Le Gall, Nolwenn

New textural and petrological data are presented on products from five paroxysms at Stromboli (Aeolian Islands, Italy) including the two from 2019 and three historical (1930, undated, sixteenth century) eruptions. The data are used to con- strain timescales associated with the initiation of paroxysms and to examine current models for their triggering. Samples were collected from the deposits and a subset selected for mineral separation and petrological and textural characterization. Minerals and glass were imaged by scanning electron microscopy (SEM), and chemical composition and zonation were analysed by electron microprobe. Trace elements in olivine were also determined. Vesicle number densities, vesicularities and vesicle diameters were measured by X-ray microCT techniques. The data were systematically compared with results of experiments simulating, on the one hand, ascent, vesiculation, degassing and crystallization of LP (low-porphyricity) magma and, on the other hand, interaction between LP and HP (high-porphyricity) magma. Paroxysm samples are mixed and include portions representative of both LP and HP magma. They host in variable proportions minerals and glass textur- ally and compositionally typical of these two magma types. Small but systematic variations in matrix glass compositions are found between each of the five eruptions considered. All samples host a population of vesicles ranging from < 15 to > 1000 μm in diameter and whose size distributions follow mixed exponential to power law distributions. Vesicularities are high (75% on average) and vesicle number densities range from 102-103 to 103-104 mm-3. Using experimental calibrations, the vesicle textural data suggest average LP magma ascent rates of 1–2 m/s (i.e. ~1.5 hours from depths between 7 and 1.5 km). The correlation between ascent rate and textures demonstrates systematic variations between eruptions, the most ener- getic (i.e. that of 1930) being associated with the highest ascent rate (~2 m/s). Widths of plagioclase reaction zones indicate that LP and HP magmas interacted for a maximum a few hours before eruption. Olivine reaction also implies durations of a few hours for LP-HP interaction and is followed by crystallization for 20 hours in the HP magma. Our results stress the fast ascent of LP magma from their storage region and their short residence times at shallow levels before being erupted. They clarify the respective roles of the deep and shallow feeding systems. An integrated phenomenological model for paroxysm initiation at Stromboli is outlined. Keywords

The majority of basaltic magmas stall in the Earth's crust as a result of the rheological evolution caused by crystallization during transport. However, the relationships between crystallinity, rheology and eruptibility remain uncertain because it is difficult to observe dynamic magma crystallization in real time. Here, we present in-situ 4D data for crystal growth kinetics and the textural evolution of pyroxene during crystallization of trachybasaltic magmas in high-temperature experiments under water-saturated conditions at crustal pressures. We observe dendritic growth of pyroxene on initially euhedral cores, and a surprisingly rapid increase in crystal fraction and aspect ratio at undercooling ≥30 °C. Rapid dendritic crystallization favours a rheological transition from Newtonian to non-Newtonian behaviour within minutes. We use a numerical model to quantify the impact of rapid dendritic crystallization on basaltic dike propagation, and demonstrate its dramatic effect on magma mobility and eruptibility. Our results provide insights into the processes that control whether intrusions lead to eruption or not.

To testmechanisms of basalticmagma degassing, continuous decompressions of volatile-bearing (2.7–3.8 wt% H2O, 600–1,300 ppm CO2) Stromboli melts were performed from 250–200 to 50–25 MPa at 1,180–1,140 C.Ascent rates were varied from 0.25 to *1.5 m/s. Glasses after decompression show a wide range of textures, from totally bubblefree to bubble-rich, the latter with bubble number densities from 104 to 106 cm-3, similar to Stromboli pumices. Vesicularities range from 0 to *20 vol%. Final melt H2O concentrations are homogeneous and always close to solubilities. In contrast, the rate of vesiculation controls the finalmelt CO2 concentration. High vesicularity charges have glass CO2 concentrations that follow theoretical equilibrium degassing paths, whereas glasses from low vesicularity charges show marked deviations from equilibrium, with CO2 concentrations up to one order of magnitude higher than solubilities. FTIR profiles and maps reveal glass CO2 concentration gradients near the gas–melt interface. Our results stress the importance of bubble nucleation and growth, and of volatile diffusivities, for basaltic melt degassing. Two characteristic distances, the gas interface distance (distance either between bubbles or to gas–melt interfaces) and the volatile diffusion distance, control the degassing process. Melts containing numerous and large bubbles have gas interface distances shorter than volatile diffusion distances, and degassing proceeds by equilibrium partitioning of CO2 and H2O between melt and gas bubbles. For melts where either bubble nucleation is inhibited or bubble growth is limited, gas interface distances are longer than volatile diffusion distances. Degassing proceeds by diffusive volatile transfer at the gas– melt interface and is kinetically limited by the diffusivities of volatiles in the melt. Our experiments show that CO2-oversaturated melts can be generated as a result of magma decompression. They provide a new explanation for the occurrence of CO2-rich natural basaltic glasses and open new perspectives for understanding explosive basaltic volcanism