3D architecture and complex behavior along the simple central San Andreas fault

Fault zones are multiscale systems with a localized slip-accommodating core, a fractured damage zone, and surrounding host rock. Slip occurs both seismically (from microearthquakes to large events and tremor) and aseismically (steady creep, slow slip, afterslip). The partitioning of seismic and aseismic slip is commonly quantified by fault coupling (slip deficit rate relative to long-term slip rate), with higher coupling generally associated with lower surface creep, fewer repeaters, lower b-values, and more clustered seismicity. Despite this framework, important gaps remain: most models assume uniform slip direction along faults, many analyses emphasize statistical seismology rather than physical processes, and interactions between on-fault and off-fault deformation are often neglected. Small-earthquake focal mechanisms can resolve fine-scale slip directions, stress orientations, and distinguish on- vs off-fault deformation, providing constraints beyond surface geodesy. The CSAF, with a long-term slip rate of ~33–35 mm/yr, variable surface creep (≤30 mm/yr), abundant microseismicity and repeating earthquakes, and few large historic ruptures on this segment, offers an ideal laboratory. This study integrates dense seismic catalogs, focal mechanisms, repeating earthquakes, geodetic surface displacements, and forward modeling to map 3D fine-scale structure and kinematics, test how coupling heterogeneity governs complex deformation, and assess implications for seismic potential along the CSAF.

Prior work established the internal structure of fault zones and the coexistence of seismic and aseismic processes, linking spatial variations in fault coupling to observable metrics such as surface creep, repeating earthquake recurrence, and seismicity characteristics. Geodetic inversions have suggested along-strike variations in slip deficit on the CSAF, but assumptions of uniform slip direction and limited sensitivity to depth-dependent heterogeneity constrain resolution. Studies on stress rotations near the San Andreas Fault invoked mechanisms including high pore fluid pressures in narrow damage zones, weak faults and/or weak lower crust, and frictional resistance under strong-fault hypotheses; however, interpretations remain debated. Repeating earthquakes have been used to estimate creep rates at depth and to illuminate time-dependent aseismic slip. Recent advances in focal mechanism determination (e.g., relative radiation pattern methods) enable high-quality mechanisms for small events, offering improved constraints on on-fault slip directions and off-fault stress. This study builds on these foundations by integrating large catalogs of small-event focal mechanisms and repeaters with boundary element modeling to jointly evaluate on-fault creep rate/direction and off-fault stress and faulting style at fine spatial scales.

Data integration and study area: The analysis targets a ~190-km-long, ~2-km-wide CSAF segment (1984–2015), integrating 75,164 Mw≥1 earthquakes, 145 Mw≥4 events, and 355 repeating earthquake sequences. Repeating earthquakes delineate main fault strands and high creep zones; Mw≥4 events indicate higher coupling. Fault geometry: The horizontal position of the main fault strands is first constrained by repeating earthquakes. Earthquakes within 1 km epicentral distance are used to estimate 3D geometry via principal component analysis (PCA) in 15 km-long by 15 km-deep segments stepped every 1 km, yielding nearly vertical planes with strike N120°E–N140°E. Focal mechanisms: Small-event mechanisms are computed using the REFOC algorithm, which uses first-motion polarities and S/P amplitude ratios to obtain initial solutions, refined by inter-event relative P- and S-wave amplitude ratios for events within 3 km hypocentral distance. Waveforms from NCEDC/NCSN are band-pass filtered 1–10 Hz; data windows and SNR thresholds are applied; velocity models assume Vp/Vs=1.732. This yields 52,211 high-quality mechanisms (≥8 polarities, uncertainty <35°). For repeating sequences, median polarities and S/P ratios across events and stations are used with the HASH algorithm to obtain robust sequence mechanisms (386 sequences). On-fault kinematics from repeaters: For each repeating sequence, the nodal plane closest in strike to the main fault is selected to represent on-fault slip. Dip and rake define local slip direction; a sign convention accounts for which side is hanging wall. Repeater occurrence rates are converted to slip using d=10^α M0^β with α=2.36, β=0.16, and log(M0)=1.6Mw+15.8. Creep rates and directions are averaged per 3×3 km fault patch (expanded to 9×9 km if sparse), with uncertainties from 200-sample bootstrap resampling. Off-fault stress and faulting style: Using 24,915 Mw≥1 mechanisms within 2 km of the fault (≥8 polarities, <35° uncertainty), stress tensors are inverted per 3×3 km patch (expanded if sparse) via STRESSINVERSE (iterative joint inversion with randomized perturbations; 200 realizations) to obtain SHmax orientation relative to the fault. Faulting style is summarized as %rake>0 (fraction of oblique-reverse events; complementary to oblique-normal). Fault slip modeling: Interseismic deformation is simulated with Poly3D (boundary element method) in a uniform elastic half-space (Poisson’s ratio 0.25, shear modulus 30 GPa). The shallow seismogenic fault (0–15 km) is modeled on a triangular mesh and analyzed on a 3×3 km grid; boundary conditions are zero-slip (locked patches) or zero-shear-stress (freely slipping). Model A: mostly freely slipping shallow fault between −17 and 166 km along strike, driven by steady deep creep of 34 mm/yr from 15 to 2000 km depth, with small locked patches near San Juan Bautista (SJB), Bitterwater (BW), and Parkfield (PK), and fully locked at the segment ends. Model B: same geometry and end locking but without internal locked patches (fully creeping interior). Modeled outputs include on-fault creep rates and directions, and off-fault SHmax orientation at 1.5 km NE of the fault. Comparisons: Modeled on-fault creep rates and directions are compared to repeater-derived values; surface creep is compared to InSAR LOS rates plus creepmeter/alignment arrays; modeled off-fault SHmax is compared with stress inversion results. Correlations and misfits are quantified; sensitivity tests assess effects of locked patch size/depth, grid resolution, and fault dip deviations.

