• Time-series analysis;
  • Numerical approximations and analysis;
  • Explosive volcanism;
  • Eruption mechanisms and flow emplacement;
  • Remote sensing of volcanoes;
  • Volcano monitoring


Observations of Strombolian volcanic explosions were carried out at Etna's southeast crater on 2001 July 4 using a ground-based pulsed Doppler radar (VOLDORAD). To obtain quantitative constraints on the geometry of the explosions, we modelled synthetic Doppler spectra by combining the outputs of a ballistic model to compute the theoretical velocities of gas and particles, and an electromagnetic scattering model to calculate the synthetic echo power. This allowed us to reproduce the shapes of recorded Doppler spectra for each volcanic explosion. We examined the geometrical distribution of ejected pyroclasts for about 200 explosions and found two main types of explosion, each showing a distinctive spectral signature. The first type, characterized by the triangular shape of their Doppler spectra, represents 34 per cent of the explosions. This spectrum shape is related to a Gaussian distribution of the pyroclast ejection angles, where most of the volcanic material is ejected vertically within a narrow cone, with the particle concentration decreasing radially. The second type represents about 12 per cent of the explosions, and is characterized by a top-hat-shaped spectrum. It is produced by a uniform distribution of pyroclast ejection angles. In this case, the bubbles tend to burst above the crater rim and eject the ballistic clasts hemispherically without preferential orientation. The majority of the Strombolian explosions analysed (54 per cent) are intermediate between these end-member shapes, and show a triangular spectra truncated by a plateau. They result from a uniform distribution of ejection angles around the jet axis.

The continuous radar recordings allowed us to carry out a statistical analysis on the geometrical features of the same 200 Strombolian explosions. Thus we find that 40° is a statistically representative aperture of the dispersion cone characterized by uniform ejecta distribution for explosions having a plateau component (i.e. 2/3 of all explosions studied). We also find that 80 per cent of the total particle load remains within this 40° dispersion cone, which in our case is oriented nearly vertically. Such quantification of the geometrical features of Strombolian explosions can provide constraints on shallow conduit processes. For example, the time evolution of the top-hat Doppler spectra during the paroxysm suggests that the magma level varies over a 15 min timescale.