Effects of zonation on the results of the application of the regional time predictable seismicity model in Greece and Japan

Recent work on time dependent seismicity is mainly based on the so-called "regional time predictable model", which is expressed by the relation: logT(t) = cM(p) + a where T-t is the interevent time, i.e. the time between two successive mainshocks of a seismogenic region, and M-p is th; magnitude of the precedent mainshock. Parameter a is a function of the magnitude of the minimum earthquake considered and of the tectonic loading and c is a positive (approximate to 0.3) constant. A problem of the method, as it has been applied till now, is its dependence on the zonation, that is, on the definition of the seismogenic regions, which is subjective to some extent. In the present work a different approach, which assumes no a-priori regionalization of the area, is attempted in order to check the validity of the model. Grids of equally spaced points at 0.5 degrees have been created for Greece and Japan and the mainshocks located within each circle with center at a point and radius varying between 30 and 150 km were considered. When the number of mainshocks within the circle was four or larger, regression was performed and the c value was calculated. In about 75 % of the cases for Greece with sufficient data and 80% for Japan the parameter c was found to be positive. This result strongly supports the validity of the model.