GEOPHYSICAL RESEARCH LETTERS, VOL. 32, L09308, doi:10.1029/2005GL022631, 2005

Source fault structure of the 2003 Bam earthquake, southeastern Iran, inferred from the aftershock distribution and its relation to the heavily damaged area: Existence of the Arg-e-Bam fault proposed Takeshi Nakamura,1 Sadaomi Suzuki,2 Hossein Sadeghi,3 Seyed Mahmoud Fatemi Aghda,4,5 Takeshi Matsushima,6 Yoshihiro Ito,7 Sayyed Keivan Hosseini,1 Arash Jafar Gandomi,1 and Mehdi Maleki5 Received 5 February 2005; revised 9 April 2005; accepted 20 April 2005; published 14 May 2005.

[ 1 ] We investigate the hypocenter distribution of aftershocks of the December 26, 2003 Bam earthquake Mw 6.5 by using a temporal seismic network. The hypocenters distribute linearly over about 20 km in parallel with a line 3.5 km west of the geological Bam fault and extend from the south of Bam city to the heavily damaged area in the eastern part of the city including the historical mud brick citadel ‘‘Arg-e-Bam’’. Based on the hypocenter distribution, we propose a schematic 3-D structural model of a new fault that we have named the Arg-e-Bam fault. We suggest that the Bam earthquake occurred not in the Bam fault but in the Arg-e-Bam fault. Bam city is located just above northern part of the Arg-eBam fault and also on the rupture propagation direction from the asperity. This may be one of the main reasons why eastern part of Bam city suffered the heaviest damage. Citation: Nakamura, T., S. Suzuki, H. Sadeghi, S. M. Fatemi Aghda, T. Matsushima, Y. Ito, S. K. Hosseini, A. J. Gandomi, and M. Maleki (2005), Source fault structure of the 2003 Bam earthquake, southeastern Iran, inferred from the aftershock distribution and its relation to the heavily damaged area: Existence of the Arg-e-Bam fault proposed, Geophys. Res. Lett., 32, L09308, doi:10.1029/2005GL022631.

1. Introduction [2] The Bam earthquake occurred in the southeastern part of Iran at 1:56 UTC on December 26, 2003 (Figure 1a). USGS reported that its hypocenter was located at 29.004°N, 58.337°E, and depth 10 km. Teleseismic focal mechanisms from several groups [e.g., Yamanaka, 2003] show a steeplydipping, right lateral strike-slip fault. This earthquake caused catastrophic damage to Bam city and neighboring villages with a collective population of about 142,000. 26,271 people were killed and tens of thousands of people 1 Department of Earth and Planetary Sciences, Faculty of Sciences, Kyushu University, Fukuoka, Japan. 2 Tono Research Institute of Earthquake Science, Mizunami, Japan. 3 Earthquake Research Center, and at Department of Geology, Faculty of Sciences, Ferdowsi University, Mashhad, Iran. 4 Geological Department, Faculty of Science, Tarbiat Moallem University, Tehran, Iran. 5 Natural Disaster Research Institute of Iran, Tehran, Iran. 6 Institute of Seismology and Volcanology, Faculty of Sciences, Kyushu University, Shimbara, Japan. 7 National Research Institute for Earth Science and Disaster Prevention, Tsukuba, Japan.

Copyright 2005 by the American Geophysical Union. 0094-8276/05/2005GL022631$05.00

