Considering spatial memory to estimate seismic risk: the case of the Azores Archipelago

Abstract

This study aims to include the “epicentre migration” effect over time in the seismic occurrence process and subsequently in the assessment of seismic hazard and risk within a region. This epicentre migration phenomenon reflects the transfer of stress among different seismic zones. This study statistically quantifies how epicentres migrate among seismic zones, and a statistical measure was proposed to evaluate this influence. Traditionally, the location of the next earthquake is considered independent of the locations of previous events. However, seismic records reveal the existence of “spatial memory”. Our objective is to statistically quantify this memory. We used data from the Azores region to investigate this phenomenon because seismic zones have already been defined for this region. These zones are the “units” to and from which successive events migrate. Markov chains were used to model the sequence of seismic events over time. A one-step transition of the spatial variable (the seismic zone of an earthquake) was studied, revealing an evident spatial dependence between consecutive occurrences. For the case of a stationary Markovian process, two-step, …, N-step transitions were also considered, but a clear attenuation of spatial memory over time was observed as we moved to higher-order step transitions. This reveals that the methodology presented herein is especially useful in the assessment of seismic risk in the short or medium term. Only epicentral migration was analysed in this study, irrespective of the magnitudes and times between events. Time and magnitude variables will be incorporated into future studies given that the time, location and magnitude of an event are not independent and need to be studied together.