km 0 0 km 50 km est Holocene geologic RMS score and has the D.40 SGP12 −116˚W Third Closest to best-fitting San Gorgonio Pass geodetic rates; Geologic Rates and (D) SGP12, which has a high geodetic χ2 30 in SGP misfit (poor fit) but whose geodetic rates match χ2 misfit to GPS: geologic rates second best in San Gorgonio Pass. 20 12.38 (poor fit) 10 Slip Rate (mm/yr) 50 km 0 0 km 50 km 50 100 150 200 Distance from western model boundary along plate boundary (km) Holocene Geologic Slip Rate (±2σ unc. SAF Mojave - San Bernardino - SGP - Coachella Geodetic Rate (±2σ unc.) or boxcar range) San Jacinto Geodetic Rate (±2σ unc.) Pleistocene Geologic Mission Creek - Mill Creek Geodetic Rate (±2σ unc.) Slip Rate (±2σ unc. Burnt Mountain - Eureka Peak Geodetic Rate (±2σ unc.) or boxcar range)

geologic observations would allow here (Fig. 9). The largest difference between multiple possible fault configurations in San Gorgonio Pass. Their two best-fitting model SGP8 and models SGP1 and SGP5 is the lack of the MMC strand running San Gorgonio Pass model geometries include one in which the Mill Creek fault is through San Gorgonio Pass. This observation indicates that in order to achieve removed (“Inactive Mill Creek” model) and is best approximated by our model sufficiently low dextral geodetic slip rates measured along the San Gorgonio SGP8 and one that includes a version of a west Mill Creek strand (“West Mill Pass thrust, strain needs to be partitioned across two faults, not just one. Creek” model), which is best approximated by our model SGP1. The model in This observation that estimated dextral slip rates are only low enough to which there is no Mill Creek fault produces dextral slip rates that are higher than match geologic data when two strands are included appears to be supported the other model and slightly higher than the geologic slip rate measurements by boundary element modeling work completed by Beyer et al. (2018); they test along the San Gorgonio Pass thrust and Banning faults (Beyer et al., 2018). The

same is seen in our model SGP8, in which estimated dextral fault-slip rates system. Further field investigations into the rugged and poorly accessible area are higher than geologic measurements there. Their second “West Mill Creek” are required to tease apart these different hypotheses of possible modern fault model, however, produces rates that match geologic rates along the San Gor- activity. However, for the moment, our geodetic slip rates support the idea of gonio Pass thrust and Banning faults; these rates do not match observations there being at least some level of activity on both known strands of this fault of zero late Quaternary motion along the MCC strand (Morton and Matti, 1993; system, though not at the high rate proposed by Fosdick and Blisniuk (2018). Kendrick et al., 2015). Our model SGP1, which closely approximates the “West Note on the San Jacinto fault. One fault that bears consideration when Mill Creek” model, produces rates that match all geologic slip-rate ranges along discussing slip being accommodated through San Gorgonio Pass is the San the San Gorgonio Pass thrust, except that of Heermance and Yule (2017). Our Jacinto fault, because it is argued to be the newly evolving optimal path of model, however, does not produce rates that match the MMC geologic obser- strain and slip accommodation to get around the San Gorgonio Pass Knot vations of Morton and Matti (1993) and Kendrick et al. (2015). Recent evidence (Janecke et al., 2010; Fattaruso et al., 2016). Across our 16 models, only two of possible activity along the Mission Creek and/or Galena Peak faults within produce rates higher than 10 mm/yr on the San Jacinto fault—SGP11 and San Gorgonio Pass (Beyer et al., 2018; Fosdick and Blisniuk, 2018; A. Morelan, SGP16—which are two of our highest reduced χ2 misfit models, indicating 2019, personal commun.) suggests that this “West Mill Creek” model may be very poor fits to modern GPS data. These two models are also highly unlikely better supported by geologic data than the “Inactive Mill Creek” model, a sug- geometries because neither model includes a SSAF San Bernardino strand, gestion that is underlined by our geodetic data and our best-fitting models here San Gorgonio Pass thrust, or Banning fault strand (all of which have Holocene (i.e., models SGP1, SGP2, and SGP5). Our second best-fitting model, SGP1, in geologic slip observations). For the rest of our models, we estimate fault-slip addition to producing rates that fit the geologic data fairly well along the San rates of ~9–10 mm/yr on the central San Jacinto, and rates increase as the fault Gorgonio Pass thrust and Banning faults, also produces rates along the MMC approaches the SSAF near Cajon Pass. These estimated rates imply that, with strand that are <2 mm/yr, showing a possible middle ground between two end- our modeling structure and chosen geometries, the San Jacinto is slipping member hypotheses of MMC activity, one in which both strands are active, but at a lower rate than the neighboring parallel SSAF Coachella strand, which the MMC is moving quite slowly in the present day. in the majority of our models (barring SGP11 and SGP16) is estimated to slip When looking at which of our 16 model geometries produce dextral geo- over a range of 12.9–17.5 mm/yr depending on model geometry. detic rates that best match geologic rates on the Banning-SGP fault in San One contributing factor to this debate is that, in our chosen geometries, Gorgonio Pass, SGP7, SGP12, and SGP5 jump to the top of the list because we use a SSAF Coachella strand that has a vertical dip from surface to depth. they produce the lowest rates of all the models even if they are not all the Recent geophysical observations (Fuis et al., 2017) and modeling efforts that best fits to geodetic data (Fig. 10). Each of these three models includes both a incorporate both GPS and InSAR data sets in this area (Lindsey and Fialko, San Gorgonio Pass thrust strand as well as a Mission–Mill Creek strand, indi- 2013; Fattaruso et al., 2014; Tymofyeyeva et al., 2019) indicate that the Coachella cating that the geologic slip-rate data are better fit by model geometries with strand has a more complicated geometry than previously assumed by earlier two strands, even when the geodetic data fit is not taken into account. SGP7 modeling efforts such as the UCERF3 model and the SCEC CFM models. In presents a model that has the closest dextral slip rates to geologic rates at San particular, these investigations suggest that this strand may be dipping from Gorgonio Pass with a slip rate of 3.0–3.8 mm/yr, which fits well in the bounds 50°–70° to the northeast. In order to assess what effect a dipping Coachella of 0.6–3.5 mm/yr calculated from Heermance and Yule (2017), the 2–9 mm/yr strand would have on our modeled slip rates, we adjust our best-fit SGP2 model rate at Burro Flats (Orozco, 2004), and the 2.3–5.9 rate at Whitewater (Gold et geometry to have a 60° dip from 5 km downward, to agree with evidence pre- al., 2015) (Fig. 10C). SGP5 has the next best-fitting slip rates in the pass (3.3– sented by Fuis et al. (2017) that this strand is steeply dipping near the surface 4.1 mm/yr geodetic, Fig. 10B) and is one of the four best-fitting RMS scores and is dipping between 50°–60° NE below 6–9 km depth. When this adjusted for overall geologic slip rates, Holocene-only geologic slip rates, and Late model is run, it decreases overall reduced χ2 misfit by only 0.03 and causes Pleistocene geologic slip rates. The third best-fitting model for San Gorgonio some estimated slip rates to change slightly, whereas rates along the San Pass rates is model SGP12, which produces slip rates of 1.4–2.2 mm/yr; these Jacinto fault rise slightly (by ~0.2 mm/yr), and rates along the Coachella strand rates overlap the range of the Heermance and Yule (2017) and Orozco (2004) remain the same. These changes, however, are within the 1σ uncertainties rate but fall short of the Gold et al. (2015) rate (Fig. 10D). In these three models, calculated by the model. If we adjust this same model to include a 60° dip all rates along the MMC strand are ~8.1 mm/yr near the intersection of the Pinto the way to the surface, running this inversion increases misfit to GPS data by Mountain fault with the MMC strand, where SGP7 produces 4.7 mm/yr, SGP5 0.07 and causes slip rates along the San Jacinto fault to increase by 0.6 mm/yr, produces 8.1 mm/yr, and SGP12 produces 6.6 mm/yr. Even though these mod- while decreasing slip rate along the Coachella strand by 0.1 mm/yr. In our els produce rates that fit well within geologic slip-rate bounds on the southern model fault structures, however, we only calculate the shear motion along a strand of the SSAF, they do not agree with direct field observations made by given fault strand, which is not as strongly affected by changes in fault dip as Morton and Matti (1993) and Kendrick et al. (2015); their observations indicate is the fault-perpendicular opening and/or closing component of deformation. that there is no late Quaternary slip on the northern MMC strand of the SSAF Ergo, changes in dip should only produce very small changes in strike-slip rate.

While adjusting the dip of the Coachella strand does produce slightly higher San Jacinto fault-slip rates, even with this adjustment, rates along the San SGP2 Lowest χ2 Misfit Model Dip-slip Rates

