Division of Geolocial and Planetary Sciences

We model seismic and aseismic slip on the Japan megathrust in the area of the Mw 9.0, 2011 Tohoku-Oki earthquake based on daily time series from 400 GPS stations of the GEONET network and campaign measurements of six sea floor displacements. The coseismic and postseismic slip distributions are inverted simultaneously using principal component analysis-based inversion method (PCAIM). Exploring a wide range of boundary conditions and regularization constraints, we found the coseismic slip distribution to be quite compact with a peak slip between 30 and 50 m near the trench. Our model shows deep afterslip fringing the downdip edge of the coseismic rupture but also a dominant zone of shallow afterslip. Afterslip over the first 279 days following the main shock represents about 40% of the coseismic moment. We compare the coseismic and postseismic models with an interseismic coupling model derived from inland and sea bottom measurements determined in a self-consistent manner. Assuming that seismic and aseismic slip had to match the long-term slip rate along the megathrust, the recurrence time of Mw 9.0 earthquakes is estimated to 100–300 years, while historical and paleotsunami records suggest a return period more of the order of 1000 years. The discrepancy is smaller if the shallower portion of the megathrust is assumed
to produce both aseismic slip, as the afterslip model suggests, and seismic slip during occasional large tsunamigenic earthquakes.

We use Interferometric Synthetic Aperture Radar (InSAR), teleseismic body waves, tsunami waveforms recorded by tsunameters, field observations of coastal uplift, subsi-
dence and run-up to develop and test a refined model of the spatio-temporal history of
slip during the Mw 8.0 Pisco earthquake of August 15th, 2007. Our preferred solution
shows two distinct patches of high slip. One patch is located near the epicenter while another larger patch ruptured 60 km further south, at the latitude of the Paracas penin- sula. Slip on the second patch started 60 sec after slip initiated on the first patch. We observed a remarkable anti-correlation between the co-seismic slip distribution and the aftershock distribution determined from the Peruvian seismic network. The proposed source model is compatible with regional run-up measurements and open-ocean tsunami records. From the latter dataset, we identified the 12 min timing error of the tsunami forecast system as being due to a mislocation of the source, caused by the use of only one tsuname- ter located in a non-optimal azimuth. The comparison of our source model with the tsunami observations validate that the rupture did not extend to the trench, and confirms that
the Pisco event is not a tsunami earthquake despite its low apparent rupture velocity
(< 1.5 km/s). We favor the interpretation that the earthquake consists of 2 subevents, each with a conventional rupture velocity (2-4 km/s). The delay between the 2 subevents might reflect the time for the second shock to nucleate or, alternatively, the time it took for afterslip to increase the stress level on the second asperity to a level necessary for static triggering. The source model predicts uplift offshore and subsidence onland with the pivot line following closely the coastline. This pattern is consistent with our obser- vation of very small vertical displacement along the shoreline when we visited the epi- central area in the days following the event. This earthquake represents, to our knowl- edge, one of the best examples of a link between the geomorphology of the coastline and the pattern of surface deformation induced by large interplate ruptures.

We evaluate the effect of coseismic stress changes on the fault slip at midcrustal
depth, assuming a velocity-strengthening brittle creep rheology. We show that this model
can help reconcile the time evolution of afterslip, as measured from geodesy, with
aftershocks decay. We propose an analytical expression for slip of the brittle creeping fault
zone (BCFZ) that applies to any dynamic or static stress perturbation, including shear
stress and normal stress changes. The model predicts an initial logarithmic increase of slip
with time. Postseismic slip rate decays over a characteristic time tr = as that does not t_
depend on the amplitude of the stress perturbation, and it asymptotically joins the long- term creep imposed by interseismic stress buildup t_ . Given that the seismicity rate might be considered proportional to the sliding velocity of the BCFZ, the model predicts a decay rate of aftershocks that follows Omori’s law, with a mathematical formalism identical to that of Dieterich [1994] although based on a different mechanical rationale. Our model also differs from Dieterich’s model in that it requires that aftershock sequences and deep afterslip, as constrained from geodetic measurements, should follow the same temporal evolution. We test this for the 1999 Chi-Chi earthquake, Mw = 7.6 and find that both sets of data are consistent with a model of afterslip due to the response of the BCFZ. The inferred relaxation time tr = 8.5 years requires a value for a = @m/@log(V) (m being the coefficient of friction) in the range between 1.3 10􏰀3 and 10􏰀2.

We use numerical modeling to investigate fault properties that explain key features of the 2011 Mw 9.0 Tohoku-Oki earthquake as well as the overall regional behavior of the NE Japan megathrust. In particular, we study the possibility that slip near the trench resulted from thermal pressurization on a shallow patch of the megathrust, and investigate whether low-velocity friction on that patch is rate- strengthening or rate-weakening. Our models also contain a deeper rate-weakening patch, not prone to efficient thermal pressurization, to account for the moderate-sized interplate seismicity. We produce earthquake sequences and aseismic slip in our models using 2D dynamic simulations that incorporate shear-induced temperature variations and the associated change in pore fluid pressure to capture thermal pressurization. We find that all our models can reproduce more frequent deeper moderate (Mw 7.5) events and less frequent larger events with substantial slip at shallow depth, as observed along the Fukushima–Miyagi segment of the Japan megathrust. However, only the scenario with a sufficiently rate-strengthening patch can match the thousand-year recurrence time of Tohoku-Oki-like earthquakes suggested by the historical and geological record, due to co-existence of seismic and aseismic slip at the shallow depths. This scenario also reproduces other characteristics of the Tohoku-Oki earthquake: the trenchward-skewed distribution of slip, the backward re-rupture of the deeper patch, as well as the weaker radiation at high frequency of the shallower portion of the rupture in spite of its larger slip.