Focus on…Marielle Malfante (GIPSA-Lab) and the Vosica project

Merapi © S Byrdina, IRD
The VOSICA project has been developed in the framework of Marielle Malfante’s thesis, supervised by Mauro Dalla Mura and Jérôme Mars of the Sigmaphy team of GIPSA-Lab (CNRS / UGA / Grenoble INP). Her work is supported by LabEx OSUG@2020 and DGA/MRIS. The aim of Marielle Malfante’s thesis is to develop automatic methods of classification in natural environments. She is currently implementing such tools in underwater acoustics and for the monitoring of volcanoes.

In situ volcano monitoring is an essential step for the safety of people. It notably uses seismometers installed at the surface of volcanoes to monitor mechanical vibrations. Being associated with natural phenomena, these vibrations are very challenging to interpret, especially for predicting volcanic eruptions. Furthermore, when a volcano is continuously monitored, a huge amount of data is produced. The detailed analysis of the acquired signals can no longer be carried out without resorting to advanced signal processing tools.

Marielle Malfante’s goal is to automatically detect and classify, in real time, the various types of recorded signals in order to provide local authorities with a decision support instrument for safety recommendations.

Marielle Malfante uses machine learning in an algorithm that precisely describse signals and classifies them : thanks to previously recorded data, this algorithm is able to analyze and recognize future signals. Three difficulties must be dealt with. Firstly, the data used to build an analysis model for a specific volcano are recorded on a shorter time scale compared to the volcano lifetime. These data are therefore hardly representative of the range of behaviors displayed by the volcano. Secondly, new event types will not be recognized by the automatic analysis since the analysis model is created from previously recorded data. Thirdly, each volcano is characterized by its own pattern of seismic activity and therefore, the tool for the automatic analysis needs to be adapted to the specific conditions encountered.

The VOSICA project, together with the Grenoble Alpes Data Institute, funded a mission for M. Malfante and R. Al Warda (an Indonesian student from Gadjah Mada University enrolled in the MEEES Master program at UGA) to develop and deploy the designed tools for the automatic analysis of seismic signals recorded on the Merapi, the most dangerous volcano of Java Island. The routines were installed, on the servers of the Indonesian BPPTKG observatory. This practical application of her work was a rewarding experience for the PhD student.

Contact scientifique local

- Marielle Malfante, GIPSA-Lab/OSUG | marielle.malfante

Cette actualité a été publiée par

- Grenoble Alpes Data Institute