Jacinto fault in our models appear to be lower than those along the SSAF 0.9 0.3 −0.4−0.0 0.1 2.9 Coachella strand. We only constrain our estimated geodetic slip rates using 0.0 −0.3−0.2 −0.5 1.1 0.5 our postseismic-reduced GPS velocity data set, while recent investigators have −0.1 −0.0 −0.1 1.4 0.9 −1.1 noted that using both GPS data and InSAR data seems to point to a more equal −0.3 −0.2 0.9 −0.3 −0.1 3.9 2.1 0.4 −1.4 distribution of slip between the two systems (e.g., Lindsey and Fialko, 2013; 4.1 −0.4 −2.7 −1.3 −1.0 1.7 −0.8 −4.5 −1.9 −1.3 1.9 0.3 0.6 −1.3 Lindsey et al., 2014; Chaussard et al., 2016). Identifying which of these two −2.2 −1.5 0.8 −2.6 −0.7 1.0 −0.0 1.2 −0.3 3.4 0.4 faults is the main conductor of slip in this region could have a major impact −2.6 0.9 0.1 0.1 1.4 −0.3 −3.3 1.9 34˚N −4.6 −2.0−2.7 on how we evaluate the tectonic evolution and present-day seismic hazard of −0.4 3.5 −3.4 0.2 −2.1 0.1 −0.3 −0.3 this complicated area. For now, more investigation is required to ascertain the −0.7 −0.1 −0.6 −0.6 −1.8 −0.6 Palm −0.1 relative degree of significance of the SSAF and the San Jacinto fault in this area. 0.1 −0.1 −0.7 Springs −0.5 −2.0 N 0.4 −0.4 −2.2 50 km Dip-Slip Rates in San Gorgonio Pass −116˚W0 km 50 km SGP9 Dip-slip Rates Closest to Geologic Rates In order to best characterize slip rates within San Gorgonio Pass, we must also consider the behavior of convergent motion along this reverse-oblique 0.8 −0.1 −0.6 0.2 0.1 3.0 fault system. We estimate both fault-parallel (dextral or sinistral) and dip-slip −0.2 −0.5−0.3 −0.6 1.2 0.4 (convergent or divergent) rates within each of our 16 block models. Our model −0.1 −0.3 −0.4 1.4 1.0 −1.2 −0.5 −0.0 0.7 −0.6 dip-slip rates are calculated as the plane-parallel dip-slip motion, and not the −0.1 4.1 2.1 0.1 −1.5 4.4 3.9 −0.4 tensional fault plane–perpendicular motion. Yule and Sieh (2003) and Heer- −2.6 −1.1 1.5 −1.3 −4.6 −1.9 0.3 −1.0 2.3 −0.3 −1.9 2.2 mance and Yule (2017) have measured dip-slip rates within San Gorgonio Pass 0.6 −2.3 2.2 0.7 2.0 3.2 0.3 0.6 −0.3 −2.7 0.2 2.2 at Millard Canyon, along two fault strands that parallel each other there. Their −4.9 0.6 −0.4 −3.5 1.0 34˚N −4.6 −4.4 measurements yield convergent rates of >2.2 mm/yr (Yule and Sieh, 2003), −0.4 3.3 −2.7 −0.4 −3.2−0.1 0.6 −0.2 −0.6 −0.7 −0.2 −0.5 and an approximated 3.6–6.6 mm/yr, with a central rate of 4.6 mm/yr using all −0.6 −1.8 −0.4 Palm 0.0 parameters averaged with their reported ranges (data from Heermance and −0.1 −0.8 Springs 0.3 −0.5 −1.9 Yule, 2017; they report a net oblique slip rate and not a true dip-slip rate. Hence, N 0.4 −0.4 we calculate an approximate dip-slip rate from their provided supplemental −2.2 50 km materials). Our best-fitting model to GPS data, SGP2, shows estimated geodetic −116˚W0 km 50 km slip rates of 3.3 mm/yr at the same location of these geologic measurements (Fig. 11). However, if we examine all of our models, six produce geodetic slip rates within the range given by the approximate slip-rate calculation using −6 −4 −2 0 2 4 6 data from Heermance and Yule (2017) (Table 6). SGP9 is presented in Figure 11 (-)Convergence (+)Divergence mm/yr because it produces geodetic slip rates that are closest to both the Heermance and Yule (2017) approximate dip-slip rate along the San Gorgonio Pass thrust Figure 11. Dip-slip rates estimated from our postseismic-reduced GPS field and plotted along fault geometries for models SGP2 and SGP9; black star is as well as the dip-slip rate measured along the Cucamonga fault by Lindvall a Holocene dip-slip rate calculated from average values given in Heermance and Rubin (2006) at Day Canyon (Table 6). and Yule (2017) supplementary table 1; the gray star is a Pleistocene dip- Interestingly, these results seem to contradict what the dextral slip-rate slip rate from work by Lindvall and Rubin (2006); SGP2 is our best-fitting estimates tell us about the activity of the MMC strand to the north of the model to the GPS data, while model SGP9 produces the closest dip-slip rate to rates from geology. San Gorgonio thrust fault. While the San Gorgonio thrust seems to require another pathway for elastic strain in the form of the MMC fault strand, in order to keep dextral rates low enough to be within geologic observations, high enough convergent dip-slip rates along the San Gorgonio thrust in our the model that produces the closest dip-slip geodetic rates to geologic mea- model framework, only one SAF strand should be present. More direct field surements is a model that has no Mill Creek strand at all—SGP9 (Fig. 11). The study of both of these fault strands is required to elucidate the interacting next best-fitting models to the geologic dip-slip rates are models SGP10 and behavior between the two, as well as their ability to rupture either on their SGP8, neither of which includes a MMC strand. This would imply that to get own, or together simultaneously.

TABLE 6. COMPARISON OF MODEL DIP-SLIP RATES rates do not agree well with values from studies such as Johnson (2013) and WITH GEOLOGIC DIP-SLIP RATES Zeng and Shen (2016), in which they used some form of geologic slip-rate con- Model San Gorgonio Residual to Cucamonga Residual to Total summed straint, either as an assigned maximum or minimum (e.g., Johnson, 2013) or as number Pass thrust Heermance fault Lindvall and residual actual data observations in an inversion (e.g., Zeng and Shen, 2016). This brings convergent and Yule convergence Rubin (2006) (mm/yr) slip rate* (2017) slip slip rate§ slip rate# to the forefront a question of what it means for geodetically estimated slip rates (mm/yr) rate† (mm/yr) (mm/yr) to match (or not match) observed geologic fault-slip rates, when postseismic (mm/yr) deformation is removed from the GPS field. If high-precision GPS instruments SGP9 4.9 0.3 2.6 0.7 1.0 are measuring high strain rates following large-magnitude earthquakes, how SGP10 4.1 −0.5 2.5 0.6 1.1 do those high strain rates affect the faults around them? Based on block-style SGP8 4.1 −0.5 2.6 0.7 1.2 modeling, ongoing postseismic deformation can increase or decrease estimated SGP3 3.9 −0.7 2.7 0.8 1.5 fault-slip rate (see comparison of rates along the Helendale fault versus a fault Note: We compare values of convergent dip-slip rates calculated by four of our such as the San Bernardino strand in Table 7); yet in the long-term scenario, models with geologic observations along the San Gorgonio Pass Thrust (Heermance any increases in fault-slip rates due to “higher than normal” postseismic-caused and Yule, 2017) and the Cucamonga fault (Lindvall and Rubin, 2006). *Model convergent rate nearest to location of Heermance and Yule (2017) study. strain rates are ephemeral and eventually decay. Particularly with respect to †Convergent rate of ~4.6 ± 2.0. geologic time, these are minor blips in the relentless march of plate motion. § Convergent rate nearest to location of Lindvall and Rubin (2006) study. However, since temporary increased strain rate leads to increased overall accu- #Model Convergent rate of 1.9 ± 0.4 mm/yr. mulated strain, could these increased strain rates (and their subsequent effect on fault-slip rates) play a role in a several-hundred- year- scale earthquake cycle? We know that co-seismic changes in stress fields are the drivers of viscoelastic Effect of Postseismic Deformation on Fault-Slip Rates postseismic deformation, and that those stress changes can affect and even trigger faults around them to produce earthquakes (Stein et al., 1997; Pollitz In the Context of Previous Investigations et al., 2003). Yet, what if these stress change–caused displacements actually impact how fast these faults accumulate elastic strain over an appreciable frac- In order to set our results in the context of past geodetic block modeling tion of an earthquake recurrence interval? Does GPS measure increased fault work completed in southern California, we present Table 7, which shows our motion caused by viscoelastic displacements that can last up to decades? Most estimated geodetic fault-slip rates, averaged along their section length, for importantly, if faults are already close to failure, can postseismic motion result our lowest reduced χ2 misfit model (SGP2) (presented with the postseismic- in time-variable seismic hazard over shorter time scales (Oskin et al., 2008)? reduced GPS data set, observed GPS data set, and partially reduced GPS data While we do not have definitive answers to the many questions engen- set) alongside the work of several other authors (Becker et al., 2005; Meade dered by comparing long-term slip rates with short-term slip-rate estimates, and Hager, 2005; Spinler et al., 2010; Johnson, 2013; Liu et al., 2015; McGill we present below our take on how our data sets and modeling speak to the et al. 2015; Evans et al., 2016; Zeng and Shen, 2016). If an author had pre- geodetic versus geologic slip-rate debate in two specific cases: the faults of sented multiple models, we either chose their preferred “best-fit” model, or the ECSZ in the Mojave Desert and the Mojave segment of the SSAF. we chose the model that most closely matched our lowest reduced χ2 misfit model geometry. Rows that appear blank mean that a particular author’s fault geometry did not include that particular fault. Where slip rates are reported Eastern California Shear Zone without uncertainties, we either measured from a figure in the relevant published work, or the authors reported a range. In addition, at the bottom of the One of the most cited examples of a geologic versus geodetic fault-slip table, we present the summed geodetic slip rates from all authors across the rate discrepancy lies within the Mojave Desert just to the northeast of San Eastern California shear zone. Lastly, while GPS data comprise the primary Gorgonio Pass along the parallel northwest-trending, right-lateral faults of data set used by these regional models, recent local investigations including the ECSZ (Fig. 1) (Meade and Hager, 2005; Oskin et al., 2008; Evans et al., both InSAR and GPS data show promise for better quantifying strain rate and 2016). The summed Late Pleistocene geologic rates documented by Oskin et velocity gradients around faults in southern California, particularly faults that al. (2008) indicate motion of ~6.2 ± 1.9 mm/yr across the Helendale, Lenwood, are more widely spaced (Lindsey et al., 2014; Chaussard et al., 2016). Camp Rock–Emerson, Calico, Pisgah-Bullion, and Ludlow faults. This range When comparing across studies, in general, our estimated fault-slip rates can be extended slightly with recent work completed by Xie et al. (2018), to seem to match studies such as Becker et al. (2005), Meade and Hager (2005), include a slightly faster slip rate on the Calico fault of 3.2 ± 0.4 to show a Spinler et al. (2010), McGill et al. (2015), and Evans et al. (2016), wherein they used range of ~4.4–9.5 mm/yr across the whole zone. Previous geodetic estimates only GPS geodetic data to invert for slip rates. On the other hand, our fault-slip generally give a much larger slip-rate sum across the southern Mojave, as

TABLE 7. OUR ESTIMATED FAULT SLIP RATES COMPARED TO OTHER PUBLISHED RATES Faults Our lowest Our lowest Our lowest Meade and Becker et al. Spinler et al. Johnson McGill et Liu et al. Evans et al. Zeng and χ2 misfit model χ2 misfit model χ2 misfit model Hager (2005) (2005) (2010) (2013) al. (2015) (2015) (2016) Shen (2016)† SGP2 SGP2 (observed, SGP2 (partially Preferred CM model Elastic preferred Group A (postseismic- unreduced reduced B=1 medium: model BM4C* reduced GPS field) GPS field) GPS field) Fig. 10*† Strike Slip Strike Slip Strike Slip Strike Slip Strike Slip Strike Slip Strike Slip Strike Slip Strike Slip Strike Slip Strike Slip (mm/yr) (mm/yr) (mm/yr) (mm/yr) (mm/yr) (mm/yr) (mm/yr) (mm/yr) (mm/yr) (mm/yr) (mm/yr) SSAF Mojave (South) 11.9 ± 0.5 18.6 ± 0.5 17.1 ± 0.5 14.3 ± 1.2 15.7 ± 12 ~20 ~12.6 18 20.1 ± 0.7 SSAF San Bernardino 6.5 ± 0.3 7.4 ± 0.2 8.1 ± 0.2 5.1 ± 1.5 0.9 ± 12 ~12 6.5 ± 3.6 ~3.4 8 10.3 ± 0.9 SSAF San Gorgonio Pass 5.8 ± 0.6 4.4 ± 0.5 4.2 ± 0.6 ~3.4 6–8 9.4 ± 0.9 SSAF Mill Creek 0.1 ± 0.6 2.6 ± 0.6 3.5 ± 0.6 8.1 ± 0.2 ~3.4 1.7 ± 0.6 SSAF Mission Creek 3.3 ± 0.7 6.9 ± 0.6 7.5 ± 0.6 9.5 ± 0.4 ~17 ~8 SSAF Banning 8.3 ± 0.5 9.6 ± 0.4 9.5 ± 0.4 ~0 ~3.4 8–9 SSAF Coachella 15.4 ± 0.2 17.8 ± 0.2 17.9 ± 0.2 23.3 ± 0.5 22.9 ± 8 18.1 ± 0.3 ~15 ~20.7 17–22 19.8 ± 0.6 San Jacinto Central 9.5 ± 0.2 12.5 ± 0.2 12.1 ± 0.2 11.9 ± 1.2 14.5 ± 9 12.6 ± 0.3 ~15 14.1 ± 2.9 ~17.2 16 13.9 ± 1.0 avg (north) Elsinore 2.7 ± 0.2 2.5 ± 0.2 2.5 ± 0.2 2.7 ± 0.6 3.7 ± 6 ~2–3 2.7 ± 1.7 2.8 3.9 ± 0.3 avg Cucamonga 0.2 ± 0.3 −0.7 ± 0.3 −0.5 ± 0.3 −4.0 ± 1.3 ~0 ~2.3 −3–−7 1.7 ± 0.3 North Frontal (West) −2.7 ± 0.2 −2.3 ± 0.2 −3.0 ± 0.2 −0.8 ± 1.7 −5–−6 North Frontal (East) 2.9 ± 0.3 3.1 ± 0.3 2.0 ± 0.3 −0.8 ± 1.7 ~1 −2–−4

