The rate and location at depth of fault creep are important, but difficult to characterize, parameters needed to assess seismic hazard. Here we take advantage of the magnetic properties of serpentinite, a rock type commonly associated with fault creep, to model its depth extent along the Bartlett Springs fault zone, an important part of the San Andreas fault system north of the San Francisco Bay, California (western United States). We model aeromagnetic and gravity anomalies using geologic constraints along 14 cross sections over a distance of 120 km along the fault zone. Our results predict that the fault zone has more serpentinite at depth than inferred by geologic relationships at the surface. Existing geodetic models are inconsistent and predict different patterns of creep along the fault. Our results favor models with more extensive creep at depth. The source of the serpentinite appears to be ophiolite thrust westward and beneath the Franciscan Complex, an interpretation supported by the presence of antigorite, a high-temperature serpent ine mineral stable at depth, in fault gouge near Lake Pillsbury.

The role of fault creep in reducing the magnitude of large earthquakes is the subject of debate (Harris, 2017; Coffey et al., 2022). This is a challenging question to address because key parameters such as how much of a fault’s slip rate is accommodated by creep and at what depth can be difficult to quantify. Furthermore, multiple factors have been proposed to facilitate fault creep, such as lithology, temperature, and fault geometry; thus, it is important to characterize rock types both within and on either side of the fault.

The San Andreas fault system (California, western United States) provides several classic examples of fault creep. In particular, the San Andreas fault zone between Parkfield and San Juan Bautista (Fig. 1) has been the target of numerous studies, including a deep drill hole at the San Andreas Fault Observatory at Depth (SAFOD) that sampled the fault zone at the top of the seismogenic zone. Creep has been documented on other prominent faults within the fault system, such as on the Maacama and Hayward fault zones east of the San Andreas fault (Fig. 1; Louderback, 1942; McFarland et al., 2016). Only fairly recently has creep been documented on the right-lateral Bartlett Springs fault zone, most notably at Lake Pillsbury (blue star in Fig. 1; McFarland et al., 2016).

Although the Bartlett Springs fault zone extends over a distance of 170 km (Lienkaemper, 2010), it is not as well studied as the San Andreas and Maacama fault zones because it traverses remote, sparsely populated, densely vegetated, and steep parts of the northern California Coast Ranges. Despite its remote location, its length suggests that it may be capable of a large earthquake (Wells and Coppersmith, 1994) that may damage infrastructure both near the fault, such as at Lake Pillsbury, and far from the fault. Even though the fault is considered one of the three principal branches of the San Andreas fault system north of the San Francisco Bay, its slip rate is poorly known, restricted to a speculative estimate of ~3 mm/yr since 12–15 Ma (McLaughlin et al., 2018) or estimated from northwestward extrapolation of Quaternary slip rates of the Concord and Green Valley faults (1–9 mm/yr; Dawson, 2013). Geodetic slip rates, which are pertinent over the past several decades, range from 4 mm/yr to 8 mm/yr (Freymueller et al., 1999; Parsons et al., 2013; Murray et al., 2014; Lienkaemper et al., 2014). The most robust estimate of the deep slip rate, derived from Bayesian modeling, is 7.5 ± 0.7 mm/yr below a depth of 13 km (Murray et al., 2014). Estimates of how much of this slip is taken up by creep varies from the fault slipping freely at all depths (Freymueller et al., 1999) to spatially variable patterns of creep along the fault length (Murray et al., 2014; Lienkaemper et al., 2014; Johnson et al., 2022a). Alinement arrays, restricted to a handful of locations along the fault (McFarland et al., 2016; Fig. 2D), indicate that shallow creep is spatially variable, with the highest and best-documented rate at Lake Pillsbury (3.5 ± 0.1 mm/yr; McFarland et al., 2017), whereas analysis of repeating earthquakes provides estimated creep rates of 2–6 mm/yr along the central part of the fault northwest of Lake Pillsbury, with higher rates at depths of 11–15 km (Shakibay Senobari and Funning, 2019). These data are augmented by creep-rate estimates from InSAR data (Xu et al., 2021), but these rates predict left-lateral slip in places along the fault attributed to noise in the data (Johnson et al., 2022a). The modeled distribution of creep has a pronounced effect on the predicted maximum earthquake magnitude and displacement (Lozos et al., 2015), with the Murray et al. (2014) model yielding in a maximum earthquake magnitude and horizontal slip of M 6.5–6.7 and ~1.6 m, respectively, as contrasted with the Lienkaemper et al. (2014) model predicting a maximum earthquake magnitude of M 7–7.2 and horizontal slip of 5.6–5.8 m.

Given the uncertainty of where creep is occurring along the fault zone and the inability to utilize geodetic methods everywhere along this remote fault, other approaches may help predict where creep may occur based on lithology. Serpentinite has been proposed to facilitate creep within fault zones (Louderback, 1942; Moore and Rymer, 2007; Moore and Lockner, 2013). Furthermore, it is magnetic with high magnetic susceptibility (Saad, 1969) and is an important source of aeromagnetic anomalies in the California Coast Ranges (Griscom and Jachens, 1989). Here we model aeromagnetic data using constraints at the surface from mapped geologic contacts to determine the extent of serpentinite along and at depth within the fault zone and characterize the rock types juxtaposed by the fault. This characterization of the fault geology can help determine where favorable rock types are located along the fault that facilitate creep. Rock densities were modeled using nearby gravity measurements.

The integrated geophysical and geologic cross sections presented here not only map inferred structures along the Bartlett Springs fault zone but also provide insights into previously unappreciated complexity in the regional geologic structure that seems to be required by the magnetic and gravity data. Serpentinite is an important component of the Franciscan Complex and forms the lower member of the Coast Range ophiolite, which is regarded as the mantle-derived basement of the western part of the Lower Cretaceous and older Great Valley sequence, also known as Great Valley Group. In particular, the models examine the subsurface interface between the Coast Ranges and Sacramento Valley as well as the Middle Mountain synform, a structure whose eastern margin was aligned with linear earthquake swarms in 2000 and 2006–2007 (Thomas et al., 2013).

