A document by Chao Song. Click on the document to view its contents.

Currently there are two end-member views of low-frequency earthquakes (LFEs). One is that they result from stick-slip behavior of Mxed brittle patches that are mechanically distinct from the surrounding, creeping, fault. The other is that they represent the high-frequency limit of stochastic accelerations and decelerations of slip on the fault, with boundaries that may vary with time. Among the many unknowns concerning LFEs are their physical dimensions, knowledge of which might place constraints on their underlying nature. Estimates of LFE source sizes range from 100 m (if they have a stress drop similar to regular earthquakes) to more than 1 km (if they rupture at local shear wave speeds, given a ~0.5-s duration). Our 4-s tremor catalogs have an estimated location uncertainty smaller than 1 km, and many consecutive and nearly consecutive detections are spaced closer than this. This suggests that if LFEs are close to the upper size limit, successive events are strongly overlapping, which seems more consistent with the stochastic acceleration and deceleration model of LFE generation. Moreover, many 4-s windows themselves contain multiple nearly co- located LFE-like arrivals. This encourages us to use LFE templates to deconvolve tremor seismograms, to obtain an LFE catalog with a temporal resolution conceivably as high as one event per LFE duration. Our preliminary catalog shows that many LFEs as close in time as 0.5 s are separated by less than 1 km. Figure 1 shows LFE detections from a 250-s window with an overall 2-km migration to the southeast. The median event has 75 other detections within 1 km. In the propagation direction, also the direction of low error, the median separation between consecutive events is 360 m. If LFEs are km-scale, possible explanations for the signiPcant overlap between events sometimes as close as 0.5 s in time (see the zoom) include reQected waves from boundaries of a low-velocity shear zone, or inertial vibrations at low normal stress (Im and Avouac, 2021). If, instead, LFEs are brittle asperities closer to 100 m in size, successive events need not overlap, but one must explain both their long duration and why, with so many sources in close proximity, nearly none are observed to grow larger in both duration and magnitude than is characteristic of LFEs.

In this study, we carried out a backprojection (BP) analysis to image the rupture process of the newly happened September 8, 2017 Mww8.1 Mexico earthquake based on a global 3D P-wave tomography model, the LLNL-G3Dv3 model. Limited to epicenter distance and data quality, only waveform observation data from Alaska (AL), USA was utilized finally, with some data from South America (SA) as supplement. First, we compared the HF BP results of 1D and 3D model to illustrate the higher resolution and reliability of the 3D one. Then we discussed the consistency among the overall rupture pattern, the main event focal mechanism and aftershocks distribution, and further inferred the possible fault geometry. After that, we explained the rationality of the setting for rupture duration based on beamforming energy pattern, normalized power variation and other previous works. We then seriously examined the creditability of stage 2 and explained why speed in stage 2 is much bigger than in 1. Finally, we obtained the coulomb stress change imparted on the faults of the subsequent September 19 Mww7.1 event and September 23 Mww6.1 event to find out if they are positively triggered by this main event. From our current research, the complete ~53s rupture process of this earthquake can be divided into two stages. In stage 1, which lasted for ~37s, the rupture propagated from the epicenter towards the NW direction (~330°measured from north clockwise) with a speed of ~2.8 km/s, and extended to a length of ~89 km. Then it made a right turn and shortly after, it continued to propagate to near N (~3°) with a higher speed of ~5.3 km/s and a scale of ~75 km. Our study intended to believe that the Mww8.1 event has almost nothing to do with the Mww7.1 event while it strongly triggered the occurrence of the latter Mww6.1 event.

Conventional back-projection (BP) method uses 1-D earth model to image rupture process of earthquakes. This may result in discrepancies while using data from arrays with different azimuths. In our study, we integrate a 3-D earth model into BP method to obtain more consistent results when data from different arrays are used for a given earthquake.

