Advanced Utilization of Tsunami Damage Estimation Method Considering Diversity of Earthquakes : Beyond the 2011 Off the Pacific Coast of Tohoku Earthquake

九州大学学術情報リポジトリ Kyushu University Institutional Repository

Advanced Utilization of Tsunami Damage Estimation Method Considering Diversity of Earthquakes : Beyond The 2011 off the Pacific coast of Tohoku Earthquake

大角, 恒雄

https://doi.org/10.15017/4060145

出版情報：九州大学, 2019, 博士（工学）, 課程博士バージョン：権利関係：

Doctoral Dissertation

Advanced Utilization of Tsunami Damage Estimation Method Considering Diversity of Earthquakes

- Beyond The 2011 off the Pacific Coast of Tohoku Earthquake -

January, 2020

Graduate School of Engineering, Kyusyu University

Tsuneo OHSUMI

DEPARTMENT OF CIVIL AND STRUCTURAL ENGINEERING GRADUATE SCHOOL OF ENGINEERING KYUSHU UNIVERSITY Fukuoka, Japan

CERTIFICATE

The undersigned hereby certify that they have read and recommended to the Graduate School of Engineering for the acceptance of this thesis entitled, "Advanced Utilization of Tsunami Damage Estimation Method Considering Diversity of Earthquakes - Beyond The 2011 off the Pacific Coast of Tohoku Earthquake -" by Tsuneo Ohsumi in partial fulfillment of the requirements for the degree of Doctor of Engineering.

Dated: January, 2020

Thesis Supervisor:

Prof. Hemanta Hazarika, Dr. Eng.

Examining Committee:

Prof. Noriaki Hashimoto, Dr. Eng.

Prof. Taiji Matsuda, Dr. Eng.

Contents

Overview 1 Chapter1: Damage related to the 2011 Tohoku Earthquake in the South-central Coastal Area of Iwate Prefecture 5 1. Motivation 5

2. Introduction 6

3. Field investigation for the 2011 Tohoku earthquake 7

3.1 Surveyed area 7

3.2 Characteristics of earthquake ground motion in surveyed area 8

3.3 Tsunami propagation analysis in Surveyed area 13

4. Damage in the South-central Coastal Area of Iwate Prefecture 16

4.1 Touni-Chou (Koshirahama and Hongo district) 16

4.2 Yoshihama 22

4.3 Ryori (Okirai) 26

4.4 Rikuzentakata City 28

5. Tsunami Evacuation of children in Unosumai District, Kamaishi City 32

5.1 Summary of the 2011 Tohoku Earthquake in Unosumai District 32

5.2 Summary of Tsunami Evacuation of Children in Unosumai District 34

5.2.1 Behaviors just after the earthquake occurrence 34

5.2.2 Evacuation to the first temporary evacuation site 35

5.2.3 Evacuation to the second temporary evacuation site 35

5.2.4 Evacuation to the third temporary evacuation site 35

5.3 Tsunami evacuation simulation 35

5.3.1 Setting the initial conditions 36

5.3.2 Setting the parameters 36

5.3.3 Results 40

6. Discussion and Conclusion 42

6.1 Field investigation and numerical investigations 42

6.2 Tsunami Evacuation 43

6.2.1 RES every second and cumulative RES 43

6.2.2 Effects of the students vs. other effects 44

6.2.3 Evacuation time 45

Chapter 2: Beyond the Tohoku Earthquake 49

1. Motivation 49

2. Comprehensive tsunami hazard assessment for Japan 50

3. What is Probabilistic tsunami hazard assessment? 51

4．An Approach to Tsunami Hazard Assessment along the Northeastern Coastal Area

in Japan 54

5. Trial of the hazard evaluation of maximum tsunami inundation flow depth 57

5.1 Probabilistic tsunami inundation flow depth map 58

5.2 Tsunami propagation analysis method 60

5.3 Results of the maximum tsunami inundation flow depth 61

6. A Study on the Utilization of the Tsunami Hazard Evaluation in Japan 63

6.1 Areas surveyed 64

6.2 Survey coverage 65

6.3 Tsunami hazard inventory survey results for municipalities 67

6.4 Current status of tsunami hazard map development 69

6.5 Proposal type hearing for policy formulation on the utilization of the

tsunami hazard evaluation 81

7. Conclusion 87

8. Adopted Recommendations 88

Chapter 3: The Comprehensive Analysis and Evaluation of Offshore Fault Informatics 95

1. Motivation 95

2. Back ground 96

3. Fault Model Database 98

3.1 The fault modelling for the Sea of Japan 99

3.2 The small islands located southeast of the main islands of Japan 103

4. Validity of the Fault Model 105

4.1 The 1940 Shakotan-Oki Earthquake 105

4.1.1 Methodology 107

4.1.2 Fault Model Setting 107

4.1.3 Validity of the Fault Model 107

4.1.4 Comparison with Previous Studies 108

4.1.5 Calculation of Tsunami Propagation Analysis 109

4.1.6 Results 111

4.1.7 Discussion 114

4.1.8 Conclusions 117

4.2 The 1983 Nihonkai–Chubu earthquake 119

4.2.1 Back ground 119

4.2.2 Methodology 120

4.2.3 Fault traces 120

4.2.4 Calculation conditions of the tsunami propagation analysis 125

4.2.5 Validity of the fault model 125

iii

4.2.6 Comparison with Previous Studies 136

4.2.7 Conclusions 140

Chapter 4: Development of a Real-Time Damage Estimation System for Embankment Using Earthquake Early Warning Information 145 1. Motivation 145

2. Introduction 146

3. System Component 147

3.1 Earthquake Early Warning (EEW) 147

3.2 Ground Amplification ratio 149

3.3 Realtime Ground-Motion Monitoring System (Kyoshin Monitor) 149

3.4 Seismic Intensity Prediction Technique 151

3.5 Estimation Method of Embankment Settlement 151

3.6 Digital Japan Portal Web Site Systems 151

3.7 Real-Time Damage Estimation Systems 152

3.8 Information output function 153

4. Conclusion 156

Chapter 5: Investigation of the effects of submarine landslide duration on tsunamis -Related to the 1771 Yaeyama/Meiwa earthquake with tsunami propagation analysis- 159 1. Motivation 159

2. Introduction 160

3. Aseismic tsunamis 162

3.1 Classification of tsunami sources 162

3.2 Water level fluctuation associated with volcanic activity 163

3.3 Water level fluctuation caused by mass movements such as landslides 163

3.4 Water level fluctuation due to falling meteorites 163

4. Classification of submarine landslides 164

5. Tsunami boulders within and around historical tsunami trace points 165

5.1 Ishigaki Island 165

5.2 Miyako Island 183

6. Seabed topographic data 195

6.1 Island shelf points 195

6.2 Analysis section 197

6.3 Submarine landslide points 198

6.4 External force from tsunami propagation analysis 198

6.4.1 Tsunami source model 199

6.4.2 Tsunami propagation analysis 201

6.4.3 Waveform of water level fluctuation amount 202

6.5 Seabed topographic data 203

6.5.1 Crustal structural data based 203

6.5.2 Soil modulus 203

6.5.3 Adhesive strength test 204

6.6 Application of submarine landslide duration using the Newmark method 206

6.6.1 Newmark sliding block method 206

6.6.2 Two-dimensional dynamic interaction analysis code: Super-FLUSH/2D 207

6.7 Analysis of tsunami propagation by submarine landslides 210

6.8 Discussion 216

Chapter 6: Seismicity in Mediterranean Sea and Evaluation for the Tsunami for the AD365 Crete Earthquake 223 1. Motivation 223

2. Back ground 224

3. Crustal movement 226

4. Trace of Upheaval 227

5. Estimating Earthquake Ground Motion 230

5.1 Setting parameters 230

5.2 Estimating the AD 365 earthquake ground motion waveforms 231

6. Estimation of Tsunami Propagation 232

6.1 Setting parameters 232

6.2 Methodology 232

6.3 Results 234

6.4 Comparison with results from previous studies 236

7. Summary and Conclusions 236

Chapter 7: Final Remarks 239

Acknowledgements 242

List of Figures

Chapter 1: Damage related to the 2011 Tohoku Earthquake in the South-central Coastal Area of Iwate Prefecture Fig. 3-1 Map of surveyed areas (GSI Web Map used） 7 Fig. 3-2 Rupture process: Several kinds of rupture processes of the main shock have been proposed in previous studies. 10 Fig. 3-3 The dense recordings of the K-NET strong ground motion networks (NIED) 11 Fig. 3-4 Comparison of the K-NET spectrum (NS) of the north and south direction inside and outside the surveyed areas. 11 Fig. 3-5 The acceleration observation waveform and the Fourier spectrum (NS) of each phase of K-NET: KAMAISHI (IWT007) 12 Fig. 3-6 The acceleration observation waveform and the Fourier spectrum (NS) of each phase of K-NET: OFUNATO (IWT 008) 12 Fig. 3-7 The acceleration observation waveform and the Fourier spectrum (NS) of each phase of K-NET: TONO (IWT 013) 12 Fig. 3-8 Mw 9.2 fault model 14 Fig. 3-9 Snapshots of tsunami propagation 15 Fig. 4-1 Tsunami inundation in the Touni-Chou surveyed area 17 Fig. 4-2 Before and after the disaster at Koshirahama port, Touni-Chou in Kamaishi City 17 Fig. 4-3 Calvert-type coastal levee. 19 Fig. 4-4 GPS (VRS method RTK-GPS) 19 Fig. 4-5 Tsunami run-up height from the coastal levee. 19 Fig. 4-6 Koshirahama port bulletin plate. 19 Fig. 4-7 Touni Sakura tunnel 19 Fig. 4-8 Houses above the inundation area in Hongo village survived intact, while everything below was destroyed by the tsunami. 20 Fig. 4-9 Monument to the Meiji Sanriku tsunami in 1896. The Monument is located in the tsunami inundation height of Meiji Sanriku tsunami. 20 Fig. 4-10 Housings moved uphill after the Meiji Sanriku tsunami in 1896 and the Showa Sanriku tsunami in 1933 20 Fig. 4-11 The difference in house damage to the housings that were moved uphill in the Koshirahama area and those that were not in the Hongo district was significant 21 Fig. 4-12 Damage in the Yoshihama area（Arrow shows tsunami direction） 23 Fig. 4-13 Tsunami inundation in the Yoshihama surveyed area. 23 Fig. 4-14 Houses above the inundation area in Yoshihama village survived intact, while everything below was destroyed by the tsunami. 24 Fig. 4-15 Monument to the Showa Sanriku tsunami in 1933. 24 Fig. 4-16 Housings moved uphill after the Meiji Sanriku tsunami in 1896 and the Showa Sanriku tsunami in 1933 24 Fig. 4-17 "Miracle Yoshihana" and inscription was made as a new monument. 25 Fig. 4-18 State of tsunami inundation in the Ryori (Okirai) surveyed area 26 vii

Fig. 4-19 Comparison the tsunami run-up height and of the Showa Sanriku tsunami in 1933 that of the 2011 Tohoku Earthquake. 27 Fig. 4-20 Tsunami run-up height of the Showa Sanriku tsunami in 1933 Yahata Shrine (arrow shows the road for uphill evacuation) 27 Fig. 4-21 The residents climbed halfway up the stairs, escaping the tsunami. 27 Fig. 4-22 Inscription of a monument in Sugishita, Okirai. 27 Fig. 4-23 State of tsunami inundation in the surveyed area of the city of Rikuzentakata. 28 Fig. 4-24 The Miracle Pine at the site of Takata-Matsubara. 29 Fig. 4-25 Comparison of coast levees and tide gate. 30 Fig. 4-26 Compulsion of foundation and mat foundation by structural damage. 31 Fig. 4-27 The ground sedimentation caused by the earthquake and an inflow of the seawater from the later river in the Kiba village. 31 Fig. 5-1 Location of the schools and the temporary tsunami evacuation sites in Unosumai district 33 Fig. 5-2 Snapshots of RES after the leading student started tsunami evacuation 41 Fig. 6-1 Changes of RES every second and cumulative RES at five site. 46

Chapter 2: Beyond the 2011 Tohoku Earthquake Fig. 4-1 Tsunami hazard assessment evaluation method. 56 Fig. 4-2 Tsunami height simulation analysis of the occurrence probability within the next 30 years. 57 Fig. 5-1 Probabilistic tsunami inundation flow depth map (image) 59 Fig. 5-2 Trial simulation setting flowchart. 60 Fig. 5-3 Simulation result of Maximum tsunami inundation flow depth. 61 Fig. 6-1 Areas in Ibaraki and Chiba Prefectures where the survey on the utilization of the tsunami hazard evaluation was conducted 64 Fig. 6-2 The survey process. 65 Fig. 6-3 Survey items 66 Fig. 6-4 Tsunami hazard map creation (following the Tohoku Earthquake). 70 Fig. 6-5 Survey item 72 Fig. 6-6 Scope of utilization of the tsunami hazard evaluation. 74 Fig. 6-7 Important items relating to the timing and accuracy of real-time information transmitted on tsunamis. 77 Fig. 6-8 Publication of "the tsunami’s arrival time" and "the evacuation distance within __ minutes". 78 Fig. 6-9 Level of content to be disclosed to residents. 79 Fig. 6-10 Formats for disseminating the tsunami hazard evaluation. 79 Fig. 6-11 Mitigation assessment that can be utilized by municipalities to prioritize disaster mitigation measures and to examine priority measures. 81 Fig. 6-12 Utilization of the mitigation assessment by municipalities to prioritize disaster mitigation measures and to assess municipalities’ utilization of priority measures for implementing disaster mitigation measures during a specific year and their business volumes. 83 Fig. 6-13 Tsunami hazard assessment in relation to the study of facilities. 85

viii

Fig. 6-14 Tsunami hazard assessment that enables the development of further detailed tsunami countermeasures. 86

Chapter 3: The Comprehensive Analysis and Evaluation of Offshore Fault Informatics Fig. 2-1 The purpose of the project is to contribute to the hazard assessment of earthquakes and tsunami is in the Japan Sea. 97 Fig. 3-1 Fault modelling for the Sea of Japan. 100 Fig. 3-2 Detail view of fault modelling for the Sea of Japan. (a-d) 101 Fig. 3-3 Detail view of fault modelling for the Sea of Japan. (e-f) 102 Fig. 3-4 Fault modelling for the small islands located southeast of the main islands of Japan. 103 Fig. 3-5 Detail view of fault modelling for the the small islands located southeast of the main islands of Japan (Area 1,2). 104 Fig. 3-6 Detail view of fault modelling for the the small islands located southeast of the main islands of Japan (Area 3). 104 Fig. 4-1 Fault modeling example 109 Fig. 4-2 Comparison of previously published fault models (Okamura, 2005 and Satake, 1986) and linkage fault models of this study for the offshore source area of the Shakotan-Oki Earthquake, Japan 110 Fig.4-3 Maximum tsunami heights determined from tsunami propagation simulations compared with results from previous studies related to the western coast of Hokkaido. 113 Fig. 4-4 Simulated maximum tsunami heights along the western coast of Hokkaido (blue lines) with tsunami traces (red circles) and tsunami heights of combined the maximum cases (gray lines) for each of the models. 113 Fig. 4-5 Large slip region fault model (HKD-2239_C). 116 Fig. 4-6 Example of a survey line cross-section with a change in the tilt angle model. 116 Fig. 4-7 Simulated tsunami heights of the additional models. 117 Fig. 4-8 Fault model used fault trace data (a, b) from observational marine seismic records (JAMSTEC). 122 Fig. 4-9 Primary model in the epicenter of the Nihonkai–Chubu earthquake. 124 Fig. 4-10 Wave source fault models for the Nihonkai–Chubu earthquake. 127 Fig. 4-11 Combination pattern of the examined large slip regions.1 127 Fig. 4-12 Example of a survey line cross-section with a change in the tilt angle model. 128 Fig. 4-13 Comparison of the one/two plats of the tilt angle model. 129 Fig. 4-14 Comparison of maximum tsunami heights and tsunami trace heights in the two tilt angles model. 129 Fig. 4-15 Comparison of maximum tsunami heights and tsunami trace heights in large slip regions 132 Fig. 4-16 Comparison of the proposed model, the Aida (1984) model and the MLIT (2014) model. 134 Fig. 4-18 Calculation area of the tsunami prediction analysis for each mesh size. 138 Fig. 4-19 Maximum tsunami propagation analysis results (T.P. 0 m). . 139 ix

Fig. 4-20 Tsunami traces (red) and tsunami propagation analysis results at the eastern part of the Sea of Japan in the Nihonkai–Chubu earthquake (Recommended model within the red dotted frame). 139

Chapter 4: Development of a Real-Time Damage Estimation System for Embankment Using Earthquake Early Warning Information Fig.3-1 The Earthquake Early Warning system provides advance announcement of the estimated seismic intensities and expected arrival time of principal motion. 148 Fig.3-2 Seismic stations used in the Earthquake Early Warning system. 148 Fig.3-3 Processing procedure of prototype real-time embankment damage estimation system using Earthquake Early Warning. 150 Fig.3-4 Realtime ground motion monitoring system (Kyoshin monitor). 150 Fig. 3-5 Operational procedure of real-time embankment damage estimation system using Earthquake Early Warning. 154 Fig. 3-6 Information disclosure server. 154 Fig. 3-7 Top page. 154 Fig. 3-8 Ground amplification ratio: (Left) entire area and (Right) embankment only. 155 Fig. 3-9 Estimated intensity Map. 155 Fig. 3-10 Estimated settlement. 155 Fig. 4-1 Real-time seismic-intensity exposed-population estimation system 157

Chapter 5: Seabed Landslides and Their Consequences related to the 1771 Yaeyama Earthquake Fig. 5-1 Ryukyu limestones in the Miyara bay, southern Ishigaki Island 166 Fig. 5-2 "Tsunami Ufuishi" at the Sakihara Park of Ohhama. 167 Fig. 5-3 Close up of "Tsunami Ufuishi". There are fossil of corals 167 Fig. 5-4 Natural Monument (Nationally Designated) 168 Fig. 5-5a Ryukyu limestone in Miyara Bay 169 Fig. 5-5b Ryukyu limestone in Miyara Bay 169 Fig. 5-5c Ryukyu limestone in Miyara Bay 169 Fig. 5-6 "Taka Koruseishi " in Ohham in Ishigaki City height: 2.4 m elevation 170 Fig. 5-7 Tsunami boulder at "Matsutou-House" height: 7.7 m elevation 171 Fig. 5-8 Meiwa-Ohtsunami Victim Memorial Monument. 173 Fig. 5-9 Distance from the Monument to the Miyaragawa River ~1 km, from the coastline 1.2 km and altitude: 63.2 m elevation 173 Fig. 5-10 Contents of Inscription panels are shown below; 173 Fig. 5-11 Inscription：According to the "Ohnaminotoki - Kakumurano - Nariyukisyo" 174 Fig. 5-12 Excerpts from records on the Meiwa Great Tsunami disaster 175 Fig. 5-13 Tacolasser Stones located on the north side of this area. 176 Fig. 5-14 The 1771 Yaeyama/Meiwa tsunami's historical tsunami trace and run-up route in the Ishigaki Island 177 Fig. 5-15 A full view of Nagura Bay and the north side of Mt. Panna 177

Fig. 5-16 View of “Amatariya-Suari ” 178 Fig. 5-17 Coral fossils are scattered around in this point 178 Fig. 5-18 Other historical tsunami trace points and current pictures 179 Fig. 5-19 Villages moved to uphill area 182 Fig. 5-20 Tsunami monument at Yonaha-Maehama. 183 Fig. 5-21 Otoiai-Gaki. 183 Fig. 5-22 Ryukyu limestones in the northern eastern Miyako Island with the Ohgami Island (a) and the Maja fisherman’s port (b). 184 Fig. 5-23 Sawada-no-Hama Coral Reef. 185 Fig. 5-24 Shimojijima Monolith (Obi-Iwa) 187 Fig. 5-25 Eastern-Cliff 189 Fig. 5-26 Other history tsunami trace points and current pictures. 190 Fig. 5-27 Brid view of the survey area of the Motojima Ruins from Miyakuni. 193 Fig. 5-28 Motojima villages moved to uphill area. 193 Fig. 5-29 Uipya-Yama Ruins with information plate. 194 Fig. 6-1 Seabed topography of the area around the Nansei Islands produced by the Red Relief Image Map visualization method 196 Fig. 6-2 Lip–surface conduits in the slope direction produced by the Red Relief Image Map visualization method. 196 Fig. 6-3 Topographic cross section of the current seabed. 197 Fig. 6-4 Model domain areas around Yaeyama Islands. 199 Fig. 6-5 Tsunami rupture models (load cases). 200 Fig. 6-6 Waveforms and Fourier spectra from flow rate. 202 Fig. 6-7 Acceleration to make an external force of the dynamic load: (Case1) 208 Fig. 6-8 Acceleration to make an external force of the dynamic load: (Case2) 208 Fig. 6-9 Analysis model 208 Fig. 6-10 Application of submarine landslide duration determined using the Newmark method 209 Fig. 6-11 Historical tsunami trace points 211 Fig. 6-12 Numerical results for propagation of the tsunami 212 Fig. 6-13 Numerical results for propagation of the tsunami 212 Fig. 6-14 Numerical results for propagation of the tsunami 213 Fig. 6-15 Close up of time history of sliding residual displacement [m] 215 Fig. 6-16 Comparison of historical tsunami traces and presentation results 215 Fig. 6-17 Selection of parameters with epistemological uncertainty. 219

Chapter 6: Seismicity in Mediterranean Sea and Evaluation of the Tsunami for the AD365 Crete Earthquake Fig. 2-1 Epicenter of the AD 365 Crete Earthquake 225 Fig. 3-1 Contours of upheaval in Crete 226 Fig. 4-1 Crustal displacements and upheaval generating areas 229 Fig. 4-2 Slip distribution (left) and crustal displacements and upheaval generating areas (right) 229 Fig. 5-1 Estimated synthetic velocity waveforms from the AD 365 earthquake 231 Fig. 5-2 Velocity response spectra comparison for the synthetic waveforms 231 Fig. 6-1 Area of the tsunami propagation simulation 233

Fig. 6-2 Snapshots of tsunami propagation simulations for the whole study area, 1,350 m mesh 234 Fig. 6-3 Snapshots of tsunami propagation simulations for Alexandria, 450 m mesh 235

xii

List of Tables

Chapter 1: Damage related to the 2011 Tohoku Earthquake in the South-central Coastal Area of Iwate Prefecture Table 3-1 Calculation conditions of tsunami propagation analysis 14 Table 3-2 Comparison of the surveyed tsunami heights and the simulated maximum tsunami heights [m] 14 Table 5-1 Main parameters and conditions in the simulation 39 Table 5-2 Parameters about the information sources creating RES 39 Table 5-3 Parameters about the velocity 39

Chapter 3: The Comprehensive Analysis and Evaluation of Offshore Fault Informatics Table 4-1 Fault parameters. 108 Table 4-2 K-κ values of the simulated maximum tsunami heights and observational records. 112 Table 4-3 Fault parameters of the additional models. 112 Table 4-4 Simulation results of the additional models (reliable at levels A, B, C, and D at 49 points). 115 Table 4-5 Parameters of initial fault model. 122 Table 4-6 Comparison of the previous models of the Nihonkai–Chubu earthquake with the model used in this study 123 Table 4-7 Calculation conditions of tsunami propagation analysis. 125 Table 4-8 Fault models by using reliability of tsunami trace with K-κ value. 137

Chapter 5: Seabed Landslides and Their Consequences related to the 1771 Yaeyama Earthquake Table 3-1 Sources of aseismic tsunami events 162 Table 5-1 Motojima villages moved to uphill area 192 Table 6-1 Details of tsunami propagation simulation. 201 Table 6-2 Soil modulus. 204 Table 6-3 Historical tsunami traces and presentation results 213 Table 6-4 Historical tsunami trace and presentation results - Doubly extended area in the eastward direction - 214 Table 6-5 Historical tsunami trace and presentation results - Slope failure duration: 200 s - 214

Chapter 6: Seismicity in Mediterranean Sea and Evaluation of the Tsunami for the AD365 Crete Earthquake Table 4-1 Parameters used in the fault model. 228 Table 5-1 Parameters used in the stochastic Green’s function method. 230 Table 6-1 Details of the tsunami propagation simulation 233 Table 6-2 Maximum tsunami heights at representative evaluation points and duration between the earthquake and arrival of different wave heights at each point. 234

xiii

xiv

Overviews

More than 8 years have passed since the 2011 off the Pacific coast of Tohoku Earthquake (the 2011

Tohoku earthquake). The 2011 Tohoku earthquake and the tsunami generated by it caused widespread damage in eastern Japan. Various studies and countermeasures are currently being conducted because of the serious damage that resulted from the tsunami. However, only earthquakes that caused the highest recorded unexpected damage have been studied. There has not been enough investigation into the effects of tsunamis caused by a diversity of earthquakes. In other words, it is necessary to evaluate the danger of the tsunamis that hit various points along the coast as a result of various levels of earthquakes. From such a viewpoint, this study concluded that it was necessary to indicate the tsunami height along the coast as part of probabilistic hazard risk assessments.

In the 2011 Tohoku earthquake, more than 20,000 people were killed by the tsunami, despite the fact that the tsunami arrived more than 30 minutes after the occurrence of the earthquake. A delay in evacuation to uphill areas, resulting from not making use of past lessons learned, was a problem. In this study, to understand the impacts of the earthquake, this study first investigated the tsunami damage to the southern coast of Iwate Prefecture. This study clarified where past lessons learned were used and where repeated losses of life occurred. In addition, this study conducted detailed investigation of a successful evacuation known as the “Kamaishi Miracle.” Using numerical simulation, this study demonstrated the effect of using lessons learned. Furthermore, this study conducted an inventory survey directly with local government officials involved in disaster prevention, and gathered opinions. There is concern regarding future tsunami vulnerability resulting from embankment damage by the earthquake.

