CORDIS - EU research results

Exploring the nucleation of large earthquakes: cascading and unpredictable or slowly driven and forecastable

Periodic Reporting for period 3 - PRESEISMIC (Exploring the nucleation of large earthquakes: cascading and unpredictable or slowly driven and forecastable)

Reporting period: 2022-01-01 to 2023-06-30

One of the key questions in fault mechanics is how do earthquakes begin. This is central to our understanding of earthquakes, including the long controversial issue of their predictability. Much of what we know about earthquake nucleation comes from the observation of foreshock events before large earthquakes. Yet, a problem is the inherent difficulty to identify earthquakes as foreshocks before the mainshock occurs.

Contrasting views have been proposed regarding the physical mechanisms that generate foreshocks and the reason why they occur. Two main competing conceptual models have been proposed. First, a “cascading model” where successive foreshock stress changes contribute to a slow cascade of random failures ultimately leading to the mainshock. Second, a “slow pre-slip model” where foreshocks are triggered by an aseismic nucleation phase in which the fault slips slowly before accelerating to a dynamic, catastrophic rupture. Implications are of uttermost importance in terms of seismic hazard assessment: In the cascading model, foreshocks randomly trigger each other and therefore randomly trigger the mainshock, hence foreshocks cannot be used to predict earthquakes. On the contrary, the presence of a mixture of seismic and aseismic slip during nucleation suggests a causal relationship in which aseismic slip drives foreshocks, hence the potential to forecast large earthquakes.

Our understanding of earthquake nucleation is currently limited by (1) the difficulty to quantitatively estimate the proportion of seismic versus aseismic slip during rupture initiation and (2) the difficulty to detect and characterize small microearthquakes (with magnitude M<2), which are essential to assess if foreshocks are triggered by aseismic preslip. Both limitations clutter our understanding of the physical mechanisms controlling the relationship between foreshocks and the onset of large earthquakes.

The overall objective of the PRESEISMIC project is to exploit the current explosion of near-fault observations to capture the genesis of earthquakes. While slow and rapid fault processes are usually studied independently, we propose a unified approach combining seismic and geodetic datasets to address the relative contribution of seismic and aseismic processes during foreshock sequences.
In the framework of the PRESEISMIC project, we are developing new tools to distinguish random cascades of foreshocks from a behavior where earthquakes are driven by aseismic pre-slip.

We are producing new time-dependent models assimilating various datasets to address the relative contribution of seismic and aseismic slip. Results for the 2017 Valparaiso earthquake in Chile have shown that this event was preceded by a transient geodetic motion that cannot be fully explained by foreshock induced displacements. Our analysis indicates that about half of the observed pre-seismic displacement is caused by aseismic pre-slip on the subducting interface.

We are also developing approaches to improve seismicity catalogs and analyze earthquake interactions before a mainshock occurs. Our first results demonstrate that non-cascading foreshock seismicity is relatively rare in southern California. Nevertheless, there is still a non-negligible portion of foreshock sequences (~18%) that cannot be explained by the natural clustering of seismicity due to earthquake interactions. Other triggering mechanisms such as aseismic pre-slip are necessary to explain such anomalous sequences.

These preliminary results highlight the fact that some earthquakes are preceded by fault pre-slip that is not induced by foreshocks (neither directly as co-seismic slip nor indirectly as after-slip). While we still need to understand why a large number of earthquakes are not preceded by any clear observable precursor, this suggests that the slow nucleation phase predicted by theoretical and laboratory experiments can sometimes be observed on natural faults.
We are now working on the final developments of probabilistic time-dependent slip models, allowing us to study the spatiotemporal evolution of earthquake preparation. This approach is currently implemented to analyze the preparatory phase preceding the 2014 Iquique earthquake (North Chile). We are also working on obtaining high-resolution seismicity catalogs in areas where aseismic pre-slip has been reported before large mainshock events. Our goal is to identify if the corresponding foreshock sequences have properties that are different from usual seismicity patterns (temporal clustering, event location, waveform properties, etc.).

While the approaches developed in this project are necessary to understand the initiation phase of earthquakes, they can also be useful in other contexts. We are planning to use our new time-dependent inversion tool to investigate slow slip events that are observed on various active faults around the world. More specifically, we aim to target the Cascadia subduction zone, where many of these slow-slip events have been reported. This region has experienced large earthquakes in the past, getting some insights on the slip behavior of this region can therefore prove to be of significant interest. Along with the study of tectonic faults, our tools are also valuable to monitor active volcanoes. Our recent applications at Piton de la Fournaise and Kilauea volcanoes demonstrates how enhanced seismicity detections can improve the forecasting of volcanic eruptions and allow us to identify early signs of impending large caldera collapses.
Two mechanisms of foreshock generation: (a) triggered by past events, (b) loaded by aseismic preslip