Abstract
Major uncertainties exist in the variability of seismicity, and yet even its general probabilistic nature has not been discussed. In particular, the value of the coefficient of variation (c. o. v. ) of seismic intensity based upon Type II extreme distribution tends to give considerable overestimation. An empirical extreme value distribution with lower and upper bound limits proposed by one of authors was applied to prov le a better probabilistic understanding of the seismic risk. The upper bound limit was determined by applying this distribution together with Kanai's attenuation formula to the historical earthquake data of 1302 years, with some considerations given to the existing upper bound earthquake magnitude map prepared for earthquake resistant design of nuclear power plants. Method's validity was examined by comparing the mean and the c. o. v. values based on this distribution, Type II distribution and those of sample data. Subsequently, relatively reliable data of 300 years (except for Hokkaido area) were used to estimate the probability distribution of the seismicity in terms of the annual maximum of the averaged velocity amplitudes at bed rock at 271 sites in Japan. Contour maps for the mean and the c. o. v. of 50 year maximum values were also provided. The comparison with available references indicates that the mean values are in fairly good agreement, while the c. o. v. value are smaller than those based on Type II distribution.