Both subduction-zone earthquakes, like the 2011 Tohoku earthquake, and active fault earthquakes occur in the sea surrounding Japan. Submarine seismogenic faults have generated earthquakes such as the 1940 earthquake off the coast of Kamui (Shakotan-Oki earthquake) and the 1983 Sea of Japan 1 earthquake, and resulted in tsunami casualties. Therefore, this study used modeling to verify the causal relationship between the fault and historical earthquakes. As an example of modeling the submarine landslide and verifying such causal relationships, this study examined a new method of reproducing the 1771 Meiwa/Yaeyama earthquake, whose tsunami is said to have caused the deaths of 10,000 people. This study also established the analysis process. Furthermore, for the AD 365 earthquake off the coast of Crete (Greece), this study conducted a survey of the 9-m ground upheaval, identified fault parameters from previous studies, and established a process for tsunami reproduction.

The results of the research are described in Chapters 1 through 7 as follows.

Chapter 1 surveys the impacts of the 2011 Tohoku earthquake on the southern coast of Iwate

Prefecture, and shows whether past earthquakes and tsunamis were used as lessons to facilitate evacuations. This study identify the current situation and discuss problems associated with relocation of evacuees to uphill areas. In this earthquake, many people could not or did not evacuate in time to escape from the tsunami. However, junior high school and elementary school students in Kamaishi

City did start tsunami evacuation quickly, and 99.8% of students were saved. This is referred to as the “Kamaishi Miracle.” The study simulates and discusses how their quick initiation of tsunami evacuation to surrounding areas affected survival rates compared with other areas.

In Chapter 2, this study describe the spatial distribution of tsunami hazards in coastal areas with a similar scale to the southern coast of Iwate Prefecture. This study use various hazard risk information to conduct probabilistic hazard risk assessments in the fields of engineering and non-life insurance.

The tsunami hazard inventory survey was performed in municipalities in Ibaraki and Chiba prefectures affected by the 2011 Tohoku earthquake. Problems were identified, and the results were proposed as a tool for tsunami hazard risk assessment. Thus, it is expected that the new tsunami hazard risk information tool will be used to improve the disaster mitigation capability of the region. 2

In Chapter 3, this study develop a comprehensive set of fault models around Japan that can be used for earthquake motion and tsunami prediction calculations. These were constructed from the fault distribution in the sea around Japan. In particular, for faults that could generate an earthquake exceeding magnitude 7, this study verify prediction calculations based on historical data and the fault model. The study uses the linkage corresponding to the closest fault trace in the epicenters of the 1940

Shakotan-Oki and the 1983 Sea of Japan earthquakes. It also verifies tsunami trace and geodetic data from tsunami traces found in the Japan Tsunami Trace Database.

In Chapter 4, this study link a real-time damage estimation system for embankments with an emergency earthquake bulletin. For linear structures such as embankments, it takes time to identify the affected area immediately after an earthquake. In addition, it is difficult to identify the affected area in advance because the ground motion changes according to the characteristics of the earthquake.

On the basis of assumed disaster areas, the proposed application could be used to select areas of priority for surveying and to identify evacuation routes immediately following an earthquake. By assuming arbitrary hypocenters and estimating the potential damage by virtual earthquakes, river administrators could consider necessary emergency measures in advance.

In Chapter 5, this study establish a method for reproducing tsunamis caused by submarine landslides. The simple technique applied by the plasticity of an experiment and the numerical analysis model, and the predictive method by the exercise dynamic submarine landslides model (kinematic model), has been used, with application for permission, for the nuclear power plant establishment.

However, there are many uncertain factors. Furthermore, using a calculated slope failure duration, a verification was conducted for the 1771 Meiwa/Yaeyama tsunami.

In Chapter 6, this study estimate tsunami propagation for the 365 Crete earthquake. In this study, the analysis parameters were defined from various papers, and the traces that still remain on the site, and the traditions about earthquakes and tsunamis in Greek mythology that were learned as lessons.

As a result, this study calculated that a 2.4-m tsunami had arrived 100 minutes later in the Nile delta area of Alexandria, Egypt, where it had caused many casualties. By reproducing historical earthquakes and tsunamis, the lessons learned have resulted in measures and preparedness for a major earthquake that is expected to occur in the near future.

Chapter 7 summarizes the results obtained in this thesis as well as future issues and prospects.

Chapter 1 Damage related to the 2011 Tohoku Earthquake in the South-central Coastal Area of Iwate Prefecture

1. Motivation

More than 8 years have passed since The 2011 off the Pacific Coast of Tohoku Earthquake

(hereinafter, the 2011 Tohoku Earthquake). Widespread damage was caused in eastern Japan by the earthquake and tsunami generated by the 2011 Tohoku Earthquake. A large tsunami struck the coastal area of eastern Japan and caused damage to buildings, breakwaters, tide embankments and river levees. The joint reconnaissance team of the Tohoku and Shikoku branches of Japanese Geotechnical

Society investigated the geotechnical damage in the south-central coastal area of Iwate Prefecture at from beginning of April to September 2011. This report summarizes the geotechnical hazards and the damage to port structures, roads, railways, river levees and buildings caused by the earthquake motion and tsunami in 2011 Tohoku Earthquake in the south-central coastal area of Iwate Prefecture. Major investigating areas are the city of Kamaishi (Koshirahama Port, Touni-Cho), the coastal area of the city of Ofunato (Sanriku-Chou Yoshihama, Sanriku-Cho Ryori), and the city of Rikuzentakata.

In the 2011 Tohoku Earthquake, many people could not or did not evacuate from the tsunami.

However, students under control of junior high schools and elementary schools Miracle”. This study focused on the tsunami evacuation of children in Unosumai district, Kamaishi City. This study simulated and discussed the effect of their quick start of tsunami evacuation on surrounding areas.

Keywords— 2011 Tohoku Earthquake, tsunami, port structures, Iwate, Kamaishi , Koshirahama Port, Touni-Cho, Ofunato, Sanriku-Chou, Yoshihama, Ryori, Rikuzentakata

2. Introduction Japan has experienced several huge disaster events in the last decade, including the Tohoku

Earthquake, which occurred on March 11, 2011. This report considers this event and the damage resulting from it according to the following questions: could a lesson of previous disasters be utilized?

Why couldn't sustained for the housings moved uphill in Touni? What is success of the housings moved uphill in Yoshihama? How was it possible to save the life of the senior persons in Okirai? In

Koshirahama port, the coastal levee was built at a total cost of 1042 million yen. However, the information plate at the port is enlightening; it states that “there is no defense superior to evacuation”.

In the 2011 Tohoku Earthquake, many people could not or did not evacuate from the tsunami.

However, students at junior high and elementary schools evacuated promptly and voluntarily, in accordance with their experiences of evacuation drills. The students decided to evacuate further to higher ground based on their own observation of the situation. This is known as the “Kamaishi

Miracle”. This study focuses on the tsunami evacuation of children in Unosumai district, Kamaishi

City. This study simulates and discusses the effect of their prompt tsunami evacuation on surrounding areas.

3. Field investigation for the 2011 Tohoku earthquake

3.1 Surveyed area The study area is shown in Fig. 3-1 Major investigating areas are the city of Kamaishi

(Koshirahama Port, Touni-Chou), the coastal area of the city of Ofunato (Sanriku-Chou Yoshihama,

Sanriku-Cho Ryori), and the city of Rikuzentakata, all areas damaged by the tsunami.

Kitakami

Kamishi

Touni

Yoshihama

Okirai

Rikuzenta

Fig. 3-1 Map of surveyed areas (GSI Web Map used）

3.2 Characteristics of earthquake ground motion in surveyed area

Previous studies have proposed several kinds of rupture processes for the main shock. Some were estimated using teleseismic waveforms, and others were obtained by measuring strong motions.

Although they focused on different frequencies or used different methods, every rupture model shows complex rupture propagations (Suzuki et al., 2011) [1] (Fig. 3-2). This area has recorded seismic waves of the main shock and its aftershocks.

Dense recordings from the K-NET strong ground motion network for the 2011 Tohoku

Earthquake indicate the occurrence of strong ground motion inside and outside of the surveyed areas.

Fig. 3-3 shows a comparison of the K-NET spectrum (NS) of the north and south direction inside and outside of the surveyed areas.

At the three northern sites (K-NET: OFUNATO (IWT 008), K-NET: KESENNUMA (MYG 001),

K-NET: TSUKIDATE (MYG 004)), the acceleration spectrum (NS) shows dominant frequencies around 3–10 Hz (Fig. 3-4; left side). However, the acceleration spectrum (NS) at the two southern sites ((K-NET SENDAI (MYG 013), K-NET HARAMACHI (FKS 005)) shows dominant frequencies around 0.6–2 Hz (Fig. 3-4; right side). It can be seen that a shift was recorded, with wavelength increasing from the north to the south. The acceleration observation waveform and the

Fourier spectrum (NS) of each phase of K-NET: KAMAISHI (IWT007) and K-NET: OFUNATO

(IWT 008) are shown in Fig. 3-5 and 3-6, respectively. The maximum acceleration appears in the first phase, the spectrum dominant short-period components, and the spectrum shapes are almost equal. However, in the observation record (NS) at the inland observation site K-NET: TONO (IWT

013), the high frequency component of 2 Hz is dominant (Fig. 3-7).

Furumura et al. (2011) [2] determined the main shock rupture process of the 2011 Tohoku

Earthquake using the acceleration record obtained from K-NET and KiK-net (K-NET, 2011).

According to their study, the first rupture occurred off Miyagi prefecture, and strong seismic waves

were released all over Tohoku (first phase). After several tens of seconds, another massive rupture 8

occurred, and strong seismic waves were released (second phase). The third rupture occurred offshore near northern Ibaraki, and strong seismic waves were radiated towards Ibaraki prefecture

(third phase). Fig. 3-5 shows the acceleration records of the main shock recorded at K-NET

IWT007 station (at Kamaishi), which recorded a maximum acceleration of 741.6 Gal. Fig. 3-6 shows the acceleration records of the main shock recorded at K-NET IWT008 station (at Ofunato), which recorded a maximum acceleration of 387.0 Gal. In both instances, the continuation time was over 220 seconds.

The Tohoku area recorded many seismic waveforms of the 2011 Tohoku Earthquake and its aftershocks. Several kinds of rupture processes of the main shock have been proposed using teleseismic waveforms (e.g., Ide et al., 2011 [3], Ammon et al., 2011 [4]) or strong motions (e.g.,

Furumura et al., 2011 [2], Kurahashi and Irikura, 2011 [5]). It is common that every rupture model shows complex rupture propagation though the results vary according to the focused frequencies or the analysis methods.

Fig. 3-2 Rupture process: Several kinds of rupture processes of the main shock have been proposed in previous studies. Some of them were estimated by teleseismic waveforms, and others were obtained by strong motions (Courtesy of Wataru Suzuki).

Fig. 3-3 The dense recordings of the K-NET strong ground motion networks (NIED)

Fig. 3-4 Comparison of the K-NET spectrum (NS) of the north and south direction inside and outside the surveyed areas.

Fig. 3-5 The acceleration observation waveform and the Fourier spectrum (NS) of each phase of K- NET: KAMAISHI (IWT007)

Fig. 3-6 The acceleration observation waveform and the Fourier spectrum (NS) of each phase of K- NET: OFUNATO (IWT 008)

Fig. 3-7 The acceleration observation waveform and the Fourier spectrum (NS) of each phase of K- NET: TONO (IWT 013)

3.3 Tsunami propagation analysis in Surveyed area

In the Tsunami Hazard Assessment [6], 1,890 characterized source models were specified, which based on the Tsunami prediction method for earthquake with specified source faults ("Tsunami

Recipe") [7]. In the model, using the model with Mw 9.2, the coastal tsunami height at the representative point of damage in the south-central coastal area of Iwate Prefecture. Using the model with Mw 9.2, the coastal tsunami height was simulated maximum tsunami heights at the representative point of damage in the south-central coastal area of Iwate Prefecture. The purpose of the characterized source model is used for the probabilistic tsunami hazard evaluation. This study tried to recurrent with the tsunami trace by the characterized source model could be recurrentend by the characterized source model used in the probabilistic tsunami hazard assessment.

The tsunami heights calculated by tsunami propagation analysis. Table 3-1 shows the calculation conditions of tsunami propagation analysis ． Fig.3-8 shows the fault zone (yellow area), characterized source model consisting of a large slip (orange area) and a ultra large slip (red area) in the fault zone. Table 3-2 shows comparison of the surveyed tsunami heights and the simulated maximum tsunami heights at Toni, Yoshihama and Okirai surveyed points. Fig.2-9 shows the snapshots of the tsunami propagation.

At Toni survey point, the tsunami run-up height was 18.2 m according to the GPS (VRS method

RTK-GPS). At Yoshihama and Okirai survey points, the tsunami run-up heights were 16 to 17 m and 17.8 m, respectively, according to the Total Station (TCR407S: Leica Geosystems). The surveyed points on the land and the simulated points on the coast are different. However, an error of 5.6 m at Touni surveyed point, an error of 2.7 m to 3.7 m at Yoshihama surveyed point, an error of 0.3 m at Okirai surveyed point.

Table 3-1 Calculation conditions of tsunami propagation analysis Governing equation Non-linear long wave theory Numeral solution Staggered Leap-frog

Calculation area Coast of the southern part of Hokkaido and the eastern part of Sea of Japan, offshore northern Honshu Size calculation area From the open ocean, 1350 m, 450 m, 150 m (Minimum 150 m) Boundary condition Considering tsunami run up in the land area Transmission border nonreflective in the sea side Structures not consider Calculation time 12 hours Initial water level Sea bed movement calculated by Okada (1992) [8] Sea water level T.P. 0 m Censored water depth 10-2 m Roughness coefficient 0.025

Table 3-2 Comparison of the surveyed tsunami heights and the simulated maximum tsunami heights [m]

Tonui Yoshihama Okirai

Surveyed tsunami heights [m] 18.20 16 ~ 17 17.13 Simulated maximum 12.63 13.31 16.79 tsunami heights [m] Error 5.57 2.69 ~ 3.69 0.34 Absolute value [m]

Fig.3-8 Mw 9.2 fault model. Fault zone (blue area), characterized source model consisting of a large slip (green area) and an ultra large slip (yellow area). The levels of the color show slip value.

Fig.3-9 Snapshots of tsunami propagation.

4. Damage in the South-central Coastal Area of Iwate Prefecture Site investigations and laboratory soil investigations were performed on two geotechnical structures (a damaged river dike and an undamaged tire retaining wall), that were located in the central and southern part of Iwate prefecture. The investigation results showed that subsidence-related failure probably had taken place in the dike body with low penetration resistance. However, in the case of the retaining wall made of tires, the flexibility and resiliency of the tires prevented any damage to the wall from scouring by the tsunami.

4.1 Touni-Chou (Koshirahama and Hongo district)

In Koshirahama port (Touni-Chou in Kamaishi-City), the tsunami washed away a residential area

following the collapse of the tide breakwater, causing serious damage. However, houses located

higher than the embankment of T.P. ;Tokyo Peil + 17.5 m were less affected by the earthquake

motion and the tsunami, as shown in Fig.4-1. The conditions at the site before and after the disaster

are shown in Figs. 4-2 (a) and (b). According to interviews with residents, “the tremor caused by

the earthquake continued for about 2 minutes, and the small first wave of the tsunami hit

immediately. The third wave was a tide from the breakwater to the offshore, it exceeded the

breakwater and the tide ridge.”

At Touni Elementary School, located about 300 m from the coast, watermarks from the tsunami

were confirmed up to the 3rd floor, where the elevation was 17.09 m according to the RTK-GPS

measurement.

In Koshirahama port, a section of the culvert-type coastal levee (height T.P. +11.3 m) partially

collapsed, and houses above the inundation area were damaged (Fig. 4-3). The tsunami run-up

height was 18.2 m according to GPS (VRS method RTK-GPS: Fig. 4-4). Fig. 4-5 depicts the side

of a house showing the tsunami run-up height from the coastal levee. Tsunami height permeated the

lintel (Kamoi) and reached the houses further up the hill. The house had been moved after the Meiji

Sanriku tsunami in 1896.

Fig. 4-1 Tsunami inundation in the Touni-Chou surveyed area.

Tsunami inundation areas

(a) Before the disaster (August 2003) (b) After the disaster (June 6 2011) Fig. 4-2 Before and after the disaster at Koshirahama port, Touni-Chou in Kamaishi City (photo taken by Ohsumi).

Fig. 4-6 shows the Koshirahama port information plate，which is located in Touni. A summary of the construction of the Culvert-type coastal levee is shown on this plaque. The coastal levee was built at a total cost of 1042 million yen. The construction took place between 1979 and 1990. This board states that the important thing is "there is no defense superior to evacuation".

The tsunami passed through Touni Sakura tunnel (opened 2006), which linked the Hongo district to the Koshirahama district, in Kamaishi-City. In the Hongo district, the tsunami overtook the coastal levee and the road embankment.

The tsunami run-up height was 20 m in this district. In the village, houses above the inundation area survived intact, while everything below was destroyed by the tsunami (Fig. 4-8).

There was a monument to the Meiji Sanriku tsunami in 1896. However, its inscription was marred by this later tsunami (Fig. 4-9).

According to tsunami history around this area (Shuto, et al., 2011a) [6] (Figs. 4-10 and 4-11), numerous forest fires occurred and spread to housings in the low areas. These housings had been moved uphill again after the Showa Sanriku tsunami in 1933. However, stores began to open in the low-lying areas two years later, and the village had been rebuilt by 1945. In addition to a forest fire, the following items are pointed out (Koshimura, 2005) [7]; the relocated places were too far from the sea ; there was a shortage of drinking water at the relocation village; the road was inconvenient; many relocated residents were preoccupied with their ancestral land; there was opportunity for good fishing at the beach, which led to temporary cabins becoming permanent housing; a large-scale fire occurred and the housing was destroyed, and many families that had never experienced a tsunami were migrating from the mountain area. n the Hongo district, the housings were moved uphill after the Showa Sanriku tsunami in 1933. This area was not damaged by this event. There is a significant difference in the amount of damage to the housings in the Koshirahama area that were moved uphill, from those in the Hongo district that were not.

Fig. 4-3 Calvert-type coastal levee. Fig. 4-4 GPS (VRS method RTK-GPS) (height T.P.+11.3m)

“There is no defense superior to evacuation.”

Fig. 4-5 Tsunami run-up height from Fig. 4-6 Koshirahama port bulletin plate. the coastal levee.

(a)Koshirahama side (b) Hongo side Fig. 4-7 Touni Sakura tunnel

(Fig.4-3 to 4-7: photo taken by Ohsumi, May 2, 2011).

Fig. 4-8 Houses above the inundation area Fig. 4-9 Monument to the Meiji Sanriku in Hongo village survived intact, tsunami in 1896. The Monument while everything below was is located in the tsunami destroyed by the tsunami. inundation height of Meiji Sanriku tsunami.

(Fig.4-8 to 4-9: photo taken by Ohsumi, May 2, 2011)

Fig. 4-10 Housings moved uphill after the Meiji Sanriku tsunami in 1896 and the Showa Sanriku tsunami in 1933. (after Shuto, et al., 2011b) [9]

Fig. 4-11 The difference in house damage to the housings that were moved uphill in the Koshirahama area and those that were not in the Hongo district was significant. (after Shuto, et al., 2011a) [9]

4.2 Yoshihama Yoshihama district has repeatedly suffered tsunami damage during big earthquakes in the past

(Iwate Nippo, 2011). Due to strong shaking of the earthquake and the tsunami that followed, the coastal dike at the mouth of Yoshihama River was damaged over a wide range with complete collapse of the concrete slabs. In some parts, the concrete slabs were observed to have been displaced by more than 30 m (Fig. 4-12 to 4-13). Unlike many other Tohoku-region coastal levees destroyed by the tsunami (Hara et al., 2012 [11], Hazarika et al., 2011 [12]), the damage to this dike in Yoshihama was not due to scouring or erosion. It is worth mentioning here that during the first phase of our investigation (May 3, 2011), the paddy fields behind the coastal levees were completely covered with deposited sands from the seashore, and the inundated waters extended over a wide area. Our survey indicated the crown height of non-damaged river dike was +5.2 m according to

Total Station (TCR407S: Leica Geosystems).

Fig.4-14 shows that houses above the inundation area in Yoshihama village remained intact, while everything below was destroyed by the tsunami. There was a monument to the Meiji Sanriku tsunami in 1896 (Fig. 4-15). The monument is located at the tsunami inundation height of the Showa

Sanriku tsunami.

According to tsunami history around this area (after Shuto, et al., 2011) [9] (Fig. 4-16), housings moved uphill after the Meiji Sanriku tsunami in 1896. These housings had been moved uphill again after the Showa Sanriku tsunami in 1933. These housings had been continued living uphill village. As a result, the 2011 Tohoku earthquake resulted in minimal damage (only 1 dead person, 2 households damaged). This area was praised as "Miracle Yoshihana" and an inscription was made as a new monument were made as shown in Fig. 4-17.

(a) Washout of dike (b) Collapse of coastal levee and inundation of field

Fig.4-12 Damage in the Yoshihama area（Arrow shows tsunami direction）

(photo taken by Ohsumi, May 3, 2011)

Yoshihama River

Damaged river dike

Fig. 4-13 Tsunami inundation in the Yoshihama surveyed area.

Fig.4-14 Houses above the inundation Fig.4-15 Monument to the Showa area in Yoshihama village Sanriku tsunami in 1933. The survived intact, while Monument is located in the everything below was tsunami inundation height of destroyed by the tsunami. Showa Sanriku tsunami.

(Fig.4-14 to 4-15: photo taken by Ohsumi, May 3, 2011)

Fig.4-16 Housings moved uphill after the Meiji Sanriku tsunami in 1896 and the Showa Sanriku tsunami in 1933. (after Shuto, et al., 2011a) [13]

1. The Village of Miracles 2. Living in Yoshimi 3. Every man for himself “Tsunami Cherish your bonds and live with I want to tell everyone in the Tendenko” hope world There was a painful thing, a sad Protecting pushing of the There will be a big earthquake in thing, Meiji, Showa predecessor and living life alive the future and a big tsunami will There was extraordinary effort uphill I'm sure to come again tsunami come move Love! To the mountain to uphill Run! To Uphill! Predictive feat the feat of everyone To protect the lives of everyone in the If you act in time else village Run away! To a high place Run away with no body! To protect Beautiful sea blue sea my hometown You can do anything if there is a your own life Like the spiral staircase tellers talk life to cherish the present Let's pass it on! A thousand years with everyone later Surely also in the next tsunami Living in Miraculous Settlement at Fig.4-17 "Miracle Yoshihana" and inscription was made as a new monument. (photo taken by Ohsumi, September 18, 2014)

4.3 Ryori (Okirai)

Sanriku-Chou Ryori is located in the coastal area of Ofunato. The tsunami run-up height of the

1933 Tohoku Earthquake tsunami is marked by a monument, depicted in (Fig.4-22). Fig.4-18 depicts the state of tsunami inundation in the Ryori (Okirai) surveyed area. At the Yahata Shrine, the Showa Sanriku tsunami reached a run-up height of 7.80 m (Fig.4-19). The yellow arrow shows the road used for uphill evacuation during the 2011 Tohoku earthquake. There is the “Sanriku great king cedar” that is ca.7,000 years old in the Yahata Shrine. According to an interview with resident

Yoshinori Sugishita in May 2011, concerned about the tsunami run-up height, the residents judged the situation of the tsunami and decided to move uphill with elderly family members. The tsunami run-up height was 16.8 m according to Total Station (TCR407S: Leica Geosystems). The residents went up as shown by the yellow arrow on Fig.4-20 and climbed halfway up the stairs, escaping from the tsunami (Fig.4-21). According to the residents, the tsunami advanced about six times and they were saved by staying on the stairs until nightfall.

Tsunami inundation area

Fig.4-18 State of tsunami inundation in the Ryori (Okirai) surveyed area

Fig.4-19 Comparison the tsunami run-up height Fig.4-20 Tsunami run-up height of the and of the Showa Sanriku tsunami in 1933 Showa Sanriku tsunami in that of the 2011 Tohoku Earthquake. 1933 Yahata Shrine (arrow shows the road for uphill evacuation)

Fig.4-21 The residents climbed halfway up the stairs, escaping the tsunami. The tsunami advanced about six times and they were saved by staying on the stairs until nightfall.

(Photo taken September 16, 2014)

After a major earthquake, a tsunami comes. If there is an earthquake, get together on high ground. If you are chased by the tsunami, go anywhere uphill. Even if you run far away, the tsunami will chase you so keep a nearby uphill area in mind in advance. Do not build a house lower than the level designated for residential areas by the prefectural government.

Fig.4-22 Inscription of a monument in Sugishita, Okirai.(photo taken by Ohsumi, May 4, 2011) 27

4.4 Rikuzentakata City

Iwate Prefecture reported that following the tsunami the number of deaths stood at 1,556, number of missing at 217 and number of injured unknown as of 28 February 2013 in the city of

Rikuzentakata. In Hirota bay, the tsunami trace was 18.3 m and it easily entered and inundated the plain area (Fig. 4-23).

Miracle Pine

Tide gate Takata-Matsubara

Kesen river

Hirota bay

Tsunami inundation areas

Fig. 4-23 State of tsunami inundation in the surveyed area of the city of Rikuzentakata.