The Bartlett Springs fault zone is the northward continuation of a system of faults that branch off of the main San Andreas fault system south of San Francisco Bay (Fig. 1). The recently active trace of the Bartlett Springs fault extends southeast from north of Round Valley to east of the southern tip of Clear Lake, where the fault continues via a right step of 2.5 km onto the Hunting Creek fault (Fig. 2D; Lienkaemper, 2010). Southeast of the Hunting Creek fault is the Berryessa fault, which Lienkaemper (2012) considered as the northern part of the recently active Green Valley fault (Fig. 1). Because these recently active fault traces are superposed on a terrain with a long and complex history of deformation (McLaughlin et al., 2018) that accommodated a large amount of displacement and is responsible for juxtaposition of different bedrock types, we include the Berryessa fault in our analysis. We do not extend our analysis north of Round Valley (Fig. 2).

Along much of its length, the right-lateral Bartlett Springs fault zone traverses bedrock exposures of the Franciscan Complex, which records a history of continental accretion in an east-dipping subduction-zone margin that predated the development and northward propagation of the San Andreas transform system (Blake et al., 1988). The overall style of deformation is eastward underthrusting of generally progressively younger and lower-grade rocks, which are divided into three belts by degree of metamorphism and structural position. The Bartlett Springs fault zone cuts through the Central and Eastern Belts of the Franciscan Complex (Fig. 2A). The Central Belt consists predominantly of mélange that encloses large blocks and slabs of various Franciscan rock types of Jurassic to Late Cretaceous age (Murchey and Jones, 1984; Blake et al., 1988). Blocks and slabs include classic rock types associated with the Franciscan Complex such as radiolarian chert, metabasalt, blueschist, eclogite, amphibolite, and serpentinite. In places, the mélange encloses slab-like bodies many kilometers in extent with distinctive tectonostratigraphy, such as the Snow Mountain volcanic terrane (Fig. 2A) ~8 km east of Lake Pillsbury (Fig. 2D), interpreted to have formed in a seamount or oceanic plateau setting (MacPherson, 1983; McLaughlin et al., 2018), and the Pomo and Marin Headlands–Geysers terranes, which are respectively slabs of dismembered seamounts and oceanic crust (McLaughlin et al., 2018). Although blocks within the Central Belt reach blueschist in grade, the mélange matrix is generally no higher than lower greenschist grade (McLaughlin et al., 2018). In contrast, the Eastern Belt, which structurally overlies the Central Belt, includes mélange with a matrix that is metamorphosed to blueschist facies (Blake et al., 1988).

The Bartlett Springs fault zone displaces another major bedrock terrane, the Upper and Middle Jurassic Coast Range ophiolite. In this area, the ophiolite consists of ultramafic and mafic rocks interpreted to be the mantle-derived, supra-subduction-related basement of the Great Valley sequence (e.g., Shervais et al., 2004), which consists of Upper Jurassic to Upper Cretaceous strata. The strata consist of an older, folded, sheared and complexly faulted sequence that is overlapped by younger Cretaceous, less-deformed sedimentary rocks. The ophiolite is considerably thinned by attenuation faults and modified by Cretaceous and younger uplift and erosion, as recorded by sedimentary serpentinite, ophiolite mélange, and megabreccia at and near the base of the Great Valley sequence (McLaughlin et al., 2018). Ophiolite is interpreted to extend east in the subsurface beneath the Sacramento Valley throughout the study area (Cady, 1975; Godfrey et al., 1997; Orme and Graham, 2018); farther east, basement consists of metamorphic and igneous rocks of the Sierra Nevada.

Structures developed during subduction appear to influence younger strike-slip deformation, exemplified by steep to moderately northeast-dipping seismicity of the Maacama and Bartlett Springs fault zones (Castillo and Ellsworth, 1993). Furthermore, subduction was not always completely orthogonal to the margin, likely producing some earlier component of right-lateral slip and transpressional folding during the early to middle Tertiary, before the development of the Bartlett Springs fault zone (McLaughlin et al., 2018).

Aeromagnetic data within the study area consist of several surveys (U.S. Geological Survey et al., 1994; U.S. Geological Survey, 1996, 1997; Excel Geophysical Services, 2008; Langenheim et al., 2011; Langenheim, 2015). The flightline spacing of these surveys ranges from 530 m (0.33 mi) to 800 m (0.5 mi), but most of the area is covered by flight lines spaced 800 m (0.5 mi) apart and flown 244 m (800 ft) to 305 m (1000 ft) above ground. The aeromagnetic data were gridded at an interval of 200 m and merged by adjusting the total-field anomaly values to a common datum (Langenheim et al., 2023; Fig. 2B). We mathematically transformed the resulting merged aeromagnetic data set into reduced-to-pole magnetic potential (or pseudogravity) anomalies (Baranov, 1957; Fig. 2E) using an inclination of 62° and a declination of 15°. This procedure effectively converts the magnetic field to the gravity field that would be produced if all magnetic materials were replaced with proportionately dense material. The transformation amplifies long-wavelength features caused by magnetically voluminous, thick sources at the expense of short-wavelength anomalies (Blakely, 1996).

Aeromagnetic anomalies reflect the distribution of magnetic minerals, primarily magnetite, in rocks above the Curie temperature, which in the study area ranges from mid-crustal to upper-mantle depths (Bouligand et al., 2009). Field and laboratory magnetic susceptibility measurements (Langenheim et al., 2023; Table 1) confirm that many of the magnetic anomalies in the study area generally reflect the presence of serpentinite either exposed or in the subsurface. Detrital serpentinite and ophiolite mélange can also be magnetic, depending on the abundance of serpentinite and other magnetic rock types (basalt, gabbro) in these units. Other rock types with high magnetic susceptibilities include basaltic rocks in the Pomo and Marin Headlands–Geysers terranes in the Central Belt, lava flows of the Clear Lake volcanic field, and magnetic sandstones within the Great Valley sequence (Langenheim et al., 2023). Cenozoic lava flows also can have a substantial component of remanent magnetization, which may not be aligned with the direction of the present-day magnetic field.

