A process-based model of pre-eruption seismicity patterns and its use for eruption forecasting at dormant stratovolcanoes

Using long-term data sets, the eruptive size, timing and style can typically be successfully forecast for dormant volcanoes using current monitoring data and knowledge of precursory eruptive patterns. The USAID-USGS Volcano Disaster Assistance Program (VDAP) utilizes seismic data over multiple eruption cycles and across volcano types to successfully forecast eruption size, style and onset times, as well as changes within ongoing eruptions. From millions of seismic event observations, we show that seismic data, a proxy for rate-dependent strain changes in the magmatic system, combined with a process-based conceptual geologic model enables eruption forecasting, especially for stratovolcanoes dormant >20 years. We show that this model often works for some open systems, ones that sufficiently reseal to plug the conduit, and also where a deep intrusion occurs after phreatic or phreatomagmatic explosions have begun. This model is divided into four seismicity and geologic stages. We describe each of the four stages of seismicity in detail and how they relate to the ongoing geologic process of that stage: Stage 1. Deep intrusion and seismicity occurring at 10 to 40 + km depth below the volcanic edifice; Stage 2. Distal VT seismicity induced at distances from 2 to >30 km laterally from the summit by magma intruding into and inflating the upper-crustal magma storage region; Stage 3. Vent-clearing seismicity associated with the initial vent creation and clearing phase of eruptive activity. Stage 4. Repetitive event seismicity associated with the final ascent of magma from 2 km to the surface. We present 36 examples of this seismic progression from 26 volcanoes. We believe that this progression generally correlates with the magma flux and the eventual explosivity and is invaluable for forecasting eruptions. (C) 2019 Published by Elsevier B.V.