Distinguishing fault offset from climatically modulated channel deflections: Insights from the Pearblossom slip-rate site, Mojave section of the San Andreas fault, California, USA

Measuring fault behavior over time is important for understanding earthquake hazard and how fault systems develop, evolve, and interact (Wallace, 1987; Field et al., 2015; Dolan et al., 2007). Paleoslip analyses reconstruct the history of fault slip using a suite of variably aged landforms displaced by multiple coseismic surface ruptures, with each individual landform recording a different interval of fault slip (e.g., Gold et al., 2017). Such analyses are particularly useful for determining the history of fault behavior over millennial time scales and for bridging the gap from single-event chronologies from paleoseismology and slip-per-event studies of offset landforms (e.g., Zielke et al., 2012; Salisbury et al., 2012) to long-term slip rates from large-magnitude geological offsets (e.g., Gold and Cowgill, 2011; Gold et al., 2017; Knuepfer, 1992; Dolan et al., 2016; Hatem et al., 2021).

Robust paleoslip analyses require many sites that span a large time interval to sample the true history of fault slip. For example, comparing sites that are young, that document few surface ruptures, or that are close in age can produce an apparent change in slip rate that is actually a result of differences in the length of the open interval or earthquake recurrence (Styron, 2019; Hatem et al., 2021). Likewise, large age differences between sites can obscure true temporal variability in slip rate because the lack of data from the intervening time results in a linear interpolation that can fail to sample actual variation in slip rate (Hatem et al., 2021).

Because sites with multiple dateable slip markers that also span a long time frame are rare, paleoslip analyses commonly combine slip rates from multiple sites along a fault. When slip rates from these different sites do not overlap, one of three explanations is likely: (1) Slip rate varied over time (i.e., the fault experienced secular variation in slip rate), (2) slip rate varied in space (e.g., due to along-strike variability in fault structure), or (3) one or more of the slip-rate sites was incorrectly interpreted (Gold and Cowgill, 2011).

The third explanation highlights a fundamental point of this study: Individual slip rates are not direct measurements but rather derived values computed from combining separate age and offset measurements. Thus, geologically determined slip rates are fundamentally reconstruction dependent and critically rely on the assumptions made when selecting the independent age and offset measurements to combine to compute a slip rate.

The most critical assumptions underpinning slip rates are made when reconstructing the site, particularly when inferring the original geometry and subsequent structural and geomorphic development of the site. Thus, every geological slip-rate determination depends upon a key set of assumptions that are embedded in the integrated structural and geomorphic reconstruction of the slip-rate site. Here, we illustrate this point by presenting a case study of a slip-rate site that permits three reconstructions: one that we interpret as a genuine age-offset pair, a second that we interpret as an imposter offset reconstruction, and a third that yields no slip rate at all.

Although conceptually simple, channel-fault interactions are quite complex. In the classic model of Wallace (1968), channels incised straight across a strike-slip fault are lengthened by progressive fault slip, decreasing the local stream gradient at the fault and causing upstream aggradation until the stream jumps its banks and avulses to cut a new, direct course (e.g., Sims, 1994; Dascher-Cousineau et al., 2021). The channel offset can also be reset by stream capture (Duvall and Tucker, 2015; Harbert et al., 2018; Gray et al., 2018; Reitman et al., 2019). Although flume experiments and field studies support this basic model of aggradation and incision in response to faulting (e.g., Ouchi, 2004, 2005), the range of behavior observed in natural channels is considerably more varied.

First, channels do not always form straight and orthogonal to the fault. Topography advected along the fault (e.g., a shutter ridge) can cause channels to incise with an initial deflection, either along or against the sense of fault slip; the characteristic direction of the initial deflections is modulated by the broad-scale topographic slope (Huang, 1993). Deflections opposing the sense of fault slip (e.g., a left-deflected stream crossing a right-lateral fault) can also form by stream capture, which occurs more frequently the closer the spacing of drainages upstream (Li et al., 2021; Huang, 1993). The fault-parallel separation of such channels is less than the true slip since incision, resulting in a reported slip rate that is slower than the true rate. In the alternative case of channels that form with an initial deflection in the sense of fault slip (e.g., a right-deflected stream crossing a right-lateral fault), the fault-parallel separation of the channel is greater than the true slip since incision, resulting in a reported slip rate that is faster than the true rate. In the case of a complex drainage network, it can be difficult to determine the original channel configuration or to correctly correlate features across the fault (Li et al., 2021). For these reasons, the fault-parallel separation between correlated channel segments is not always an accurate record of fault slip, resulting in slip rates that are too fast, too slow, or entirely wrong (e.g., due to matching of unrelated features across the fault) as the present study demonstrates.

Second, geomorphic processes modify offset channels after they form. Deflected stream channels erode laterally over time, increasing the angle of deflection between the fault and the along-fault channel reach, especially if stream power is large (Huang, 1993; Ouchi, 2004, 2005). Landscape evolution models indicate that when geomorphic processes are particularly efficient relative to fault slip, channels and ridges upstream of the fault shift in the direction of movement of the downstream fault block, reducing offset distances and the frequency of stream captures (Duvall and Tucker, 2015; Harbert et al., 2018). On a larger scale, surface processes work to straighten drainages that have become curved due to off-fault deformation, reducing the preservation of this diffuse deformation in the landscape (Gray et al., 2018). Such modification of offsets by geomorphic processes produces offset measurements that underestimate the true slip since channel formation and thus slip rates that are too slow (Reitman et al., 2019).

Third, the set of offset channels present along a fault generally records only a fraction of the total amount of fault slip and, even then, only incompletely. While some deeply entrenched channels can maintain connectivity through many earthquake cycles and thus accrue large offsets, even these channels capture only the youngest part of the slip history on a long-lived fault (Walker and Allen, 2012; Ouchi, 2005). Instead, channel avulsion and stream capture act to reset channel offset. In landscape evolution models, the major influences on the maximum offset that channels can record are shutter ridge length, drainage spacing, and rates of sediment aggradation, although the erodibility of the fault zone and length of the earthquake recurrence interval also play a role (Reitman et al., 2019; Harbert et al., 2018; Dascher-Cousineau et al., 2021; Li et al., 2021). Because the frequency of channel avulsion to form new offset features depends on aggradation rates and thus on sediment supply, changes in the rates of geomorphic processes—due to climate variability, for example—may complicate the reconstruction of slip-rate sites and slip histories (Reitman et al., 2023).

Although some assumption-violating channels are readily identified, of greater concern are “imposter” offset channels that initially look promising as markers of fault slip but ultimately do not record useful information (e.g., Akçiz et al., 2014). We suspect that aborted investigations of such imposter channels that fail to yield a slip rate are likely far more common than is currently documented. Scrutinizing imposter offset channels is valuable for understanding both the processes that give rise to them and how they can be avoided in future slip-rate studies. Here, we begin to address this problem by presenting a case study of an imposter offset at a site along the Mojave section of the San Andreas fault near Pearblossom, California (USA). The Pearblossom site records both a genuine offset and an imposter offset. From the genuine offset, we calculate a slip rate for the site, while other reconstructions of the site motivate us to address the influence of climate on potential piercing lines. We then revisit other imposter offsets to develop two criteria for identifying imposter versus genuine slip-rate sites.

