Time‐Invariant Late Quaternary Slip Rates Along the Agua Blanca Fault, Northern Baja California, Mexico

Fault slip rates inform models of strain accumulation and release, which over geologic time may vary or remain constant depending on factors like structural complexity, fault strength, deformation rates, and proximity to other faults. In this study, we present a Late Pleistocene–Holocene slip history based on four new geologic slip rates for the Agua Blanca Fault (ABF), which transfers Pacific‐North American dextral plate boundary motion across the Peninsular Ranges of northern Baja California. Time‐averaged slip rates from three sites are 2.8 + 0.8/−0.6 mm/a since ~65.1 ka, 3.0 + 1.4/−0.8 mm/a since ~21.8 ka, 3.2 + 1.0/−0.6 mm/a since ~12.5 ka, and 3.5 + 5.1/−2.0 mm/a since ~1.4 ka; however, the actual slip rate may be closer to 4 mm/a when off‐fault slip and age interpretation uncertainties are considered. Significantly, although the ABF has more in common in terms of length, net offset, and slip rate with known variable slip rate faults, the most straightforward age and offset interpretations for the ABF suggest constant slip rates over ~10 kyr time scales. As with other constant slip rate faults, comparable neighboring faults that might modulate the ABF slip rate are absent, suggesting that fault interaction, or lack thereof, may be a more significant factor controlling fault behavior on this and potentially other faults. The new rates indicate that the ABF accommodates at least half of total slip across the Peninsular Ranges, clarifying strain partitioning for seismic forecasting models that previously lacked modern geologic slip rate constraints for this domain of the plate boundary.


Introduction
Earthquakes occur when locked faults surpass a critical elastic strain threshold after a period of loading by tectonic plate motion. The cycle of gradual strain accumulation and abrupt release as first described by Reid's (1910) elastic rebound theory remains the conceptual foundation for models of fault mechanics and earthquake recurrence on which seismic hazard forecasts rely (Field et al., 2014(Field et al., , 2015. Although regularity in earthquake occurrence and the assumption that post-rupture reset should significantly lower the probability of another earthquake are known oversimplifications of elastic rebound theory, the classic characteristic, slip-predictable, and time-predictable models for earthquake recurrence incorporate this assumption by restricting variability such that fault slip rates over two to three earthquake cycles approximate the long-term average (Shimazaki & Nakata, 1980). However, a growing number of paleoseismic and paleoslip (slip rate) studies have measured slip rate variations of several orders of magnitude lasting multiple millennia (Dolan et al., 2016;R. D. Gold et al., 2017;Goldfinger et al., 2013;Ninis et al., 2013;Onderdonk et al., 2015;Wechsler et al., 2018;Weldon et al., 2004;Zinke et al., 2017Zinke et al., , 2019. Identifying the factors that cause strain to be released periodically or in punctuated bursts is critical for understanding fault mechanics and requires reconstructing slip histories from the geologic record for a variety of faults. Worldwide, faults are characterized by a diverse range of parameters that may influence earthquake recurrence. First-order factors include fault length (e.g., Mouslopoulou et al., 2009), structural complexity (e.g., Berryman et al., 2012;Fletcher et al., 2020), net offset (e.g., Wesnousky, 1988Wesnousky, , 1990, slip rate (e.g., Anderson et al., 1996), fault loading rates (e.g., Chery & Vernant, 2006), connectivity with or proximity to subparallel active structures (e.g., Bennett et al., 2004;Berryman et al., 2012;Dolan et al., 2007; R. D. Kirby et al., 2006;Onderdonk et al., 2015Onderdonk et al., , 2018, proximity to ephemeral surface loads (e.g., lakes, glaciers) (e.g., Hetzel & Hampel, 2005), variations in fault strength (e.g., Chery & Vernant, 2006;Dolan et al., 2007;Oskin et al., 2008), and orientation with respect to principal stress directions (e.g., Fletcher et al., 2020). Varying these parameters is likely to result in a wide range of fault behaviors, so a practical goal is to determine whether different categories of faults that share several common characteristics are more or less likely to exhibit constant or variable slip over geologic time scales. Realizing this goal will require substantially more data to reconstruct slip histories from a diverse range of faults than are presently available. In particular, because competition between coseismic surface deformation and erosion disproportionately obscures the near surface earthquake record along slower faults (<5 mm/a) that may rupture less frequently, concerted efforts to construct slip histories for such faults are essential for a more complete view of earthquake recurrence.
In this paper, we report a 65-kyr paleoslip history for the Agua Blanca Fault (ABF), which transfers dextral slip across the Peninsular Ranges of northern Baja California, Mexico ( Figure 1) (Allen et al., 1960;Gastil et al., 1975Gastil et al., , 1981. Superficially, the ABF is structurally simple, and clear, well-preserved tectonic geomorphology records a long-term history of dextral displacement (Allen et al., 1960;Hatch, 1987;Schug, 1987), making the ABF an ideal candidate for characterizing earthquake behavior on a slow slip-rate fault. Proximity to population centers in southern California and northern Baja California, as well as connectivity to faults that parallel the Pacific coast further incentivizes study of recent faulting along the ABF in order to evaluate its contribution to earthquake hazard. We review the seismotectonics of the region before describing geochronologic strategies, site tectonic-geomorphologic histories, and displacement, age, and slip rate interpretations. We evaluate the strengths and limitations of the new slip rates, compare the new rates to previous estimates, reassess regional strain partitioning and slip transfer across the Peninsular Ranges, and finally, discuss factors that may control slip variability in strike slip faults, and the evidence for and against long-term slip rate variations on the ABF.

