Spatiotemporal Variations of In Situ Vp/Vs Ratios During the 2019 Ridgecrest Earthquake Sequence Suggest Fault Zone Condition Changes

The 2019 Mw 7.1 Ridgecrest earthquake was the largest event in California over the past 20 years. The earthquake was preceded by a sequence of foreshocks. However, the physical processes leading to the mainshock remain unclear. Here, we image the ratios of compressional (P)‐ to shear (S)‐wave velocity (Vp/Vs) in the fault zones and examine the spatial and temporal evolution of near‐source material properties during the Ridgecrest earthquake sequence. We find that the Vp/Vs ratios are spatially homogeneous in the rupture zones, indicating a lack of fault‐zone material difference along strike. We identify an anomalously low Vp/Vs ratio fault patch near the mainshock hypocenter before its occurrence, which returned to the background value after the earthquake. This low Vp/Vs ratio suggests fluid overpressure, which may have facilitated the nucleation of the 2019 Ridgecrest mainshock.


Introduction
The 2019 Ridgecrest earthquake sequence comprises the most significant earthquakes in California for the past 20 years.On 4 July 2019, a moment magnitude (Mw) 6.4 earthquake struck approximately 13 km southwest of Searles Valley, California (Ross et al., 2019).Nearly 34 hr later, an Mw 7.1 earthquake ruptured a complex fault system extending over 75 km to the northwest direction.This event marks the largest earthquake in southern California since the 1992 Mw 7.3 Landers and the 1999 Mw 7.1 Hector Mine earthquakes.Approximately 2,500 aftershocks occurred in between the Mw 6.4 foreshock and the Mw 7.1 mainshock, including an Mw 5.4 earthquake about 4 hr before and 2.4 km southeast of the mainshock (Hauksson & Jones, 2020).The Ridgecrest earthquakes ruptured multiple segments of the fault system, including non-vertical, interlaced orthogonal conjugate faults (Lin, 2020;Shelly, 2020).The close spatiotemporal correlation between the foreshocks and the mainshock suggest a causal triggering relationship.However, the foreshocks did not instantaneously trigger the Mw 7.1 mainshock, and there is an approximately 34 hr delay between the Mw 6.4 foreshock and the mainshock.
The static or dynamic stress perturbation from the Mw 6.4 foreshock appear to be insufficient to trigger the mainshock rupture (Jin & Fialko, 2020;Lozos & Harris, 2020;Taufiqurrahman et al., 2023), suggesting that additional fault zone processes might have helped nucleate the Mw 7.1 mainshock.Foreshocks are often the most accessible observations for inferring mainshock nucleation processes, which are likely due to stress interactions that eventually lead to the triggering of mainshocks (Abercrombie & Mori, 1996;Kato et al., 2012;Ruiz et al., 2014).For example, accelerating aseismic slip events have been hypothesized to nucleate subsequent earthquakes and can also produce foreshocks as by-products during their propagation (Ellsworth & Beroza, 1995).Foreshocks may also result from a sequence of stress-triggering events among themselves, eventually cascading into a larger event known as the mainshock (Ide, 2019;McLaskey, 2019;Meng & Fan, 2021).Direct laboratory observations of multi-scale damage evolution in the failure zone (fault zone) reveal the interplay between various deformation mechanisms; and their competing effects determine the path to macroscopic failures (Renard et al., 2017(Renard et al., , 2018)).Such experiments suggest that fault nucleation and propagation processes exist; however, their evolution depends not only on the states of fault stress but also, more importantly, on the strength conditions of the fault zone, which can evolve rapidly near the earthquake hypocenters.Therefore, resolving the material properties and conditions of the fault zone at the scales of earthquake nucleation is critical for understanding the evolution of the Ridgecrest earthquake sequence.
The ratio of compressional (P)-to shear (S)-wave velocity (V p /V s ) is sensitive to both fault strength and stress state due to its direct connection with Poisson's ratio, fluid content, and pore fluid pressure (e.g., Brantut & David, 2019;Christensen, 1996;Takei, 2002).In fault systems, V p /V s ratios reflect the composition of rocks, the presence of cracks, and the degree of pore fluid saturation (Lin et al., 2022;Liu et al., 2023), all of which directly control earthquake sequences (e.g., Zhu et al., 2020).However, resolving V p /V s ratios at seismogenic depths, where great earthquakes occur, is challenging.Seismic tomography can image V p /V s ratios at depth, but conventional passive imaging techniques are strongly influenced by seismic wave propagation.As a result, these models may be overly smooth when compared to the dimensions of most earthquake ruptures (Tong et al., 2021;Yang et al., 2022).Furthermore, tomographic models often lack the temporal resolution needed to detect rapid spatiotemporal variations that occur during the earthquake preparation stages.In this study, we apply a highresolution estimation method that uses waveform cross-correlation data to calculate in situ V p /V s ratios in the near-source regions of the 2019 Ridgecrest earthquake sequence.We directly resolve fault-zone material properties, complementing existing velocity structure models.Additionally, we synthesize a collection of geophysical evidence to evaluate the faulting environments and discuss the possible physical causes of the Ridgecrest earthquake sequence.