The Mojave section of the San Andreas fault (MSAF; Fig. 1A) in Southern California is an interesting region to investigate both secular variation in slip rate and the potential influence of climate variability on slip-rate sites. Based on paleoseismic evidence (Dawson et al., 2003; Dolan et al., 2007, 2016; Rockwell et al., 2000) and apparent discrepancies between geologic (e.g., Matmon et al., 2005; Weldon et al., 2004) and geodetic slip rates (Becker et al., 2005; Meade and Hager, 2005), periods of faster slip on the MSAF and the Garlock fault have been argued to alternate with times of faster slip on the Eastern California shear zone, suggesting deformation in Southern California alternates between slip on the MSAF–Garlock fault–Los Angeles Basin system versus on the Eastern California shear zone and southern San Andreas fault (Guns et al., 2020; Dolan et al., 2007). Although the possibility of temporal variations in slip rate on the MSAF is intriguing, alternate interpretations of geodetic and paleoseismic data do not require secular variation in slip rate (e.g., Chuang and Johnson, 2011; Johnson et al., 2007; Scharer et al., 2017). In a recent study, Young et al. (2021) reported new slip rates for the MSAF in a paleoslip analysis, which shows that slip-rate variability is permitted, but not required, by the data. Additional paleoslip data are needed to further investigate the proposed temporal variation in slip on the MSAF.

The potential for unambiguously documenting secular variation in fault slip on the MSAF is clouded by the possibility that incision and deposition have been strongly modulated by variations in climate, as has been inferred from observations from across the Southern California region. The specific concerns are that climatic modulation could lead to either periods of time during which few landforms develop across the fault, thereby introducing gaps in the record of fault slip, or periods of time with frequent formation of imposter offsets. Sediment cores from the Santa Barbara Basin show evidence of a large flooding and sediment transport event in Southern California ~400 years ago (Schimmelmann et al., 1998), with a series of unusually thick, gray clay deposits recording several additional large terrestrial flooding events over the last ~1800 years (Schimmelmann et al., 1998, 2003). Northwest of the MSAF, on the Carrizo section of the San Andreas fault, these climate events have been linked to changes in the frequency of channel incision and local patterns of sediment deposition at slip-rate sites (Grant Ludwig et al., 2010; Schimmelmann et al., 1998). For example, a post-1857 sheet flow event on the Bidart Fan produced a right-deflected swale that appears to be tectonically offset but, when excavated, showed no sign of disturbance by the fault (Akçiz et al., 2014; Grant Ludwig et al., 2010). Likewise, to the southeast of the MSAF at Mission Creek, a depositional history from optically stimulated luminescence samples along the San Andreas fault indicates eight pulses of deposition over the last 30 k.y. that appear to be correlated with periods of increased precipitation in the region (Saha et al., 2021). Establishing how climate variability influences the production and preservation of offset features on the MSAF is important for understanding how effectively the question of secular variation in slip rate can be addressed in this region.

We investigated the Pearblossom site remotely using the B4 project lidar data that has been classified to identify and remove vegetation returns (Bevis et al., 2017, 2005) and mapped the surficial geology in the field at a scale of 1:3000 on base maps produced from the B4 lidar digital elevation model. We hand excavated specific units to establish stratigraphy at the site and to sample for radiocarbon and infrared stimulated luminescence (IRSL) dating. We selected excavation sites based on field mapping, preliminary site reconstruction, and land access.

In each excavation, we sampled for both radiocarbon and IRSL dating, collecting at least three IRSL samples from different stratigraphic levels with the highest proportion of sand and greatest bed thickness. We collected each sample by hammering an ~25-cm-long, ~5-cm-diameter section of metal electrical conduit into the wall of the excavation then extracting the sample and sealing the ends of the tube with lightproof material. IRSL sample preparation and analyses were performed at the Desert Research Institute Luminescence Laboratory (Reno, Nevada, USA). We used an infrared stimulated luminescence at 50 °C (IR₅₀) approach on feldspar because, similarly to previous studies in Southern California (e.g., Roder et al., 2012; Rhodes, 2011), we found an unsuitably dim signal from quartz. We dated single grains because the samples were collected from alluvial fans, where sediment commonly experiences incomplete bleaching, which can result in age overestimation (Rhodes, 2015). To model burial dose and calculate age, we used the minimum age model for all samples, which estimates the age of the completely bleached population of grains (Galbraith and Laslett, 1993). Additional details of IRSL methods are provided in the Supplemental Material.¹

For radiocarbon dating, we collected both charcoal and other organic material. Radiocarbon analyses were performed at the University of California, Irvine, W.M. Keck Carbon Cycle Accelerator Mass Spectrometer facility. We report radiocarbon ages as the average of the minimum and maximum ages at 2σ for all calibrated age peaks.

We use the Monte Carlo sampling code of Styron (2015) to calculate the 95% confidence range of linear slip rates for the Pearblossom site and to calculate piecewise linear slip rates for the MSAF by combining the new data from Pearblossom with data reported for the X-12, Key Slide, and Ranch Center sites in Young et al. (2021; see also Fig. 1A), for which we use the reported uncertainty distributions. Following the approach of Young et al. (2021), we use trapezoidal distributions to characterize the age and offset uncertainties at Pearblossom, where the long base of the age trapezoid is the difference between the outer bounds of the bracketing samples at 2σ, and the short base is the difference in their mean ages; the long base of the offset trapezoid is the difference in the maximum and minimum reconstructed offsets, and the short base is the same less the 1 m uncertainty in the lidar for each measurement.

The Pearblossom site formed by the interaction of three northward-draining catchments south of the San Andreas fault with a shutter ridge of older alluvium north of the fault (Figs. 1B and 2; Barrows et al., 1985). The shutter ridge blocks sediment transport except through two gaps where active channels in the western and central catchments cross the fault at elevations of 1111.0 m (western gap) and 1108.0 m (central gap) above sea level, respectively. A third, eastern gap in the shutter ridge hosts an alluvial fan (Qf3-i in Fig. 2B) and a right-deflected channel, the head of which has an elevation of 1111.8 m at the fault. Flow from the eastern catchment crosses the fault at a fourth gap outside and east of the study area (Fig. 1B). Alluvial fans have aggraded behind the shutter ridge, supplied by sediment from all three catchments.

The western catchment is ~22,000 m² and fed the abandoned alluvial fan Qf1-w. The northeastern corner of this alluvial fan connects with a fault-parallel channel that drains eastwards along the upstream side of the shutter ridge to connect with the active central channel at the central gap. The western side of alluvial fan Qf1-w is incised by the western channel.

The central catchment is ~46,000 m² and fed alluvial fans Qf1-c and Qf3-c upstream of the fault and alluvial fan Qf2-c downstream of the fault. The abandoned surfaces of these alluvial fans have mature California juniper trees and lack distributary channel networks or other features that would indicate that they are presently active. All three alluvial fans are incised by the active central channel, which is ~280 m long from its headwaters to the fault and increases in depth across the alluvial fan complex, from ~1 m at the head to ~3 m near the toe. Thalweg-perpendicular topographic profiles across the central channel delineate the top of the channel walls on both the western and eastern banks (Fig. 3A). Upstream of the fault, the central channel contains one generation of inset stream terraces (unit Qt). The ~30-m-long section of the central channel immediately upstream of the fault trends ~40° more westerly than the ~50-m-long section farther upstream (Fig. 2).

The eastern catchment is ~22,000 m² and feeds alluvial fan Qf4-e. Unlike Qf1-c and Qf3-c, the presence of distributary channel networks and absence of California junipers on the surface of this alluvial fan indicates that it is active (Fig. S1 [see ^{footnote 1}]). The toe of Qf4-e merges with other alluvial fans sourced from catchments farther east, all of which drain through the fourth gap in the shutter ridge outside and east of the study area (Fig. 1B).

The central channel is deflected ~7 m to the east where it crosses the fault at the central gap in the shutter ridge at the head of the Qf2-c fan (Figs. 1B and 2). On the south side of the fault at the central gap, the western bank of the central channel terminates in a hillside facet that strikes ~343°, oblique to the fault (Fig. 2A). North of the fault, a northwest-draining channel segment joins the central channel from the eastern side of the central gap (Fig. 2B); the northeastern wall of the channel segment trends ~345°, similar to the strike of the hillside facet south of the fault. The ~345°-trending channel is anthropogenically modified along its southwestern side by an off-road vehicle track that locally deepened the channel. Both the ~345°-trending channel and vehicle track are evident in a series of channel-perpendicular topographic profiles (Fig. 3A). The ~345°-trending channel terminates upstream at the San Andreas fault at a drainage divide with a minimum elevation of ~1111.8 m.

