Ongoing development of pattern classification techniques applied to volcanic tremor data at Mt. Etna

Abstract

Timely identification of changes in the state of volcanoes and onset of potentially dangerous eruptive phenomena requires efficacious surveillance methods. In the case of an active volcano like Mt Etna, the continuous background seismic signal called volcanic tremor is of paramount importance. The huge amount of continuously acquired digital data entails the necessity of data reduction and parameter extraction. For this purpose, techniques of automatic analysis of volcanic tremor were applied by INGV for the real time monitoring of this signal. We checked the possibility to identify regimes of volcanic activity based on pattern classification of volcanic tremor. A specific software named “KKAnalysis” was developed. It combines various unsupervised classification methods (Kohonen Maps and fuzzy cluster analysis) and forms the backbone of an automatic alert system at INGV-OE. Besides its near real time application, it can be operated off-line, allowing an efficient a-posteriori processing of data and tuning of the alarm criteria to match specific needs of sensitivity and robustness. An ongoing development of this tool will allow us to include a large number of seismic stations in a multistation-alarm system. The new system will be more robust in case of failure of single sensors, and will achieve a better coverage of the various eruptive craters. In an off-line test, we exploited a dataset covering eight years of seismic records, and analysed the performance of the new system in terms of “trigger timing” and spatial distribution of the stations. Intriguing results were obtained throughout periods of renewal of volcanic activity at Bocca Nuova-Voragine and North East Crater, and in the absence of paroxysmal phenomena at South East Crater and New South East Crater.