Zefeng Li

Low‐velocity zones along the San Jacinto Fault, Southern California, from body waves recorded in dense linear arrays

We derive high-resolution information on low-velocity fault zone (FZ) structures along the San Jacinto Fault Zone (SJFZ), Southern California, using waveforms of local earthquakes that are recorded at multiple linear cross-fault arrays. We observe clear across-fault delays of direct P and S waves, indicating damage zones at different segments of the SJFZ. We then compute synthetic traveltimes and waveforms using generalized ray theory and perform forward modeling to constrain the FZ parameters. At the southern section near the trifurcation area, the low-velocity zone (LVZ) of the Clark branch has a width of ~200 m, 30–45% reduction in Vp, and ~50% reduction in Vs. From array data across the Anza seismic gap, we find a LVZ with ~200 m width and ~50% reduction in both Vp and Vs, nearly as prominent as that on the southern section. We only find prominent LVZs beneath three out of the five arrays, indicating along-strike variations of the fault damage. FZ-reflected phases are considerably less clear than those observed above the rupture zone of the 1992 Landers earthquake shortly after the event. This may reflect partially healed LVZs with less sharp boundaries at the SJFZ, given the relatively long lapse time from the last large surface-rupturing event. Alternatively, the lack of observed FZ-reflected phases could be partially due to the relatively small aperture of the arrays. Nevertheless, the clear signatures of damage zones at Anza and other locations indicate very slow healing process, at least in the top few kilometers of the crust.

Spatial variations of shear‐wave anisotropy near the San Jacinto fault zone in Southern California

We examine crustal anisotropy at several scales along and across the San Jacinto Fault Zone (SJFZ) by systematically measuring shear wave splitting (SWS) parameters. The analyzed data are recorded by 86 stations during 2012–2014, including five linear dense arrays crossing the SJFZ at different locations and other autonomous stations within 15 km from the main fault trace. Shear phase arrivals and SWS parameters (fast directions and delay times) are obtained with automated methods. The measurement quality is then assessed using multiple criteria, resulting in 23,000 high-quality measurements. We find clear contrast of fast directions between the SW and NE sides of the SJFZ. Stations on the SW side have fast directions consistent overall with the maximum horizontal compression direction (SHmax), while stations on the NE side show mixed patterns likely reflecting lithological/topographic variations combined with fault zone damage. The fast directions in the Anza gap section with relatively simple fault geometry agree with the inferred SHmax, and the delay times at an array within that section are smaller than those observed at other across-fault arrays. These indications of less pronounced damage zone in the Anza section compared to other segments of the SJFZ are correlated generally with geometrical properties of the surface traces. Significant variations of fast directions on several across-fault arrays, with station spacing on the orders of a few tens of meters, suggest that shallow fault structures and near-surface layers play an important role in controlling the SWS parameters.

High-resolution seismic event detection using local similarity for Large-N arrays

We develop a novel method for seismic event detection that can be applied to large-N arrays. The method is based on a new detection function named local similarity, which quantifies the signal consistency between the examined station and its nearest neighbors. Using the 5200-station Long Beach nodal array, we demonstrate that stacked local similarity functions can be used to detect seismic events with amplitudes near or below noise levels. We apply the method to one-week continuous data around the 03/11/2011 Mw 9.1 Tohoku-Oki earthquake, to detect local and distant events. In the 5–10 Hz range, we detect various events of natural and anthropogenic origins, but without a clear increase in local seismicity during and following the surface waves of the Tohoku-Oki mainshock. In the 1-Hz low-pass-filtered range, we detect numerous events, likely representing aftershocks from the Tohoku-Oki mainshock region. This high-resolution detection technique can be applied to both ultra-dense and regular array recordings for monitoring ultra-weak micro-seismicity and detecting unusual seismic events in noisy environments.

An Automatic Phase Picker for Local Earthquakes with Predetermined Locations: Combining a Signal‐to‐Noise Ratio Detector with 1D Velocity Model Inversion

We present a simple method to pick additional P and S phases for local earthquakes with predetermined locations. Different from conventional characteristic function techniques, our method incorporates 1D velocity inversion into phase picking. It first predicts phase arrivals using initial velocity models and available event locations and then applies a detector function to search genuine phase arrivals around the initial predictions. Using the newest searched phase arrivals (picks), the velocity models are updated accordingly and then used to predict more accurate arrival times. Such a procedure can be iterated multiple times, during which both phase picking and velocity model are improved. We perform a synthetic test with a 1D velocity model. The resulting phase picks are consistent with true arrivals, and true velocity models can be well recovered. Another test with real data recorded by the Anza Seismic Network in southern California shows that the resulting velocity models agree with the average Southern California Earthquake Center community models in this region. Out of 23,932 event–station pairs, 23,770 P picks and 21,935 S picks are obtained from this method. After four iterations, 90% of our P picks and 80% of our S picks have differences within 0.15 s from the phase picks of the Southern California Earthquake catalog. Given its simplicity and efficiency and ability to produce robust P and S picks and 1D velocity models, the technique is particularly suitable for a wide range of seismological research in which phase picks at additional stations and/or refined 1D velocity models are needed.

Stress‐ and Structure‐Induced Anisotropy in Southern California From Two Decades of Shear Wave Splitting Measurements

We measure shear wave splitting (SWS) parameters (i.e., fast direction and delay time) using 330,000 local earthquakes recorded by more than 400 stations of the Southern California Seismic Network (1995–2014). The resulting 232,000 SWS measurements (90,000 high-quality ones) provide a uniform and comprehensive database of local SWS measurements in Southern California. The fast directions at many stations are consistent with regional maximum compressional stress σ_(Hmax). However, several regions show clear deviations from the σ_(Hmax) directions. These include linear sections along the San Andreas Fault and the Santa Ynez Fault, geological blocks NW to the Los Angeles Basin, regions around the San Jacinto Fault, the Peninsular Ranges near San Diego, and the Coso volcanic field. These complex patterns show that regional stresses and active faults cannot adequately explain the upper crustal anisotropy in Southern California. Other types of local structures, such as local rock types or tectonic features, also play significant roles.

Machine Learning Seismic Wave Discrimination: Application to Earthquake Early Warning

Performance of earthquake early warning systems suffers from false alerts caused by local impulsive noise from natural or anthropogenic sources. To mitigate this problem, we train a generative adversarial network (GAN) to learn the characteristics of first‐arrival earthquake P waves, using 300,000 waveforms recorded in southern California and Japan. We apply the GAN critic as an automatic feature extractor and train a Random Forest classifier with about 700,000 earthquake and noise waveforms. We show that the discriminator can recognize 99.2% of the earthquake P waves and 98.4% of the noise signals. This state‐of‐the‐art performance is expected to reduce significantly the number of false triggers from local impulsive noise. Our study demonstrates that GANs can discover a compact and effective representation of seismic waves, which has the potential for wide applications in seismology.