In a 2 km section of lagoon and the border area of the Gulf of Hirota there was a tsunami control forest,

Takata-matsubara, comprising 70,000 Japanese umbrella and black pine. Apart from the Miracle Pine at the site of Takata-Matsubara (Fig. 4-24), the majority of this forest was washed away by the tsunami.

Fig. 4-24 The Miracle Pine at the site of Takata- Matsubara.

(photo taken by Ohsumi, May 3, 2011).

Coastal levees were washed away by the tsunami. A Takada-Matsubara sandbar located at the mouth of a bay was mostly destroyed the tsunami inundation and subsidence of the ground surface.

The land lock tide gate equipment survived, in contrast to those that collapsed in the coastal levees

(Fig. 4-25). As the tide gate equipment had good foundations only the frame section suffered slight damage. In the area surveyed the tide gates had survived the tsunami.

Full view Close up

Fig.4-25 Comparison of coast levees and tide gate. (photo taken by Ohsumi, April 6, 2011). The tsunami completely washed away wooden houses. The level of damage sustained by the building structures varied depending on foundation type. Most structures with mat foundations collapsed (Fig. 4-26 (a)). In contrast, the structures with pile foundations, which were the buildings located on the marsh of an old river in Takada-Matsubara park, avoided collapse and only suffered scouring of the foundations (Fig. 4-26 (b)).

Similar forms of damage were seen in the structures close to the coast. In Rikuzentakata, most of the structures located in the coastal area suffered crushing damage in the tsunami (Fig. 4-26).

The Kiba district area was flooded owing to the ground sedimentation caused by the earthquake and an inflow of the seawater from the later river (Fig. 4-27). The inundation was seen 2 months after the tsunami occurred. Influence of long-term inundation such as the damage from salt breeze

on rice fields is a concern. Back current that occurred at the time of high water is shown by the blue

arrow in Fig. 4-27.

（Arrow shows back current direction）

a b

a: Mat foundation’s structure located in 1,500 m away from the coast. b: Pail foundation’s structure located in only 10 m away from the coast.

Fig. 4-26 Compulsion of foundation and mat foundation by structural damage. (photo taken by Ohsumi, April 6, 2011)

April 6, 2011 May 3, 2011 Fig. 4-27 The ground sedimentation caused by the earthquake and an inflow of the seawater from the later river in the Kiba village. (photo taken by Ohsumi) 31

5. Tsunami Evacuation of children in Unosumai District, Kamaishi City

In the 2011 Tohoku Earthquake and Tsunami, many people could not or did not evacuate from the tsunami. More than 18,000 people were killed or reported missing (excluding earthquake-related deaths) (Fire and Disaster Management Agency, 2016 [14]; Reconstruction Agency, 2016 [15]), and more than 90% of the people were killed due to the tsunami (National Police Agency, 2012 [13]). In

Kamaishi City, about 1,000 people were killed or missing (excluding earthquake-related deaths)

(Iwate Prefecture, 2017 [16]).

On the other hand, students at junior high schools and elementary schools started tsunami evacuation quickly, and consequently, they saved not only themselves but also people around them from the tsunami in Kamaishi City. This is called the “Kamaishi Miracle” and has gained much attention as an example of the benefits of disaster prevention education (Katada and Kanai, 2016 [18];

NHK, 2015 [19]; Katada, 2012 [20]).

Here, this study focus on the tsunami evacuation of children in Unosumai district, Kamaishi City and discuss the effect of their prompt evacuation on surrounding areas.

5.1 Summary of the 2011 Tohoku Earthquake in Unosumai District

Unosumai District (7-19 and 23-29 Chiwari, Unosumai-Chou) in Kamaishi City was heavily

damaged by the 2011 Tohoku Earthquake. Of the 3,276 residents in this district (Feb. 2011), 348

people were killed or reported missing due to the tsunami (Kamaishi City, 2015 [21]). Fig. 5-1

shows the location of the schools and the temporary tsunami evacuation sites in the district (based

on Kamaishi City, 2015 [22]). Most of the area around the schools was inundated.

Fig. 5-1 Location of the schools and the temporary tsunami evacuation sites in Unosumai district (Based on Kamaishi City, 2015 [22])

5.2 Summary of Tsunami Evacuation of Children in Unosumai District

This study summarize the tsunami evacuation of students from Kamaishi Higashi Junior High

School and Unosumai Elementary School. For more information on their tsunami evacuation, refer to Katada and Kanai (2016) [18], Kamaishi City (2015) [21], or Katada (2012) [22].

Kamaishi Higashi Junior High School was located near the sea and Unosumai Elementary School was located opposite from the junior high school (Fig. 5-1).

5.2.1 Behaviors just after the earthquake occurrence The earthquake occurred at 14:46 and strong shaking hit Unosumai District. In Ofunato City,

located directly south of Kamaishi City, the Japan Meteorological Agency (JMA) observed strong

shaking higher than level 4 on the JMA seismic intensity scale, for a period of 160 seconds at

Ofunato City located right south of Kamaishi City (JMA, 2011 [23]).

At 14:46 at Kamaishi Higashi Junior High School, all the classes were finished and students had

started club activities and were preparing for the graduation ceremony. Just after the earthquake

occurrence, the vice principal tried to give students instructions to evacuate over the school's

broadcasting system, but it did not work because of the blackout. The students protected

themselves from danger and waited for the earthquake shaking to stop. After that, they gathered in

the schoolyard on their own judgment without the instruction of teachers.

In Unosumai Elementary School, the classes were not over and the students were still inside the

school building. After the earthquake shaking stopped, the teachers tried to give the students

instructions to move to the third floor (higher floor).

5.2.2 Evacuation to the first temporary evacuation site

In Kamaishi Higashi Junior High School, a teacher shouted, “Run away!” to the students

gathered in the schoolyard. The students started to run to the first temporary evacuation site (an

elderly care facility, not higher place), which was designated as the evacuation site located further

inland from the school (Fig. 5-1).

In Unosumai Elementary School, the teachers and the students saw running junior high school

students. This witnessing triggered the start of tsunami evacuation in the elementary school. They

followed the junior high school students and moved to the first temporary evacuation site.

5.2.3 Evacuation to the second temporary evacuation site

After they reached the first temporary evacuation site safely, an old woman who had followed

students and some junior high school students found the cliff behind the building was collapsing.

This witness led people gathered at the first temporary evacuation site to move to the second

temporary evacuation site (other elderly care facility, higher place) (Fig. 5-1).

5.2.4 Evacuation to the third temporary evacuation site After they reached the second temporary evacuation site safely, they saw the tsunami had

overflowed the seawalls. They moved to the third temporary evacuation site (the stone store, higher

place).

5.3 Tsunami evacuation simulation

This study apply the tsunami evacuation simulation tool (Dohi et al., 2016 [24]) to the tsunami evacuation of children in order to discuss the effect of their behavior on surrounding areas. This simulation tool consists of two models: Evacuee Generation Model and Evacuee Behavior Model.

It can simulate not only the evacuation behaviors but also the sense of urgency (Reality of

Evacuation Start: RES). For more information on the simulation tool, refer to Dohi et al. (2016)

[24] or Dohi et al. (2017) [25]. This study summarize the main parameters and conditions of the

simulation in Table 5-1.

5.3.1 Setting the initial conditions

This study focus on the tsunami evacuation of 212 Kamaishi Higashi Junior High School and

361 Unosumai Elementary School students from their schools to the first temporary evacuation

site.

According to Kamaishi City (2015) [21], Kamaishi Higashi Junior High School had 217

students. Three of them were absent and another two of them left school early. Unosumai

Elementary School had 362 students. One of them left school early and another one of them left

with the family member at the first temporary evacuation site.

Regarding the start of tsunami evacuation, this study set the following conditions.

 Two students went through the school gate every second and moved to the first temporary

evacuation site.

 Junior high school students started tsunami evacuation after the earthquake shaking stopped.

 Elementary school students started tsunami evacuation a minute after the leading junior high school

students start tsunami evacuation.

This study set the computational area around the route of the schools to the first temporary

evacuation site as shown in Fig. 5-2 (a) (Based on Kamaishi City, 2015 [2]).

5.3.2 Setting the parameters

i. Evacuee Generation Model

This study consider four information sources creating RES: evacuating people, strong shaking,

municipal RCS (Radio Communication Systems), and TV/Radio. Table 5-2 summarizes the 36

parameters of each source.

The source of evacuating people implies that the witnessing of their behavior creates RES in the viewer. This study set the effective area for evacuating people to be a circular shape around them.

This study set the effective areas for the other three sources to be the whole Unosumai district.

The weight assigned to an information source implies its effectiveness level and indicates which information source is more important in our society or local communities. These parameters are based on Dohi et al. (2016) [24].

To determine the radius of the circular effective area (evacuating people as the information source), we used the relationship between a visual distance (m) and the size of an object

(m), which is defined as follows: 𝐷𝐷 𝑆𝑆

= (1) where is a vision factor. According 𝐷𝐷to a visual𝑉𝑉𝑉𝑉 acuity test with the use of a Landolt C, if a test subject 𝑉𝑉can see the symbol (7.5 cm wide) from 5 m away, they are judged to have 20/20 vision. In this case, the vision factor is 66.7. According to e-Stat (2018) [26], mean height of Japanese junior high school students 𝑉𝑉is 1.57 m (mean height of each grade in 2011) and that of Japanese elementary school students is 1.31 m (mean height of each grade in 2011). In the case of Japanese junior high school students, using Eq. (1), = 66.7, and = 1.57 m, this study can obtain the visual distance = 105 m. Similarly, in the𝑉𝑉 case of Japanese𝑆𝑆 elementary school students, this study obtain the𝐷𝐷 visual distance = 87 m, using = 1.31 m. In this simulation, for the effective range (radius) of evacuating𝐷𝐷 people, this study𝑆𝑆 use 105 m for junior high school students and 87 m for elementary school students.

ii. Evacuee Behavior Model

In this model, a human body is modelled as a double circle element with a physical radius and

a psychological radius, and the interaction between people is modelled by a spring and damping.

The movement of each person is determined by solving the equation of motion. For more information on the model, refer to Kiyono et al. (1994) [27].

This study summarize parameters about the velocity in Table 5-3 based on Okada et al. (1977)

[28]. In this model, each reference velocity of the evacuee is automatically determined using the mean, maximum, and minimum velocity. Without external forces, an evacuee can move at the reference velocity. When an evacuee is subjected to external forces, solving the equation of motion, the acceleration occurs and the velocity is changed.

This study set the evacuation route of students based on Katada (2012) [20] (Fig. 5-2 (a)). The distances between the first temporary evacuation site and Kamaishi Higashi Junior High School or Unosumai Elementary School are 1,000~1,100 m on this evacuation route (The linear distances are about 800 m).

Table 5-1 Main parameters and conditions in the simulation The number of students 573 Total time for analysis (min) 30 Time step of the Evacuee Generation Model (s) 0.10 Time step of the Evacuee Behavior Model (s) 0.05

Table 5-2 Parameters about the information sources creating RES (The weights are set based on Dohi et al., 2016 [24]) Information Strong Municipal Evacuating people TV/Radio source shaking RCS

Spatial Around an evacuee characteristic (round shape) (Effective area) Whole Whole Whole

Spatial characteristic Junior high school students: 105 m (Radius of round Elementary school students: 87 m - - - effective area) After the After the start of start of From the start of tsunami evacuation During tsunami tsunami Time until the finish of it strong characteristic (From going through the school gate until evacuation evacuation shaking (Effective time) reaching the first temporary evacuation of the of the (160s) site) leading leading student student 7.66 × 2.51 × 2.25 × Weight 2.82 × 10-3 10-3 10-3 10-3 212 Junior high school students Total number 361 Elementary school students - - -

Table 5-3 Parameters about the velocity (Based on Okada et al., 1977 [28]) Mean velocity (m/s) 1.21 Standard deviation of the velocity 0.30

Minimum velocity (m/s) 0.67 Maximum velocity (m/s) 2.40

5.3.3 Results

The simulation results showed the leading student reached the first temporary evacuation site

7.6 minutes after the start of tsunami evacuation. Half of the group (287 students) reached it within

9.8 minutes of the leading student reaching it.

Fig. 5-2 shows the snapshots of RES after the leading students began tsunami evacuation. In

Fig. 5-2, this study consider only evacuating people as the information source in order to focus on the spatio-temporal changes of RES created by evacuating students. RES created by them moves toward the lower side in Fig. 5-2 with their evacuation.

Fig. 5-2 Snapshots of RES after the leading student started tsunami evacuation (In the figures, this study considered only evacuating people as the information source. The map is based on Kamaishi City, 2015 [22] and the evacuation route of students is based on Katada (2012) [20].)

6. Discussion and Conclusion 6.1 Field investigation and numerical investigations

Based on the field and numerical investigations, the following conclusions could be drawn.

The Yoshihama river dike was damaged by subsidence-related failure in the dike body with low

soil density and high water table. Furthermore, the tsunami and backrush led to further reduction in

levee body strength, despite the existence of concrete blocks at the back.

According to the rupture process of the main shock of the 2011 Tohoku Earthquake, using the

acceleration record obtained from K-NET and KiK-net (K-NET, 2011), the first rupture occurred

off Miyagi prefecture, and strong seismic waves were released all over Tohoku (first phase). After

several tens of seconds, another massive rupture occurred, and strong seismic waves were released

(second phase). The third rupture occurred offshore near northern Ibaraki, and strong seismic waves

were radiated towards Ibaraki prefecture (third phase).

In the Touni area, residences were moved uphill after the Meiji Sanriku tsunami in 1896 and the

Showa Sanriku tsunami in 1933. Numerous forest fires occurred and spread to residences in the low

areas. These residences had been moved uphill again after the Showa Sanriku tsunami in 1933.

However, stores began to open in the low-lying areas two years later, and the village had been rebuilt

by 1945.

In Yoshihama area, residences moved uphill after the Meiji Sanriku tsunami in 1896. These

residences had been moved uphill again after the Showa Sanriku tsunami in 1933. As a result, 2011

Tohoku earthquake resulted in minimal damage (only 1 dead people, 2 households damaged). This

area was praised as “Miracle Village” and an inscription was made as a new monument.

6.2 Tsunami Evacuation

This study discuss spatio-temporal changes of RES by comparing RES at five sites in three residential areas around the evacuation route of students (Fig. 5-1 (a) (b)). In residential areas B and

C (Fig. 6-1 (a)), this study consider sites on the far side of the evacuation route (B1 and C1) as well as on the near side of it (B2 and C2) (Fig. 6-1 (b)).

6.2.1 RES every second and cumulative RES

Fig. 6-1 (c) to (g) show the changes of RES every second and cumulative RES at five sites. In

each figure, left graphs show the per-second and cumulative RES during the 160 seconds of strong

shaking. Right graphs show them after the leading student had begun tsunami evacuation. The

right graphs (solid lines) in the figures show that strong RES was created by students on the near

side of the evacuation route of students (A1, B2, and C2), 5-20 minutes after the leading student

started tsunami evacuation. In particular, the peak of RES at A1 (around 12 minutes) is

significantly high because a lot of students were evacuating nearby, at the peak time.

RES at B2 was lower than at A1 because B2 was always out of the effective area of RES created

by evacuating elementary school students. It means the distance between B2 and the evacuation

route of students is longer than that of evacuating elementary school students (87 m) and shorter

than the radius effective area of evacuating junior high school students (105 m).

RES at C2 was also lower than at A1 because the students reached the first temporary evacuation

site soon after they passed near C2 and stopped acting as information sources in this simulation.

This means RES at C2 was created by students for a shorter time than it at A1.

On the other hands, RES was no longer created by students on the far side of the evacuation (B1

and B2). The distance between the two sites and the evacuation route is longer than the radius of

effective area of evacuating students (87 m and 105 m).

Considering the above, this study can assume that the tsunami evacuation of students created strong RES in the residential areas near the evacuation route of students. In other words, it is assumed that in these areas tsunami evacuation easily began within 10 to 20 minutes of the leading student beginning evacuation because of the students’ prompt tsunami evacuation.

6.2.2 Effects of the students vs. other effects

So far, this study have only discussed evacuating people (students) as the information source creating RES. Here, this study compare the effects of the students with other effects.

First, this study discuss the effect of strong shaking. Table 6-2 summarizes the parameters of strong shaking as the information source. According to JMA (2011), strong shaking lasted for 160 seconds in Ofunato City. This study therefore simulated RES created by strong shaking for 160 seconds at five sites (Fig. 6-1 (b)). The left graphs in Fig. 6-1 (c) to (g) show the RES created by it at five sites. Compared with the right graphs (solid lines) in Fig. 6-1 (c) to (g), which show the

RES created by evacuating students, the effects of strong shaking are significantly smaller than the effects of evacuating students at A1, B2, and C2. However, in terms of RES at B1 and C1, this study can assume that strong shaking was the important information source because RES was not created by students at B1 and C1 which were beyond the effect of students.

Second, this study discuss the effect of municipal RCS and TV/Radio. As there was a blackout, municipal RCS could not have worked. Here, this study discuss how much of a difference municipal RCS and TV/Radio could have created at the five sites. Table 6-2 summarizes the parameters of them as the information source. In the simulation, this study set the effective time to be after the leading student began evacuation. In the right graphs in Fig. 6-1 (c) to (g), the dashed lines show RES created by both evacuating students and municipal RCS and TV/Radio. Compared with the solid lines, which show RES created by evacuating students only, the effects of municipal

RCS and TV/Radio are significantly small at A1, B2, and C2. However, similar to the effects of 44

strong shaking, this study can assume that municipal RCS and TV/Radio could have been important information sources at B1 and C1, which were beyond the radius of effect of students.

6.2.3 Evacuation time There is a possibility that more students reached the first temporary evacuation site earlier than the simulation results for the following reasons.

First, this tsunami evacuation simulation considers each evacuation behavior separately and the acceleration changes caused by physical or psychological interaction. On the other hands, the photo of the tsunami evacuation of the students showed their evacuation in line (Katada and Kanai,

2016 [18]; NHK, 2015 [19]). The students could have moved smoothly without physical or psychological interaction among them.

Second, the parameters about the velocity are based on the walking velocity of adults (Okada et al., 1977 [28]) in this simulation. Our simulation does not consider the following situation: according to Katada and Kanai (2016) [18] and NHK (2015) [19], the students ran to the first temporary evacuation site. In addition, they say that some Junior high school students assisted with the evacuation of nursery school children and staff who they met on the way to the first temporary evacuation site.

If the students moved more smoothly than simulation results, it is assumed that the peaks of

RES were higher and peak times were shorter at A1, B2, and C2 (Fig. 6-1 (b)). On the other hand, there is no change in RES at B1 and C1 between the smooth evacuation case and the simulation results, because these sites were beyond the effective area of RES created by evacuating students.

(a) Distribution of buildings (b) Location of the target sites near the evacuation route of the and the evacuation route of the students students (c) Changes of RES (upper) and cumulative RES (lower) at A1

(d) Changes of RES (upper) and cumulative RES (lower) at B1 (e) Changes of RES (upper) and cumulative RES (lower) at B2

(f) Changes of RES (upper) and cumulative RES (lower) at C1 (g) Changes of RES (upper) and cumulative RES (lower) at C2

Fig.6-1 Changes of RES every second and cumulative RES at five site. In the figures (c) to (g), left graphs show the RES every second and cumulative RES during strong shaking for 160s. Right graphs show them after the leading student start the tsunami evacuation. Two lines almost overlap in the right graphs in the figures (c) to (g). (The figure (a) is based on Kamaishi City, 2015 [21]. In the figure (b), the map is based on Kamaishi City, 2015 [22] and the evacuation route is based on Katada (2012) [20].) 46

References [1] Suzuki, W., Aoi, S., Sekiguchi, H., Takashi K. (2011) : Rupture process of the 2011 Tohoku‐ Oki mega‐thrust earthquake (M9.0) inverted from strong‐motion data, Geophysical Research Letters, 38, L00G16, doi:10.1029/2011GL049136. [2] Furumura, T., Takemura, S., Noguchi, S., Takemoto, T., Maeda, T., Iwai, K. and Padhy, S. (2011), “ Strong ground motions from the 2011 off-the Pacific-Coast-of-Tohoku, Japan (Mw = 9.0) earthquake obtained from a dense nationwide seismic network.” Landslides, 8, 333 – 338. [3] Ide, S., Baltay, A. and Beroza, G. C. (2011), “Shallow Dynamic Overshoot and Energetic Deep Rupture in the 2011 MW 9.0 Tohoku-Oki Earthquake.” Science, 332, No. 6036, 1426 – 1429. [4] Ammon, C. J., Lay, T., Kanamori, H. and Cleveland, M. ( 2011), “A rupture model of the 2011 off the Pacific coast of Tohoku Earthquake.” Earth Planets Space, 63, 693 – 696. [5] Kurahashi, S. and Irikura, K. (2011), “Source model for generating strong ground motions during the 2011 off the Pacific coast of Tohoku Earthquake.” Earth Planets Space, 63, 571 – 576. [6] An Approach to Tsunami Hazard Assessment along the Northeastern Coastal Area in Japan - Method and Preliminary Results -, Technical Note of the National Research Institute for Earth Science and Disaster Prevention, No.400, [in Japanese] [7] Tsunami prediction method for earthquake with specified source faults ("Tsunami Recipe") (2017), Headquarters for Earthquake Research Promotion [in Japanese] [8] Okada, Y. (1992), Internal Deformation due to Shear and Tensile in a half-space, Bull. Seismol. Soc. Am., 85, 1018–1040.

[9] Shuto, N. (2011b) ,Tohoku Earthquake Tsunami Survey No.5, Coastal Engineering Committee, Journal of Japan Society of Civil Engineers, pp.1-11. [10] Koshimura, S. (2005), Learn from a past disaster, series no.4., The 1896 Maiji Sanriku Earthquake/Tsunami, No.28, pp.18-19 [in Japanese] [11] Hara, T., Okamura, M., Uzuoka, R., Ishihara Y. and Ueno, K. (2012), Damages to river dikes due to tsunami in south-central coastal area of Iwate prefecture in 2011 off the Pacific Coast of Tohoku Earthquake, Japanese Geotechnical Journal, Japanese Geotechnical Society, Vol. 7, No. 1, 25-36 [in Japanese] [12] Hazarika, H., Kasama, K., Suetsugu, D. and Kataoka, S. (2011), Damage to waterfront structures in northern Tohoku area due to the March 11, 2011 tsunami, Invited Paper, Proc. of the First Indo- Japan Workshop on Geotechnical Engineering, Kochi, India, CD-ROM. [13] Shuto, N. (2011a), Tohoku Earthquake Tsunami Survey No.1, Coastal Engineering Committee, Journal of Japan Society of Civil Engineers, pp.1-12. [14] Fire and Disaster Management Agency (2016), “News Releases”, http://www.fdma.go.jp/neuter/topics/houdou/h28/03/280308_houdou_1.pdf [Last accessed January, 47

2020] [15] Reconstruction Agency (2016), “Disaster related deaths due to The 2011 Great East Japan Earthquake” [in Japanese] [16] National Police Agency (2012), White Paper on Police 2012 [in Japanese] [17] Iwate Prefecture (2017), “List of the human damage and the building damage related to the Great East Japan Earthquake” [in Japanese] [18] T. Katada and M. Kanai (2016), “The School Education to Improve the Disaster Response Capacity”, Journal of Disaster Research, Vol.11, No.5, pp.845-856. [19] NHK (2015), “Kamaishi Miracle”, East Press, 260 pp. [in Japanese] [20] T. Katada (2012), “Education for protecting the life”, PHP Institute, 206 pp. [in Japanese] [21] Kamaishi City (2015), “Kamaishi East Japan Earthquake verification report [tsunami evacuation ed.] (FY2013 Edition)” [in Japanese], http://www.city.kamaishi.iwate.jp/fukko_joho/torikumi/shinsai_kensyo/detail/1196093_3066.html [Last accessed March 6, 2019] [22] Kamaishi City (2015), “2014 1st Kamaishi East Japan Earthquake verification committee school Subcommittee held a result” [in Japanese], http://www.city.kamaishi.iwate.jp/shisei_joho/shingikai/kaisai_kekka_26_/detail/1192167_3428.ht ml [Last accessed March 6, 2019] [23] Japan Meteorological Agency (2011), “News Releases”， http://www.jma.go.jp/jma/press/1103/25a/201103251030.html [Last accessed January 1, 2020] [24] Y. Dohi, Y. Okumura, M. Koyama, and J. Kiyono (2016), Evacuee Generation Model of the 2011 Tohoku Tsunami in Ishinomaki, Journal of Earthquake and Tsunami, Vol.10, No.2, pp.1640010_1- 1640010_17. [25] Y. Dohi, Y. Okumura, and J. Kiyono (2017) : Spatio-temporal Analysis of the Start of Tsunami Evacuation in the 2011 Tohoku Tsunami in Shidugawa, Minamisanriku, Journal of Japan Society of Civil Engineers, Ser. A1, Vo l . 73, No.4, pp.I_742-I_752 [in Japanese] [26] e-Stat (2018), “School health statistics” [in Japanese], https://www.e-stat.go.jp/dbview?sid=0003147022 [Last accessed January 1, 2020] [27] Kiyono, J., Miura, K., Takimoto, K. and Nakajima, Y. (1994), “Evacuation Simulation in Emergency by Using DEM”, Papers of the Annual Conference of the Institute of Social Safety Science, Vol.4, pp.321-327 [in Japanese] [28] Okada, K., Yoshida, K.. Kashihara, S. and Tsuji, M. (1977), “Ergonomics on an architecture and a city”, Kajima Institute Publishing, 313pp. [in Japanese]

Chapter 2

Beyond the 2011 Tohoku Earthquake

1．Motivation

More than 8 years have passed since the Tohoku Earthquake, which together with the subsequent tsunami generated off the Pacific coast of Tohoku, caused widespread damage in eastern Japan. A tsunami hazard inventory is important for mitigation of tsunami disaster due to earthquakes, and also for tsunami hazard evaluation. The tsunami hazard evaluation project for Japan was started in 2012 by NIED (Fujiwara, et al., 2013, Hirata, et al., 2014). This project carried out an inventory survey on the application of tsunami hazard evaluation in Ibaraki and Chiba prefectures, municipalities that were damaged in the tsunami after the Tohoku Earthquake. A committee was established that compiled an assessment of several categories. This included summaries of the status of tsunami hazard assessment, the feasibility of its actual application, case studies, and a survey on the utilization of this tool by municipalities. The committee conducted the assessment to establish medium-to-long- term objectives and to create a proposal for the implementation of those objectives. There is hope that the new tsunami hazard assessment will be applied via a research and development project for improving local disaster mitigation support.