Eastern California Shear Zone Blue Cut −3.5 ± 0.2 −4.4 ± 0.2 −3.8 ± 0.2 −1.7 ± 0.4 −4.6 Pinto Mountain −2.6 to +0.9 −2.5 to +5.6 −2.6 to +3.1 −9.4 ± 0.9 −2.2–5.2 ± 0.3 ~−1 −2.3 to −3.4 −2–−5 3.9 ± 0.4 Burnt Mountain/Eureka Peak 5.4 ± 0.4 6.5 ± 0.4 5.4 ± 0.4 21.3 ± 1.6 8.6–9.4 ± 0.4 ~3.4 7–9 Johnson Valley/Camp Rock/Emerson 0.6 ± 0.3 2.6 ± 0.3 2.7 ± 0.3 1.9 ± 0.6 10.4 ± 0.2 ~4 ~4.6 0.5–4.7 1.2 ± 0.3 Helendale 6.9 ± 0.3 6.6 ± 0.3 5.8 ± 0.3 2.2 ± 1.2 ~2 1.9 ± 2.7 2–4 0.9 ± 0.2 Lenwood - - 0.4 ± 2.9 1.0 ± 0.3 Calico 2.5 ± 0.4 3.7 ± 0.5 4.3 ± 0.4 11.7 ± 2.6 ~4.6 7.6 2.5 ± 0.3 Pisgah-Bullion 0.7 ± 0.4 5.00 ± 0.4 2.9 ± 0.4 11.0 ± 0.3 ~4.6 0 1.2 ± 0.3 Ludlow/Cleghorn 1.8 ± 0.2 2.1 ± 0.2 2.1 ± 0.2 −2.5 ± 0.8 −4.6 ± 0.2 ~1 1.7 ± 1.4 ~5.7 1–3 0.5 ± 0.3 Eastern California Shear Zone Sum§ 12.0 18.9 16.8 15 13.9 16.8 ~7 15.7 ~19.5 17.6 ~7.8+ Note: First three columns provide our estimated rates from inversions of our three datasets. In the last row we compare summed geodetic slip rates across the ECSZ (see Fig. 12 for approximate location of summed cross section). SSAF—southern San Andreas fault. *Slip rates measured from figures rather than tables within above references. †These authors used some form of geologic slip-rate constraint in their modeling. §Geologic slip-rate range summed over the eastern California shear zone (ECSZ) in the Mojave is 4.4–9.5 mm/yr including uncertainties, measured by Oskin et al. (2008) and Xie et al. (2018).

evidenced in Table 7, where authors who did not utilize geologic observa- Our postseismic-reduced GPS velocity field, however, produces the lowest tions as constraints calculate summed rates of 13.9–19.5 mm/yr. Work done summed slip rate of 12.0 mm/yr (Fig. 12), which still lies above the Oskin et by Spinler (2014) showed that geodesy could estimate a summed fault-slip al. (2007) and Xie et al. (2018) combined range of 4.4–9.5 mm/yr, but is also rate of 7.0 ± 0.6 mm/yr across the Northern Mojave Desert just south of the smaller than the previous estimates from other studies. These comparisons Garlock fault. This rate is well within the range of geologic slip rate; however, indicate that our postseismic reduction has in fact made a measurable differ- the same has not yet been shown in the southern Mojave Desert. When we ence in the southern Mojave Desert, as might be expected given the recent compare our work for our lowest reduced χ2 misfit model, SGP2, we see large-magnitude earthquakes that have happened there over the past three that with our observed, unreduced GPS velocity field, we also calculate a decades (e.g., 1992 Landers and 1999 Hector Mine events). similar summed rate of ~18.9 mm/yr across the ECSZ (Fig. 12). When we use The geodetic-geologic slip-rate discrepancy in the ECSZ can perhaps be our partially postseismic-reduced field, that rate dips slightly to 16.8 mm/yr, explained, and indeed, even resolved, with the additional mechanism of per- which is still in agreement with other published summed slip rates (Table 7). manent off-fault deformation (Herbert et al., 2014; Milliner et al., 2016). This

A. SGP2, Postseismic-reduced GPS Field B. SGP2, Observed GPS Field

3.2 0.4 3.2 0.4 1.8 1.8 7.0 1.4 2.6 1.8 6.7 1.4 3.8 2.2 0.2 1.00.8 1.7 1.05.3 0.8 0.8 6.4 6.2 0.2 2.50.7 1.7 1.7 3.84.8 1.9 ECSZ 7.0 ECSZ 6.7 2.5 3.7 35.6 11.9 0.2 0.7 1.8 35.618.6 1.7 4.9 2.5 11.9 7.1 18.6 6.8 30.0 3.6 30.0 11.9 −3.9 0.2 2.5 1.8 −2.8 1.6 2.4 −0.2 −3.2 0.7 18.7 −0.5 −2.8 4.8 3.0 3.2 25.014.5 3.2 25.014.5 4.0 0.1 0.1 2.4 1.7 −0.8 2.7 3.4 1.6 12.8 6.5 0.1 12.8 7.4 −0.9 2.4 −1.8 2.4 6.3 12.46.3 0.2 −1.8 12.4 7.4 −1.8 9.7 0.1 −2.6 0.9 12.6 2.6 −2.5 4.6 19.56.4 −3.0−1.65.0 1.1 19.5 7.4 −3.0−1.46.0 5.4 2.8 9.3 1.1 2.6 12.4 1.0 4.5 5.4 5.5 34˚N 4.5 4.5 6.5 34˚N 15.6 1.9 15.6 2.1 4.6 5.1 2.1 4.15.9 4.14.6 2.8 9.0 8.2 −1.66 −3.5 2.6 11.6 9.1 −1.66 −4.4 15.2 −3.5 −3.5 −3.5 17.2 −4.4 −4.4 −4.4 2.6 9.7 22.5 −0.9 2.5 12.7 22.5 −0.3 17.8 2.6 9.5 15.9,15.515.4 2.5 12.5 Palm −0.9 Palm Springs15.9,15.5 −0.4 5.6 15.4 5.6 17.8 2.6 9.6 Springs 2.5 12.7 15.3 17.7 9.2,12.19.6 9.2,12.112.6 4.92.6 4.92.5 N N 9.6 15.4 12.6 17.9 50 km 8.9 8.9 −116˚W0 km 50 km 50 km −116˚W0 km 50 km

Summed ECSZ Geodetic Rate: 12.0 mm/yr Summed ECSZ Geodetic Rate: 18.9 mm/yr Removing Off-fault Deformation: 7.2 mm/yr Removing Off-fault Deformation: 11.3 mm/yr

Holocene Preferred or Median Geologic Slip Rates -25 -10 0 10 25 Pleistocene Preferred or Median Geologic Slip Rates (-)Left lateral (+)Right lateral mm/yr

Figure 12. Comparison of estimated fault-slip rates between the postseismic-reduced GPS velocity field (A) and the observed GPS velocity field (B) of our best-fit geodetic model SGP2. The biggest difference between these two model slip-rate data sets are (1) the majority of slip rates are lower using the postseismic-reduced velocity field, and (2) within the eastern California shear zone (ECSZ) (black arrows), fault-slip rates from the postseismic-reduced velocity field sum to a rate of ~12.0 mm/yr, while in the observed velocity field the sum across the ECSZ is ~18.9 mm/yr. When off-fault deformation is taken into account, these rates drop to 7.2 mm/yr and 11.3 mm/yr, respectively, indicating that our postseismic reductions in conjunction with off-fault deformation lead to agreement with known geologic rates in the ECSZ (4.4–9.5 mm/yr).

explanation suggests deformation that occurs near or in between distinct fault If we account for this off-fault deformation using our observed, unreduced zones is being measured by geodesy, but that this measured strain is mapped GPS velocity field or just our partially reduced GPS field, it leaves 11.3 mm/yr directly onto model fault planes because of the assumption of zero intra-block and 10.1 mm/yr as the respective summed ECSZ geodetic slip rate, neither of deformation in elastic fault-block modeling methods. This strain, however, is which are sufficiently low to reach the geologically observed slip-rate range. Thus, not measured by geologic slip-rate observations because it is accommodated accounting for both ongoing postseismic deformation, as well as off-fault defor- through other mechanisms around the fault system (e.g., pressure solution, mation, may be critical to estimated geodetic fault-slip rates in southern California. folding, fracturing, rigid body rotation; Shelef and Oskin, 2010). Herbert et al. When we turn our attention to our other block-model geometries, 4 of our (2014) use boundary element modeling methods to determine that off-fault 16 models produce summed ECSZ fault-slip rates that, even without account- deformation of this form can account for ~40% ± 23% of total strain in the ing for off-fault deformation, lie within the range of geologic measurements ECSZ region. While they include the entire ECSZ system from south of the (4.4–9.5 mm/yr). These include models SGP12, SGP13, SGP14, and SGP16, with Garlock fault to north of the Pinto Mountain fault, if we apply their findings to rates that span from the lowest sum at 3.9 mm/yr for SGP12 to 9.2 mm/yr with our best-fit geodetically summed rate across the southern ECSZ (12 mm/yr), SGP16. All four of these models share simplified ECSZ fault geometries, as well we find that accounting for an inferred 40% of deformation occurring off-fault as an absence of a Blue Cut fault within the Eastern Transverse Ranges. Models produces a summed slip rate of 7.2 ± 2.8 mm/yr. This rate falls directly within SGP12, SGP14, and SGP16 take the form of one fault carrying all the slip for the the bounds of the geologically measured summed slip rate across the ECSZ region, analogous to the hypothesis about a Landers-Mojave Earthquake Line and provides testimony for the importance of accounting for off-fault defor- (Nur et al., 1993), while SGP13 includes two faults. Yet, all four of these models mation in geodesy-based elastic fault-block modeling studies. have very poor misfits to the geodetic data, for all velocity fields. One other