• The CSAF main fault is nearly vertical with along-strike variations in strike from ~N120°E to ~N140°E. Repeating earthquakes predominantly delineate the main fault and high-creep zones, while Mw≥4 events cluster near SJB and PK transition zones and locally near BW, indicating higher coupling there.
• A simple fault coupling model (Model A) with small locked patches embedded in an otherwise freely creeping shallow fault driven by 34 mm/yr deep creep reproduces first-order observations: • Modeled vs repeater-derived on-fault creep rates: correlation coefficient ≈ 0.47. • Modeled vs repeater-derived on-fault creep directions: correlation coefficient ≈ 0.55. • Modeled vs observed surface creep rates (InSAR/creepmeter/alignment arrays): correlation coefficient ≈ 0.91. • Modeled off-fault SHmax vs stress from focal mechanisms: correlation coefficient ≈ 0.33.
• Locked patches produce localized reductions in creep rate and rotations in slip direction; around SJB and PK, observed creep rates are <25 mm/yr at depth (repeaters) and <15 mm/yr at surface. In the central creeping section (Melendy Ranch to SAFOD), creep approaches the deep rate (~34 mm/yr).
• A deep locked patch near Bitterwater (BW) is supported by abrupt spatial changes in vertical slip components (creep direction), elevated angles between SHmax and fault strike to the NW of BW and reduced angles to the SE, and the presence of Mw≥4 sequences—features not evident from creep rate alone.
• Off-fault faulting style correlates with stress orientation: there is a significant negative correlation between θ (angle of SHmax to fault) and %rake>0 (fraction of oblique-reverse events), with r ≈ −0.43. High θ favors oblique-normal events on high-angle secondary faults; low θ favors oblique-reverse events; in all cases, small faults tend to slip in fault-parallel directions.
• Model B (no internal locked patches) matches creep rates only in a first-order sense but fails to reproduce observed creep directions and stress orientations, highlighting the sensitivity of vertical slip components and stress rotations to fine-scale coupling heterogeneity.
• Seismic potential: Assuming deep slip of 34 mm/yr, the modeled moment deficit rate is ~1.32×10^18 N·m/yr with ~1.48×10^18 N·m/yr released aseismically; over 150 years, stored moment is equivalent to Mw 7.5. If partially coupled areas catch up only via aseismic slip (e.g., afterslip), the deficit rate is ~4.04×10^17 N·m/yr (Mw ~7.2 over 150 years). Locked patches distributed within the creeping section are likely to rupture as moderate earthquakes, with surrounding slip taken up by transient afterslip of major ruptures on adjacent SAF segments.
• The observed stress rotation across the SAF is unlikely caused by large-scale aseismic creep alone; instead, it is governed by fine-scale heterogeneity in coupling and mixed seismic/aseismic processes in a narrow, mechanically weak fault zone.
• The narrow weak zone and localized stress concentrations favor right-lateral ruptures aligned with the main fault for moderate-to-large events on the CSAF, while smaller events display more diverse focal mechanisms including high-angle subsidiary faults.

