10.1007/s11434-011-4885-1 Seismicity characteristics of the 2011 M 9.0 Tohoku earthquake near the East Coast of Honshu in

2011 The spatiotemporal evolution of the M 9.0 Tohoku earthquake sequence off the East Coast of Honshu in Japan on March 11, 2011 and precursive seismic activity near the Japan Trench show that the earthquake sequence has foreshock-main shock-aftershock characteristics. Its foreshock sequence is characterized by a concentrated spatial distribution, low b value and the same focal mechanisms. Half an hour after the main shock, the two greatest aftershocks, with magnitudes of M 7.9 and M 7.7, occurred, fol-lowed by a rapid reduction in the strength of events. The aftershock activity was enhanced roughly two weeks and one month after the main event. This great earthquake ruptured bilaterally. Five hours after the main shock, the aftershock zone extended over a range that was 500 km in length and 300 km in width. A day later, the long axis of the aftershock area had expanded to about 600 km. Nine years prior to the 2011 earthquake, the seismicity in the location of the seismic source for this event enhanced significantly, with the extent of this area of enhanced seismicity being roughly equivalent to the aftershock zone. foreshock sequence, aftershock activity, accelerating moment release

According to the China Earthquake Networks Center (CENC), the M9.0 Tohoku earthquake occurred near the east coast of Honshu in Japan (longitude 142.6°E, latitude 38.1°N, focal depth 20 km) at 13:46 on March 11, 2011 (Beijing time) (www.ceic.ac.cn). The earthquake triggered a strong tsunami, and by April 12 the number of missing and deceased reached 27475 (www.sina.com.cn).
The global earthquake catalog of the U.S. National Earthquake Information Center (NEIC) reports the occurrence of a M7.3 earthquake and 22 shocks with M   5 in the region of the M9.0 earthquake during the two days (March 9-11) prior to the main M9.0 earthquake. The M7.3 event was only 46 km away from the epicenter of the M9.0 main shock (http://earthquake.usgs.gov/). In retrospect, these events can be classified as foreshocks in the real sense. The sequence also has had a great number of aftershocks, including three with M   7.0 within one month after the main earthquake. Therefore, the great earthquake sequence can be characterized as a foreshock-main shock-aftershock type.
Of all of the short-term and imminent precursors, foreshocks are recognized as one of the most effective indicators of earthquake prediction [1]. The contribution of foreshocks is obvious among all the effective short-term and imminent precursors in China, such as the M7.3 Haicheng event in 1975 and the M5. 6 Xiuyan earthquake in 1999. However, the effective recognition of foreshocks is difficult. The 1996 Xingtai earthquake in China, led to an increased interest in earthquake sequences [2,3]. Chen [4] and Chen et al. [5] found foreshocks to be concentrated in space and with focal mechanisms that are relatively similar; however, aftershocks were found to be more distributed in space and differed widely in their mechanisms. They also proposed that the consistency of the source parameters can indicate whether a sequence is a type of foreshock sequence or a normal main shock-aftershock sequence. Ni et al. [6] studied the temporal and spatial sequence distribution and waveform characteristics of the M4.7 earthquake occurring about 2 h before the M7.1 Yushu shock on April 14, 2010 in Qinghai at a distance of only 2 km from the epicenter of the main earthquake. They suggest that the M4.7 sequence is in line with the theory proposed by Chen [4], i.e. that it is a typical foreshock sequence. In addition, statistical methods based on the b value have been used to identify foreshock sequences [7]. In recent years, a number of foreshock investigations have been conducted based on high-precision locations of earthquake sequences. These studies have systematically analyzed the temporal and spatial evolution of foreshock-main shock-aftershock sequences, and provided models to explain the relationship between foreshocks and a main event [8]. Previous foreshock research mainly involved the statistical characteristics of moderate-strong earthquake sequences [9]. Examinations of earthquakes with M   8 earthquakes were uncommon, and in particular, the study of earthquake sequences with a foreshock of M   7.0 was rare.
Many studies have been made of the great earthquakes [10][11][12]. Mogi pointed out that 10 to 20 years before great earthquakes, seismic activity obviously strengthened and a doughnut-shape distribution of earthquakes often formed around the source regions [11]. Mei et al. [12] analyzed the precursive seismicity for 16 great earthquakes on the Chinese mainland and adjacent regions. They noted enhanced seismic activity of moderate earthquakes in the vicinity of the source regions has certain universal characteristics over a long-term preparation stage leading to a main shock earthquake.