Between the central and eastern catchments south of the fault, the northern margins of the Qf1-c and Qf3-c alluvial fans truncate at a north-facing scarp along the southern edge of the ~345°-trending channel (Figs. 1B, 2, and 3). The scarp height increases systematically westwards along the fault, from 0 m near point E in Figures 2 and 3 to ~1.7 m at the central channel. This variation in scarp height results from the westward slope of the ~345°-trending channel floor combined with the Qf3-c fan surface at the top of the scarp maintaining a roughly constant elevation of ~1113 m, similar to the elevation of the Qf3-c fan surface where it abuts the shutter ridge north of the fault and east of the eastern gap.

The eastern gap in the shutter ridge contains a right-deflected channel incised into alluvial fan Qf3-i, north of the fault and east of the ~345°-trending channel (Figs. 1B, 2, and 3). The eastern-gap channel is ~3.5 m wide and terminates upstream at a drainage divide that separates it from the head of the ~345°-trending channel to the west. The catchment for the Qf3-i alluvial fan extends across and south of the fault onto the Qf3-c alluvial fan, and both topographic contours and profiles indicate that the eastern-gap channel steepens at the fault (Fig. 3).

Based on topographic data alone, the most obvious cross-fault correlation of offset features is between the eastern-gap channel in the Qf3-i alluvial fan and the active central channel (Fig. 2). However, age and stratigraphic data presented in the Unit Descriptions and Geochronolgy section indicate this is an imposter offset. As explained in the following section, the feature we interpret as a genuine offset is more subtle. Specifically, we interpret the ~345°-trending channel on the east bank of the central gap as a beheaded paleochannel, the head of which has been offset from the active central channel to the west and defines the primary offset marker that we date and reconstruct to determine a slip rate at this site (Fig. 2; Fig. S2). An alternative reconstruction is that the ~345°-trending channel was produced by overland or distributary flow across the upstream low-relief surface coalescing into the fault zone at some point in the last few hundred years, in which case it formed approximately in its present configuration and does not yield a slip rate.

To bracket the offset at the Pearblossom site, we restored the head of the ~345°-trending channel (inferred to be a beheaded paleochannel for this reconstruction) on the downstream side of the fault with the western and eastern margins of the central channel on the upstream side of the fault (Fig. 3). Upstream of the fault, we used thalweg-perpendicular profiles to identify and map the slope breaks at the top of each channel wall where the Qf1-c and Qf3-c fan surfaces are present (Fig. 3A). We then estimated linear fits to the slope break points, projecting the fit lines to the fault to obtain piercing-point locations (points A and B in Fig. 2A), noting that the high angle (~70°) between the channel and the fault minimizes projection uncertainty.

Downstream of the fault, an eastern bound on the head of the ~345°-trending channel is provided by the crest of the northwest-trending drainage divide that separates the ~345°-trending channel from the eastern-gap channel (point D in Figs. 2 and 3). Topographic contours define the ~345°-trending channel with distinct Vs in the channel bottom to elevations within 0.5 m of the divide crest. Thus, for a western bound, we use the topographic contour that is ~0.5 m below and west of the divide crest (point C in Figs. 2 and 3).

Using these piercing lines, we measure an offset of 24–49 m (Fig. 3B). The maximum bound is based on matching the divide crest north of the fault (point D) with the western margin of the active central channel to the south (point A); the minimum bound is based on matching the contour 0.5 m below the divide crest north of the fault (point C) with the eastern margin of the active central channel to the south (point B). We choose the tops of the channel walls as the offset markers to account for channel widening as incision depth increased. While the channel thalweg at the time of incision must have been somewhere between the tops of the present-day channel walls, it was not necessarily located between the bases of the present-day channel walls. We choose the conservative pairings of point D with A and C with B to ensure that we include all possible locations and orientations with which the channel could have incised. The minimum bound in particular is also conservative due to the likelihood of eastward erosion of the east bank south of the fault and the expected focusing of erosion following right-lateral coseismic offset of the active channel (e.g., Huang, 1993; Ouchi, 2004; Cowgill, 2007).

For comparison, we also estimate the amount of fault-parallel transport of the eastern-gap channel across the path of the composite Qf1-c–Qf3-c alluvial fan. For the minimum bound north of the fault (point D in Figs. 2 and 3), we use the crest of the drainage divide between the central and eastern gaps. For the maximum bound north of the fault (point E), we use the break in slope at the base of the shutter ridge on the eastern side of the gap to define the eastern boundary. We reconstruct these positions north of the fault to the same maximum and minimum bounds south of the fault used in the reconstruction of the genuine offset. This approach likely overestimates the width of the gap because it does not account for probable erosion within the catchment. Matching the western and eastern edges of the eastern gap with the eastern and western margins of the active central channel, respectively, gives an imposter offset of 36–88 m (Fig. 3B).

We hand dug four excavations (PT1–PT4) to establish the depositional sequence at the site, determine the age of the Qf3-i alluvial fan in the eastern-gap channel, and bracket the age of incision of the ~345°-trending channel. We did not excavate in the ~345°-trending channel or in the Qf2-c fan because we did not have permission to dig in those locations. The excavations exposed deposits emplaced before (Qf1-c fan in PT3 and PT1) and after (unit Qt in PT4, Qf3-c fan in PT3 and PT1, and Qf3-i fan in PT2) incision of the central channel (Fig. 4). To date these deposits, we analyzed ten IRSL samples, four each from PT3 and PT2 and two from PT4, all of which successfully produced ages. Results are summarized in Figure 5 and Table 1 and detailed in the Supplemental Material (see ^{footnote 1}). Overall, ~13% of grains gave a signal, comparable to the proportion in other samples from Southern California (Dolan et al., 2016). In each excavation, the IRSL ages are in stratigraphic order within error (Figs. 4 and 5). In addition, we analyzed eight radiocarbon samples from PT2 and PT3, all of which successfully produced ages (Table 2). The radiocarbon dates are in stratigraphic order within error in all excavations except for sample PT215 in excavation PT2 (Figs. 4 and 5).

In the Qf3-c alluvial fan south of the fault, excavations PT3 (Fig. 4) and PT1 (Fig. S3a) exposed poorly indurated, fine to coarse, subangular to subrounded massive pebbly sand with localized zones of fine horizontal bedding and minor gravel lenses, especially toward the base of the excavation. In PT3, the three stratigraphically lowest IRSL samples overlap within 1σ error and date to ~1.4 ka (Figs. 4A and 5; Table 1). In contrast, the uppermost IRSL sample yielded a younger date of 0.68 ± 0.18 ka. The five radiocarbon analyses from excavation PT3 show the same pattern as the IRSL results, with an abrupt upwards decrease in age at ~60 cm depth below the surface, from ~3100 calibrated years before present (cal BP; present is 1950 CE) in the Qf1-c fan to ~400 cal BP in the Qf3-c fan. We do not observe differences in sedimentological characteristics across the divide in deposit ages.

We interpret the break in IRSL and radiocarbon ages as a depositional hiatus between the end of deposition of the Qf1-c alluvial fan at ~1.44 ka (the age of the youngest sample in Qf1-c) and emplacement of the overlying Qf3-c alluvial fan at ~0.68 ka (the age of the oldest sample in Qf3-c; Fig. 5). We interpret the topmost ~60 cm of PT3 as a separate deposit rather than a zone of bioturbation because, with the exception of sample PT303 (231 ± 75 cal BP), both IRSL and radiocarbon ages within this horizon are clustered around ~0.55 ka and transition abruptly downward to older ages over an interval ≤40 cm (Fig. 5). These patterns are more consistent with rapid deposition of a discrete younger deposit than a bioturbated horizon, which we expect would have a broader distribution of ages and a more gradational transition to older ages at depth. Furthermore, we do not observe a significant population of grains with low D_e values in the IRSL sample from this interval (PT309), which we would expect if bioturbation were present (Table 1; Fig. 5).

