Tsunami inundation hazard across Japan

Abstract

Japan faces the world's highest tsunami hazard and risk due to its tectonic environment, high population density and exposure concentration along its coastlines. It is therefore of paramount interest to quantify and differentiate absolute and relative tsunami hazard and risk on a countrywide scale. We quantify tsunami hazard in terms of inundation depth for the entire Japanese coastline with a probabilistic tsunami hazard assessment utilizing the earthquake source information of the 2017 Japanese seismic hazard model. We simulate a stochastic event set for offshore earthquake sources and generate, for each single source with M_W ≥ 7.5, non-uniform finite-fault slip models. We calculate elastic seafloor and landmass deformations that serve as initial conditions to high-resolution numerical modelling of tsunami wave propagation and coastal inundations by solving the non-linear shallow water equation. Variable land surface roughness based on land cover data is used to simulate accurate hydraulics of coastal inundation.

We differentiate tsunami hazard by inundation depth hazard curves and inundation depth return period maps aggregated to city ward polygons as meaningful administrative boundaries. We find mega-thrust events on the subduction interfaces constituting the largest hazard. However, events down to M_W = 7.5 can contribute to substantial hazard in several regions along the coast. In particular for city wards within the Tokyo bay area, we find that earthquakes occurring on the Sagami trough and not the largest mega-thrust events on the Nankai trough, contribute to the highest inundation hazard. Our results also illustrate that tsunami hazard on the western Japanese coast is not negligible.

Introduction

Tsunamis triggered by earthquakes are perceived as infrequent compared to earthquake ground shaking hazard and therefore treated as tail risk by emergency planners and the insurance industry. Rare mega-thrust subduction earthquakes are the major contributors whereas ground shaking hazard and risk have a more diverse set of contributing earthquake sources down to moderate magnitudes. Despite some pioneering works in the early 1980s mainly on Japan [1,2], detailed large-scale Probabilistic Tsunami Hazard Assessments (PTHA) have only recently been developed [[3], [4], [5], [6], [7]]. This is primarily due to the scarcity of large tsunamis, the challenges of assessing tsunamogenic earthquake sources, difficulties in differentiating between ground shaking and tsunami impacts as well as the need for detailed inundation modelling. In this study, we perform a PTHA for the entire Japanese coastline for comparative hazard assessment.

Evaluating tsunami hazard has improved in recent years through the introduction of probabilistic tsunami hazard assessment [3,4,8]. Many local, regional and global applications have been published aiming to improve the understanding of inundation depths, tsunami wave heights, and flooding areas using different approaches [4]. In particular following the 2004 Andaman-Sumatra and the 2011 Tohoku earthquakes, many lessons have been learned about requirements for near-field tsunami assessments in comparison to far-field assessment and countermeasures for tsunami hazard (e.g. Ref. [9,10]). Geist et al. [11] initially pointed to the importance of rupture complexity for the near-field tsunami hazard assessment, the uncertainty of incoming tsunami wave heights and inundation depths due to the rupture complexity. This has been further quantified in detailed studies in several papers mostly following the 2011 Tohoku event [10,[12], [13], [14]].

Japan is prone to tsunami hazard and risk as large portions of the population, commercial and industrial exposure are located close to the coastline. More than 50% of the total population of Asia exposed to tsunami hazard lives in Japan according to Ref. [7]. This specific exposure distribution increases the ratio of tsunami-to-earthquake hazard and risk in comparison to many other countries. For Japan it is essential to understand the tsunami hazard as basis for all avenues of risk assessment, disaster mitigation planning and upscaling community resilience for earthquake and tsunami risk. Several studies for Japan targeted near-field tsunami hazard due to earthquakes [12,[14], [15], [16]], however, a PTHA on a national scale that considers a comprehensive set of tsunamogenic earthquakes is missing.

In this study, we present a PTHA with results for the entire Japanese coast. Our analysis is based on offshore earthquake sources as characterized in the 2017 release of the National Map of Earthquake Prediction of the Earthquake Research Committee (ERC) of the Headquarters for Earthquake Research Promotion (HERP) under the Japanese ministry of Education, Culture, Sport, Science, and Technology. We use the information about earthquake sources and rate models available from the Japanese Seismic Hazard Information Station [[17], [18], [19]]. (J-SHIS, http://www.j-shis.bosai.go.jp/en/). We include one additional source geometry for the Tohoku section to complement possible tsunamogenic sources and substitute time-dependent rates for the largest earthquake sources assuming time-independence following [20].

We differentiate tsunami hazard on a national scale in terms of inundation depth for coastal regions of Japan utilizing cities/city wards located within the first 10 km as a set of nationally recognized administrative boundaries. We focus the PTHA on local earthquake tsunami sources and characterize uncertainty in tsunami generation due to the variability of possible rupture scenarios on causative offshore faults. Sources include subduction interface and normal faulting outer-rise earthquakes with a moment magnitude M_W ≥ 7.5. Each earthquake either is associated with one (M_W < 8.0) or multiple (M_W ≥ 8.0) non-uniform slip distributions on its source geometry. The non-uniform slip-distributions are generated with the method of [21,22]. With these specifications, more than 700 tsunami sources are created for which we model tsunami wave propagation at a 50 m² resolution along the entire Japanese coastline with an inland extent of a few kilometers. The number of earthquake ruptures is limited as the numerical simulation of tsunami waves along the entire Japanese coast is time consuming. We limit the input for tsunami generation to the final slip distributions, though we acknowledge that the temporal evolution of slip on a rupture plane can play a strong role when the initial tsunami waves are generated [23,24].

This PTHA uses only earthquake sources that also cause earthquake shaking hazard, i.e., we disregard tsunamis from other sources around the Pacific rim as well as all other possible sources of tsunamis. The total amount of inundation footprints per location partially limits robust calculations of statistics as not each individual grid cell is sampled with the same amount of inundation values. We treat this shortcoming by using the city wards as boundaries for our analysis so that the data used for our inferences is based on a large enough number of modelled solutions.

Our PTHA aims to model tsunamis that are prone to cause a substantial amount of damage. Thus we dismiss solutions that do not provide inundations depths with a maximum inundation depth of I_max = 0.5 m over larger coastal areas. We introduce this threshold based on engineering considerations and fragility/vulnerability functions developed with data of the 2011 Tohoku event [25,26]. This a priori threshold definition is not ideal, however, necessary to limit the immense computational demand for detailed tsunami-wave propagation and inundation modelling.

Our interest is to quantify recurrence for different thresholds of inundation depth hazard within city wards enabling to differentiate between them as information for potential stakeholders. We identify the major contributors to the inundation hazard and provide insight into which sources are relevant for a set of selected city wards across Japan and in the Tokyo bay area. Our analysis shows that the largest mega-thrust events are mostly, but not always, the key drivers of inundation hazard. In particular for the Tokyo region, the key drivers are events in the Sagami trough, a particularly complicated tectonic triple junction off-shore Tokyo and the source region of large historic events such as the 1703 Genroku and the 1923 Great Kanto earthquake [17,18,27,28].