S5: Synchronous Slow Slip & Seismic Swarm – S_5

Identifying new S5 and following their evolution in time and space requires not only dense observation networks but also new approaches in order to derive images of slip and seismicity patterns.

A first workpackage of the project aims at developing new methods for a full time-dependent inversion of slip, using the displacement time series from GNSS stations (GNSS is the acronym for satellite positioning systems including GPS and Galileo). The novelty of our approach is to solve for daily or even sub-daily rates of the slip taking place at the fault.

The second approach will apply, adapt and develop new Machine Learning algorithms to detect, to locate small earthquakes occurring during any S5 in order to produce earthquake catalogues with an enhanced completeness and accuracy compared to the ones derived from conventional approaches. Machine Learning will also be used to detect new seismological signals potentially induced by slip acceleration. Machine Learning algorithms, trained on existing data sets are well suited to pick-up new signals that are hidden in noisy seismograms and open the perspective to precisely document how seismicity is organized in space and time during S5.

Once we will get precise models of slip and seismicity for a few events, the second part of the project aims at understanding the relationship between the time dependent evolution of slip and the seismicity location and rate. In a first step, we will analyze spatial and temporal correlations, before moving to joint dynamical models where the stress evolution will be driving both slip and seismicity. Finally, these results will be compared to dynamical simulations using space variable friction parameters to reveal the anatomy of the subduction interface hosting S5.

A first result is the detailed observation in September 2020 of an original sequence mixing seismic and aseismic slip in the Copiapo segment off-shore Chile. By chance, the sequence developed during the period of deployment of a temporary seismological network we had planned for the S5 project in addition to the existing continuous permanent GNSS and seismological stations.

We were expecting either S5 or a regular SSE in this area, but here the sequence started with a moderate size earthquake of magnitude Mw 6.8, followed a few hours later by an aftershock of magnitude Mw 6.4. But was this sequence a usual mainshock/aftershocks process, as many others around the world? Not exactly. Our analysis shows that this sequence shares some similarities with S5: the size of the largest aftershocks is unusually large, and large/rapid aseismic slip, similar to SSEs migrated from the mainshock area towards the location of the aftershocks. Then, abnormally large aseismic slip, equivalent in magnitude to the mainschock developed during two weeks. This sequence highlights an original mechanism, perhaps a case intermediate between a S5 and a classical mainschock/afterschoks sequence. In that sense, it illuminates the complexity and diversity of frictional behaviors at subduction interfaces.

Better understanding the faults behavior requires to document the many processes contributing to the accumulation, redistribution and release of stress at faults and within the surrounding medium. S5 is one of such processes. S5 usually occur at the edges of highly coupled fault area, where energy available for future earthquakes accumulates at the highest rate. S5 are interesting processes because they induce stress perturbations, perhaps with the potential to initiate large rupture.

Because S5 includes regular seismicity, rather than tremors that are more difficult to detect and to locate, the perspective to get description precise enough to constrain dynamical models appears more reachable.

The first results obtained are encouraging: despite the restriction for field work due to the COVID pandemics, a first process has been caught by our geodetic and seismological networks. Un second process occurred in June 2021 in central Ecuador during June 2021. It is currently being analyzed, in close collaboration with the S5 project partners from IG-EPN (https://www.igepn.edu.ec/servicios/noticias/1885-informe-sismico-especial-no-2021-006).

First results of Machine Learning and new slip modelling approaches seem promising.

1. Klein, E., Potin, B., Pasten-Araya, F., Tissandier, R., Azua, K., Duputel, Z., ... & Vigny, C. (2021). Interplay of seismic and a-seismic deformation during the 2020 sequence of Atacama, Chile. Earth and Planetary Science Letters, 570, 117081.

2. Bletery, Q., & Nocquet, J. M. (2020). Slip bursts during coalescence of slow slip events in Cascadia. Nature communications, 11(1), 1-6.

Slow slip events (SSE) are transient processes releasing stress at faults without significant earthquake. Their discovery about two decades ago in subduction zones demonstrates a complex dynamics of the megathrust controlled by spatially variable friction at the plate interface. While deep SSEs occurring downdip of highly locked areas have been extensively studied, other subduction zones highlight another transient process where slip occurs at the same depths as large earthquakes and is synchronous to intense micro-seismicity. We refer to this type of transient as S5 for Synchronous Slow Slip & Seismic Swarm, which is the focus of our proposal.

With recurrence time of a few years, S5 periodically induce stress perturbation at the megathrust and might be precursors to an incipient large earthquake, as observed for the 2011 Japan giant earthquake. However, most S5 are not followed by a large earthquake. A major challenge is to know whether some characteristics (e.g. seismicity increase, acceleration of slip, penetration of slip into highly locked areas) could be indicators of the nucleation phase of an incipient large earthquake. As a step required to answer this question, this project aims at (1) precisely observing S5 (2) using novel analysis methodologies to better document the slip and the seismicity during S5 and (3) developing new modelling approaches to consistently integrate the different observations to decipher the underlying physics.

We selected four areas along the South America subduction zone where (1) the probability of observing S5 during the duration of the project is high (2) thanks to previous efforts and partnerships with local institutes, existing seismological and geodetic infrastructure enables to deploy a dense network at a lower cost. Two areas are located in the northern Andes in Ecuador and Peru and two are in Central Chile. At each targeted area, we will install additional continuous GNSS stations and broadband seismometers that will be recording during the 4 years of the project. In addition, we will regularly survey dense networks of GNSS benchmarks and perform a 6-months long seismological experiment with 10 additional broadband seismometers at both targeted areas in Chile. Together with existing data sets from dense seismological networks that recorded S5 in Ecuador and Peru, this observational effort will provide spatially and temporally high resolution data to apply novel methods.

We will process the GNSS data and investigate new methods to separate non-tectonic contributions in GNSS time series. Then we will derive a velocity field used to perform refined modelling of the interseismic coupling to understand the environment of S5. A novelty is that we will develop a generalized full time-dependent slip inversion from GNSS time series opening the way for a kinematic imaging of slip and slip rate at the subduction interface.
For the seismological data, we will (1) search for repeating earthquakes (2) search for tremors and low frequency earthquakes (3) systematically calculate focal mechanisms and source time functions for the largest events. An additional novelty of our proposal is to use Machine Learning (ML) techniques in order to speed up and to improve micro-seismicity analysis.

Finally, we will integrate the geodetic and seismological results in a modelling approach where the spatial and temporal evolution of the seismicity and recurrence time for repeating earthquakes are consistent with the stress evolution induced by the slip developing through time at the plate interface. Simultaneously, forward numerical modelling of a frictionally heterogeneous fault will provide a synoptic view of the relations between friction parameters and observable slip behaviors. Finally, physical modelling will examine the physical conditions and characteristics required for an S5 to lead to a large earthquake.