Agua Blanca Fault Geologic Background
Roughly 50 mm/a of NW-SE directed Pacific-North America plate margin shearing (DeMets et al., 2010;Kreemer et al., 2014) is partitioned on to a regionally distributed network of active faults across southern California and northern Baja California. A NE-SW transect across northern Baja California crosses several NW-SE oriented faults or fault zones that accommodate dextral plate motion at this latitude (Figure 1). At the eastern boundary of the plate margin, the Cerro Prieto Fault, which is the southern extension of the San Andreas and San Jacinto Faults, has a present day (from GPS) slip rate of~40 mm/a (Bennett et al., 1996;González Ortega et al., 2018). The Laguna Salada Fault and faults within the Sierra El Mayor-Sierra Cucapah (Axen et al., 1999;Axen & Fletcher, 1998;Fletcher et al., 2014;Fletcher & Spelz, 2009) are southern extensions of the Elsinore Fault (Suarez-Vidal et al., 1991) and accommodate at least an additional~2-3 mm/a (Mueller & Rockwell, 1995) immediately west of the Cerro Prieto Fault. Farther west, roughly~7-8 mm/a of slip is accommodated by the Agua Blanca and San Miguel-Vallecitos Faults across the Peninsular Ranges (Allen et al., 1960;Bennett et al., 1996;Dixon et al., 2002;Hirabayashi et al., 1996;Rockwell et al., 1993;Wetmore et al., 2018). Slip from these faults in part feeds into faults off the Pacific coast of northern Baja and southern California (Legg, 1991), which also accommodate a combined 7-8 mm/a of dextral slip (Larson, 1993;Platt & Becker, 2010). Multiple faults project northward into southern California, but south and east of the Agua Blanca Fault the plate margin converges into a comparatively narrow transtensional system within the Gulf of California. This investigation focuses on the ABF, which based on surface expression alone appears to accommodate the greatest proportion of slip across the Peninsular Ranges; an expanded description of the tectonic framework of northern Baja California and a summary of notable historical seismicity in this region is provided in Section S1 in the supporting information.
The trace of the ABF trends WNW for~120 km, maintaining an overall orientation oblique to present day relative plate motion (Allen et al., 1960;Gastil et al., 1975;Wetmore et al., 2018), and is segmented into five named sections oriented from~280°to~300° (Figure 1b; Section S2). The surface trace of the ABF is readily delineated in aerial and topographic datasets by fault scarps, offset constructional and erosional geomorphology, deflected streams, uphill facing scarps, and local releasing and restraining geometries (Figures 2 and S1-S6). Tectonic geomorphologic expression progressively grows more pronounced along strike to the west but is most clearly expressed near the center of the fault in Valle Agua Blanca, where lower topographic relief, sparse vegetation, and a more arid climate compared to segments farther west combine to create conditions more favorable for geomorphic preservation (Allen et al., 1960).
Disrupted Quaternary landforms unambiguously demonstrate a history of Late Pleistocene and Holocene dominantly dextral surface displacement along the ABF (Figure 2), although historically it has been nearly devoid of microseismicity (Frez et al., 2000;Frez & González, 1991;Gonzalez & Suárez, 1984). The ABF has Numbers following fault abbreviations are slip rates given in units of mm/a; citations for slip rate are provided in sections 2 and 6.3. Roughly~7-8 mm/a of relative plate motion is transferred across northern Baja California by the San Miguel-Vallecitos Fault zone and the Agua Blanca Fault. Stars indicate locations of notable Southern and Baja California earthquakes (refer to Section S1 of the supplement for a summary of these events  (Jarvis et al., 2008). Pacific-North American relative plate velocity from MORVEL (DeMets et al., 2010). accommodated 7-11 km of total dextral slip (Allen et al., 1960;Wetmore et al., 2018), and if existing slip rate estimates are more or less representative of the long-term rate, slip on the ABF would have commenced ca. 2-3 Ma (Wetmore et al., 2018). Although minor secondary fault traces paralleling the ABF are evident from vegetation lineaments and subtle offsets, no major structures diverge from the main trace except in the Valle Santo Tomas near the western end of the fault where the Maximinos and Santo Tomas Faults branch off and follow a more westerly path south of Punta Banda Ridge (Figure 1b). Kinematics and slip histories for these faults are poorly understood, though a marine terrace displaced by the Maximinos Fault that formed ca. 120 ka (MIS 5e based on U-series dating of corals) may record~1 mm/a of dextral slip diverted away from the ABF (Rockwell et al., 1989). The mapped trace of the ABF intersects the coast, where it, and the Maximinos Fault, is presumed to feed slip into the offshore Palos Verdes-Coronado Bank Fault, although offshore connectivity has not been definitively established (Legg, 1991;Legg et al., 1987Legg et al., , 2007. At its eastern end, the ABF intersects, but does not visibly cross, the Main Gulf Escarpment, which separates the Peninsular Ranges from the Gulf of California (Axen, 1995), and south of the ABF intersection is defined by the active Sierra San Pedro Martir Fault (SSPMF) (Section S1; Brown, 1978;O'Connor & Chase, 1989). Slip is transferred between the ABF and SSPMF by a distributed network of faults that can be identified by vegetation lineaments and subtle tectonic geomorphology across the Valle de la Trinidad and Valle San Matias south of the ABF (Figure 1; Section S2) (Hilinski, 1988;Wetmore et al., 2018).
The orientation of the ABF is oblique to current (post-5 Ma) relative plate motions, which was originally explained by initiation along a preexisting structural fabric or terrane boundary (Gastil et al., 1981;Sedlock, 2003). However, based on recent detailed structural mapping, Wetmore et al. (2018) concluded that it was more likely the result of linkage to the extensional SSPMF and transtensional off-shore Continental Borderlands faults at the east and west termini of the ABF. Previous investigators have suggested that the misorientation of the ABF may have prompted the initiation of the more favorably oriented San Miguel-Vallecitos Fault (Grant & Rockwell, 2002), which is characterized by irregular and discontinuous fault sections, a more subtle geomorphic expression (Harvey, 1985) and less than a kilometer total offset (Giroux, 1993), but has ruptured historically (Doser, 1992;Hirabayashi et al., 1996;Shor & Roberts, 1958). However, implicit in this interpretation is the assumption that the orientation of the ABF inhibits slip, which is inconsistent with observations of secondary normal slip along nearly the entire length of the fault (e.g., Wetmore et al., 2018).
Multiple faults between the ABF and the SMVF can be identified, most notably the Tres Hermanos Fault, none of which are well studied, but based on surface expression are likely active (Figure 1a). The amount of slip distributed across secondary and tertiary faults between the ABF and SMVF is unknown, and the geologic slip rate for the SMVF (<<1 mm/a) has been estimated at only one site near its southern terminus (Hirabayashi et al., 1996). Three geodetic (GPS) studies are consistent in measuring~7-8 mm/a of cumulative slip across the Peninsular Ranges (~7 mm/a- Bennett et al., 1996;Wetmore et al., 2018;4-8 mm/a-Dixon et al., 2002), but how this is partitioned between the ABF and SMVF varies based on the crustal rheological model. Specifically, using a simple elastic half space model, slip is split approximately evenly between the ABF (4 ± 2 mm/a) and the SMVF (3 ± 3) (Bennett et al., 1996) whereas if the rheological model is varied, possible solutions place 2.2-3.1 mm/a and 2.4-3.7 mm/a (elastic half-space model), or 6.2 ± 1.0 mm/a and 1.2 ± 0.6 mm/a on the ABF and SMVF, respectively (viscoelastic coupling model) (Dixon et al., 2002); this latter estimate is consistent with both tectonic geomorphologic observations (Section S2; Figures S1-S6) and earlier geologic slip rate estimates of~4-6 mm/a along the ABF (Hatch, 1987;Schug, 1987).
The existing slip rate and earthquake timing estimates for the ABF were first reported in three San Diego State University masters theses (Hatch, 1987;Hilinski, 1988;Schug, 1987). Schug (1987) mapped the geomorphology along the Punta Banda Ridge section of the ABF for~3 km along strike east of Rancho Mirador, a location~5 km west of the site we investigate (Figure 1b). Primarily using 1:50,000 scale air photos, they measured 19 geomorphic features recording offsets ranging from 13 to 1845 m. Surface ages based on soil development range from <2.5 to 840 ka but due to a scarcity of datable material were calibrated with only two radiocarbon dates. Schug (1987) proposed a preferred Late Pleistocene-Holocene (post-28 ka) slip rate for this section of the fault of 4.1 mm/a. Hatch (1987) mapped the tectonic geomorphology along 13 km of the Valle Agua Blanca section of the ABF between Arroyo San Jacinto and Cañada Paredes Coloradas ( Figure 1b) using similar methods. They measured six geomorphic landforms recording 50-300 m of slip. Ages estimated from a soil chronosequence range from~10 to 255 ka but again were calibrated by just two radiocarbon dates. Hatch (1987) proposed a preferred Late Pleistocene-Holocene (post-55 ka) slip rate of 4-6 mm/a. The full ranges of rates permissible by these measurements is~2-10 mm/a (Schug, 1987) and 3-12 mm/a (Hatch, 1987). The westward decrease from 4-6 mm/a to 4.1 mm/a may reflect transfer of 1 ± 0.6 mm/a to the Maximinos Fault opposite Punta Banda Ridge from the ABF (Figure 1b) (Rockwell et al., 1989). Along the eastern, Valle San Matias section of the ABF, Hilinski (1988) measured a post-50 ka slip rate of~1 mm/a, which is some portion of the total amount of slip distributed on to multiple subparallel faults between the ABF and the SSPMF.

