Presenter: Chagnon Glynn
Date: 2016/03/17
Abstract
Volcanoes are complex and dynamic systems, which for decades have been monitored using seismological, geochemical and geophysical techniques (e.g. GPS, strain and tilt monitors), all of which have very limited real-time monitoring and forecasting applications and can become quite expensive to implement and maintain. Using ambient seismic noise, an omnipresent feature in recorded seismograms that are generated by the interaction of atmospheric disturbances and human activities (stochastic sources), allows for characterization of the medium properties using only one seismic station. This study determines measured complexity variations in ambient seismic noise using permutation entropy (PE) to forecast volcanic eruptions. To do this we used records from the HOTSPOT seismic network for the 1996 Gjálp eruption (the eruptive phase lasted from 30th SEPT to 4th OCT). Results show PE reached a minimum ~7-9 days before an earthquake (Mw=5.6) and ~9-10 days before the subglacial eruption. The onset earthquake immediately triggered PE to rebound and steadily increase to maximum until the subglacial eruption. This indicates that the ambient noise was stochastic up until ~7-9 days before the earthquake whereby it’s characteristics changed to a more deterministic nature and once the subglacial eruption occurred the ambient noise reverted to it’s “normal” stochastic nature. However, only stations located inland within 50-90km of the ring fault (near Bardarbunga and Grímsvötn volcanoes) were able to capture the complexity variations suggesting a spatial limitation. The observed complexity variations are likely due to changes in the scattering properties in the shallow crust surrounding the volcanoes, which expand and contract spatially depending on the caldera’s extent. The pressurization, caused by magma injection into the caldera changes the medium’s properties from a heterogeneous to more homogeneous nature, thereby reducing scattering (and stochasticity) and allowing a more volcanic deterministic source signal to dominate.