Résumé
The ATLAS research project, led by the French agency for risk mitigation (DGPR), aims to update the probabilistic seismic hazard assessment (PSHA) for the French territories in the Lesser Antilles (Guadeloupe, Martinique, Saint Martin, Saint Barthélemy) by 2027. Within this framework, we evaluate Ground-Motion Models (GMMs) and their associated uncertainties for crustal, subduction interface, and intraslab earthquakes. As a first step, a regional strong-motion database has been compiled, as no comprehensive dataset previously existed. It includes 393 earthquakes with magnitudes (Ml or Md) ≥ 4.0 recorded between 2003 and 2021, corresponding to > 6424 recordings from 90 seismic stations across the French Lesser Antilles. The dataset was enriched with the metadata required for GMM evaluation, including moment magnitudes (Mw), focal mechanisms, source-to-site distances, Vs30 estimates, and earthquake source-type classification. Several parameters rely on simplifying assumptions, contributing to uncertainties. In addition, the limited distribution of seismic stations across the islands results in reduced azimuthal coverage, leading to uncertainties in earthquake locations, particularly in focal depth. We evaluate in this study existing GMMs because limited flatfile data does not yet allow to define a reliable model specific for the French Lesser Antilles. The performance of several models for each tectonic setting (crustal, subduction interface, and intraslab) is analyzed against the observations using mixed-effects regressions and common performance metrics. The set of GMMs tested was chosen due to their recent application in PSHA studies, prioritizing global NGA models and including a few regional models relevant to similar tectonic contexts. Sensitivity analyses are performed to quantify how uncertainties in metadata (e.g., Mw, focal depth, Vs30) and alternative choices of recordings for comparison affect GMM performance evaluation. These results provide the basis for defining a set of candidate GMMs and associated epistemic uncertainties to be considered in future PSHA calculations for the French Lesser Antilles.