Data
We use over 164 million P-and S-wave differential times, along with the corresponding correlation coefficients, from 1 million event pairs to resolve the in situ V p /V s ratios in the fault zones of the 2019 Ridgecrest earthquake sequence.These waveform cross-correlations and differential times originate from 24,925 relocated earthquakes that occurred between 1 January and 31 August 2019 (Figure 1a; Lin, 2020).The catalog is accurate in both absolute and relative locations, obtained using ray tracing through a three-dimensional (3-D) seismic velocity model (Zhang & Lin, 2014) and a waveform cross-correlation relocation technique (Lin, 2018;Lin & Shearer, 2006).

Methodology
The high-resolution V p /V s ratio estimation method uses differential times from waveform cross-correlation to study spatial and temporal variations of near-source V p /V s ratios in different tectonic regions (Lin & Shearer, 2007).For a compact cluster of events, the waveforms for each pair of events observed at the same station will be similar enough so that the cross-correlation can be computed.This method assumes that the scale length of changes in the V p /V s ratio is greater than the size of the similar event clusters so that the V p /V s ratio can be treated as a constant within the event cluster.The best-fitting V p /V s ratio for the entire cluster can be obtained by fitting all the points of the demeaned differential P-and S-arrival times simultaneously.A hybrid L 1 -L 2 fitting method is adopted for the procedure (Lin & Shearer, 2007), which is more robust to outliers in the observations than the traditional least squares fitting.
The most accurate V p /V s results for real data can be obtained for clusters with a three-dimensional distribution of events (Lin & Shearer, 2007).In order to estimate the spatial distribution of events in each cluster, we use the method of principal component analysis (e.g., Kirschvink, 1980) to compute eigenvalues for the covariance matrix of the earthquake locations for all similar event clusters.Clusters are considered to have nearly spherical distribution if λ 1 /λ 3 ≤ K, where eigenvalues λ 1 ≥ λ 2 ≥ λ 3 , and K is a constant (Lin & Shearer, 2009;Michelini & Bolt, 1986;Shearer et al., 2003).Standard errors in the V p /V s estimates are calculated by applying a bootstrap approach (Efron & Gong, 1983;Efron & Tibshirani, 1991), in which each pair of suitable differential P and S times in the same cluster may be sampled multiple times or not sampled at all.This process is repeated for many subsamples for each cluster and the standard deviation of these subsamples is used as the standard error of the V p / V s ratio.In this study, we use differential times with a waveform correlation coefficient of 0.6 or greater in the calculation and require each event pair have at least eight individual differential times with correlation coefficients of 0.65 or greater.Furthermore, we select the clusters with λ 1 /λ 3 ≤ 25 = K and discard V p /V s ratios with the estimated standard errors greater than 0.03.The standard errors are computed from 100 bootstrap resamples.

Results and Discussion
The V p /V s estimates are most sensitive to materials near the earthquake sources; complex wave propagation effects due to surrounding 3-D seismic structures have minimal impact on the results (Lin & Shearer, 2007;Liu et al., 2023).Theoretically, the resolution can be as high as the spatial and temporal inter-event separation.In this study, we first cluster the earthquakes and apply a sequence of quality control steps before estimating V p /V s ratios to ensure their robustness (see Figures S1 and S2 in Supporting Information S1).As a result, we identify 80 similar event clusters satisfying our selection criteria, comprising a total of 24,082 events.The sizes of these clusters range from 10 to 1,356 events, and they are nearly uniformly distributed across the Ridgecrest fault system.These clusters are centered at depths ranging from the surface to 19 km, and the majority of average interevent distances are less than 3 km.The in situ V p /V s ratios for the 80 clusters are estimated in the range of 1. 2a,and 2b).Example plots of the demeaned differential times are shown for two clusters (see Figure S3 in Supporting Information S1).

