Lower edge of locked Main Himalayan Thrust unzipped by the 2015 Gorkha earthquake

Large earthquakes are thought to release strain on previously locked faults. However, the details of how earthquakes are initiated, grow and terminate in relation to pre-seismically locked and creeping patches is unclear1–4. The 2015 Mw 7.8 Gorkha, Nepal earthquake occurred close to Kathmandu in a region where the prior pattern of fault locking is well documented5. Here we analyse this event using seismological records measured at teleseismic distances and Synthetic Aperture Radar imagery. We show that the earthquake originated northwest of Kathmandu within a cluster of background seismicity that fringes the bottom of the locked portionof theMainHimalayanThrust fault (MHT).The rupture propagated eastwards for about 140 km, unzipping the lower edge of the locked portion of the fault. High-frequency seismic waves radiated continuously as the slip pulse propagated at about 2.8 kms−1 along this zone of presumably high and heterogeneous pre-seismic stress at the seismic–aseismic transition. Eastward unzipping of the fault resumed during the Mw 7.3 aftershock on 12 May. The transfer of stress to neighbouring regions during the Gorkha earthquake should facilitate future rupture of the areas of the MHT adjacent and updip of the Gorkha earthquake rupture. On 25 April 2015, an earthquake with moment magnitude Mw 7.8 occurred along the Himalayan front close to Kathmandu (Fig. 1). The epicentre was located 80 km to the west–northwest of Kathmandu within a long-identified zone of clustered seismicity that runs beneath the front of the high Himalaya6. The focal mechanism7 indicating thrusting on a subhorizontal fault dipping about 10 northwards and the 15 km hypocentral depth7 make it likely that this earthquake ruptured the MHT, the main fault along which northern India underthrusts the Himalaya at a rate of approximately 2 cm yr−1 (ref. 8). A Mw 7.3 aftershock with a very similar focal mechanism9 occurred on 12 May, 75 km east of Kathmandu (Fig. 1). The geometry of the MHT in the hypocentral area is relatively well known from various geophysical experiments10,11. Geodetic measurements collected over the past 20 years revealed that this fault has remained locked over this time period5,12 and the pattern of locking is nowwell constrained5 (Fig. 1), allowing a detailed comparison with the rupture process during the Gorkha earthquake. We imaged the rupture process by back-projecting13 teleseismic P-waves recorded by the Australian seismic network (Fig. 2a and Supplementary Fig. 1) using the Multitaper-MUSIC array processing technique. The technique tracks the spatio-temporal evolution of the sources of high-frequency radiation (0.5–2Hz) during the rupture process (Supplementary Fig. 2; see Methods). The back-projection forms coherent sources for about 60 s after initiation of the rupture. The high-frequency sources are almost linearly distributed for about 45 s, and their timing indicates a 2.72±0.13 km s−1 eastward propagation (Fig. 2b). They follow remarkably well the downdip edge of the locked zone (Fig. 1) and the cluster of background seismicity (Fig. 2a), including a local kink northwest of Kathmandu. The amplitude rises sharply from 10 to 20 s, peaks from 20 to 40 s, and decays abruptly after about 45 s (Fig. 2c). High-frequency radiation persists after 45 s, but migrates updip in a southeastward direction. The 12May aftershock occurred a few tens of kilometres east of where the initial phase of along-strike propagation of the rupture stopped (Fig. 2a). We also determined a finite source model of the rupture from the joint inversion14 of teleseismic waveforms in the 0.01–1Hz frequency band and static surface displacements measured from SAR image offsets. The fault is assumed planar and its dip angle was adjusted to 7 by trial and error. The model assumes that, once initiated, slip accrues over a certain duration (rise time) in the wake of the rupture front. The inversion solves for the final slip amplitude, rake, rise time and rupture front velocity at each grid point (see Methods). The source model is determined so as to best fit the static surface displacements (Supplementary Fig. 3) and teleseismic waveforms (Supplementary Fig. 4). The static surface displacements were measured using two pairs of European Space Agency’s Sentinel-1 radar images acquired on 17 and 29 April, and 9 April and 3 May. We ignored the possibility of post-seismic deformation over the four and eight days following the event (see Methods). The finite source model (Fig. 2) shows that the rupture propagated eastwards at 3.0±0.5 km s−1 on average (Supplementary Fig. 5). The slip area is about 120 km in length along strike and 50 km in width along dip. The implied moment tensor is nearly identical to the W-phase moment tensor (Fig. 1). Altogether the earthquake released a total moment of 7.2 × 1020 Nm, corresponding to Mw 7.84. The moment rate function shows a simple rupture with a single major pulse of 50 s duration (Fig. 2c). The 12 May aftershock falls in a gap of relatively low slip at the eastward termination of the mainshock. The results from the back-projection and finite source inversion are in remarkable agreement during the first 45 s of the rupture. The moment release rate and the power of the high-frequency sources show the same temporal pattern (Fig. 2c). Both source imaging techniques reveal a unilateral pulse-like rupture with a narrow strip of active slip, 20–30 km wide along strike, propagating eastwards at about 2.7 to 3.0 km s−1 (Fig. 3 and Supplementary Animation).