New High-Resolution Modeling of the 2018 Palu Tsunami, Based on Supershear Earthquake Mechanisms and Mapped Coastal Landslides, Supports a Dual Source

The Mw 7.5 earthquake that struck Central Sulawesi, Indonesia, on September 28, 2018, was rapidly followed by coastal landslides and destructive tsunami waves within Palu Bay. Here, we present new tsunami modeling that supports a dual source mechanism from the supershear strike-slip earthquake and coastal landslides. Up until now the tsunami mechanism: earthquake, coastal landslides, or a combination of both, has remained controversial, because published research has been inconclusive; with some studies explaining most observations from the earthquake and others the landslides. Major challenges are the numerous different earthquake source models used in tsunami modeling, and that landslide mechanisms have been hypothetical. Here, we simulate tsunami generation using three published earthquake models, alone and in combination with seven coastal landslides identified in earlier work and confirmed by field and bathymetric evidence which, from video evidence, produced significant waves. To generate and propagate the tsunamis, we use a combination of two wave models, the 3D non-hydrostatic model NHWAVE and the 2D Boussinesq model FUNWAVE-TVD. Both models are nonlinear and address the physics of wave frequency dispersion critical in modeling tsunamis from landslides, which here, in NHWAVE are modeled as granular material. Our combined, earthquake and coastal landslide, simulations recreate all observed tsunami runups, except those in the southeast of Palu Bay where they were most elevated (10.5 m), as well as observations made in video recordings and at the Pantoloan Port tide gauge located within Palu Bay. With regard to the timing of tsunami impact on the coast, results from the dual landslide/earthquake sources, particularly those using the supershear earthquake models are in good agreement with reconstructed time series at most locations. Our new work shows that an additional tsunami mechanism is also necessary to explain the elevated tsunami observations in the southeast of Palu Bay. Using partial information from bathymetric surveys in this area we show that an additional, submarine landslide here, when simulated with the other coastal slides, and the supershear earthquake mechanism better explains the observations. This supports the need for future marine geology work in this area.

Two global ensemble seismicity models obtained from the combination of interseismic strain measurements and earthquake-catalogue information

SUMMARY Global seismicity models provide scientific hypotheses about the rate, location and magnitude of future earthquakes to occur worldwide. Given the aleatory variability of earthquake activity and epistemic uncertainties in seismicity forecasting, the veracity of these hypotheses can only be confirmed or rejected after prospective forecast evaluation. In this study, we present the construction of and test results for two updated global earthquake models, aimed at providing mean estimates of shallow (d ≤ 70 km) seismicity for seismic hazard assessment. These approaches, referred to as the Tectonic Earthquake Activity Model (TEAM) and the World Hybrid Earthquake Estimates based on Likelihood scores (WHEEL) model, use the Subduction Megathrust Earthquake Rate Forecast (SMERF2), an earthquake-rate model for subduction zones constrained by geodetic strain measurements and earthquake-catalogue information. Thus, these global ensemble seismicity models capture two independent components necessary for long-term earthquake forecasting, namely interseismic crustal strain accumulation and sudden lithospheric stress release. The calibration period for TEAM and WHEEL extends from 1977 January 1 to 2013 December 31. Accordingly, we use m ≥ 5.95 earthquakes recorded during the 2014–2019 period to pseudo-prospectively evaluate the forecasting skills of these earthquake models, and statistically compare their performances to that of the Global Earthquake Activity Rate (GEAR1) model. As a result, GEAR1 and WHEEL are the most informative global seismicity models during the pseudo-prospective test period, as both rank with the highest information scores among all participant earthquake-rate forecasts. Nonetheless, further prospective evaluations are required to more accurately assess the abilities of these global ensemble seismicity models to forecast long-term earthquake activity.

Complex spatiotemporal patterns of intracontinental earthquakes: a different game

<p>Intracontinental earthquakes show complex spatiotemporal patterns. In North China, no large (M>7) earthquakes ruptured the same fault segments in the past 2000 years; instead they roamed among widespread fault systems. In Australia, morphogenic evidence indicates clusters of earthquakes separated by tens of thousands of years of dormancy. In central and eastern United States, paleoseismic studies suggest that large Holocene earthquakes occurred in places that are seismically inactive today. Such seismicity does not fit existing earthquake models that assume steady tectonic loading and cyclic stress release on fault planes. Intracontinental fault systems are widespread and collectively accommodate slow tectonic loading. A major fault rupture both transfers stress to the neighboring faults and perturbs loading conditions on distant faults. Thus, the loading rate on each individual fault can be variable. Slow tectonic loading means that local stress variations from fault interaction or nontectonic processes, or changes of fault strength, could trigger an earthquake. Furthermore, large intracontinental earthquakes usually rupture multiple fault segments or faults, which vary for each event. For these earthquakes, commonly used concepts such as recurrence intervals and characteristic earthquakes, all based on earthquake models assuming cyclic elastic rebound, are inadequate or inapplicable. On the other hand, the general patterns of intracontinental earthquakes can be described by the theory of complex dynamic systems, in which all faults interact with each other. The rupture of individual fault or fault segment cannot be predetermined, but the system behavior can be studied based on the records of previous events. We found that large intracontinental earthquakes, either on a fault system or in a region, are usually clustered and separated by long but variable periods of quiescence. The lengths of the quiescence periods inversely correlate with tectonic loading rates, and the characteristics of earthquake clusters depend on fault geometry and crustal rheology, through fault interaction and viscoelastic relaxation. Spatially, large intracontinental earthquakes are not limited to faults that are active recently, although weak regions tend to have more earthquakes. Intracontinental earthquakes require a different approach, one that focuses on stress interactions between faults in a complex dynamic system rather than stress accumulation and release on individual faults.</p>