Abstract
A statistical study of the earthquake catalogue compiled by the Japan Meteorological Agency during 1961-1979 suggests empirical rules on earthquake prediction in Japan. In each earthquake sequence, the magnitude of the largest event which has already occurred by time t is defined as M1. Similarly, those of the largest fore- and aftershocks with respect to M1 are defined as F1 and A1, respectively. The proposed rules are: (1) When seismic activity occurs with M1-F1≤0.4 within about a week, it may be a foreshock sequence followed by a larger event. The probability is 25-30% for M1-F1≤0.2, and 20% for 0.3≤M1-F1≤0.4. (2) A similar rule is also applicable for M1-A1≤0.2. The probability is 20%. (3) If 0.5≤M1-F1 and 0.3≤M1-A1, the probability is, roughly speaking, 5-10% or less. In these cases, M1 is expected with high probability to be a main shock. (4) The magnitude of the expected event is larger than M1 by about 0.5 on the average. (5) About 40-45% and 80% of the expected events occur within a day and a week, respectively, after criterion (1) or (2) is satisfied. Time intervals of F1-M1 and M1-A1 may also help to estimate the time of the expected event. (6) If no larger events actually occur within a day or a week after criterion (1) or (2) is satisfied, the probability that a larger event is still expected drops to 3/5 or 1/5 of the initial value, respectively.