In this study, we have used the global catalog, beginning in 1973, from the United States National Earthquake Information Center (NEIC; http://earthquake.usgs.gov/) to obtain data for an analysis of the characteristics of the spatiotemporal evolution of the M9.0 Tohoku great earthquake sequence and, in particular, the features of seismic activity near the Japan Trench before the large event. Through analysis of the Gutenberg-Richter (G-R) relationship, the records of M   5 earthquakes near the Japan Trench are considered to be basically complete. Taking into account the agency bias in earthquake locations, this work has used both the Japan Meteorological Agency catalog (JMA; http:// www.jma.go.jp/jma/index.html) and NEIC catalog for comparative analysis of the foreshock sequence and the b-value for the earthquake sequence. In addition, we make use of part of the global catalog of M W   7.0 earthquakes since 1900 in the discussion (which mainly comes from the global catalog from 1900 to 1999, revised by Engdahl et al. [13], and the catalogue from Harvard University after 2000 (http://www.seismology.harvard.edu/)), in which the rec-ord of M   7.0 earthquakes is complete [13].
1 Characteristics of the evolution of the M9.0 Tohoku earthquake sequence

Spatiotemporal evolution of foreshocks and first three days of aftershocks
At 02:45 UTC on March 9, 2011, a M7.3 earthquake was the first event to occur in the great earthquake sequence. By the time of the occurrence of the main shock at 05:46 on March 11, 23 M   5.0 foreshocks had occurred, of which two were M   6.0. Viewed temporally (Figure 1(a)), the foreshock sequence can be divided into three active periods: from 02:00 to 11:00 on March 9 (Figure 1 . Before the main event, the foreshock sequence was quiescent for 13 h, the most prominent quiescence for the entire foreshock sequence. Note that some previous studies of foreshocks have shown that the closer to the main shock, the more frequent the earthquakes are, but a short calm period would appear before the main shock [2, 14,15]. In contrast, the calm phenomenon before this main shock was prominent; however, the earthquakes were not frequent. Whether this was affected by the minimum magnitude needs further study. The spatial distribution of the foreshock sequence is concentrated, with the epicenters concentrated in a range of 40-50 km (Figure 1(b),(e),(f)). From the point of view of depth, the initial rupture point was 30 km deep (the focal depth of the M7.3 was 32 km), then subsequent earthquakes generally occurred at more shallow depths. The focal depths of the foreshock sequence mainly are distributed over the range of 10-40 km (Figure 1(i)).
After the main event, the aftershock zone rapidly extended bilaterally to the north and south, with a greater distribution to the south. Twenty-eight minutes after the main earthquake, the greatest aftershock, with M7.9, occurred about 260 km to the south of the main shock (06:15). Ten minutes later, the second greatest aftershock, with M7.7, occurred about 200 km away from the main shock in the east of the trench (06:25). Five hours later, the NNEdirected aftershock zone was well formed, about 500 km long and 300 km wide. Eighteen hours after the main shock (i.e. by 24:00 on March 11), 120 aftershocks of M   5.0 had (b) epicenter distribution (solid circles for the foreshocks, hollow circles for aftershocks, five-pointed star for the main shock); (c) foreshocks distribution (hollow circles, solid circles and triangles respectively stand for earthquakes occurring from 02:00 to 11:00 am on March 9, from 18:00 to 23:00 on March 9 and from 08:00 to 16:00 on March 10; five-pointed star represents the main shock); (d) epicenter distribution of foreshock sequences from NEIC catalog and the JMA catalog (circles represents the NEIC catalog, triangles for the JMA catalog, hollow circles and triangles for foreshock sequences, solid circle and triangle represent the main shock); (e) epicenter migration map in AB direction (i.e. projecting the epicenter of the earthquake on the line AB and calculating the change of earthquake projection distance from point A over time); (f) epicenter migration map in CD direction (i.e. projecting the epicenter of the earthquake on the line CD, and calculating the change of earthquake projection distance from point C over time); (g) focal depth distribution in profile AB; (h) focal depth distribution in profile CD; (i) focal depths varying with time.
occurred, including 21 earthquakes with M   6.0 and two M   7.0 events. Compared to the five hours after the main shock, there was no significant change in the size of the aftershock area. After March 12, there was a slight expansion of the long axis of the aftershock area to a length of about 600 km, with the width unchanged (Figure 1(e),(f)).