In alluvial fan Qf3-i along the eastern-gap channel north of the fault, excavation PT2 exposed two lithologically distinct units separated by a very sharp, channelized, and scoured contact at a depth of ~2 m below the surface of Qf3-i (Fig. 4B). Beneath this contact, unit Qoa is pale brown, matrix-supported conglomerate with subrounded to rounded clasts that range in size from gravel to small cobbles in a clay and silt matrix; this clast-size distribution is similar to that of the deposit that makes up the adjacent shutter ridges. Unit Qoa is very well indurated and contains pedogenic carbonate veins, nodules, and coatings on clast bottoms (Fig. 4B). The overlying upper unit (Qf3-i deposit) lacks such pedogenic carbonate and is poorly indurated, matrix-supported sandy conglomerate with subangular to subrounded grains of coarse gravel in a silt matrix. The deposit also contains gravel lenses as much as 15 cm thick, ranging from local lenses a few tens of centimeters long to layers nearly continuous across the entire excavation. The unit fines upwards overall, with larger and more numerous gravel lenses at the bottom of the deposit than at the top. Bioturbation is significant in this deposit, especially within the upper ~50 cm; a second heavily bioturbated zone occurs at ~150 cm depth on the north and east walls of the excavation.

In PT2, all four IRSL samples are from the upper unit (Qf3-i deposit); the lower (Qoa) deposit was too well indurated to drive in IRSL sample tubes, so we collected no IRSL samples from this deposit. We interpret the Qoa deposit to be older than the units exposed in other excavations based on its coarse grain size that matches that of the shutter ridges at the site, its significant carbonate development, and its high degree of induration. The three stratigraphically lowest samples overlap within 1σ error and date to ~0.6 ka, which is consistent with the uppermost IRSL sample in PT3. The uppermost sample in PT2 at 0.51 ± 0.06 ka is only slightly younger than the lower samples (Figs. 4B and 5; Table 1). The three radiocarbon dates are younger than those of the IRSL samples, which is explained by the sample locations and materials: charcoal sample PT208 is from a heavily bioturbated horizon, and samples PT215 and PT230 are organic materials, not charcoal, that may have been emplaced by burrowing rather than buried during deposition. The correlation in age of IRSL samples from Qf3-i in PT2 and Qf3-c at the top of PT3 indicate that these alluvial fans were deposited at the same time. Importantly, however, PT2 did not expose a massive pebbly sand unit equivalent to unit Qf1-c nor any deposit with an age matching Qf1-c or intermediate in age between Qf1-c and Qf3-i.

In the central channel to the south and upstream of the fault, excavation PT4 in an inset terrace (unit Qt) exposed three layers (Fig. 4C). The lowest (Qf1-c) is composed of moderately well-indurated silt to coarse sand with rare pebbles, significant clay development, and fine carbonate veins in the uppermost 10 cm; its sedimentological characteristics match those of Qf1-c in PT3. IRSL sample PT405 yielded a date of 1.97 ± 0.30 ka for layer Qf1-c in PT4 (Figs. 4C and 5; Table 1), which overlaps within uncertainty the three IRSL dates from unit Qf1-c in PT3 (1.67 ± 0.51 ka, 1.31 ± 0.32 ka, and 1.44 ± 0.43 ka). The middle and upper layers are fill-terrace deposits overlying the Qf1-c alluvial fan into which the channel is incised. The middle layer (Qt-l) consists of very poorly indurated, orange-brown, subangular to angular gravel and sand lenses divided by very thin (<0.5 cm) layers of well-indurated fine sand and silt. IRSL sample PT404 yielded a date of 1.27 ± 0.18 ka for layer Qt-l (Figs. 4C and 5; Table 1). The upper layer (Qt-u) is gray-brown, medium to coarse sand that lacks clay development. Other than this difference in color and upwards reduction in mean grain size, units Qt-l and Qt-u have similar characteristics. No IRSL samples from the upper unit (Qt-u) were analyzed because the poor induration and shallow sampling depth resulted in a high probability that sediment mixed during sampling. There are no radiocarbon data from PT4 to compare with the IRSL ages in this excavation because we did not find any material suitable for radiocarbon dating. The excavation contains a near-vertical fault that strikes roughly parallel to the San Andreas fault and offsets the contact between Qf1-c and Qt-l with 40 cm of south-side-down vertical separation (Fig. S3). This fault also offsets strata near the base of unit Qt-l, but layers near the top of this unit and in overlying Qt-u appear undisturbed.

Stratigraphic and geochronologic data presented above reveal four different depositional periods at the Pearblossom site. From oldest to youngest these events are recorded by: (1) unit Qoa, exposed at the base of excavation PT2; (2) alluvial fan Qf1-c, exposed at the base of PT4 and in PT3, deposition of which began by ~1.97 ka (IRSL sample PT405) and continued through ~1.44 ka (IRSL sample PT310); (3) inset terrace Qt, exposed in PT4 and deposited within the central channel at ~1.27 ka, based on IRSL sample PT404; and (4) alluvial fans Qf3-c and Qf3-i exposed at the tops of PT3 and PT2 and deposited at ~0.6 ka, based on IRSL sample PT309 and samples PT236, PT237, PT238, and PT239, respectively.

The apparent absence of a deposit of the age of unit Qf1 in PT2 is particularly notable and was an unexpected result. The simplest explanation is that the eastern gap is an inherited notch in the shutter ridge that was neither cut nor occupied by the central channel during Qf1 deposition between ~1.97 ka and ~1.44 ka. Alternatively, if the central channel did flow through the eastern gap, a deposit matching the age of unit Qf1 may be missing here if the deposit was scoured away prior to Qf3-i deposition; if the basal Qf3-i unconformity was a bedrock channel at Qf1 time; or if the unit Qoa deposit was emplaced at this time and is not older shutter-ridge material as we interpret.

There are several possible sources for the ~0.6 ka Qf3-c alluvial fan deposit. One is the central catchment as a result of lateral migration of the locus of deposition, occasional debris flows escaping the central channel in the upstream reach where it is relatively shallow (~1 m deep), or a single large flood event that jumped the eastern bank of the central channel where it is more deeply incised downstream and then spilled out onto the Qf1-c alluvial fan surface. Alternatively, the hiatus could represent a shift from an initial phase with debris flows sourced from the central catchment to a later phase, when flows sourced from the eastern catchment prograded onto the Qf1-c alluvial fan.

Likewise, there are several possible sources for the ~0.6 ka Qf3-i alluvial fan in the eastern gap, although the exact source remains unclear. One possibility is that it is derived from the steep hillsides that define the eastern gap in the shutter ridge, with additional local input from the Qf3-c alluvial fan south of the fault, as indicated by the modern local catchment above the Qf3-i alluvial fan (Fig. 2A; Fig. S1 [see ^{footnote 1}]). Another is that the Qf3-i alluvial fan shares the same source as the Qf3-c and Qf4-e alluvial fans south of the fault and results from recent (post ~0.8 ka) hydraulic connection across the fault of the eastern channel with the central catchment, the eastern catchment, or both.

We consider three potential site reconstructions that explain the field relationships, stratigraphy, and deposit ages at the site; one invokes only alluvial fan dynamics, while the other two involve interaction between the San Andreas fault and the catchments at the site. Which reconstruction is correct depends upon the interpretation of the depositional hiatus between Qf1 and Qf3-c and the details of the topography at the site. Although we prefer the second reconstruction for reasons explained in the following three sections, we acknowledge that this solution is non-unique.

