Many applications in seismology require to isolate earthquake clusters from a background activity. Relative declustering methods essentially find a 2D representation of an earthquake catalogue that distinguishes between two classes of events: crisis and non-crisis events. However, the number of statistical and/or physical parameters to be used is often limited due to the difficulty of concatenating the information onto a physically meaningful 2D grid. In this study, we propose to alleviate the declustering task by using the ability of unsupervised artificial intelligence to model complex spatio-temporal relationships directly from data. Through a data-driven approach, we define an easily transferable declustering model that provides declustering results with fewer assumptions and no prior selection of thresholds. We first obtain this model by training a self-organising neural network (SOM) that learns to cluster data points according to their feature similarity on a 2D map. We then assign each SOM cluster a label (crisis or non-crisis class) using an agglomerative clustering procedure. We quantify the classification uncertainty by developing a probabilistic function based on the projection learned by SOM. Our method is applied to a synthetic dataset and to real catalogues from the Gulf of Corinth, Central Italy and Taiwan. We discuss the validity of the method by estimating its classification accuracy. For real data, we qualitatively compare our results to previous declustering attempts. We show that our approach is easy to handle, provides a fairly new representation of earthquake catalogues and has the potential to reduce classification ambiguities between nearby events.

Four transform faults and three intra-transform segments located at Equatorial Atlantic form the Saint Paul Transform System (SPTS), with a long-offset of 630 km. In the northern transform, the 200 km long and 30 km wide Atoba Ridge is a major topographic feature that reaches the sea level at the St. Peter and St. Paul Archipelago island (SPSPA, 00º 55 ‘0’ ‘N and 29º 20 ”43”W). The islets have an average uplift rate of approximately 1.5 mm/year. The southern and northern flanks of the Atobá Ridge are marked by a series of large thrust faults visible in the bathymetry and clearly imaged through seismics and correspond to an exceptionally serpentinized mantle. We have determined the hypocentral location of 62 minor-moderate earthquakes of SPTS, with magnitudes 1.9 ≥ M ≤ 5.3. The earthquakes occurred in 2013 and were recorded by a seismometer installed in SPSPA and three autonomous hydrophones deployed during the COLMEIA cruise. The HYPOCENTER software and Seismic Analysis Code (SAC) were used for data analysis and hypocenter location. The depth range is from 0.2 to 17.5 km and are concentrated in three different zones: the East Shear Zone (ESZ), the Atobá Ridge Zone (ARZ) and the Central Fracture Zone (CFZ). A seismogenic zone with a deep britle-ductile transition was identified in SPTS, with hypocenters reaching 18 km beneath the seafloor. We observed that this lithospheric structure presents relation with the offset age and controls the maximum hypocentral depths of oceanic transform faults. Besides, the earthquakes indicated the existence of a broad serpentinization depth reaching 18 km beneath the ARZ. This was interpreted as the effect of deep water percolation into the mantle in the SPTS, which caused a fluid-mantelic rocks interaction and allowed the expansion of faults into the mantle. Some hypocenters were located in the central fracture zone (CFZ) segment of SPTS and their depths reached 8.8 km beneath the seafloor. We interpreted this seismicity as reactivation of a weakness zone existent in CFZ due to the transpressive load-induced stress originated in Atobá Ridge.