More than 410 gravity measurements (Langenheim et al., 2023) were collected to augment existing measurements (Snyder et al., 1981; Langenheim et al., 2007a, 2019b; Peacock et al., 2020; Fenton et al., 2022). Together, more than 5800 gravity measurements (Fig. S15 in the Supplemental Material1) were processed using standard techniques (Blakely, 1996) to produce an isostatic residual gravity map of the region (Fig. 2C) using a crustal density of 2670 kg/m3, a crust-mantle density contrast of 400 kg/m3, and a sea-level crustal thickness of 25 km. Gravity measurements are non-uniformly distributed in the study area, with an average 0.3 measurement/km2, although spacing is as low as 0.1 measurement/10 km2 within the mountainous areas. Isostatic gravity anomalies should reflect density variations in the upper and middle crust (Simpson et al., 1986), although broad anomalies could arise from the lower crust or mantle. One of the most significant density contrasts in the upper crust is that between dense Mesozoic bedrock and overlying sedimentary cover, although density measurements indicate some variation within the bedrock units (Langenheim et al. 2023; Table 1). Density considered together with magnetic susceptibility can be a proxy for the degree of serpentinization of ultramafic rocks, with increasing serpentinization generally resulting in decreasing density and increasing magnetic susceptibility values (Saad, 1969).

The most prominent magnetic anomalies in the Coast Ranges coincide with outcrops of the Coast Range ophiolite along the western margin of Sacramento Valley (Fig. 2A). Magnetic rocks extend east beneath the Great Valley complex and younger cover, as demonstrated by the eastward increase in magnetic potential values (Fig. 2E). The magnetic basement completely concealed beneath Sacramento Valley, known as the Great Valley ophiolite, is interpreted to be the intact crust and mantle of the Coast Range ophiolite (Jachens et al., 1995; Godfrey et al., 1997). Exposures of Coast Range ophiolite indicate substantial thinning. Two magnetic models disagree on the westward extent of the Great Valley ophiolite, with one model indicating that it does not extend west of the Coast Range ophiolite (Godfrey et al., 1997), while the other model shows it extending significantly west beneath the Coast Ranges (Jachens et al., 1995). For our purposes, we acknowledge the uncertainty in differentiating Coast Range ophiolite and the Great Valley ophiolite and model the magnetic material that extends at depth west of the exposed ophiolite as Great Valley ophiolite with the same properties as the Coast Range ophiolite.

The western boundary of this magnetic terrane trends north-northwest and indicates a fundamental change in basement between the generally weakly magnetic rocks of the Franciscan Complex and magnetic Great Valley ophiolite. This boundary must dip westward given the gentle westward-facing magnetic gradient, which has been confirmed by previous modeling of magnetic anomalies (Jachens et al., 1995; Godfrey et al., 1997) along a profile nearly coincident with our profile 5 (Fig. 2).

West of the Coast Ranges–Sacramento Valley interface, northwest- to west-northwest-trending magnetic highs (Fig. 2B) are in places bounded by Quaternary faults such as the Maacama, Collayomi, and Bartlett Springs fault zones (Fig. 2A). South of Ukiah (Fig. 2D), the Maacama fault zone truncates the eastern margin of magnetic highs that trend obliquely to the fault zone. The Collayomi fault zone generally bounds the eastern margin of a thick section of Coast Range ophiolite; the magnetic high associated with this section extends and becomes more attenuated to the east, suggesting that the ophiolite section and the fault dip to the northeast. This magnetic high continues northwestward beyond the mapped surface extent of the Quaternary fault trace, suggesting a possible bedrock structural connection to the faults that produced the earthquake swarms in the Middle Mountain area. Magnetic highs associated with the Coast Range ophiolite also continue southeast toward the Bartlett Springs–Green Valley fault zones.

The Bartlett Springs fault zone can be divided into two segments based on where it crosses the pronounced Coast Ranges–Sacramento Valley magnetic potential gradient between profiles 8 and 9 (Fig. 2E). The magnetic highs southeast of the gradient indicate a substantial volume of magnetic rock along the southeastern part of the fault zone. The fault zone lies entirely within this volume of magnetic material, except for the northern part of the Hunting Creek fault. There, the fault coincides with the eastern margin of a band of prominent magnetic highs in the vicinity of profile 12 (Fig. 2B). To the northwest of the Coast Ranges–Sacramento Valley magnetic potential gradient, the fault zone traverses a region characterized by lower magnetic potential values (Fig. 2E) and magnetic anomalies with values less than ~300 nT (Fig. 2B).

Despite low magnetic-potential values northwest of the Coast Ranges–Sacramento Valley gradient (Fig. 2E), magnetic anomalies extend along the fault zone nearly continuously to Lake Pillsbury (Fig. 2B). Northwest of Lake Pillsbury, magnetic anomalies are located dominantly west of the Quaternary strands of the fault zone and appear to be truncated by the fault zone.

Generally, sedimentary deposits are weakly magnetic and do not produce measurable aeromagnetic anomalies in the study area. An exception is the pattern of long, north-trending, narrow magnetic highs that coincide with outcrops of the Great Valley sequence (Figs. 2A and 2B) along the western margin of the Sacramento Valley. The asymmetry of the anomalies is consistent with an east dip of the magnetic beds within the sequence. Another exception is detrital serpentinite that is mapped within the lower part of the sequence, most notably near the northeastern end of profile 12 (Figs. 2A, 2B, and 3L).