injured. The well-known historical citadel ‘‘Arg-e-Bam’’, which is the biggest adobe complex in the world and the World heritage sites by UNESCO, was also severely damaged by the Bam earthquake. [3] The main reason for such massive damage may be the weakness of adobe and brick houses. However, the damage was disproportionately and unexpectedly large in comparison with the magnitude of the earthquake. This fact was the first motivation for our study. The Bam fault (dashed line in Figure 1b), which was well known before the earthquake, extends along the west side of Baravat village, about 5 km southeast of Bam city. Just after the earthquake it was supposed that the main shock had occurred in the geological Bam fault [e.g., Ahmadizadeh and Shakib, 2004]. However, nobody could find any clear evidence of dislocation on this fault. This raised the question of where the exact source fault of this earthquake was. This was the second motivation. Moreover, the area of heaviest damage occurred not around Baravat village close to the Bam fault but in the eastern half of Bam city 2 – 4 km away. To discover why Bam city suffered the worst damage was our third motivation. We therefore went to the Bam area to carry out seismological investigations taking seismometers and paid particular attention to the source fault structure of the Bam earthquake by analyzing aftershock data recorded using a temporal seismic network. Analyzing accurate hypocenters of the aftershocks and referring SAR results [Talebian et al., 2004], we propose the existence of the ‘‘Arg-e-Bam fault’’, so named to distinguish it from the Bam fault and present an effective result for evaluating seismic hazards of the Bam earthquake.

2. Aftershock Observation and Analysis [4] A seismic network consisting of 9 temporal stations (Figure 1b) was installed in and around Bam city on February 6, 2004, and monitoring continued until 7 March 2004 [Suzuki et al., 2004]. Each station was equipped with a high sensitivity, velocity type, three-component seismometer (LE-3D) with a natural frequency of 1 Hz. We also installed a three-component strong motion accelerometer (JEP-6A3) in the station at Arg-e-Bam citadel (yellow triangle with circle in Figure 1b). The waveform data were continuously recorded at a sampling rate of 100 Hz by a 16-bit data-logger (LS-8000SH). [5] We picked manually onsets of clear P- and S-waves recorded at more than 3 stations. The reading errors for Pand S-waves are less than several 10 ms. The 2908 initial hypocenter and origin time were determined from these onset data using the HYPOMH software [Hirata and

L09308

1 of 5

L09308

NAKAMURA ET AL.: SOURCE FAULT STRUCTURE OF THE 2003 BAM EARTHQUAKE

L09308

Figure 1. (a) Location of the 2003 Bam earthquake in Iran. Focal mechanism of the main shock [Yamanaka, 2003] is also shown. (b) Epicenter distribution (red dots) of aftershocks as shown on the NASA satellite map (http:// earthobservatory.nasa.gov/). Yellow triangles indicate stations of the temporal seismic network we installed. The station in Arg-e-Bam is also marked by a circle. The black dashed line indicates the traced line of the Bam fault inferred from the geological map supplied by the National Geoscience Database of Iran. (c) The cross-section projected on the N2°W – S2°E direction from the reference point (cross mark in (b)), parallel to the source fault. (d) The cross-section projected on the S88°W – N88°E direction, perpendicular to the source fault. Matsu’ura, 1987]. Selecting neighboring events within 3 km, we then relocated 2789 aftershocks by the doubledifference method [Waldhauser and Ellsworth, 2000]. A 1D local velocity model in this area by Sadeghi et al. [2004] (Figure 2) was used for the hypocenter determination. Their root mean square residuals of the double-difference times decreased from 180 to 50 ms. By using the bootstrap method, the median standard errors of hypocenters are estimated to be 0.06, 0.05, and 0.15 km for the N-S, E-W, and depth, respectively. The magnitudes of aftershocks are from M 0.1 to 3.1.

Bam fault itself but are distributed along a line parallel and about 3.5 km to the west of the fault. The epicenter distribution is nearly linear in the direction of N2°W – S2°E. The dip angle of the hypocenter distribution is nearly

3. Aftershock Distribution [6] Figure 1b shows epicenter distribution of aftershocks overlapping on a satellite image taken by a NASA satellite. Surprisingly, most of epicenters are not on the geological

Figure 2. P- (solid line) and S-wave (dashed line) velocity model used in this study. The VP/VS = 1.73 is used.

2 of 5

L09308

NAKAMURA ET AL.: SOURCE FAULT STRUCTURE OF THE 2003 BAM EARTHQUAKE

L09308

The difference of the epicenter distribution between our 1D and his 3D velocity models is within 1.0 km and the hypocenter inaccuracy is quite unlikely to affect the offset of faults. Considering the aftershock distribution in Figure 1 and the location error (at most 100 m) against its offset (at least 3 km), we may suggest that the main shock occurred not in the Bam fault but in the Arg-e-Bam fault.

