A data driven approach to mineral chemistry unveils magmatic processes associated with long-lasting, low-intensity volcanic activity.

The most frequent volcanic eruptions are of low-intensity and small magnitude. They produce abundant ash-sized (< 2 mm) clasts, which are too small to establish quantitative links between magmatic processes and eruptive dynamics using classic approaches. This inhibits our ability to study the past behaviour of frequently erupting volcanoes, essential to predict their future activity and mitigate their impact. The Palizzi unit (10-13th century, Vulcano, Italy) includes a prototype sequence of ash deposits resulting from prolonged Vulcanian eruptions punctuated by those of two larger sub-Plinian events. We apply Hierarchical Clustering to chemical analyses of clinopyroxene collected along the stratigraphy to decipher magma dynamics during this eruptive period. We identify periods of magma input and we link deep magmatic processes to eruptive dynamics, also showing that our approach can be used to connect magma and eruptive dynamics in any volcanic sequence. This is essential to track the processes occurring during frequent eruptions and to identify the build-up to larger explosive events.