Were changes in stress state responsible for the 2019 Ridgecrest, California, earthquakes?

The 2019 Ridgecrest, California, earthquake sequence comprised an M6.4 event followed 34 hours later by an M7.1 mainshock, occurring within the Eastern California Shear Zone (ECSZ). A key question is whether changes in the regional and local stress state associated with these earthquakes influenced rupture initiation, propagation, and subsequent seismic hazard. Prior Coulomb stress transfer studies suggested that historic large earthquakes and the M6.4 event could have loaded the future rupture areas, but such physics-based forecasts have struggled to outperform statistical models due to uncertainties in fault geometries. This study proposes a statistics-based approach using the Gutenberg–Richter b-value to infer spatio-temporal stress changes before, between, and after the M6.4 and M7.1 events, and relates these to Coulomb stress modeling and aftershock decay characteristics to assess implications for future seismicity, particularly near the Garlock fault.

Previous work used Coulomb stress transfer to link stress increases to subsequent events, including loading from the 1872 Owens Valley, 1992 Landers, and 1999 Hector Mine earthquakes, and stress transfer from the M6.4 to the M7.1 Ridgecrest hypocenter. However, predictive skill has been limited due to unknown receiver faults. The b-value of the Gutenberg–Richter law is inversely related to differential stress, a relationship supported by laboratory and field observations. Studies have also shown that microseismicity patterns and b-value mapping can forecast rupture areas and reflect asperities, frictional properties, and interface locking. Temporal decay of aftershocks follows the Omori–Utsu (OU) law with spatial variability in the p-parameter related to stress relaxation rates, and spatial organization metrics (η) have been used to identify foreshock-type clustering preceding large events in southern California. These strands of literature motivate combining b-value mapping, Coulomb stress modeling, OU/ETAS analyses, and spatial clustering diagnostics to evaluate stress evolution and triggering in the Ridgecrest sequence.

Data: Southern California Seismic Network (SCSN) catalog, focusing on events since 1980 within the Ridgecrest region and depths <20 km. Relocated hypocenters (Hauksson–Yang–Shearer) were used for accurate mainshock depths (M7.1 at ~1.9 km; M6.4 at ~11.8 km). Completeness magnitude Mc was evaluated, with typical levels ~1.5±0.5 since the 1980s, and time-varying Mc assessed after M6.4 to ensure homogeneity for temporal analyses (using M≥3 between M6.4 and M7.1).

b-value estimation and mapping: The Entire-Magnitude-Range (EMR) technique in ZMAP was used to jointly estimate a, b, and Mc, applying maximum likelihood above Mc and modeling detection rates below Mc. Spatial grids (0.01°×0.01° for maps; 0.5×0.5 km for cross-sections) were sampled within r=5 km (tested 5–12 km to select optimal coverage vs. smoothing). At least 20 events per node and good GR fits were required. Uncertainties in b were obtained via bootstrapping; differences in b between regions were tested using the Utsu test (log10P≤−1.3 indicates significance). Sensitivity tests increased local Mc by +0.2 and +0.5 to verify stability of spatial/temporal b patterns.

Temporal analyses: Moving-window b-value time series (100-event windows) were constructed around the M6.4 and M7.1 hypocentral regions at selected depths. Aftershock decay between M6.4 and M7.1 was modeled using the OU law (λ=k(c+t)^(−p)) with maximum-likelihood estimation and bootstrap uncertainties for p, using M≥3 events to ensure recording homogeneity, and subdividing the rupture into northern (including M7.1 hypocenter) and southern areas. ETAS modeling (SASeis2006) provided a check on OU-based inferences using target and precursory time intervals and confirmed OU results.

Coulomb stress modeling: Static stress changes were computed with the USGS Coulomb software in an elastic half-space (Poisson’s ratio 0.25, Young’s modulus 8×10^5 bar, shear modulus 3.3×10^5 bar, friction μ=0.4). Source faults included the finite-fault models of the M6.4 (left-lateral, strike 226°, dip 90°, rake ~6°) and M7.1 (right-lateral, strike 136°, dip 90°, rake ~176°) earthquakes (Xu et al.), and M≥4.5 events between M6.4 and M7.1 using SCSN moment tensors converted to realistic rectangular sources. Stresses were resolved on receiver faults corresponding to the M7.1 geometry and on generic left-lateral/right-lateral planes to assess stress changes at 8–12 km depths.

Spatial organization (η): Following Lippiello et al., η=R−1/R1 was computed from M≥3 seismicity between M6.4 and just before M7.1 along cross-sections of the M7.1 and M6.4 ruptures to diagnose seismic concentration (η>1) or dispersion (η<1). Robustness was evaluated for n=15–35 events; uncertainties were estimated with a bootstrap procedure resampling distances to compute ση and error bars for η time series.

