AUTHOR=Mitchinson Sam , Johnson Jessica H. , Milner Ben , Lines Jason 

TITLE=Identifying earthquake swarms at Mt. Ruapehu, New Zealand: a machine learning approach

JOURNAL=Frontiers in Earth Science

VOLUME=Volume 12 - 2024

YEAR=2024

URL=https://www.frontiersin.org/journals/earth-science/articles/10.3389/feart.2024.1343874

DOI=10.3389/feart.2024.1343874

ISSN=2296-6463

ABSTRACT=Mt. Ruapehu is an active andesitic stratovolcano, consisting of several peaks with the summit plateau at 2,797 m, making it the tallest active volcano in New Zealand. The extent of the volcano spreads 40 km across with a series of complex faults encompassing almost the entire base of the volcano. A series of earthquakes occurring 20 km west of the summit of Mt. Ruapehu, near the small town of Erua, which preceded the 1995/1996 major volcanic eruption sequence has been proposed as a medium-term precursor for eruptions at Mt. Ruapehu. We use unsupervised machine learning clustering algorithms HDBSCAN and DBSCAN to define anomalous earthquake swarms in the region and determine whether the Erua swarm was unique by identifying key characteristics in space, time and magnitude distribution. HDBSCAN found six spatial cluster zones to the west of Mt. Ruapehu, which have temporal seismic bursts of activity between 1994-2023. DBSCAN identified the seismic anomaly that preceded the 1995/1996 major eruption, along with one other similar event in the same region, which did not coincide with any documented volcanic activity. We found that the temporal evolution of the earthquake clusters west of Mt. Ruapehu share similar characteristics to variable inter-event bursts, related to variability in fluid migration typical of fault-valve models. Therefore, suggesting that the occurrence of a distal seismic swarm cannot be used as a reliable eruption precursor at Mt. Ruapehu when considered in an unbiased way with no a priori knowledge.