The boundary region between Alxa Block and Ordos Block is an area of stress concentration with strong seismicity and frequent small earthquakes. However, the knowledge of this area is limited since only a few seismic stations were deployed in this area. The 2015 Ms5.8 Alxa Left Banner Earthquake on April 15 is the largest one occurred in the surroundings since the 1976 Ms6.2 Bayinmuren Earthquake. Abundant stations built in the northern part of Chinese North-South Seismic Belt recorded this event sequence well within short distances, which provides us a great opportunity to carry out studies. We use these data to obtain a mean 1-D layered velocity structure via iterative inversion based on both travel time and waveform misfits. Then we use the travel time difference between data and synthetic seismograms to relocate the epicenter. Finally we invert the best double-couple focal mechanism and centroid depths of the source. As the result, the source was located at (39.7663°N, 106.4304°E) with a depth of 18 km and Mw 5.25. Nodal plane 1 had strike 176°, dip angle 85°and slip angle-180°, while plane 2 had strike 86°, dip angle 90°and slip angle-5°. Considering the tectonic characteristics of regional fault zone, we believe this earthquake was caused by a nearly pure left-lateral strike-slip fault with a slight normal component, while the nodal plane 2 striking towards NEE (near E-W) was the fault plane. The seismogenic structure was likely to be an E-W striking buried fault nearby. From our study, the corresponding fault of this event may indicate all groups of faults with same E-W strike has the common character of large-dip left-lateral strike slip. Moreover, there may be some buried faults being newly born or not found yet. These results could be an important supplement to the future research of regional seismicity and modern fault zone structure.

Although tremor is believed to consist of myriad Low-frequency Earthquakes (LFEs), it also contains longer-period signals of unknown origin. We investigate the source of some of the longer-period signals by locating tremor windows independently in relatively high-frequency (’HF’, 1.25–6.5 Hz, containing typical LFEs) and low-frequency (’LF’, 0.5–1.25 Hz) bands. We hypothesize that if tremor consists entirely of LFEs, such that the lower-frequency signals come from the non-uniform timing of higher-frequency ($\sim$2 Hz) LFEs, then contemporaneous LF and HF signals should be nearly co-located. Here we search for a systematic offset between the locations of contemporaneous LF and HF detections during rapid tremor migrations (RTMs). This first requires correcting for apparent offsets in location that arise simply from filtering in different passbands. To guard against possible errors in our empirical filtering effect corrections, we focus on a region of the subduction interface beneath southern Vancouver Island that hosts migrations propagating in nearly opposing directions. We find that the LF energy appears to occur roughly 500 m farther behind the propagating fronts of RTMs than the HF energy, whether those fronts propagate to the ENE or to the WSW. This separation is small compared to the location error of individual LF detections, but the result seems robust owing to the large number of detections. If this result stands, it suggests that tremor consists of more than just a collection of LFEs, with longer-period energy being generated farther behind the migrating fronts of RTMs, where slip speeds are presumably lower.

The broadband stacks (templates) of velocity seismograms of nearly co-located low-frequency earthquakes (LFEs) detected using a 1-8 Hz passband beneath southern Vancouver Island tend to exhibit a simple dipolar shape with a characteristic duration of ~0.3-0.5 s, which is also found to be nearly independent of the seismic moment. An important question left unanswered is whether the duration is due to the nature of the source, is set by attenuation near the source region, or is just a bias introduced by the narrow passband used to detect LFEs. In tremor catalogs detected using a relatively low-frequency passband, 0.5-1.25 Hz, we have found some tremor windows that contain relatively isolated dipole arrivals similar to LFEs. A few of these have a duration apparently longer than that of the LFE templates. Notably, the same location on the fault also seems capable of generating signals with a shorter duration at other times. Figure 1 shows seismograms, at 3 stations, of one such example in the vicinity of LFE family 001 of Bostock et al. (2012), in which the main arrival has a duration of ~1 s, whereas another signal 3 s earlier with a duration of only ~0.4 s comes from roughly the same location (same move-out between the stations). This significant variability in duration at approximately the same location suggests that the long-duration events owe their duration to source processes and not attenuation, provided that attenuation does not vary on extremely short time and space scales during the episodic tremor and slip episode. The relative isolation in time also makes the longer duration less likely to result from the temporal clustering of multiple typical LFEs. We will undertake a more systematic search of our longer- and shorter-period tremor catalogs to assess this possibility. Addressing this question will shed more light on the factors that control the apparent duration of LFEs. Figure 1 The top panel shows the long-duration tremor signal in a relatively lower-frequency band, 0.5-1.25 Hz, whereas the second panel from the top is the same 32-s segment in a higher-frequency band, 1.25-6.5 Hz. The third panel shows the trace in a broader passband, 0.5-6.5 Hz. The bottom panel shows the stacked LFE templates of the same family filtered through 0.5-6.5 Hz.