In 2021, a period of volcanic unrest began on the Reykjanes Peninsula (Iceland). To date, eight eruptions have taken place with durations ranging from one day to half a year. The cumulative eruption volume is about 400 x 106 m3 and covers an area of about 46 km2. The eruptions can be divided into the Fagradalsfjall eruptions, directly fed by a deep-seated reservoir, and the Sundhnúkur eruptions, where the uprising magma is temporarily stored in a sill.
Magma travels in dykes, and breaks its way through the rock as long as the fluid pressure exceeds the host-rock strength. The earthquakes reflect this fracturing. Together with the inflation of the sill, earthquake swarms serve as predictors for imminent eruptions in the study area.
All the eruptions have taken place in the very close vicinity of the fishing town of Grindavík. One eruption started in the area protected by dams. Under the premise that magma transport can be traced by the earthquakes, I am improving the accuracy with which the location of the eruption site can be predicted.
I apply a fully automated and thus unbiased algorithm. The algorithm detects the number of clusters and assigns earthquakes from the subsequent time windows, according their spatial distribution, to the clusters. It then analyses the geometric properties of the clusters, ranks them and finally determines a location that presents the most likely eruption site for each time window. Finally, I compare the results with real world data and determine the applicability of my approach.