My PhD and Postdoc fieldwork mainly focused on the Mount Pelée volcano (Martinique, Lesser Antilles). The analysis and interpretation of eruptive deposits from this volcano (observed and sampled during four field campaigns in 2017, 2019, 2021 and 2022) allowed the reconstruction of its eruptive history for the past 25,000 years with the discovery of 6 new eruptions (Michaud-Dubuy, 2019; Carazzo et al., 2020; Michaud-Dubuy et al., 2023; Carazzo et al., 2025). Additional field data acquired in collaboration with colleagues from the LMU (Germany) also allowed exploring the fragmentation efficiency through rapid decompression experiments (Huebsch et al., 2023). 

As several eruptions seemed to have dispersed their products in peculiar directions (that did not match the usual wind conditions in Martinique), it was necessary to perform tephra dispersal simulations using 2-D models. Our results allow identifying peculiar atmospheric circulations associated to a modification of the subtropical jet-stream path, thus producing northerly winds over Martinique and spreading tephra towards the most populated areas of the island (Michaud-Dubuy et al., 2019).

Determining the source conditions leading to a stable plume or a collapsing column is a crucial step to better understand the dynamics of explosive eruptions. Using the 1-D model of volcanic column PPM (Paris Plume Model) allowed quantifying the impact of the grain-size distribution of pyroclastic fragments on plume stability. The results show that the drastic effect of gas entrapment during magma fragmentation in the volcanic conduit is reduced when considering open porosity, which thus contributes to stabilize the eruptive column and limit the formation of pyroclastic currents. We also predicted various grain-size distributions in pyroclastic currents depending on the Total Grain-Size Distribution at the volcanic vent (Michaud-Dubuy et al., 2018).

In the framework of my CNES postdoctoral grant, and after the 2021 eruption of Soufriere of St Vincent and Grenadines (see Monitoring of on-going eruptions), I adapted PPM to estimate the mass eruption rates (MER) from the evolution of the geometry of downwind plumes. The model is calibrated and tested using GOES-16 satellite images of the 2021 La Soufrière St Vincent eruption. We then show our ability to estimate source parameters in near real-time on Mt Etna eruptions by using a new tool on the HOTVOLC platform (Michaud-Dubuy & Gouhier, in press).

Field-based studies allowed to reconstruct the past eruptive history of the Mount Pelée volcano (Michaud-Dubuy et al., 2019; Michaud-Dubuy, 2019; Carazzo et al., 2020, Michaud-Dubuy et al., 2023, Carazzo et al., 2025) and to have a better insight of a potential future eruption. 

The study on the Bellefontaine eruption demonstrated that daily winds must be taken into account in a future volcanic hazard assessment in Martinique. We thus considered daily wind profiles between 1979 and 2021 to compute two new probability maps, based either on the maximum eruptive scenario, or on 16 eruptive scenarios including new field constraints (Michaud-Dubuy et al., 2021). These new maps are now integrated to the revised version of evacuation plans in case of an eruption in Martinique (ORSEC plan).

In the context of a multiparameter volcanic unrest at Mt Pelée since April 2019 (Fontaine et al., 2025), we also computed new probabilistic maps for tephra fallout in case of a phreatic eruption in Martinique (included in the ORSEC plan) and in Guadeloupe, including a Monte Carlo method (Michaud-Dubuy et al., 2025). Current work includes a similar methodology in Mayotte, for tephra fallout (Michaud-Dubuy et al., 2024) and for ballistic hazard assessment (ClerVolc postdoctoral fellowship granted in 2024).

On April 8, 2021, a few hours before the explosive phase of La Soufriere, St Vincent started, I used volcanic tephra dispersal simulations to inform the French volcanological and seismological observatories in Guadeloupe (OVSG-IPGP) and Martinique (OVSM-IPGP) about the possibility of receiving volcanic ash from the eruption. The results were also shared within the international volcanic crisis group constituted during the eruption, and in the PREST project report (pages 104-105, in french). Even if the probability of receiving ash in Martinique was low, the OVSM team built several ash collectors and asked to the population through social media to send their testimonies if they did receive ash. We then built a map showing the ash dispersion from the Soufriere of St Vincent eruption in Martinique (PREST project report, pages 106-107, in french). 

After the explosive phase started, I also used satellite images to estimate in real-time the eruptive source parameters of the eruption (such as the mass eruption rate) and communicate the results to the crisis group. These estimates allowed to rapidly classify the eruption as a VEI 4. In 2022, I received a two years postdoctoral research grant from CNES to keep on working on this topic, in collaboration with the LMV-OPGC in Clermont-Ferrand (see Theoretical modeling). The new version of our volcanic column model (PPM), combined with umbrella measurements on the HOTVOLC platform (OPGC), now allows estimating in near real-time the mass eruption rate of observed eruptions (Michaud-Dubuy & Gouhier, in press).

The predictions made with 1-D models of volcanic columns accounting for the presence of wind strongly depend on a wind entrainment coefficient, a parameter whose value varies greatly in the literature. New laboratory experiments on turbulent jets with reversing buoyancy rising in a crossflow allowed to better constrain this wind entrainment coefficient. Parameterizing the 1-D model PPM with this result, we propose a new transition diagram between stable plume and collapsing fountain regimes (Michaud-Dubuy et al., 2020).