Applying the Nearest Neighbor Method to the Analysis of Volcanic Swarms Based on Data for Bárðarbunga and Fagradalsfjall Volcanoes

Abstract

This study presents an analysis of event clustering in volcanic seismicity, especially in volcanic swarms. We have analyzed the volcanic swarms occurring during the eruptions of Bárðarbunga (2014) and Fagradalsfjall (2021) volcanoes in Iceland, attempting to use the nearest neighbor method for this goal. The method enables identification of clusters with different scales of generalized distance; as an example, the method generally identifies two clusters of events in tectonic seismicity and is widely used in aftershock identification. We have detected considerable differences in the distributions of generalized distance to the nearest neighbor for volcanic seismicity compared with tectonic seismicity. To be more specific, we have discovered two types of unimodal distribution, with the one mostly occurring before eruptions and the other during eruptions. The former type is probably due to coalescence of two near modes in the distribution, and reflects an internal inhomogeneity in seismicity during such periods. However, the unimodal nature of the distributions makes it difficult to identify events viewed as interrelated (clustered) or independent (background events). It can be surmised based on the results obtained that, before an eruption, the percentage of background seismicity fluctuates around 70%, while being between 90 and 100% during the eruption. This can be considered as providing evidence of different sources of seismicity during different phases of an eruption.