feature that these particular models share is the fact that they all estimate larger observed, unreduced, GPS velocity field, is ~18.6 mm/yr, which agrees well slip rates along the Mojave segment of the San Andreas fault. This may imply with rates from earlier geodetic modeling. Yet, when we apply our full post- that strain measured in most of our models to map onto ECSZ faults can poten- seismic reduction, this rate decreases to 11.9 mm/yr for this model geometry. tially be mapped onto the San Andreas fault on the western side of the Mojave This decrease implies that long-term geodetic velocities are detecting a very Desert, thereby improving a second geologic-geodetic rate discrepancy there. slow rate along this section of the plate boundary, an observation that is in This may also suggest the possibility that there could be give and take between direct conflict with geologic measurements along the same section. In addition, these two systems, a concept that we explore in more detail in the next section. the postseismic reductions increase the gap here, rather than help to partially resolve the gap as they did in the ECSZ. Off-fault deformation, though lower along mature faults than immature faults (Titus et al., 2011; Herbert et al., 2014), Mojave and San Bernardino Sections of the SSAF also does little to solve this issue, since removing its effect from the current geodetic slip rate will only lower it, further widening the discrepancy here. Along the Mojave and San Bernardino sections of the SSAF, geodetic rate One limitation in our estimation of slip rates along this Mojave segment estimates from previous studies are systematically lower than their geologic in our model is that our fault segment is very close to the edge of our model counterparts, in contrast to the ECSZ, where reported geodetic slip rates have domain and therefore has a more limited amount of input data. This limitation hitherto been higher than geologic rate estimates. A number of detailed geo- can cause an effect wherein the uncertainties and noise levels coming from logic slip-rate studies have been completed along both the Mojave and San the GPS velocities nearby can affect estimated slip rates. In particular, our Bernardino strands through intensive paleoseismic trenching such as the study Mojave segment is quite short, and the block rotations on either side of it are at Wrightwood (Weldon et al., 2002, 2004), through detailed mapping and dat- constrained by 12 GPS station velocities in block WMOJ, but only 4 GPS station ing of offset surfaces (Weldon and Sieh, 1985; Harden and Matti, 1989; Matmon velocities in block SGAB to the southwest. The fewer stations in a block, the et al., 2005; McGill et al., 2013; Young et al., 2019; S. McGill, personal commun., higher the uncertainties are in its calculated rotation rate, which is applied to 2020), and through paleomagnetic investigation at Pallett Creek (Salyards et al., estimate fault-slip rates. In addition, choices surrounding locking depth and 1992). These geologic rates indicate a Holocene slip rate of ~30 mm/yr along fault dip can make a larger impact on final estimated fault-slip rates when the the southern Mojave section (Salyards et al., 1992; Weldon et al., 2002; Matmon data are sparse. In order to quantify this impact, we run two additional adjusted et al., 2005; Young et al., 2019, at X12 site); this rate transitions to ~25 mm/yr scenarios of our best-fit model SGP2: one in which the locking depth is set at at Cajon Pass (Weldon and Sieh, 1985). At Cajon Pass, the SSAF transitions a lower depth of 10 km, and one in which the locking depth is set at a deeper from the Mojave section to the San Bernardino section, which runs along the depth of 20 km. A change in five kilometers depth shallower along this short San Bernardino Mountains range front. In addition, this location is where the section of our model causes the estimated slip rates to decrease by 1.7 mm/yr San Jacinto fault originates and breaks off to run parallel to and slightly to the (in addition to raising the overall reduced χ2 misfit by 0.11). On the other hand, south and west of the San Bernardino strand. Geologic observations along the a change in five kilometers depth deeper than our chosen 15 km locking depth San Bernardino strand indicate a drop from the high slip rates in Cajon Pass results in raising the estimated slip rate by 1.2 mm/yr (and lowering the overall to rates between 7 and 18 mm/yr for four well-constrained estimates (McGill reduced χ2 misfit by 0.08). These changes with just a slight adjustment of locking et al., 2013; S. McGill, personal commun., 2020), with a couple of other sites highlight the point that assumed fault geometries and elastic parameters can having rates that overlap this range but with larger uncertainties (Harden have an effect on final estimated fault-slip rates, particularly near the edge of and Matti, 1989; S. McGill, personal commun., 2020, Matthews Ranch/Pitman the model domain where data become less encompassing. Canyon site). The drop in SSAF slip rate within a couple of km southeast of A similar story to the Mojave segment is represented along the San Ber- Cajon Pass presumably is due to sharing plate boundary motion and slip with nardino section, where most geologic observations show a rate in the teens the San Jacinto fault. (7–18 mm/yr), but rates estimated from geodetic data are in the single digits, Previously reported geodetic rate estimates for the southern Mojave sec- ~0.9–8 mm/yr, depending on the chosen model (Table 7). Our model with the tion are in the range of 14–18 mm/yr for models derived without geologic lowest reduced χ2 misfit to the full postseismic-reduced GPS field produces constraints (Table 7) (Becker et al., 2005; Meade and Hager, 2005; Liu et al., fault-slip rates of 6.5 mm/yr, which is another decrease that widens the gap 2015; Evans et al., 2016); these estimated rates comprise just over half of the more when compared to our unreduced observed GPS data set fault-slip rate measured geologic rate. When geologic slip-rate constraints are included of 7.4 mm/yr (Table 7). The rate produced with our unreduced, observed GPS (Johnson, 2013; Zeng and Shen, 2016), estimated rates rise only to ~20 mm/yr, field falls within the range of geologic slip rates provided from the literature which reaches the lower bound of some of the geologic rate uncertainties (7–18 mm/yr), but our postseismic-reduced field rate falls just under the low- (Weldon et al., 2002; Matmon et al., 2005; Young et al., 2019) but still lies est end of this geologic range, illustrating the persistence of this particular ~10 mm/yr below the preferred geologic rate in Pleistocene and Holocene time. geodetic-geologic discrepancy and that postseismic reductions increase this Our estimated fault-slip rate for our lowest reduced χ2 misfit model, using the discrepancy, rather than help resolve it.

Possible implications for alternating SSAF/ECSZ hypothesis? In attempting

to deduce possible explanations for this discrepancy between geodetic and 3.2 0.4 1.8 geologic rates along the Mojave and San Bernardino sections, we turn again to 1.4 Mojave Section 1.0 our 16 different model geometries. In only one model geometry do rates along 0.8

the Mojave section and San Bernardino sections match preferred geologic 33.6 35.6 rates of ~30 mm/yr and 7–18 mm/yr, respectively: model SGP15, one of our 33.5 San Gorgonio Pass 30.0 33.6 end-member scenarios (Fig. 13). This model geometry was included to test the 11.8 25.014.5 13.7 −2.2−0.8 12.8 11.7 13.7 feasibility of just the SSAF system accommodating plate motion (along with −2.2 11.4 13.9 12.4 11.6 −1.8 9.9 13.7−3.0 the known parallel fault systems to the west, including the San Jacinto and 2.3 19.5 −1.6 9.6 13.5 Elsinore faults). The main component that this model geometry lacks is active 2.3 4.5 −3.7 34˚N 15.6 2.1 4.1 2.3 −3.2 −1.95 faulting in the ECSZ. This is an unrealistic model for modern-day fault motion, 9.2 10.6 12.3 2.1 10.0 22.5 given that small- and large-magnitude earthquakes have been occurring in 2.1 Palm9.8 Springs15.9,15.512.8 the ECSZ over the past century. However, since our postseismic-reduced GPS 5.6 2.1 9.9 13.0 9.2,12.19.8 12.8 field approximates steady-state motion, it is feasible to test what slip rates 4.92.1 N 9.8 emerge when all elastic strain in our field is applied to an end-member SSAF 13.0 8.9 fault geometry. Geodetically estimated fault-slip rates for this model (with our 50 km −116˚W0 km 50 km postseismic-reduced GPS velocity field) lie between 33.5–33.6 mm/yr along the Mojave section and 11.6–11.8 mm/yr along the San Bernardino section, both Figure 13. Slip-rate map plot showing the along strike, strike-slip sense geodetic fault-slip rates for model SGP15 (slip-rate scale is same as Figs. 8 and 12). within the range of preferred geologic rates (Fig. 13). However, estimates along White transparent box illustrates the general location of San Gorgonio Pass. the San Gorgonio Pass thrust show low rates of left-lateral motion, which is This is the only fault geometry model to produce geodetic slip rates along the highly improbable, and likely contributes to the fact that this model has poorest Mojave section of the southern San Andreas fault (SSAF); these rates match the fit to the geodetic data of all 16 models. ~30 mm/yr measured geologically over most recent Holocene time (Salyards et The fact that our end-member SSAF scenario fits these two areas of al., 1992; Weldon, et al., 2002; Matmon et al., 2005; Young et al., 2019) and the Holocene and Late Pleistocene rates range along the San Bernardino section geologic-geodetic discrepancy presents an intriguing observation in support (McGill et al., 2010, 2013; S.F. McGill, personal commun., 2020); however it has of the hypothesis that southern California has been experiencing periodic the poorest fit to modern GPS observations. alternating levels of activity between the SSAF system and the ECSZ system (Dolan et al., 2007; Oskin et al., 2008). If our postseismic-reduced GPS velocity field approximates long-term plate boundary crustal velocities, we should be because this model does not have an active ECSZ system), but an alternative able to at least better match long-term geologic slip rates with our estimated model with an ECSZ fault zone included produces slip rates along the Mojave block motions, if we assume that fault-slip rates generally remain constant section of 19 mm/yr (model SGP14, Fig. S3N [footnote 1]) and lower (SGP13, over long periods of geologic time (Tapponnier et al., 2001). In the case that Fig. S3M; SGP3, Fig. S3C; SGP12, Fig. S3L) depending on which faults in the the removal of transient earthquake processes from GPS velocities does not ECSZ are included. resolve the discrepancy, we then turn to the possibility that perhaps these fault- However, with a model that removes the connecting SSAF through San slip rates may have changed through time. Dolan et al. (2007) discuss this in Gorgonio Pass and the San Bernardino Mountains, resulting fault-slip rates are the context of their hypothesis pointing out that geologic observations from quite unreasonable compared to geologic observations in the models we have the Wrightwood paleoseismic investigation (Weldon et al., 2002; Weldon et al., constructed, with estimated geodetic Coachella section rates of ~9.0 mm/yr 2004, see their figure 9) show time-averaged slip rates of 31 mm/yr over the (SGP11, Fig. S5K; SGP16, Fig. S6P). Perhaps the mechanism proposed by past 1500 years, but that the most recent slip rate over the past 1100 years is Dolan et al. (2007) is less an alternating fault system switch but rather the only 24 mm/yr, which is much closer to the estimated geodetic rate of Argus ECSZ sharing the load of the SSAF for periods of time, while the SSAF still et al. (2005) of 20 mm/yr and to the previous block modeling work estimated remains partially active. Alternatively, our data could simply be documenting rates of 14–18 mm/yr. This transition from a very large slip rate over a period deformation in a period of time in which the deformation field is transitioning of earthquake clustering along the SSAF (89 mm/yr between 600–900 A.D.; from one end member to another, rather than seeing a picture of one or the Weldon et al., 2004), to a slower rate over more recent time corresponds with other fault system being fully active. a change in active seismicity toward the ECSZ system (Rockwell et al., 2000b; To explain the switching behavior between the ECSZ and the SSAF system, Dolan et al., 2007). Turning back to our different fault geometries, this transition Dolan et al. (2007) propose that ECSZ activity could be modulated by fluctu- may be reflected in the fact that our end-member, SSAF only, scenario SGP15 ations in fault loading rate from deep ductile shear zones arising from cycles produces 34 mm/yr on the Mojave section (faster rates than other models, of strain hardening (during periods of increasing fault activity) and annealing

