Skip to Main content Skip to Navigation
Theses

Identification and characterization of Mass Transport Deposits from seismic data. Application to the Amazon River Mouth basin

Abstract : Understanding the processes leading to the formation of a Mass Transport Deposit (MTD) allows a better knowledge of the potential resources and geohazards associated to a sedimentary basin. Seismic data interpretation is used to study MTDs. Two methodologies are developed in this PhD thesis, in order to infer the physical processes that may have impacted the current aspect of an MTD in a seismic image. The “identification” methodology allows to locate the position and extent of MTDs in seismic images, while preserving their internal seismic facies variability. This methodology is based on textured image segmentation, jointly with a weakly-supervised learning of probabilities of MTD occurrence on seismic images. The “interpretation” methodology provides hypotheses to explain the various characters of an MTD, in terms of physical processes. These hypotheses are retrieved through a knowledge base built from the literature and used as an inference engine, thus highlighting the interpretation process. Both methodologies are successfully applied to a seismic dataset acquired in the Amazon basin (Brazil). They enhance the joint use of data-, knowledge- and model-driven approaches.
Complete list of metadatas

Cited literature [361 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02613626
Contributor : Abes Star :  Contact
Submitted on : Wednesday, May 20, 2020 - 11:30:10 AM
Last modification on : Tuesday, May 26, 2020 - 8:44:13 AM

File

LE_BOUTEILLER_Pauline_2018.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02613626, version 1

Citation

Pauline Le Bouteiller. Identification and characterization of Mass Transport Deposits from seismic data. Application to the Amazon River Mouth basin. Earth Sciences. Sorbonne Université, 2018. English. ⟨NNT : 2018SORUS481⟩. ⟨tel-02613626⟩

Share

Metrics

Record views

24

Files downloads

2