Gravity gradients coincide in places with strands of Quaternary fault zones in the Coast Ranges. Because of the large density contrast between Cenozoic deposits and underlying Mesozoic bedrock, some of the most pronounced gravity anomalies coincide with basins developed within the fault zones. For example, local gravity lows (values of −12 mGal to −16 mGal in Fig. 2C) near Willits, Ukiah, Hopland, and ~5 km north of Healdsburg (Fig. 2D) correspond with strike-slip basins within the Maacama fault zone (Langenheim et al., 2010, 2019a). Other notable local gravity lows in Round Valley, Lake Pillsbury, and Cache Valley delineate strike-slip basins in the Bartlett Springs Fault zone. A three-dimensional inversion of gravity data assuming a density contrast of −400 kg/m3 indicates ~400 m of Quaternary deposits in the Lake Pillsbury basin (Langenheim et al., 2007a).

Other gravity anomalies clearly show important density contrasts within Mesozoic bedrock. The large, prominent gravity high beneath Sacramento Valley is attributed to the intact Great Valley ophiolite composed of gabbro and unserpentinized peridotite 14–16 km thick (Godfrey et al., 1997). Within the Coast Ranges, less-prominent gravity highs reflect dense mafic volcanic slabs or terranes within the Franciscan Complex, such as the Snow Mountain volcanic terrane and the north-northwest-striking gravity highs on either side of Clear Lake (Fig. 2C). The high on the west side of Clear Lake is obscured by a semicircular gravity low southwest of Clear Lake that has been attributed to a thick section of Great Valley sedimentary rocks (Stanley et al., 1998) or, earlier, to a large low-density melt body extending below a depth of ~7 km (Chapman, 1975). Three-dimensional gravity inversions, however, indicate that the low likely reflects mid-crustal intrusions and some percentage of partial melt (Mitchell et al., 2023). Gravity gradients coincident with the Bartlett Springs fault zone indicate density contrasts within the bedrock. Along the Berryessa strands of the Bartlett Springs fault zone, a northwest-striking gravity gradient indicates juxtaposition of lower-density bedrock to the northeast against denser bedrock to the southwest. The local gravity lows (Fig. 2C) at Round Valley and Lake Pillsbury (Fig. 2D) are superposed on broader gravity lows that extend along the Bartlett Springs fault zone across exposed Mesozoic bedrock; these lows are in places accompanied by magnetic highs, suggesting the presence of subsurface serpentinite.

In the northern part of the study area, west of 122°45′W longitude, gravity values decrease even more (<−28 mGal; Fig. 2C) along a roughly east-trending gradient, all over exposed Franciscan Complex (Fig. 2A) and west of the gentle magnetic gradient along the Coast Ranges–Sacramento Valley interface (Fig. 2B). The source of these low values is not explained by geology at the surface and was attributed by Jachens and Griscom (1983) to the presence of the Gorda plate at depth beneath the Coast Ranges and northern part of Sacramento Valley.

To test models of the depth extent of serpentinite within the Bartlett Springs fault zone, we constructed a series of 14 geologic cross sections (Fig. 2A) that we then modified using magnetic and gravity modeling. The sections were constructed using geologic maps of McLaughlin et al. (2018), Blake et al. (1992), and Melosh et al. (2023). Geologic units were assigned magnetic susceptibility and density values guided by physical property measurements (Table 1; Langenheim et al., 2023); note that remanent magnetization was included only for the Clear Lake volcanics (which was incorporated into the magnetic susceptibility value as the remanent directions are nearly the same as that of the Earth’s magnetic field). Although we do not have paleomagnetic measurements of serpentinite from the study area, Saad (1969) showed that increasing serpentinization led to increasingly unstable remanent magnetization. Because of this instability, the component of remanent magnetization is difficult to characterize and we do not attribute any remanent magnetization to the serpentinite.

The predicted aeromagnetic and gravity anomalies were then computed using a 2.5-dimensional modeling program (GM-SYS v. 4.9, Seequent, Christchurch, New Zealand). All the profiles are nearly perpendicular to the anomalies associated with the Bartlett Springs fault zone except for profile 1, which intersects the magnetic high where it changes strike; there we modeled the anomaly assuming its two different orientations. Modeling across other anomalies that are not perpendicular to the line of section do not fully account for the three-dimensional shape of those structures. The predicted and observed magnetic anomalies were compared to then modify the unit geometry until a visually good fit with a geologically reasonable geometry was achieved. A geologically reasonable geometry is not unique given the complex and long history of deformation, particularly where the source of the anomaly is concealed; thus, alternate models were constructed to provide a sense of the range of possible body geometries that also fit the observed magnetic anomaly. The amplitude of the anomaly was not the only attribute to match; matching its gradients and inflections were critical parameters that provide important information on the depth to the top of the source and its shape. Franciscan Complex units were generally assumed to have a density of 2670 kg/m3 (Langenheim et al., 2023); however, in some regions, the density of this unit was modified slightly to fit gravity measurements within 1 km of the profile. Although magnetic susceptibility can vary by an order of magnitude (or more) at the outcrop scale, we note that density may also vary considerably because of blocks and slabs within the various mélange units within the Franciscan Complex. Given the variable and commonly sparse distribution of gravity measurements along many of the profiles, more care was taken to fit the magnetic anomalies because of their uniform data distribution.