4. 3D Structure of the Arg-e-Bam Fault

Figure 3. Detailed cross-sections of the aftershock distribution projected on the S88°W – N88°E direction. Red diamonds and black inverted triangles show the surface displacement by Talebian et al. [2004] and the Bam fault, respectively.

vertical or appears to lean slightly in the S88°W – N88°E vertical cross-section (Figure 1d). The trends of epicenteral and hypocenteral distribution are in accord with the strike and dip angle of the focal mechanism (strike, dip, slip) = (175°, 85°, 153°) of the main shock [Yamanaka, 2003] as shown in Figure 1a. The horizontal length of the distribution is about 20 km with a depth range from 0 km to 16 km in the N2°E – S2°W vertical cross-section (Figure 1c). The source dimension is therefore roughly estimated to be 20 km  16 km. This coincides with the teleseismic result of Yamanaka [2003]. This means that the hypocenter distribution of aftershocks occurring during the observation term presents a general view of the source fault structure of the Bam earthquake. [7] The distribution exhibits different features in the southern, middle and northern areas. In the southern area, most of the epicenters are aligned along a weak curve with a slight scatter. Their density is higher in the 4 – 10 km depth range. In the middle area, most of epicenters are distributed along a thin 5 km line. In the cross-section (Figure 1c) a seismic gap can be distinguished at a depth from 5 km to 10 km. Das and Henry [2003] have suggested that generally few, and usually smaller, aftershocks occur in the high-slip regions of the fault. The seismic gap in Figure 1c, therefore, may correspond to the higher slip region (asperity) of 80 cm to 1 m proposed by Yamanaka [2003]. [8] The northern part of the epicenter distribution shows dispersion. There is even the possibility of fault branches in the NNW and NNE directions. In this area, some of the aftershocks occurred at a shallower depth near ground surface. We note that the northern part of the epicenter distribution corresponds to the populated area of eastern Bam city. And Arg-e-Bam is also located in the epicenter distribution (Figure 1b). We therefore named the source fault inferred from this epicenter distribution the ‘‘Arg-e-Bam fault’’ to distinguish it from the Bam fault. And we confirm the offset with 3 to 4 km between the Arg-e-Bam fault and the Bam fault even using another hypocenter determination method of a preliminary 3D velocity model (H. Sadeghi, personal communication).

[9] Figure 3 shows detailed slices of the seismic crosssection in contrast with the location of the geological Bam fault on the ground surface (black inverted triangle). We also show the location of surface displacement modeled from Envisat radar data by Talebian et al. [2004] (red diamond in Figure 3). In sections of B, C, and D of the northern region the pattern of aftershocks is very complex with three main linear branches. In section E of the central region, the main cluster of aftershock distribution clearly shows a direct vertical trend like a wedge, facing not to the Bam fault but toward the surface displacement. The trend in section F of the southern region is not clear but seems to be extending to the Bam fault. However, referring the relationship between a dense cluster of aftershocks and the surface displacement in section F, we may suggest that a dense cluster deeper than 4 km in section F which is continued in section E shows a main fault section under the surface displacement rather than a shallower cluster. [10] By using dense clusters of aftershocks in Figure 3 a schematic model of the 3-D structure of the Arg-e-Bam fault is shown in Figure 4. This 3D fault structure is modeled as follows. First, we identify the strike and the length of each fault section from the epicenter distribution. Then, we estimate the dip angle and the width of fault section by

Figure 4. Schematic 3-D model of the Arg-e-Bam fault (blue rectangle) distinguishing it from the Bam fault (dashed line). The fault sections are numbered from 1 to 5. Yellow and green triangles show the locations of the Arg-e-Bam citadel and Bam accelerograph station, respectively.

