Focused Seismic Monitoring on a 3D Laboratory Dataset Internship

Ultra-sparse acquisition makes it possible to monitor CO₂ storage sites more frequently and at a fraction of the cost of conventional surveys. To support its deployment at scale, the next step is to characterize its performance in a fully controlled setting and to define best practices for survey design. 3D acoustic lab data offer a unique opportunity to do this: its 4D dataset can be used to benchmark the focused seismic approach against a known plume.

The internship aims to answer three questions. How well does sparse acquisition perform, and which parameters control where and when it performs well? How much uncertainty in the survey and in the plume forecast can be tolerated? And what are the consequences of detection errors, particularly false negatives, for CO2 monitoring?

Missions

1. Build a top-view CO2 detection map from spots

• Import and standardize the CO2 lab dataset, reconciling pickle and SEG-Y formats and resolving missing-header issues. Careful data management is an integral part of this work.

• Apply the SpotLight workflow (de-migration, optimal selection, focused seismic processing and detection) to produce, for each survey, a top-down map of the lab model in which each pixel is a single focused seismic spot from one source–receiver pair.

2. Benchmark the focused seismic approach against the lab ground truth

• Evaluate the quality of each spot relative to the ground truth, using attributes such as local topography, signal-to-noise ratio (SNR) and NRMS.

• Benchmark CO2 detection capability by quantifying how much of the plume is resolved as a function of the number of spots, comparing random, regular and optimized down-sampling strategies to establish a cost-versus-detection (value of information) relationship.

3. Assess uncertainty in focused seismic monitoring

• Introduce controlled perturbations to SNR and NRMS and analyze their impact on monitoring results, in relation to spot properties: noise level, geology and structural dip, and acquisition geometry (offset, azimuth, repositioning errors).

• Perform uncertainty quantification on plume forecasting and assess its effect on detection risk, including false positive and false negative rates.

Way forward:

• Repeat the workflow on the public Sleipner dataset, to test the method with real case scenario.

Throughout the internship, the intern will document methods and results, present progress regularly to both supervisors, and deliver a final report and presentation. Depending on results, the work may contribute to a conference abstract or publication.

We are looking for a final-year MSc or engineering student seeking an end-of-studies internship, in geophysics, geoscience, applied physics or a related field. 

Required 

• Solid background in active-source seismic methods: acquisition geometry, seismic processing and imaging fundamentals (migration, wavelets, SNR, 4D/time-lapse concepts). 

• Good programming skills in Python (NumPy, SciPy, Matplotlib), and comfort working with large datasets. 

• Rigorous, organized approach to data handling and documentation. 

• English: fluent or native, French: basic proficiency or conversational.  

Appreciated 

• Knowledge of time-lapse seismic, FWI or CO2 storage monitoring. 

• Familiarity with uncertainty quantification, statistics or signal detection theory. 

• Interest in the energy transition and in bridging academic research and industrial applications.

1️⃣ Entretien avec deux membres de l’équipe, dont la CTO
2️⃣ Cas pratique court