The models of the original, alternate, and final geologic cross sections are shown in Figures S1–S14 (see footnote 1); Figure 3 shows the final model for each profile. Note that four of the profiles (3, 4, 9, and 10) have geologic cross sections independently constructed by McLaughlin and Melosh to assess differences in the projection of geologic contacts into the subsurface and its effect on the predicted magnetic and gravity anomalies (Figs. S3, S4, S9, and S10). To fit the long-wavelength decrease in magnetic anomalies west of Sacramento Valley that affected all the models except along profiles 1 and 2, we assumed that the magnetic source beneath the valley extended to a depth of 15 km and extended west of exposures of Coast Range ophiolite. For all the profiles except for 9, 10, and 14, the assumed magnetization and density of Great Valley ophiolite was the same as that assumed for the Coast Range ophiolite and other serpentinite bodies. Aside from the constraints provided by the geologic mapping of contacts at the surface, only a handful of other constraints are located along or near the cross sections. A seismic refraction line nearly coincident with profile 5 is the only independent geophysical information across the Bartlett Springs fault zone (dashed green line in Fig. 2D; Godfrey et al., 1997). Bouguer gravity and magnetic anomalies were also modeled along that line (Fig. 4); however, Godfrey et al. (1997) made no attempt to model the magnetic signature of the Bartlett Springs fault zone because of their emphasis on modeling of the prominent magnetic anomalies at the Coast Ranges–Sacramento Valley interface. A deep geothermal well (Table 2; McLaughlin et al., 1990) near the end of profile 11 helped locate the top of the ophiolite, and geothermal wells (Table 2; California Division of Oil, Gas, and Geothermal Resources, 2022) southeast of profile 10 provided insight into rock type and density (Fig. 2D). A seismic-reflection profile (Ramirez, 1992; Constenius et al., 2000) east of Clear Lake near the eastern end of profiles 11 and 12 (dashed gray line in Fig. 2D) helped constrain the thickness of the Great Valley sedimentary sequence.

Modeling of the magnetic and gravity anomalies combined with mapped geologic relations provides important constraints on fault geometry (Saltus and Blakely, 2011), even if potential-field anomalies provide mathematically non-unique solutions. We show alternate models along the profiles that illustrate sensitivity of the geometry and magnetic properties of the bodies, particularly within and near the Bartlett Springs fault zone, in Figures S1–S14. For example, increasing the magnetization while decreasing the thickness of a proposed serpentinite body would generally not produce an appreciable change in the computed field (Figs. S1C, S2D, S3E, S4E, S5F, S5G, S6E, S7D, S8D, and S10E). We also show different extents of magnetic bodies within the Bartlett Springs fault zone that also can fit the observed magnetic anomalies (e.g., Figs. S2C, S4D, S5C–S5E, S6C, S6D, S7C, S8C, S9D, and S14D). Although the body modeled in Figure S5E fits the observed magnetic anomaly, the geometry of such a shallow body implies little or no strike-slip offset along the Bartlett Springs fault zone. Nonetheless, the final models presented here are based on numerous iterations and have been tested to produce models that are geologically reasonable with minimum structural complexities. The synthesis of the potential-field data and independent constraints from the surface geology should lead to a family of similar models, regardless of the starting model, with characteristics that support our major conclusions regarding the subsurface geology, even if the depth extents have considerable uncertainty.

A major finding of this analysis is that more magnetic material resides in the subsurface than observed at the surface along much of the Bartlett Springs fault zone. Comparison of predicted magnetic anomalies from original geologic cross sections with those of the final models illustrates this result for all the profiles (Figs. S2–S14 [see footnote 1]) except for profile 1 (Fig. S1) where a small increase in the thickness of exposed serpentinite (unit Josp) can account for the location and amplitude of the magnetic anomaly west of the Bartlett Springs fault zone (Fig. 3A).

Profile 1 also illustrates that the ophiolite mélange unit is not appreciably magnetic given that its exposed extent predicts a magnetic high west of the observed magnetic high (Fig. S1A). This mismatch in predicted and observed magnetic anomalies is also obvious along profiles 5, 6, 7, 8, and 9 (Figs. S5–S9), indicating that the mélange does not consist of significant amounts of serpentinite in these areas but rather that the matrix is likely dominated by argillite. However, this unit appears to be more magnetic to the south along profiles 12–14 (Figs. 3L3N), indicating that the matrix contains a higher percentage of serpentinite or incorporates more serpentinite or basalt blocks. The general absence of a corresponding gravity high argues against basalt. This means that the geologic identity of the magnetic material modeled near or within the Bartlett Springs fault zone must incorporate an important volume of serpentinite, regardless of whether it is mapped as unit Josp or as ophiolite mélange (unit KJom). Lower gravity values along much of the Bartlett Springs fault zone are also consistent with the identity of the magnetic material as mostly serpentinite rather than other denser magnetic rock types such as peridotite or mafic volcanic rocks. Along profile 5, the absence of higher seismic velocities at the Bartlett Springs fault zone (Godfrey et al., 1997; Fig. 4D) and presence of a gravity low are also consistent with serpentinite being the source of the magnetic anomaly at the fault zone, rather than unserpentinized ultramafic or mafic volcanic rocks with expected velocities of >6 km/s.

Our models highlight the extent of Great Valley ophiolitic basement west of the Sacramento Valley–Coast Range interface. The absence of a corresponding regional gravity high (like the one beneath the Sacramento Valley) with this large volume of highly magnetic rock indicates that the bulk of the material may be serpentinite or serpentinized ultramafic rocks of the Great Valley ophiolite. This interpretation is also consistent with the findings of Godfrey et al. (1997) for their joint interpretation of geophysical data approximately along our profile 5 (Fig. 4). Although the geometry of the Great Valley ophiolite west of exposed Coast Range ophiolite varies from that of three independent magnetic models (Fig. 4), the magnetic data are consistent with a substantial volume of ophiolitic rock in the upper to middle crust at least 20 km west of exposed ophiolite at the Coast Ranges–Sacramento Valley interface along profile 5. Note that the observed magnetic anomalies along profiles 11–13 can be fit without a deep magnetic body; however, the absence of such a deep body still indicates magnetic material to depths of 10 km in the Bartlett Springs fault zone (Figs. S11D, S12C, and S13C).

