Fig. 1. Map of Stromboli volcano with the INGV-CT permanent monitoring network before 2000. In the inset are reported the 9 IGS stations included in the continuous GPS data processing. General tectonic lineaments and volcano-tectonic collapse are also reported.

Fig. 2. Cumulative strain release in Stromboli area from June 1985 to December 2006 (modified from Falsaperla et al., 2003).

Fig. 3. Hypocentral location obtained by ray-tracing method. On the left: theoretical Ts–Tp vs. epicentral distance (top) and Ts–Tp vs. depth (middle) plots on the basis of four different velocity models. Grey colours mark epicentral distance and depth ranges achieved by the observed Ts–Tp data for an average angle of incidence of 6° or 12°. At the bottom-left: the resulting epicentral area (grey) for the 1999 seismic swarm, compared to the best analytical epicentres (stars) reported in Falsaperla et al., 2003. At the top and middle right: the same as on the left for the 2006 earthquakes. An average angle of incidence of 12° was assumed. The 2006 epicentral area (grey at bottom-right) is compared to best location (at about 6 km of depth) based on travel-time residuals method (black circle). Grey circle indicates the localization of the same earthquake achieved with a seismic network set as in 1999. Data from Jeffreys and Bullen (1967), Chiarabba and Frepoli (1997), La Rocca et al. (2004) and Patane (pers. comm., 2007).

Fig. 4. On the left: correlation matrix for the 78 analyzed earthquakes at STR station. A colour scale bar, related to correlation coefficient values, is shown. The squares mark the two detected families (see text). On the right: as an example, two earthquakes of duration magnitude of 3.2 (grey) and 2.4 (black) with cross-correlation value of 0.77 are reported. Amplitudes of seismic signals are normalized.

Fig. 5. Top: sequence of three earthquakes (yellow) occurring on June 13, 1999. In red, the same sequence after applying a 2.5 Hz low pass filter. The different amplitude attenuation of seismograms evidences that the third seismic event is characterized by lower frequencies than the other ones. Bottom: spectrogram obtained by 128-points FFT with 50% overlap. The different frequency content of the last event is enhanced. This event, characterized by high-frequency onset followed, a few seconds later, by lower frequencies than the other ones, is presumed to be a “hybrid” event. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 6. Y₁ and e_xx strain tensor components (the squares represent daily values and the grey line is an interpolation curve). Dashed vertical lines indicate the 1999 June swarm period.

Fig. 7. SPLB tilt signal recorded from 1993 to 2006.

Fig. 8. Three-dimensional model sketch of the intrusion producing the Stromboli 2000 geodetic anomaly. Projection of the estimated four phases in plane and cross-sectional views are also reported.

Fig. 9. Spatial and temporal distribution of dike opening component for each phase. Modelled and recorded signals at the SPLB tilt station are also reported.

Fig. 10. Modelled temporal change of the volume of the intrusion for all phases.

Estimated parameter and uncertainties

Coordinates in UTM (WGS84). The reference surface (z = 0) has been assumed at the sea level, which is approximately the altitude of stations used in the inversion.