New Slip Rate Sites
We used newly acquired airborne lidar topographic data to identify offset geomorphology at three new slip rate sites along the western half of the Agua Blanca Fault (Figures 2 and S6). West to east these are the Las Animas site, located along the Punta Banda Ridge section, and the Arroyo San Jacinto and Valle Agua Blanca sites, both located along the central Valle Agua Blanca section (Figure 1b). We mapped the broad-scale geomorphology at these sites remotely using the lidar and verified our observations and 10.1029/2019TC005788 Tectonics measurements in the field. Remote exploration, mapping, and offset measurements were made using ArcGIS or with LidarViewer . Lidar data were collected in 2014 by the National Center for Airborne Laser Mapping (NCALM). The dataset covers an area of~76 km 2 with a point density of~7 pts/m 2 (all data) and~2.7 pts/m 2 (ground classified). Detailed descriptions of each site are provided in section 5.

Geochronology
We use either in situ cosmogenic 10 Be surface exposure dating or optically stimulated luminescence dating to estimate stabilization and abandonment timing for the offset geomorphic surfaces at the Las Animas, Arroyo San Jacinto, and Valle Agua Blanca slip rate sites. Dating results are given in the context of the age interpretations presented in section 5. Surface type and limited availability of necessary sample material prevented a multichronometer dating strategy (Behr et al., 2010;Blisniuk et al., 2012;P. O. Gold et al., 2015).

10
Be cosmogenic radionuclide geochronology is used to estimate the duration of surface exposure by measuring the concentration of 10 Be isotopes primarily in quartz. The rate of 10 Be production in quartz at and near the earth's surface is known, as is the decay rate, so in theory the measured nuclide concentration is proportional to the time since initial exposure of the sample, which is assumed to coincide with deposition. Detailed descriptions of cosmogenic dating methods can be found in several books and reviews (Dunai, 2010;Gosse & Phillips, 2001;Granger et al., 2013;Ivy-Ochs & Kober, 2008).

Sample Collection and Processing Procedures
We followed established sample collection and processing procedures for extracting 10 Be from quartz in crystalline rocks. When selecting surface samples, we chose clasts that exhibited the least possible evidence of erosion or prior shielding and either used a masonry chisel and hammer to remove <5 cm of material from the tops of boulders (Las Animas Site) or collected whole cobbles (Arroyo San Jacinto Site). At the Valle Agua Blanca site, we excavated a 2-m-deep pit into the surface of Qaf and collected several kg of pebbles and small cobbles (n ¼~20-30 clasts) in 25-cm intervals.
Initial mineral separation was completed at the University of Texas at Austin and quartz separation, cleaning, and chemistry to isolate 10 Be were completed at the Arizona State University School of Earth and Space Exploration. Boulder and cobble samples thicker than 5 cm were reduced manually with a hammer and chisel, and material from the upper <5 cm of each sample was crushed, milled, and sieved. Samples were demagnetized by hand and using a Frantz electromagnet separator to isolate non-magnetic quartz and feldspar separates. Quartz was isolated using froth floatation, leached in HF/HNO 3 solutions, and tested for chemical purity using ICP-OES. A 9 Be carrier was added to samples and blanks prior to digestion in concentrated HF, cleaning in HClO 4 , and chloride conversion. Anion and cation exchange chromatography were used to separate Fe, Ti, Mg, Al, and Be, which was then precipitated as a hydroxide, oxidized to BeO, mixed with Nb, and pressed into cathodes for measurement at the Center for Accelerator Mass Spectrometry at Lawrence Livermore National Laboratory. A detailed description of the 10 Be separation chemistry is provided in Section S3.

10
Be cosmogenic dates were calculated using time-independent production scaling (Table 1) (Lal, 1991;Stone, 2000). We dated 16 boulders at the Las Animas site (Figures 2a, 3, and S20), 12 cobbles at the Arroyo San Jacinto site (Figures 2b, 7, and S20), and eight intervals in the depth profile at the Valle Blanca site (Figures 2c,8,and S21). Uncertainties for all 10 Be concentrations include the analytical (AMS measurement) error, the blank analytical error, a 1% carrier mass error, and a 2% sample preparation error, combined in quadrature. We used the online CRONUS exposure age calculator version 2.3 (https://hess.ess. washington.edu) to calculate exposure ages (Balco et al., 2008) (Table 1). Dates calculated using this version of CRONUS are based on the latest time-invariant, sea level/high latitude 10 Be reference production rate of 4.01 atoms g −1 year −1 (Borchers et al., 2016;Lal, 1991;Stone, 2000). The total uncertainty in the dates combines in quadrature a 1% 10 Be decay constant error and the external error calculated by CRONUS, which includes an~8% production rate error.

Optically Stimulated Luminescence Dating
Optical dating of quartz or feldspar grains estimates the time since last exposure to sunlight, which when dating buried sediments is interpreted to correlate with the timing of last transport and final deposition. The measured signal accumulates over time as ionizing radiation primarily from isotopes of uranium, thorium, rubidium, and potassium frees electrons from parent nuclei in the mineral crystal lattice that then become trapped in crystal lattice defects (Aitken, 1998). Stimulation, or freeing, of trapped electrons by exposure to a controlled light source resets (bleaches) the cumulative radiation and releases energy as a luminescence signal in an amount proportional to the number of trapped electrons, which is in turn positively correlated to the length of time over which the analyzed grains were shielded from sunlight by burial. Under laboratory settings, the light emitted by exposing grains to certain wavelengths of light is measured using a photomultiplier and converted to a date assumed to be representative of depositional timing (Aitken, 1998).

Sample Collection and Processing Procedures
We followed established sample collection and processing procedures for dating quartz sands using optically stimulated luminescence (OSL). Samples were collected by driving 30 cm lengths of 5 cm diameter plastic pipe into sand lenses exposed in the walls of cross-fault trench excavations; trench walls were scraped clean to produce fresh surfaces to avoid sampling collapsed material. Samples were processed and measured at the Luminescence Geochronology Laboratory at the University of Cincinnati.
Sediments within~5 cm of the ends of the sample tube were used to determine dose rates. Sediment was weighed and dried to estimate water content, then crushed to <90 μm, digested in HF, dissolved in HCl and analyzed to measure U, Th, K, and Rb concentrations. Sediment from the center of the sample tubes was isolated for OSL dating and was treated with 10% HCl and 10% H 2 O 2 to remove carbonate and organic material, rinsed, dried and sieved to separate the 90-150 μm fraction. This fraction was etched in concentrated HF and HCl to remove meteoric contamination, other silicates, and fluorides. Magnetic minerals were removed with a Frantz magnetic separator and aliquots of each sample were subjected to IRSL (infrared stimulated luminescence) and OSL to test for feldspar presence and evaluate quartz quality. Aliquots loaded into multi-grain steel disks were measured using a Risø TL-DA-20 OSL reader and a single aliquot regeneration method was used to determine the equivalent dose for the age estimates (Murray & Wintle, 2000).

Error Propagation and Date Calculations
We dated four OSL samples from the Qaf fan, and two additional samples from clastic material deeper in the trenches (Table 2). Between 28 and 52 aliquots were read for each sample, of which 24 to 36 were used to calculate the weighted mean equivalent dose (D e ). Dose rates were calculated using the Aberystwyth University DRAC online calculator; details of dose-rate calculations are provided in Table 2 (Adamiec & Aitken, 1998;Durcan et al., 2015;Mejdahl, 1979;Prescott & Hutton, 1994). The mean D e was divided by the cosmogenic-corrected total dose rate to yield weighted mean ages with standard errors. Dispersion exceeded 20% in all cases, so the preferred dates for all samples are those calculated using a Gaussian 2-mixing model (Galbraith et al., 1999;Vermeesch, 2009) (Table 2).  (Gastil et al., 1975;Wetmore et al., 2018) and Quaternary alluvial sediments deposited across the fault to the north into and over a minor braided river system. Topographic contours reveal the conical form of a large, dissected alluvial fan (F1) north of the fault, the west shoulder of which has been incised by two broad (80 to 100 m-wide), low-relief channels, c1 and c2, that are separated by a remnant of F1 (F1'). A  (Balco et al., 2008). Time-independent spallation production rate based on Lal (1991) and Stone (2000).