Comparison of original and final models across the Middle Mountain synform indicates the need for magnetic material beneath Great Valley sedimentary rocks (Figs. S3–S8) to match the observed magnetic anomalies. Given that Coast Range ophiolite forms the basement of the Great Valley sequence, we interpret the source of the magnetic anomalies to be the Coast Range ophiolite. The Coast Range ophiolite is modeled to be 1–2 km thick; in places it appears to form the base of a symmetric synform (Figs. 3C, 3F, and 3G) or an asymmetric structure that deepens to the northeast (Figs. 3D, 3E, and 3H). Nowhere does the magnetic basement of the synform extend below a depth of 4–5 km, and thus it may not play an obvious role in influencing the location of two earthquake swarms along the eastern margin of the synform, although it is interesting that seismicity along profile 4 (Fig. 3D) appears to be concentrated near the margins of the modeled serpentinite.

The magnetic models have limited resolution for small or deep magnetic bodies. Forward models of thin (<~500 m), horizontal serpentinite bodies at >10 km depth indicate that these bodies produce anomalies that are very broad and very low amplitude and thus difficult to isolate from anomalies produced from adjacent sources (Figs. S3A and S4A). Even vertical bodies near the surface can produce anomalies that may be difficult to resolve, given the height of the magnetic sensor. For example, a thin serpentinite body in the southwestern part of profile 2 produces a magnetic high with an amplitude of <20 nT and a width of ~2 km; the sensor is more than 1 km above the terrain (Fig. S2A). Bodies that are less than 50–100 m in width may not be resolvable with aeromagnetic data, depending on the height of the magnetic sensor and the position and spacing of the flight lines.

Implications for Creep

The magnetic modeling is consistent with a varying amount of serpentinite along the Bartlett Springs fault zone (Fig. 5A), with the northernmost two profiles suggesting that the serpentinite is restricted to the upper 1 km or mostly southwest of the fault zone in contrast to the rest of the models. If the presence of serpentinite predicts creep, our models predict that much of the Bartlett Springs fault zone is creeping, particularly between profiles 3 and 6 and profiles 8 and 14, where the extent of serpentinite in our models extends to 10 km depth or more. Laboratory studies indicate that chemical reactions along contacts between serpentinite and quartzofeldspathic rocks (such as clastic rocks of the Franciscan Complex) reduce the coefficient of friction significantly and stabilize slip (velocity-strengthening behavi or) (Moore and Lockner, 2013). Our models indicate this juxtaposition along profiles 3–9, whereas in profiles 10–14 this juxtaposition exists only locally at depth within the fault zone.

Two published geodetic models of creep along the Bartlett Springs fault zone differ primarily in the extent of creep at depth. While the Lienkaemper et al. (2014) model predicts the Bartlett Springs fault zone to be locked at depths ranging from 0 km to 10–11 km (Fig. 5B), the Murray et al. (2014) model predicts that the fault is creeping everywhere below a depth of 5 km (Fig. 5C). Our model results are generally more consistent with the geodetic model of Murray et al. (2014). They do not predict a fully locked zone between 55 km and 85 km along the fault southeast of Round Valley (profiles 7–10) as modeled by Lienkaemper et al. (2014). However, our model results also do not support a locked zone in the upper 5 km at 45–65 km along the fault southeast of Round Valley (profiles 6–8) as per Murray et al. (2014). Instead, our model results for profiles 6–8 predict creep in the upper 5 km, which appears to be consistent with creep rates derived from InSAR in this region (Xu et al., 2021).

Actively creeping faults are commonly associated with small earthquakes (M < 3), some of which appear to repeat regularly in time and location (Nadeau et al., 1995) and have been used to infer fault slip rates (Shakibay Senobari and Funning, 2019). These earthquakes are interpreted to result from small locked patches within a fully creeping fault (see Harris, 2017, their figure 4A). Such locked patches within the serpentinite are likely, given that this unit incorporates blocks of mafic rocks and other rock types. Such locked patches may also be localized along the margins of the serpentinite where it is juxtaposed against more rigid rock types. Comparison of modeled serpentinite bodies with seismicity suggests that there is some correlation between the modeled extent of serpentinite and microearthquakes (Fig. 3). Note that seismicity was not used to constrain the geophysical models. Seismicity (M > 1; NCEDC [2021] using the methods of Waldhauser and Schaff [2008] and Waldhauser [2009]) projected onto the profiles mostly coincides with serpentinite modeled in or bounded by the Bartlett Springs fault zone, except along profiles 1 and 8 (Figs. 3A and 3H). A cluster of seismicity is located along profile 2 at the intersection of a serpentinite body with the Bartlett Springs fault zone at a depth of ~8 km (Fig. 3B), whereas seismicity is located within or bounds serpentinite bodies along profiles 4–6 and the upper part of 7 (Figs. 3D3G). A handful of microearthquakes are located near the serpentinite contact along profile 9 (Fig. 3I) and within the magnetic basement along profiles 10–14 (Figs. 3J3N). Seismicity is more abundant southwest of the mapped trace of the Bartlett Springs fault zone along the four southernmost profiles (Figs. 3K3N), all close to modeled serpentinite or ophiolitic basement, except for profile 14, where seismicity is located above and aligned with the southwestern edge of ophiolitic basement. This may imply a transfer of stress in this region to a different splay of the fault zone that lies to the southwest or that these earthquakes accommodate a small component of compression across the fault zone (Melosh et al., 2022).

Comparison of the creep data set compiled by Johnson et al. (2022a, 2022b; Fig. 5, top panel) with geodetic models of Lienkaemper et al. (2014; Fig. 5B) and Murray et al. (2014; Fig. 5C) indicates the difficulty of assessing the variation of creep along the Bartlett Springs fault zone due to the different spatial and temporal scales of alinement, GPS, and InSAR data and apparent noise indicated by apparent left-lateral creep measurements. The locked zone in the upper 5 km modeled by both Murray et al. (2014) and Lienkaemper et al. (2014) coincides with creep rates as high as 4 mm/yr as derived from InSAR (Xu et al., 2021). Incorporating information about rock types juxtaposed across the fault may provide insight into the spatial pattern of creep. However, aeromagnetic data are not capable of resolving thin (<50–100 m) magnetic bodies, highlighting the limitations of our approach.

