Institut für Geophysik, Stuttgart, Germany

Broadband seismic measurements performed in 1995 and 1996 in the summit region of Stromboli are analyzed. The experiment in 1995 used an array of four Guralp seismometers and one Wielandt-Streckeisen seismometer. The stations were installed around the craters in a semicircle with a radius of about 500 m. This implies that the seismic signals are dominated by near field motions up to frequencies of about 2 Hz. The observed Strombolian explosions are preceded by long-period ground motions occurring between 20 s and 70 s prior to the ejections. They are obviously generated by a slow pressure increase within the magma conduits. The long-period signals are simple compared to the short period wave forms. Four classes of pulse-shaped seismograms can be distinguished. The radiation pattern is radially symmetric with respect to the crater region. Particle motion analysis indicates that the seismic sources are located between 50 and 200 m below the crater terrace. Hydrostatic model sources were studied by means of finite element calculations with different geometries, i.e. ellipsoids, in a solid cone modeling the topography of Stromboli. The results suggest that the explosive events on Stromboli originate from a shallow vertically elongated volume source.

This investigation deals with the nature of the long-period seismic signals (>1 s) observed at Stromboli and addresses the question whether they are of volcanic origin or produced by sources such as Ocean Microseisms (OMS). We present results from the analysis of seismic broadband data recorded during 1992 by an array of 9 Guralp CMG-3T seismometers. The determination of the Array Response Function (ARF) shows that array techniques like delay-and-sum beamforming cannot be applied for this purpose, as the extension of the array is limited by the geographical constraint of the island of Stromboli volcano, being simply too small. Spectral analysis reveals three main peaks with periods at 4.8 s, 6 s and 10 s which are not stable in time but vary according to the regional meteorological situation. Whereas 4.8 s and 10 s show up in amplitude spectra calculated during rainy and stormy weather, the 6 s period can be observed during a period of good weather. The signals were first narrowly filtered and then cross correlation, particle motion and amplitudes of the main long periods studied in detail. Relative arrival times as well as seismic amplitudes of the filtered traces do not show any systematic feature but vary with time. Particle motion analysis demonstrates that all long-period signals are recorded by the array as plane waves and that the main propagation direction of the 10 s signal is parallel to the wind direction. No correlation with volcanic activity is obvious. We conclude therefore that the three main long periods are not generated by a close volcanic source. We assume a local cyclone to be the seismic source at 4.8 s and 10 s, which represent the Double Frequency (DF-band) and the Primary Frequency (PF-band), respectively. Concerning the 6 s peak, we speculate a cyclone near the British Isles to act as a seismic source.