Spatial variations of rockfalls from a growing lava dome: Soufriere Hills Volcano, Montserrat

The lava dome at Soufriere Hills Volcano poses a physical hazard to nearby inhabitants. When the lava dome collapses, pyroclastic flows are generated that can travel several kilometers from their source. Understanding the behavior of lava domes is often performed using rockfalls produced by gravitational instabilities on the lava dome. Rockfalls generate seismic signals from which seismic energy can be extracted. Previous work has demonstrated that rockfalls may be locatable based upon their seismic energy, and this work develops a continuation of that approach. Rockfall direction and intensity is most likely determinable by seismic energy distribution as measured by the seismic station network, thus not requiring direct visual analysis, as clouds often cover the lava dome at Soufriere Hills Volcano. In this work, a technique was developed, which allows seismic data to be used to both identify and locate small rockfalls, a predecessor to most lava dome collapses.;Digital photographs of the lava dome from remote cameras and rockfall seismic energy data from 11 February 2006 - 1 May 2007 were utilized to visually confirm rockfall seismic event locations. From a set of 25, 297 seismic events, 1180 of these were visually confirmed and assigned a controlled location based in one of four quadrants surrounding the lava dome (i.e. quadrant one represented north, quadrant two represented east, etc.). From this dataset, a seismic energy-based mapping system was implemented; contours of equal energy were calculated using a cubic spline interpolation based on the total energy recorded for each seismic event. The energy mapping displayed each rockfall location in one of four quadrants, displaying results from both the cubic spline interpolation and the control dataset. Two different energy maps were created to compare the initial and final directions of rockfall runout.;Results of the seismic energy-based mapping yielded event initial rockfall location accuracies of 58%, 48%, 9% and 13% for quadrants one through four respectively. Final rockfall location accuracies were 61%, 41%, 10%, and 14% for quadrants one through four respectively. Ratios of seismic energy between seismic stations often produced two distinct trend lines of strong data scattering for seismic events located in quadrant one; both phenomenon were believed to be caused by path effects on seismic energy. Visually, the energy mapping was able to accurately indicate the direction of dome collapse prior to collapse events on 20 May 2006, 30 June 2006 and 8 January 2007. Actual rockfall runout direction was found to be controlled by current lava dome dimension. Rockfall frequencies prior to collapse were variable and not necessarily indicative of collapse immanency. Sources of inaccuracy were identified to be path effects on seismic energy distribution, and the availability and quality of seismic data. With improved accuracy, the method of seismic energy-based mapping of rockfalls could be a useful tool in real-time hazard mitigation, having already demonstrated an ability to forecast the direction of dangerous lava dome collapse material runout.