IRAP – OMP, Université Paul Sabatier, Toulouse, France
Contact: Jean Letort | jean.letort@irap.omp.eu
2-Year Postdoctoral Position
Deep Learning for Shallow Earthquake Depth Estimation: Teleseismic Methods and Applications to Stable Continental Regions
Scientific Context
Reliable depth estimation of shallow earthquakes (< 15–20 km) remains one challenging problems in observational seismology, particularly for small-to-moderate magnitude events (M < 5) in regions with sparse instrumentation. This postdoctoral position is part of the ANR-funded project ShallowDepthDL, which aims to develop novel Deep Learning (DL) approaches to improve depth estimation across spatial scales. The selected candidate will be primarily responsible for A: developping a new IA-based teleseismic depth estimation method and B: analysing the seismogenic depth of Stable Continental Regions, in close collaboration with the broader project team.
Task A – Deep Learning for Teleseismic Depth Estimation of moderate magnitude events (Magnitude 4-6)
At teleseismic distances (30°–90°), focal depth can be constrained from the delay between the direct P wave and its surface reflections (pP, sP). However, detecting these depth phases at low signal-to-noise ratios across thousands of stations remains labour-intensive and difficult for shallow events (< 20 km). The postdoc will develop and train Convolutional Neural Networks (CNNs) operating on images of stacked teleseismic envelope sections, building on a methodology developed at IRAP. One interrest of this approch is that no reference depth phases picks are needed, but a trustable reference database of well located events which can be obtained by other ways (local/regional network, focal mechanism inversion, INSAR). Different tasks are targeted: (1) learning to identify coherent pP/sP alignments in the teleseismic sections and convert them into depth estimates; (2) evaluating the posssibility to identify Very Shallow Earthquakes (VSEs, < 2–3 km) based on the absence of resolvable depth phases; and (3) integrating focal mechanism information and synthetic seismogram comparisons to improve depth phase identification in ambiguous cases. A central component is therefore the construction and curation of a global reference database of well-constrained shallow and very shallow events, aslo a key point for task B.
Task B – Studying the seismogenic depth in Stable Continental Regions
The CNNs developed in A will be applied to investigate the depth distribution of seismicity in Stable Continental Regions, where depth uncertainties are currently large and debated. Especially, past moderate magnitude events (1990-2010), where fewer local networks are availlable will be investiguated. Constrained local/regional depth estimation, depth solutions provided by INSAR (very efficient mostly for indentifying very shallow events) will also be combined to teleseismic depth estimation method to provide a reference catalog for the seismogenic depth of stable regions.
Across these different stable regions, the postdoc will contribute to a reassessment of whether seismicity is confined to the upper crust or shows a secondary peak near the Moho, of which conditions (temperature, age of the crust, Moho depth..) control the depth of the seismogenic detph, of whether the number of very shallow earthquakes (<3km) could have been underestimated over the past, with direct implications for seismic hazard in the intraplate context. Focus could be made on specific intraplate regions with important shallow seismcity (Mongolia, Australia, Eastern US).
Candidate Profile
• PhD in seismology, geophysics, signal processing, or a closely related field
• Experience in seismic data processing and/or earthquake location methods, programming, experience with deep learning frameworks (PyTorch or TensorFlow)
• Aiming/ready for several missions between the different partenairs of the ANR (GFZ, Potsdam, Germany, CEA – French Atomic Agency, Paris, France, EDF, Aix-en-provence, France & CSIC, Barcelona, Spain)
• Good communication skills in English;
Position Details
Duration: 2 years | Start date: October 2026 | Location: IRAP – Institut de Recherche en Astrophysique et Planétologie, OMP, Toulouse, France | Funding: ANR ShallowDepthDL
Applications (CV, cover letter) to: jean.letort@irap.omp.eu