Assessing the rock density distribution of the La Soufrière de Guadeloupe volcano lava dome with a 4-panel scintillator-based muon detector

The intense hydrothermal activity occurring at La Soufrière de Guadeloupe, French Lesser Antilles, is a major concern for volcanologists at the Volcanological and Seismological Observatory of Guadeloupe and for the local population living nearby. Partly hosted within the andesitic lava dome, this hydrothermal system, continuously fueled with meteoric waters, is indeed responsible for both violent phreatic eruptions (the last major event occurred in 1976-1977), and fast rock alteration due to hot and acid fluid circulation that worsens the risk of a partial volcano flank collapse. Since 2015, the deployment of particle trackers built in IP2I, Lyon, to perform dynamic muon imaging of the lava dome structure has allowed to increase the knowledge of the hydrothermal system dynamics. Recently a new generation of hodoscopes equipped with 4 plastic scintillator matrices have been installed around the volcano to scan different areas of the lava dome. Here we will present a new versatile method to process the data acquired by these detectors, and estimate the corresponding density radiographies. Particle trajectories are reconstructed by performing a fit of the recorded hits in the impacted scintillator bars using a Random Sampling Consensus procedure (RANSAC). This algorithm is specifically built for discriminating outlier points ("noisy" hits) in the data. Thus, it allows to significantly improve the muon signal/noise separation, and obtain higher quality estimates of the particle trajectories. To measure the performances of this RANSAC-based processing algorithm, and its ability to reconstruct muon tracks, we have developed a dedicated GEANT4 simulation of our 4-panel detector, interfaced with the MC generator CRY to simulate the incident flux of each component of the atmospheric showers (muonic, hadronic and electromagnetic).In addition to the simulation outputs, we will notably show preliminary results obtained from six months of real data (∼107 reconstructed events), recorded in a telescope, equipped with four 16x16 scintillator bars matrices, located at the volcano south-west. This configuration allows to increase the precision on the 2D density radiography of the central part of the scanned region, and offers two additional 3-panel sub-configurations, which leads to three different radiographies.The 2D density radiographies will then be combined to serve as input for numerical modelling, conjointly with other geophysics data sets (e.g from gravimetry surveys), in order to obtain a 3D bulk density distribution model of the lava dome. An unprecedented resolution is expected thanks to the uniqueness of the different data sets available and the quality of the event reconstruction. This will lead to a better characterization of the hydrothermal fluid circulation impact on rock alteration and the resulting partial flank collapse hazard.