22 years of satellite imagery reveal a major destabilization structure at Piton de la Fournaise

Flank destabilization in volcanoes is a significant hazard, capable of triggering tsunamis and large earthquakes, accounting for a substantial portion of volcanic fatalities. This phenomenon is frequently observed on oceanic islands, evidenced by bathymetric studies revealing widespread debris avalanche deposits. Geological and geophysical observations, along with physical and analog models, suggest that magmatic activity is a key trigger for flank destabilization. Magma often follows preferential intrusion paths, known as rift zones, controlled by the edifice's stress field and structural weaknesses. Two main mechanisms for magma-induced flank destabilization have been proposed: the first involves subvertical dyke intrusions coupled with low-angle faulting, as seen at Kilauea, Mount Etna, and Cumbre Vieja; the second involves subhorizontal sill intrusions undergoing coeval opening and slip, as suggested for Piton des Neiges. Piton de la Fournaise volcano on Réunion Island presents a significant hazard due to potential eastern flank collapse, with evidence of ongoing flank slip. This study leverages the volcano's high activity and extensive monitoring data (InSAR and GNSS since 1998) to create a 3D image of the shallow plumbing system and the structures accommodating flank slip, aiming to understand the link between magmatic activity and potential catastrophic flank collapse.

Previous research has established two primary mechanisms for magma-induced flank destabilization. The first involves subvertical dyke intrusions within rift zones interacting with low-angle faults, leading to flank movement. This mechanism has been observed in volcanoes like Kilauea, Mount Etna, and Cumbre Vieja, where geodetic data reveal intrusions pushing the flanks seaward, occasionally resulting in flank failure. The second mechanism involves subhorizontal sill intrusions that undergo simultaneous opening and slip, as observed at Piton des Neiges. Studies at Piton des Neiges identified a detachment-type fault intruded by sills, suggesting that sill injections induce rapid co-intrusive slip. At Piton de la Fournaise, the eastern flank's potential for collapse is a major concern, supported by bathymetric evidence of large debris avalanche deposits and observed ongoing flank slip, with significant eastward movement and upward motion during the 2007 eruption. InSAR and GNSS data have revealed continuous slip, accelerating with magmatic activity, and the presence of a fault under the eastern flank has been proposed to explain observed deformation.

This study utilizes a comprehensive dataset of InSAR and GNSS measurements collected between March 1998 and December 2020 to image the shallow plumbing system and associated flank slip structures at Piton de la Fournaise. A total of 57 magmatic intrusions, including both eruptions and failed eruptions, were analyzed, with 12 events discarded due to insufficient deformation. The analysis focused on 57 events with detectable deformation (>2cm). Inverse modeling was employed using a 3D Mixed Boundary Element Method to determine the geometry of each intrusion. This method allows for curved intrusions, a key feature observed at Piton de la Fournaise, and accounts for stress changes (overpressure and shear stress) on fractures, unlike most previous studies that focused on displacement. The models utilized triangular elements to represent intrusion geometry, enabling the modeling of complex shapes. The inversion process involved a search stage to explore parameter space and an appraisal stage using Bayesian inference to compute mean models, marginal probability density functions (PDFs), and confidence intervals. The statistical representation of models employed involved randomly generating meshes within the 95% confidence interval of the best-fit model, enabling the analysis of spatial distribution, dip, and strike of intrusion elements. InSAR data were primarily from the Indian Ocean InSAR Observatory (OI²), supplemented by GNSS data from OVPF-IPGP for events lacking InSAR coverage.

Inverse modeling revealed six preferential intrusion zones radiating from the central cone, five coinciding with previously identified rift zones. A sixth zone, a previously unidentified sill intrusion zone at depth, was discovered. The most significant intrusion zone (NE-SE, 45% of total volume) forms a spoon-shaped structure with horizontal and vertical curvatures, connecting a subvertical dyke system at the surface to a subhorizontal sill system at depth. The second largest intrusion zone (35% of total volume) consists of low eastward-dipping sills beneath the eastern flank, connecting at depth to the main NE-SE zone, extending the spoon-shaped structure eastwards. The NE-SE and sill zones form a continuous structure accounting for 80% of the total intruded volume. Slip within the spoon-shaped structure increases eastward, showing a continuum from pure opening (westernmost sills) to pure slip (easternmost intrusion behaving like a fault). The transition is consistent with field observations of sheared intrusions accommodating flank slip. The base of the spoon-shaped structure aligns with the pre-eruptive seismic swarm, indicating a critically stressed volume underneath. Eastern flank seismicity, however, is not directly related to slip on the main structure but might be triggered by stress transfer from magma intrusions. Four secondary intrusion zones were identified, possibly induced by the main structure, rather than inherited crustal fractures. The overall spoon-shaped geometry suggests a rotational landslide-like structure, but one accommodating magma, with a continuum of deformation from pure opening to coeval opening and slip, and finally to pure slip eastward. The easternmost part of the structure appears to be locked, accumulating stress that is sporadically released, potentially triggering earthquakes and tsunamis. The estimated volume of material above the potential failure surface is 45 km³.

The findings reveal a hybrid flank slip mechanism, combining aspects of models proposed for shield volcanoes: dyke intrusions coupled with low-angle fault slip, and sheared sills intruding faults. The continuous deformation within the spoon-shaped structure, ranging from pure opening to pure slip, highlights the complex interplay between magma intrusion and pre-existing structures. The spoon-shaped geometry and the observed eastward increase in slip suggest a potential for catastrophic flank collapse. The locked easternmost part of the structure indicates a potential for the build-up and release of significant stress, which can manifest as slow slip events or sudden large earthquakes and tsunamis. The study's results are consistent with observations at Piton des Neiges and other volcanoes worldwide, where sheared intrusions and low-angle faults play a role in flank instability. The revealed mechanism emphasizes the significance of comprehensive long-term studies of intrusive activity for hazard assessment.

This study provides a detailed 3D model of the shallow plumbing system at Piton de la Fournaise, revealing a major spoon-shaped structure that accommodates flank slip through a continuum of deformation. This structure represents a hybrid mechanism of flank destabilization, combining elements of dyke-induced faulting and sill-induced slip. The findings highlight the potential for catastrophic flank collapse due to the locked eastern section of the structure. Future research could explore the long-term evolution of this structure, focusing on the precise role of hydrothermal alteration and creep in accelerating flank movement, and improve the understanding of stress transfer mechanisms between different parts of the edifice.

The study's interpretations are based on the assumption of a linearly elastic, homogeneous, and isotropic medium for the inverse modeling. While the model effectively captures the observed deformation, heterogeneities in the volcano's structure could affect the accuracy of the determined intrusion parameters. The reliance on InSAR and GNSS data could limit the detection of small-scale intrusions or those occurring at depths beyond the sensitivity of these techniques. The 95% confidence intervals for the model parameters, while providing uncertainty estimates, do not completely encapsulate the full range of possible intrusion geometries.