Keywords— 2011 Tohoku Earthquake, tsunami hazard assessment, Earthquake Research Committee(ERC),

2. Comprehensive tsunami hazard assessment for Japan

NIED have started a research project for tsunami hazard assessment for the whole of Japan, based on the lessons learned from the Tohoku Earthquake. In this project, this study is comprehensive probabilistic tsunami hazard assessment in which this study considers all earthquakes that could be tsunami sources and this study will study detailed tsunami analysis of scenario type for speciﬁed earthquakes [1, 2].

Tsunami hazard assessment is the most important information to take effective measures against possible tsunami attacks in future. After the national tragedy caused by the 11st March 2011 off the

Pacific Coast of Tohoku Earthquake (Mw9.0), NIED started a research project regarding tsunami hazard assessment in Japan to support various kind of measures by sectors such as municipalities, life-line companies, and so on, (Fujiwara et al., 2013, JpGU) [1]. Our research project consists of two components; (A) a research of probabilistic tsunami hazard assessments in which this study considers all of possible tsunamis that may affect coastal regions in future and (B) a research to forecast coastal tsunami heights and inundation ﬂow depths based on speciﬁed earthquake scenarios.

3. What is Probabilistic tsunami hazard assessment? For probabilistic tsunami hazard assessment for the whole of Japan, NIED has been making tsunami source model for all possible earthquakes based on the long-term evaluation of earthquakes by the Headquarters for Earthquake Research Promotion, taking into account of the various types of uncertainty (Fujiwara et al., 2013, JpGU) [1]. Based on the model, using probabilistic assessment methods, this study evaluates the height of tsunami at the coast. In the calculation for the nationwide tsunami hazard assessment, the minimum mesh size for the land side is 50 m and set larger mesh size as 150 m, 450 m, 1350 m for the ocean side. NIED conducted test calculation of probabilistic tsunami hazard assessment for earthquakes in the Japan Trench. In addition, by limiting the area, this study calculates tsunami hazard by using the ﬁne mesh (10m as minimum mesh size). This study proposes a new method to express probabilistic tsunami hazard information in a form of chart like medical records. By indicating probability of tsunami height, inundation depth, and the arrival time, NIED has been considering how representation can show the regional tsunami risk in the form of chart for tsunami hazard. In the tsunami analysis of scenario type, this study has been evaluating the height of the tsunami, inundation area and the inundation depth, for speciﬁed earthquakes that are assumed in each region. By comparison with the previous record, this study has been planning to examine the validity of the calculation results and tsunami source parameters. To implement the tsunami hazard evaluation, NIED has been making terrain model for sea area and coastal terrain model needed to calculate the tsunami for the whole of Japan. Summarized the concept of source model for tsunami hazard assessment, this study organizes materials on various surveys used in modeling into database.

To strengthen regional cooperation and improving the reliability of the hazard assessment. This study has been collecting and organizing information about the tsunami hazard maps of municipalities and this study will reﬂect them into the model for calculation. Based on these efforts on tsunami hazard assessment, this study will summarize typical tsunami hazard assessment methods. This study has been considering the utilization of the information on tsunami hazard. This study has been prepared as contributing to the consideration by the Tsunami Evaluation Subcommittee of Headquarters for 51

Earthquake Research Promotion of Japan.

This study investigated a method for wide-area probabilistic hazard assessment of tsunamis accompanying for the subduction-zone earthquakes along the Japan Trench occurring along with subduction plate motion that occupies most of the cause of the tsunami. This study tried to estimate the tsunami hazard evaluation by taking a trench type earthquake along the trench as an example. The summary of the tsunami estimation procedure is as follows;

1) The Earthquake Research Committee (ERC) /Headquarters for Earthquake Research Promotion

(2011) assessed that earthquake source models will occur there and defined the possible extent of

the Japan trench. In this case, an earthquake with several earthquake source models assessed by

the ERC (2011) as the seismic source area, an earthquake of magnitude that is evaluated as a

frequently occurring earthquake and occurred in a single area by considering the earthquake of the

seismic activity than the earthquake which may be possible, from Mw 7.0 to Mw 9.4.

2) Evaluate the occurrence probability of modeled earthquake. Here, this study considers two kinds

of earthquake occurrence probability models to set the occurrence probability. The first model is an

occurrence model assuming that all earthquakes occur in accordance with probability process

(stationary Poisson process or simply Poisson process) expressed by stationary Poisson distribution.

The second model is to use the probability for the earthquake given the occurrence probability by

the update process following BPT (Brownian Passage Time) distribution in the earthquake

evaluation of ERC, assuming the stationary Poisson process for the other earthquakes. This is a

highbred model using both occurrence probabilities. Probabilities were assumed for each of these

two models.

3) As the method to estimate the tsunami height due to the modeled earthquake, the maximum

amount was adopted from the calculation value of the tsunami height on the coast obtained from

simulation analysis.

Even if the location and magnitude of earthquake occurrence are same, the slip distribution differs,

the coast tsunami character is different.

As the tsunami propagation is cased strong nonlinearity in the coastal tsunami behavior, and so on, simple tsunami estimation methods are not adopted.

4) Numerical simulation analysis obtains number of N value maximum tsunami height increases at a coastal point corresponding to the number of N value of assumed wave source fault models.

Within a return period T, the probability P (excess probability) that a certain coastal point is hit by a tsunami height H exceeding a certain reference height h can be analyzed as the occurrence probability of each wave source fault model and the predicted value of the tsunami height at each coastal point (Maximum tsunami level increase amount).

This composite assessment is expressed as a hazard curve obtained by plotting the excess probability

P on the vertical axis and the reference height h corresponding to the horizontal axis. In synthesizing the hazard curve, since this study set the above two types as the probability model of the earthquake, this study applied each occurrence probability model and calculate two kinds of hazard curves.

4．An Approach to Tsunami Hazard Assessment along the Northeastern Coastal Area in Japan NIED is working from April 2012 to the study of “Probabilistic tsunami hazard assessment”, which was an overview of Japan. Here, an overview of “Probabilistic tsunami hazard assessment” will be explained by taking a tsunami hazard in the Japan Trench as an example (Fig. 4-1).

Tsunami hazard assessment (THA) is the most important information to take effective measures against possible tsunami attacks in future (Hirata et al., 2017, AGU) [3]. After the national tragedy caused by the 11st March 2011 Tohoku earthquake (Mw9.0), NIED started a research project regarding THA in Japan to support various kind of measures by sectors such as municipalities, life- line companies and so on (Fujiwara et al., 2013, JpGU) [1]. Our research project consists of two components; (A) a research of probabilistic tsunami hazard assessments (PTHA) in which this study considers all possible tsunamis that may affect coastal regions in future and (B) a research to forecast coastal tsunami heights and inundation flow depths based on specified earthquake scenarios. In the research (A) of PTHA, began working on subjects of (1) nation-wide probabilistic tsunami hazard assessment (NWPTHA) and (2) detailed probabilistic tsunami hazard assessment for a specific region

(DPTHASR). Outlines of (1) NWPTHA are as follows;

(i) NIED considers all of possible earthquakes in future including earthquakes that the

Headquarters for Earthquake Research Promotion (HERP), already assed.

(ii) NIED constructs a set of simpliﬁed earthquake fault models, called “characterized earthquake

fault models (CEFMs)”, for all of the earthquakes mentioned above by following prescribed rules

(Toyama et al., 2014, JpGU [4]; Korenaga et al., 2014, JpGU [5]).

(iii) NIED solves a non-linear long wave equation, using staggered leap-flog, ﬁnite difference

method (FDM), including inundation calculation as coastal boundary condition, over a nesting

grid system with the minimum grid size of 50 meters, to calculate tsunamis for each of initial water

surface distributions (under research for initial water surface calculation by Akiyama et al., 2014,

JpGU [6]) generated from a large number of the CEFMs. 54

(iv)This study integrates information about coastal tsunami heights from the numerous CEFMs to

get nation-wide tsunami hazard curves, deﬁning excess probability, for coastal tsunami heights,

incorporating uncertainties inherent in tsunami forward calculation and earthquake fault slip

heterogeneity (Abe et al., 2014, JpGU [7]). In the present step, this study is revising a prototype

of NWPTHA in the case where possible tsunami sources are located along the Japan Trench as

well as this study are constructing a set of CEFMs in the case where possible tsunami sources are

located along the Nankai Trough.

As for the study of (2) DPTHASR, this study is going to develop new methods to assess inundation probability and inundation time, and so on, through tsunami inundation simulations for a set of

CEFMs using the same FDM over a nesting grid system with the minimum grid size of 10 meters including information of seawalls and breakwaters. Some of results from DPTHASR will be represented in a similar format of “Karte”(medical chart) to help understandings of tsunami hazard information by residents. In the present step, this study are constructing a new method to assess probabilistic inundation depth distribution along with calculation of hazard curves for inundation depth at speciﬁed points on land (Saito et al., 2014, JpGU [8]). In the study (B), this study are planning to construct a deterministic method to forecast coastal tsunami heights, inundation area and depth and so on, in speciﬁed sites in the scenarios that possible maximum-sized tsunamis strike there. These deterministic forecasts should be examined through comparisons with tsunami deposits distribution, historical materials, and instrument records. Also, this study are making a lot of effort to utilize probabilistic and deterministic tsunami hazard information by investigating actual usages of domestic/oversea tsunami hazard information (Osada et al., 2014, JpGU [9]) and by investigating opinions and ideas from person-in-charge of measures by municipalities for tsunami disasters thorough questionnaire surveys with direct interviews (Ohsumi et al., 2014, JpGU [10]).

Fig. 4-2 shows simulation analysis, which assessed 70% and 0%~5% of the excess probability

within the next 30 years (from Jan. 1, 2014), respectively, for the M7 and M8 class earthquakes.

This study set the within 30 years of estimated tsunami height, which associated with all the seismic activity assumed along the Japan Trench off the Sanriku coast off the Sanriku region into eight categories at various locations. In Fig. 4-2, for each classification, for each tsunami generated by these earthquakes with the calculated hazard curve of the 30-year probability of the tsunami height at each category of the coast.

This study synthesized the excess probability of tsunami heights, which enveloped for maximum cases (gray lines) for each of the models.

There is a possibility that the earthquake group contributing to the excess probability may differ from the seismic activity along the Japan Trench from the Sanriku off to Boso off areas, assuming both long-term evaluation by ERC (2011) and other seismic activity respectively. Fig. 4-2 shows the case where the Poisson model is adopted as the earthquake occurrence probability model.

Fig. 4-1 Tsunami hazard assessment evaluation method.

Fig. 4-2 Tsunami height simulation analysis of the occurrence probability within the next 30 years.

5. Trial of the hazard evaluation of maximum tsunami inundation flow depth.

Using an example as a basis this study assessed the usability of the tsunami hazard evaluation applied to predict inundation flow depth. This study used the technique to express the hazard assessment of the tsunami inundation flow depth as one aspect within the tsunami hazard evaluation of the seas around Japan. Topography data and the simulation lattice size were set at 10 m to construct an example in a city area of the Tohoku district. A hazard curve of inundation flow depth in each simulation lattice of the tsunami run-up range was constructed and the inundation flow depth distribution was predicted in a probabilistic manner. The earthquake that was assumed was set at a large number of seismic centers and outbreak frequency based on long-term evaluation of earthquakes.

The age excess probability of inundation flow depth which reached a certain threshold level demanded it from age frequency and earthquake occurrence probability of wave source fault. 57

5.1 Probabilistic tsunami inundation flow depth map Fig. 5-1 shows a probabilistic tsunami inundation flow depth map. The hazard curve indicated by the gray box on the left shows the distribution map of the coastal tsunami height in the coastal area of the Tohoku region of Japan during the return period t years. This is an example of a coastal section of the Rikuzentakata area, where the target district considered as a model area is a city area in Tohoku, Iwate Prefecture. In the coastal area, the distribution map of the probabilistic tsunami height from this hazard curve is plotted and the enlarged view of this urban area is shown in the left image in Fig. 5-1. This study expects that the hazard evaluation of tsunami inundation flow depth can be quantified to probabilistic.

沿岸における、確率論津波ハザードの陸前高田市市街地周辺における、最大浸水深のハザードカーブ Probabilistic tsunami height Probabilistic tsunami inundation flow depth 試検討例を用いた、確率論的な浸水深ハザードの面的な図化例 during the return period map in the city of Rikuzentakada during the t 再現期間years.t (image)年相当の沿岸津波高さの分布図 return period t years. (image)

ハザードカーブHazard curve

years

depth (m) eturnperiod 30 R

Tsunami height Inundation flow

再現期間t年相当の津波水位浸水深ハザード評価を確率論的に定量化できるThe hazard evaluation of tsunami inundation flow depth can be quantified to probabilistic. 広田湾（陸前高田市）The city of Rikuzentakada

Fig. 5-1 Probabilistic tsunami inundation flow depth map (image)

5.2 Tsunami propagation analysis method.

Fig. 5-2 shows the trial tsunami propagation analysis flowchart. The left side of the figure shows hazard curves, which analyzed a 10 m mesh model of the inundation flow depth, using the precise topographical data of the urban periphery. The upper right chart is the target area for tsunami inundation flow depth understanding. For the conditions of the facilities, the line data provided from

Iwate Prefecture are incorporated into the terrain. This study used a simple method for tender study, this study does not consider over flow or structural damage and so on.

項目陸前高田市周辺

①震源モデルの設定※ Wave source fault model setting 震源モデル・日本海溝の地震、1,333シナリオ（検討中）

地形データ・航空レーザスキャナ測量によって作成 ②10mメッシュの地形 10 m mesh model setting ・非線形長波理論式： Staggered grid, データを作成 Leap-frog差分スキーム・計算格子： 2430、810、270、90、30、津波伝播遡 10m の6領域を外洋から順に接続 ③浸水深の計算上シミュレー・初期水位： Okada (1992)で算出した海底 Tsunami inundation flow ション地盤変位量 depth analyzes ・施設の破壊条件：越流破壊なし（震災前の施設データを使用）・潮位条件： T.P.=0m ※ ④ハザードカーブを作成 The hazard curves calculation 項目沿岸

震源モデル・日本海溝の地震、約1,800シナリオ（検討中） ⑤ハザードカーブを面的に図化 The hazard curves distribution ・非線形長波理論式： Staggered grid, Leap-frog差分スキーム ※沿岸の津波ハザード評価と同様の方法で設定・作成津波伝播遡・計算格子： 1350、450、150、50の4領域上シミュレーを外洋から順に接続ション・初期水位： Okada (1992)で算出した海底地盤変位量・潮位条件： T.P.=0m

Fig. 5-2 Trial simulation setting flowchart.

5.3 Results of the maximum tsunami inundation flow depth

In Fig. 5-3 the results of the predicted maximum tsunami inundation flow depth calculated from the simulation topography data are shown. The results that were calculated for 1,333 scenarios in

1,333 wave source fault models was output. The inundation distribution map shows the highest magnitude earthquake. This map shows situations where there is little to no inundation in urban areas.

This kind of small scale earthquake will rarely lead to flooding. In urban areas, 196 scenarios out of

1,333 scenarios resulted inundation following breeches of seawalls.

Source models設定した震源モデル (1, 333 scenarios) were全1,333 calculatedシナリオを対象に最大浸水深を計算 for maximum tsunami inundation flow depth.

………

……… ………

………

陸前高田市市街地では、1,333シナリオのうち 196シナリオが堤防を超えて浸水

Fig. 5-3 Simulation result of Maximum tsunami inundation flow depth.

Comments from questionnaire object persons · The dispersion values, which is the dispersion found on the coast, used this time are still in the trial stage. It is a task to consider and give the dispersion for tsunami inundation flow depth.

· Since the Ministry of Land, Infrastructure, Transport and Tourism has issued tsunami inundation guidance, users hope that the countries do not make discrepancies with another ministry as much as possible. Please pay attention so as not to cause confusion to the site or society.

· Users want to know how the prediction figures concerning levels of tsunami inundation may relate to their communities so they can develop investment and evacuation plans.

· Since wavelengths greatly affect inundation, matching the fault geometry setting and the probability theory is problematic.

· Structures were damaged, which led to loss of life. Litigation may occur in cases where inconsistent results are issued against resulting in open hazard assessment results. Does the defect in that case borne?

· Users do not have to be reassured from information and must carefully learn radiality for information.

· It is necessary to refer also to the expression method of dispersion, the method of expressing flooding depth hazards, and the method of expressing coastal tsunami height.

· Users are essentially experts, and experts must be engaged to ensure the security and safety of the public. Regarding the information on hazards, it is necessary to deepen discussion first about how to use hazards from the viewpoint of experts if possible.

6. A Study on the Utilization of the Tsunami Hazard Evaluation in Japan

A tsunami hazard inventory is an important element of both earthquake-related tsunami disaster mitigation and tsunami hazard assessments. The Japanese tsunami hazard assessment project was initiated in 2012 by NIED (Fujiwara, et al., 2013 [1], Hirata, et al., 2014[2]). For this project, an inventory survey was conducted on the utilization of the tsunami hazard evaluation in Ibaraki and

Chiba Prefectures where several municipalities were damaged by the tsunami that followed in the wake of the Tohoku Earthquake. The crisis management departments of these two prefectures were initially surveyed. Subsequently, 10 municipalities in Ibaraki Prefecture and 18 municipalities in

Chiba Prefecture were surveyed. A tsunami hazard inventory entailing a description of the tsunami hazard evaluation was conducted. The current status of tsunami measures, identified issues, opinions, and criticism relating to evacuation procedures were assessed to explore the possibility of applying tsunami hazard assessments within municipalities.

6.1 Areas surveyed

In 2012, NIED initiated the tsunami hazard evaluation project in Japan (Fujiwara, et al., 2013 [1],

Hirata, et al., 2014[2]). For this project, the crisis management departments in the prefectures of

Ibaraki and Chiba were surveyed. An inventory survey was conducted on the utilization of the tsunami hazard evaluation in these two prefectures. Fig. 6-1 depicts the surveyed areas.

Fig. 6-1 Areas in Ibaraki and Chiba Prefectures where the survey on the

utilization of the tsunami hazard evaluation was conducted.

6.2 Survey coverage

The survey items covered the following points.

· Tsunami countermeasures considered through the experiences of the Tohoku Earthquake, and of

the municipalities themselves, as well as the prevailing situation and constraints.

· The feasibility of utilizing a probabilistic tsunami hazard assessment in the medium to long term.

· Inputs elicited on methods for promoting a probabilistic tsunami hazard assessment.

Fig. 6-2 depicts the survey process.

During the introductory phase of the interviews conducted for the survey, the tsunami hazard evaluation was explained to respondents.

◆ Introduction section： Introduce probabilistic tsunami hazard assessment

◆Step1 Disaster mitigation measure considering tsunami hazard assessment Issues when doing Extraction type hearing

◆Step 2 Use tsunami hazard assessment measures Proposal type hearing

Fig. 6-2 The survey process

1) The survey objectives

Direct interviews were conducted with the following objectives.

· To consider tsunami countermeasures based on the experiences of the Great East Japan

Earthquake, and of the municipalities themselves, and in light of the current situation and

constraints.

· To assess the possibility of utilizing a probabilistic tsunami hazard assessment in the medium to

long term.

· To elicit inputs regarding methods of promoting a probabilistic tsunami hazard assessment.

Fig. 6-3 depicts the survey items.

◆ Introduction section： Introduce probabilistic tsunami hazard assessment.

◆Step 1: Issues to be solved when discussing disaster prevention measures by utilizing tsunami hazard assessment Extract type hearing. 1. Viewpoint of considering various plans, measures and systems related to disaster mitigation. 1-1 About municipalities' tsunami inundation assumption (tsunami hazard map) . 1-2 Issues in creating municipalities' tsunami inundation assumption (tsunami hazard map). 1-3 Direction of utilization of tsunami hazard evaluation. 1-4 Way of disclosure of tsunami hazard assessment. 1-5 Is disclosure of information on tsunami hazard assessment more useful information for persons in charge of disaster prevention? 1-6 Is disclosure of information on tsunami hazard assessment more useful information for people in charge of disaster prevention? 2. Perspectives for implementing countermeasures (general residents, various groups, enterprises, etc.). 2-1 Contents of information for general public such as residents, distribution / acquisition method and usage. 2-2 How to display and distribute tsunami hazard information?

◆Step 2 Proposal type hearing for tsunami hazard evaluation utilization policy.

Fig. 6-3 Survey items

6.3 Tsunami hazard inventory survey results for municipalities Review and implementation of non structural measures against current tsunamis and tasks required for developing countermeasures

Based on the municipalities’ tsunami inundation maps (tsunami hazard maps), current conceptions of tsunami disaster mitigation measures were investigated.

1) Study and current implementation status of non structural measures to counter the impacts of tsunamis

a) The following categories of non structural measures against tsunamis were identified: plan

formulation, information provision relating to public relations and public hearings, training, and

single making.

b) Cities, towns, and villages generally apply structural measures in response to tsunamis of the

L1 category the L1 category tsunami (L1), which is high occurrence probability tsunami, and L2

category tsunami (L2), which is maximum level tsunami.

c) All of the surveyed municipalities have already prepared tsunami hazard maps or are in the

process of preparing them.

2) Issues that need to be considered in relation to tsunami disaster mitigation measures

a) Tsunami disaster mitigation measures are generally aimed at mitigating the impacts of

maximum class tsunami. The results of the survey indicated that municipal authorities are thinking

about situations in which it is not possible to respond to tsunamis belonging to the most frequent

and highest categories.

b) One of the opinions expressed was that the objectives of tsunami disaster mitigation measures

need to be clarified.

3) Opinions expressed on the tsunami hazard evaluation

a) Many municipalities are currently considering the possibility of utilizing the tsunami hazard

evaluation and evacuation measures to counter the impacts of tsunamis belonging to the L1 and 67

L2 categories. They are using the gradual tsunami hazard evaluation results, and consequently pointed out that the results of a thorough evaluation of tsunami hazards would make these efforts more effective. b) One of the issues associated with an assessment of tsunami hazards relates to the multiplicity of citizens’ inputs and evaluation results, leading to confusion. Whereas municipalities are considering worst-case scenario. Some of the respondents felt that utilization of the damage is smaller than the situation is difficult. c) There were many voices call for the expectation for these research, because of the difficulty of conducting simulations at the village level.

6.4 Current status of tsunami hazard map development

This study investigated the current situation based on the findings of the survey conducted in

2013 during the process of preparing the project of the tsunami inundation map (tsunami hazard

map) at the level of the municipalities.

(1) The situation regarding the development of tsunami hazard maps (Fig. 6-4)

1) Tsunami hazard maps have been created for many cities, towns, and villages based on the

assumption of prefectural or unique tsunami inundation through the Toh oku Earthquake. Eight

of the municipalities that have not yet created hazard maps of the tsunami inundation are

currently constructing maps or are planning to revise their maps in the future.

2) Before the supposed prefecture's announcement was made, there were some municipalities

creating original tsunami hazard map on the assumption of tsunami inundation area. The

municipal tsunami hazard maps present three cases of tsunamis of different heights.

3) Some hazard maps show the flow rate rather than the tsunami height. Wide tsunami

inundation areas and tsunami height levels have consequently generated anxiety.

4) Some municipalities specify evacuation ranges at the time of the announcement of the height

of tsunamis of 3, 5, or 10 m as inundation and evacuation guidelines. In addition, because of

damage incurred in the past, some municipalities have set the tsunami hazard zone at 5 m above

sea level in the inner bay area and 10 m above sea level in the outer bay area.

5) A map depicting altitudes for tsunami evacuation was prepared showing contour lines and

the tsunami inundation range at the time of the Tohoku Earthquake. This provided a flash

measure prior to the projection of the inundation and was distributed to residents within some

of the municipalities.

9 municipalities

Prepared Not prepare

19 municipalities Number of municipalities, who responded, is 28.

Fig. 6-4 Tsunami hazard map creation (following the Tohoku Earthquake)

(2) Residents’ responses to questions

1) Residents in many of the surveyed municipalities did provide any particular opinions or raise

questions after the creation of tsunami hazard maps.

2) Based on experiences associated with the Great East Japan Earthquake, it has been posited that

widespread recognition of the damage caused by the tsunami resulted in a lack of questions.

(3) Methods for creating a map based on the projection of tsunami inundation (Fig. 6-5)

Information dissemination: Major items listed in the tsunami hazard map relate to information dissemination. A total of 44% of municipalities indicated information on tsunamis on topics such as the mechanisms of tsunamis, tsunami evacuation procedures, and residents’ mental attitudes.

Geographical and tsunami-related information: All of the municipalities (100%) listed evacuation sites and 67% of the municipalities posted information relating to the altitude/ground level.

Frequency of occurrence of tsunamis: None of the municipalities described the annual rate of occurrence of tsunamis.

Characteristic description content included general guidelines for evacuation and traffic hazards, steep slopes and narrow roads, tsunami evacuation target lines, and distance from the coast.

