Automatic seismic swarm analyzer system based on template matching algorithms and Master-Cluster relative location methods

Seismic swarms may have periods of intense activity with a high number of earthquakes per hour, with overlapping events and/or low signal-to-noise ratio seismic records. During these intervals, the manual characterization of the activity can become very complex to perform by seismic or volcanic observatories, resulting in inhomogeneous seismic catalogs. In order to tackle this problem, we have developed a set of automatic algorithms capable of detecting earthquakes, picking their P and S arrivals and locating the events with absolute and relative methodologies. Detections are performed over the filtered seismic energy while phase picking is based in the correlation of new events with a set of previous well-characterized templates. Absolute locations are computed using traditional algorithms as Hypoellipse and for relative locations we introduce a novel technique Master-Cluster, which is a hybrid between the double differences and the master event. The algorithms have been tested on real data of two series corresponding to two different tectonic regimes: the volcanic pre-eruptive swarm of El Hierro, Spain (2011) and the tectonic seismic series of Torreperogil, Spain (2012-2013). Both data sets are considerably different in terms of epicentral distances and distribution of the network varying from stations very close to the activity at El Hierro (5-20 km) to regional distances in the case of Torreperogil (10-180km). The templates were taken as a partial dataset of 3 600 (El Hierro) and 800 (Torreperogil) relocated earthquakes from the manual regional catalog. Based on these datasets, the algorithm was able to improve the number of events by a factor of 6 in El Hierro and 10 in Torreperogil, producing a seismic catalog between 3 and 4.5 times larger than the manually obtained one. An additional test was performed with the smaller earthquakes (local magnitude<1.5) which were not included in the set of templates, resulting not only in a good factor of success –larger than 65% of events were retrieved in both series– but also an enhancement in their automatic locations was observed with a more clustered seismicity than the previous catalog.