A statistical method of detecting unusual seismic activities has been developed. Earthquake occurrences are modeled by a point process, and the number of shocks that occurred in a unit time is assumed to be distributed in accordance with Poisson's law when the seismic activity is normal. The method consists of a test of rejecting outlying observations in the past seismic activity, a goodness-of-fit test and a variance test for Poisson's distribution, and a prediction of future seismic activity based on the normal activity in the past.
The number of microearthquakes observed at the Shiraki Microearthquake Observatory are divided into a number of groups according to the S-P interval. The method advanced in this paper is then applied to them. As a result, it is suggested that there was unusual seismic activity 20 days before the commencement of the Kutsugahara Earthquake Swarm of 1970.