bandeau
Accueil Inscription Soumission des résumés Programme Liste des participants Liste des résumés soumis

Automatic detection of Interplanetary Coronal Mass Ejections from in-situ data: a deep learning approach

Auteur

Nguyen Gautier

Institution

LPP

Thème

Theme1
Auteur(s) supplémentaire(s)Nicolas Aunai, Dominique Fontaine

Abstract

Interplanetary coronal mass ejections (ICME) are the interplanetary manifestation of Coronal Mass Ejections (CME). They are generally identified by characteristics such as an enhanced and smoothly rotating magnetic field, low proton temperature, declining velocity profile and low plasma beta. However, these features are not all observed for each ICME due to their strong variability. Manual detection of ICMEs through visual inspection is thus a time-consuming and fastidious task biased by the observer interpretation leading to non exhaustive, subjective and hardly reproducible catalogs. Moreover, the use of empirical thresholds on all or a subset of these criteria to automatize the detection is not flexible enough and leads to lots of missed detection or false positives.

Here, we use convolutional neural networks on different windows of data to provide a fast and automatic detection of ICMEs that uses no knowledge specific to ICMEs but only learns from the data itself.

The method was tested on the in-situ measurements provided by WIND between 1997 and 2015 and on the 657 ICMEs that were recorded during this period. In addition to providing automatic ICME catalogs with few errors and with a fair number of ICMEs, the method offers a unambiguous visual proxy of ICMEs. Although less accurate, the method also works with one or several missing input parameters and has the advantage of improving its performance by just increasing the amount of input data. Finally, the generality of the method paves the way for automatic detection of many different event signatures in spacecraft measurements.


Retour à la page précédente