Texturally constrained machine learning thermobarometry and chemometry of the Cumbre Vieja 2021 eruption, La Palma

<p>Magma has a dynamic and often-complex journey from source to surface, the record of which is largely encoded in the chemistry of minerals. Its storage conditions prior to eruption and modifications during ascent can influence eruptive dynamics and eruption duration. We present quantitative 2D chemical maps of clinopyroxene crystals from the Cumbre Vieja eruption (La Palma, Canary Islands; 19th September 2021 – 13th December 2021). The histories of individual crystals are constrained using novel thermobarometric (pressure, temperature) and chemometric (equilibrium melt composition) machine learning algorithms. We identify the remobilisation of colder (~950 ˚C), deeper (2 – 3.5 kbar), and more evolved (1 – 2 wt% MgO) cores by a hotter (1050 – 1100 ˚C) and less-evolved (3.5 – 4.5 wt% MgO) carrier melt. Textural evidence shows resorption of these antecrystic cores suggesting an uninterrupted ascent through the crustal column followed by upper-crustal (~ 1kbar) crystallisation and eruption. By using both quantitative maps and reliable single-phase thermobarometric and chemometric calibrations, we overcome several issues associated with acquiring statistically representative mineral chemistry via single spot analyses. In doing so we precisely track the syn-eruptive evolution of storage pressure-temperature and magma composition. These parameters are then related to the variation of geophysical signals (seismicity, gas monitoring) recorded during the La Palma eruption.</p>