Although it is not our preferred reconstruction, one possibility is that the site stratigraphy and formation of the ~345°-trending channel both result from lateral migration of deposition on the alluvial fan surface upstream of the fault, and thus provide no piercing relations. In this reconstruction, the central channel incised sometime before ~1.2 ka (the age of the inset terrace) but was abandoned at or after ~1.2 ka by eastward avulsion near the head of the stream, directing sediment from the central catchment somewhere eastward along the fan surface. Westward avulsion at this time appears to be precluded by the ridge separating the western and central catchments (Fig. 1). At ~0.6 ka, another avulsion event emplaced the Qf3 alluvial fans and returned flow from the central catchment to the central channel, by which time the central gap in the shutter ridge had been transported into alignment with the central channel to allow flow directly through the central gap.

Importantly, in this reconstruction, the ~345°-trending channel is a fan toe gully (Hsu and Petellier, 2004) that formed since ~0.6 ka emplacement of the Qf3 deposits and never received flow from the central channel. In this view, the ~345°-trending channel formed at some point in the last ~600 years by coalescence into the fault zone of overland or distributary flow across the low-relief surface of the Qf3-c fan upstream of the fault. Such an interpretation does not yield a slip rate because the ~345°-trending channel formed in essentially its present configuration, meaning that both the location of the central gap opposite to the central channel at the time of reoccupation and the match in orientation between the beheaded paleochannel and topographic features upstream of the fault are pure coincidence.

Although this purely geomorphic reconstruction makes several testable predictions, the data required to complete those tests are currently lacking. One key prediction is the presence of a channel or outlet north of the fault that was occupied or cut prior to ~1.2 ka deposition of terrace Qt within the central channel. We see no obvious location for this outlet. If the outlet were the ~345°-trending channel, then this reconstruction would become our preferred beheaded paleochannel reconstruction presented in the following section. Flow through the eastern gap at this time is problematic considering that the central channel is 4 m lower than the eastern gap at the fault (i.e., 1108.0 m versus 1111.8 m elevation) and the absence of Qf1-equivalent deposits in excavation PT2. Erosional removal of such deposits by scour at the beginning of Qf3-i deposition seems unlikely considering how small the modern catchment area is above PT2. A second key prediction of this reconstruction is that a deposit that was emplaced between ~1.2 ka and ~0.6 ka and sourced from the central catchment exists somewhere to the east of the central channel. There are two barriers to the identification of this predicted deposit. The first is that the present-day geomorphology provides no indication of the location of such a deposit. If present, the deposit appears to have been buried beneath more recent deposits from catchments to the east of the central catchment (Fig. S1). The second challenge is non-uniqueness: Even if a deposit of the right age is identified, it is unclear how to distinguish whether it derived from the central catchment as opposed to the eastern catchment.

Several observations are better explained by interpreting the ~345°-trending channel as a beheaded paleochannel rather than a fan toe gully that never received flow from the central channel. The width and depth of the ~345°-trending channel are similar to those of the central channel (Fig. 3A) and suggest that similar magnitudes of flow were required to cut the channels. The watershed upstream of the ~345°-trending channel comprises only a small portion of the Qf3-c alluvial fan surface south of the fault (Fig. S1) and seems too small to have provided the flow needed to cut the ~345°-trending channel. Because the eastern catchment drains to the east, it is unlikely to have cut the ~345°-trending channel. Absent another large catchment to provide the necessary flow, we view the ~345°-trending channel as most likely to have been cut by connection with the central catchment via the central channel. The matching orientation of the hillside facet near the central channel upstream of the fault and the preserved paleochannel wall downstream of the fault (Fig. 2) are consistent with this correlation. Although these features support the beheaded paleochannel reconstruction over the purely geomorphic interpretation, they do not uniquely require it, and if the purely geomorphic interpretation is correct, then no channels at the site record a meaningful offset.

We interpret the depositional hiatus in excavation PT3 between the youngest deposition of the Qf1-c alluvial fan at ~1.44 ka and emplacement of the Qf3-c deposit at ~0.6 ka to result from a reduction in sediment supply to the alluvial fan surface due to aggradation of the Qf1-c fan surface to the point that it was captured by a gap in the shutter ridge, triggering rapid incision of the central channel. Deposition on the surface of Qf1-c was less frequent after incision, although the Qf3-c deposit in PT3 clearly indicates the surface was not abandoned entirely. We consider two different possible channel configurations at the time of Qf1-c abandonment and incision of the central channel. Our preferred interpretation is Reconstruction 2, in which the central channel was captured by the central gap with an initial left deflection. We interpret this scenario as the genuine reconstruction of the site and present it using a detailed sequence of events that explains the formation and evolution of the structural and geomorphic features at the site.

During Time 1 (Fig. 6A), the Qf1-w and Qf1-c alluvial fans aggraded upstream of the shutter ridge due to blockage of flow from the western and central catchments. The specific process that produced the shutter ridge topography is uncertain. At some point during this interval, the western catchment likely drained out the central gap, as indicated by the east-draining, fault-parallel channel at the northeastern corner of the Qf1-w alluvial fan (Fig. 2), while sediment from the central catchment aggraded upstream of the shutter ridge to form the Qf1-c alluvial fan to the east of the eastern gap. Importantly, we infer that fault movement during this stage carried the eastern gap in the shutter ridge past the toe of the Qf1-c alluvial fan without capturing the central channel on the basis of the lack of Qf1-aged deposits in PT2. If the gap passed without capture, then we presume the surface of the Qf1-c alluvial fan south of the fault was lower than the head of the eastern-gap channel at the fault. At present, the Qf3-c alluvial fan south of the fault is ~1 m higher than the eastern gap, as indicated by comparison of the 1113 m elevation of the Qf3-c surface at the top of the fault scarp with the ~1112 m elevation of the eastern-gap channel at the fault. Mechanisms to produce this elevation difference include a small component of south-side-up deformation (~0.6 mm/yr to ~1.8 mm/yr), later aggradation of Qf3-c (~0.4 m to ~0.8 m thick) during Time 3, and local lowering of the eastern-gap channel at the fault by headward erosion or coseismic faulting or slumping.

By Time 2 (Fig. 6B), additional lateral fault offset brought the central gap in the shutter ridge and the Qf1-c alluvial fan close enough for the central gap to capture the central catchment, forming the ~345°-trending channel along an initial left deflection at the fault. Additional support is the ~40° leftward bend of the ~30-m-long section of the central channel immediately upstream of the fault (Fig. 2). Hydraulic connection of the Qf1-c alluvial fan surface with the lower base level in the area now buried by the Qf2-c alluvial fan led to abandonment of the Qf1 alluvial fan surface and incision of the central channel upstream of the fault, along with deposition of the Qf2-c alluvial fan downstream. Rapid incision along the left deflection at the fault formed parallel channel walls along the faceted hillside and northeastern wall of the ~345°-trending channel to the southwest and north of the fault-channel intersection, respectively. Subsequent right slip progressively shortened this left-deflected, ~345°-trending channel segment, driving further lowering of local base level, upstream incision of Qf1-c, and downstream deposition of Qf2-c. Shortening of the left-deflected, ~345°-trending channel resulted in progressive, westward-younging abandonment of this channel north of the fault, producing a west-sloping beheaded paleochannel from prolonged westward incision. Juxtaposition of this west-sloping, ~345°-trending channel against the Qf1-c alluvial fan to the south produced the fault scarp that increases in height westward along the south side of the ~345°-trending channel. Within the central channel upstream of the fault, inset terraces (unit Qt) formed shortly after incision.

During Time 3 (Fig. 6C), the Qf3-c alluvial fan was deposited upstream of the fault and the Qf3-i alluvial fan and eastern-gap channel formed downstream of the fault within the eastern gap. We assume that a pulse of sediment from the eastern catchment was also deposited at this time in the area now buried by the Qf4-e alluvial fan, although additional excavations and geochronology would be needed to test this inference. As noted in the Depositional History and Sediment Sources section, the source(s) for the veneer of Qf3-c on top of the Qf1-c alluvial fan south of the fault is uncertain and include encroachment by an alluvial fan sourced from the eastern catchment or debris flows from the central catchment. Likewise, the Qf3-i alluvial fan either could have been sourced from the shutter ridge within a local catchment above the alluvial fan or could have resulted from the same depositional event(s) that deposited Qf3-c, which would indicate that a combination of aggradation and possible south-side-up fault motion finally allowed flow across the fault through the eastern gap in the shutter ridge.

