Unsupervised learning exposes magma dynamics during the 2014-2015 Holuhraun-Bardarbunga eruption (Iceland)
Abstract
The Holuhraun-Bardarbunga eruption was intensely monitored, which makes it ideal to investigate relationships between observed geodetic and geophysical signals and the evolution of magma chemistry. We collected a series of samples of tephra and lava flow lobes emplaced between September 2014 and February 2015. Previous investigations focused on whole rock and melt inclusion geochemistry and did not evidence any significant temporal evolution of magma throughout the eruption. Here we collected trace element maps and performed major and trace element analyses along core-to-rim profiles in clinopyroxene crystals. The clinopyroxenes are zoned with chemistries that can be separated into 5 compositional clusters on the basis of hierarchical clustering. One of the clusters represents the chemistry of hourglass sectors, two groups are associated with chemical variations that can be reconciled with moderate to small variations of cooling rates, one cluster identifies crystallisation at high rates of undercooling; A final group characterising the clinopyroxene rims formed by co-crystallisation of clinopyroxene and plagioclase at shallow depth. We quantify the variance of crystal chemistry from core to rim and the relative fraction of transitions from one cluster to another that crystals experience during eruption. These variations reflect main changes in the magma dynamics within the plumbing system which should also be reflected by geophysical measurements. We find that during the eruption, the chemical variance of the clinopyroxene is proportional to the number of recorded earthquakes per day. We suggest a direct relationship between the number of earthquakes and the extent of interaction between magmas at moderately different temperature. This approach allows more accurate interpretation and correlation of geophysical signals with processes occurring within the volcano plumbing system.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2019
- Bibcode:
- 2019AGUFM.V11C..09C
- Keywords:
-
- 1099 General or miscellaneous;
- GEOCHEMISTRY;
- 1916 Data and information discovery;
- INFORMATICS;
- 1942 Machine learning;
- INFORMATICS;
- 3699 General or miscellaneous;
- MINERALOGY AND PETROLOGY