The presence of serpentinite is not the only factor that promotes creep, as indicated by a cursory examination of magnetic anomalies along the Maacama fault zone, where prominent magnetic highs that characterize the fault zone south of Hopland are absent to the north. North of Hopland, surface creep rates exceeding 4 mm/yr are well documented near Ukiah and Willits (McFarland et al., 2017). This may reflect the limited resolution of the aeromagnetic data to resolve narrow magnetic bodies, but other factors likely promote creep along the Maacama fault zone, as discussed in Lienkaemper et al. (2014). The Maacama fault zone north of Hopland has a trend that is clockwise to the relative plate motion, resulting in fault-normal extension that may promote creep. However, fault obliquity to relative plate motion is not always a reliable guide to assess the occurrence of creep; the central San Andreas fault has a trend that is counterclockwise relative to relative plate motion (Lienkaemper et al., 2014, their figure 2a) resulting in fault-normal compression, yet this stretch of the fault is one of the best-known and best-documented creeping faults in the world.

Implications for Regional Structure

Our results support a nearly continuous extent of serpentinite within the fault zone to depths of >10 km along profile 5 near Lake Pillsbury (Fig. 3E). Although the aeromagnetic data do not resolve a thin lens of serpentinite in the upper 400 m, magnetic data collected by boat on Lake Pillsbury do show a linear magnetic high that suggests a thin lens of serpentinite at or near the lake bottom (Langenheim et al., 2007a). An interesting implication of the magnetic model is that it documents a pathway for antigorite, a high-temperature serpenti nite mineral stable at depth, to ascend to the surface in fault gouge at Coyote Rocks about 1 km north of Lake Pillsbury (Moore et al., 2018). The presence of this mineral in the fault gouge at Lake Pillsbury may have been facilitated by the releasing step in the fault zone. Silica-carbonate sinter deposits are locally exposed along the Logan Spring fault, a strand of the Bartlett Springs fault zone ~5 km southwest of Coyote Rocks (Ohlin et al., 2010), indicating a shallow hydrothermal system that could be an alternative source for high-temperature serpenti nite. However, the absence of hydrothermal alteration just north of Lake Pillsbury argues against this interpretation for the presence of antigorite in the Bartlett Springs fault zone.

The presence of significant magnetic material in the subsurface west of exposed Coast Range ophiolite at the Coast Ranges–Sacramento Valley interface provides some support for the source of the serpentinite in the Bartlett Springs fault zone being deeply sourced by wedging of Coast Range ophiolite or Great Valley ophiolite beneath the Franciscan Complex. The three magnetic models along or approximately along profile 5 (Fig. 4) show different possible extents of magnetic bodies west of exposed Coast Range ophiolite at the Coast Ranges–Sacramento Valley interface to account for the westward decrease in aeromagnetic values, but all show magnetic material at depths of 5 km or deeper that extend either beneath the Bartlett Springs fault zone (Fig. 4C) or to within 10 km of the fault zone (Figs. 4A and 4B). The lack of a high-velocity body (Fig. 4D) corresponding with these modeled extents of magnetic material west of the Coast Range ophiolite supports the interpretation that the magnetic material is mostly serpentinite.

In addition to the intermediate-velocity (~5.5 km/s) magnetic material extending west of the exposed Coast Range ophiolite at the Coast Ranges–Sacramento Valley interface, Godfrey et al. (1997) also modeled as magnetic a deep high-velocity (~7 km/s) body that dips shallowly eastward beneath the Bartlett Springs fault zone (Fig. 4B). It is possible that this slab is a source for the serpentinite within the Bartlett Springs fault zone. However, their modeled body produces a broad aeromagnetic high with an amplitude of ~100 nT along the western part of the profile that does not obviously match the observed aeromagnetic anomalies along profile 5 (Fig. 4B). Furthermore, such a broad magnetic feature is not evident in the aeromagnetic or magnetic potential data (Figs. 2B and 2E), and thus this body does not appear to be the source of the serpentinite in the Bartlett Springs fault zone or, for that matter, to be as magnetic as modeled by Godfrey et al. (1997).

The exact geometry of the serpentinized Great Valley ophiolite that extends westward beneath the Franciscan Complex is incompletely constrained by the aeromagnetic data. Long-wavelength anomalies can, in theory, be fit by deep or shallow sources. However, our models support the interpretation of a deep (>10 km depth) body of serpentinized ophiolite that extends beneath the Franciscan Complex (Fig. 4A), rather than the shallower body as modeled by Godfrey et al. (1997) above 10 km depth (Fig. 4B). Although the shallower magnetic body fits in general the long westward decrease in magnetic values along profile 5, the predicted magnetic anomaly curve has a bench in the gradient along the profile between 20 km and 40 km coincident with the western edge of the shallow body, which is not evident in the observed magnetic data. The dipole nature of magnetic sources requires that if the sources are shallow, (1) the upper surfaces of these bodies be extremely smooth and precisely shaped and (2) their magnetic properties vary only over distances comparable to the width of the gradients. We argue that these constraints are too restrictive to explain the gentle, observed magnetic gradient over a length of >75 km and a width of 25 km by using sources in the upper 5 km.

Our modeled geometry of the serpentinite beneath the Franciscan Complex has implications for the magnitude of tectonic wedging along the Coast Ranges–Sacramento Valley interface. If the deep serpentinite is Great Valley–Coast Range ophiolite, this implies a large magnitude of post-Cretaceous crustal shortening that has been accommodated beneath the Coast Ranges. Our modeled horizontal extents of serpentinite west of the Coast Ranges–Sacramento Valley interface beneath the Franciscan Complex range from ~10 km to 30 km and imply at least that amount of crustal shortening. Interestingly, the westernmost extent of the deep serpentinite appears to extend to near or at the Bartlett Springs fault zone along profiles 6, 7, 8, 10, 12, and 13 (Figs. 3F3H, 3J, 3L, and 3M), whereas along profiles 9, 11, and 14 it appears to extend beyond the fault zone. Along profile 14, a cluster of seismicity southwest of the mapped fault traces is aligned with the western margin of the deep body. These spatial relationships may indicate influence of this deep serpentinized ophiolitic body on active structures in the current predominantly strike-slip regime.