Comparison With Tomographic Velocity Model
For comparison, we interpolate the 3-D tomographic models by Zhang and Lin (2014) (Figure S4 in Supporting Information S1) at the earthquake locations to extract both V p and V p /V s values.We then calculate the average V p and V p /V s value in each cluster and compare these average values with our in situ V p /V s ratios (Figures 2c-2f).
The average velocity values from all the seismicity along profiles A-B and C-D, 1.737 for V p /V s and 5.962 km/s for V p , are used as the references for the percentage plots.Both the tomographic V p /V s (Figures 2c and 2d) and V p (Figures 2e and 2f) models are more spatially heterogeneous than the in situ V p /V s ratios in the Ridgecrest fault zones, suggesting material variations from one segment to another.The tomographic V p /V s model (Figures 2c and  2d) have values varying from 1.69 to 1.76, which are generally higher than the in situ results.The only exception occurs near the Airport Lake Fault Zone (ALFZ) (50-60 km distance, below 5 km depth in Figure 2c), where the tomographic V p /V s model has lower ratios of ∼1.66.The tomographic V p model (Figures 2e and 2f) shows great variations along the two profiles with increasing V p values with depth, a pattern which is absent in either the tomographic and in situ V p /V s ratios.
The tomographic V p model suggests significant material variability along the fault strike, which may be related to the complex tectonics of the Eastern California Shear Zone (ECSZ; Dokka & Travis, 1990).However, such complexities are not observed in our in situ V p /V s results.The differences between the in situ V p /V s ratios and the tomographic models may primarily stem from the contrast between the seismic structures surrounding the fault zone and the fault zone material properties themselves (Qiu et al., 2021;Tong et al., 2021;White et al., 2021;Yang et al., 2022).These observed differences suggest that although the Ridgecrest fault system is situated in the complex ECSZ and surrounded by host rocks with strong variations, the fault zones themselves are relatively homogeneous.This is different from the fault zone material segmentation observed in other strike-slip fault systems (Lin et al., 2022;Liu et al., 2023).For example, the westernmost Gofar transform fault, an analogous site to the Ridgecrest fault system, a barrier zone associated with anomalous V p /V s have repeatedly stopped adjacent M6 earthquakes to rupture further (Gong & Fan, 2022;Liu et al., 2023;McGuire et al., 2012).This barrier zone is likely highly damaged with active seawater infiltration, causing material and fault condition contrasts (Kohli et al., 2021;Roland et al., 2012).
The differences between the in situ and tomographic results are likely due to the data sets, inversion techniques, and varying sensitivities to different spatial scales.The tomographic models are constructed using 30 years of body wave travel-time data (Zhang & Lin, 2014) and they represent both the spatial and temporal average of the velocity structures in the study area.In contrast, the in situ V p /V s estimates are obtained using the 2 months aftershock data and are most sensitive to the fault zone structures.In the tomographic inversion, velocity perturbations, instead of absolute values, are solved relative to the initial model, showing strong dependence on the starting values (e.g., Lin et al., 2007Lin et al., , 2010Lin et al., , 2014Lin et al., , 2021;;Zhang & Lin, 2014).The in situ V p /V s estimation method is based on a robust least squares linear fitting and does not rely on the initial value (Lin & Shearer, 2007).In addition, tomographic models are dependent on inversion parameters, such as damping and smoothing factors.Tomographic V p /V s inversions also limit the model perturbations to improve inversion stability and convergence.

