Ph.D. Student: “Determining the Geometry and Slip of an Earthquake from Surface Observations Using Machine Learning" M/F

Job Description

Thesis Subject

Determining the geometry and slip of an earthquake from surface observations using machine learning.

Your Work Environment

During an earthquake, most of the co-seismic slip occurs along faults deep underground. However, most of our observations are limited to the surface via remote sensing. Reconstructing the geometry of the rupture and the three-dimensional slip from these surface observations is a fundamental yet complex problem in geophysics. The ERC BE_FACT project (Boxing Earthquakes and Faults in Active Tectonics) aims to address this problem through laboratory experiments and numerical simulations to study the relationship between fault geometry, rupture dynamics, and surface deformation. This thesis topic is part of the BE_FACT project.

Objectives :
The goal of this thesis is to explore the extent to which, for a given earthquake, the geometry of the rupture and the distribution of slip during that rupture can be inferred from observed surface displacement fields—possibly supplemented by seismological data—using machine learning methods (statistical or deep learning). As part of this thesis, we will use a dataset of approximately 1,000 numerically simulated earthquakes (already partially available) for which the source characteristics and 3D displacement fields (including at the surface) are known. The student will analyze these datasets to determine the degree of complexity of the inverse problem as a function of the target accuracy level, identify the most relevant characteristics of the surface signals, and develop machine learning methods to reconstruct the slip (or its main characteristics) at depth.

Research Plan
The thesis will begin with an exploration of simulated earthquakes to understand the patterns of displacement at the surface and at depth. The student will first extract and analyze simple characteristics (average slip, affected area, geometry, orientation) and apply dimension reduction techniques to identify the main patterns of variation in the data. Based on these characteristics (and, if necessary, more advanced characteristics), prediction models will be developed, ranging from simple statistical relationships to more flexible models such as artificial neural networks. The performance and uncertainty of the models will be evaluated using events hidden during the training phase to test the models' ability to generalize and avoid overfitting. In a second step, depending on the performance of the initial models, the goal will be to increase the complexity of the source models to better approximate more realistic geometries. While an initial approach will not make any assumptions about the rheology of the modeled medium, once the prediction models are more advanced, we may also consider introducing some a priori information into the learning process (physics-informed neural networks) to gain a better understanding of both the seismic source and the modeled physical medium. In the third part, the objective will be to use the models. In the third part, the objective will be to use the models developed to analyze actual surface deformation measurements for earthquakes, obtained from InSAR or optical measurements.

Profile ;
This thesis is intended for students with a Master 2 degree in geophysics, data science, or mechanical engineering who are interested in deformation processes. Proficiency in Python for data analysis is required. Experience with geodetic data or inverse problems is a plus but not required. The Ph.D. student must be able to work independently and discuss results with the team.

Compensation and benefits

Compensation

2300 € gross monthly

Annual leave and RTT

44 jours

Remote Working practice and compensation

Pratique et indemnisation du TT

Transport

Prise en charge à 75% du coût et forfait mobilité durable jusqu’à 300€

About the offer

About the CNRS

The CNRS is a major player in fundamental research on a global scale. The CNRS is the only French organization active in all scientific fields. Its unique position as a multi-specialist allows it to bring together different disciplines to address the most important challenges of the contemporary world, in connection with the actors of change.

CNRS

The research professions