Size and format of maps: The maps for depicting each purpose were poster-size and presented in a brochure format.

Visual features: Several municipalities have designed visual elements for their maps, including color schemes, character size, and map size.

Depiction of inundation level classification: Flow velocity indicated cases of inundation level in the map, and there were also examples of integrated classification that facilitated comprehension, due care for residents’ anxiety.

Ground level: The municipalities provided clear guidelines on the ground levels of the evacuation sites.

Evacuation procedures: Local residents’ familiarity with local roads evidently facilitated evacuation procedures. The municipalities were that residents’ own thinking about disaster mitigation would lead to improved awareness and effectiveness of evacuation procedures without need to dwell upon determined route of the municipalities. Moreover, at the time of a disaster, it is possible that such evacuation procedures may not be followed as planned. A relatively large number

of municipalities did not specify concrete evacuation routes; rather, they merely provided the direction of the evacuation.

Tsunami arrival time: Some of the municipalities did not mention the tsunami arrival time but instead called for immediate evacuation.

Publish knowledge on tsunami

Be prepared and ready to take action

Altitude · Ground height

Evacuation site

Escape route

Evacuation direction

Tsunami height

Arrival time Number of municipalities, Frequency of occurrence who responded, is 27. Flow velocity

Distance from coast

History

Fig. 6-5 Survey item

(4) Application of the tsunami hazard evaluation as a proactive measure

For this study, opinions were elicited on how municipals’ tsunami inundation maps (tsunami hazard maps) are being incorporated within disaster mitigation-related planning, urban planning administration, and the generation of awareness and knowledge about disaster mitigation among residents.

1) Information on tsunami hazards is being applied in several administrative areas such as preparing regional disaster mitigation plans, designating evacuation sites, formulating evacuation plans, and relocating facilities to areas at higher altitudes.

2) Some of the company is used for notifying visitors, checking omnibus routes, and examining the route of a courtesy car.

3) In tsunami induration areas, this information is used to raise awareness through evacuation training, disaster mitigation training, and voluntary organizations’ disaster mitigation activities.

(5) Scope of application of the tsunami hazard evaluation and issues relating to its utilization (Fig. 6-6) For this study, opinions were sought on the tsunami hazard evaluation developed by NIED to determine how it could be utilized within disaster mitigation planning, urban planning administration, and fostering public awareness and knowledge regarding disaster mitigation. This feedback also enabled the identification of issues relating to its application.

1) More than half of the respondents from the surveyed municipalities affirmed that they could use the tsunami hazard evaluation.

2) It was suggested that information on frequent tsunamis should be provided separately from that on maximum class tsunami when attempting to raise public awareness. In addition, some

respondents felt that the application of the tsunami hazard evaluation would be more effective if the results were more realistic.

3) However, some respondents noted that multiple information sources and results could lead to confusion among residents. Administrative officials also felt that the public would find the data difficult to use and that smaller results than the current assumption (maximum class). There were also opinions that it was difficult.

13 municipalities Possible

15 municipalities Negative

Number of municipalities, who responded, is 28.

Fig. 6-6 Scope of utilization of the tsunami hazard evaluation.

(6) Providing information for tsunami hazard assessment personnel on tsunami hazard assessment

(Fig. 6-7)

This survey asked how to open to public this Tsunami hazard assessment to make for utilizing scientifically grounded data as the basis of internal administrative documents and plans should be explored.

1) Respondents identified an issue that affects the utilization of the tsunami hazard evaluation.

This was the need to develop a better understanding among administrative staff who are the

recipients of the disseminated information.

2) Some respondents pointed to the assessment within enterprises, schools, and among residents.

They suggested that the focus should be on more familiarly experienced and frequent tsunamis

according to factors such as the height of the tsunami . One of the opinions was that it would be enlighten evacuation without abundant.

3) A further issue related to utilization that was identified by respondents was the prevailing assumption that the difference between L1 and L2 categories was generally understood.

To avoid underestimating the scale of a tsunami caused by an earthquake of a very high magnitude, the tsunami warnings issued by the Japan Meteorological Agency (JMA) are modified after the occurrence of such an earthquake. For the first tsunami warning, the expected tsunami height is reported, and words such as “huge” or “high” are used in the announcement to convey that an emergency situation exists. Even when a high magnitude earthquake occurs, the JMA has the capacity to determine the precise magnitude of the earthquake within about 15 minutes of its occurrence

(Meteorological Agency leaflet titled “Tsunami warning changed" [11]). The National Research

Institute for Earth Science and Disaster Mitigation is currently developing the Nihon Trench Ocean

Bottom Earthquake Tsunami Observation Network that can undertake real time detection. This is possible through direct detection of a trench-type earthquake and the tsunami that immediately follows it, resulting in the transmission of more accurate information on earthquakes compared with previously available information. This advance will contribute to disaster mitigation measures such as reduced damage caused by the tsunami and more effective evacuation behavior. Based on these circumstances, we requested respondents for their opinions on real-time tsunami information.

1) About 80% of the municipality respondents stated that they wanted present tsunami warning

systems, the expected of the tsunami height, to be issued earlier than expected. As discussed further

along in this paper, the prevailing assumption is that real-time information is used to broadcast an

alert immediately after the occurrence of an earthquake and tsunami and as a trigger signaling the

beginning of evacuation activities. Consequently, speed is considered to be more important than

any other factor.

2) The majority of municipality respondents (80% of the total) felt that the use of the conventional

warning system was acceptable but that a higher degree of accuracy was required.

3) Many municipality representatives felt that “real-time tsunami information is used as an aid to

prompt action immediately after the occurrence of an earthquake and tsunami” and to “trigger

evacuation action.”

4) “Daily disaster mitigation measures” and “a grasp of the damage situation caused by earthquake

motions and tsunamis” were also reported.

【The composition ratio of the most important item】【The composition ratio of the second most important item】

Number of municipalities, Number of municipalities, who responded, is 29. who responded, is 25 1) We want the information to be issued earlier than expected," such as "the expected height of the tsunami etc. 2) It is acceptable level even with conventional warning, but we want you to deliver it with higher accuracy. 3) Real time tsunami information is used as an aid to prompt attention immediately after earthquake and tsunami occurrence. 4) Others.

Fig. 6-7 Important items relating to the timing and accuracy of real-time information transmitted on tsunamis.

(7) Transmission of information on tsunami hazards to an agency that implements countermeasures

This study asked respondents for their opinions from the perspective of transmitting information

on tsunami hazards to individuals and agencies implementing countermeasures (members of the

public and various organizations and enterprises).

Respondents pointed out that “ingenuity that does not give a sense of security”, "importance of

understanding the dangers posed by a tsunamis of a considerable height" and “a proper

understanding of probability”.

1) About publication of "the tsunami’s arrival time" and "the evacuation distance within __

minutes" (Fig. 6-8).

a) A total of 60% of representatives of municipalities responded that announcing "the tsunami’s

arrival time" to residents in advance was an effective strategy, and 4% or respondents noted the

“the evacuation distance within __ minutes ”.

b) Respondents pointed out that the publication of "the tsunami’s arrival time" dissipated tension,

whereas publication of a short arrival time caused giving up refuge.

Number of municipalities, who responded, is 25.

(%)

rate Response

Evacuation distance within __ minutes Tsunami arrival times

Fig. 6-8 Publication of "the tsunami’s arrival time" and "the evacuation distance within __ minutes"

2) Opinions on the level of content available to the public (Fig. 6-9)

a) About 60% of the municipality representatives reported that minimal explanations such as

simply results were easy to understand in relation to the contents released to the public.

b) In some cases, the public announcement of data entailed a high degree of preparation. Some

respondents also suggested that a mechanism for providing detailed data was necessary in cases

where residents asked for details (Fig. 6-10).

1 municipalities 3 municipalities

Minimum explanation

Explanation including process 4 municipalities 15 municipalities According to the case Others

Number of municipalities, who responded, is 23

Fig. 6-9 Level of content to be disclosed to residents.

Website

Smart phone,etc.

Paper medium

Disaster mitigation e-mail/Area e-mail

Distribution method adapted to the age Number of municipalities, who responded, is 21.

Fig. 6-10 Formats for disseminating the tsunami hazard evaluation.

3) Formats for disseminating the tsunami hazard evaluation

a) A majority (70%) of municipal authorities stated that a website was the preferred format for the

general dissemination of the tsunami hazard evaluation.

b) Some municipal authorities felt that a paper medium (a booklet) was also required for

information dissemination because of the aging population.

c) Some municipal authorities felt that individual or area-wise emails should be considered for the

effective dissemination of information on disaster mitigation to individuals with low levels of

interest or awareness.

4) Method of presenting the results of the tsunami hazard evaluation

a) Many respondents felt that the results of the tsunami hazard evaluation were well presented and

included important information such as the probability of tsunamis occurring once every 100 years.

b) Some respondents felt that further explanations of terms such as “levels” were required to

facilitate residents’ understanding.

6.5 Proposal type hearing for policy formulation on the utilization of the tsunami hazard evaluation Proposal type hearing for policy formulation on the utilization of the tsunami hazard evaluation

Proposal type hearings were conducted on the following themes.

1) Presentation of menu for utilization policy

2) How to use the tsunami hazard evaluation and municipal tsunami flooding projection,

3) Tsunami hazard chart

4) Provision of GIS-applicable information.

Fig. 6-11 shows the mitigation assessment, with examples of its application in relation to basic materials.

Assistance for prioritization of disaster mitigation Support for review and improvement of disaster facilities development mitigation measures

●Information by tsunami hazard evaluation (example) ●Information by tsunami hazard evaluation (example) Probability of occurrence Probability of occurrence Municipality Tsunami height Arrival time of tsunami of 10 m or of tsunami less than 5 m A city Max. 10 m 10 minute more within ◯ years within ◯ years Within 10% About 60% B city Max. 3 m 3 minute

● City A disaster reduction countermeasure policy ● B prefecture disaster reduction countermeasure (draft) policy (draft) Priority disastermitigation measures: Priority 1: Measures to prevent flooding in low-altitude A city: Although there is relatively evacuation time, urban areas, measures to evacuate, etc. since the tsunami height is high, emphasis is placed Priority 2: Training to prompt evacuation in on soft countermeasures such as evacuation drills preparation for the 10 m class tsunami from harder Evacuation location designation for high altitude B City: Since there is no evacuation time, strengthen districts coast levees, emphasize hard measures such as designation of evacuation building

Fig. 6-11 Mitigation assessment that can be utilized by municipalities to prioritize disaster mitigation measures and to examine priority measures.

1) Comment from the tsunami hazard evaluation object persons #1:

This information will provide a basis for developing an overall policy on disaster prevention measures. If simple results can be provided, then it will be easy to set up mitigation measures. This information can be utilized for assessing the level of evacuation. However, this assessment may prove difficult in practice.

This information can be used as a basis for making policy decisions and providing the public with explanations.

This information can be utilized as the tsunami hazard evaluation analysis.

This information is only necessary that the probability will become advantageous to subsidies subject and disaster mitigation plans. For example, even if this information to resolve a bottleneck against evacuation, it will not be easy way. In such a case, it is expected to be utilized as a basis for disaster mitigation plans. Thus, respondents requested the development of a mechanism for promoting business using these simulation results.

Fig. 6-12 shows how the mitigation assessment can be utilized to prioritize municipalities’

disaster mitigation measures.

Tsunami heights occurred with a probability of 50% within the next 30 years in A prefecture. 0 5 10 15 ○○ Prf. Ａ市 A city 30 Disaster Mitigation Plane B cityＢ市 15 C cityＣ市 5 Comme Ｄ市 D city 10 1. Disaster commer Ｆ市 E city 1 A City: Since maximum tsuna we will implemen speed up "ev embankm

Fig. 6-12 Utilization of the mitigation assessment by municipalities to prioritize disaster mitigation measures and to assess municipalities’ utilization of priority measures for implementing disaster mitigation measures during a specific year and their business volumes.

2) Comment from the tsunami hazard evaluation object persons #2:

Respondents suggested that the mitigation assessment results could be used to facilitate the

formulation of policies for prioritizing tsunami countermeasures.

This plan will serve as the basis for developing an overall policy on disaster resilience measures.

This plan can be utilized as supplementary explanatory material and can be incorporated within

city planning as well.

A long-term goal for town development is required. However, developing long-term disaster

prevention plan, compared with medium-term and short-term plans, is difficult.

If there is probability theory, not only disaster mitigation plans but also comprehensive plans

and so on, must be addressed will be encouraged, it will be a measure of the roadmap for tsunami

countermeasures.

Fig. 6-13 shows the tsunami hazard assessment, which is setting of the tsunami level to be considered that conforms to the occurrence probability.

NIED Tsunami Hazard Assessment Coastal conservation facility Design level

Coastal area Tsunami ○ A height●m Relatively frequent tsunami (L1 tsunami) is is defined by calculating the tunami

Coastal area Tsunami height of probabilistic evaluation for the B height●m class historical earthquake occurring once in the past decades to hundreds of years, and design the tsunami height of L1 . Coastal area Tsunami C height●m ○ The maximum class tsunami (L2 tsunami) is defined by calculating probabilistic evaluated tsunami height for the historical earthquake which has Coastal area Tsunami D height●m extremely low occurrence frequency, even thought causes serious damage occurs.

Utilization for setting Tsunami height assumed to arrive at a certain frequency (return period: "Design tsunami water level" several decades to once every hundred and several decades)

Fig. 6-13 Tsunami hazard assessment in relation to the study of facilities.

3) Comment from the tsunami hazard evaluation object persons #3:

This assessment provides a reference for thinking about practical countermeasures and will also serve as a basis for requesting funds to construct coastal levees for prefectures.

Within various countries as well as in Japanese prefectures, a growing trend of utilizing gradual tsunami hazard assessment results when considering disaster prevention measures is apparent.

These results can be used within municipalities as well.

A limited budget was available to the institution that developed the L1 category tsunami (L1), which is high occurrence probability tsunami, and L2 category tsunami (L2), which is maximum level tsunami in Ibaraki Prefecture after the Tohoku Earthquake. Ongoing efforts should be based on a thorough investigation by NIED that generates detailed data.

Fig. 6-14 shows examples of customized utilization of the results of the tsunami within

municipalities.

✔ Wave source model ✔ Tsunami height ✔ Inundation area etc.

Validation of simulation results performed by local government using NIED provided data Tsunami height 1m≦H< 2m 2m≦H< 3m 3m≦H< 5m 5m≦H<10m 10m≦H<20m 20m≦H<30m 30m≦H

Fig. 6-14 Tsunami hazard assessment that enables the development of further detailed tsunami countermeasures.

4) Feedback from the tsunami hazard evaluation object persons:

The detailed tsunami frequency data along with data on the height and arrival time of a

tsunami that are presented in the hazard evaluation results are easily comprehensible.

The administration must respond to the simulation results obtained for the prefecture if

differences from the simulation results in the prefecture are known.

The results could be used in delivery courses conducted by municipalities.

7. Conclusion The results of the survey conducted on the utilization of tsunami hazard information by municipalities are summarized below.

1) A tsunami hazard assessment is expected not only for the tsunami level of only the largest class

but also for expectation that gradual evaluation can be performed. However, it is not possible to

estimate tsunami for the general population against the tsunami of high occurrence probability (e.g.,

L1 category tsunami). Many of the respondents felt that the explanation was difficult to comprehend.

2) A majority of municipalities (70%) have published tsunami hazard maps. These maps are used

for developing guidelines for raising public awareness and conducting evacuation drills for residents.

Further, they specifically describe actions to be taken by residents and evacuation procedures.

3) The survey revealed equal numbers of municipalities that have actively incorporated and utilized

information on tsunami hazards and those that have not strongly pursued this approach. Active users

suggested that an evaluation based on scientific evidence could be used to promote countermeasures

against tsunamis of a scale that is more common by promoting or prioritizing disaster mitigation

measures. However, those who were more critical felt that staff and residents found it difficult to

understand the information presented. The majority of municipalities (80%) suggested that

workshops and meetings should be organized to provide explanations and disseminate information.

4) A widely held view was that submerged cables are required to transmit real-time tsunami

information to relevant municipalities at an earlier stage than was previously done. More than 90%

of the surveyed municipal authorities were in favor of utilizing this information for alerting the

public and evacuating residents immediately after the occurrence of a disaster.

5) A total of 60% of the municipal authorities who wanted to know the arrival time of a tsunami

were expected to provide residents with information using the relevant website, smart phones, and

brochures as tools for disseminating the assessment results.

8. Adopted Recommendations

This study was conducted to assess the possibility of utilization of tsunami hazard assessments by municipalities. The study was also aimed at determining the current implementation status and issues relating to tsunami measures and compiling the views and experiences relating to evacuation, including negative ones, of municipalities.

(1) Recommendations for agencies that develop the tsunami hazard evaluation 1) Introduction of an evaluation of epistemological uncertainty relating to hazards．

· An assessment of epistemological uncertainty is also necessary for evaluating hazards.

· In general, random variation (fluctuations attributed to fundamentally unpredictable random

properties) and epistemological uncertainty (resulting from incomplete knowledge and data)

feature in evaluations of uncertainty. Theoretical uncertainty has not been sufficiently addressed,

even in areas such as ground motion prediction maps. An assessment of epistemological

uncertainty is important in the evaluation of tsunami hazards that entail a high degree of uncertainty.

· Hazard assessments that entail an evaluation of uncertainty should not be conceptualized in

restricted scientific terms but should also incorporate engineering concepts.

· If the earthquake headquarters performs the hazard assessment, not only a physical evaluation

but also an engineering-based evaluation of uncertainty should be emphasized.

· A lesson derived from the experience of the Tohoku Region Pacific Offshore Earthquake is that

efforts to evaluate hazards conducted at headquarters should focus on evaluating epistemological

uncertainty from the front.

· Logic trees can be used as an analytical tool for evaluating epistemological uncertainty. However,

it is also necessary to develop a methodology for constructing a logic tree that can appropriately

express uncertainty.

2) Clarification of preconditions and the process of developing a tsunami hazard assessment.

· It is necessary to clarify prerequisites and the process of developing a tsunami hazard assessment 88

to enable the evaluation of risks based on the hazard assessment.

· To ensure accountability to users who conduct risk assessments based on the results of a tsunami hazard assessment, it is necessary to maintain transparency by developing the assessment methodology, including preconditions. Consequently, users will be able to create their own hazard assessments by changing the prerequisites.

· This study believe that it is important to prepare a tsunami hazard assessment based on disaster prevention and engineering perspectives. However, it is important to pay attention for a tsunami hazard assessment that is closer to the safety margin, depending on the user's position, risk, it is necessary to keep in mind that it may be difficult to use for evaluation.

·A lesson derived from the experience of the Tohoku Earthquake is that efforts to evaluate hazards should focus on the evaluation of epistemological uncertainty from the front.

· Logic trees can be used as an analytical tool to evaluate epistemological uncertainty. However, it is also necessary to develop a methodology for constructing a logic tree that can appropriately express uncertainty.

3) Establishment of a method for assessing the flooding depth hazard in the inundation area.

· Conducting a probabilistic tsunami hazard assessment of the inundation area and depth in a terrestrial environment with relates to structural damage, is more important than obtaining information on the height of coastal tsunamis.

· As damage inflicted by coastal tsunamis rather than their height is the subject of a hazard assessment, land inundation areas and flooding depth are assessed. For infrastructure design, it can be informed at the structure design site with having high utility value. However, because the tsunami inundation is influenced by structural factors, including the detailed topography and conditions of coast and land areas, it is important to structural design a probabilistic evaluation method for the uncertain problem solving.

4) Development of a released to the public system to enable the utilization of a variety of

tsunami hazard assessments.

Experts who conduct tsunami risk assessments would benefit from an open system that enables a variety of tsunami hazard assessments (including the prerequisites for developing an assessment) to be utilized.

Various experts can utilize the tsunami hazard evaluation, requiring different information for different purposes. Users need to develop a system that enables them to properly utilize various tsunami hazard assessments according to their purpose.

5) Periodic updating of tsunami hazard assessments.

· Maintenance announcement is critical for sustained utilization of a tsunami hazard assessment.

· A hazard assessment entails forecasting based on scientific knowledge and can change in the future as a result of the acquisition of new knowledge and data regarding the probability of earthquake occurrence, as well as updated topographical data and more advanced and sophisticated hazard evaluation methods. Thus, it is necessary to continue to incorporate new knowledge into the development of observation and analytical technologies on an ongoing basis and to appropriately adjust and refine hazard assessments.

6) Points to be covered and incorporated in the hazard assessment to avoid misplaced confidence on the part of those designing the assessment.

·The absence of inundation across the range shown in the hazard map, the designers of the hazard assessment should clarify the purpose and limits of the hazard assessment to avoid misplaced confidence on the part of users of the assessment. Specifically, they should specify the scope of application and points of caution at the time of application as conditions relating to the provision of information.

· One of the lessons learned from the Great East Japan Earthquake relates to the issue of misplaced confidence. Authorities were confident that flooding would not occur above the range indicated in the hazard map. The developers of the hazard assessment should be aware of the danger of

misplaced confidence among users associated with the transmitted information. They should clearly state the purpose and scope of application of the assessment and should outline the method of setting preconditions such as how to handle structural conditions.

(2) Recommendations for users of the tsunami hazard evaluation

1) Proper understanding and use of the tsunami hazard evaluation can be fostered through the

promotion of consultations between the developers of the assessment and users who conduct risk

assessments that entail highly specialized information needs.

Because a tsunami hazard assessment entails highly specialized information, members of the

public, municipalities, and the personnel of disaster resilience companies may find it difficult to

directly apply this information. Experts who conduct risk assessments, and are therefore direct

users of the tsunami hazard evaluation, should collaborate with the developers of the tsunami

hazard evaluation. Consequently, the assessment can be tailored, stepwise, to achieve the

intended purpose, thereby encouraging its wider application.

Municipality representatives and residents who want to know about effective countermeasures

to mitigate the impacts of maximum class tsunami should be provided with appropriate

information.

Users should be provided with concrete examples of the assessment’s application to further

their understanding. Specifically, these examples should be incorporated into guidelines on the

use of the assessment by the general public, local public entities, and enterprises. In addition,

educational materials and supplementary course readings on tsunami hazard assessments should

be developed for schools to raise the awareness of students.

2) Risk assessments based on the tsunami hazard evaluation and their utilization within regional

disaster mitigation efforts.

Public dissemination of the tsunami hazard evaluation will enable the development of regional

measures based on risk assessments relating to different kinds of tsunamis. Plans and

countermeasures based on information on a wide range of tsunami heights, arrival times,

inundation ranges, immersion depths, flow velocity and direction, duration, time-series, and water

level fluctuations are also anticipated.

Through the conduct of a risk assessment based on an evaluation of tsunami hazards, information relating to risks can be incorporated into assessments of land use, structural planning, and investments. Establishing tsunami categories that are commensurate with local tsunami measures

It is necessary to establish tsunami categories in consultation with stakeholders in target areas that are susceptible to tsunamis. In addition, the planning of tsunami-resistant facilities to help mitigate disasters as well as non structural measures relating to land use and evacuation are necessary.

In the United States, probabilistic assessments have included a consideration of the probability of earthquake occurrence. In Japan, Level 1 and Level 2 categories have been established based on a probabilistic conception. However, a discussion is still insufficiently to contrast with hazard assessment. In addition, because residual risk exists regardless of what kind of occurrence probability setting is established, the development of non structural measures is imperative.

References [1] Fujiwara, H., Nakamura, N., Morikawa, S., Aoi, S. and Kawai1, S. et al., (2013), Tsunami hazard assessment project in Japan, JpGU, HDS28-04. [2] Hirata, K., Fujiwara, H., Nakamura, H., Osada, M. and Ohsumi, T. et al. (2014), Tsunami hazard assessment project in Japan, JpGU, HDS28-04. [3] Hirata, K., Fujiwara, H., Nakamura, H., Osada, M. and Ohsumi, T. et al. (2017), Probabilistic tsunami hazard assessment based on the long-term evaluation of subduction-zone earthquakes along the Sagami Trough, Japan, AGU. [4] TOYAMA, N., Hirata, K., Fujiwara, H., Nakamura, H. and Morikawa, N. (2014), A set of characterized earthquake fault models for the probabilistic tsunami hazard assessment in Japan, JpGU, HDS28-03. [5] Korenaga M., Hirata, K., Fujiwara, H., Nakamura, H. and Morikawa, N. et al. (2014), Large slip area in characterized Tsunami source model toward Tsunami Hazard assessment, JpGU, HDS27-17. [6] Akiyama N., Hirata, K., Fujiwara, H., Nakamura, H. and Morikawa, N. et al. (2014), A new calculation method for seabed displacement due to fault slip by boundary integration, JpGU, HDS27- 14. [7] Abe, Y ., Hirata, K., Fujiwara, H., Nakamura, H. and Morikawa, N. et al. (2014), Uncertainty for sunami hazard caused by heterogeneous slip on the characterized source model, JpGU, HDS27-16. [8] Saito, R., Hirata, K., Fujiwara, H., Nakamura, H. and Morikawa, N. et al. (2014), Inundation hazard mapping toward probabilistic tsunami hazard assessment, JpGU, HDS27-15. [9] Osda, M. Nakamura, H., Hirata, K., Ohsumi, T. and Fujiwara, H. et al. (2014), An overview on current status of public disclosure for tsunami hazard information in and around Japan, JpGU, HSC25-P10. [10] Ohsumi, T., Nakamura, H. Hirata, K., Osda, M. and Fujiwara, H. et al. (2014), Tsunami hazard inventory survey of utilize for municipalities, JpGU, HSC25-P08. [11] Meteorological Agency’s leaflet "Tsunami warning changed" http://www.jma.go.jp/jma/kishou/books/tsunamikeihou/

Chapter 3

The Comprehensive Analysis and Evaluation of Offshore Fault Informatics

1. Motivation

Chaper2 distributed the tsunamis which occur rerated to the movement of the plates. This chapter distributes the tsunamis which occur rerated to the movement of the faults. In 2013, the Japan Ministry of Education, Culture, Sports, Science and Technology (MEXT) launched “Project for the

Comprehensive Analysis and Evaluation of Offshore Fault Informatics” referred to here as the Project.