Homogeneous Fault Zone Materials Facilitating Earthquake Rupture Propagation
The Ridgecrest fault zones exhibit relative homogeneity in V p /V s ratios (1.71-1.73, Figure 1b, 3-19 km distance in Figure 2a and 3-46 km distance in Figure 2b), particularly where the majority of slip occurred during the Mw 6.4 foreshock and the Mw 7.1 mainshock (Jin & Fialko, 2020;Ross et al., 2019).The Mw 6.4 foreshock was dominated by slip on a northeast-trending, left-lateral fault (Jia et al., 2020;Wang et al., 2020b), where the slip propagated toward the Little Lake Fault Zone (LLFZ, 20-26 km distance in Figure 2a).We find that its rupture arresting boundary correlates with lower V p /V s ratios (1.66-1.69)at the southwestern end of the LLFZ.Similarly, we observe lower V p /V s ratios near the Garlock fault (GF, ∼3 km distance in Figure 2b) and in the ALFZ (46-60 km distance in Figure 2b) along the northwest-trending, right-lateral fault where the Mw 7.1 mainshock halted its rupture propagation (Jia et al., 2020;Wang et al., 2020b).Similar to Lin et al. (2022), we identify anomalous V p /V s ratios in areas where fault segments intersect, terminate, or change orientations (e.g., ALFZ), where the ratios range from 1.52 to 1.89.Intriguingly, the areas surrounding the hypocenters of both the Mw 6.4 and Mw 7.1 earthquakes have sporadic clusters with relatively high V p /V s ratios, peaking at a value of 1.97 for one cluster near the Mw 7.1 mainshock's hypocenter (33-36 km distance in Figure 2b).Similar high V p /V s ratios are also observed near the Mw 5.4 earthquake, albeit with subdued variations.These clusters include earthquakes that occurred both before and after the three major earthquakes, and the anomalous values likely associate with the changes of the fault environment due to the earthquakes (Trugman et al., 2020).The overall homogeneous V p /V s ratio distribution in the Ridgecrest fault system distinctly differs from those observed in the Parkfield segment of the San Andreas Fault in Northern California, as well as the San Jacinto Fault Zone and its adjacent regions in Southern California (Lin et al., 2022).
The homogeneous material within the fault zone may suggest relatively homogeneous fault conditions, and the lack of material segmentation may have encouraged the earthquake rupture propagation, thereby facilitating the development of both the Mw 6.4 foreshock and the Mw 7.1 mainshock (Wibberley et al., 2008).The 2019 earthquake sequence ruptured numerous immature and largely unmapped faults, which were not previously recognized as a connected system capable of hosting major earthquakes (Ponti et al., 2020).The earthquake sequence cascaded across hierarchically interlaced, antithetic faults (Ross et al., 2019), which are likely statically strong (Fialko, 2021;Fialko & Jin, 2021).These geological and structural complexities seem to likely inhibit the rupture propagation across the entire fault system.Therefore, various weakening mechanisms might have played critical roles in facilitating the observed rupture propagation (Milliner et al., 2022;Taufiqurrahman et al., 2023).As shown in our in situ V p /V s ratios, the Ridgecrest fault zones are surprisingly homogeneous.This structural simplicity may have facilitated rupture propagation due to a lack of seismic structural barriers (McGuire et al., 2012), specifically on the northwest-trending fault where the mainshock occurred (Figure 2b).Variations in the resolved V p /V s ratios near the edges of the fault zones indicate transitions in material properties and fault zone environments, which may have contributed to arresting the foreshock and mainshock rupture propagation (Wibberley et al., 2008), in addition to the effects of prestress-to-strength distribution in the fault zones (Taufiqurrahman et al., 2023).

