Improving the Accuracy of Volcanic Earthquake Locations in Montserrat, West Indies

Abstract:

Volcanic activity is usually accompanied by an increase in seismicity linked to magma movement and ascent. This is important not only for monitoring, but also for improving our understanding of the magmatic system. At present, the Montserrat Volcano Observatory (MVO) locates seismicity using a 1D seismic velocity model based on the neighbouring island Guadeloupe, with the location program Hypocenter. This research examines the effect of different velocity models and location methods on hypocenter location to determine whether we can improve locations over the current MVO method.

P and S travel times for the current MVO network of nine stations were calculated from a grid of 336 synthetic earthquakes throughout the island of Montserrat. The synthetic travel times were used to relocate the synthetic earthquakes with the current MVO methodology. These were compared to relocations using other seismic velocity models of Montserrat with the location method NonLinLoc. To have confidence in the location method, synthetic hypocentres should be relocated back to their original earthquake location, within the calculated error – referred to as ‘trusted’ locations. The quality of the relocated hypocentres was assessed by comparing RMS, change in location, calculated errors and the number of ‘trusted’ locations.

Results indicate a low proportion of hypocentres with trusted locations when using the current MVO method. Use of NonLinLoc with the Oct-Tree search method yields smaller errors and changes in location, and a greater percentage of ‘trusted’ locations for all depths tested. Use of an alternative 1D velocity model further increased the number of ‘trusted’ locations and performs better at shallower depths compared to other velocity models.

The whole MVO catalogue, from July 1995 to February 2018, was relocated using NonLinLoc and the Rowe velocity model. Relocations show the majority of seismicity to be beneath Soufrière Hills Volcano at depths less than 5 km bsl, with clusters of seismicity at 1 and 3.5 km. These locations have significantly lower errors than those calculated previously, allowing the seismicity to be constrained with greater confidence during this timeframe. This will be applied to time periods of heightened seismicity such as July 2008 to further understand volcanic processes beneath Soufrière Hills.