Thèse Explorer les Liens Séismes - Glissements Lents avec des Données Sismo-Géodésiques Denses au Chili et au Pérou H/F - Doctorat.Gouv.Fr

Établissement : Université Grenoble Alpes École doctorale : STEP - Sciences de la Terre de l'Environnement et des Planètes Laboratoire de recherche : Institut des Sciences de la Terre Direction de la thèse : Anne SOCQUET ORCID 0000000292087136 Début de la thèse : 2026-10-01 Date limite de candidature : 2026-05-04T23:59:59 Les connaissances actuelles sur le glissement lent restent principalement limitées aux grands épisodes (Slow Slip Events, SSE), généralement d'une magnitude équivalente Mw 6, détectables par GNSS. Les épisodes accompagnés de trémor ou de séismes de basse fréquence sont mieux contraints grâce aux méthodes de corrélation des formes d'onde. En revanche, dans les zones de subduction sans trémor régulier, les épisodes de faible amplitude échappent largement à l'observation. Comme en sismologie, où la détection des petits séismes a révolutionné la compréhension des failles, l'identification systématique des glissements lents de petite amplitude est indispensable pour appréhender pleinement la dynamique des zones de subduction.

Ce projet vise à détecter et caractériser ces transitoires asismiques et à analyser leurs interactions avec la sismicité dans deux régions densément instrumentées du Chili et du sud du Pérou. Nous supposons que le glissement lent participe au chargement des aspérités sismiques, que les séismes répétés correspondent à la rupture de petites zones verrouillées au sein de domaines en fluage, et que le spectre de glissement reflète l'hétérogénéité frictionnelle et la rugosité de l'interface. Comprendre la relation spatio-temporelle entre déformation asismique, sismicité de fond, essaims et séismes répétés est essentiel pour contraindre les modèles de préparation sismique.

La thèse s'appuiera sur les jeux de données sismo-géodésiques acquis dans le cadre du projet ERC DEEPtrigger : séries GNSS et InSAR, catalogues sismiques haute résolution et méthodes avancées de détection. Les objectifs sont de : (1) construire des catalogues de séismes répétés, d'essaims et de glissements lents ; (2) localiser précisément les événements sismiques et asismiques sur l'interface et caractériser leurs paramètres ; (3) quantifier le couplage spatio-temporel entre glissement lent et sismicité ; (4) rechercher d'éventuels processus précurseurs ; et (5) tester par modélisation les conditions mécaniques favorisant l'interaction sismique-asismique, notamment le rôle de l'hétérogénéité frictionnelle et des fluides.

L'analyse géodésique consistera à identifier des transitoires subtils dans les séries GNSS et InSAR après retrait des tendances intersismiques, offsets cosismiques et relaxation post-sismique. Les signaux résiduels seront étudiés par filtrage, stabilisation en référentiel local, empilement et décomposition (PCA, ICA), puis inversés afin d'estimer les distributions de glissement sur l'interface.

L'analyse sismologique reposera sur l'identification et la relocalisation de séismes répétés, utilisés comme jauges du glissement, ainsi que sur l'étude statistique de la sismicité déclustérisée afin de suivre l'évolution des taux de sismicité de fond, indicateurs sensibles des variations de contrainte.