Theoretical Crack Models
In order to understand the observed V p /V s anomalies, we explore the effects of porosity and aspect ratio on in situ V p /V s ratios by applying the Kuster-Toksöz theoretical crack model (Kuster & Toksöz, 1974).The Kuster-Toksöz elastic wave-scattering method appears to be more accurate than some of the self-consistent approaches, even though it does not explicitly consider inclusion interactions (Zimmerman, 1984).We construct idealized porousmedium models with different inclusions in ellipsoidal shapes and assume an over-pressurized condition to compare our observed in situ V p /V s ratios with theoretical predictions.We use V p of 5.962 km/s and V p /V s ratio of 1.737 from the tomographic model as the background values for the host rock.Because of the presence of large number of carbonate fragments in the area (Thompson Jobe et al., 2020), we examine water and carbon dioxide (CO 2 ) inclusions for our theoretical modeling (Figure S5 in Supporting Information S1) and assume water velocity and density as V p = 1.5 km s 1 and ρ = 1.0 Mg m 3 and CO 2 velocity and density as V p = 0.4 km s 1 and ρ = 0.75 Mg m 3 (Adam & Otheim, 2013;Lemmon et al., 2023;Takei, 2002).The ellipsoidal crack shapes are evaluated with their aspect ratios ranging from 0.001 to 0.3.For water-filled cracks, V p reduces with increased porosity, however, V p /V s can either increase or decrease, depending on aspect ratio.For very flat water-filled cracks (aspect ratio ≤0.01), the V p /V s ratio increases with porosity; whereas for more spherical cracks, V p /V s decreases with porosity.Including CO 2 would reduce both the V p and V p /V s , regardless of aspect ratio.Similar results are yielded by applying Walsh (1969) or O 'Connell and Budiansky (1974) approaches.Our idealized modeling can help explain the correlations between the tomographic V p and in situ V p /V s ratio, although interpreting their absolute values require adopting more sophisticated models (e.g., Brantut & David, 2019).The model suggests a positive correlation between the tomographic V p and in situ V p /V s ratio if CO 2 and/or more spherical water-filled cracks are involved.In contrast, a negative correlation is expected if flat water-filled cracks exist.It is important to note that the presence of water is often associated with high V p /V s ratios as a result of water reducing the shear modulus.This leads to decreases in both V p and V s , but with a greater effect on V s , resulting in a high V p /V s ratio, showing as a negative correlation.This correlation is commonly used to interpret the low V p and high V p /V s ratio as the existence of fluids in tomographic models.However, laboratory experiments (Christensen, 1996;Takei, 2002) and our crack model demonstrate that water inclusions can either increase or decrease the V p /V s ratio (Figure S5b in Supporting Information S1), depending on crack porosity and geometry.According to the crack model, thick cracks (i.e., large aspect ratios) can lower V p /V s ratios, while thin cracks are more likely to result in high V p /V s ratios.We notice that the aspect ratio and porosity cannot reduce the V p /V s ratio to below 1.5 if only water inclusions are considered.This suggests that another type of inclusion might have been present, leading to the low value of 1.45 observed in this study, such as CO 2 only or a combination of water and CO 2 .

Imaging Fault Zone Environments in Space and Time
The abundant seismicity of the Ridgecrest earthquake sequence offers a rare opportunity to examine short-term changes of in situ V p /V s ratios.We segment the earthquake sequences into four time periods, using the timings of the Mw 6.4 and Mw 5.4 foreshocks and the Mw 7.1 mainshock as delineation points.These periods are termed "BEFORE," "MIDDLE1," "MIDDLE2," and "AFTER" (Table 1 and Figures 3 and 4).There are few earthquakes before the Mw 6.4 foreshock to form reliable clusters.We conduct separate earthquake cluster analyses for each period and then estimate the in situ V p /V s ratios for event clusters within each period.We use the same parameters and selection criteria for V p /V s estimation as for the entire data set (Table 1).We find that the V p /V s ratios near the Mw 7.1 mainshock increased from 1.45 to 1.72 between the "MIDDLE2" and "AFTER" periods (see Figure S3 in Note.ALL: all the earthquakes from 1 January to 31 August 2019; BEFORE: events before the Mw 6.4 earthquake on July 4; MIDDLE1: events between the Mw 6.4 and 5.4 earthquakes; MIDDLE2: events between the Mw 5.4 and 7.1 earthquakes; and AFTER: earthquakes after the Mw 7.1 mainshock on July 6.Note that we only show information for clusters with robust V p /V s ratios. Supporting Information S1 for best-fitting V p /V s ratios), suggesting an abrupt change in the fault zone environment near its hypocenter (Figures 3c and 3d).From "MIDDLE1" to "MIDDLE2", the V p /V s ratios all increased for the clusters located above the Mw 6.4 foreshock, whereas those near the Mw 5.4 and Mw 7.1 earthquakes all decreased (Figures 3 and 4).We also observe a high V p /V s ratio of 1.97 for a nearby cluster in the entire data set (Figure 4a2), in which only one earthquake occurred 3 hr before the mainshock and the rest after it.
The differences in the values between the "MIDDLE2" and entire data set suggest a possible structural change or reactivation caused by the earthquake.The aftershocks in the nearby cluster from the entire data set are grouped into a bigger cluster in "AFTER" (Figure 3d) with a normal V p /V s value of 1.72.This indicates that if the cluster size is too big, the average V p /V s will be captured over a relatively large spatial scale and any small-scale variations may be missed (see Supporting Information S1).

