The “Jerk” Method for Predicting Intrusions and Eruptions of Piton De La Fournaise (La Réunion Island) from the Analysis of the Broadband Seismological Rer Station

Thursday, 18 December 2014
Genevieve C Roult, Francois Beauducel, Valerie Ferrazzini, Patrice Boissier and Nicolas Villeneuve, Institut de Physique du Globe de Paris, Paris, France
The predictability of volcanic eruptions remains a challenging problem and forecast of volcano behavior (intrusion versus eruption) is a difficult task. Since 1979, the Volcano Observatory of Piton de la Fournaise (OVPF) maintains SP and BB seismic stations, tiltmeters, extensometers and GPS stations. In addition, the RER GEOSCOPE broadband station was installed in 1986, 8.5 km north of the summit crater. The analysis of 83 seismic crises from December 1985 to December 2010 (preceding 54 eruptions, 26 intrusions, 2 summit pit craters,1 caldera collapse) allowed us to identify short-term long period seismic transients (period > 100 s) for most events (Roult et al., 2012). These precursors observed on the horizontal components are tilt signals induced by the inflation/deflation of the volcano. We analyzed 17 eruptions and 7 intrusions spanning the 2005-2010 period with the aim of distinguishing whether an injection of magma will stop or if it can evolve towards an eruption. Transient signature is an acceleration step that can be large or not, with slopes more or less steep according to the acceleration rate. We show a clear differentiation between the acceleration rate of the intrusions (low rate) and the acceleration rate of the eruptions (high rate). With a ratio estimated to 7, the acceleration rate allows to determine a threshold value and to discriminate between intrusive and eruptive events. The real-time calculation of the ground acceleration of the horizontal components of the RER station after removal of the theoretical tide effect is integrated since April 2014 to the Piton de la Fournaise volcano monitoring. In June 2014, the “JERK”method predicted an eruption 50 minutes before the eruption onset.

We applied the "material failure prediction" of Voight and Cornelius (1991) with the aim to predict the onset time of the eruptions. Preliminary tests on the 17 eruptions of the 2005-2010 period have shown that the summit eruptions were relatively well predicted (error of a few minutes). The distal eruptions seem to be systematically predicted too early. This approach will be added to the monitoring system.

Roult et al., 2012. JVGR, 241-242, 78-104

Voight and Cornelius, 1991. Letters to Nature, 350, 695-698.