The fault width formation of impending large earthquakes: Its observation near the base of the crustal seismogenic zone by the time series analysis of seismicity.

Dividing Japan into meshes of about 5°, we collect earthquakes (EQ's) for each mesh-area from a focus catalog of Japan with a regionally dependent magnitude window of M >= 3-3.5. The time history of each mesh-collection is a string of EQ events, which draw a trajectory in a physical space. The space coordinates are the EQ epicenter, focal depth (DEP), inter-EQ time interval (INT), and magnitude (MAG). Thus, each coordinate component of the trajectory is the time series of the corresponding EQ source parameter where time is the chronological event index. Taking a moving-average of the series over 15-25 events, we find only two different triple phase couplings of the averaged DEP, INT, and MAG fluctuations precursory to every large EQ (M >= about 6) throughout Japan [Takeda, 2003; Takeda and Takeo, 2004]. Each triple phase coupling begins the MAG with medium MAG of about 4.1 at either small (shallow) DEP and large INT or large (deep) DEP and small INT, then change it to small MAG of about 3.8 at either deep DEP and small INT or shallow DEP and large INT. The transition of the EQ state creates a large linear DEP variation (W) on its series, which becomes comparable to the fault width of large EQ's except for only a very few cases. Thus, the precursory variation W appears to load the corresponding stress into the local region to prepare for an impending large EQ whose fault width becomes W. With the assumption of the fault length (L in km) being L = 2W, we can successfully forecast the M of the impending large EQ by an empirical relation given by Utsu [2002], log L = 0.5M - 1.8 for 6 =< M =< 8.5, [Takeda and Takeo, 2007]. Each triple phase coupling can draw its own physical picture of the seismogenic process in the earth lithosphere consisting of the brittle (B), brittle ductile transition (B-D) and ductile (D) layers. The B-D layer is at the base of the crustal seismogenic zone. The plate driving force of about 3x1012 Nm-1 creates steady state creep in the D part. According to Aki [2004], Zoback and Zoback [2002] gave a new perspective on the role of D part during the EQ loading process as follows. If the creep deformation rate is high, the stress in the B part builds up and the region becomes tectonically active by the coupling of three layers. Thus, the EQ's of various sizes in magnitude (M) occur in the B part. Aki [2004] and Jin and Aki [2005] have proposed a local version of the B and D coupling as "B-D interaction hypothesis" by a time series analysis (completely different from ours) of the decay rate of coda waves and the local selective seismicity. It is extremely intriguing that the hypothesis is in perfect harmony with physical models of the seismogenic zone inferred from the triple phase couplings.