Des approches d'apprentissage automatique permettront de caractériser les relations empiriques entre glissement lent et signatures sismologiques à partir de descripteurs physiquement interprétables. Enfin, la modélisation mécanique contraindra les propriétés frictionnelles, la contrainte normale effective et la pression de fluide afin d'explorer les conditions contrôlant l'interaction entre glissement lent et rupture dynamique dans les zones de subduction. Present knowledge of slow slip phenomena is restricted to large Slow Slip Events (SSE), typically with equivalent Mw of 6 or greater, which can be measured by geodetic observations, typically GPS. Episodes of slow slip with seismic signatures (e.g. low frequency earthquakes or non-volcanic tremor) are much better characterized because template-matching methods allow their detailed monitoring. For subduction zones without regular tremor activity, we face the challenging task of deciphering the slow slip kinematics only knowing the largest slow slip events. The history of seismology has shown that the detection of small to very small earthquakes from the 1960s onwards, with the deployment of better, denser networks, and the further improvement of seismic processing methods (from multiplet relocation to noise cross-correlation) is of paramount importance: knowledge of small events is needed to understand and monitor faulting. The same is true with SSE in regions without tremor: knowledge of the time history of SSEs, and their space-time pattern of occurrences, in relation with «regular» earthquakes in the region (i.e. background seismicity rate and repeating earthquakes that are other markers of slow slip on faults), will allow the development of models that account for the full dynamics of convergence in subduction zones, which are still lacking.
It is needed to develop and systemize our detection of small slow slip events to understand the mechanics of subduction zones, in particular their interactions with seismic ruptures, and potentially impending destructive earthquakes. How aseismic slip can trigger earthquakes should be investigated, and how this triggering depends on depth, duration, migration, periodicity, and total amount of slip. We posit that deformation and seismicity are related, and that their relationship can be seen as a proxy for friction on the interface. More precisely, we suppose that: slow slip events contribute to the loading of seismic asperities, with repeating earthquakes representing the rupture of small asperities resisting this slow slip; slow slip events preferentially occur in partially locked areas, that we assume to be highly heterogeneous in terms of friction mode (interfingering of seismic asperities and metastable areas); the frequency content or slip spectrum of slow or dynamic ruptures is representative of the roughness of the fault surface. The objectives of this PhD project are to:
1. Build catalogues of repeating earthquakes, earthquake swarms and slow slip events
2. Precisely map the seismic and aseismic events on the plate interface, characterize their slip spectrum and stress drop
3. Depict the link and its time-space evolution between slow slip pulses and the characteristics of the seismicity
4. Look for potential premonitory processes leading to dynamic failure
5. Test what mechanical conditions favour the emergence of the observed seismic-aseismic interplay by building a model with heterogeneous frictional properties to model asperities, and by exploring the role of fluid pressure on the initiation of slow slip and seismic rupture.
This PhD will build on data sets, catalogues and methods developed in the frame of the DEEPtrigger ERC project: dense seismo-geodetic data acquired in Atacama (Chile) and in south Peru, dense catalogues of seismicity, GNSS and InSAR time-series mapping the evolution of the surface deformation, and on methods to for detecting and characterizing slow slip episode.

Geodesy
To look for hidden transients in GNSS and InSAR, the residual time series will be analysed, from which know tectonic and non-tectonic component will beforehand be removed by applying a trajectory model accounting for the average interseismic loading, the main co-seismic offsets and post-seismic relaxation. The residual time-series will then be carefully characterized and explored. Noise will be reduced by removing a common mode, or by stabilizing the solution with respect to a local reference frame. Signal processing techniques such as decomposition with Independent Component Analysis (ICA) or Principal Component Analysis (PCA), filtering, and/or stacking will be applied.
The obtained values of interseismic rates, co-seismic offsets, afterslip and slow slip events will be inverted to map the slip associated with the main phases of the seismic cycle using PCAIM/ICAIM, a geodetic time series inversion software.

Seismology
The existing repeating earthquakes will be grouped together, by relocating the highly similar waveforms. The goal is to use REs as in-situ slip gauges, since short inter-rupture times between two co-located earthquakes implies a rapid loading of the corresponding asperity. Quasi-periodic RE time series that display sudden departure from periodicity are typical signatures that will be tracked for. Systematic trends in the relative magnitude of REs will be searched for, as it has been shown that the slip rate of the surrounding fault partly controls the size of the subsequent rupture.
Seismicity catalogues will be statistically analyzed to extract the dynamics of seismicity. To do so, we will estimate whether the observed earthquake time series display stable characteristics or not, in particular regarding the earthquake rate in the absence of notable aftershock sequences. Changes in background rates are a very useful in-situ observable of the evolution of the loading conditions on the fault. Background seismicity will be extracted by discarding aftershock sequences (declustering), using 3D space-time stochastic models of seismicity parameterized to the targeted areas.

Characterization with Machine Learning
We also want to characterize the empirical relationship between these slow slip events and their seismological signature, and to monitor how this relationship evolves with time and space. This can be approached as a regression problem that can help detecting anomalous trends in the observations, or changes in the underlying physics. A particular attention will be given to the optimal representation of our signals by selecting relevant descriptors. A large variety of features will be used in different domains (e.g., temporal, spectral and cepstral domains). Data descriptors should have a physical sense that can be directly related to specific characteristics of the physical phenomena considered, so that they are more easily interpretable and later used to constrain models of the processes.

Mechanical modelling
We will test functioning mechanisms for the initiation of megathrust earthquakes through mechanical modeling using our observations as a constraint, notably by exploring the potential role of fluids in the fostering of the observed seismic-aseismic interactions. For this purpose, the inverted slip patterns will be used to infer frictional properties, effective normal stress and pore fluid pressure using the code UNICYCLE.

Diplôme de Master en physique, géophysique ou dans un domaine connexe, avec une capacité à l'analyse de données et la programmation (Python, MATLAB ou équivalent). Une curiosité pour les géosciences et les aléas naturels est requise, ainsi qu'une motivation à travailler à l'interface entre sismologie observationnelle et modélisation géodésique.