• Citer

On the use of the machine learning algorithms for automatically classifying cross-correlation functions: Application to the seismic data obtained around the Vatnajökull icecap, Iceland

par Forestier, Justine
Président du jury Mattielli, Nadine
Promoteur Lecocq, Thomas
Co-Promoteur Nowé, Sylvain
Publication Non publié, 2022-06-27

Mémoire

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

Ambient seismic noise interferometry allows us to see patterns in correlograms. These patterns have been interpreted as different sources of noise. One of them is seasonal, coming back each year in spring and shutting off just before winter. It is supposed to be linked to glaciers movements, either glacial quakes or more probably rivers and associated waterfalls fed from the seasonal melting. A better understanding of the characteristics of these processes and the ability to detect them in near real time could help prevent natural disasters as flooding or jökulhlaups.ewlineHowever, the study of such events is made on a very large amount of data, difficultly treated by a human operator. Artificial intelligence is a way to process big data. Machine learning, a branch of artificial intelligence, is more and more used in earth science, especially when it comes to data classification. This work aims to assess if these seasonal signals can be discriminated from other signals automatically through a random forest supervised machine learning algorithm. In order to find relevant data features that will be used to classify the data, exploratory data analysis will be done through unsupervised learning (similarity and clustering). A secondary goal in this work will be to examine the features and information extracted from these classification preliminary steps to see if a better understanding and characterisation of seasonal ambient noise can be achieved.ewlineThis work shows that machine learning algorithms succeeded in identifying the difference between seismic noise sources. They can be used on a daily basis to discriminate a seasonal source of seismic noise while assigning it a probability rate.