3 of 5

L09308

NAKAMURA ET AL.: SOURCE FAULT STRUCTURE OF THE 2003 BAM EARTHQUAKE

L09308

projecting in the direction perpendicular to the fault strike. The shape of the fault in the central and southern region (fault section 4 and 5 in Figure 4) is rather simple but seems to be slightly twisted. In the northern region there are three branching fault sections with different depth ranges. In this figure we also plot the locations of the Arg-e-bam (yellow triangle), and the Bam accelerograph station (green triangle) of BHRC (Building and Housing Research Center), respectively. We suppose that the asperity locates in the fault section 4 from the reason as mentioned in the previous section.

5. Relation Between Fault Structure and Damaged Area [11] Figure 5 shows a map of the damaged area in and around Bam city. The schematic structure of the Arg-e-Bam fault shown in Figure 4 has also been projected onto the damage map. Comparing the heavily damaged area with the projection of the Arg-e-Bam fault taking into consideration the average error in determining epicenters of aftershocks, we notice that most of the heavy damage which concentrates in the eastern region of Bam city is located just or nearly on the three fault branches in the northern part of the Arg-e-Bam fault. In this area, most of the houses (more than 80%) collapsed, and very strong ground motion of 979.95 gals was recorded at the Bam BHRC accelerograph station. [12] Looking at the source fault (Figure 5) and its rupture process [Yamanaka, 2003], we suggest that the most heavily damaged areas in eastern Bam city are located not only on the northern part of the Arg-e-Bam fault but also along the rupture propagation direction from the asperity (fault section 4 in Figure 4). In general, strong seismic energy arrives due to the directivity [e.g., Lay and Wallace, 1995, p. 368]. The relationship between the location of Bam city and the rupture propagation direction must be one of the causes of the devastating damage. In contrast, Baravat village close to the Bam fault was not so strongly damaged compared with eastern Bam city. Baravat village is not located on the Arg-e-Bam fault and not on the rupture propagation direction from the asperity. Those spatial conditions may have been the reason that Baravat village did not suffer the highest level of damage.

6. Conclusion [13] We determined the accurate hypocenters of 2789 aftershocks of the Bam earthquake by using a temporal seismic network. The distribution of aftershocks reveals the following features. [14] 1. The overall trend of the epicenter distribution is virtually linear along an approximately 20 km axis in the N2°W – S2°E direction, parallel to a line about 3.5 km west from the geological Bam fault on the ground surface. [15] 2. The hypocenter distribution shows a nearly vertical trend or a slight tendency to lie farther west with increasing depth from 0 km to 16 km. This distribution does not correspond to the Bam fault. [16] 3. Most of the heavily damaged area, which was concentrated in the eastern region of Bam city including Arg-e-Bam, is on the northern part of the source fault inferred from the aftershock distribution. We propose the

Figure 5. A map of the damaged area in and around Bam city (National Geoscience Database of Iran, http:// www.ngdir.ir/Downloads/Downloads.asp, Copyright # 2004 National Geoscience Database of Iran). Purple, red and yellow correspond to damage rates of 80– 100%, 50– 80% and 20– 50%, respectively. Blue dashed lines indicate projected sections of the schematic Arg-e-Bam fault as shown in Figure 4. The fault section numbers are the same as those shown in Figure 4. The Bam fault is shown by a black dashed line.

‘‘Arg-e-Bam fault’’ as the source fault to distinguish it from the Bam fault. The Bam earthquake occurred not in the Bam fault but in the Arg-e-Bam fault. [17] 4. The heavily damaged area in eastern Bam city is located not only on the Arg-e-Bam fault but also on the rupture propagation direction from the asperity. This may be one of the main reasons why the eastern part of Bam city suffered the heaviest damage. [18] From the aftershock results, which show the structure and location of the source fault of the Bam earthquake, we were able to resolve qualitatively the question of why the eastern part of Bam city suffered the heaviest damage. As the next step, we need to make a more detailed and quantitative study to fully evaluate the seismic hazards of the Bam earthquake. And the relationship between the Arg-e-Bam fault and the geological Bam fault is one of the most interesting questions for seismogenesis. [19] Acknowledgments. We greatly thank the Natural Disaster Research Institute of Iran for all supports. Naoshi Hirata, Taku Urabe, Kenji Uehira, Tomomi Okada, Tamao Sato, Hiroshi Takenaka, and Nobuki Kame helped in our seismic observation and data analysis. Seismological members of Kyushu University and Ferdowsi University also helped in our data analysis. We are grateful to two anonymous reviewers for their remarks and helpful comments. This research is mainly supported by the Grant-in-Aid for Scientific Research No.15800013 from the Ministry of Education, Science, Culture, Sports and Technology of Japan.

