Evidence for subduction beneath Gibraltar Arc and Andean regions from k-means earthquake centroids

Abstract

k-Means algorithms are widely used for determining clusters in broad types of datasets. Since zones of high seismic activity as plate boundary present diffuse seismicity patterns, the use of k-algorithm is a way to classify earthquakes in terms of centroids. Mapping centroids improves seismic visibility for further tectonic interpretation. We used selected datasets of earthquakes and determined the number of clusters or values of k by introducing the silhouette index method to check the validity of cluster numbers. By introducing magnitude size in the vectorial attributes, k-means algorithm provides a map of centroids that represents the location of high seismic energy, which is useful in seismic risk assessment. By including the depth of seismic events as the main attribute, we obtained spatiotemporal variations of centroids, which improve the image resolution of seismicity at depth to find out the underlying dynamic process. This has been achieved in subduction zone of Chile where the presence of slab is reflected by centroid distribution. The method is particularly relevant to complex seismic zones where controversial geodynamic models are reported such as the Gibraltar Arc. Resulting model supports W-oriented subduction underlying many parts of the Gibraltar zone.