10.1029/2019TC005788
Tectonics subtler channel (c0) east of c1 widens before arcing west where it meets the eastern F1-c1 channel wall; c2 is the presently active channel. The north-northeast-trending bed of c1, nearly equivalent in width to the mouth of the catchment, is incised~5 m below the surface of F1, and c2, which is oriented~45 degrees closer to west, is incised~5 m below c1. Upstream of the fault, a flight of fill-cut terraces perched along the east wall of the catchment climbs in elevation with distance from the fault. The lowest terrace is partially capped by a boulder debris flow deposit. No terraces are preserved on the west wall of the catchment, which is steepened and almost uniformly scarred by landslides.

F1 Surface Age
Three dates from boulders occupying a topographically high position near the apex of F1 cluster tightly between 64.2 and 66.2 ka (Table 1; Figures 3 and 4a). A 52.2 ka date was measured from a fourth boulder (ABF3), although we exclude this from the age interpretation because compared to the three older boulders, ABF3 is not nested within F1 but sits atop the fan surface and is noticeably smoother and lighter in color ( Figure S20b), potentially signifying different transport, deposition and exposure histories (e.g., colluvial deposition from steep topography across the fault). However, the older boulders display some evidence of surface weathering and none rise more than~40 cm above the surface ( Figures S20a, S20c, and S20d), so it is probable that the scatter in these dates reflects some variable balance of an inherited nuclide surplus  Estimated fractional day water content for whole sediment is taken as 10% and with an uncertainty of ±5%. c Estimated contribution to dose-rate from cosmic rays calculated according to Prescott and Hutton (1994).
Uncertainty taken as ±10%. d Total dose-rate from beta, gamma and cosmic components. Beta attenuation factors for U, Th, and K compositions incorporating grain size factors from Mejdahl (1979). Beta attenuation factor for Rb is taken as 0.75 (cf. Adamiec & Aitken, 1998). Factors utilized to convert elemental concentrations to beta and gamma dose-rates from Adamiec and Aitken (1998)

10.1029/2019TC005788
Tectonics from pre-emplacement exposure and a nuclide deficit from erosion and/or shielding. We therefore assume that the dates should have a normal distribution (Bird, 2007;Zechar & Frankel, 2009) Figures 3 and 4b). The boulders show little sign of surface weathering, and are situated along a topographically higher axial bar (referred to as the 'c1 bar') into which they are embedded but still higher than the surrounding surface ( Figures S20e-S20j). The outlying c1 boulder dates (54.7 and 66.9 ka) are similar in age to those from F1 and may have been recycled from exposed positions atop or shielded positions within the older deposit. The four younger dates define a recognizable cluster with a median date of 21.8 + 8.0/−6.8 ka (95% confidence) (Figure 4b). This date

10.1029/2019TC005788
Tectonics also assumes a normal distribution, but given the shorter exposure duration (compared to F1) and the positions of these boulders well above the bed of c1, nuclide deficits from shielding or erosion may contribute proportionally less to the~10 kyr range in the individual dates than do inherited nuclide concentrations. Therefore, in addition to the 21.8 ka median, we consider the 17.1 ka minimum date as a viable c1 abandonment age in some slip history interpretations. The estimates of c1 abandonment must predate c2 incision and partial stabilization, and so are consistent with the c2 boulder dates, the most reliable of which (those without unexpectedly wide analytical errors) range from 1.1 to 4.4 ka (Figures S15 and S20k-S20p). The c1-c2 riser has been displaced, but how this feature reconstructs across the fault is unclear ( Figure S18; Section S4), so the c2 dates are not used in the slip rate calculations.

Tectonic-Geomorphologic History and Lateral Displacements
The geomorphology of the Las Animas site records a post-aggradation history of incision and channel abandonment with progressive eastward lateral translation of the F1 alluvial fan (Figures 3 and 5a). The more westerly orientation of c2 relative to c1 maintains a downstream trend roughly perpendicular to elevation contours, illustrating the influence of original radial fan topography on subsequent geomorphic Tectonics modification. Dextral slip is recorded most clearly by the positions of the F1 fan axis and the c1 channel east of the catchment mouth, but relating the cosmogenic dates to these features necessitates unraveling the geomorphic history of the site after deposition of F1 ceased at roughly 65 ka. Uncertainty in the geomorphic history is due primarily to several key parameters including (1) the total offset recorded by F1, (2) the amount of time and displacement that accrued between F1 abandonment and c1 incision, (3) the age of the c1 axial boulder bar (21.8 ka or 17.1 ka), (4) the position of the c1 bar when deposited relative to the catchment walls, (5) the amount of time between c1 incision and bar deposition, (6) the position of the active stream within the upstream catchment, and (7) the position of the eastern wall of c2 when incised relative to the catchment walls. As discussed below, adjusting these variables leads to four end member models of the site geomorphic history that predict a range of slip rates over 65-21.8(17.1) ka and 21.8(17.1)-0 ka time frames that result in the present-day configuration. When attempting to choose which model likely provides the best approximation, it is important to consider the following six observations and interpretations: 1. Topographic contours derived from lidar data reveal the conical form of the original F1 surface and show that the axis of the F1 fan is positioned near the present-day incised eastern margin of F1. 2. The walls of c1 are topographically abrupt, the channel is roughly symmetric in profile, and the width of c1 is comparable to the width of the catchment mouth immediately across the fault. This suggests rapid incision and abandonment relative to seismic cycle length because slow incision and prolonged stream occupation should carve a broader more asymmetric channel as the leading (east) wall is translated to a position protected from erosion and the trailing (west) wall is continually exposed to refreshment. 3. The upstream channel is asymmetrical. The eastern wall is characterized by a flight of terraces cut into alluvial fill, whereas the western wall is steepened and nearly uniformly scarred by landslides. This asymmetry suggests concentrated incision and erosion following post-c1 westward migration of the stream to the west catchment wall. 4. Eastward translation of F1 continually brings lower elevations on the trailing, west side of F1 to the mouth of the channel, perpetually lowering base level where the west catchment wall intersects the fault. This provides the impetus for the prolonged stream occupation at the base of the west catchment wall. 5. Total post-21.8(17.1) ka slip is~35,~65, or~95 m. Restoring the c1 boulder bar to the center of the catchment mouth predicts~65 m of dextral offset (Figure 6a). Restoring c1 to the east and west walls of the catchment mouth suggests lateral displacement of~35 and~95 m, respectively (Figures S16a and S16b). 6. Total post-65 ka slip is either~180 or~140 m. Restoring the F1 axis to the center of the catchment mouth predicts~180 m of dextral offset, with a maximum of 210 m and minimum of 150 m when restored to the west and east catchment walls, respectively ( Figure 6b). Restoring F1 so that the position of the boulders is immediately downstream of the catchment mouth suggests a lesser lateral displacement of~140 m (range: 110-170 m) ( Figure S16c).
Four end member models can explain the present-day configuration of the Las Animas site ( Figure S17; Section S4). Our favored model (Figure 5a; "Model 1" in Figure S17) is that which most easily reconciles these multiple factors without requiring overly complex or unlikely geomorphologic interpretations. In this version of events (Figure 5a), deposition on F1 ceases at~65 ka when the axis of F1 is aligned with the center of the catchment mouth, 180 m from its present-day position (Figure 6b). Approximately 115 m of slip accrues over~43 kyr before channel c1 is carved. Rapid incision of c1 and deposition of the c1 axial bar are coeval at 21.8 ka (Figure 6a). An additional~65 m of slip accrues as the stream migrates to the west side of the catchment in response to continued base-level lowering where the lower elevation trailing edge of F1 meets the west catchment wall. The west wall is steepened as c2 is incised and the c1-c2 riser carved. The ±30-m range for the F1 offsets reflects the width of the present-day channel mouth; this is probably an overly conservative estimate because while the channel mouth may have widened since 65 ka, it likely has not narrowed. The same uncertainty when applied to the~65 m c1 bar offset would be unacceptably high, so we assign an uncertainty of ±5 m that stems from the width of the c1 bar. However, we consider the maximum (95 ± 5 m) and minimum (~35 ± 5 m) c1 bar offsets, as well as the 17.1 ka c1 bar age and 140 ± 30 m F1 offset in alternative, more geologically complex models that are illustrated in Figure S17 and discussed in Section S4.

