The coupling at the interface between tectonic plates is a key geophysical parameter to capture the frictional locking across plate boundaries, and provides a means to estimate where tectonic strain is accumulating through time. Here, we use both interferometric radar (InSAR) and GNSS data to investigate the plate coupling of the Hikurangi subduction zone beneath the North Island of New Zealand, where multiple slow slip cycles are superimposed on the long-term loading. We estimate the plate coupling across the subduction zone over different observational periods (2, 4, and 10 years) targeting different stages of the slow slip cycles. Our results highlight the importance of the observational period when interpreting coupling maps, notably highlighting the temporal dependence of plate coupling. Through our analysis of multiple geodetic datasets, we demonstrate how InSAR provides powerful constraints on the spatial resolution of plate coupling, even in a region where a dense GNSS network exists.

The 2016 M7.8 Kaikōura earthquake is one of the most complex earthquakes in recorded history, with significant rupture of at least 21 crustal faults. Using a matched-filter detection routine, precise cross-correlation pick corrections, and accurate location and relocation techniques, we construct a catalog of 33,328 earthquakes between 2009 and 2020 on and adjacent to the faults that ruptured in the Kaikōura earthquake. We also compute focal mechanisms for 1,755 of the earthquakes used as templates. Using this catalog we reassess the rupture pathway of the Kaikōura earthquake. In particular we show that: (1) the earthquake nucleated on the Humps Fault; (2) there is a likely linking offshore reverse fault between the southern fault system and the Papatea Fault, which could explain the anomalously high slip on the Papatea Fault; (3) the faults that ruptured in the 2013 Cook Strait sequence were reactivated by the Kaikōura earthquake and may have played a role in the termination of the earthquake; and (4) no seismicity on an underlying subduction interface is observed beneath almost all of the ruptured region suggesting that if deformation did occur on the plate interface then it occurred aseismically and did not play a significant role in generating co-seismic ground motion.

Slow slip events have previously been observed along the Hikurangi subduction zone beneath the North Island of New Zealand. These slow slip episodes occur both on the shallow plate interface (< 15km depth) and at the deeper end of the seismogenic zone (> 30km depth). We present the first catalog of low-frequency earthquakes (LFEs) in the Hikurangi subduction zone, located beneath the Kaimanawa Range on the central Hikurangi margin, downdip of a region that regularly (every 4-5 years) hosts M7 slow slip events. To systematically detect LFEs using continuous seismic data recorded by GeoNet, we developed a matched-filter technique with template waveforms derived from previous observations of tectonic tremor. The workflow presented in this work is composed of two iterations of a matched-filter search. In each iteration, the detections were gathered into families and their common waveforms postprocessed with machine-learning methods to extract high-quality waveforms, allowing us to pick seismic phase arrivals with which to locate the LFEs. We found that LFEs occur in episodes of intense activity during the neighboring updip M7 slow slip events. We also observe a recurrence time of 2 years between other large bursts of LFEs, suggestive of a shorter cycle of slow slip. We hypothesize that these and other frequent LFE episodes highlight smaller slow transients that have not yet been geodetically observed.

How faulting processes lead to a large earthquake is a fundamental question in seismology. To better constrain this pre-seismic stage, we create a dense seismic catalog via template matching to analyze the precursory phase of the Mw 6.3 L’Aquila earthquake that occurred in central Italy in 2009. We estimate several physical parameters in time, such as the coefficient of variation, the seismic moment release, the effective stress drop, and analyze spatio-temporal patterns to study the evolution of the sequence and the earthquake interactions. We observe that the precursory phase experiences multiple accelerations of the seismicity rate that we divide into two main sequences with different signatures and features: the first part exhibits weak earthquake interactions, quasi-continuous moment release, slow spatial migration patterns, and a lower effective stress drop, pointing to aseismic processes. The second sequence exhibits strong temporal clustering, rapid spatial expansion of the seismicity and larger effective stress drop typical of a stress transfer process. We interpret the differences in the seismicity behavior between the two sequences as distinct physical mechanisms that are controlled by different physical properties of the fault system. We conclude that the L’Aquila earthquake is preceded by a complex preparation, made up of different physical processes taking place over different time scales on faults with different physical properties.

The rate of background seismicity, or the earthquakes not directly triggered by another earthquake, in active seismic regions is indicative of the stressing rate of fault systems. However, aftershock sequences often dominate the seismicity rate, masking this background seismicity. The identification of aftershocks in earthquake catalogs, also known as declustering, is thus an important problem in seismology. Most solutions involve spatio-temporal distances between successive events, such as the Nearest-Neighbor-Distance algorithm widely used in various contexts. This algorithm assumes that the space-time metric follows a bi-modal distribution with one peak related to the background seismicity and another peak representing the aftershocks. Constraining these two distributions is key to accurately identify the aftershocks from the background events. Recent work often uses a linear-splitting based on nearest-neighbor distance threshold, ignoring the overlap between the two populations and resulting in a mis-identification of background earthquakes and aftershock sequences. We revisit this problem here with both machine-learning classification and clustering algorithms. After testing several popular algorithms, we show that a random forest trained with various synthetic catalogs generated by an Epidemic Type Aftershock Sequence model outperforms approaches such as K-means, Gaussian-mixture models, and Support Vector Classifications. We evaluate different data features and discuss their importance in classifying aftershocks. We then apply our model to two different actual earthquake catalogs, the relocated Southern California Earthquake Center catalog and the GeoNet catalog of New Zealand. Our model capably adapts to these two different tectonic contexts, highlighting the differences in aftershock productivity between crustal and intermediate depth seismicity.