Influence of bedrock structure on younger deformation is also implied from the spatial relationship of Quaternary fault traces and magnetic anomalies northwest of Clear Lake. Quaternary fault traces of the Collayomi and other faults mapped along the northeastern side of Clear Lake coincide with northwest-trending magnetic anomalies. The Quaternary traces terminate within young deposits, yet their associated magnetic anomalies continue to the northwest along faults mapped within the Mesozoic bedrock. These faults may have been active during the Quaternary, as suggested by the earthquake swarms along the eastern margin of the Middle Mountain synform (Figs. 3C3E). Robust evidence of Quaternary slip on these structures is limited by the paucity of suitably aged deposits and difficult access; however, this spatial relationship of faults and magnetic anomalies may prove to be an area to investigate with other types of analyses, such as geomorphic analyses to assess Quaternary deformation.

Implication for Fault Offset along the Bartlett Springs Fault Zone

Correlations of magnetic anomalies across strike-slip faults can provide estimates of cumulative fault offset. Based on older aeromagnetic data, Langenheim et al. (2007b) posited ~8–9 km of apparent right-lateral offset near Lake Pillsbury. This estimate seemed plausible, given a similar estimate of offset of the Eel River drainage (Langenheim et al., 2007a). This offset amount, however, is considerably lower than the 37–50 km of offset proposed from correlations of distinctive lower Cretaceous Great Valley sequence strata and a restoration of transpressional folds that trend obliquely to the Bartlett Springs fault zone (McLaughlin et al., 2018). This discrepancy can be explained by attributing the lower estimate of offset of the Eel River drainage to displacement by the Bartlett Springs fault zone proper (the current transform margin) and the higher estimate to incorporation of right-lateral shear within the subduction-zone margin prior to the development of the transform margin as the Mendocino triple junction migrated north after 5 Ma. However, the offset magnetic anomalies near Lake Pillsbury should reflect the total right-lateral offset before and after development of the transform margin. These magnetic anomalies may not be reliable offset markers because they are based on older aeromagnetic data and serpentinite appears to be focused along and within the fault zone. Magnetic anomalies west of the Bartlett Springs fault zone near profiles 1 and 2 appear to be truncated, but no equivalent magnetic anomalies are evident 8–9 km or 37–50 km to the east side of the fault.

Another method to estimate offset along a strike-slip fault uses the length of basins developed within releasing steps as a proxy for offset. Gravity anomalies indicate that the basins developed within right steps in the Bartlett Springs fault zone at Round Valley, Lake Pillsbury, and Cache Valley are ~7 km in length. These estimates are somewhat lower than the offset of the Eel River drainage, but the length of the basin at Lake Pillsbury may not reflect the total displacement. The actively creeping trace of the fault bisects the basin, reflecting a structural reorganization in which a right step evolved to a straight trace. A different structural reorganization applies to the basin beneath Round Valley, where the right-stepping geometry appears to have been abandoned, with the active trace making a left (compressional) step as it enters the basin from the southeast. This part of the fault zone consists of multiple strands, which may also contribute to a higher Quaternary offset than the length of the pull-apart basin. Lastly, the width of the Cache Valley basin as defined by both the gravity low and extent of Cache Formation is greater than that of the right step in the Quaternary fault strands, suggesting the step may be narrowing with increased displacement. The estimates based on basin length appear to support the original correlation of magnetic anomalies across Lake Pillsbury, whereas it is difficult to reconcile these estimates with earlier estimates of 37–50 km.

New integrated geologic and geophysical models across the Bartlett Springs fault zone provide estimates of the distribution of serpentinite along the fault zone and at the interface of the Coast Ranges and the Sacramento Valley. The models in general show more serpentinite within the Bartlett Springs fault zone than predicted by geologic exposures. Given that serpentinite is commonly associated with fault creep, these models would predict a variable amount of creep at depth and are more consistent with the geodetic model of Murray et al. (2014) than the model of Lienkaemper et al. (2014), suggesting a possible reduction in the maximum amount of slip from a single earthquake rupture on the fault zone. We note that our approach is limited by the inability of aeromagnetic anomalies to delineate thin (<50–100 m) bodies at distances of 1 km or more from the magnetic sensor.

The models also show a substantial volume of serpentinite extending west of the Coast Ranges–Sacramento Valley interface, consistent with previous studies (Jachens et al., 1995; Godfrey et al., 1997). Although the exact geometry of serpentinite cannot be constrained by these models, the modeled serpentinite extends 10–30 km beyond the easternmost exposures of Coast Range ophiolite, implying a similar amount of crustal shortening since the Cretaceous. The serpentinite beneath the Lake Pillsbury region, whether of Coast Range or Great Valley ophiolite affinity, is at least 5 km deep (in our preferred model it is >10 km deep) and is likely the source of antigorite, a high-temperature serpenti ne mineral stable at depth found in fault gouge there.

1Supplemental Material. Includes original, preferred, and alternate geophysical models across the Bartlett Springs fault zone and a figure showing the distribution of gravity measurements in the region. Please visit to access the supplemental material, and contact with any questions.
Science Editor: Andrea Hampel
Associate Editor: Craig H. Jones

We gratefully acknowledge support for this study from the National Cooperative Geologic Mapping Program of the U.S. Geological Survey and Pacific Gas and Electric Company. Seismicity data in this study were accessed through the Northern California Earthquake Data Center (NCEDC; Constructive reviews by Michael Mitchell, Richard Saltus, Sarah Titus, and Diane Moore were very helpful in clarifying and refining the presentation of this research. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Gold Open Access: This paper is published under the terms of the CC-BY-NC license.