(during periods of decreasing fault activity). Postseismic deformation follow- faults of the Los Angeles area or the Garlock fault. If the former explanation is ing large-magnitude earthquakes may also contribute to cyclic behavior in applied, it could have important implications for our treatment of postseismic fault activity, and has been promoted as a possible self-loading mechanism motions as a transient process that has little bearing on earthquake recurrence. for clustered earthquake recurrence on individual faults (e.g., Kenner and If the latter explanation is correct, we would expect to find another portion of Simons, 2005; Michael, 2005; DiCaprio et al., 2008). However, the extent to the plate boundary system that is accommodating ~5–8 mm/yr of motion in which postseismic deformation generated by slip on one fault zone may affect excess of the geologic rates there. More investigation is required to further elastic loading and earthquake potential on other neighboring fault zones is understand how these systems may interact. currently poorly understood. The question is complicated further because postseismic deformation can last longer than the roughly decades-long time period between major earthquakes in the southern California region, such that ■■ CONCLUSIONS earthquakes on the San Andreas fault, Los Angeles basin faults, Garlock fault, and the ECSZ produce postseismic deformation fields that constructively and San Gorgonio Pass in southern California is one of the main loci of fault destructively interfere with one another over a very broad area (Fig. 4B; also complexity in the San Andreas plate boundary system, and characterizing see Guns and Bennett, 2020, their figure 9). In essence, the question is: could the slip rates of the separate fault strands in the area is critical to better delin- viscous flow caused by earthquake-induced stress changes alter the rate of eating seismic hazards for the region. We apply elastic fault-block modeling ductile shear in middle- to lower-crustal shear zones below mapped faults, to address this challenge, and to improve upon past results of block model- potentially enhancing or hindering accumulation of strain at the base of the ing, we incorporate a new postseismic-reduced GPS velocity field data set to locked zone of a given neighboring fault or even fault system? In the context assess its effect on estimated fault-slip rates. We find that multiple model fault of southern California, whether or not the cumulative postseismic effects geometries fit different data observations in different ways. Models that best fit from a series of earthquakes on the SSAF and Los Angeles Basin systems or geologic slip-rate observations within San Gorgonio Pass itself indicate that a Garlock fault could promote increased activity within the ECSZ or vice versa Mission–Mill Creek fault strand is needed to keep geodetically estimated dex- is unknown. Further research is needed to evaluate the possible role that tral fault-slip rates on the Banning-SGP fault low enough to match geologic postseismic deformation plays in the apparent mode switching between the observations. A key result is that accounting for postseismic deformation, in SSAF, ECSZ, and other faults in and around San Gorgonio Pass and elsewhere addition to accounting for off-fault deformation in the Mojave Desert, leads in southern California. to a summed geodetically estimated fault-slip rate of 7.2 ± 2.8 mm/yr, which is One last observation to take into consideration is that the sum of block- directly in line with known geologic observations from this area. On the other model slip rates across the Mojave section of the SAF and the Eastern California hand, we find that our postseismic reductions do not fill the known discrepancy shear zone of 28–31 mm/yr is smaller than the 36 mm/yr obtained by summing for the Mojave and San Bernardino sections of the SAF, and indeed only widen the geologic slip-rate estimates, even for the model that produces geodetic this gap between estimated geodetic rates and observed geologic rates. The rates matching observed geologic rates along the Mojave and San Bernardino implications of this lack of success indicate that this particular discrepancy sections of the SSAF. If the SSAF Mojave segment–ECSZ system is a closed might have a different source, and that our results might support the Dolan system, we might expect the total rate accommodated across the system to et al. (2007) fault system switching hypothesis and the idea that Mojave SAF be constant regardless of how slip is partitioned between the SSAF and the slip rates may have potentially changed through time and thereby may not be ECSZ over time. However, our best-fitting GPS model, SGP2, only seems to represented well by long-term steady-state geodetic velocities. In addition, we recover ~24 mm/yr across these systems (Fig. 12). Our best-fitting Mojave– suggest that short- and long-term viscoelastic postseismic motions may have San Bernardino model, SGP15, comes much closer because it attributes a role to play in explaining how shifts in fault-slip rate can happen through 34 mm/yr across those two systems (Fig. 13), though it is still just short of the time, and how our estimation of fault-slip rate can be affected. In summary, preferred geologic summed rate. However, when we examine our fault-slip our elastic fault-block modeling using a postseismic-reduced GPS velocity field rates estimated from the observed GPS velocity field (no postseismic reduc- introduces a promising new data set for fault modeling in California, and our tion performed), the rates across the Mojave Section and ECSZ amount to suite of 16 models has allowed us to examine regional fault activity and fault ~38 mm/yr, which matches the summed geologic rate quite well. This could interaction in this complex area. be an indication that accumulated postseismic strain contributes in some way to the earthquake cycle, or alternatively it could indicate that the SSAF Mojave segment–ECSZ system is not closed and accommodating a constant total rate ACKNOWLEDGMENTS through time together. That is, the SSAF Mojave segment–ECSZ system might We wish to thank reviewers Eric Lindsey and Jack Loveless, and Associate Editor Michele Cooke for their constructive comments and valuable feedback, which greatly strengthened this manu- itself trade off with other components of the plate boundary zone through script. We also want to thank all those from the University of Arizona who helped in the collection time, similar to what has been suggested by Dolan et al. (2007) with regard to of campaign GPS data from Joshua Tree National Park (thanks to all those who assisted Joshua

Spinler in collecting JOIGN campaign data in the early years, as well as Lisa Knowles, Phillip Blisniuk, K., Scharer, K.M., Sharp, W.D., Burgmann, R., Rymer, M.J., and Williams, P.L., 2013b, New McFarland, Clinton Koch, Carson Richardson, Audrey Dunham, Lauren Reeher, Ken Gourley, Ter- slip rate estimates for the Mission Creek strand of the San Andreas fault zone: San Francisco, rance Delisser, Wit Nantonoi, Sean Callahan, Lauren Ward, Tommy Yong, Maria Snyder, Brooke California, American Geophysical Union 2013 Fall Meeting, Abstract T42A-01. Elser, and Gunnar Speth for assisting in years 2016–2018). We also thank all of the many students Bos, M.S., Bastos, L., and Fernandes, R.M.S., 2010, The influence of seasonal signals on the esti- who assisted with collecting campaign data in the San Bernardino Mountains. In addition, we mation of the tectonic motion in short continuous GPS time-series: Journal of Geodynamics, thank Dr. Jay Theuer and Luke Sabala from the National Park Service at Joshua Tree National v. 49, p. 205–209, https://doi.org/10.1016/j.jog.2009.10.005. Park for their support and assistance in permitting our research in the park. All campaign data Broermann, J., 2017, Appendix B: Time-independent and time-varying surface velocity field for are archived at UNAVCO and are available for download under the campaign names “San Ber- the Colorado Plateau and adjacent Basin and Range inferred from viscoelastic modeling, in nardino Mountains” or “Joshua Tree.” We prepared many of our figures using GMT software Alignment of post-Atlantic-rifting Volcanic Features on the Guinea Plateau, West Africa, and (Wessel et al., 2013). We thank University NAVSTAR Consortium (UNAVCO) and Plate Boundary Present-Day Deformation in the Southwest United States from GPS Geodesy [Ph.D. disser- Observatory facilities for providing the GPS observation equipment. The collection of campaign tation]: Tucson, University of Arizona, 218 p. GPS data in 2017 and 2018 was partially funded by the Southern California Earthquake Center Bürgmann, R., and Thatcher, W., 2013, Space geodesy: A revolution in crustal deformation mea- (Award #17161) as well as by two Geological Society of America Graduate Student Research surements of tectonic processes, in Bickford, M.E., ed., The Web of Geological Sciences: Grant awards (the ExxonMobil Specialized Award and the John T. and Carol G. McGill Special- Advances, Impacts, and Interactions: Geological Society of America Special Paper 500, p ized Award) to K. Guns. We also received support from the U.S. Geological Survey Earthquake 1–34, https://doi.org/10.1130/2013.2500(12). Hazards Program (grant G19AP00054). In addition, we are grateful for the support for this work Cadena, A.M., 2013, Paleoseismologic evidence for Holocene activity on the Pinto Mountain Fault, provided to K. Guns by the University of Arizona Department of Geosciences Dr. H. Wesley Peirce Twentynine Palms, California [M.S. thesis]: Ellensburg, Central Washington University, 54 p. and Maxine W. Peirce Scholarship. Carter, J.N., Luyendyk, B.P., and Terres, R.R., 1987, Neogene clockwise tectonic rotation of the eastern Transverse Ranges, California, suggested by paleomagnetic vectors: Geological Society of Amer- ica Bulletin, v. 98, p. 199–206, https://doi .org /10 .1130 /0016 -7606 (1987)98 <199: NCTROT>2 .0 .CO;2 . Castillo, B.A., McGill, S.F., Scharer, K.M., Yule, D., McPhillips, D., McNeil, J.C., Saha, S., Brown, REFERENCES CITED N.D., and Moon, S., 2019, Ages of Prehistoric Earthquakes on the Banning Strand of the San Anderson, K.B., and Conder, J.A., 2011, Discussion of Multicycle Hubbert modeling as a method Andreas Fault, Near North Palm Springs, California: Poster Presentation at 2019 Southern for forecasting future petroleum production: Energy & Fuels, v. 25, p. 1578–1584, https://doi California Earthquake Center Annual Meeting, Poster #134. .org/10.1021/ef1012648. Chaussard, E., Johnson, C.W., Fattahi, H., and Bürgmann, R., 2016, Potential and limits of InSAR to Argus, D.F., Heflin, M.B., Peltzer, G., Crampé, F., and Webb, F.H., 2005, Interseismic strain accu- characterize interseismic deformation independently of GPS data: Applications to the south- mulation and anthropogenic motion in metropolitan Los Angeles: Journal of Geophysical ern San Andreas Fault system: Geochemistry, Geophysics, Geosystems, v. 17, p. 1214–1229, Research, v. 110, B04401, https://doi.org/10.1029/2003JB002934. https://doi.org/10.1002/2015GC006246. Argus, D.F., Gordon, R.G., Heflin, M.B., Ma, C., Eanes, R.J., Willis, P., Peltier, W.R., and Owen, Chuang, R.Y., and Johnson, K.M., 2011, Reconciling geologic and geodetic model fault slip-rate S.E., 2010, The angular velocity of the plates and the velocity of Earth’s centre from space discrepancies in Southern California: Consideration of nonsteady mantle flow and lower geodesy: Geophysical Journal International, v. 180, p. 913–960, https://doi.org/10.1111/j.1365 crustal fault creep: Geology, v. 39, p. 627–630, https://doi.org/10.1130/G32120.1. -246X.2009.04463.x. d’Alessio, M.A., Johanson, I.A., and Bürgmann, R., 2005, Slicing up the San Francisco Bay Area: Becker, T.W., Hardebeck, J.L., and Anderson, G., 2005, Constraints on fault slip rates of the south- Block kinematics and fault slip rates from GPS-derived surface velocities: Journal of Geo- ern California plate boundary from GPS velocity and stress inversions: Geophysical Journal physical Research, v. 110, B06403, https://doi.org/10.1029/2004JB003496. International, v. 160, p. 634–650, https://doi.org/10.1111/j.1365-246X.2004.02528.x. Dawson, T.E., and Weldon, R.J., II, 2013, Appendix B: Geologic Slip-Rate Data and Geologic Defor- Behr, W.M., Rood, D.H., Fletcher, K.E., Guzman, N., Finkel, R., Hanks, T.C., Hudnut, K.W., Kendrick, mation Model, in Uniform California earthquake rupture forecast, version 3 (UCERF3)—The K.J., Platt, J.P., Sharp, W.D., Weldon, R.J., and Yule, J.D., 2010, Uncertainties in slip-rate esti- time-independent model: U.S. Geological Survey Open-File Report 2013-1165, California mates for the Mission Creek strand of the southern San Andreas fault at Biskra Palms Oasis, Geological Survey Special Report 228, and Southern California Earthquake Center Publication southern California: Geological Society of America Bulletin, v. 122, p. 1360–1377, https://doi 1792, 97 p., http://pubs.usgs.gov/of/2013/1165/. .org/10.1130/B30020.1. DiCaprio, C.J., Simons, M., Kenner, S.J., and Williams, C.A., 2008, Post-seismic reloading and Bennett, R.A., 2008, Instantaneous deformation from continuous GPS: Contributions from quasi- temporal clustering on a single fault: Geophysical Journal International, v. 172, p. 581–592, periodic loads: Geophysical Journal International, v. 174, p. 1052–1064, https://doi.org/10.1111 https://doi.org/10.1111/j.1365-246X.2007.03622.x. /j.1365-246X.2008.03846.x. Dolan, J.F., Bowman, D.D., and Sammis, C.G., 2007, Long-range and long-term fault interactions Bennett, R.A., Rodi, W., and Reilinger, R.E., 1996, GPS constraints on fault slip rates in southern in Southern California: Geology, v. 35, p. 855–858, https://doi.org/10.1130/G23789A.1. California and northern Baja, Mexico: Journal of Geophysical Research, v. 101, p. 21,943–21,960, Dong, D., Fang, P., Bock, Y., Cheng, M.K., and Miyazaki, S., 2002, Anatomy of apparent seasonal https://doi.org/10.1029/96JB02488. variations from GPS-derived site position time series: Journal of Geophysical Research, v. 107, Bennett, R.A., Friedrich, A.M., and Furlong, K.P., 2004, Codependent histories of the San Andreas https://doi.org/10.1029/2001JB000573. and San Jacinto fault zones from inversion of fault displacement rates: Geology, v. 32, no. 11, Douilly, R., Oglesby, D.D., Cooke, M.L., and Hatch, J.L., 2020, Dynamic models of earthquake p. 961–964, https://doi.org/10.1130/G20806.1. rupture along branch faults of the eastern San Gorgonio Pass Region in California using Beyer, J., Cooke, M.L., and Marshall, S.T., 2018, Sensitivity of deformation to activity along the complex fault structure: Geosphere, v. 16, p. 474–489, https://doi.org/10.1130/GES02192.1. Mill Creek and Mission Creek strands of the southern San Andreas fault: Geosphere, v. 14, Elliott, J.L., Larsen, C.F., Freymueller, J.T., and Motyka, R.J., 2010, Tectonic block motion and glacial no. 6, p. 2296–2310, https://doi.org/10.1130/GES01666.1. isostatic adjustment in southeast Alaska and adjacent Canada constrained by GPS measure- Blewitt, G., and Lavallée, D., 2002, Effect of annual signals on geodetic velocity: Journal of Geo- ments: Journal of Geophysical Research, v. 115, B09407, https://doi .org /10 .1029 /2009JB007139 . physical Research, v. 107, https://doi.org/10.1029/2001JB000570. Elósegui, P., Davis, J.L., Mitrovica, J.X., Bennett, R.A., and Wernicke, B.P., 2003, Crustal load- Blisniuk, K., Rockwell, T., Owen, L.A., Oskin, M., Lippincott, C., Caffee, M.W., and Dortch, J., 2010, ing near Great Salt Lake, Utah: Geophysical Research Letters, v. 30, https://doi.org/10.1029 Late Quaternary slip rate gradient defined using high-resolution topography and 10Be dating /2002GL016579. of offset landforms on the southern San Jacinto Fault zone, California: Journal of Geophysical Evans, E.L., Loveless, J.P., and Meade, B.J., 2015, Total variation regularization of geodetically and Research, v. 115, B08401, https://doi.org/10.1029/2009JB006346. geologically constrained block models for the Western United States: Geophysical Journal Blisniuk, K., Oskin, M., Mériaux, A.S., Rockwell, T., Finkel, R.C., and Ryerson, F.J., 2013a, Stable, International, v. 202, p. 713–727, https://doi.org/10.1093/gji/ggv164. rapid rate of slip since inception of the San Jacinto fault, California: Geophysical Research Evans, E.L., Thatcher, W.R., Pollitz, F.F., and Murray, J.R., 2016, Persistent slip rate discrepancies in the Letters, v. 40, p. 4209–4213, https://doi.org/10.1002/grl.50819. eastern California (USA) shear zone: Geology, v. 44, p. 691–694, https://doi .org /10 .1130 /G37967 .1 .