By Time 4 (Fig. 6D), further fault offset first aligned the central channel with the deepest part of the central gap, allowing the stream to fully abandon the left deflection and pass straight across the fault. Additional slip then moved the central gap slightly past the central channel, producing the present ~7 m right deflection at the fault. Upstream of the fault, alluvial fan Qf3-c was abandoned, and the eastern gap began to feed the presently active alluvial fan Qf4-e to the southeast. These changes in deposition are unlikely to have resulted from deformation because no scarp or shutter ridge has cut off the Qf3-c alluvial fan from the eastern catchment and the contact between Qf3-c and Qf4-e is entirely upstream of the fault. On the western side of the site, the western gap captured the western catchment, leading to incision of the Qf1-w alluvial fan.

Based on the topographic data alone, the simplest integrated structural-geomorphic reconstruction of the Pearblossom site is one in which the Qf3-i alluvial fan was deposited when the eastern gap was directly downstream of the central channel. In this reconstruction, the eastern-gap channel incised straight across the fault from the central channel at the time of Qf3-i alluvial fan deposition. These features were then displaced along the fault to develop a right deflection in the eastern gap before the central gap captured the central channel via a left deflection, beheading the eastern-gap channel. This reconstruction results in a total offset of the eastern-gap channel of 36–88 m since deposition of Qf3-i.

However, we view this as an imposter reconstruction due to inconsistencies with the geochronologic, stratigraphic, and topographic details of the site. One key problem is the implied accumulation of 36–88 m of slip since ~0.6 ka deposition of the Qf3-i fan, which conflicts with the paleoseismic record at Pallett Creek (e.g., Sieh, 1984; Salyards et al., 1992) and the slip history at Wrightwood (Weldon et al., 2004), the paleoseismic sites closest to the Pearblossom site (Fig. 1A). Another key problem is the absence of alluvial fan Qf1-c deposits in excavation PT2, which is most simply interpreted as indicating that the eastern gap bypassed the central catchment without occupation. Alternatively, the central channel could have flowed through the eastern gap as a bedrock stream, or the Qf1 deposits could have been removed by erosional scour prior to Qf3-i deposition. A third concern is that the topography of the eastern gap does not clearly preserve a paleochannel that matches the central channel upstream of the fault in terms of size, shape, or orientation. If the central channel once flowed through the eastern gap, no clear evidence of that channel remains because no downstream features in the eastern gap pair with the central channel. For these reasons, we view restoration of the Qf3-i fan to a position downstream of the central channel as an imposter reconstruction. In contrast, the preserved northern wall of the ~345°-trending channel matches the hillside facet along the western margin of the central channel south of the fault, and the trend of the central channel nearest the fault better matches the orientation of the beheaded channel than that of the eastern channel.

To determine a slip rate at the Pearblossom site, we combine the age constraints on the incision of the central channel with the 24–49 m offset between the margins of this channel and the head of the beheaded paleochannel north of the fault according to our preferred reconstruction. The central channel incised upstream of the fault after deposition of the uppermost part of the Qf1-c alluvial fan at 1.44 ± 0.43 ka (sample PT310 near the top of excavation PT3) but before deposition of unit Qt-l on a terrace inside the channel at 1.27 ± 0.18 ka (sample PT404 in PT4). Provided that our preferred reconstruction of the site is correct, this yields a slip rate at Pearblossom of 16–38 mm/yr (95% confidence range; Fig. 7A), which partially overlaps with the combined linear slip rate of Young et al. (2021), giving an updated combined linear slip rate for the Pearblossom, X-12, Key Slide, and Ranch Center sites of 33–38 mm/yr (Figs. 1A and 7A). This range is also broadly consistent with other published slip rates for the MSAF (~30 mm/yr; Matmon et al., 2005; Rust, 2005; Salyards et al., 1992; Sieh, 1984; Weldon et al., 2004; Moulin et al., 2023).

To investigate the possibility of secular variation in slip rate on the MSAF, we also model the Pearblossom site in combination with the sites of Young et al. (2021) using a two-part piecewise linear fit (Styron, 2015). Young et al. (2021) found that the paleoslip data from the X-12, Key Slide, and Ranch Center sites permit, but do not require, phases of faster and slower slip at younger and older times in the history, respectively. This history coincides with a period of faster than average slip rate on the Garlock fault (Dolan et al., 2016; Dawson et al., 2003; McGill and Sieh, 1991), consistent with conceptual models of Southern California fault interactions (Dolan et al., 2007, 2016). With the Pearblossom site included, the data still allow for a change in slip rate, although the mean rate during the younger phase is reduced from 52 ± 19 mm/yr to 39 ± 13 mm/yr and the mean age of the transition between phases is increased from ~1.2 ka to ~1.4 ka (Figs. 7C and 7D). These changes result from the similarity of the Pearblossom slip rate (16–38 mm/yr) with that at Ranch Center (17–39 mm/yr) but younger overall age of the Pearblossom site. Thus, inclusion of data from the Pearblossom site weakens the argument for a temporal change in slip rate on the MSAF around 1.2–1.4 ka.

The Pearblossom site provides a particularly clear example of how a pulse of sediment deposition can generate an imposter offset (i.e., Qf3-c and Qf3-i alluvial fan deposits at ~0.6 ka). But what produced this pulse of sediment if deposition was not tectonically modulated?

To explore this question, we compared the timing of depositional and erosional events at the Pearblossom site with a record of regional climate from Southern California (Fig. 8). Sediment cores from Zaca Lake in the western Transverse Ranges (Fig. 1A inset) record significant changes in precipitation in the region over the last ~3000 years (Kirby et al., 2014), with increases in the percentage of 125–2000 μm sand serving as a qualitative proxy for increased precipitation (Anderson, 2001). The Zaca Lake cores indicate four periods of increased precipitation as indicated in Figure 8A: at present (peak A), ~500 cal BP (peak B), ~1900 cal BP (peak C), and ~2700 cal BP (peak D; Kirby et al., 2014). At the Pearblossom site, the two key depositional periods correlate with this record, with deposition of the Qf3-c and Qf3-i alluvial fans at ~0.6 ka coinciding with the wet phase of peak B and the onset of emplacement of the Qf1-c alluvial fan between 1.9 ka and 1.4 ka correlating with the wet phase of peak C (Fig. 8). We interpret the correlation with peaks B and C to indicate that climate, rather than tectonics, exerted the dominant control on deposition at the Pearblossom site.

To further investigate the potential influence of climate variability on the geomorphic development and modification of other slip-rate sites in the Mojave Desert region, we further compare the Zaca Lake proxy precipitation record with a kernel density estimate of the distribution of 128 charcoal radiocarbon dates that we compiled from studies of late Holocene deposits sampled at 11 sites in the Mojave Desert region. The compilation is limited to sites with depositional records extending at least as far back as 3000 cal BP (Barr, 2016; Dawson et al., 2003; Feakins et al., 2014; Fumal et al., 2002; Rockwell et al., 2015; Young et al., 2021; Young, 2017). We focus on the Zaca Lake record because its resolution and duration are suitable for comparison with the paleoseismic and slip-rate sites. Peaks in the distribution of charcoal dates are broadly synchronous with peaks in the percent sand record, with a shoulder peak at the present day (peak A in Fig. 8A), at ~600 cal BP (peak B), at ~2,000 cal BP (peak C), and at ~2600 cal BP (peak D). In detail, we note that there is only one broad peak in the charcoal record at peak B as opposed to the two sharp sub-peaks in the sand data and that the peaks at B and C are shifted ~100 years older in the charcoal data than in the sand record. We envision that the positive correlation between the abundance of charcoal ages and the precipitation proxy (percent sand) could result from changes in either (1) sediment deposition, with increased rates of deposition increasing charcoal capture and preservation, or (2) the rate of charcoal production via changes in a combination of the density and type of vegetation and the duration and severity of the fire season (Iglesias et al., 2015; Gavin et al., 2007). Increased deposition is the most likely mechanism, given the correspondence between precipitation and deposition reported at Mission Creek (Saha et al., 2021). Saha et al. (2021) similarly compared climate records with the distribution of single-grain luminescence ages at Mission Creek, finding a correlation between increased deposition and wet periods at ~0.6 ka, ~1.6 ka, and ~2.4 ka (plus five older periods), from which they conclude that wetter-than-average periods likely exert a first-order control on deposition in Southern California at millennial time scales.