Most Probable Slip Rates
Our best estimates suggest that the Punta Banda Ridge section of the ABF has accommodated 65 ± 5 m of slip since 21.8 + 8.0/−6.8 ka at a rate of 3.0 + 1.4/−0.8 mm/a and 180 ± 30 m of slip since 65.1 + 12.2/−12.1 ka at a rate of 2.8 + 0.8/−0.6 mm/a at the Las Animas site (Table 3). Ages and rates are given with uncertainties that 10.1029/2019TC005788 Tectonics represent 95% confidence bounds and the ages and offsets are both represented by Gaussian PDFs. Although we estimate offset uncertainty from maximum and minimum bounds in this and the following calculations, our use of a Gaussian rather than a rectangular or trapezoidal model reflects our interpretation that in these cases the maximum and minimum are possible, but improbable, displacements.

Site Description
At the western end of Valle Agua Blanca, the ABF strikes~285°across an incised alluvial deposit that fills the mouth of an embayment at the convergence of four small (~0.4-1 km 2 ) catchments that drain southward into Arroyo San Jacinto (Figures 7, S9, and S10). Channels incising the alluvial surface, designated Q5 (~20 ka) by Hatch (1987) and distinguished by a soil with a reddened argillic horizon, are laterally displaced by the fault, which also generates both uphill-and downhill-facing scarps (Figure 2b). A channel incising the western flank of Q5 is filled by a clastic deposit (Q2) into which the modern channel has incised by up tõ 2 m. Soil development in Q2 is negligible and the low-relief surface is defined by very subtle bar and swale topography and primarily cobble-sized clasts. Where the modern thalweg crosses the fault, it diverges from a single channel downstream into two roughly parallel branches-a less incised channel that continues upstream straight across the fault and the presently active channel, which bends right and parallels the fault before continuing upstream. Dextral slip at this site is recorded by the active stream channel incised into the Q2 channel fill deposit.

Q2 Surface Age
The range in the cobble dates from Q2 spans~4 kyr (Figures 7, 4c, and S21a-S21l). Although this is relatively precise for a cosmogenic clast dataset, it is wide relative to the dates themselves, none of which greatly exceed 4 ka ( Table 1). The scattered dates do not define any single clear cluster, although several dates provide unlikely approximations of Q2 deposition and can be excluded. The 325-year date from sample ABF25, which is bracketed by abnormally high analytical errors (~87%, 1σ), almost certainly represents too short a period of time for Q2 deposition and channel incision followed by one to two recent surface ruptures for which there is no historical record. Sample ABF27 (4.2 ka) is significantly older than the others, likely containing an uncharacteristically high inherited nuclide component. Deposition as young as 697 years (sample ABG18) cannot be absolutely ruled out, so we consider this to be a lower limit to the Q2 age and 2.6 ka (sample ABF16-the oldest nonoutlying sample) to represent the upper limit. Because at least one of the eight dates demonstrably underestimates Q2 exposure (ABF25, 325 years), the scatter in the six dates between <1 and 2.6 ka may not be attributable to inheritance alone, so we again assume a normal distribution of dates and interpret the 1.4 + 1.6/−0.8 ka (95% confidence) median to approximate timing of Q2 deposition. Four samples (ABF 20-23) from the active channel upstream of the fault yielded about the same range in dates, suggesting that they may be recycled from the Q2 surface and not representative of background inheritance; due to uncertainty in their origin, we do not consider these dates further.

Tectonic-Geomorphologic History and Lateral Displacements
The channel incised into Q2 jogs to the right~5 m as it crosses the ABF (Figures 5b and S9c), which could be the result of dextral displacement or simply a deflection of the channel around topography. However, in the absence of any impediment, a headward-eroding channel should incise through Q2 upstream and directly across, not parallel to, the fault. The coherence of Q2 does not change abruptly at the fault, and incision of the second, more recent channel paralleling the main channel to the west demonstrates that cross-fault headward erosion is not impeded. We therefore interpret the jog in the channel to record post-incision lateral offset of 5 ± 1 m (Figure 6d), with uncertainties based on field estimates of the maximum and minimum possible offset ( Figure S9c).

Most Probable Slip Rate
Based on these measurements, the western Valle Agua Blanca section of the ABF has accommodated 5 ± 1 m offset since 1.4 + 1.6/−0.8 ka at a rate of 3.5 + 5.1/−2.0 mm/a at the Arroyo San Jacinto site (Table 3). All uncertainties are again reported at 95% confidence, and age and offset distributions treated as Gaussian.

Site Description
In the central Valle Agua Blanca, the 290°-striking ABF makes an~350 m right step to bound the northern side of the valley (a releasing stepover), where it truncates the toe of an extensive, incised alluvial fan complex designated Q8 (255-495 ka) by Hatch (1987) (Figures 8 and S11). A Figure 7. Map of the Arroyo San Jacinto site. The channel incised into the Q2 surface is offset laterally by the ABF; Q2 was dated to measure thẽ 1.4 ka slip rate. View direction of the photographs in Figure S9 is indicated by the cameral symbol. Schematic topographic profiles along lines A-A′ and B-B′ are illustrated in Figure S10.

10.1029/2019TC005788
Tectonics prominent~600-m-long shutter ridge that just overlaps the western corner of the Q8 fan is the only apparent remnant of this surface on the south side of the fault (Figures 2c and 8). Opposite the fault from Q8 east of the shutter ridge, the valley floor is blanketed by irregular, asymmetric alluvial deposits designated Q4-Q6 (soils characteristic of Latest Pleistocene-Holocene deposition) by Hatch (1987). Lidar-derived 1 m topographic contours reveal the nearly symmetric form of the most westerly of these alluvial surfaces, which we refer to as Qaf. The axis of this fan appears to curve slightly to the SSW, with an average azimuthal orientation of 015-020°. East of the apex of Qaf the westernmost of five incised channels intersects the fault but is at present deflected away from Qaf by an intervening sag pond formed at a smaller right (releasing) stepover. Down-to-the-north slip on the southern fault of this local structural complexity constructs an upstream-facing barrier to prolonged refreshment or modification of the Qaf surface. Lateral slip at this site is recorded by the offset axis of Qaf.

