Key points are not available for this paper at this time.
Abstract Although not specifically conceived for tackling short-term aftershock incompleteness (STAI), earthquake detection methods such as template matching (TM) and machine learning (ML) can help mitigate the under-reporting of aftershocks after large earthquakes by detecting low-magnitude events hidden in seismic noise. So far, the ability of TM and ML to address STAI has not been evaluated against benchmark data sets reconstructed by independent methods. In this study, we use events reconstructed by RESTORE (REal catalogs STOchastic REplenishment), a Python toolbox specifically designed to tackle STAI, as a stochastic benchmark to assess the ability of TM and ML in recovering the bulk statistical properties of aftershocks missed during STAI period. Our results show overall good compatibility between the TM/ML detections and the RESTORE benchmark in the space–time–magnitude domain, though some discrepancies in detection rates and in the upper bounds of magnitudes are noted. This study also highlights the complementary use of stochastic and enhanced detection techniques. Stochastic algorithms like RESTORE can be implemented for immediate STAI mitigation in short-term forecasting and operational earthquake forecasting, whereas enhanced detection techniques can be used over longer time scales to precisely recover unrecorded events.
Angela Stallone (Wed,) studied this question.