A final and more speculative comparison is between the Zaca Lake record and periods of landform development on the MSAF inferred from clusters in the magnitude of landform offsets along the MSAF (Fig. 8). Specifically, we converted 60 individual channel offsets measured remotely along the MSAF (Compton, 2012) to their implied ages using the average slip rate of 36 mm/yr for the last 3 k.y. reported by Young et al. (2021). A kernel density estimate of the implied landform ages results in a distribution with peaks at ~1.2 ka, ~2 ka, ~4 ka, and ~8 ka (Fig. 8C). These peaks correspond roughly to interludes between pulses of deposition at Mission Creek, which may indicate that periods of increased incision and channel formation are anticorrelated with periods of increased deposition during wet periods. Thus, we infer that both deposition and channel incision appear to be tied to variations in regional precipitation, with wetter phases enhancing deposition and drier periods enhancing incision in the Mojave Desert region.

The Pearblossom site demonstrates the potential role that climatically modulated deposition can play in forming imposter offsets. The time, effort, and expense involved in determining slip rates from faulted landforms highlights the value of identifying imposters before investing significant resources in their study. Deflected and beheaded channels are commonly used in paleoslip studies, as is particularly well exemplified by the classic slip-rate site at Wallace Creek, to the north of the MSAF on the Carrizo section of the San Andreas fault (Sieh and Jahns, 1984). In general, paleoslip studies of deflected and beheaded channels implicitly assume that fault slip exerts the dominant control on the location and timing of deposition and incision upstream of the fault. Under this assumption, deposition occurs when a shutter ridge blocks or progressively deflects a channel. Subsequent incision occurs when aggradation upstream of the shutter ridge allows the stream to cut a new channel or when fault movement allows a downstream outlet to capture the upstream drainage (e.g., Wallace, 1968; Ouchi, 2005). However, false piercing lines may form when deposition or incision are climatically modulated and thus occur independently of fault activity, as we show here for the imposter offset at the Pearblossom site and as has been documented for a channel that incised with an original deflection on the Bidart Fan on the Carrizo Plain (Akçiz et al., 2014). In the following sections, we use examples of imposter offsets at the Garden Gulch, Collins Canyon, and Webber Ranch sites on the MSAF (Rust, 2005; Barr, 2016; Young, 2017; Figs. 1 and 9) to develop two criteria that can help identify imposter channel offsets on less well-understood faults.

The first criterion is the upstream position of a channel relative to the line of bisection of an alluvial fan or valley-filling terrace into which the channel is incised. The clearest case is that of a steep, symmetrical, conical alluvial fan blocked by a shutter ridge, such as the one shown in Figure 10A, for which the line of bisection is also the axial crest of the alluvial fan. For the case of a valley-filling terrace, the line of bisection is the midpoint of the valley at the elevation of the terrace tread. A channel on the left flank of the alluvial fan or terrace between the line of bisection and left valley wall is likely to incise with an initial left deflection at the fault. For the case of a right-lateral fault, subsequent slip must first remove the left deflection before right-lateral offset accumulates (Fig. 10C). Thus, upstream left-edge channels would tend to provide robust offsets. Such offsets are likely to be a minimum bound on the true offset if the channel is within the alluvial fan, with a higher probability of more closely approximating the true offset if the channel incised along a left valley wall and continues up to the fault, minimizing the space for an initial left deflection along the fault between the toe of the alluvial fan and the shutter ridge. The genuine offset at the Pearblossom site is an example of an upstream left-edge channel along a valley wall that likely captures a true offset because three of the four channel walls appear to be preserved, including both on the upstream side.

In contrast, a channel on the right flank of an alluvial fan or valley-filling terrace is likely to incise with an initial right deflection (Fig. 10D). On a right-lateral fault, offsets of deflected right-edge channels are therefore maximum offsets because some (or even all) of the apparent offset results from this initial deflection. The initial geometry is particularly unclear in the case of a channel incised along the line of bisection because the downstream segment could have formed with a wide range of initial geometries, from orthogonal to right or left deflection, depending on whether it followed one of the depressions that can form where an alluvial fan abuts a shutter ridge (Fig. 10A). Of these three upstream channel configurations, left-edge channels such as the one documented here at the Pearblossom site are the best candidates to pursue for slip-rate studies of right-lateral faults because they are the most likely to record true tectonic offsets.

The Garden Gulch site on the northwestern MSAF provides an unambiguous example of an imposter offset of a right-edge channel, so no slip rate has been reported (Rust, 2005; Fig. 9A). Here, a narrow active channel incises the eastern edge of alluvial and lacustrine deposits that ponded upstream of a bedrock shutter ridge. Given that the youngest horizons postdate the 1857 earthquake (Rust, 2005), the channel postdates the most recent rupture. Thus, the active channel at Garden Gulch formed with an initial right deflection that records no tectonic offset.

The Collins Canyon site provides another example of a right-edge imposter offset (Fig. 9B; Barr, 2016). Here the channel deflects to the right by ~75 m before exiting through a fault-perpendicular notch in a shutter ridge on the northeastern (downstream) side of the fault. To the southwest (upstream) of the fault, the channel incises two debris flow deposits: an older one (unit Qd2 in Fig. 9B) with radiocarbon dates of 2227 ± 79 cal BP to 1969 ± 68 cal BP, and a younger one (Qd3) with dates of 611 ± 53 cal BP to 144 ± 144 cal BP. The younger (Qd3) deposit onlaps the shutter ridge and crosses the fault trace without clear evidence of offset, although it should be noted that the area is vegetated and has been anthropogenically modified (Barr, 2016). Northeast (downstream) of the fault, the channel incises two other deposits: undated older basin fill (Qbf), which is depositionally overlain by a younger alluvial fan (Qf1) with radiocarbon dates of 4037 ± 105 cal BP and 1697 ± 82 cal BP in the lower part and 1347 ± 42 cal BP and 601 ± 54 cal BP in the upper part (Barr, 2016). Because the modern channel incises Qd3 and the Qd3 toe is not offset, the modern channel presumably formed in its current configuration and thus records little to no tectonic offset. We infer that Qd2 and Qd3 aggraded upstream of the shutter ridge until the stream spilled through the notch and deposited at least the upper part of Qf1 before incising in its current geometry. The aggradation and spillover events appear to be climatically modulated, based on the synchroneity between wet periods in the precipitation-proxy record and the ages of the Qd3 (peak C) or Qd2 and uppermost Qf1 (peak B) deposits (Fig. 8A).

This criterion is also generally applicable to fault-crossing alluvial fans in areas that lack uphill-facing topography. A channel that incises on the left side of the alluvial fan is still more likely to incise with a left deflection than a right one because to deflect right would involve the channel following a less steeply sloped—or even uphill—path along the alluvial fan. Similarly, a right-side channel may incise with a right deflection to form an imposter offset. However, the likelihood of such a deflection forming is reduced if there is no change in slope or slope direction at the fault.

