Evaluación y comparación de algoritmos para la detección automática de eventos sísmicos

Abstract

Early seismic events detection might reduce the number of people affected and save lives. In particular, the seismic activity in Ecuador is high, given its location along the zone named the Pacific Belt of Fire. In this context, this paper presents a solution to compare algorithms for detecting seismic events. This comparison was performed both concerning the functionality and the configuration of the parameters required for each algorithm. This solution was implemented on an SBC platform (Single Board Computer) to obtain a portable, scalable, economical, and low-cost computational tool. The methods compared were: Classic STA/LTA, Recursive STA/LTA, Delayed STA/LTA, Z Detector, Baer and Kradolfer picker, and AR-AIC (Autoregressive-Akaike-Information-Criterion-picker). For the evaluation and comparison, multiple experiments were carried out using real data provided by the Regional Seismological Network (RSA). In particular, such registers were used as input data to the seismic algorithms. Results revealed that the algorithm with the best performance was the Classic STA/LTA, since from the total number of real events (58), only one was not detected. In addition, 6 false negatives were obtained, achieving 98,2% of precision. Finally, the software used for the comparison of the algorithms has been released for free usage, which represents another contribution of this work in the context of seismic analysis.