It is still debated whether earthquake occurrence can be described as a single process or whether foreshocks are different phenomena. If foreshocks behaved differently, this would suggest a change of physical processes in the mainshock preparation phase that would boost hopes of forecasting large earthquakes. Most research on foreshocks focuses on case studies or uses global datasets in which recordings of small earthquakes are incomplete and are thus neglected. We do comprehensive foreshock statistics on all mainshocks in a regional earthquake catalog that is complete above M 2.5. To detect possible differences between foreshocks and seismicity that follows a uniform triggering model (the epidemic‐type aftershock sequence [ETAS] model), we perform a null‐hypothesis test. We also estimate the size of the differences between observed and ETAS‐simulated foreshocks.
We define different sets of foreshocks using two different methods, because there is no unique definition: a nearest‐neighbor declustering technique (Zaliapin et al., 2008) and a variety of space–time windows (e.g., Agnew and Jones, 1991). We use data from southern California, northern California, and Italy. For each region, we first search an appropriate null model: an ETAS model that describes aftershock numbers well. In southern California, we find an appropriate spatiotemporal model that is characterized by a large productivity parameter . After performing a null‐hypothesis test for different mainshock and foreshock magnitudes, we find foreshock signals () for all mainshocks sizes and independent of the foreshock’s lower magnitude threshold. Observed mainshocks have more foreshocks than the ETAS model predicts.