A catalog of volcanic earthquake swarms in the United States: comparative analysis and use in eruption forecasting

Swarms of earthquakes are often associated with eruptive processes, but many swarms occur near volcanoes that are not easily associated with eruptions, complicating their use in eruption forecasting.  In some cases, swarms may be caused by hydrothermal processes and can be considered as part of a volcano’s normal background seismic activity. Other swarms near volcanoes may be considered purely tectonic in origin, or some combination of tectonic, magmatic, and hydrothermal processes. Distinguishing driving processes for a volcanic swarm is often difficult using seismic data alone and yet seismic data may be the only monitoring stream available at many volcanoes.  Even when other monitoring data are available, seismic unrest often manifests itself first in the run-up process to an eruption.  Thus, tools that help distinguish whether a swarm is magmatic or not are desirable for observatories to improve forecasting efforts.

Determining when an eruption will follow a swarm is non-trivial, even if a swarm can be confidently linked to a magma intrusion.  Statistical comparison of an ongoing swarm to prior swarms at the volcano or at other volcanoes provides baseline probabilities for forecasting efforts and may reveal patterns in precursory activity in general. Swarms are often easy to visually identify in an earthquake catalog, but as no standard approach exists to consistently define and detect swarms across time and space, it is difficult to statistically compare them. At individual volcanoes, temporal changes in monitoring networks present challenges, while other factors, such as varying rates of background seismicity, complicate inter-volcano swarm comparisons.

To address some of these challenges, we have created a catalog of earthquake swarms covering 62 eruptions at 79 active volcanoes in the United States. The catalog balances consistency in methodology with the inherent variations between and among the volcanoes and their monitoring networks and is calibrated such that only concerning swarms, significantly above background levels, are retained.  We compute a suite of statistical attributes for each swarm and compare these attributes among various subgroups of swarms and volcanoes. Overall, we find that ~25% of the swarms in the catalog are associated with eruptions but only ~10% began prior to the eruption. In addition, though ~35% of eruptions are associated with swarms, only ~20% of eruptions have swarms that began prior to the eruption.  The eruptive and non-eruptive swarms show significant differences in evolution of moment release and event rate, but these differences vary depending on the types of volcanoes and eruptions considered. When tailored to the specifics of an ongoing swarm, analog swarm comparisons may help decipher driving process and likelihood of eruption.