A large amount of buildings was damaged or destroyed by the 2011 Great East Japan tsunami. Numerous field surveys were conducted in order to collect the tsunami inundation extents and building damage data in the affected areas. Therefore, this event provides us with one of the most complete data set among tsunami events in history. In this study, fragility functions are derived using data provided by the Ministry of Land, Infrastructure and Transportation of Japan, with more than 250,000 structures surveyed. The set of data has details on damage level, structural material, number of stories per building and location (town). This information is crucial to the understanding of the causes of building damage, as differences in structural characteristics and building location can be taken into account in the damage probability analysis. Using least squares regression, different sets of fragility curves are derived to demonstrate the influence of structural material, number of stories and coastal topography on building damage levels. The results show a better resistant performance of reinforced concrete and steel buildings over wood or masonry buildings. Also, buildings taller than two stories were confirmed to be much stronger than the buildings of one or two stories. The damage characteristic due to the coastal topography based on limited number of data in town locations is also shortly discussed here. At the same tsunami inundation depth, buildings along the Sanriku ria coast were much greater damaged than buildings from the plain coast in Sendai. The difference in damage states can be explained by the faster flow velocities in the ria coast at the same inundation depth. These findings are key to support better future building damage assessments, land use management and disaster planning.

In 2011, Japan was hit by a tsunami that was generated by the greatest earthquake in its history. The first tsunami warning was announced 3 min after the earthquake, as is normal, but failed to estimate the actual tsunami height. Most of the structural countermeasures were not designed for the huge tsunami that was generated by the magnitude M = 9.0 earthquake; as a result, many were destroyed and did not stop the tsunami. These structures included breakwaters, seawalls, water gates, and control forests. In this paper we discuss the performance of these countermeasures, and the mechanisms by which they were damaged; we also discuss damage to residential houses, commercial and public buildings, and evacuation buildings. Some topics regarding tsunami awareness and mitigation are discussed. The failures of structural defenses are a reminder that structural (hard) measures alone were not sufficient to protect people and buildings from a major disaster such as this. These defenses might be able to reduce the impact but should be designed so that they can survive even if the tsunami flows over them. Coastal residents should also understand the function and limit of the hard measures. For this purpose, non-structural (soft) measures, for example experience and awareness, are very important for promoting rapid evacuation in the event of a tsunami. An adequate communication system for tsunami warning messages and more evacuation shelters with evacuation routes in good condition might support a safe evacuation process. The combination of both hard and soft measures is very important for reducing the loss caused by a major tsunami. This tsunami has taught us that natural disasters can occur repeatedly and that their scale is sometimes larger than expected.

We present analysis of the December 2018 Anak Krakatau tsunami in Sunda Strait, Indonesia, from a combination of post-tsunami field surveys, bathymetric changes and spectral analysis of the tsunami tide gauge records. Post-tsunami surveys revealed moderate tsunami height along the coast of Sumatra and Java with maximum surveyed runup of 13.5 m and maximum inundation distance of 330 m. At small islands located close to the volcano, extreme tsunami impacts were observed indicating not only a huge tsunami was generated by large amounts of collapse material which caused notable changes of seafloor bathymetry, but also indicates the role of those small islands in reducing tsunami height that propagated to the mainland of Indonesia. Our spectral analysis of tide gauge records showed that the tsunami’s dominant period was 6.6–7.4 min, indicating the short-period nature of the 2018 Sunda Strait tsunami.

Rapid and accurate hazard forecasting is important for prompt evacuations and reducing casualties during natural disasters. In the decade since the 2011 Tohoku tsunami, various tsunami forecasting methods using real-time data have been proposed. However, rapid and accurate tsunami inundation forecasting in coastal areas remains challenging. Here, we propose a tsunami forecasting approach using convolutional neural networks (CNNs) for early warning. Numerical tsunami forecasting experiments for Tohoku demonstrated excellent performance with average maximum tsunami amplitude and tsunami arrival time forecasting errors of ~0.4 m and ~48 s, respectively, for 1,000 unknown synthetic tsunami scenarios. Our forecasting approach required only 0.004 s on average using a single CPU node. Moreover, the CNN trained on only synthetic tsunami scenarios provided reasonable inundation forecasts using actual observation data from the 2011 event, even with noisy inputs. These results verify the feasibility of AI-enabled tsunami forecasting for providing rapid and accurate early warnings.