From the focal depth profile, we see that along the ocean trench, the distribution of aftershock depths changed little, mainly concentrated to 0-40 km (Figure 1(g)). The main aftershock zone was located in the western part of the trench, some aftershocks also occurred in the eastern part of the trench, including the second greatest aftershock of M7.7 and four aftershocks with magnitudes 6.1-6.4. These five earthquakes with M   6.0 all occurred within the first 2.5 h after the main shock. In addition, compared with the western side of the trench, the focal depths of aftershocks in the eastern part of the trench were shallow, concentrated in 0-40 km, but there were some earthquakes deeper than 40 km ( Figure  1(h)).
From the time of the main shock until 15:00 on March 12, most of the aftershocks were concentrated in the 0-50 km depth range, and some were in the 50-70 km range. Subsequently, few aftershocks had hypocenters below 40 km, and they were concentrated mainly in the 0-40 km range (Figure 1(i)).
In view of some systematic bias in earthquake locations, this paper comparatively analyzed the NEIC and JMA catalogs. The results show that the magnitudes determined from the JMA catalog are generally higher, and the epicenters shift slightly to the east. However, the use of the JMA catalog in this research has not affected the results obtained using the NEIC catalog. From Figure 1(d), the epicenters determined by the JMA catalog on the whole are more to the east than those from the NEIC catalog. The epicenter of the main shock determined by the two catalogs has a discrepancy of about 50 km. The JMA positioned foreshocks provided a more concentrated spatial distribution, and the main shock was located to the southwestern end of the foreshock sequences.

Spatiotemporal evolution of the earthquake sequence in the month after the main shock
Aftershock activity in the month after the main event showed obvious fluctuations. The M   6.0 aftershocks can be divided into three periods: March 11 to 17, March 22 to April 1 and April 7 to 12 (Figure 2(a)). During the first three days of the first stage (March 11-13), there were 27 aftershocks with M   6.0, including two M   7.0 earthquakes, and these events were distributed across the entire aftershock zone. Four aftershocks with M   6.0 occurred from March 14 to 17, the greatest being M6.1, moving from the central part to the northern part of the aftershock area. The activity of M   6.0 aftershocks then calmed and became quiet for four days, before the second active phase began. During the second period, there were eight aftershocks with M   6.0, the greatest one being M6.4, concentrated in the south-central part of aftershock zone, close to the epicenter of the main shock. Subsequently, the activity of M   6.0 earthquakes calmed again, this time for 6 d, until a M7.1 aftershock occurred at 14:00 on April 7 about 60 km west of the epicenter of the main shock. This marked the start of the third active phase. By April 12, four M   6.0 aftershocks had occurred, with the seismicity migrating to the south. The M7.1 event occurring on April 7 was the strongest aftershock since March 12. Compared with the first two M   7.0 aftershocks, its epicenter was the closest to that of the main shock (Figure 2(b)).

Parameter calculation for the earthquake sequence
The G-R relationship, lgN(M)=abM, as proposed by Gutenberg and Richter [16], is a statistical description of earthquake intensity and frequency distribution, where N is the number of earthquakes with magnitude M and a and b are statistical factors, respectively reflecting the level of seismic activity and the intensity distribution. Studies of Chinese mainland earthquake sequences [9] show that, for the foreshock-main shock-aftershock type earthquake sequence, aftershocks and foreshocks have obvious differences in b value. Figure 3 shows the G-R curves calculated using the NEIC catalog for the foreshock sequence and first three days of aftershocks (March 11-13) and in the first month (March 11 to April 12) after the M9.0 main shock. As shown in the figure, the b values are 0.825, 1.111 and 1.115 for the foreshocks, the three-day aftershock record and the one-month aftershock record, respectively. Clearly, the b value for the foreshocks is significantly lower than that of the aftershocks. This is consistent with previous research [9,17]. We also made similar calculations using the JMA catalog, and the results are shown in Figure 4. In the figure, the b value is 0.616 for foreshocks, 0.886 for the three-day aftershock record and 0.907 for the one-month aftershock record. In contrast with Figure 3, the overall value of b calculated from the JMA catalog is lower than that from the NEIC catalog. This is because the magnitudes determined by JMA are slightly higher than those by NEIC, except for the main shock and the greatest foreshock. Therefore, this difference is caused by the system bias in the magnitude measurement. Nevertheless, Figure 4 clearly shows that the b value for the foreshock sequence is significantly lower than that of the aftershock sequence.
Based on the Omori aftershock attenuation law [18,19], Liu [20] proposed a method for calculating the h value, the earthquake frequency attenuation coefficient, to determine the sequence type for earthquake prediction studies involving foreshocks. When h   1.0, the sequence is a foreshock sequence. When h > 1, the sequence is generally an aftershock sequence, and the occurrence of subsequent greater earthquake is unlikely. Based on calculations for the Tohoku sequence, the h value is 1.9 for the foreshocks and 1.41 and 1.27, respectively, for the three-day and one-month aftershock sequences. Thus, according to the value of h, we cannot identify the foreshock sequence in this case.