The purpose of this project is to assess the hazards of earthquakes and tsunamis in the Sea of Japan.

Currently, single-channel and multichannel reflection seismic data (SCS and MCS, respectively) and structural information for the area obtained from previous surveys conducted by various agencies around the Sea of Japan is insufficient arranged integrally information. The collection of reflection seismic data and their re-analysis with the latest data processing technology in a unified manner is also intended to develop structural information in a standard format. In this project, the Japan

Agency for Marine-Earth Science and Technology (JAMSTEC) collected offshore fault survey data, analyzed the data using a uniform method, and constructed a database. This study, conducted as part of the Project, constructed a fault model based on the geological fault information analyzed by

JAMSTEC.

Keywords: offshore fault, geological fault, tomographic, hazard assessment, Sea of Japan

2. Back ground

In 2013, the Japan Ministry of Education, Culture, Sports, Science and Technology launched "the

Project ". The purpose of the project is to contribute to the hazard assessment of earthquakes and tsunamis in the Sea of Japan. Currently, single-channel and multichannel reflection seismic data (SCS,

MCS), and structural information for the area obtained from previous surveys conducted by various agencies around the Sea of Japan is insufficient. The collection of reflection seismic data and their re-analysis with the latest data processing technology in a unified manner is also intended to develop structural information in a standard format.

In this project, the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) collected offshore fault survey data, analyzed the data using a uniform method, and constructed a database. This study was carried out as part of “ the Project”. The purpose of this work was to construct a fault model based on the geological fault information analyzed by JAMSTEC. Two fault model types are presented: a primary model and a linkage model. The primary model is based on information from “the Project”, while the linkage model considers potential fault distribution continuity and consolidation at considerable fault depths. The linkage model therefore represents a combination of specific primary models.

To evaluate the linkage effect of the faults interpreted in this study. This study used the linkage corresponding to the closest fault trace in the epicenter of the 1940 Shakotan-Oki earthquake. This study compared previous study models and linkage fault models. This study verified for tsunami traces of the tsunami traces and geodetic data from tsunami traces by Japan Tsunami Trace data base

Tsunami trace height information.

This study attempted to explain the differences in fault models using the application of appropriate slip magnitudes in the fault models as shown Fig.2-1: Theme3. The primary model is based on information from a new fault model using marine seismic industry data and geological and geophysical data compiled by the Offshore Fault Evaluation Group by JAMSTEC as shown Fig. 2-

1: Theme1 and Theme2, while the linkage model considers potential fault distribution continuity 96 and consolidation at considerable fault depths. The linkage model therefore represents a combination of specific primary models.

Sub-theme

Fig. 2-1 The purpose of the project is to contribute to the hazard assessment of earthquakes and tsunami is in the Japan Sea.

3. Fault Model Database

To construct the primary model, this study defined the following guidelines for each parameter setting that consider epistemic and aleatory uncertainties.

The position, length, and strike of the fault are based on geological fault information from the

Project. This study constructed two types of settings for fault dips. According to the strong ground motion prediction method, referred to as the Recipe, for earthquakes with specified source faults

(2009) [1], one type of setting defines the basic values of a thrust fault, normal fault, and strike/slip fault as 45°, 60°, and 90°, respectively. The second setting is used for the upper layer dip values, which were obtained in the shallow parts of the faulted region.

The fault evaluation project also determined that the shallow part of the fault dips steeply, as indicated by data of the upper layer dip value. The deeper parts of the fault dips have a gradual dip, as determined by the Recipe. The bottom depths of the fault in the model consist of two patterns, one of which uses the Conrad discontinuity level of the three-dimensional velocity structure model provided by this project; the other is based on a previous study of the Sea of Japan. The “fault width” is based on the relationship between the bottom depth of the fault and the dip angle and the fault rake was set to 90°, 270°, 0°, and 180° for thrust, normal, right lateral, and left lateral faults. The average of the fault slips is determined by the empirical relationship of the fault area and the moment from large slip areas that account for 30% of the fault area and a twofold average of slip [1].

The “position, length, and strike of the fault” is based on geological fault information from project for fault evaluation in the sea around Japan and “depth at the top of fault” is at the seabed. We constructed two types of settings for “fault dips”. One type of setting defines the basic values of a thrust fault as 45 °., normal fault as 60 °., and strike/slip fault as 90 °. Another setting is used for the

“apparent dips” from project for fault evaluation in the sea around Japan where the shallow part of the fault steeply dips, set by the apparent dips from geological data, and the deeper part has gradual dipping that is adjusted to 45 ° or 60 °., based on the average of all of the fault dips. The “bottom depths of the fault” are two patterns, one of which uses a three-dimensional velocity structure model 98 provided by this project, and another that is based on a previous study of the Japan Sea area. “Fault width” is set based on the relationship between the bottom depth of the fault and the dip angle, and the “fault rake” is set as 90 °.for a thrust fault, 270 ° for a normal fault., 0 ° for a right lateral fault., and 180 ° for a left lateral fault. The “average of fault slips” is set by the empirical relationship between the fault area and Mw by Irikura and Miyake（2001）[2]. We also considered large slip areas to account for 30% of the fault area and a twofold average of slip.

3.1 The fault modelling for the Sea of Japan

Figs. 3-1 to 3-3 shows the fault modelling for the Sea of Japan. Solid line indicates the fault

trace by JAMSTEC, and the shaded rectangle shows the fault model set by the present study, which

examines active faults of the land area based on active faults in Japan. The geological fault

information is provided by Project for the Comprehensive Analysis and Evaluation of Offshore

Fault Informatics (JAMSTEC setting). Rectangles showing hatches indicate fault models by NIED.

Fig. 3-1 Fault modelling for the Sea of Japan.

Fault trace by JAMSTEC

Fault model set by the present td

100

Fig. 3-2 Detail view of fault modelling for the Sea of Japan. (a-d)

101

Fig. 3-3 Detail view of fault modelling for the Sea of Japan. (e-f)

102

3.2 The small islands located southeast of the main islands of Japan

Fig. 3-4 to 3-6 shows the fault modelling for the small islands located southeast of the main

islands of Japan. Solid line indicates the fault trace by JAMSTEC.

Fig. 3-4 Fault modelling for the small islands located southeast of the main islands of Japan.

Fault trace by JAMSTEC

Fault model set by the present study

103

Area1 Area2 249 lines

Fig. 3-5 Detail view of fault modelling for the the small islands located southeast of the main islands of Japan (Area 1,2).

Area3 142lines

Fig. 3-6 Detail view of fault modelling for the the small islands located southeast of the main islands of Japan (Area 3).

The faults of the back arc basin, island arc basin and fore arc basin are targeted for modeling, and the fault which seems to be a spray fault near the trench axis is not covered.

104

4. Validity of the Fault Model

This study indicated a three-dimensional fault distribution of around Sea of Japan obtained from reflection seismic data. A main fault with large spread of fault plane was aimed at providing basic data for tsunami hazard evaluation, and the fault models were selected. The seismic source fault of the earthquakes, which are presumed to be ~Ⅿ 7 /more and has the record of the tsunami and the earthquake record, were verified. This study tried to explain historical tsunami records of the 1940

Shakotan-Oki earthquake and the 1983 Nihonkai-Chubu earthquake.

4.1 The 1940 Shakotan-Oki Earthquake

To evaluate the linkage effect of the faults interpreted in this study. This study used the linkage

corresponding to the closest fault trace in the epicenter of the 1940 Shakotan-Oki earthquake. This

study compares previous study models and linkage fault models, and this study verify the data by

tsunami traces of the tsunami traces and geodetic data from tsunami traces by the Japan Tsunami

Trace database and tsunami trace height information.

This study explains the differences in fault models by using the application of appropriate slip

magnitudes in these models. The primary model is based on information from a new fault model

using marine seismic industry data and geological and geophysical data compiled by the Offshore

Fault Evaluation Group by JAMSTEC, whereas the linkage model considers potential fault

distribution continuity and consolidation at considerable fault depths. The linkage model therefore

represents a combination of specific primary models.

105

Tsunami heights analysed by fault models proposed in previous studies [3] tend to be lower than observed heights. The comprehensive of reproduced model is difficult to explain the historical tsunami records of the Shakotan-Oki earthquake. The purpose of this study verifies the fault modeling of the Shakotan-Oki earthquake source region.

In this study, a new fault model was constructed using SCS and MCS, respectively; information from a new fault identified in previous studies; and geological and geophysical data compiled from the Project. This fault informatics data includes information from a new fault located in the Sea of

Japan that was investigated in previous studies. To verify the tsunami records with comprehensive faults, the arrangement of the geometrical continuity of these faults was adjusted for the several combinations. This study applied standard scaling laws based on strong ground motion for the fault parameters, and the validity of the model was examined by comparing historically observed tsunami heights from the tsunami propagation analyses on the coastline. The verification was quantified using scale and variance parameters referred to as Aida’s K and κ [4]. According to the Nuclear

Civil Engineering Committee [5], these values are recommended for use under the following conditions: K values in the 0.95–1.05 range and κ values below 1.45. A κ value of Table 3-1 is considered to be well adapted and is a measure of reproducibility [6].

The tsunami heights analysed using our new model were consistent with those shown by historical records, which suggests that the model is accurate for the source region.

106

4.1.1 Methodology

The fault model covers the epicentral area of the Shakotan-Oki earthquake that occurred on

August 2, 1940, with a moment magnitude of Mw7.7. The fault model used fault trace data from observational marine seismic reflection survey records (JAMSTEC) to approximate the fault plane in a rectangular model. The validity of the fault model was then investigated by comparison with observed tsunami records during the Shakotan-Oki earthquake and the maximum tsunami heights by simulation based on the Sea of Japan coastline. These water level data were obtained from tsunami propagation analysis using the fault model. The validity of the fault model was also assessed by quantifying the size and dispersion of tsunami height along the Sea of Japan coastline between historical events and tsunami propagation analyses.

4.1.2 Fault Model Setting

The rectangular fault model was considered to approximate the fault plane based on offshore fault informatics. The fault model group by NIED, as the minimum unit, was considered to be the basic model. We constructed basic fault models labelled HKD-22, -38, and -39 off the western coast of Hokkaido based on fault traces reported by JAMSTEC [4]. The basic model HKD-38 is located at the epicenter of the Shakotan-Oki earthquake, and HKD-39 is located to the north of the epicenter. The south end of HKD-39 corresponds to the north end of HKD-38. HKD-22 is located to the south of the epicenter; the south end of HKD-38 corresponds to the north end of

HKD-22.

4.1.3 Validity of the Fault Model

To validate the fault model, we used the linkage fault trace in the epicenter of the Shakotan-Oki earthquake.

The Shakotan-Oki earthquake was caused by interlocking neighboring faults. We considered an interlocking model with linkages between HKD-38 and other basic models in the vicinity. 107

Essentially, a rupture model was applied to reproduce the Shakotan-Oki earthquake. Further considerations included HKD-38 with the linkage model, HKD-39 and two linkage faults north of

HKD-38 (HKD-3839), and HKD-22 with two linkage faults south of HKD-38 (HKD-2238). HKD-

38 was also compared with observations recorded to the north, three of the fault linkage patterns, and three linkage faults to the south (HKD-2239).

4.1.4 Comparison with Previous Studies

A comparison of the model presented in this study and two previously published models [3, 4] revealed different trends at the coastal tsunami height scale. These differences can be attributed to differences in fault model geometries. The parameters of the fault model used in this study are listed in Table 4-1. The comparison of the previous models and the fault models is shown in Fig.

4-1. In the reanalysis of historical earthquakes, it is necessary to refer to other records such as those of magnitude and fault area. However, the fault models used in this study included the fault geometry from JAMSTEC and the magnitude described in previous surveys.

Table 4-1 Fault parameters. Fault Rakes, Dips, Slip, Length, Width, Slip ave., Mw Name deg deg deg km km m 184 22.3 HKD-38 7.1 45 90 16.9 1.50 170 30.3 176 22.3 HKD-2238 7.2 189 45 90 30.3 16.9 2.20 167 28.9 189 33.9 HKD-3839 7.2 162 45 90 15.1 16.9 2.10 189 30.3 189 33.9 162 15.1 HKD-2239 7.4 45 90 16.9 3.00 189 30.3 167 28.9 22 42.0 1.64 184 42.0 2.23 Okamura 2005 7.5 45 90 16.0 162 37.0 2.74 0 53.0 0.58 Satake1986 7.4 167 50 90 100.0 35.0 1.50 All faults are outcropping. Thus, all faults of upper depth are 0 km

108

Fig. 4-1 Fault modeling example. Solid line indicates the fault trace by JAMSTEC, and the shaded rectangle shows the fault model set by the present study, which examines active faults of the land area based on active faults in Japan. The geological fault information is provided by Project for the Comprehensive Analysis and Evaluation of Offshore Fault Informatics (JAMSTEC setting). Rectangles showing hatches indicate fault models by NIED.

4.1.5 Calculation of Tsunami Propagation Analysis The equations governing the tsunami propagation analysis with non-linear long-wave theory consider friction and advection on the seabed. This simulation uses the leapfrog finite-difference method (FDM) of a staggered grid. The computational time interval for each grid spacing in the FDM is appropriately set according to considerations of the CFL and the stability of the calculation.

In such analysis, displacement is normally neglected. However, Tanioka and Satake (1996) [8] showed the effect of horizontal deformation. When the tsunami source is on a steep slope, the horizontal displacement is large relative to the vertical displacement, and the effect becomes significant. Thus, the initial water level of the tsunami is set as the vertical component obtained from the vertical direction, and horizontal deformation effects are usually neglected. When the wave source is on a steep slope and the horizontal displacement is large relative to the vertical displacement, the

109 effect becomes significant [6]. Thus, we calculated the seabed variation considering the horizontal level. The tide level condition of the tsunami propagation analysis was set to T. P. = 0.0 m.

The maximum tsunami height was calculated at the target area. To verify the maximum tsunami height, the calculated and observed values in the coastal zone were compared (Fig. 4-2). This level was corrected for the tidal height during the tsunami that was caused by the Shakotan-Oki earthquake.

The computation time was defined from the arrival time of the reflected wave of the tsunami at the coast. The first wave of the tsunami, based on the tide gauge records [9, 10, 11], was confirmed to have reached the Noto Peninsula and Shimane Prefecture coasts within 2 h of the earthquake.

Fig. 4-2 Comparison of previously published fault models (Okamura, 2005 and Satake, 1986) and linkage fault models of this study for the offshore source area of the Shakotan-Oki Earthquake, Japan (HKD-2238: HKD-38 + HKD-22, HKD-3839: HKD-39+HKD-38, HKD-2239: HKD-39+HKD-38 + HKD-22).

110

4.1.6 Results The maximum tsunami heights determined from tsunami propagation simulations were compared with results from previous studies related to the western coast of Hokkaido (Fig. 4-2 and Table 4-1). The green numbers in Fig. 4-3 represent coastal areas. The maximum tsunami height on the coast was fairly high from Area 9 to Area 11. However, the heights were lower in parts of Area 10. It can be concluded that the wave direction line is dependent on the sea topography around Teuri Island; other models showed the same trends.

The simulated levels of the arithmetic mean of the maximum tsunami height along the western coast of Hokkaido (blue lines) and the observation records (red circles) with combined the maximum cases (gray lines) of tsunami heights are shown in Fig. 4-4. The maximum tsunami height of the Shakotan Peninsula is equal to or greater than 3 m along the coast; the tsunami produced by the fault was concentrated along the west coast of Teuri Island. The maximum value of 5 m was evident along the western coast of Teuri Island (Area 11 in Fig. 4-3) according to fault models HKD-2238, HKD-2239, near HKD-39 and previous models [4] and was located slightly to the north. The simulated maximum tsunami height was 2 m on Okushiri Island (Area 17 in Fig.

4-3) according to models HKD-3839, HKD-2239 and previous surveys [2]. This result is considered to be influenced by parameter settings of the fault model, such as rake and slip, set near

HKD-22 (Fig. 4-2) located to the south.

The values of K and κ from the simulated maximum tsunami heights and observational records are shown in Table 4-2. In any fault model in which the K value is >1.0, the simulated maximum tsunami height is less than the observed level. The K value of model HKD-38, which is not a linkage model, was 1.9, whereas that of the linkage models (HKD-2238, -3839 and -2239) were in the 1.1–1.5 range. When linkage is considered, the rupture area is wider because the seismic moment and average slip increase in accordance with the scaling law.

111

Table 4-2 K-κ values of the simulated maximum tsunami heights and observational records. Fault Model Name K κ

HKD-38 1.90 1.67 HKD-2238 1.43 1.66 HKD-3839 1.51 1.61

HKD-2239 1.17 1.61 Okamura 2005 1.34 1.64 Satake 1986_W2 1.36 1.63

Table 4-3 Fault parameters of the additional models.

Upper Fault Rakes Dips Slip Length Width Slip ave. depth Name ° ° ° km km m km 189 33.9 162 15.1 HKD-2239 0.0 45 90 16.9 3.00 189 30.3 167 28.9 189 33.9 162 15.1 0.0 45 90 16.9 2.62 189 30.3 167 28.9 HKD-2239_C 189 33.9 162 15.1 0.0 45 90 16.9 4.50 189 30.3 167 28.9 0.0 70 5.3 189 33.9 5.0 30 14.0 0.0 70 5.3 162 15.1 5.0 30 14.0 HKD-2239_2pt 0.0 70 90 5.3 3.40 189 30.3 5.0 30 14.0 0.0 70 5.3 162 28.9 5.0 30 14.0

HKD-2239 reproduced from Table 1; HKD-2239_C = large slip region model; HKD-2239_2pt = change tilt angle model

112

Satake1986

Fig.4-3 Maximum tsunami heights determined from tsunami propagation simulations compared with results from previous studies related to the western coast of Hokkaido.

MA Maximum tsunami heights (m) Fig. 4-4 Simulated maximum tsunami heights along the western coast of Hokkaido (blue lines) with tsunami traces (red circles) and tsunami heights of combined the maximum cases (gray lines) for each of the models.

113

4.1.7 Discussion

This study applied two additional fault models. Model HKD-2239_C is a large slip region model.

Model HKD-2239_2pt consists of two fault angle models located in the shallow and deep parts of the faulted region. Next, tsunami numerical simulation was conducted. The simulated tsunami height compared tsunami heights from the western coast of Hokkaido, which were based on the

Japan Tsunami Trace database [7]. When constructing a marine fault model, it is necessary to include uncertainties of the fault parameters because of imperfections in the model. To obtain these uncertainties, a parameter study was conducted to examine the sensitivity of the maximum tsunami heights to differences in fault parameters (Table 4-1) in the Shakotan-Oki earthquake source region.

To validate the fault model presented, we used the basic HKD-38 model that corresponded to the closest fault trace at the epicenter of the Shakotan-Oki earthquake. Reproducibility was evaluated by comparing the model with observational records and observations from the tide gauge records.

The linkage model HKD-2239 did not include a large slip region or changes in tilt angles. In the subsequent stage of the model, we used a model with two tilt angles. The shallow parts of the fault were high fault angles using the reflection seismic data. The deep parts of the fault showed a low fault angle, as defined by the Recipe. The simulated tsunami height, which compared tsunami heights from the west coast of Hokkaido using the Japan Tsunami Trace database [7], related to a large slip region fault model (HKD-2239_C in Fig. 4-5) and a change tilt angle model (HKD-

2239_2pt in Fig. 6). The change tilt angle model was set to 70° on the basis of reflection seismic data of 0–5 km. In the deepest parts, the tilt of the fault area was set to 30°.

This study verified the additional models based on HKD-2239. The validity of the fault model was examined by comparing tsunami simulation analysis with observational records of the

Japanese coastline (Table 4-3). Table 4-3 and Fig. 4-7 show simulation results of the additional models. The variation κ was approximately 1.6 for all models, which is reliable at 49 points at levels A, B, C, and D. As previously mentioned, model HKD-2239 does not include a large slip 114

region, and the average slip amount, based on a scaling law, was used to develop a strong ground

motion scaling law. Depending on the settings of the fault model, the resulting K value of 1.17

needs to be increased to improve reproducibility. In model HKD-2239, the resulting K value of

1.17 needs to be increased to reliability reproducibility. In model HKD-2239_C, the resulting K

value of 0.75 is 1.2 times greater than that placed in a large slip region model, which almost

exceeds the observational records. In model HKD-2239_2pt, the resulting K value was 1.03, which

of all the fault models is closest to 1.0. The effect of model HKD-2239_2pt is to increase the

tsunami height because of the proportional relationship between the fault area and the moment

magnitude used to estimate the fault slip. In this model, the fault area increases with a gradual

decrease in the slope angle of the deep parts of the faulted region. As the fault area increases, the

moment magnitude increases; thus, the amount of estimated slip increases.

Although fault models HKD-2239_C (large slip region model) and HKD-2239_2pt (change tilt

angle model) did not improve κ much greater than 1.45, K became as large as 0.75, and the large

slip region model and change tilt angle model were close to 1.03 and 1.0. Therefore, changing the

fault slope angle proved to be effective, thus demonstrating the importance of the shallow parts of

the fault using marine seismic data.

Table 4-4 Simulation results of the additional models (reliable at levels A, B, C, and D at 49 points).

Fault Model Name K Κ HKD-2239 1.17 1.61 HKD-2239_C 0.75 1.60 HKD-2239_2pt 1.03 1.61

HKD-2239 reproduced from Table 1 HKD-2239_C = large slip region model; HKD-2239_2pt = change tilt angle model

115

45°N

Fault area 44°N Large slip area

139°E 140°E

Fig. 4-5 Large slip region fault model (HKD-2239_C).

Seabed b apparent dips: c.a. 5 km 70 deg.

30 deg. Conrad discontinuity

Fig. 4-6 Example of a survey line cross-section with a change in the tilt angle model.

(a) P-wave velocities are used to identify the marine reflection seismic prospecting image collar; (b) fault modeling of change tilt angle case (HKD- 2239_2pt).

116

HKD-2239 HKD-2239_C HKD-2239_2pt

MA Maximum tsunami heights (m) Fig. 4-7 Simulated tsunami heights of the additional models.

4.1.8 Conclusions

This study aimed to verify the validity of a fault model based on the rupture area of the 1940

Shakotan-Oki earthquake. To test the model’s validity, tsunami propagation analysis was conducted on a fault model that had previously been presented [8, 9]. The maximum tsunami heights were used calculate the values of K and κ [5] to select the fault model; these data were obtained from the Japan Tsunami Trace database [7].

All of the simulation results in this study showed K values >1.0. Moreover, for six fault models, verification revealed that all cases of tsunamis were underestimated. The linkage model HKD-

2239 was not placed in a large slip region, and the average slip amount used was based on the same scaling law [1] as that used for the strong ground motion. The resulting K value was 1.17, which indicates the need to increase the reproducibility of the fault model. This value is higher than that placed in a large slip region model and almost exceeded the historical observation records.

117

These results require further investigation in terms of reproducibility, particularly for different fault model settings. In future work, seabed topography should also be incorporated to enhance the fault modelling.

The effect of the interlocking fault model was confirmed in this study. This study set the parameters of the fault model on the basis of reflection seismic data. Changes in the fault slope angle and in the shallow parts of the fault were identified by using marine seismic data.

Two fault angle model (HKD-2239_2pt), which located in the shallow and deep parts of the faulted region is not enough to explain the tsunami traces, precisely. However, two fault angle model is recommendation model.

118

4.2 The 1983 Nihonkai–Chubu earthquake

4.2.1 Back ground The fault model used fault to trace data from observational marine seismic records

(JAMSTEC) to approximate the fault plane in a rectangular model. This study, conducted as part

of the“Project” [4], constructed a fault model based on the geological fault information analyzed

by JAMSTEC [12, 13]. This study carried out an examination of the 1983 Nihonkai–Chubu

earthquake as a representative example of earthquakes generating tsunamis in the Sea of Japan.

In JAMSTEC, the review of low density data is carried out based on the small islands located

southeast of the main islands of Japan (Nansei Islands) using topography data and seismic activity.

In this study, an examination was conducted of the fault trace in consideration of topography data

and seismic activity with seismic reflection survey records about the Sea of Japan area.

There were 15 fault planes in fault traces (Fig. 4-8 a) of the Aomori offing shown in

JAMSTEC in 2014, and these analyses were all reverse faults, which were near the coastline Each

fault was described in the Red Relief Image Map prepared by No, et al. (2016) [7] (Fig. 4-8 b).

These fault traces were surveyed by the multi-channel seismic reflection survey and it was

confirmed that the relatively deep upper crust, lower part and even the sea topography corresponds

with the sea topography map (No, et al. (2016)) [14]. However, the AOM-09 fault, the most

characteristic fault, was in the epicenter of the main shock in the offshore faults, and the length of

the fault was approximately 55 km in the north and south direction. This survey used multi-channel

seismic reflection with high precision of the air gun and seismic reflection wave to the fault deep

part was provided. Including the cross section, a continuation of these faults was confirmed that

corresponded with the result (No, et al. (2012) [15]) of the Seismic intensive study in the vicinity

of the deformed zone in the eastern part of the Sea of Japan [5]. Ocean bottom seismographs from

the speed structured model were based on source points of the air gun, which were installed in

OBS. Wide-angle seismic reflection/refraction data analyzed the deep part at 15 km. These surveys

119

became clear to arrive at the Moho, which is a border of upper crust and lower crust, which was

provided from seismic reflection survey records.