The study demonstrates that the CSAF’s complex seismic and aseismic behavior along a geometrically simple, nearly vertical fault can be explained by fine-scale heterogeneity in fault coupling. Integrating small-earthquake focal mechanisms with repeaters and geodetic data enables resolution of depth-dependent creep rates and directions, as well as off-fault stress orientations and faulting styles, that are not uniquely constrained by surface displacements alone. The success of Model A in reproducing observed on-fault and off-fault kinematics—versus the failure of a fully creeping interior (Model B) to match slip directions and stress orientations—highlights the diagnostic value of vertical slip components and stress patterns for identifying locked asperities. The identification of a deep locked patch near BW, supported by coordinated signals (creep direction rotation, SHmax variation, and local Mw≥4 activity), refines previous geodetic inferences that suggested deep coupling in the central creeping section without pinpointing location. From a hazard perspective, distributed locked patches embedded in a generally weak, creeping fault suggest frequent moderate events and afterslip-mediated release of surrounding slip deficits; while the BW asperity could reduce the barrier effect of the creeping section and potentially enable through-going ruptures, there is no direct evidence that great ruptures have crossed the CSAF creeping section. The analysis further informs the long-standing stress rotation debate: observed rotations near the SAF are better explained by fine-scale, heterogeneous deformation within a narrow weak zone than by large-scale creep alone or solely by strong/weak fault end-member models. Off-fault kinematics reveal that even when secondary faults strike at high angles to the main fault, their slip is often fault-parallel, indicating stress-guided deformation within a mechanically distinct zone that promotes linkage and multi-scale rupture complexity.

By combining high-resolution earthquake catalogs, improved focal mechanism determination for small events, repeating earthquake analyses, and boundary-element modeling, this study maps fine-scale 3D architecture and interseismic kinematics of the CSAF. A simple coupling model with small locked patches embedded in a narrow, mechanically weak and largely creeping fault zone reconciles observed variations in creep rates and directions, off-fault stress orientations, and earthquake faulting styles. The approach reveals a deep locked patch near Bitterwater and refines the understanding of stress rotations and multi-scale rupture behavior along the CSAF. These insights inform seismic hazard by quantifying moment deficit accumulation and suggesting that moderate earthquakes and afterslip are likely to dominate deficit release within the creeping section, while the possibility of through-going major ruptures remains uncertain. Future work should constrain absolute stress levels and fault strength by leveraging time-dependent changes in small-earthquake focal mechanisms around moderate events (e.g., Parkfield 2004), incorporate more realistic 3D geometry, finer meshing, and variable rheology, and extend the methodology to other transform and subduction zones to improve assessments of fault coupling, stress fields, and earthquake potential.

• Model simplifications: The boundary-element models impose zero-slip or zero-shear-stress conditions without explicit frictional laws; fault geometry is assumed nearly vertical within segments, and the grid resolution (3×3 km analysis; ~1–1.12 km mesh edges) may under-resolve sharp gradients near asperity edges.
• Sensitivity to parameterization: Modeled creep direction rotations are smaller (±6°) than observed from repeaters (±30°), indicating sensitivity to locked patch size/depth, adjacent freely creeping area, grid resolution, and deviations from vertical dip not fully captured by the first-order model.
• Data uncertainties and selection: Focal mechanism uncertainties (up to ~35°) and nodal-plane selection (assuming the plane closest in strike to the main fault) can bias inferred slip directions and ignore left-lateral high-angle faults. Stress inversions require ≥100 mechanisms per patch, potentially smoothing heterogeneity compared to sparser repeater constraints.
• Temporal variability: Differences between modeled and observed surface creep (e.g., MR–BW) may reflect temporal oscillations not represented in steady interseismic models.
• Incomplete rupture history: Interpretation of seismic potential (e.g., through-going ruptures) is limited by the absence of direct evidence for great ruptures crossing the CSAF creeping section and by historical catalog incompleteness.