Qaf Depositional Age
OSL dates constrain the timing of alluvial deposition near the apex of Qaf, which is exposed in trenches excavated across the fault. Four samples (10.6, 12.5, 12.9, and 13.0 ka) from two excavations were taken from depths of~50-220 cm below the ground surface and~20-80 cm below the top of the buried fan surface (Figures S12 and S13; Table 2). Two additional samples from deposits farther into the sag pond yielded similar ages of 8.7 and 13.2 ka; however, establishing a correlation between these deposits and Qaf is difficult, so we do not include these dates in our age interpretation. Of the four remaining samples, the older three could have been incompletely bleached during transport, in which case deposition is best estimated by the youngest 10.6 ka date; however, this sample (T1-OSL-1) was taken from a deposit dipping towards the source ( Figure S12), and we cannot rule out the possibility that it was partially re-bleached during later transport and redeposition from some higher position on Qaf. We therefore interpret the 12.5 + 1.7/−2.8 ka (95% confidence) median of the four OSL dates to constrain the timing of Qaf fan deposition (Figure 4d). 10 Be nuclide concentrations from a depth profile are highly scattered and do not decrease exponentially with depth (Figures 8 and S14), likely the result of some combination of submeter incremental deposition, variable catchment residence times and inheritance. No exposure age could be modeled from the depth profile.

Qaf Displacement
The axis and apex of the Qaf fan are readily apparent from the curvature of the 1 m lidar-derived elevation contours (Figure 8). Projected north, the fan axis intersects the fault~40 m east of the mouth of the nearest

10.1029/2019TC005788
Tectonics channel in Q8, which is the nearest viable piercing point for reconstructing slip (Figures 5c and 6c). We make the simplifying assumption that the shortest route between the catchment mouth and the fan apex approximates the stream path at the time of deposition and measure a post-deposition dextral offset of 40 ± 5 m, with uncertainties reflecting the width of the mouth of the source channel (Figure 6c).

Most Probable Slip Rate
Although this site is coincident with a right step in the ABF (Figure 2c), the dominant geomorphic signature of the fault along the north side of the valley suggests that dextral slip has probably been concentrated here since Qaf deposition. Therefore, the central Valle Agua Blanca section of the ABF has accommodated at least 40 ± 5 m of slip since 12.5 + 1.7/−2.8 ka at a rate of 3.2 + 1.0/−0.6 mm/a at the Valle Agua Blanca site ( Table 3). All uncertainties are again reported at 95% confidence, and age and offset distributions treated as Gaussian.

Slip Rate Interpretations
The new slip rate measurements should be considered minimum estimates for reasons related to geochronology or fault zone architecture (Table 3). First, the surface ages at the Las Animas and Arroyo San Jacinto sites are probably more biased by inherited cosmogenic nuclide surpluses gained during clast transport than they are by nuclide deficits from post-deposition erosion or shielding. As is common, we have no specific knowledge of the relative impact of these processes on our dates, but the most logical explanation for observed scatter is inherited signal, since post-deposition erosion and shielding should reduce nuclide concentrations uniformly. If significant erosion and shielding can be ruled out, this balance is dominated by inheritance (more so for the OSL dates from the Valle Agua Blanca site, which are primarily biased by inherited signal due to incomplete bleaching during transport). As a result, the assumption of a normal distribution would be invalid, causing median ages meant to account for these competing processes to overestimate the true timing of abandonment. This is consistent with recent results from modeling of larger and more spatially distributed clast datasets (D'Arcy et al., 2019;Prush & Oskin, 2020), which suggest that surface age might be better approximated by the minimum clast date. We consider this possibility for the c1 bar at the Las Animas site but ultimately reject it because several studies have reported U-series dates from pedogenic carbonate, which in theory provide a minimum constraint on surface stabilization (Sharp et al., 2003), that are older than the youngest cosmogenic dates measured on the same surfaces (Behr et al., 2010;Blisniuk et al., 2012;P. O. Gold et al., 2015). While the assumption of a normal distribution is imperfect, without a greater number of dates, complimentary dates from different geochronometers, or more knowledge of sample transport and exposure histories, interpreting the median as a probable upper bound on surface age in the context of the site-specific observations from F1, c1, Q2, and Qaf described in section 5 is less uncertain than alternative interpretations.
Second, the new slip rates cannot account for off-fault distributed deformation or slip on secondary faults and overlapping strands of the ABF. For example, the Maximinos Fault splits from the ABF in Valle Santo Tomas and may divert~1 mm/a of dextral slip away from the Las Animas site (Figure 1b) (Rockwell et al., 1989). The Qaf fan at the Valle Agua Blanca site lies within a releasing step-over in the ABF and only records slip along the more prominent northern fault strand (Figure 2c). Although the tectonic geomorphologic expression of the southern strand is comparatively subtle, a secondary amount of offset must nevertheless periodically bypass this site. We did not observe any minor faults parallel to the Arroyo San Jacinto site, but offset is recorded by an active stream channel that post-dates deposition of Q2, and there is no way to be sure that displacement did not occur between deposition and channel incision. A third related consideration is that distinguishing between apparent dip-slip from lateral translation of topography and true dip-slip was not possible at the sites we investigated because in all cases but Arroyo San Jacinto, the displacement is recorded by the separation of an abandoned deposit from its still-active source, rather than by a discrete landform identifiable on both sides of the fault. For that reason, we only measure horizontal displacement, but the true slip vector may be somewhat longer when dip-slip is accounted for. The degree to which the slip rates are more or less biased by these factors is unclear, so we interpret the~3 mm/a to be a minimum rate at all sites, and suggest that the true slip rates are likely closer to 4 mm/a.

Comparison to Previous Geologic Slip Rate Estimates
Within error, slip rates averaged over 65.1, 21.8, 12.5, and 1.4 ka of~3-4 mm/a are somewhat lower on average than, but well within uncertainty of, the previous estimates by Hatch (1987) and Schug (1987) for the same sections of the ABF. Both studies follow the same fundamental principles for measuring paleoslip from geomorphic surfaces, but the new rates rely on updated imagery, high-resolution lidar topographic data, and modern direct surface dating strategies and so are more appropriately incorporated into hazard models (e.g., Field et al., 2014).
Defining uncertainties associated with offsets remains difficult (P. O. Scharer et al., 2014), and our approach does not differ significantly from that of Hatch (1987) and Schug (1987), although the quality of the lidar data (compared to the 1:50,000 scale air photos used in the earlier studies) should minimize uncertainties in feature identification and correlation. However, methods for dating geologic surfaces have improved substantially since the earlier investigations of the ABF. Hatch (1987) and Schug (1987) relied primarily on soil development metrics for age control, although soil development is a function of multiple natural processes (Jenny, 1941;Johnson et al., 1990) that can cause soil characteristics to vary substantially between soils of the same age or even from the same surface (Harrison et al., 1990). Due to a scarcity of datable material, soil ages interpreted by Hatch (1987) and Schug (1987) along the ABF were loosely calibrated to just four radiocarbon dates and to soils of known age in southern California. These ages were then projected to other surfaces along the ABF with similar soils, but accounting for and propagating errors was practically impossible. Although geologic processes introduce different sources of uncertainty in modern dating methods that may be difficult or impossible to quantify, an important advantage of cosmogenic and OSL geochronology is that they directly date the offset landforms, removing the need to project ages from qualitatively similar locations. For that reason, the new slip rates provide a stronger foundation for discussions of regional kinematics, seismic hazard, and long-term patterns of strain release.