References Ahmadizadeh, M., and H. Shakib (2004), On the December 26, 2003, southeastern Iran earthquake in Bam region, Eng. Struct., 26, 1055 – 1070, doi:10.1016/j.engstruct.2004.03.006. Das, S., and C. Henry (2003), Spatial relation between main earthquake slip and its aftershock distribution, Rev. Geophys., 41(3), 1013, doi:10.1029/ 2002RG000119.

4 of 5

L09308

NAKAMURA ET AL.: SOURCE FAULT STRUCTURE OF THE 2003 BAM EARTHQUAKE

Hirata, N., and M. Matsu’ura (1987), Maximum-likelihood estimation of hypocenter with origin time eliminated using nonlinear inversion technique, Phys. Earth Planet. Inter., 47, 50 – 61. Lay, T., and T. Wallace (1995), Modern Global Seismology, Elsevier, New York. Sadeghi, H., S. Suzuki, S. M. Fatemi Aghda, T. Matsushima, Y. Ito, S. K. Hosseini, T. Nakamura, J. Gandomi, and M. Maleki (2004), Source fault and source parameters of the 2003 Bam, Iran earthquake, 5th ASC General Assem., Asian Seismol. Comm., Yerevan, Armenia. Suzuki, S., T. Matsushima, Y. Ito, S. K. Hosseini, T. Nakamura, A. J. Gandomi, H. Sadeghi, M. Maleki, and F. Aghada (2004), Source fault of the 2003/12/26 Bam earthquake (Mw 6.5) in southeastern Iran inferred from aftershock observation data by temporal high-sensitive-seismograph network, Eos Trans. AGU, 85(17), Jt. Assem. Suppl., Abstract S23A-07. Talebian, M., et al. (2004), The 2003 Bam (Iran) earthquake: Rupture of a blind strike-slip fault, Geophys. Res. Lett., 31, L11611, doi:10.1029/ 2004GL020058. Waldhauser, F., and W. L. Ellsworth (2000), A double-difference earthquake location algorithm: Method and application to the northern Hayward fault, Bull. Seismol. Soc. Am., 90, 1353 – 1368.

L09308

Yamanaka, Y. (2003), Seismological note No. 145, Earthquake Inf. Cent., Earthquake Res. Inst., Univ. of Tokyo, Tokyo.

S. M. Fatemi Aghda and M. Maleki, Natural Disaster Research, Institute of Iran, Shabdiz Avenue No. 1, Tehran 19395-4676, Iran. A. J. Gandomi, S. K. Hosseini, and T. Nakamura, Department of Earth and Planetary Sciences, Faculty of Sciences, Kyushu University, Hakozaki 6-10-1, Fukuoka 812-8581, Japan. ([email protected]) Y. Ito, National Research Institute for Earth Science and Disaster Prevention, Tennodai 3-1, Tsukuba 305-0006, Japan. T. Matsushima, Institute of Seismology and Volcanology, Faculty of Sciences, Kyushu University, Shin-yama 2-5643-29, Shimbara 855-0843, Japan. H. Sadeghi, Earthquake Research Center, Department of Geology, Faculty of Sciences, Ferdowsi University, Vakil-abad St., Mashhad 91775-1436, Iran. ([email protected]) S. Suzuki, Tono Research Institute of Earthquake Science, Yamanouchi Ayeko-cho 1-63, Mizunami 509-6132, Japan. ([email protected])

5 of 5