.4.2.2 Methodology The fault model covers the epicentral area of the Nihonkai–Chubu earthquake, which occurred

on May 26, 1983, with a moment magnitude of Mw 7.7 (–7.9). The fault model used faults trace

data from observational marine seismic records (JAMSTEC) to approximate the fault plane in a

rectangular model. The validity of the fault model was then investigated by comparison with

observed tsunami records during the Nihonkai–Chubu earthquake and the maximum tsunami

heights by simulation based on the Sea of Japan coastline. These water level data were obtained

from tsunami propagation analysis using the fault model. This study was an examination of the

fault trace, which took into consideration topography data and seismic activity.

4.2.3 Fault traces The fault model used fault trace data (Fig. 4-8 a) from observational marine seismic records

(JAMSTEC). In addition, the existing fault AOM-09 is shown from observational marine seismic records. The fault used for the fault trace, which considered topography data and seismic activity for the modeling on Red Relief Image Map (Fig. 4-8 b). At Tohoku and Hirosaki universities, the unified hypocenter catalogs with more than 8,000 after the main shock and aftershock database. The catalogs related to about 2 months of the end of July 1983. The aftershock catalogs were located on the east edge of the Japan Basin. The fault plane is a distribution shape of "L" shape extending north and south. Fig. 4-8 c shows position of the main shock (stars) and aftershocks (Takagi, et. al, (1984) [16]).

To construct the primary model, this study defined the following guidelines for each parameter

setting, which considered epistemic and aleatory uncertainties. The position, length, and strike of

the fault were based on geological fault information from the Project. According to the strong

ground motion prediction method, referred to as the Recipe, for earthquakes with specified source

120 faults (2009) [1], one type of setting defines the basic values of a thrust fault, normal fault, and strike/slip fault as 45°, 60°, and 90°, respectively. The fault depth was 15 km and the fault plane was a rectangular model.

The combination and shape of the fault traces glided as the initial fault model and verified the fault modeling of the Nihonkai–Chubu earthquake source region such as large slip region and so on. The parameters of the fault model used in this study are listed in Table 4-5. A comparison of the previous study models of the Nihonkai–Chubu earthquake [5, 17-22] and the fault model used in this study is shown in Table 4-5. Fig. 4-9 shows primary model in the epicenter of the Nihonkai–

Chubu earthquake.

121

Table 4-5 Parameters of initial fault model.

a b c The fault traces

AOM-09 Fault

Fig. 4-8 Fault model used fault trace data (a, b) from observational marine seismic records (JAMSTEC).

a: The fault trace data provided in the cross sections across the Off-Aomori-Prefecture fault and the AOM-09 fault. b: The fault used for the fault trace, which considered topography data and seismic activity for the modeling on Red Relief Image Map. c: Aftershock distribution of the Hokkaido–Nansei–Oki earthquake, which occurred from May 31 to the end of June 1983 (Takagi, et. al, (1984) [16])

122

Table 4-6 Comparison of the previous models of the Nihonkai–Chubu earthquake with the model used in this study. a) Aida（1984）[5] Top Bottom L W D Lat. Lon. Strike Dip Rake Mw (km) (km) (km) (km) (km) 40.21 138.84 2.0 22 40 90 40 30 4.87 100 7.7 40.54 139.02 3.0 355 25 80 60 30 b) Tada（1984）[17] Top Bottom L W D Lat. Lon. Strike Dip Rake Mw (km) (km) (km) (km) (km) 40.20 138.80 0.0 13 20 90 60 40 3.5 120 7.7 40.73 138.96 0.0 335 20 90 60 30 c) Tanaka, et al.（1984）[18] Top Bottom L W D Lat. Lon. Strike Dip Rake Mw (km) (km) (km) (km) (km) 40.37 138.87 0.0 15 20 120 30 35 4.61 90 7.8 40.61 139.06 0.0 350 20 90 60 40 d) Satake (1985) [19] Top Bottom L W D Lat. Lon. Strike Dip Rake Mw (km) (km) (km) (km) (km) 40.37 138.87 0.0 15 20 120 30 35 4.61 90 7.8 40.61 139.06 0.0 350 20 90 60 40 e) Kanamori and Astiz (1985) [20] Top Bottom L W D Lat. Lon. Strike Dip Rake Mw (km) (km) (km) (km) (km) 40.10 138.70 - 21 30 115 150 40 2.0 150 7.7 f) Sato (1985) [21] Top Bottom L W D Lat. Lon. Strike Dip Rake Mw (km) (km) (km) (km) (km) 40.27 138.86 0.0 15 20 90 35 35 40.58 138.97 0.0 15 20 90 35 35 6.07 105 7.9 40.84 139.06 0.0 345 20 90 35 35 g) Kosuga, et al. (1986) [22] Top Bottom L W D Lat. Lon. Strike Dip Rake Mw (km) (km) (km) (km) (km) 40.79 139.09 0.0 345 25 90 40 40 40.65 139.03 0.0 20 25 90 30 40 4.60 100 7.8 40.39 138.91 0.0 20 25 90 30 40

123

Fig. 4-9 Primary model in the epicenter of the Nihonkai–Chubu earthquake.

124

3. 4.2.4 Calculation conditions of the tsunami propagation analysis

Table 4-7 shows the calculation conditions of tsunami propagation analysis．

. Table 4-7 Calculation conditions of tsunami propagation analysis. Governing equation Non-linear long wave theory Numeral solution Staggered Leap-frog

Calculation area Coast of the southern part of Hokkaido and the eastern part of Sea of Japan, offshore northern Honshu

Size calculation area From the open ocean, 1350 m, 450 m, 150 m (Minimum 150 m) Boundary condition Considering tsunami run up in the land area Transmission border nonreflective in the sea side Structures not consider Calculation time 12 hours Initial water level Sea bed movement calculated by Okada (1992) [9] Sea water level T.P. 0 m -2 Censored water depth 10 m Roughness coefficient 0.025

4.2.5 Validity of the fault model

There were three phases of the parametric studies as the topography model of 50 m. The first

stage inspected the parametric studies of the large slip regions positions, the second stage

calibrated a model to control the rise in water level in the northern region, and the third stage

compared the fault models of the previous studies.

(1) First stage inspection The fault model is the epicenter near the Nihonkai–Chubu earthquake as shown in Fig. 4-9. The

simulated tsunami heights compared tsunami heights from the parametric studies of the large slip

regions positions for comparing the maximum tsunami heights in the coastline regions with wave

source fault models as shown in Fig. 4-10. The combination of eight patterns of large slip regions

is shown in Fig. 4-11. In addition, this study used a model with two tilt angles. The shallow parts

of the fault showed high fault angles using reflection seismic data. The deep parts of the fault

showed a low fault angle, as defined by the Recipe as shown in Fig. 4-12. The change tilt angle

125 model was set to 70° based on reflection seismic data of 0–5 km. In the deepest parts, the tilt of the fault area was set to 30°.

Fig. 4-13 shows the comparison of the two plates of the tilt angle model. There is a tendency that tsunami height fluctuates widely in the area of the opposite shore of the region of the main two faults (Nakadomari and Fukaura) excluding the northern end (enclosed in a red dotted line in

Fig. 4-13). This study compared this tendency with tsunami trace heights in the coastal area (Fig.

4-14). Maximum tsunami heights were used to select reliability level A, B; these data were obtained from the Japan Tsunami Trace database [7]. The tsunami trace points in the inland area correspond to the 150-m mesh of the surrounding coast. The blue line in the figure is the calculation result of the fault model of Chubu - 11, and the gray line shows the maximum value of all calculation results at each evaluation point.

In the Matsumae region, by setting the large slip region of the northern fault to the south of the two main faults, the tsunami heights decreased and tended to approach the trace heights. Even in the Kojima region, similar to the Matsumae region, when the large slip regions of the northern fault plate was set to the southern end, tsunami tended to approach the trace heights. In the Nishi-

Tsugaru region, setting the large slip region of the northern fault plate to set to the southern end made the tsunami higher than the trace heights. Setting the large slip region of the northern fault plate to set to the northern end made closer to the trace heights. In the Noshiro region, the tsunami trace heights were locally high, and it was impossible to reproduce the tsunami trace heights in all fault models. In addition, in the Oga Peninsula region, the estimated tsunami heights tended to be higher than the tsunami trace heights in all fault models.

126

Fig. 4-10 Wave source fault models for the Nihonkai–Chubu earthquake.

Chubu-33 Chubu-11（L/R） Chubu-12（L/L） Chubu-21（R/R）

Chubu-13(L/C) Chubu-32（C/L） Chubu-23 (R/C) Chubu-31（C/R）

Fig. 4-11 Combination pattern of the examined large slip regions.

127

Seabed b apparent dips: c.a. 5 km 70 deg.

30 deg. Conrad discontinuity

Fig. 4-12 Example of a survey line cross-section with a change in the tilt angle model. (a) P-wave velocities are used to identify the marine reflection seismic prospecting image collar (b) Fault modeling of a change tilt angle (image).

128

Fig. 4-13 Comparison of the one/two plats of the tilt angle model.

Matsumae

Koima

Kitatsugaru

NIshitsugar u Noshiro

Oga

Tsunami trace ● Level A ● Level BC

Fig. 4-14 Comparison of maximum tsunami heights and tsunami trace heights in the two tilt angles model. Simulated maximum tsunami heights along the eastern part of the Sea of Japan (blue lines) with tsunami traces (level A: red circles; level B: orange circles) and tsunami heights of combined the maximum cases (gray lines) for each model.

129

(2) Second stage inspection Based on the first stage inspection, this study reduced the tsunami heights in the northern region and examined how to set a high fault model in the Noshiro region. If large slip regions were not set in the fault on the north side, the fault shape from JAMSTEC, the maximum tsunami heights in the coast, and the tsunami trace heights were compared when the large slip region’s position, area, aspect ratio were changed (4 cases: Fig. 4-15). Using Chubu - 11 as a comparative model, the amount of moment released by large slip regions is the basic model. The study case is as follows.

· Chubu - 11 W–S: The fault width was set to about half, and it was set only on the south side of the fault, to not change the area of large slip regions.

· Chubu - 11 W: The fault width was set to about half, and it was set on the south and north sides, to not change the area of large slip regions. · Chubu - 11h - S: Half the area of large slip regions was set only on the south side of the fault.

In the Matsumae region, in all cases the maximum tsunami heights (gray lines) surmount tsunami traces (level A: red circles; level B: orange circles). The maximum tsunami heights (blue lines) of the selected four cases reproduce tsunami traces. In the Kojima region, approaching the trace heights was the large slip region of the northern fault plate to shorten in the southern end (Chubu-11W (L/R),

11(L/R)). However, the average slip value of the whole fault when the large slip regions were not set was larger than the slip amount of the background area when large slip regions were set. Thus, the slip amount of the fault in the area closest to Kojima was getting larger, and the analyzed tsunami heights got higher. The simulated value tended to exceed the tsunami trace heights in the

Nishitsugaru region. This study approached the tsunami trace heights by not setting large slip regions in the fault on the north plate (Chubu-11h-S (/R), 11W-S (/R)). In this case, however, it was difficult to satisfy the conditions of trace heights in the Kojima and Nishitsugaru regions at the same time because the trace heights were higher in the Kojima region. In the Noshiro region, trace heights could not be reproduced. In the Oga Peninsula region, the trace heights reproduced in case of the

130 length of the large slip regions on the south plate were shorter (Chubu-11W (L/R), 11W-S (/R)).

Thus, the tsunami in Matsumae region explained the trace heights largely by the large slip regions of the north fault not being set (Chubu-11h-S (/R), 11W-S (/R)). In addition, in the Nishitsugaru region, the analyzed tsunami heights tended to be somewhat higher than the trace heights, and in the Noshiro region. The results that reproduced the trace heights obtained in this area were obtained by setting large slip regions, and the fault was not so significant even changed fault shape.

131

Tsunami Hight Tsunami Hight Tsunami Hight Tsunami Hight (m) (m) (m) (m)

Matsumae

Kojima

Nishitsugaru

Noshiro

Oga peninsula

Tsunami trace Tsunami trace Tsunami trace Tsunami trace ● Level A ● Level A ● Level A ● Level A ● Level BC ● Level BC ● Level BC ● Level BC

Fig. 4-15 Comparison of maximum tsunami heights and tsunami trace heights in large slip regions Simulated maximum tsunami heights along the eastern part of the Sea of Japan (blue lines) with tsunami traces (level A: red circles; level B: orange circles) and tsunami heights of combined the maximum cases (gray lines) for each model.

Max Max Max Max Chubu4 Chubu4 Chubu4 Chubu4 Level A Level A Level A Level A Level B Level B Level B Level B

132

(3) Third stage inspection

This model had longer faults than the Aida (1984) [8] model and the research organization meeting about large-scale earthquakes in the Sea of Japan (MLIT (2014)) [17] model, which was organized by the Japan Ministry of Land, Infrastructure, Transport and Tourism (MLIT). The moment magnitudes of the fault models set in the MLIT study is Mw 7.91 for two fault models and Mw 7.97 for three fault models. However, the size of the fault model of the study for the Aida model was Mw

7.7. The Sato (1985) [21] model used Mw 7.9, which corresponded with the previous study of the two fault models covered in this study. The length of the fault of this study is much longer than that of the previous study. In the previous study, the fault plane of the Nihonkai–Chubu earthquake was set to a low angle, so the fault width was larger. Compared with the fault shape used in this study and the aftershock distribution, the southern tip of the fault model by the JAMSTEC trace reached the southern part of the Oga Peninsula region. However, the aftershock distribution was within the northern part of the Oga Peninsula region. The northern end of the fault model reached the Matsumae

Peninsula in Hokkaido. The area where there were frequent aftershocks was off the coast of the southern part of Aomori Prefecture. Fig. 4-16 shows the fault model.

133

MLIT（2014）

（） Aida 1984

Proposed model

Fig. 4-16 Comparison of the proposed model, the Aida (1984) [5] model and the MLIT (2014) [17] model.

The fault length was set by considering the combination of aftershock faults. As a result, the length to be set was about 10 km longer than the fault length of the Aida (1984) [5] study, which was close to the previous study (about 110 km) by Aida (1984) [8]. The fault angle in the previous study, at 30°, was considered to be low, according to the Recipe [1] developed by the Headquarters for Earthquake Research Promotion.

The angle of inclination was changed so that the fault angle was the high angle in the shallow parts of the fault as high and the low angle in deep parts. The magnitude was Mw 7.61 and the scale became slightly smaller than that in the previous study.

The model setting of the fault showing with a change in the tilt angle model the result in Fig. 4-

17 is shown below. 134

· Chubu - 30 deg - BLR: Fault angles were set at 30°. The large slip region was set to the lower part of the fault on the north plate and the south plate.

· Chubu - 30 deg - BR: Fault angles were set at 30°. Only the fault on the south plate and large slip region on the lower part were set.

Chubu - 30 deg - BR - 2 pt: The fault angles with shallow parts and deep parts were changed to make an average of 30°. Only the fault on the south plate set the large slip region on the lower parts.

Chubu - 45 deg - BR - 2 pt: The fault angles with shallow parts and deep parts were changed, making an average of 45°. Only the fault on the south side set the large slip region on the lower parts.

The calculations of the four simulated maximum tsunami heights shown in Fig. 4-17 were almost the same as the tsunami trace heights in the Matsumae and Kojima regions, and the same as the tsunami trace heights in the Nishitsugaru region. A high degree of maximum tsunami height was calculated. In the Noshiro region, the maximum tsunami heights were lower than the tsunami trace heights. However, in the Oga Peninsula region, the same result as maximum tsunami height and the tsunami trace heights were obtained in Chubu - 30 deg – BR case. From these results, this study could explain the tsunamis from the Nihonkai–Chubu earthquake by considering the combination of faults based on aftershock distributions. The above discussion is an outline calculation using a topography model with a minimum mesh size of 150 m. In this study, a fault model was chosen for detailed calculation using a topological model of a 50-m mesh for a tsunami propagation analysis

135

4.2.6 Comparison with Previous Studies

The fault model that contributes to the detailed calculation of the model of 50-m mesh selected as the result of preliminary calculation of the tsunami propagation analysis using the 150-m mesh model. The calculated area of the tsunami prediction analysis is shown in Fig. 4-18. The calculation area was subdivided at a ratio of 3:1 from the open ocean to the coast. The mesh size of each calculation area was 1350 m, 450 m, 150 m, and 50 m from the open ocean toward the coast. The topological model from FY2005 was used in this project [1,2]. For the calculation conditions, the mesh size and the calculation time under the conditions shown in Table 4-8 were changed to 6 hours.

The results of the tsunami prediction analysis are shown in Fig. 4-19. Fig. 4-20 shows the maximum tsunami heights and trace heights. To validate the fault model in the present study, we compared the coastal maximum tsunami heights and trace heights of tsunami propagation analysis.

The maximum tsunami heights were used to calculate the values of K and κ [5] to select the fault model; these data were obtained from the Japan Tsunami Trace database [7] as shown Table 4-8.

The maximum tsunami heights were compared the values of K and κ based on the Aida (1984)

[8] fault model. The K values were in the 1.35 range and the κ values were below 1.58. For the fault model set in this study, high relative K–κ was obtained with the following two patterns. The first model of case 3 was set to a low angle (30°) on the south side large slip region model, which had K values in the 1.01 and κ values below 1.56. In the fault model of Aida (1984) [5] and the fault model set in this study, the value of K had a large range of fitness from 0.89 to 1.51. However, the value of

κ varied greatly from 1.53 to 1.62 (Table 4-8). According to the Nuclear Civil Engineering

Committee [7], these values are recommended for use under the following conditions: K values in the 0.95–1.05 range and κ values below 1.45. Results in this range were not obtained in these cases.

However, the value of K obtained multiple calculation results falling within the range 0.95

Regarding the reproducibility of the tsunami heights, this study demonstrated a fault model that reproduced the tsunami heights caused by the Nihonkai–Chubu earthquake.

136

Table 4-8 Fault models by using reliability of tsunami trace with K-κ value. Model name Fault area Angle Large slip K κ Aida ① 40°, 25° － 1.35 1.58 Chubu-AIDA (1984) ② Chubu-3f_30deg Aftershock 30° － 1.45 1.53 Upper end of the ③ Aftershock 30° 0.97 1.53 Chubu-3f-R South Fault Lower end of the ④ Chubu-30deg-BLR Aftershock 30° North and South 0.92 1.58 Fault Lower end of the ⑤ Aftershock 30° 1.01 1.56 Chubu-30deg-BR South Fault ⑥ Chubu-30deg-2pt Aftershock 69°, 22.6° － 0.98 1.60 Lower end of the ⑦ Aftershock 69°, 22.6° 0.89 1.62 Chubu-30deg-BR-2pt South Fault ⑧ Chubu-3f-2pt_45deg Aftershock 69°, 37.4° － 1.51 1.59 Upper end of the ⑨ Aftershock 69°, 37.4° 1.32 1.56 Chubu-3f-R-2pt South Fault Lower end of the ⑩ Aftershock 69°, 37.4° 1.38 1.60 Chubu-45deg-BR-2pt South Fault

137

Red 50 m mesh Blue 150 m mesh Green 450 m mesh Whole area 1350 m mesh

Fig. 4-18 Calculation area of the tsunami prediction analysis for each mesh size.

138

Fig. 4-19 Maximum tsunami propagation analysis results (T.P. 0 m).

① ② ③ ➃ ⑤ ⑥ ⑦ ⑧ ⑨ ⑩

Fig. 4-20 Tsunami traces (red) and tsunami propagation analysis results at the eastern part of the Sea of Japan in the Nihonkai–Chubu earthquake (Recommended model within the red dotted frame).

139

4.2.7 Conclusions

This study validated the fault model in the source area of the 1983 Nihonkai–Chubu earthquake.

Validity verification was conducted using tsunami propagation analysis with the comparison of the project [4] fault models and previous models [5, 17-21]. Observation records of maximum tsunami heights, trace heights and geodetic data on the coastline of K–κ were obtained from the tsunamis, and the scale of the tsunamis was quantified and evaluated.

In the selection of the fault model were detailed for the calculations by parametric studies. This study examined the modeling method of fault data obtained in the JAMSTIC project. The following results were obtained.

1) For the fault model two cases were prepared: a constant fault angle and a case where the

fault angle was high in the shallow part and the low in the deep part. In addition, a parametric

study was conducted with eight patterns of large slip region positions. As a result, there was a

difference in the fault parameter setting in the area near the opposite shore, which was the

boundary between the north and the south faults. These areas increased the variation in coastal

tsunami height rise. By setting the large slip region of the fault on the north side from the south,

the maximum tsunami heights in the Matsumae region approached the trace heights. In the

Noshiro region, trace heights were locally high, so this study could not reproduce them.

2) Based on the first stage inspection, this study performed multiple patterns and parameter

studies of fault models with different large slip region setting methods. This study considered a

fault model that sustained tsunami heights in the northern part of the calculation area low and

gathered high waves in the Noshiro region. As a result, the trace heights in the Matsumae region

could be explained largely by not setting a large slip region at the north fault. Regarding Noshiro's

surroundings, it was impossible to achieve a tsunami height that could reproduce. In the Noshiro

region, it was reported that a soliton wave was observed by the previous study, which may have

possibly resulted in local high tsunami heights. 140

3) Based on the first stage to third stage inspections of this study, the difference between the

fault model in the previous study and the aftershock distribution of the Nihonkai–Chubu

earthquake were compared. As a result, a fault model referencing the aftershock distribution was

set up. Tsunami propagation analysis was carried out and compared where the fault angle was set

low (30°) and where it was normal (45°). The fault angle was changed in the tilt angle model.

This study examined how to set the large slip region. As a result, by setting the fault model by

considering the aftershock distribution, we generally explained the tsunami trace heights caused

by the Nihonkai–Chubu earthquake. This study selected the fault model for verification by

quantitatively judging the fitness of the model using the result of K–κ parameter study.

In the tsunami propagation analysis of the Nihonkai–Chubu earthquake, this study used the 50-m

mesh topographic model based on the fault model selected by the above parameter study. The

reproducibility of the tsunamis created by Nihonkai–Chubu earthquake were investigated.

4) The maximum tsunami heights were compared the values of K and κ based on the Aida

(1984) [8] fault model. The K values were in the 1.35 range and the κ values were below 1.58.

・K–κ was obtained with the following two patterns among fault models set in this study. a) The slope angle was set low (30°) and a large slip region was set only on the south fault plate in the

north end portion. b) In the slope angle was low, the large slip region was set from the north end of the south fault plane.

For a), K–κ obtained was K = 0.97, κ = 1.53. For b), K = 1.01, κ = 1.56.

5) In the fault model of Aida (1984) [5] and the fault model set in this study, the value of K had

a large range of fitness from 0.89 to 1.51. However, the value of κ was 1.53 to 1.62, which was a

large difference. As a criterion of conformity in the Civil Engineering Society Nuclear Power

Civil Engineering Committee, κ <1.45 was set. However, results falling within this range were

not obtained in this study. For K, several calculation results falling within the range of 0.95

<1.05 were obtained.

141

6) Based on the fault data obtained by referring to geological, topological and seismological

data, such as aftershock distribution, it is possible to estimate tsunami height past earthquakes.

The possibility of explaining the tsunami trace heights caused by an earthquake was demonstrated.

142

Reference [1] Strong ground motion prediction method ("Recipe") for earthquakes with specified source faults (2009), Headquarters for Earthquake Research Promotion [in Japanese]. [2] Irikura K. and Miyake H. (2001), Prediction of Strong Ground Motions for Scenario Earthquakes, J., of Geography, 110(6), 849-875.

[3] Satake K (1986), Re-examination of the 1940 Shakotan-oki earthquake and the fault parameters of the earthquakes along the eastern margin of the Japan Sea, Phys. Earth and Planetary Int, 137- 147, 1986.

[4] Project for the Comprehensive Analysis and Evaluation of Offshore Fault Informatics (2015), The Headquarters for Earthquake Research Promotion, [in Japanese]. [5] Aida, I. (1978), Reliability of a tsunami source model derived from fault parameters, Journal of Physics of Earth, Vol.26, pp. 57–73.

[6] Tsunami Assessment Method for Nuclear Power Plants in Japan (2006), The Tsunami Evaluation Subcommittee, The Nuclear Civil Engineering Committee, JSCE (Japan Society of Civil Engineers), 321p.

[7] Japan Tsunami Trace database, Tsunami trace height information, International Research Institute of Disaster Science (IRIDeS), Tohoku University,

[8] Tanioka, Y. and Satake, K. (1996), Tsunami generation by horizontal displacement of ocean bottom, Geophys. Res. Letters 23, 861-864.

[9] Okada, Y. (1992): Internal Deformation due to Shear and Tensile in a half-space, Bull. Seismol. Soc. Am., 85, 1018–1040.

[10] Geospatial Information Authority of Japan, Measurement of the sea level at a tide station, [in Japanese]

[11] Miyabe N (1941), Tunami associated with the Earthquake of August 2, Bull. Earthq. Res. Inst., Univ. Tokyo, 104–114 [in Japanese].

[12] Kaneda, Y., Takahashi, N., Oikawa, N.,Ohsumi, T. and Fjiwara, H. (2014), Comprehensive Analysis and Evaluation of Offshore Fault Informatics，Seismological Society of Japan, autumn meeting in 2014.[in Japanese with English abstract]

[13] Ohsumi, T., Norimatsu, K.,, Matsuyama, H. and Fujiwara, H. (2015), Consideration of Fault Modelling for the Japan Sea Area based on the “Off Shore Faults Research Project”, Seismological Society of Japan, autumn meeting in 2015. [in Japanese with English abstract]

[14] No, T., Hiramatsu, T, Sato, T., Miura1, S., Chiba, T., Kamiyama, S., Iki, S. and Kodaira, S. (2016)，Red relief image map and integration of topographic data in and around the Japan Sea.

