Abstract
The persistent seismic signals recorded on basaltic volcanoes, known as volcanic tremor, have proven to be one of the most significant measures for volcanic surveillance. Nevertheless, the continuous acquisition of the tremor signal produces masses of data (something like 80 MB per station per day) which are cumbersome to handle. Data condensation and parameter extraction, which can be carried out automatically, are highly recommended for the study of the long-term behavior of a volcano. In this context it is necessary to identify parameters which can be correlated with volcanic activity. The surveillance can be thus reduced to the monitoring of key parameters which speed up the data processing as well. In the light of these goals, we present the analysis of seismic data recorded on Stromboli between 1990 and 1998. We describe how we obtained a drastic data reduction carrying out multivariate statistical analysis, and our choice of the parameters to monitor. The problem of data reduction concerning transient signals – mostly explosion quakes at Stromboli – may be tackled by carrying out an automatic supervised classification. For this purpose we propose the application of Artificial Neural Networks. Finally, we discuss the relationship between the parameters considered and volcanic activity visible at the surface. We believe that the basic concepts outlined here could be generalized in a sense that they may be used for the definition of a monitoring system based on data reduction and parameterization.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have