Regional Kinematics and Strain Partitioning
Present day slip rates for most named active faults offshore and across the Peninsular Ranges of Northern Baja California have been estimated from geodetic (GPS) measurements of interseismic strain (Bennett et al., 1996;Dixon et al., 2002;Larson, 1993;Platt & Becker, 2010;Plattner et al., 2007). Although direct comparison of these to long-term geologic slip rates is not particularly meaningful because GPS rates for individual faults may vary by rheological model and are measured over geologically instantaneous time spans, geodetically constrained cumulative slip across multiple faults may be more reliably projected over the Late Pleistocene-Holocene. Three GPS surveys spanning the Peninsular Ranges largely agree that net slip across the region is~7-8 mm/a (Figure 1a; Bennett et al., 1996;Dixon et al., 2002;Wetmore et al., 2018); thus, the new minimum slip rates of 3-4 mm/a imply that the ABF has accommodated at least half of total slip across the Peninsular Ranges since~65 ka, consistent with its clear tectonic geomorphologic expression. Although the SMVF is clearly active, having produced a M6.8 earthquake in 1956 (Doser, 1992;Hirabayashi et al., 1996;Shor & Roberts, 1958), its surface expression seems too subtle by comparison for it to have accommodated all of the remaining slip across this region over tens of thousands of years. A significantly lower geologic slip rate (<<1 mm/a) for the SMVF supports this view (Hirabayashi et al., 1996), as do the geodetic measurements of Dixon et al. (2002), which account for seismic cycle effects. However, determining whether the subtle surface expression of the SMVF is the result of distributed near-surface displacement rather than a low slip rate, whether our new rates underestimate the ABF slip rate more than we have estimated, or whether slip is more broadly distributed among other active faults across the Peninsular Ranges (e.g., the Tres Hermanos Fault- Figure 1), will require additional long-term slip rate measurements.
Offshore, GPS measurements place~8 mm/a of net slip across the three primary Continental Borderland faults (Figure 1a), which strike subparallel to the west coast at the latitude of northern Baja California (Platt & Becker, 2010). This rate accounts for the difference between net plate boundary slip and slip accommodated by onshore faults and is consistent with an earlier 5.9 ± 1.8 mm/a geodetic rate measured between San Diego and San Clemente Island to the north (Larson, 1993). Roughly half of this slip (4.3 ± 0.8 mm/a from GPS models) is accommodated by the San Clemente-San Isidro Fault (Plattner et al., 2007) and subsea geologic measurements place 1.5 ± 0.3 mm/a on the San Diego Trough-Bahia Soledad Fault over the Holocene (Ryan et al., 2012). No geologic rate has been measured for the Coronado Bank Fault, and the geodetic rate cannot be differentiated from the San Diego Trough-Bahia Soledad Fault because no land exists 10.1029/2019TC005788 Tectonics between them on which to install GPS stations. Subtracting the San Clemente-San Isidro and San Diego Trough-Bahia Soledad rates from the 8 mm/a total (Platt & Becker, 2010) leaves a remainder of~1.1-3.3 mm/a, but a better approximation for the slip rate comes from the slip rates of the faults to which the Coronado Bank Fault connects: the Palos Verdes Fault to the north and the ABF to the south (Legg, 1991;Legg et al., 1987Legg et al., , 2007. Near Los Angeles, the Palos Verdes Fault geologic slip rate is 2.7-3.0 mm/a over the Holocene (McNeilan et al., 1996) or 2.5-3.8 mm/a since 80-120 ka (Stephenson et al., 1995), practically the same as the new 3.0 + 1.4/−0.8 mm/a (22 ka) and 2.8 + 0.8/−0.6 mm/a (65 ka) minimum slip rates from the Las Animas site along the ABF. Including the~1 mm/a contributed by the Maximinos Fault (Rockwell et al., 1989), this suggests a long-term slip rate for the Coronado Bank Fault of 3-4 mm/a.

Slip Rate Variability
The most probable slip rates for the western ABF suggest constant post-65 ka displacement, but variable slip rates over the measured time frames are theoretically possible (Figures 9, S17, and S19). Various displacement-time histories (Figures 9 and S19) result from the different potential age and offset interpretations at the Las Animas site (Figures 5a and S17), specifically, the age of channel c1 (17.1 or 21.8 ka; Figure 4b), the offset recorded by c1 (35,65,or 95 m;Figures 6a,S16a,and S16b), and the offset recorded by the F1 fan axis (140 or 180 m; Figures 6b and S16c). Slip histories that take into account all four slip rates (western and central sites) all predict some degree of deviation from the long-term average regardless of the variables used, but in general, displacement-time variability increases with greater complexity in the geomorphic interpretation of the Las Animas site (Figures S17 and S19; Section S4). The least variable slip histories are those based on Model 1 (c1 offset ¼~65 m; Figure 5a), assuming our best estimates for F1 offset (180 m) and c1 age (21.8 ka); this model predicts minimal deviation from the long-term average within error (Figure 9a). Models 2 and 3 (c1 offset ¼~35 m; Figure S17) differ in the manner of c1 incision and the timing of westward upstream channel migration, but predict the same displacement time history (Figures 9b and S19; Section S4). The defining feature of these models is a sustained period of zero (or negative, i.e., leftlateral) slip required by the roughly equivalent post-12.5 ka and post-21.8 ka offsets (Figure 9b). In contrast, slip histories based on Model 4 (c1 offset ¼~95 m, Figure S17) predict accelerated slip over this same time frame, with rates increasing by a factor of~2.5 to~10 followed by a factor of~1.5 or~3 decrease depending on the F1 offset and c1 age parameters (Figure 9c; Section S4). These slip histories are also differentiated by the 65-21.8 ka slip rate, which never differs from the long-term average by more than a half millimeter per year in Models 2 and 3, but for Model 4 differs by~1.5 to 3.5 mm/a. Given the along-strike separation of the western and central slip rate sites, variability in Models 2, 3, and 4 could be due to an along-strike gradient rather than temporal changes; however, this cannot explain the most extreme variations predicted in these models, which occur between the 65-22 ka and 22-0 ka time frames, since these rates were measured at the same site. Although our interpretation is that the ABF is characterized by long-term time-invariant slip, except when negative (left-lateral) slip rates are predicted (Figure 9b), nearly all of the potential ABF slip histories (Figures 9 and S19) have some precedent in the published literature (Table 4).   Frizzell et al. (1986) 10.1029/2019TC005788