Changes in Fault Environment Near the 2019 M7.1 Ridgecrest Earthquake Hypocenter
We identify a transient change in the in situ V p /V s ratio at the hypocenter of the Mw 7.1 mainshock, which increased from 1.45 to 1.72 before and after the earthquake (Figure 3).Prior to the Mw 6.4 foreshock, the Ridgecrest fault system is seismically quiescent.Between the Mw 6.4 and Mw 5.4 foreshocks, seismicity forms an L-shaped distribution in the fault zones without extending beyond the location of the Mw 5.4 foreshock (Hauksson & Jones, 2020).During this period and the subsequent time between the Mw 5.4 and Mw 7.1 earthquakes, the in situ V p /V s ratios maintain a nominal value around 1.70-1.74for the same fault segment (Figures 3b and 3c).Intriguingly, V p /V s ratios appear to increase for the clusters located around the connection of the northeast-trending and northwest-trending faults from "MIDDLE1" to "MIDDLE2."In contrast, V p /V s ratios near the Mw 5.4 and Mw 7.1 hypocenters seem to decrease during these periods (Figure 4).Moreover, a low V p / V s ratio patch of 1.45 exists near the Mw 7.1 mainshock during "MIDDLE2".This low value of V p /V s ratio likely reflects over-pressurized pores filled with water and CO 2 in the fault zone (Lin, 2019;Lin et al., 2022).This patch, with a spatial extent of 4 km, seems to disappear after the Mw 7.1 mainshock (Figure 3c), suggesting a fault environment change caused by the mainshock.Abundant aftershocks over the 2 months illuminate the entire Ridgecrest fault system, which is characterized by a homogeneous distribution of V p /V s ratios with a nominal value of 1.72.The homogeneous distribution as well as the values of the V p /V s ratios align closely with those obtained from the complete earthquake data set because aftershocks of the Mw 7.1 mainshock dominate the data set.
We interpret the low V p /V s ratio of 1.45 near the Mw 7.1 mainshock's hypocenter between the Mw 5.4 foreshock and the mainshock suggests as evidence of the region enriched with fluids including both water and CO 2 prior to the rupture of the mainshock.The fluid might have been trapped in ellipsoidal pockets, which subsequently escaped through pathways created during the mainshock rupture.These fluid pockets may have caused high pore pressure at the Mw 7.1 hypocenter, thus reducing the effective normal stress and resulting in a locally weak fault (Parsons et al., 2017;Sibson, 1992).This hypothesis aligns with the physics-based rupture simulations that indicate elevated pore fluid pressure is required for nucleating the Ridgecrest earthquakes, and the spatial footprint of the low-V p /V s -ratio patch matches the hypothesized rupture nucleation size (Taufiqurrahman et al., 2023).Specifically, neither the static nor the dynamic stress induced by the Mw 6.4 foreshock immediately triggered the mainshock.It is also unclear whether the static stress from the Mw 5.4 foreshock have promoted or inhibited the onset of the Mw 7.1 mainshock (Barnhart et al., 2019;Huang et al., 2020;Jin & Fialko, 2020), and its dynamic stress changes also fail to instantaneously trigger the mainshock (Jin & Fialko, 2020).Such delayed triggering has been observed in other strike-slip earthquake sequences, such as the Landers-Hector Mine pair and the 2023 Turkey earthquake doublet, which could be a reflection of the rate-and-state friction behaviors (Jia et al., 2023;Parsons, 2005;Pollitz & Sacks, 2002;Price & Bürgmann, 2002).In addition, the enclosed fluid inferred from the low V p /V s ratios may have provided one necessary weakening mechanism to help nucleate the Mw 7.1 mainshock (Lambert et al., 2021).Once nucleated, this enclosed fluid could further weaken the fault due to thermal pressurization effects (Andrews, 2002;Noda & Lapusta, 2010;Vyas et al., 2023), leading to the observed significant slip near the hypocenter of the Mw 7.1 mainshock (Chen et al., 2020;Goldberg et al., 2020;Wang et al., 2020a).
The earthquakes occurring between the Mw 6.4 foreshock and the Mw 7.1 mainshock did not follow an Omori's law-type decay, indicating that this group of earthquakes might deviate from typical stress-triggered aftershock sequences (Hauksson & Jones, 2020;Ogata & Zhuang, 2006).Foreshocks of other M > 7 earthquakes in Southern California have also shown anomalous behavior, suggesting preparatory processes that eventually led to the M > 7 mainshocks (Dodge et al., 1995;Yoon et al., 2019).These foreshock sequences bear similarities to earthquake swarms, much like the earthquakes that occurred between the Mw 6.4 and Mw 7.1 events.We also observe changes in the pattern of in situ V p /V s ratios from "MIDDLE1" to "MIDDLE2" in the northwest-trending main fault (Figures 3b, 3c, and 4).These changes might indicate that the northeast-trending fault, which ruptured during the Mw 6.4 foreshock, was fluid-rich (Taufiqurrahman et al., 2023); the increase at the fault intersection point reflects a similar process that occurred near the Mw 7.1 hypocenter between "MIDDLE2" and "AFTER." The V p /V s ratio increase suggests that there was either fluid movement or aseismic slip within the northwesttrending fault zone around the Mw 5.4 earthquake and between the periods "MIDDLE1" and "MIDDLE2."Given the spatiotemporal scale, this V p /V s ratio increase is unlikely due to fluids migrating from the fault intersection to the Mw 7.1 hypocenter for realistic diffusivity of 0.1-1 m 2 /s (e.g., Barbour et al., 2017;Chen et al., 2012;Goebel & Brodsky, 2018).However, disconnected, localized fluid movements could possibly cause the observed changes.Alternatively, early afterslip or aseismic slip triggered by the Mw 6.4 earthquake (e.g., Golriz et al., 2021;Huang et al., 2020;Jiang et al., 2021;Premus et al., 2022) could alter the stress conditions in the northwest-trending fault zone and cause the changes in the V p /V s ratios.Foreshocks of previous California M > 7 earthquakes exhibit lower stress drops and are depleted of high-frequency energy compared to their mainshocks and associated aftershocks (Chen & Shearer, 2013).These features in foreshocks could be indicative of aseismic transients occurring near and prior to the mainshocks (Dodge et al., 1996;Huang et al., 2020).Additionally, they could be attributed to a fluid-rich fault environment or migrating fluid flow, both of which could reduce fault strength and lower differential stress; these processes could also occur concurrently with