Quality control: Catalog homogeneity and Mc over time were quantified using EMR and compared against MAXC and GOF methods, with EMR yielding more conservative Mc. Sensitivity checks on sampling radius and Mc ensured stability of key patterns.

• Before the M6.4 earthquake, a pronounced low b-value zone (b≈0.6–0.7) at 7–13 km depth surrounded the future M6.4 hypocenter (depth 10.7 km), while shallow depths (0–7 km) did not show such a low-b patch. b-values around the M6.4 epicentral area decreased gradually since ~2010 to ~0.7.
• Cross-sections showed low b (b<0.9) near the M6.4 hypocenter and high b (b>1) near the eventual M7.1 hypocenter before M6.4, indicating a weakly stressed barrier near the future M7.1 site that impeded M6.4 rupture propagation.
• In the 34 hours between M6.4 and M7.1, b increased near the M6.4 hypocenter and decreased near the M7.1 hypocenter (to ~0.66), consistent with stress relaxation at the M6.4 site and stress loading at the M7.1 site; differences in b were statistically significant (Utsu test).
• Coulomb stress calculations indicate that the M6.4 event plus its subsequent M≥4.5 events increased Coulomb stress near the M7.1 hypocenter by about 2 bars (≈1 bar for the M6.4 event alone), promoting failure.
• Aftershock decay analysis (M≥3) showed OU p-values smaller in the northern area (including the M7.1 hypocenter) than in the southern area, indicating slower stress decrease in the north; ETAS modeling confirmed these trends.
• Spatial organization (η) immediately before M7.1 revealed seismic concentration (η≈1.5) near the M7.1 hypocenter, while other regions showed η≈1 or lower; results were robust for n=15–35 with quantified uncertainties.
• Post-M7.1, low b-values (b<0.9) emerged in a zone near, but not on, the Garlock fault, spatially complementary to high-slip patches (4–5 m peak slip near M7.1 hypocenter). This low-b zone lies in areas of modeled stress promotion for both left-lateral (M6.4-type) and right-lateral (M7.1-type) receiver faults at depths ~8–12 km.
• Over the first 8 months post-M7.1, ~30,000 events with M≥1 occurred, including >90 with M≥4. The identified low-b zone exhibited a decreasing b trend toward ~0.8, suggesting increasing stress in an unruptured volume that could influence the Garlock fault if further loaded.

The combined statistical (b-value mapping, OU/ETAS decay analysis, spatial clustering η) and physics-based (Coulomb stress transfer) analyses coherently indicate that stress redistribution from the M6.4 event eroded a previously weakly stressed barrier near the M7.1 hypocenter, increasing local stress (≈2 bars) and fostering conditions for M7.1 nucleation. Low b-values prior to both M6.4 and M7.1 events suggest high-stress nucleation zones are a common feature of this sequence. Post-M7.1, the emergence of a low-b patch near the Garlock fault, not coincident with high-slip regions, implies elevated stress in adjacent unruptured volumes. Modeled Coulomb stress changes on both left-lateral and right-lateral receiver faults at 8–12 km depths align with this observation, indicating enhanced failure potential in that zone. The concordance among independent indicators (b, p, η, and Coulomb) strengthens the interpretation that temporal-spatial b-value variations track evolving stress states and can highlight regions of heightened seismic hazard in the ECSZ. While these results inform qualitative hazard assessments and potential fault interactions (including possible influence on the Garlock fault), quantitative forecasting remains challenging.

This study demonstrates that spatio-temporal b-value mapping, integrated with Coulomb stress transfer and aftershock decay/spatial organization analyses, provides insight into evolving stress conditions during the 2019 Ridgecrest sequence. Low b-value zones anticipated the nucleation of both the M6.4 and M7.1 events; stress transfer from M6.4 and subsequent events likely advanced the M7.1 rupture. Following M7.1, a low-b, unruptured zone near the Garlock fault coincides with areas of modeled stress promotion, suggesting potential for future activation. Continued monitoring of b-values, alongside seismological and geodetic observations, is recommended to track stress evolution and inform seismic hazard in the ECSZ. Future work should pursue quantitative risk assessment, for example via nowcasting frameworks, and refine catalog completeness and detection to resolve the smallest-event contributions to b-value signals.

• Quantitative predictive power of b-value mapping is not established; low-b patches may persist or dissipate without producing large earthquakes.
• Coulomb stress models depend on assumed source and receiver fault geometries; unknown or unmapped faults introduce uncertainty.
• Catalog completeness (Mc) varies temporally, especially immediately after large events; although EMR and M≥3 thresholds mitigate biases, residual under-detection may influence b and p estimates.
• Spatial resolution depends on sampling radius and event density; smaller radii reduce coverage, larger radii smooth heterogeneity.
• Depth assumptions (e.g., analyses at 8–12 km) and choice of friction coefficient (μ=0.4) affect stress-change estimates.
• Spatial organization metric η sensitivity to sample size (n) requires bootstrap uncertainty assessment; very small n produces unstable results.