We propose an empirical alarm criterion for the occurrence of mainshocks (M>5or M>6) in the Izu region on the basis of the clustering activities of earthquakes which appear immediately before the mainshochs. The criterion consists of three parts: 1) elimination of aftershocks, 2) selection of candidates for foreshocks on the basis of the grade of clustering which is measured by the number of earthquakes (N1) with magnitude larger than or equal to M in unit time (two days) and in unit space

more »

... titude)XD(longitude)), 3) setting up the alarm period with Ta days and the alarm segments with the unit spaces where candidates for foreshocks are observed. In this criterion N M, D and Ta are treated as parameters. After the investigation of the effects of the parameters on the alarm results, the value of each parameter was selected as Nf>10, M, D=3.0, D=0.2 and Ta=4.0 days. This set of values gives us the following practically useful alarm results: the alarm rates are 68% and 71% for mainshocks with magnitude M>5 and M>6, respectively; the occurrence rates of mainshocks in the alarmed time-space per Ta (=4 days) and unit space (0.2X0. 2 are 0.24 and 0.08, respectively; the probability gain, the ratio of the occurrence rate of mainshocks in alarmed time-space to that in all time-space concerned, for either magnitude range is as high as about 850. As for Ta, it is found that four days are enough to give the alarm. This means that the proposed foreshocks are not true foreshocks but merely seismic swarms if no mainshock follows during the four days after the first alarm is raised by the proposed foreshocks. All these results are obtained from the data of earthquakes almost homogeneously observed in the Izu region by J. M. A. through 1977 to 1991, with the range of depth <50km and magnitude >3.0. The numbers of mainshocks contained in this data with magnitude M>5 and M>6 are 22 and 7, respectively.