Geophysical Research Letters
10.1029/2024GL109171 aseismic transients.While such foreshocks are not consistently observed before mainshocks (van den Ende & Ampuero, 2020), the transient change in V p /V s ratio near the Mw 7.1 mainshock may reflect an underlying fault zone condition that could facilitate the nucleation of other M > 7 earthquakes in California and beyond.

Conclusions
We apply a high-resolution method to estimate in situ V p /V s ratios for similar earthquake clusters during the 2019 Ridgecrest seismic sequence.The majority of the rupture zone is governed by relatively homogeneous V p /V s ratios.Slightly low values are resolved in the ALFZ, at the terminus of the LLFZ and the intersection with the GF.Fine-scale V p /V s anomalies are revealed near the source regions of the Mw 6.4, 5.4, and 7.1 shocks, with low value before and high one after the mainshock, indicating a fluid rich fault zone environment prior to the earthquake nucleation process, which is likely altered by the mainshock after its rupture.

Figure 1 .
Figure 1.(a) Map view of the waveform cross-correlation relocations by Lin (2020) for the first 8 months of 2019 in the adjacent area of the Ridgecrest earthquake sequence.Pink dots represent the background seismicity before 1 July 2019.Events from July 1 to August 31 are colored by days since 1 July 2019.Stars mark the Mw 6.4, 5.4, and 7.1 earthquakes on July 4, 5, and 6 2019 (UTC).The inset map shows the location of the study area in California.(b) Map view of the earthquakes in the similar event clusters with robust V p /V s ratios.Events are colored by the corresponding in situ V p /V s ratio.The red straight lines (A-B and C-D) and dotted boxes are the profiles and boundaries for the cross-sections in the following figures.

Figure 2 .
Figure 2. (a), (b) Depth distribution of seismicity, colored by the in situ V p /V s ratio for the corresponding cluster, along profile A-B and C-D and enclosed by the corresponding dotted box in Figure 1b, respectively.(c), (d) Tomographic V p /V s and (e), (f) tomographic V p values for events in the similar earthquake clusters along the two profiles.Events are colored by the average velocity value for each cluster.Stars represent the 3-D relocations of the Mw 6.4, 5.4, and 7.1 earthquakes by Lin (2020).The purple lines at the top of the cross sections show the local topography.Zero depth refers to mean sea level.Abbreviations are LLFZ, Little Lake Fault Zone; and ALFZ, Airport Lake Fault Zone.

Figure 4 .
Figure 4. Depth distributions of seismicity along profiles A-B (a1-d1) and C-D (a2-d2) for different time periods.Symbols are the same as those in Figure 2.

Table 1
Details of the Seismicity Sequences in Different Time Periods