Fattaruso, L.A., Cooke, M.L., and Dorsey, R.J., 2014, Sensitivity of uplift patterns to dip of the San Herring, T.A., King, R.W., Floyd, M.A., and McClusky, S.C., 2018, Introduction to GAMIT/GLOBK, Andreas fault in the Coachella Valley, California: Geosphere, v. 10, p. 1235–1246, https://doi Release 10.7: GAMIT/GLOBK Documentation, http://geoweb.mit.edu/gg/docs.php. .org/10.1130/GES01050.1. Janecke, S.U., Dorsey, R.J., Forand, D., Steely, A.N., Kirby, S.M., Lutz, A.T., Housen, B.A., Bel- Fattaruso, L.A., Cooke, M.L., Dorsey, R.J., and Housen, B.A., 2016, Response of deformation garde, B., Langenheim, V.E., and Rittenour, T.M., 2010, High Geologic Slip Rates since Early patterns to reorganization of the southern San Andreas fault system since ca. 1.5 Ma: Tec- Pleistocene Initiation of the San Jacinto and San Felipe Fault Zones in the San Andreas Fault tonophysics, v. 693, p. 474–488, https://doi.org/10.1016/j.tecto.2016.05.035. System: Southern California, USA: Geological Society of America Special Paper 275, 48 p., Fialko, Y., 2006, Interseismic strain accumulation and the earthquake potential on the southern https://doi.org/10.1130/2010.2475. San Andreas fault system: Nature, v. 441, https://doi.org/10.1038/nature04797. Johnson, K.M., 2013, Slip rates and off-fault deformation in Southern California inferred from Field, E.H., and 2014 Working Group on California Earthquake Probabilities, 2015, UCERF3: A new GPS data and models: Journal of Geophysical Research, Solid Earth, v. 118, p. 5643–5664, earthquake forecast for California’s complex fault system: U.S. Geological Survey Open-File https://doi.org/10.1002/jgrb.50365. Report 2015-3009, 6 p. Jones, L.M., Bernknopf, R., Cox, D., Goltz, J., Hudnut, K., Mileti, D., Perry, S., Ponti, D., Porter, K., Fosdick, J.C., and Blisniuk, K., 2018, Sedimentary signals of recent faulting along an old strand Reichle, M., Seligson, H., Shoaf, K., Treiman, J., and Wein, A., 2008, The ShakeOut Scenario: of the San Andreas Fault, USA: Scientific Reports, v. 8, p. 12,132, https://doi.org/10.1038 U.S. Geological Survey Open-File Report 2008-1150 and California Geological Society Pre- /s41598-018-30622-3. liminary Report 25, 312 p. Freed, A.M., Ali, S.T., and Bürgmann, R., 2007, Evolution of stress in Southern California for the Kendrick, K.J., Morton, D.M., Wells, S.G., and Simpson, R.W., 2002, Spatial and Temporal Defor- past 200 years from coseismic, postseismic, and interseismic stress changes: Geophysical mation along the Northern San Jacinto Fault, Southern California: Implications for Slip Rates: Journal International, v. 169, p. 1164–1179, https://doi.org/10.1111/j.1365-246X.2007.03391.x. Bulletin of the Seismological Society of America, v. 92, no. 7, p. 2782–2802, https://doi.org Fuis, G.S., Bauer, K., Goldman, M.R., Ryberg, T., Langenheim, V.E., Scheirer, D.S., Rymer, M.J., /10.1785/0120000615. Stock, J.M., Hole, J.A., Catchings, R.D., Graves, R.W., and Aagaard, B., 2017, Subsurface Kendrick, K.J., Matti, J.C., and Mahan, S.A., 2015, Late Quaternary slip history of the Mill Creek Geometry of the San Andreas Fault in Southern California: Results from the Salton Seismic strand of the San Andreas fault in San Gorgonio Pass, southern California: The role of a Imaging Project (SSIP) and Strong Ground Motion Expectations: Bulletin of the Seismological subsidiary left-lateral fault in strand switching: Geological Society of America Bulletin, v. 127, Society of America, v. 107, https://doi.org/10.1785/0120160309. no. 5/6, p. 825–849, https://doi.org/10.1130/B31101.1. Fumal, T.E., Rymer, M.J., and Seitz, G.G., 2002, Timing of large earthquakes since A.D. 800 on the Kenner, S.J., and Simons, M., 2005, Temporal clustering of major earthquakes along individual Mission Creek Strand of the San Andreas Fault Zone at Thousand Palms Oasis, near Palm faults due to post-seismic reloading: Geophysical Journal International, v. 160, p. 179–194, Springs, California: Bulletin of the Seismological Society of America, v. 92, no. 7, p. 2841–2860, https://doi.org/10.1111/j.1365-246X.2005.02460.x. https://doi.org/10.1785/0120000609. Laske, G., Masters, G., Ma, Z., and Pasyanos, M., 2013, Update on CRUST1.0 - A 1-degree Global Gabriel, K., 2017, Late Pleistocene slip rate for the Western Pinto Mountain Fault, Morongo Valley, Model of Earth’s Crust: Geophysical Research Abstracts, p. 15, Abstract EGU2013–EGU2658. Southern California [M.S. thesis]: Northridge, California State University, 64 p. Lindsey, E.O., and Fialko, Y., 2013, Geodetic slip rates in the southern San Andreas Fault system: Gold, P.O., Behr, W.M., Rood, D., Kendrick, K., Rockwell, T.K., and Sharp, W.D., 2015, Holocene Effects of elastic heterogeneity and fault geometry: Journal of Geophysical Research, Solid geologic slip rate for the Banning strand of the southern San Andreas Fault, southern Cal- Earth, v. 118, no. 2, p. 689–697, https://doi.org/10.1029/2012JB009358. ifornia: Journal of Geophysical Research. Solid Earth, v. 120, p. 5639–5663, https://doi.org Lindsey, E.O., Sahakian, V.J., Fialko, Y., Bock, Y., Barbot, S., and Rockwell, T., 2014, Interseismic /10.1002/2015JB012004. Strain Localization in the San Jacinto Fault Zone: Pure and Applied Geophysics, v. 171, p. 2937– Guns, K.A., 2020, Appendix C: Evaluating the late Quaternary record of fault slip in the Eastern 2954, https://doi.org/10.1007/s00024-013-0753-z. Transverse Ranges, CA: New time-variable geologic slip rates along the Blue Cut Fault, in Lindvall, S.C., and Rubin, C.M., 2006, Slip rate studies along the Sierra Madre–Cucamonga Fault Exploring Long-term Earthquake Deformation in Southern California using Field-based Geologic System using geomorphic and cosmogenic surface exposure age constraints: Collaborative and Space-based Geodetic Methods [Ph.D. dissertation]: Tucson, University of Arizona, 306 p. Research with Central Washington University and William Lettis & Associates, Inc., U.S. Guns, K.A., and Bennett, R.A., 2020, Assessing long-term postseismic transients from GPS time Geological Survey National Earthquake Hazard Reduction Program Final Report for External series in Southern California: Journal of Geophysical Research. Solid Earth, v. 125, no. 4, Grant Award Number 03HQGR0084, https://earthquake.usgs.gov/cfusion/external_grants https://doi.org/10.1029/2019JB018670. /reports/03HQGR0047.pdf (accessed 19 January 2020). Harden, J.W., and Matti, J.C., 1989, Holocene and late Pleistocene slip rates on the San Andreas Liu, S., Shen, Z., and Bürgmann, R., 2015, Recovery of secular deformation field of Mojave Shear fault in Yucaipa, California, using displaced alluvial-fan deposits and soil chronology: Geo- Zone in Southern California from historical terrestrial and GPS measurements: Journal of logical Society of America Bulletin, v. 101, p. 1107–1117, https://doi.org/10.1130/0016-7606 Geophysical Research, Solid Earth, v. 120, p. 3965–3990, https://doi .org /10 .1002 /2015JB011941 . (1989)101<1107:HALPSR>2.3.CO;2. Lozos, J.C., Oglesby, D.D., Brune, J.N., and Olsen, K.B., 2015, Rupture propagation and ground Hauksson, E., 2000, Crustal structure and seismicity distribution adjacent to the Pacific and North motion of strike-slip stepovers with intermediate fault segments: Bulletin of the Seismological America plate boundary in southern California: Journal of Geophysical Research, v. 105, no. B6, Society of America, v. 105, no. 1, p. 387–399, https://doi.org/10.1785/0120140114. p. 13,875–13,903, https://doi.org/10.1029/2000JB900016. Lundgren, P., Hetland, E.A., Liu, Z., and Fielding, E., 2009, Southern San Andreas–San Jacinto Hearn, E.H., 2019, Kinematics of southern California crustal deformation: Insights from finite-el- fault system slip rates estimated from earthquake cycle models constrained by GPS and ement models: Tectonophysics, v. 758, p. 12–28, https://doi.org/10.1016/j.tecto.2019.02.016. interferometric synthetic aperture radar observations: Journal of Geophysical Research, v. 114, Hearn, E.H., Onishi, C.T., Pollitz, F.F., and Thatcher, W.R., 2013, How do “ghost transients” from no. B02403, https://doi.org/10.1029/2008JB005996. past earthquakes affect GPS slip rate estimates on southern California faults?: Geochemistry, Matmon, A., Schwartz, D.P., Finkel, R., Clemmens, S., and Hanks, T., 2005, Dating offset fans along Geophysics, Geosystems, v. 14, https://doi.org/10.1002/ggge.20080. the Mojave section of the San Andreas fault using cosmogenic 26Al and 10Be: Geological Society Heermance, R.V., and Yule, D., 2017, Holocene slip rates along the San Andreas Fault System in the of America Bulletin, v. 117, no. 5/6, p. 795–807, https://doi.org/10.1130/B25590.1. San Gorgonio Pass and implications for large earthquakes in southern California: Geophysical Matti, J.C., and Morton, D.M., 1993, Paleogeographic evolution of the San Andreas fault in southern Research Letters, v. 44, p. 5391–5400, https://doi.org/10.1002/2017GL072612. California: A reconstruction based on a new cross-fault correlation, in Powell, R.E., Weldon, Herbert, J.W., Cooke, M.L., Oskin, M., and Difo, O., 2014, How much can off-fault deformation R.J., and Matti, J.C., eds., The San Andreas Fault System: Displacement, Palinspastic Recon- contribute to the slip rate discrepancy within the eastern California shear zone?: Geology, struction, and Geologic Evolution: Geological Society of America Memoir 178, https://doi v. 42, p. 71–75, https://doi.org/10.1130/G34738.1. .org/10.1130/MEM178-p107. Herring, T.A., Melbourne, T.I., Murray, M.H., Floyd, M.A., Szeliga, W.M., King, R.W., Phillips, D.A., Matti, J.C., Kendrick, K.J., Yule, D., and Heermance, R.V., 2019, Is the Mission Creek Fault in the Puskas, C.M., Santillan, M., and Wang, L., 2016, Plate Boundary Observatory and related San Gorgonio Pass region of southern California a long-abandoned strand of the San Andreas networks: GPS data analysis methods and geodetic products: Reviews of Geophysics, v. 54, Fault?: Or is it a major player in the San Andreas Fault’s late Quaternary strain budget?: Palm p. 759–808, https://doi.org/10.1002/2016RG000529. Springs, California, 2019 Southern California Earthquake Center Annual Meeting, Poster #133.

