Bottom-up vs top-down drivers of eruption style: Petro-geochemical constraints from the holocene explosive activity at La Soufrière de Guadeloupe

Signals of volcanic unrest have been used successfully to provide insights into the timing, magnitude, intensity, and style of future eruptions. However, in order to provide context for the subsequent activity, analysis of past eruptions is required. This provides useful information in order to understand processes of magma genesis, storage, evolution and ascent which lead to the onset of eruptions. Here, we examine basaltic-andesitic to andesitic deposits from La Soufrière de Guadeloupe Holocene eruptions, covering a range of explosive eruption styles, ages and magnitudes. Our work is timely given unrest at this system has increased over the last 25 years, with a potential eruption capable of directly impacting up to 80,000 people in Southern Basse-Terre and potentially thousands more indirectly on a regional scale. We report on the geochemistry of pre-eruptive magmas using detailed analyses of glass (melt inclusions and groundmass glass) from four Holocene explosive eruptions: 1657 Cal. CE (Vulcanian, VEI 2), 1530 Cal. CE (sub-Plinian, VEI 3), 1010 Cal. CE (Plinian, VEI 4), and 5680 Cal. BCE (Plinian, VEI 4). Major element concentrations vs SiO2 in whole rock (WR), groundmass glass (GM) and melt inclusions (MI) show a strong linear trend. MIs reveal a relatively homogenous melt composition from the first to the most recent eruptions, ranging from 63.6–78.7 wt% SiO2. Volatiles, including H2O (2.3–4.4 wt%), CO2 (35–866 ppm) and sulphur (30–202 ppm), are also consistent across the various eruptions. The major element and volatile compositional homogeneity across the eruptions indicates that composition and volatiles do not have a direct control on eruption explosivity at this system. Instead, we find differences in ascent rate, groundmass glass viscosity and microlite volume percentage, indicating that explosive eruptive style at La Soufrière is controlled by a combination of ascent rate and top-down controls affecting rock strength, stress distribution and the development of fluid overpressure. Rapid ascent in the absence of top-down controls (processes with a cause external to the magma but affecting the plumbing system) will result in explosive eruptions driven from the bottom-up (internal to magma dynamic response with varying pressure and temperature, e.g., 1010 Cal. CE in the case of very rapid ascent or 1657 Cal. CE in the case of rapid ascent). However, we also highlight the importance of top-down controls, such as conduit sealing which can promote the onset of explosive eruptions, even in the case of slow magma ascent (e.g., 5680 Cal. BCE). External effects (including ingress of water and rapid edifice unloading) can also favour explosive eruptions with flank collapses involved in some scenarios (e.g., 1530 Cal. CE). The multiple controls on explosive eruption style make this system more hazardous and complex to model and monitor. In order to improve early-warning system efficiency, forecast models, eruption scenario crisis response and long-term risk reduction planning, we stress that internal processes such as fracture and host-rock sealing (fluid pore pressure) as well as external processes such as water moving into the system and the mechanical stability of the edifice should be monitored and modelled closely.