Spatiotemporal variations of interplate slip rates in northeast Japan inverted from recurrence intervals of repeating earthquakes

Repeating earthquakes, the sequence of stress accumulation and release at isolated small asperities on a plate interface, can be regarded as a renewal process in statistics. From such a point of view, we modelled a sequence of repeating earthquakes and developed an objective Bayesian method to estimate the space-time distribution of interplate slip rates from the recurrence intervals of repeating earthquakes. The space-time distribution of slip rates is represented by the superposition of tri-cubic B-splines. The knots of B-splines in time are unequally allocated for representing co-seismic abrupt and post-seismic rapid changes in slip rates. In addition, to avoid overfitting, smoothness constraints are imposed and their optimal weights are determined by Akaike's Bayesian Information Criterion. We applied this method to the complete data set of repeating earthquakes in northeast Japan for about 18 yr before the 2011 Tohoku-oki earthquake, and revealed spatiotemporal variations of interplate slip rates off the Hokkaido-Tohoku region, where the 1994 Sanriku-oki (M7.6), 2003 Tokachi-oki (M8.0), 2004 Kushiro-oki (M7.1), and 2008 Ibaraki-oki (M7.0) earthquakes occurred. First, we confirmed the reciprocal correlation between the spatial distribution of average slip rates for a seismically calm period (1996-2000) and that of average slip-deficit rates, which has been estimated from GPS array data. Then, we examined the temporal variations of slip rates associated with the large interplate earthquakes in detail.