McCaffrey, R., 2005, Block kinematics of the Pacific–North America plate boundary in the south- Oskin, M., Perg, L., Blumentritt, D., Mukhopadhyay, S., and Iriondo, A., 2007, Slip rate of the Calico western United States from inversion of GPS, seismological, and geodetic data: Journal of fault: Implications for geologic versus geodetic rate discrepancy in the Eastern California Shear Geophysical Research, v. 110, B07401, https://doi.org/10.1029/2004JB003307. Zone: Journal of Geophysical Research, v. 112, B03402, https://doi .org /10 .1029 /2006JB004451 . McCaffrey, R., Qamar, A.I., King, R.W., Wells, R., Khazaradze, G., Williams, C.A., Stevens, C.W., Oskin, M., Perg, L., Shelef, E., Strane, M., Gurney, E., Singer, B., and Zhang, X., 2008, Elevated Vollick, J.J., and Zwick, P.C., 2007, Fault locking, block rotation and crustal deformation in the shear zone loading rate during an earthquake cluster in eastern California: Geology, v. 36, Pacific Northwest: Geophysical Journal International, v. 169, p. 1315–1340, https://doi.org/10 no. 6, p. 507–510, https://doi.org/10.1130/G24814A.1. .1111/j.1365-246X.2007.03371.x. Plesch, A., Shaw, J.H., Bryant, W.A., Carena, S., Cooke, M.L., Dolan, J.F., Fuis, G.S., Gath, E.M., McGill, S., and Sieh, K., 1993, Holocene slip rate of the Central Garlock Fault in southeastern Grant Ludwig, L.B., Hauksson, E., Jordan, T.H., Kamerling, M.J., Legg, M.R., Lindvall, S.C., Searles Valley, California: Journal of Geophysical Research, v. 98, no.B8, p. 14,217–14,231, Magistrale, H., Nicholson, C., Niemi, N.A., Oskin, M.E., Perry, S.C., Planansky, G., Rockwell, https://doi.org/10.1029/93JB00442. T.K., Shearer, P.M., Sorlien, C.C., Suess, M., Suppe, J., Treiman, J.A., and Yeats, R.S., 2007, McGill, S., Owen, L., Kent, E., Rockwell, T., Kendrick, K., Onderdonk, N., and Rhodes, N., 2012, Slip Community Fault Model (CFM) for Southern California: Bulletin of the Seismological Society rate of the Northern San Jacinto Fault from offset landslides in the San Timoteo Badlands: of America, v. 97, no. 6, p. 1793–1802, https://doi.org/10.1785/0120050211. Seismological Research Letters, v. 83, no. 2, p. 436. Pollitz, F.F., Peltzer, G., and Bürgmann, R., 2000, Mobility of continental mantle: Evidence from McGill, S., Owen, L., Weldon, R.J., and Kendrick, K., 2013, Latest Pleistocene and Holocene slip postseismic geodetic observations following the 1992 Landers earthquake: Journal of Geo- rate for the San Bernardino strand of the San Andreas Fault, Plunge Creek, Southern Cali- physical Research, v. 105, p. 8035–8054, https://doi.org/10.1029/1999JB900380. fornia: Implications for strain partitioning within the southern San Andreas Fault system for Pollitz, F., Vergnolle, M., and Calais, E., 2003, Fault interaction and stress triggering of twentieth the last ~35 k.y: Geological Society of America Bulletin, v. 125, no. 1/2, p. 48–72, https://doi century earthquakes in Mongolia: Journal of Geophysical Research, v. 108, https://doi.org .org/10.1130/B30647.1. /10.1029/2002JB002375. McGill, S.F., Spinler, J.C., McGill, J.D., Bennett, R.A., Floyd, M.A., Fryxell, J.E., and Funning, G.J., Powell, R.E., 1993, Balanced palinspastic reconstruction of pre-late Cenozoic paleogeology, south- 2015, Kinematic modeling of fault slip rates using new geodetic velocities from a transect ern California: Geologic and kinematic constraints on evolution of the San Andreas fault across the Pacific–North America plate boundary through the San Bernardino Mountains, system, in Powell, R.E., Weldon, R.J., and Matti, J.C., eds., The San Andreas Fault System: California: Journal of Geophysical Research, Solid Earth, v. 120, p. 2772–2793, https://doi Displacement, Palinspastic Reconstruction, and Geologic Evolution: Boulder, Colorado, Geo- .org/10.1002/2014JB011459. logical Society of America Memoir 178, https://doi.org/10.1130/MEM178-p1. Meade, B.J., and Hager, B.H., 2005, Block models of crustal motion in Southern California con- Richards-Dinger, K.B., and Shearer, P.M., 2000, Earthquake locations in southern California obtained strained by GPS measurements: Journal of Geophysical Research, v. 110, B03403, https://doi using source-specific station terms: Journal of Geophysical Research, v. 105, no. B5, p. 10,939– .org/10.1029/2004JB003209. 10,960, https://doi.org/10.1029/2000JB900014. Michael, A.J., 2005, Viscoelasticity, postseismic slip, fault interactions, and the recurrence of Rockwell, T., Loughman, C., and Merifield, P., 1990, Late Quaternary rate of slip along the San large earthquakes: Bulletin of the Seismological Society of America, v. 95, no. 5, p. 1594–1603, Jacinto Fault Zone near Anza, Southern California: Journal of Geophysical Research, v. 95, https://doi.org/10.1785/0120030208. p. 8593–8605, https://doi.org/10.1029/JB095iB06p08593. Milliner, C.W.D., Dolan, J.F., Hollingsworth, J., Leprince, S., and Ayoub, F., 2016, Comparison of Rockwell, T., Bergmann, M., and Kenny, M., 2000a, Holocene slip rate of the Elsinore fault in coseismic near-field and off-fault surface deformation patterns of the 1992 Mw 7.3 Landers Temecula Valley, Riverside County, California, in Birnbaum, B., and Cato, K., eds., Geology and 1999 Mw 7.1 Hector Mine earthquakes: Implications for controls on the distribution of and Enology of the Temecula Valley: San Diego, California, San Diego Association of Geol- surface strain: Geophysical Research Letters, v. 43, no. 19, p. 10,115–10,124, https://doi.org ogists, p. 105–118. /10.1002/2016GL069841. Rockwell, T., Lindvall, S., Herzberg, M., Murbach, D., Dawson, T., and Berger, G., 2000b, Paleo- Millman, D.E., and Rockwell, T.K., 1986, Neotectonics of the Elsinore fault in Temescal Valley, Cal- seismology of the Johnson Valley, Kickapoo, and Homestead Valley Faults: Clustering of ifornia, in Neotectonics and faulting in Southern California: Geological Society of America, Earthquakes in the Eastern California Shear Zone: Bulletin of the Seismological Society of Cordilleran Section Annual Meeting Field Trip Guidebook, p. 159–166. America, v. 90, no. 5, p. 1200–1236, https://doi.org/10.1785/0119990023. Morton, D.M., and Matti, J.C., 1993, Extension and contraction within an evolving divergent Rockwell, T.K., Seitz, G., Dawson, T., and Young, J., 2006, The long record of San Jacinto fault paleo- strike-slip fault complex: The San Andreas and San Jacinto fault zones at their convergence earthquakes at Hog Lake: Implications for regional patterns of strain release in the southern in southern California, in Powell, R.E., Weldon, R.J., II, and Matti, J.C., eds., The San Andreas San Andreas fault system: Seismological Research Letters, v. 77, no. 2, p. 270. Fault System: Displacement, Palinspastic Reconstruction, and Geologic Evolution: Geological Ross, Z.E., Hauksson, E., and Ben-Zion, Y., 2017, Abundant off-fault seismicity and orthogonal struc- Society of America Memoir 178, https://doi.org/10.1130/MEM178-p217. tures in the San Jacinto fault zone: Science Advances, v. 3, https://doi .org /10 .1126 /sciadv .1601946 . Nicholson, C., Plesch, A., and Shaw, J.H., 2017, Community Fault Model Version 5.2: Updating & Salyards, S.L., Sieh, K.E., and Kirschvink, J.L., 1992, Paleomagnetic measurement of nonbrittle expanding the CFM 3D fault set and its associated fault database: Palm Springs, California, coseismic deformation across the San Andreas Fault at Pallett Creek: Journal of Geophysical 2017 Southern California Earthquake Center Annual Meeting, Contribution 7735. Research, v. 97, p. 12457–12470, https://doi.org/10.1029/92JB00194. Nur, A., Ron, H., and Beroza, G.C., 1993, The Nature of the Landers-Mojave Earthquake Line: Savage, J.C., and Burford, R.O., 1973, Geodetic determination of relative plate motion in central Science, v. 261, p. 201–203, https://doi.org/10.1126/science.261.5118.201. California: Journal of Geophysical Research, v. 78, no. 5, p. 832–845, https://doi.org/10.1029 Okada, Y., 1985, Surface deformation due to shear and tensile faults in a half-space: Bulletin of /JB078i005p00832. the Seismological Society of America, v. 75, no. 4, p. 1135–1154. Scharer, K., and Streig, A., 2019, The San Andreas Fault system: Complexities along a major Okada, Y., 1992, Internal deformation due to shear and tensile faults in a half-space: Bulletin of transform fault system and relation to earthquake hazards, in Duarte, J.C., ed., Transform the Seismological Society of America, v. 82, no. 2, p. 1018–1040. Plate Boundaries and Fracture Zone: Elsevier, p. 249–269, https://doi.org/10.1016/B978-0-12 Olsen, K.B., Day, S.M., Minster, J.B., Cui, Y., Chourasia, A., Faerman, M., Moore, R., Maechling, P., -812064-4.00010-4. and Jordan, T., 2006, Strong shaking in Los Angeles expected from southern San Andreas earth- Shelef, E., and Oskin, M., 2010, Deformation processes adjacent to active faults: Examples from quake: Geophysical Research Letters, v. 33, p. L07305, https://doi.org/10.1029/2005GL025472. eastern California: Journal of Geophysical Research, v. 115, B05308, https://doi.org/10.1029 Onderdonk, N.W., McGill, S.F., and Rockwell, T.K., 2015, Short-term variations in slip rate and size /2009JB006289. of prehistoric earthquakes during the past 2000 years on the northern San Jacinto fault zone, Shen, Z.-K., King, R.W., Agnew, D.C., Wang, M., Herring, T.A., Dong, D., and Fang, P., 2011, A uni- a major plate boundary structure in southern California: Lithosphere, v. 7, no. 3, p. 211–234, fied analysis of crustal motion in Southern California, 1970–2004: The SCEC crustal motion https://doi.org/10.1130/L393.1. map: Journal of Geophysical Research, v. 116, B11402, https://doi.org/10.1029/2011JB008549. Orozco, A.A., 2004, Offset of a mid-holocene alluvial fan near Banning, CA: Constraints on the Smith, B., and Sandwell, D., 2004, A three-dimensional semianalytic viscoelastic model for time- slip rate of the San Bernardino strand of the San Andreas fault [M.S. thesis]: California State dependent analyses of the earthquake cycle: Journal of Geophysical Research, v. 109, B12401, University, Northridge, 56 p. https://doi.org/10.1029/2004JB003185.

