Characteristics of Foreshocks Revealed by an Earthquake Forecasting Method Based on Precursory Swarm Activity

We have developed an empirical earthquake forecast method, Maeda's method, based on the statistical features of precursory seismic swarm activity, that is foreshocks, which sometimes appear before a mainshock, and issuing an alert of a mainshock occurrence within a certain period of time. In this study, we investigated the effectiveness of earthquake forecast of Maeda's method by applying it to seismicity under various tectonic environments of Japan such as regions characterized by interplate seismic activity, a tectonic fault line (concentrated deformation zone), and an island arc area of seismic and volcanic activity. As a result, we confirmed that Maeda's method yielded generally higher scores than a forecast model based on a stationary space‐time epidemic‐type aftershock sequence (ETAS) model. We also found that foreshocks detected along the Japan Trench were distributed along the edges of low‐velocity anomalies and among areas with background swarms related to slow slip events (SSEs). The foreshocks may have been caused by a heterogeneous stress distribution associated with the existence of a plate‐bending axis and a subducted seamount. Foreshocks off Iwate prefecture, in particular, were excited by periodic SSEs. In an inland tectonic zone and an island arc, swarm activity associated with magmatic or fluid activity related to low‐velocity anomalies tended to be followed by a mainshock. Maeda's method is a simple and efficient counting‐number‐based earthquake forecast model and may capture characteristics of foreshocks that reflect a physical phenomenon, such as a nucleation process involving precursory slip, which the stationary ETAS model is not able to represent. Plain Language Summary We have developed an empirical earthquake forecast method, Maeda's method, based on the statistical features of precursory seismic swarm activity, that is foreshocks. In this study, we investigated the effectiveness of the earthquake forecast of Maeda's method by applying it to seismicity under various tectonic environments of Japan. As a result, we confirmed that Maeda's method yielded generally higher scores than a forecast model based on a stationary space‐time epidemic‐type aftershock sequence model. We also found that foreshocks detected along the Japan Trench were distributed along the edges of low‐velocity anomalies and among areas with background swarms related to slow slip events (SSEs). In addition, we found that foreshocks off Iwate prefecture in particula