SEMINARIS DE LA FACULTAT DE GEOLOGIA I L'INSTITUT DE CIÈNCIES DE LA TERRA 'JAUME ALMERA'
Hora: 12 h.
Lloc:Sala d'actes de l'Institut Jaume Almera

Roberto CARNIEL (Lab di misure e trattamento dei segnali, DICA, Università di Udine, Itàlia; Earthquake Research Institute, The University of Tokyo, Japó)

Starting from speech recognition, Non-negative Matrix Factorization (NMF) is showing an increasing number of applications to different fields, and recently its application was proposed also for volcanic signals. Volcanic tremor often consists of a mixture of signals having different and independent sources, both natural - volcanic and non-volcanic - and possibily anthropogenic. In this talk we show that NMF is a suitable technique to separate such a mixture of foreground / interesting / target "signals" from background / interference / undesired "noise" where the definition of "signals" and "noise" depends of course on what we want to investigate in each specific case. The first encouraging results were obtained separating high convection foreground signal from low convection background noise at Erta 'Ale lava lake in Ethiopia, and support a wider applicability in volcanic signals separation also where only a single station is available, a common situation especially w

In this talk we will focus on the possible application of NMF as a tool to remove the disturbances caused by wind on seismic and especially infrasonic data recorded close to volcanoes. Examples will be shown from Villarrica volcano in Chile and Mt. Shinmoe-dake, Kirishima, Japan.

*Autors de la conferència: G. Cabras (1), R. Carniel (2,4) J. Jones (3), M. Takeo (4), M. Ichihara (4)
1 Dipartimento di Chimica, Fisica e Ambiente, Università di Udine, Italy
2 Laboratorio di misure e trattamento dei segnali, DICA, Università di Udine, Italy
3 Dept. of Earth and Space Sciences, University of Washington, Seattle WA, USA
4 Earthquake Research Institute, The University of Tokyo, Japan