[1] We investigate seismic scaling relations and seismic activity using two Japanese earthquake sequences, the 2004 mid Niigata prefecture earthquake (MNPE) and the 2005 west off Fukuoka prefecture earthquake (WOFE) sequences. Although the mainshocks of the two sequences have the same M W of 6.6, the aftershock activities are different from each other in term of moment release. The MNPE produced a large number of aftershocks including six events with M W ! 5.5 over a period of about two weeks. On the other hand, the WOFE sequence did not produce aftershocks that exceeded M W 5.5. In the scaling relationship of seismic moment (M 0 ) versus corner frequency ( f c ) obtained from the MNPE sequence, the f c 's tend to decrease with decreasing M 0 between M W 3.5 and 6.6 and the best fit line determined in a least squares sense is M 0 / f c À3.41 . In the WOFE sequence, the best fit line is M 0 / f c À3.06 in the same M W range. The f c range of small aftershocks is lower in the MNPE sequence than in the WOFE sequence. Moreover, the M 0 À f c relations for small events (3.5 M W 4.0) in the MNPE sequence are different between the events on the main faults (M 0 / f c À3.53 ) and the off-main fault events (M 0 / f c À4.44 ). These results suggest that there is difference (or variation) in the scaling relation even in the same M W range, probably reflecting different tectonic or seismogenic conditions.

We investigated seismic scaling relations of rupture area S, average slip D, combined area of large-slip area (LSA) SaL, that of strong motion generation area (SMGA) SaS, and acceleration source spectral shortperiod level A versus seismic moment M0 for six crustal earthquakes with 7.5≤Mw≤7.9 in inland crusts and six plate-boundary earthquakes with 8.4≤Mw≤9.1 in subduction zone. We compared our relations obtained above with previous studies, and discussed the difference between source parameters derived from source models based on relatively long-period (10 to 100 s) and short-period (0.1 to 10 s) data. For the crustal earthquakes we confirmed that the scaling relations fit in the relationships suggested by previous studies. On the other hand, for the plate-boundary earthquakes we found a new M0 -S scaling relation of great earthquakes with Mw≳8.4 where S is proportional to M0 1/2 . This corresponds to the second stage in the "three stage scaling model" that is caused by the saturation of fault width due to the restriction of the seismogenic layer in the subduction zone. Moreover we found that SaS of great plate-boundary earthquakes is approximately three times as small as SaL, although M0 -SaL relation fits in the scaling relation, SaL＝0.20･S, suggested by Murotani et al. (2008). These results indicate that source characteristics of great earthquakes on subducting plateboundaries are different from those in inland crusts, especially in the short-period range related to strong ground motions, and that the great plate-boundary earthquakes require some new source models to predict strong ground motions from the scenario earthquakes.

The relationship of seismic moment (M 0 ) to corner frequency ( f c ) of the 2007 Off Mid Niigata prefecture (Chuetsu-oki) earthquake sequence is examined in comparison with that of two other earthquake sequences. Thec . The relatively large deviation from f −3 c may reflect the influence of fluids identified in tomographic studies around the source region of the tectonic strain concentration. Moreover, the lower f c s of the small events (3.4 ≤ M w ≤ 3.8) in this sequence are similar to those of off-main-fault events in the 2004 Mid Niigata prefecture earthquake sequence. The characteristics in the scaling relation may be interpreted in the context of the event locations, which are either on-or off-main faults, and may have implications for the seismogenic conditions.

S U M M A R YThe seismic velocity structure in and around the source area of the 2004 mid-Niigata earthquake, which featured complicated heterogeneities, was investigated by combining waveform modelling and traveltime tomography inversion using low-frequency (0.05 ≤ f ≤ 0.2 Hz) and high-frequency (f ≥ ∼1 Hz) data, respectively. On the footwall of the main shock that includes multiplanar faults, 3-D finite-difference waveform modelling using only a previously proposed tomography model was not sufficient to synthesize the observed waveforms in 0.05 ≤ f ≤ 0.2 Hz at most stations. Thus, we derived a final 3-D model 3DM-28, examining body wave amplitudes, phases and traveltimes. The image produced by model 3DM-28 shows a clearer contrast between low-and high-velocities than that seen in the original tomography models due to the higher velocity on the footwall. This increase in velocity, particularly in the seismogenic zone, also indicates that short-wavelength low-velocity anomalies revealed in the revised tomography image may be more localized in the vicinity of the multiplanar faults than that shown in the original one. Moreover, the low-velocity anomaly zone within a depth range of 15-20 km beneath the seismogenic zone (lower crust) appears to be associated with the short-wavelength low velocity anomalies at shallower depths (upper crust). These characteristics may support the hypothesis of infiltration of pressurized fluids from the lower crust into the multiplanar fault system. The volumes or distributions of such fluids may be clarified by further examination of 'unsatisfactory fit waveforms' recorded at stations along the strikes of the major faults. We suggest that the clear velocity contrasts between the hangingwall and footwall and the upper crust and lower crust, including the effects of fluids, all seem to be essential characteristics of the seismogenic conditions in this earthquake sequence.

The locations and scales of the seismic sources of inland crustal earthquakes without surface fault traces (Mw≲6.5 in Japan) are difficult to identify in advance, even by conducting detailed surveys, and in such a case, it seems rational to uniformly evaluate ground-motion levels in the regions with similar seismogenic conditions. For such earthquakes, we first developed a technique to estimate ground-motion levels in a specific area by calculating the response spectra corresponding to nonexceedance probabilities (NEPs) based on probability density functions derived using strong-motion records. These records were used in the analysis after adjustments to the condition of being and on hard bedrock (VS≈2000–3000  m/s) in the source vicinity. Next, we developed an empirical method to estimate the correspondence between the NEP spectrum levels and their annual exceedance probabilities (AEPs) by considering annual occurrence frequencies for the target event group. Moreover, we showed an example that applied our approach to all over Japan, where a large number of downhole records on stiff baserock (VS≈700–3000  m/s) have been obtained by the KiK-net, a dense nationwide network of vertical array stations (pairs of surface and downhole recordings). In the example, we demonstrated that the empirical AEP spectral levels using our method are consistent with AEP response spectra, that is, uniform hazard spectra, derived from the probabilistic seismic hazard analysis using the kinematic fault rupture modeling method in a previous study.