The tsunami observations produced by the 2018 magnitude 7.5 Palu strike‐slip earthquake challenged the traditional basis underlying tsunami hazard assessments and early warning systems. We analyzed an extraordinary collection of 38 amateur and closed circuit television videos to show that the Palu tsunamis devastated widely separated coastal areas around Palu Bay within a few minutes after the mainshock and included wave periods shorter than 100 s missed by the local tide station. Although rupture models based on teleseismic and geodetic data predict up to 5‐m tsunami runups, they cannot explain the higher surveyed runups nor the tsunami waveforms reconstructed from video footage, suggesting either these underestimate actual seafloor deformation and/or that non‐tectonic sources were involved. Post‐tsunami coastline surveys combined with video evidence and modeled tsunami travel times suggest that submarine landslides contributed to tsunami generation. The video‐based observations have broad implications for tsunami hazard assessments, early warning systems, and risk‐reduction planning.

The slip distribution of the 1 April 2014 Iquique earthquake is obtained by using the least squares inversion of tsunami data at three Deep-Ocean Assessment and Reporting of Tsunamis stations. Most of the slip is concentrated along a 60 km by 40 km slip patch near the hypocenter, with magnitude ranging from 5 to 7 m and a depth of 23 km. The earthquake magnitude from the inversion is estimated as M_w 8.0. The slip distribution is converted into seafloor displacement based on Okada's formula. A nonlinear shallow water equation model is used to simulate tsunami wave propagation, and the simulated water surface elevations are compared with the measured data at 10 tide gauges along the Chilean coast. The agreement is excellent at gauges where the local bathymetry data are complete and the gauges are open to the ocean; otherwise, mismatches of up to 10 min in arrival time and 1.0 m in amplitude are seen.

In this paper, we have conducted a probabilistic tsunami hazard assessment (PTHA) for Hong Kong (China) and Kao Hsiung (Taiwan), considering earthquakes generated in the Manila subduction zone. The new PTHA methodology with the consideration of uncertainties of slip distribution and location of future earthquakes extends the stochastic approach of Sepúlveda et al. (2017). Using sensitivity analyses, we further investigate the uncertainties of probability properties defining the slip distribution, the location, and the occurrence of earthquakes. We demonstrate that Kao Hsiung and Hong Kong would be significantly impacted by tsunamis generated by M_W > 8.5 earthquakes in the Manila subduction zone. For instance, a specific M_W 9.0 earthquake scenario is capable of producing tsunami amplitudes exceeding 4.0 and 3.5 m in Kao Hsiung and Hong Kong, respectively, with a probability of 50%. Despite the significant tsunami impact, great earthquakes have long mean return periods. As a result, the PTHA shows that Kao Hsiung and Hong Kong are exposed to a relatively small tsunami hazard. For instance, maximum tsunami amplitudes in the assessed locations of Kao Hsiung and Hong Kong exceed 0.32 and 0.18 m, respectively, with a mean return period of 100 years. The inundation hazard in populated areas is small as well, with mean return periods exceeding 1,000 years. Sensitivity analyses demonstrate that the PTHA can be affected by the uncertainties of the probability properties defining the slip distribution, the location, and the occurrence of earthquakes. However, PTHA results are most sensitive to the choice of the earthquake occurrence model.

We face a new era in the assessment of multiple natural hazards whose statistics are becoming alarmingly non‐stationary due to ubiquitous long‐term changes in climate. One particular case is tsunami hazard affected by climate‐change‐driven sea level rise (SLR). A traditional tsunami hazard assessment approach where SLR is omitted or included as a constant sea‐level offset in a probabilistic calculation may misrepresent the impacts of climate‐change. In this paper, a general method called non‐stationary probabilistic tsunami hazard assessment (nPTHA), is developed to include the long‐term time‐varying changes in mean sea level. The nPTHA is based on a non‐stationary Poisson process model, which takes advantage of the independence of arrivals within non‐overlapping time‐intervals to specify a temporally varying hazard mean recurrence rate, affected by SLR. The nPTHA is applied to the South China Sea (SCS) for tsunamis generated by earthquakes in the Manila Subduction Zone. The method provides unique and comprehensive results for inundation hazard, combining tsunami and SLR at a specific location over a given exposure time. The results show that in the SCS, SLR has a significant impact when its amplitude is comparable to that of tsunamis with moderate probability of exceedance. The SLR and its associated uncertainty produce an impact on nPTHA results comparable to that caused by the uncertainty in the earthquake recurrence model. These findings are site‐specific and must be analyzed for different regions. The proposed methodology, however, is sufficiently general to include other non‐stationary phenomena and can be exploited for other hazards affected by SLR.