Focal mechanism solutions of foreshocks and three days of aftershocks after the main shock (March 11-13) were obtained from Harvard University (http://www.seismology. Harvard.edu/). They show that the faulting mechanisms, including the strike, dip and the slip directions of the fault planes, are consistent for the foreshocks and main event. In contrast, the focal mechanism solutions for the three-day aftershock sequence are chaotic, with both thrust faulting and normal faulting. Furthermore, the strikes and dips of the fault planes are quite different ( Figure 5). This is in line with the theory that the consistency of source parameters can be taken as a precursive indicator in forecasting an earthquake sequence [4].    Only four shocks occurred with an average annual frequen-cy of 0.25, significantly lower than during the active periods ( Figure 6(b)). Figure 6(a) shows the epicenter distribution of earthquakes with M   6.0 from 1988 to March 2011. From 1988 to November 2001, the seismicity was weak in the vicinity of the M9.0 earthquake epicenter (the circular region in Figure 6(a)). Only eight earthquakes of M   6.0 occurred (hollow circles in the circular area of Figure 6(a)). The annual frequency was 0.6 and the maximum magnitude was 6.7. From December 2001 until the 2011 M9.0 Tohoku earthquake, seismic activity significantly increased, with 32 earthquakes of M   6.0 in total and an annual average frequency of 3.2, of which there were four earthquakes with M   7, with the largest of those a M7.2 (the solid circle in the circular area in Figure 6(a)). The March 11, 2011, M9.0 earthquake occurred right in the same location as the region of enhanced seismicity (the triangle in Figure 6(a)).
Before some great earthquakes, the activity of mid-strong magnitude earthquakes has been observed to increase. This has been described by Accelerating Moment Release [21,22] (AMR), expressed as the measure of seismicity, using the Benioff strain as in this paper; A and B are constants; the power exponent m is called the scaling constant, and t f is the occurrence time of the great earthquake. When m < 1, the release curve is accelerating; when m > 1 the release curve is decelerating; and  when m = 1, the release curve is linear [23]. The seismic energy of all earthquakes with M   6.0 occurring from January 1, 1988 to March 8, 2011, near the epicenter of the M9.0 event (the circular area in Figure 6(a)) was calculated using the formula of earthquake magnitude and energy proposed by Gutenberg and Richter [24]. The strain released by the earthquake was then determined by extracting the square root of the energy. The best-fit curve is  

Discussion
Small to moderate earthquake activity before the devastating Tohoku earthquake has drawn wide attention from seismologists. In particular, the significance of identifying foreshocks for earthquake forecasts and their potential role in earthquake prediction have been an academic focus for a long time. Over the past 50 years, multiple series of obvious foreshocks occurred in China, such as those before the 1975 M7.3 Haicheng earthquake and the 1999 M5.6 Xiuyan earthquake. Statistically however, the number of moderate and strong earthquakes preceded by foreshocks is less than 10% of the total [4,25]. Records of strong earthquakes preceded by foreshock sequences are also very limited globally, the most notably one is the significant foreshock activity recorded before the M9.6 Chile earthquake of May 22, 1960 [17]. Statistical records show that 17 earthquakes with M w   8.5 have occurred since 1900 globally. Five of these were preceded by a foreshock of M   7.0, accounting for 29.4%. This is significantly higher than the proportion of foreshocks identified before moderate or strong earthquakes. Table 1 shows the basic parameters of the five great shocks and their foreshocks. The time interval between these foreshocks and the main shocks is from several hours to four days, and the distance between foreshocks and main events is 20 to 145 km. If the foreshock could have been identified before the occurrence of the main event, it would have been valuable for forecasting the main earthquake. The foreshock sequence of the M9.0 earthquake along the east coast of Honshu, Japan is characterized by a concentrated spatial distribution, low b value and consistent focal mechanisms. The traditional method of h value determination is not an effective way to identify this particular foreshock sequence. It would be more meaningful if we were able to precisely relocate the sequence, and on that basis, make systematic analyses of the spatiotemporal evolution of the earthquake sequence [8].
Nine years prior to the M9.0 earthquake, seismic activity increased markedly in the vicinity of the source region. Characteristics of regional seismic activity before great events such as this are valuable for the identification of future great earthquakes.