The second criterion applies to the case of a deflected channel incised into a geomorphic surface, such as a fluvial terrace or alluvial fan, and is based on the width of the deflection relative to that of the fault zone, where both are measured across fault strike (Fig. 10B). Reitman et al. (2019) distinguished the geomorphic fault zone, which is the region within which landscape evolution modifies channel offsets following surface rupture, from the structural fault zone, which is where coseismic slip and off-fault deformation occur. This distinction is important because the leading channel walls that are transported into the path of the stream by fault slip are likely to experience focused erosion, in contrast to the trailing walls (Bull, 1991), which are progressively abandoned as they are transported away (e.g., Cowgill, 2007). As a result, the geomorphic fault zone can be wider than the structural fault zone where a stream laterally erodes the leading channel walls, relaxing a sharp coseismic offset at the fault into a more sinuous deflection (Huang, 1993; Ouchi, 2004, 2005; Cowgill, 2007; Reitman et al., 2019). If a channel is deflected over a geomorphic fault zone that is much wider than the width of the channel itself, the sinuous channel trajectory across the fault is likely a primary feature that does not record fault slip. In this case, the wide geomorphic fault zone was produced by a former channel that has since been overprinted in the landscape, and the active deflected channel has simply followed this pre-existing path through the landscape. The clearest evidence of such an imposter offset is where the deflected channel is cut into a single geomorphic surface, such as an alluvial terrace or fan, because preservation of the surface within the trailing corner between the channel and the fault provides evidence of an original curved geometry that did not develop by the time-transgressive lateral erosion and channel migration that are required for the deflection to have formed by widening of the geomorphic fault zone.

The Webber Ranch site on the central MSAF provides an example of this second criterion (Young, 2017). Here, valley-filling alluvium (unit Qf3a in Fig. 9C) is incised by an ~15-m-wide channel that is deflected right-laterally at the MSAF by ~230 m in a geomorphic fault zone ~80 m wide across fault strike (Fig. 9C). Radiocarbon dates from Qf3a range from 954 ± 23 cal BP to modern, indicating that aggradation and incision of Qf3a occurred with an initial right deflection and do not record tectonic offset (Young, 2017). The Garden Gulch and Collins Canyon sites also serve as additional examples of the second criterion, with narrow channels incised with wide deflections into alluvial surfaces preserved between the channel and the fault with no clear evidence of progressive abandonment (Figs. 9A and 9B). The channel offsets at all three sites are imposters because they incised with geometries either the same as or close to their present configurations and thus fail to record measurable tectonic offsets.

Here we use a case study of the Pearblossom site along the Mojave section of the San Andreas fault (MSAF) to introduce the concept of imposter offsets, which are landforms that develop across a fault with an initial deflection that is easily misread as tectonic displacement caused by fault slip. Investigation of the Pearblossom site using 1:3000-scale surficial geologic mapping, analysis of devegetated B4 lidar data, stratigraphic studies of hand excavations, and radiocarbon and infrared stimulated luminescence (IRSL) dating reveals four depositional periods recorded by: (1) unit Qoa at the base of excavation PT2; (2) alluvial fan Qf1-c, deposition of which began by ~1.97 ka (sample PT405) and continued through ~1.44 ka (sample PT310); (3) inset terrace Qt that was deposited within the central channel at ~1.27 ka (sample PT404); and (4) alluvial fans Qf3-c and Qf3-i that were both deposited at ~0.6 ka (samples PT309, PT236, PT237, PT238, and PT239). Key landforms at the site include a central catchment and downstream channel, central and eastern gaps in a shutter ridge north of the San Andreas fault, an ~345°-trending channel north of the fault in the central gap, and a Qf3-i alluvial fan north of the fault in the eastern gap.

We present three reconstructions of the site. The first is a purely geomorphic reconstruction in which the ~345°-trending channel is interpreted as a fan toe gully that never received flow from the central channel and thus provides no piercing relations across the fault. Although this reconstruction is permitted by the data, it makes several problematic predictions and provides unsatisfying explanations for several observations.

The second reconstruction is our preferred model, in which we interpret the ~345°-trending channel as a subtle, formerly left-deflected beheaded paleochannel that records 24–49 m of slip. In this integrated structural-geomorphic reconstruction, a capture-induced switch to channel incision and offset accumulation started after ~1.44 ka, which is the youngest date from Qf1-c, but before ~1.27 ka deposition of inset terrace Qt within the central channel. This reconstruction yields a slip rate of 16–38 mm/yr. Combining this rate from the Pearblossom site with the linear MSAF slip rate of Young et al. (2021) slightly narrows the average rate on the MSAF to 33–38 mm/yr over the last 3 k.y. The updated MSAF paleoslip data still permit, but do not require, phases of faster and slower slip at younger and older times in the history, respectively. Inclusion of the Pearblossom data reduces the magnitude of the temporal change in rate that is permitted by reducing the rate during the younger phase from 52 ± 19 mm/yr to 39 ± 13 mm/yr and increasing the mean age of the transition between phases from ~1.2 ka to ~1.4 ka, with a rate of ~16 mm/yr between ~4 ka and ~1.4 ka.

The third reconstruction correlates the Qf3-i fan and an associated right-deflected channel in the eastern gap with the central channel south of the fault, yielding an apparent offset of 36–88 m. However, geochronologic, topographic, and stratigraphic relations are inconsistent with this reconstruction, which we describe as an imposter offset.

Comparison of the timing of depositional and erosional events at the Pearblossom site with a record of regional climate from Southern California indicates that formation of the imposter offset temporally coincides with both a period of increased precipitation recorded in lake sediment cores (Kirby et al., 2014) and the most recent pulse of climatically modulated deposition at Mission Creek (Saha et al., 2021). The Pearblossom site is unlikely to be the only site thus affected, based on a marked temporal correlation between periods of increased precipitation and the frequency of charcoal ages compiled from across the region (Fig. 8A). Offsets may also be clustered temporally, with more offset markers forming in between depositional periods, though this relationship is speculative. These relationships suggest that, as at the Pearblossom site, climate variability may drive the formation of imposter offsets at other sites on the MSAF, which has been documented at the Collins Canyon and Webber Ranch sites. Systems in which alluvial channels interact with shutter ridges appear to be particularly prone to producing spurious imposter offsets that do not record information about fault slip. Using examples of imposter offsets at the Pearblossom, Garden Gulch, Collins Canyon, and Webber Ranch sites on the MSAF, we develop two criteria that can help identify imposter channel offsets by showing that channels that are incised on the upstream left edges of alluvial fans on right-lateral faults or channels that are wide relative to their deflections are the most likely to record reliable information about fault activity.

This research was supported by the Southern California Earthquake Center (SCEC) (contribution 13380). SCEC is funded by U.S. National Science Foundation (NSF) Cooperative Agreement EAR-1600087 and U.S. Geological Survey Cooperative Agreement G17AC00047. Funding for this work was also provided by NSF award 1220588 to Cowgill and by awards to Anderson-Merritt from the EarthScope AGeS program and the University of California (UC) Davis Department of Earth and Planetary Sciences Cordell Durrell fund. This work is an outgrowth of Anderson-Merritt's senior thesis research project, which was made possible with the support of the UC Davis Department of Earth and Planetary Sciences Matthews Fund and the Association of Environmental and Engineering Geologists Foundation Tilford Fund. The W.M. Keck Carbon Cycle Accelerator Mass Spectrometer facility was founded with support from the W.M. Keck Foundation and UC Irvine. We thank Nadine Reitman and an anonymous reviewer for helpful and constructive reviews. We thank Michael Shoaf for generously granting site access, Bruce Buyer and the Angeles National Forest Mojave Work Center firefighters for logistical assistance with our fieldwork, and Yuri Kawashima and Adrien Moulin for assistance in the field. This work has benefitted from a number of helpful discussions with Katherine Scharer, Dawn Sumner, Michael Oskin, Devin McPhillips, Charles Trexler, Alexander Morelan, and Dylan Vasey. All photographs were taken by Anderson-Merritt or Cowgill.