Spinler, J.C., 2014, Appendix C: Present-day loading rate of faults in the northern Mojave and Wang, Q., Jackson, D.D., and Kagan, Y.Y., 2009, California earthquakes, 1800–2007: A unified cata- southern Walker Lane region, California: From GPS geodesy, in Investigating Crustal Defor- log with moment magnitudes, uncertainties, and focal mechanisms: Seismological Research mation Associated with the North America-Pacific Plate Boundary in Southern California with Letters, v. 80, no. 3, p. 446–457, https://doi.org/10.1785/gssrl.80.3.446. GPS Geodesy [Ph.D. dissertation]: Tucson, University of Arizona, 214 p. Wang, R., Lorenzo-Martín, F., and Roth, F., 2006, PSGRN/PSCMP—A new code for calculating Spinler, J.C., Bennett, R.A., Anderson, M.L., McGill, S.F., Hreinsdóttir, S., and McCallister, A., 2010, co- and post-seismic deformation, geoid and gravity changes based on the viscoelastic- Present-day strain accumulation and slip rates associated with southern San Andreas and gravitational dislocation theory: Computers & Geosciences, v. 32, p. 527–541, https://doi.org eastern California shear zone faults: Journal of Geophysical Research, v. 115, B11407, https:// /10.1016/j.cageo.2005.08.006. doi.org/10.1029/2010JB007424. Wei, S., Fielding, E., Leprince, S., Sladen, A., Avouac, J., Helmberger, D., Hauksson, E., Chu, R., Spinler, J.C., Bennett, R.A., Walls, C., Lawrence, S., and González Garcia, J.J., 2015, Assessing long- Simons, M., Hudnut, K., Herring, T., and Briggs, R., 2011, Superficial simplicity of the 2010 El term postseismic deformation following the M7.2 4 April 2010, El Mayor–Cucapah earthquake Mayor Cucapah earthquake of Baja California in Mexico: Nature Geoscience, v. 4, p. 615–618, with implications for lithospheric rheology in the Salton Trough: Journal of Geophysical https://doi.org/10.1038/ngeo1213. Research, v. 120, p. 3664–3679, https://doi.org/10.1002/2014JB011613. Weldon, R.J., II, and Sieh, K.E., 1985, Holocene rate of slip and tentative recurrence interval for Stein, R.S., Barka, A.A., and Dieterich, J.H., 1997, Progressive failure on the North Anatolian fault large earthquakes on the San Andreas fault, Cajon Pass, southern California: Geological since 1939 by earthquake stress triggering: Geophysics Journal International, v. 128, p. 594–604, Society of America Bulletin, v. 96, p. 793–812, https://doi.org/10.1130/0016-7606(1985)96<793: https://doi.org/10.1111/j.1365-246X.1997.tb05321.x. HROSAT>2.0.CO;2. Sykes, L.R., and Seeber, L., 1985, Great earthquakes and great asperities, San Andreas fault, southern Weldon, R.J., II, Fumal, T.E., Powers, T.J., Pezzopane, S.K., Scharer, K.M., and Hamilton, J.C., 2002, California: Geology, v. 13, p. 835–838, https://doi .org /10 .1130 /0091 -7613 (1985)13 <835: GEAGAS>2 Structure and earthquake offsets on the San Andreas Fault at the Wrightwood, California, .0.CO;2 . paleoseismic site: Bulletin of the Seismological Society of America, v. 92, no. 7, p. 2704–2725, Tapponnier, P., Ryerson, F.J., Van der Woerd, J., Mériaux, A.-S., and Lasserre, C., 2001, Long- https://doi.org/10.1785/0120000612. term slip rates and characteristic slip: Keys to active fault behavior and earthquake hazard: Weldon, R., Scharer, K., Fumal, T., and Biasi, G., 2004, Wrightwood and the earthquake cycle: What Comptes Rendus de l’Académie des Sciences, Series IIA, Earth and Planetary Science, v. 333, a long recurrence record tells us about how faults work: GSA Today, v. 14, no. 9, https://doi p. 483–494, https://doi.org/10.1016/S1251-8050(01)01668-8. .org/10.1130/1052-5173(2004)014<4:WATECW>2.0.CO;2. Thatcher, W., Savage, J.C., and Simpson, R.W., 2016, The Eastern California Shear Zone as the Wesnousky, S.G., 2008, Displacement and geometrical characteristics of earthquake surface northward extension of the southern San Andreas Fault: Journal of Geophysical Research, ruptures: Issues and implications for seismic-hazard analysis and the process of earthquake Solid Earth, v. 121, p. 2904–2914, https://doi.org/10.1002/2015JB012678. rupture: Bulletin of the Seismological Society of America, v. 98, no. 4, p. 1609–1632, https:// Titus, S.J., Dyson, M., DeMets, C., Tikoff, B., Rolandone, F., and Bürgmann, R., 2011, Geologic doi.org/10.1785/0120070111. versus geodetic deformation adjacent to the San Andreas fault, central California: Geolog- Wessel, P., Smith, W.H.F., Scharroo, R., Luis, J., and Wobbe, F., 2013, Generic mapping tools: ical Society of America Bulletin, v. 123, no. 5–6, p. 794–820, https://doi.org/10.1130/B30150.1. Improved version released: Washington, D.C., Eos (Transactions, American Geophysical Tymofyeyeva, E., Fialko, Y., Jiang, J., Xu, X., Sandwell, D., Bilham, R., Rockwell, T.K., Blanton, C., Union), v. 94, no. 45, p. 409–410, https://doi.org/10.1002/2013EO450001. Burkett, F., Gontz, A., and Moafipoor, S., 2019, Slow slip event on the southern San Andreas Xie, S., Gallant, E., Wetmore, P.H., Figueiredo, P.M., Owen, L.A., Rasmussen, C., Malservi, R., and Fault triggered by the 2017 M w 8.2 Chiapas (Mexico) earthquake: Journal of Geophysical Dixon, T.H., 2018, A new geological slip rate estimate for the Calico Fault, eastern California: Research, Solid Earth, v. 124, no. 9, p. 9956–9975, https://doi.org/10.1029/2018JB016765. Implications for geodetic versus geologic rate estimates in the Eastern California Shear Zone: U.S. Geological Survey and California Geological Survey, 2006, Quaternary fault and fold database International Geology Review, https://doi.org/10.1080/00206814.2018.1531272. for the United States: http//earthquake.usgs.gov/hazards/qfaults/ (accessed 28 October 2019). Young, E.K., Scharer, K.M., Keen-Zebert, A., Anderson-Merritt, E., Cowgill, E.S., and Weldon, R.J., van der Woerd, J., Klinger, Y., Sieh, K., Tapponnier, P., Ryerson, F.J., and Mériaux, A., 2006, Long- 2019, New interpretations at the Ranch Center and Key Slide slip rate sites, Mojave section of term slip rate of the southern San Andreas Fault from 10Be-26Al surface exposure dating of the San Andreas fault: Palm Springs, California, 2019 Southern California Earthquake Center an offset alluvial fan: Journal of Geophysical Research, Solid Earth, v. 111, B04407, https://doi Annual Meeting, Poster #130. .org/10.1029/2004JB003559. Yule, D., and Sieh, K., 2003, Complexities of the San Andreas fault near San Gorgonio Pass: Vergnolle, M., Pollitz, F., and Calais, E., 2003, Constraints on the viscosity of the continental crust Implications for large earthquakes: Journal of Geophysical Research, v. 108, p. 2548, https:// and mantle from GPS measurements and postseismic deformation models in western Mon- doi.org/10.1029/2001JB000451. golia: Journal of Geophysical Research, v. 108, 2502, https://doi.org/10.1029/2002JB002374. Yule, D., Matti, J.C., Kendrick, K.J., and Heermance, R.V., 2019, Evidence for inactivity since ~100 ka Waco, J., and Blisniuk, K., 2019, Geologic map from Mission Creek Preserve to Whitewater on the northern route of the San Andreas fault, southern California: Palm Springs, California, River, San Bernardino and Riverside counties, California: Evidence for an active and evolv- 2019 Southern California Earthquake Center Annual Meeting, Poster #132. ing left-stepping San Andreas Fault (Mission Creek Strand) in the eastern San Bernardino Zeng, Y., and Shen, Z.-K., 2016, A fault-based model for crustal deformation, fault slip rates, and Mountains: Geological Society of America Abstracts with Programs, v. 51, no. 5, https://doi off-fault strain rate in California: Bulletin of the Seismological Society of America, v. 106, no. 2, .org/10.1130/abs/2019AM-337624. p. 766–784, https://doi.org/10.1785/0120140250.