A detailed analysis of the eruption behaviour of the island volcanoes Stromboli, Mount Etna (Italy), Yasur (Vanuatu), and Whakaari (New Zealand) was published by Darius Fenner from Nishtha Srivastava's team together with Patrick Laumann and Georg Rümpker from the Frankfurt Institute of Advanced Studies (FIAS). Such analyses of past major and minor events can help to understand volcanic eruptions, including the underlying physical and chemical processes.

The group employs a recently developed approach to detect seismo-volcanic events. It catalogues all small and large seismo-volcanic events including major eruptions continuously at stations near volcanoes. The automated and very powerful method "Adaptive-Window Volcanic Event Selection Analysis Module" (AWESAM), presented by the FIAS Seismology & Artificial Intelligence (SAI) working group last year (News / FIAS), allows volcanic events to be detected quickly.

In their current research, the group extended their research by analysing the available data from up to fifteen years in detail, for example the time intervals between events, the amplitudes and the relationship between amplitude and frequency. This enabled them to identify differences and common patterns in volcanic events.

For example, they observed that there are more large eruptions on Stromboli than one might expect based on known distributions. Building on their previous findings, this study assesses and extends the understanding of this phenomenon based on a decade of data. For example, a certain pattern was found before and after the two heavy Stromboli eruptions in 2019, and the expanded dataset confirms the statistical significance of the results. However, so far this pattern has only been observed for Stromboli, which raises questions about its uniqueness.

Furthermore, the study classifies event types for Stromboli using an unsupervised machine learning approach. It reveals alternating patterns before and after paroxysms. The group was able to subdivide these patterns in more detail for the first time. Based on a clustering algorithm, they classified the frequencies of the events more precisely, for example. These patterns can be important for predicting large eruptions.

Using an identical approach for all four volcanoes, the group found similar behaviour despite different types and activities. At Whakaari, a certain pattern of repetition of large events was evident. As this observation is based on data from a single station, further in-depth investigations are needed as more data become available. "In a next step, we want to investigate if there are signatures prior to major eruptions" says first author Darius Fenner. "Our method offers a promising basis for more accurate volcano monitoring and for a better understanding of the underlying processes."

Publication: Darius Fenner, Georg Rümpker, Patrick Laumann, and Nishtha Srivastava, Amplitude and inter-event time statistics for the island volcanoes Stromboli, Mount Etna, Yasur, and Whakaari. Front. Earth Sci. 11:1228103.

doi: 10.3389/feart.2023.1228103, https://www.frontiersin.org/articles/10.3389/feart.2023.1228103/full