Subevent location and rupture imaging using iterative backprojection for the 2011 Tohoku Mw 9.0 earthquake

S U M M A R Y Knowledge of the rupture speed and spatial–temporal distribution of energy radiation of earthquakes is important for earthquake physics. Backprojection of teleseismic waves is commonly used to image the rupture process of large events. The conventional backprojection method typically performs temporal and spatial averaging to obtain reliable rupture features. We present an iterative backprojection method with subevent signal stripping to determine the distribution of subevents (large energy bursts) during the earthquake rupture. We also relocate the subevents initially determined by iterative backprojection using the traveltime shifts from subevent waveform cross-correlation, which provides more accurate subevent locations and source times. A bootstrap approach is used to assess the reliability of the identified subevents. We apply this method to the Mw 9.0 Tohoku earthquake in Japan using array data in the United States. We identify 16 reliable subevents in the frequency band 0.2–1 Hz, which mostly occurred around or west of the hypocentre in the downdip region. Analysis of Tohoku aftershocks shows that depth phases can often produce artefacts in the backprojection image, but the position and timing of our main shock subevents are inconsistent with depth-phase artefacts. Our results suggest a complicated rupture with a component of bilateral rupture along strike. The dominant energy radiation (between 0.2 and 1 Hz) is confined to a region close to the hypocentre during the first 90 s. A number of subevents occurred around the hypocentre in the first 90 s, suggesting a low initial rupture speed and repeated rupture or slip near the hypocentre. The rupture reached the coastal region about 106 km northwest of the hypocentre at 43 s and the region about 110 km north of the hypocentre at 105 s with a northward rupture speed ∼2.0 km s−1 at 60–110 s. After 110 s, a series of subevents occurred about 120–220 km southwest of the hypocentre, consistent with a 3 km s−1 along-strike rupture speed. The abundant high-frequency radiation in the downdip region close to the coast suggests intermittent rupture probably in the brittle–ductile transition zone. The lack of high-frequency radiation in the updip region suggests the rupture near the trench was more continuous, probably due to more homogeneous frictional properties of the shallow slab interface. The lack of early aftershocks in the updip region indicates that most of the accumulated slip in the updip region during the interseismic period was probably released during the main shock.