Tectonics
Time-invariant slip rates have been documented along sections of other dominantly lateral faults like the San Jacinto Fault (Blisniuk et al., 2013) in southern California, the central and eastern sections of Kunlun Fault in China (R. D. , the Alpine Fault in New Zealand (Sutherland et al., 2006), and the central section of the North Anatolian Fault in Turkey (Kozacı et al., 2009). However, a growing body of evidence demonstrates that some faults or fault sections experience highly variable slip rates (Table 4). Prolonged periods of zero slip lasting multiple average earthquake cycles have been proposed for the Garlock Fault in southern California (Dolan et al., 2016), and apparent accelerations and decelerations exceeding a factor of 10 have been measured along the Warm Springs Valley Fault in western Nevada (R. D.  and the Awatere Fault in New Zealand (Zinke et al., 2017), although analysis of an earlier slip rate dataset from the Awatere Fault found a more moderate (factor of~2-3) change in slip rate (R. D. . Less extreme but still resolvable slip rate variations (factor of 3-8) have been measured on sections of the Clarence (Zinke et al., 2019) and Wellington (Ninis et al., 2013) Faults in New Zealand, as well as on a section of the Dead Sea Transform in Israel (Wechsler et al., 2018) and the Mojave section of the San Andreas Fault (Weldon et al., 2004). Most of these examples document multimillennial deviations from a long-term average-the longest slip rate record is 760 kyr, but most are less than 100 kyr, with a median of 17 kyr-but short pulses of accelerated slip lasting just several hundred years have been identified in dense datasets from the Altyn Tagh Fault in China (R. D. Gold et al., 2017) and the San Andreas Fault (Weldon et al., 2004) (Table 4).
This variety of displacement-time histories likely reflects significant differences in seismogenic behavior between different types of faults. Large compilations aimed at understanding the influence of various fault characteristics on rupture propagation and uniformity in strain release have shown that with lower cumulative displacement, structural complexities, and fault plane heterogeneities (e.g., bends, step-overs, transitions between locked, and creeping fault sections) are more likely to complicate elastic strain release (Wesnousky, 1990(Wesnousky, , 2006Wesnousky & Biasi, 2011), potentially resulting in phenomena such as earthquake super-cycles, earthquake clustering, and variable slip rates (Dolan et al., 2016;Goldfinger et al., 2013).
Most faults are segmented by structural complexities at some scale, and the datasets in Table 4, whether variable or constant, were measured along sections or segments of broader faults or fault systems, with the partial exception of the Alpine Fault. As predicted, faults with constant slip rates have greater cumulative offsets (median of 100 km, excluding the ABF) compared to those with variable slip rates (median of 15 km). Constant slip rate faults are also somewhat longer than variable slip rate faults (segment median of 140 vs. 90 km; fault median of 1,050 vs. 250 km), and slip faster (median of 12.1 vs. 5.6 mm/a). Interestingly, the ABF has more in common with variable slip rate faults in terms of net slip (7-11 km), segment and fault length (65 and 120 km, respectively), and slip rate (3-4 mm/a). What most notably differentiates the ABF from variable rate faults is that it is not close to any faults of apparently comparable prominence that might affect its slip rate though long-term interaction, as has been proposed for the San Jacinto and San Andreas Faults in southern California (Bennett et al., 2004).
Generally, faults with constant rates are more structurally isolated than those with variable rates (Table 4). Constant rates (and near-regular earthquake recurrence-see Berryman et al., 2012) on the linear and structurally isolated Alpine Fault give way to variable slip rates as it splays into the Marlborough fault system, a 150-km-wide zone of four primary faults that includes the variable slip rate Awatere and Clarence Faults (Sutherland et al., 2006;Zinke et al., 2017Zinke et al., , 2019. The variable slip rate Garlock Fault has high-angle intersections with both the San Andreas Fault and the many faults of the Eastern California Shear Zone-Walker Lane (Dolan et al., 2016). Similarly, active faults parallel, intersect, or overlap the Warm Springs Valley (R. D. , Jordan Gorge (Dead Sea transform) (Wechsler et al., 2018), Wellington (Ninis et al., 2013), and San Andreas Faults (Weldon et al., 2004). In most of these examples, fault interaction is proposed as an explanation for slip rate variability. In contrast, the constant rate Kunlun (R. D.  and central North Anatolian (Kozacı et al., 2009) faults are largely isolated from other comparable structures. The only notable fault subparallel to the ABF that might modulate its slip rate is the San Miguel-Vallecitos (SMVF) system; however, the SMVF has accommodated an order of magnitude less slip than the ABF (Giroux, 1993), and the geologic slip rate along the best expressed section of the fault is <<1 mm/a over the past~110 kyr (Hirabayashi et al., 1996). It seems exceedingly unlikely, therefore, that the SMVF could periodically assume a dominant enough role across the Peninsular Ranges to affect detectable change in the slip rate of the ABF to the degree suggested by our Models 2, 3, and 4 (Figures 9b and 9c).

Tectonics
The ABF is more similar to variable slip rate faults in length, net offset, and slip rate, but like other constant slip rate faults is structurally isolated, suggesting that this characteristic has a dominant influence on whether slip rates are constant or variable.
An important caveat to this discussion is that whether a fault is understood to slip constantly or variably over time is partially a matter of time scale and dataset resolution. The key question is: over what time frames, or over how many average earthquake cycles, are slip rates likely to represent the long-term average rather than a shorter-term deviation? For example, the constant rates measured by Blisniuk et al. (2013) along the San Jacinto Fault are averaged over~12, 45, and <760 ka time frames, between about 1 and 45 times the median 16-kyr duration of the variable slip rate datasets (Table 4), leaving wide open the possibility that the San Jacinto Fault slip rate does vary over shorter time scales. This is a reasonable expectation given its position within a complex multifault shear zone, and variable slip has been documented over the past~2 kyr along a different section of this fault nearer its intersection with the San Andreas Fault (Onderdonk et al., 2015). Slip rates along the central Altyn Tagh Fault are on average uniform over the Holocene (R. D. , which is consistent with its length, net offset, and slip rate. But with the nearly unparalleled density of displacement-time measurements along this section of the fault, a pulse of strain release lasting~400 years has been resolved (R. D. Gold et al., 2017, and references therein). Such short-term deviations from the average slip rate could easily be concealed in almost all of the slip histories listed in Table 4 in which offset is averaged over more than a few thousand years. These examples affirm the importance of dataset resolution in understanding fault behavior over millennial time frames: of the slip histories listed in Table 4, the median number of slip rate measurements per thousand years for variable slip faults is twice that for apparently constant slip faults (0.4/kyr vs. 0.2/kyr, excluding the ABF), implying that interpretations of constant slip rate could in some cases just be an artifact of poorer measurement resolution. While we interpret the prolonged deviations from the average rate predicted by Models 2, 3, and 4 to be unreasonably extreme for reasons related to structural proximity and geomorphic history, the ABF dataset falls at the lower end of this spectrum (four measurements/65 kyr or <0.1/kyr). The period over which time-averaged slip rates approximate the long-term average along the ABF is~10 kyr according to our measurements, but whether relative structural isolation keeps slip rates constant along the ABF, or any other comparable fault, over millennial scales must be confirmed with a higher resolution dataset.

Conclusions
We report the first Late Pleistocene-Holocene displacement-time history for the Agua Blanca Fault based on geologic slip rates measured with modern high-resolution lidar topographic data and direct dating of displaced landforms. The conclusions of this study are as follows: 1. We measured rates over four time frames from three sites along the western and central sections of ABF. Specifically, we measured rates of 3.5 + 5.1/−2.0 mm/a since~1.4 ka at the Arroyo San Jacinto site and 3.2 + 1.0/−0.6 mm/a since~12.5 ka at the Valle Agua Blanca site, both located along the Valle Agua Blanca section, and rates of 3.0 + 1.4/−0.8 mm/a since~21.8 ka and 2.8 + 0.8/−0.6 mm/a sincẽ 65.1 ka at the Las Animas Site along the Punta Banda Ridge section. 2. The~3 mm/a slip rates should be considered minima because at each site landform ages may be biased by inherited signal, and displacement measurements do not include potential off-fault deformation or slip on secondary faults. We suggest that the actual slip rate is likely closer to 4 mm/a. 3. 3-4 mm/a accounts for half of the present day~7-8 mm/a slip budget across the Peninsular Ranges as determined from GPS. Assuming this net rate is not anomalous, the ABF has accommodated more slip since 65 ka than any other fault included in the geodetic rate, which is consistent with tectonic geomorphologic evidence. 4. The new rates imply that the time-averaged slip rate for the ABF has not varied over~10-kyr time scales since at least 65 ka. Variable slip rates over these times scales are theoretically possible, but require less straightforward tectonic-geomorphologic interpretations and/or age interpretations based on only a single date; slip variability over time frames shorter than~10-kyr cannot be ruled out. 5. The ABF has more in common with variable slip rate faults in terms of length, total offset and slip rate, but like many known constant slip rate faults is not paralleled by any faults apparently significant enough to modulate its long-term behavior. This suggests that of the many variables affecting fault 10.1029/2019TC005788 Tectonics slip, interaction with neighboring faults may be a primary mechanism responsible for slip rate variability, or lack thereof. 6. The new slip rates help clarify on-and off-shore strain partitioning at this latitude and should constitute important additions to hazard models for this domain of the Pacific-North American plate boundary system, which until now has lacked long-term slip rates based on modern techniques.