143

JAMSTEC Report of Research and Development, 22, p. 13-29, doi: 10.5918/jamstecr.22.13. V [in Japanese with English abstract]

[15] No, T., Hiramatsu, T, Sato, T., Kodaira, S. and Kaneda, Y. (2012), Seismic Reflection Survey in the eastern margin of the Japan Sea, Ultrasonic Technology, 24, 6, 15-20. [in Japanese]

[16] Takagi, A. Hasegawa, A., Saijyo, T., Yamamoto, A. and Misada, M., et al. (1984), General study of the disaster by the 1983 Nihonkai–Chubu earthquake, 2.2 Seismic activity before and after the main shock ， Ministry of Education’s Res. Grant Program (No.58022002) , Catastrophic failure disaster results of research of Research in Natural Disaster Report, pp. 24- 30. [in Japanese]

[17] Tada, T. (1984), Nihonkai–Chubu earthquake and crustal movements, The Earth Monthly, Vol. 6, pp.18-21. [in Japanese]

[18] Tanaka, K., Sato, T., Kosuga, M. and Sato, H. (1984), General study of the disaster by the 1983 Nihonkai–Chubu earthquake, 2.4 Characteristic of Nihonkai–Chubu earthquake， Ministry of Education’s Res. Grant Program (No.58022002) , Catastrophic failure disaster results of research of Research in Natural Disaster Report，pp. 39-45. [in Japanese]

[19] Satake, K. (1985), The mechanism of the 1983 Japan Sea earthquake as inferred from long- period surface eaves and tsunamis, Physics of the Earth and Planetary Interiors, 37:249-260.

[20] Kanamori, H. and Astiz, L. (1985), The 1983 Akita-Oki Earthquake (Mw = 7. 8) and Its Implications for Systematics of Subduction Earthquakes, Terra Scientific Publishing Company (Terrapub), Tokyo, Japan. Earthq. Predict. Res. 3 (1985) 305 317.

[21] Sato, T. (1985), Rupture Characteristics of the 1983 Nihonkai Chubu (Japan Sea) Earthquake as Inferred from Strong Motion Accelerograms, J. Phys. Earth, 33:525-557.

[22] Kosuga, M., Ikeda, H., Kamazuka, Y. and Sato, H. (1986), Static Fault Model of the 1983 Nihonkai-Chubu (Japan Sea), Earthquake as Inferred from Aftershock Distributions, Crustal Deformation, and Tsunami Data，Journal of the Geodetic Society of Japan，Vol.32, No.4, 290- 302.

[23] Investigation study meeting about the large-scale earthquake in the Sea of Japan (2014), Report of the investigation study meeting about the large-scale earthquake in the Sea of Japan. [in Japanese]

144

Chapter 4 Development of a Real-Time Damage Estimation System for Embankment Using Earthquake Early Warning Information

1. Motivation

In cases where river embankments are damaged by an earthquake, any damage could be exacerbated by subsequent tsunami run-up. Following the occurrence of an earthquake, it takes time to detect damage done to linear structures such as river embankments. Moreover, because of the characteristics of seismic waves, it is difficult to identify in advance those areas likely to be affected most. Based on both the seismic intensity distribution identified immediately following an earthquake early warning and the seismic intensity measured by seismographs, the application proposed herein could be used to identify potential disaster sites. Therefore, the application could be a valuable asset in the coordination of initial disaster response in regions where tsunamis occur frequently.

Specifically, the application has two primary elements: (1) issue an emergency earthquake bulletin for affected basins and (2) estimate the magnitude of embankment subsidence using relational expressions between seismic intensity and subsidence. Based on assumed disaster areas, the proposed application could be used to select areas of priority for surveying and to identify evacuation routes immediately following an earthquake. By assuming arbitrary hypocentres and estimating the potential damage by virtual earthquakes, river administrators could consider necessary emergency measures in advance.

Keywords— 2011 off the Pacific Coast of Tohoku Earthquake, Earthquake Early Warning (EEW), embankment, settlement

145

2. Introduction

Certain limitations of the Earthquake Early Warning (EEW) system in Japan became apparent

during and after the 2011 off the Pacific Coast of Tohoku earthquake. To overcome these problems,

this study propose a method that uses “Earthquake Damage Estimation Tables” for automatic

analysis and correction of detection errors evident when embankment settlement occurs. Thus,

embankment damage attributable to large earthquakes could be evaluated.

The proposed application offers a solution to the problems of underestimation of the magnitude

and seismic intensities of major earthquakes, and the fact that alarms were announced to areas

within a certain definite range of the epicentre without consideration of the seismic intensities

estimated using empirical equations. Our method could be used by licensed operators in the event

of major earthquakes. For example, it has been applied to a prototype of a decision-making support

system for an expressway company. This study have also considered the possibility of false alarms

in cases when two earthquakes might occur close to one another. The probability of false alarms

has been quantified and a decision-making support system developed to improve the operational

characteristics of the proposed disaster prevention system.

146

3. System Component

3.1 Earthquake Early Warning (EEW)

Given the hypocentre location provided by the EEW system, it is possible to acquire useful

earthquake information a few minutes faster than via traditional methods. The EEW (in Japanese,

“Kinkyū Jishin Sokuhō”) is issued just after the detection of an earthquake in Japan [1] (Figs. 2-1,

2-2). The warnings are issued by the Japan Meteorological Agency (JMA). The JMA also issues tips

on how to react to the warnings. The purpose of the EEW is to inform the public as soon as possible

that an earthquake has occurred and to alert the population to the estimated seismic intensity several

seconds (or more) before the arrival of the strong motions caused by the earthquake. However, in

areas close to the earthquake hypocentre, the EEW might not be transmitted before the arrival of the

strong motions.

Before the S-wave motion arrives, it is possible to utilize supplementary information, gained from

ground motion duration times, to help in decision-making procedures and to plan the actions that

should be adopted immediately after the earthquake.

147

Fig. 3-1 The Earthquake Early Warning system provides advance announcement of the estimated seismic intensities and expected arrival time of principal motion. http://www.jma.go.jp/jma/en/Activities/eew.html [1]

Fig. 3-2 Seismic stations used in the Earthquake Early Warning system.

148

3.2 Ground Amplification ratio

The major factor that controls site effects is the S-wave velocities of the surficial sediments. This study can therefore estimate the site amplification characteristics based on the S-wave velocities of the near-surface layers. The averaged S-wave velocity in the uppermost 30 m (AVS30) has been used widely to assess site conditions that can then be used to predict earthquake ground motion [2].

This study used an amplification map based on 30-second JEGM in grid cells of 7.5 arc-seconds

JEGM, which is openly available from the website operated by the Japan Seismic Hazard

Information Station (J-SHIS) [3] of the National Research Institute for Earth Science and Disaster

Prevention (NIED). A ground amplification ratio was calculated for AVS30 by Fujimoto and

Midorikawa [4] using velocity–amplification relationships.

3.3 Realtime Ground-Motion Monitoring System (Kyoshin Monitor)

Analysis of the EEW issued at the time of the 2011 Tohoku-oki earthquake reveals it was inadequate for emergency reporting of this earthquake because of the underestimation of its magnitude (i.e., a class M8.0 earthquake was expected according to the EEW information).

Therefore, our proposed system automatically analyses and corrects detection errors when the server receives earthquake information for events of magnitude >7.5 (Fig. 3-3). The correction for the detection error is performed using seismometer observation data recorded by the K-NET and KiK- net networks. These data are provided in real time by the ground motion monitoring systems (Fig.

3-4).

149

Fig. 3-3 Processing procedure of prototype real-time embankment damage estimation system using Earthquake Early Warning.

Fig. 3-4 Realtime ground motion monitoring system (Kyoshin monitor). http://www.kyoshin.bosai.go.jp/

150

3.4 Seismic Intensity Prediction Technique The seismic intensity prediction based on the EEW’s determined hypocentre uses the attenuation relationship with fault distance of Si and Midorikawa [5]. The “fault distance” is defined by subtracting half the estimated dislocation length from the hypocentre distance [6]. A ground amplification ratio is used, as shown in chapter II.B. Moreover, the relationship between averaged shear wave velocity and site amplification, as defined by Fujimoto and Midorikawa [7], is also adopted. This relation is obtained using the peak ground motions recorded at two nearby seismic stations. The KiK-net Kainan station, which is the nearest NIED station to the targeted river embankment, is used as a reference.

3.5 Estimation Method of Embankment Settlement

The method for estimation of embankment settlement uses “Earthquake Damage Estimation

Tables” that are calculated using seismic intensity and embankment settlement information. The

“Earthquake Damage Estimation Tables” are calculated by combining the results from analyses of the effective stress and previous earthquake damage information in relation to embankments within the Tokushima area. Thus, embankment damage due to large earthquakes can be evaluated.

Furthermore, these “Earthquake Damage Estimation Tables” could be upgraded and improved in the future based on calculations that are more reliable and refined.

3.6 Digital Japan Portal Web Site Systems

This system uses the GEBCO Digital Atlas, which the Geographical Survey Institute of Japan has provided to the Digital Japan Portal Website. The data files (containing, for example, predicted seismic intensities) should follow the Digital Japan Portal Website format [8], i.e., text (XML) format. An HTML file that contains these data is uploaded to an information disclosure server. The

Digital Japan Portal Website system is described in the HTML file by a set of API functions using

JavaScript. If a user accesses the HTML file on an information disclosure server via the Internet

151

(e.g., using a web browser), the data are visualized overlain on the background map of the

Geographical Survey Institute; thus, they can be examined in a convenient way.

3.7 Real-Time Damage Estimation Systems

Hypocentre information is acquired from the final announcement of the EEW. The system

automatically improves the estimated seismic intensity through consideration of the difference

between the estimated and the observed accelerations, as recorded by the real-time ground motion

monitoring system (Kyoshin Monitor).

The amount of subsidence at the top of a river embankment is evaluated from the results of the

estimated seismic intensity and the system automatically sends a notification to the river

administrator.

The calculated results for a certain earthquake can be seen on the analysis server, which outputs a

file in a format that can be displayed by the Digital Japan Portal Website system. Thus, a river

administrator can peruse the analysis results by accessing this server using an Internet web browser.

152

3.8 Information output function

The content provided by an information disclosure server is shown in Fig. 3-5. A river administrator can access the information disclosure server using an Internet web browser. The accessed page has the composition shown in Fig. 3-6. When powerful tremors that exceed a pre-set seismic intensity threshold occur, the top page will provide links to the data table (Fig. 3-7). This table structure is automatically added to a list.

The page displaying the ground amplification ratio is shown in Fig. 3-8. According to the magnitude of the ground amplification ratio, various levels are distinguished using different colours.

The information output includes information-by-forecast web pages for seismic intensity and the amount of embankment subsidence. On the information-by-forecast web page for seismic intensity

(Fig. 3-9), the related map displays the seismic intensity distribution in coloured boxes along the river. Using the checkboxes in the lower part of the screen, the user can chose whether to display the river embankment and/or the coloured boxes. The estimated amount of embankment subsidence is shown on its own information-by-forecast page (Fig. 3-10).

153

Fig. 3-5 Operational procedure of real-time embankment damage estimation system using Earthquake Early Warning.

Fig. 3-6 Information disclosure server. Fig. 3-7 Top page.

154

Real-Time Damage Estimation System for Embankment Real-Time Damage Estimation System for Embankment

Ground amplification ratio Ground amplification ratio

Fig.3-8 Ground amplification ratio: (Left) entire area and (Right) embankment only.

Real-Time Damage Estimation System for Embankment Real-Time Damage Estimation System for Embankment

Ground amplification ratio Ground amplification ratio

Fig. 3-9 Estimated intensity Map. Fig. 3-10 Estimated settlement.

155

4. Conclusion

In this study, this study developed a prototype decision-making support system that enables river administrators to take decisions quickly in response to the occurrence of an earthquake. As a proactive measure, river embankments require thorough safety checks immediately after the occurrence of a large earthquake, even if the structures are designed with seismic resilience. To support emergency checks after the occurrence of a large earthquake, the prototype decision-making support system provides an estimation of damage to river embankments (and other structures) based on EEW seismic parameters.

Some limitations of the EEW system were also examined. While these limitations have been identified, their influence and ways in which their effects could be mitigated have not been discussed thoroughly. Underestimation of the magnitude and seismic intensities of large earthquakes, as well as the occurrence of false alarms that can happen when two large events occur in close spatiotemporal proximity, were found to be serious problems for EEW users following the 2011 Tohoku-oki earthquake. Similar problems might recur when the next earthquake strikes on the Nankai Trough.

Some countermeasures, which could be performed by operators under license of the ground motion forecasting services were proposed and discussed.

“The Seismic Performance Evaluation Chart-system for Coastal Structures” was designed by the

Ministry of Land, Infrastructure, Transport and Tourism of the Kinki Regional Development Bureau

[9]. This system can be appraised for embankment settlement, performed using “Earthquake Damage

Estimating Tables”, which are calculated with high accuracy.

In the future, the Japan Real-time Information System for earthQuake (J-RISQ) [10], which can estimate the real-time seismic intensity and exposed population, will be implemented throughout the islands of Japan using 250-m mesh boxes (Fig. 4-1). This system, which is intended to provide earthquake vulnerability analysis and other related information as quickly as possible after the occurrence of a major earthquake, is under construction at NIED. In the event of a large earthquake, 156

J-RISQ will be able to assist with estimations of damage, exposed population and other hazards by combining observational data such as real-time ground motion data recorded by the K-NET and KiK- net networks, site-specific intensity and ground motion amplitude data, as well as relevant population and building-related data. Therefore, as a contribution, this study would like to propose the integration of our real-time embankment damage estimation system with J-RISQ. Also, Estimated acceleration damage extent through Fragility curves [11,12], would like to propose the integration of our real-time building damage estimation systems.

Fig. 4-1 Real-time seismic-intensity exposed-population estimation system Aoi, S., et al. (2013)[10].

Is distribution (Left): Seismic Intensity (Is) distribution is estimated from the observed data (circles) of NIED K-NET, KiK-net, JMA, and local governments that had been collected by 2013/04/13 05:43:32. Estimated Is for Major Cities (Right): The histogram shows frequency distribution of estimated Is derived from interpolation of the observation with 250-m mesh. The daytime and nighttime correspond to 9:00-18:59 and 19:00- 8:59, respectively. The distance is measured from the epicenter to the center of the municipality.

157

References [1] Earthquake Early Warnings, EEW, http://www.jma.go.jp/jma/en/Activities/eew.html [2] Mataoka, M. and Wakamatsu, K. (2008), Amplification map based on 30-second JEGM in grid cells of 7.5 arc-seconds JEGM, Ground-condition map containing the attributes of geomorphologic classification latitude 11.25 arc-seconds longitude for all of Japan, intellectual property management system: H20PRO-936. [3] Japan Seismic Hazard Information Station, J-SHIS, (2020.01.10 access) http://www.j-shis.bosai.go.jp [4] Fujimoto,K. and Midorikawa, S. (2006), Relationship between Average Shear-Wave Velocity and Site Amplification Inferred from Strong Motion Records at Nearby Station Pairs, Journal of Japan Association for Earthquake Engineering, Vol. 6, No. 1, pp. 11 - 22. [5] Si, H., and Midorikawa, S. (1999), New Attenuation Relationships for Peak Ground Acceleration and Velocity Considering Effects of Fault Type and Site Condition, Journal of Structural and Construction Engineering, 523, pp. 63 - 70. [6] Utsu, T. (1977) Seismology, kyoritsu_pub． [7] Fujimoto, K. and Midorikawa, S. (2005) Empirical Method for Estimation J.M.A. Instrumental Seismic Intensity from Ground Motion Parameters Using Strong Motion Records during Recent Major Earthquakes, Journal of Social Safety Science, No. 7, pp. 241 – 246. [8] Digital Japan Portal Web Site Web.NEXT http://portal.cyberjapan.jp/site/mapuse/index.html [9] Kimura, H. (2008), Seismic Performance Evaluation Chart-system for Coastal Structures, Ministry Land, Infrastructure, Transport and Tourism Kinki Regional Development Bureau, Kobe Research and Engineering office for Port and Airport [in Japanese] (2020.01.10 access) http://www.mlit.go.jp/chosahokoku/h18giken/program/kadai/pdf/new/new4-01.pdf [10] Aoi, S., Nakamura, H., Kunugi, K., Suzuki, W. and Fujiwara, H. (2013), Real-time seismic- intensity exposed-population estimation system (J-RISQ), SSJ 2013 Fall Meeting, 2013. [11] Ohsumi, T. and Hazarika, H. (2013), Development of Real-Time Damage Estimation System for Embankment Using Earthquake Early Warning, 2013 International Conference on Signal- Image Technology & Internet-Based Systems, pp.854-859. [12] Ohsumi, T. and Hazarika, H. (2013), Flash Report on Damage Caused in Mexico City, Mexico, by the 2017 Puebla-Morelos Earthquake, Proceedings of the 1st International Conference on Press-in Engineering, 2013 International Conference on Signal-Image Technology & Internet- Based Systems, pp.854-859.

158

Chapter 5 Investigation of the effects of submarine landslide duration on tsunamis -Related to the 1771 Yaeyama/Meiwa earthquake with tsunami propagation analysis-

1. Motivation The Japanese Ministry of Education, Culture, Sports, Science, and Technology initiated the “Project for the Comprehensive Analysis and Evaluation of Offshore Fault Information” (The Project) in 2013. The objective of The Project is to contribute to assessment of earthquake and tsunami hazard in Japan’s seas by collecting off-shore fault information. The Project collects and consolidates reflection seismic data using the latest data processing technology where offshore fault information is missing. The data are collected from various institutions that have conducted fault evaluation surveys in the seas around Japan. This study reanalysed the data and evaluated the faults using a consistent approach. NIED has reanalysed the data of underground structures that were acquired by multiple institutions for the Japan Agency for Marine-Earth Science and Technology (JAMSTEC) and it has analysed offshore fault information in the area of the southern Nansei Islands. In assuming the tsunami caused by the 1771 Yaeyama/Meiwa earthquake was associated with a submarine landslide, the present study reconstructed the shape of the shelf area around Yaeyama Islands and calculated the seabed deformation before the submarine landslide occurred. The maximum tsunami heights in the coastal area were calculated by applying the change directly in tsunami propagation analysis. The duration of the collapse of the assumed submarine landslide, about which little is known, was calculated using the Newmark sliding block method. A map of the geological structure of the seabed revealed that older Paleozoic age rocks are exposed from the Paleogene rocks on the slope. This study analysed the propagation of tsunamis caused by landslides using the duration of the landslides and this study found that the influence of duration has considerable effect on maximum coastal tsunami height.

Keywords— Yaeyama/Meiwa earthquake, submarine landslides, Shimajiri-mudstone, Newmark sliding block method

159

2. Introduction

Offshore faults exist in the sea areas surrounding Japan. However, tsunamis due to submarine landslides are generally considered much bigger than caused by offshore fault deformation. It is believed that the tsunami associated with the 1771 Yaeyama/Meiwa earthquake was caused by a submarine landslide. This tsunami caused devastating damage and over 10,000 fatalities in the

Miyako Islands archipelago.

In 2018, a strong earthquake with a moment magnitude of 7.5 occurred on the island of Sulawesi,

Indonesia. This earthquake caused extensive liquefaction and, subsequently, submarine landslides characterized as liquefaction-induced flows caused tsunamis. The features of this earthquake have been referenced in explaining historical tsunami records.

For example, Imamura et al. (2001) [1] used information from this event to explain the historical tsunami records of the 1771 Yaeyama/Meiwa earthquake that devastated large areas of the Miyako and Yaeyama islands in the southwestern part of the Ryukyu Arc. It was assumed that this event was caused by an earthquake with a maximum class magnitude of Mw 8.8. Conversely, based on detailed bathymetric and reflection seismic data, Okamura et al. (2018) [2] proposed a model for the source of the tsunami as a large-scale collapse on the accretionary prism along the trench, whereby a submarine landslide on the prism caused the aseismic tsunami. Using the circular slip method,

Hiraishi et al. (2001) [3] also attributed the source of the tsunami to submarine landslides through numerical calculations based on an assumed duration for slope failure of 30–90 s.

This study investigated the effect of seabed structure on the generation of tsunamis triggered by submarine landslides through investigation of the shape of the island shelf in the Yaeyama area. By applying changes directly in the tsunami propagation analysis, this study calculated the maximum tsunami heights in the coastal area. The duration of the collapse of the assumed submarine landslide, about which little is known, was calculated using the Newmark sliding block method.

The hazard presented by storm surges and flooding associated with global warming is increasing at a rapid rate. Therefore, it is highly important that regional studies contribute to disaster mitigation 160 by securing safe urban function, while enhancing awareness of disaster prevention without causing concern for local residents. Internationally, it is important for Japan to demonstrate its contribution to disaster prevention as a nation facing many serious hazards.

161

3. Aseismic tsunamis

In relation to the 1771 Yaeyama/Meiwa earthquake, evidence of a submarine landslide has been confirmed through precise topographic investigation by Matsumoto and Kimura (1993) [4], who found it impossible to explain the height of tsunami traces based only on earthquake-scale fault displacement. Moreover, Hiraishi et al. (2001) [3] used the circular slip method to attribute the tsunami, associated with the 1771 Yaeyama/Meiwa earthquake, to mass movements generated by earthquake-induced landslides. In this study, this study introduced a submarine landslide model using the circular slip method applied to the calculation of slope stability.

3.1 Classification of tsunami sources

A seismic sea wave (i.e., a tsunami) is a water level fluctuation attributable to seabed crustal movement due to fault motion at the time of an earthquake. An aseismic tsunami is classified as a tsunami caused by other factors. Representative examples of aseismic tsunamis are shown in Table

3-1. Table 3-1 Sources of aseismic tsunami events (after Aylsworth, J.M. (2012) [5])

162

3.2 Water level fluctuation associated with volcanic activity

Active volcanic eruptions can cause aseismic tsunamis, as can the collapse of a volcanic edifice triggered by an earthquake. The Sakurajima eruption is an example of an undersea volcano near

Japan known to have caused a tsunami (Toshiji and Ueda, 1996) [6]. Similarly, the 1883 eruption of

Krakatoa in Indonesia was another such event (Normanbhoy N. and K. Satake, 1995 [7], Lander, J.

F. and P. A. Lockridege, 1989 [8]). The onshore volcanic eruption of Komagatake (1640) in

Hokkaido is an example where a tsunami was triggered by a landslide caused by the eruption

(Yoshimoto et al., 2008) [9], as also occurred following the Oshima-Oshima event in 1741 (Satake,

2007) [10], and the Matsubara Peninsula Mayuyama collapse in 1791–1792 (Tsuji and Hino, 1993)

[11].

3.3 Water level fluctuation caused by mass movements such as landslides

Mass movements attributable to earthquakes can trigger landslides that produce tsunamis. In

Japan, the Surugawan earthquake in 2009 was one such event (Baba, et al., 2012) [12], as was the

1958 Lituya Bay earthquake in Alaska (USA) (Friz, et al., 2009 [13], Weiss, et al., 2009 [14]).

3.4 Water level fluctuation due to falling meteorites

Although it is unknown whether the occurrence of a tsunami in Japan has ever been caused by meteorite impact, it is thought that the extinction of the dinosaurs at the Cretaceous–Tertiary boundary was related to meteorite impact [15].

163

4. Classification of submarine landslides

Ohyagi (2004) [16] cited Heezen and Ewing (1952) [17] who introduced the idea of a submarine landslide based on a study of the sea floor following the Grand Banks earthquake in southern

Newfoundland (Ms 7.2) in 1929, during which an undersea cable was broken by a submarine landslide.

Macroscopically, a submarine landslide can be classified into one of five types of flow: a sediment gravity flow, turbidity current (turbulence), fluidized sediment flow, grain flow, or debris flow. Gibo et al. (2003) [18] classified the collapse form of Shimajiri-mudstone rock slopes as follows: strongly weathered fractured mudstone, crushed mudstone, normal consolidated mudstone, and mudstone containing a weak plane. This classification was directed to inland slopes and weathering of the seabed slope was not considered. Excluding the strongly weathered fractured mudstone and crushed mudstone, failure of the sedimentary rock can be assumed to have the strength degradation of normal consolidated mudstone. Thus, assuming that surface rupture contributed to the failures, this study assumed weak plane strength, this study assumed the surface of rupture weak-plane strength, and set the surface of rupture.

164

5. Tsunami boulders within and around historical tsunami trace points

Representative points of historical tsunami traces, which were arranged by Hatori (1988) [19] based on Kato (1987) [20] and Kawana et al. (1987) [21] and so on as shown, were compared with calculated values for Ishigaki Island (ten points) and Miyako Island (three points).

Kawana et al. (1994) [21] were dated much older than the age of the 1771 Yaeyama/Meiwa

Tsunami about 200 yr. BP. The boulders suggest that the Yaeyama Islands had been affected by devastating tsunamis around 600, 1,100, 2,000 and 2,400 yr. BP during the last 3,000 years with intervals of several hundred to one thousand years in the study area.

5.1 Ishigaki Island

1) Ryukyu limestones in the Miyara Bay, southern Ishigaki Island: Fig. 5-1

In the vicinity of the airport in southern Ishigaki Island, Ryukyu, limestone thought to be

tsunami boulders is scattered in the Miyara Bay. At that point the 1771 Yaeyama/Meiwa tsunami

has reported 30 m tsunami traces.

165

（GSI Web Map used）

Fig. 5-1 Ryukyu limestones in the Miyara bay, southern Ishigaki Island. (Aerial photo taken by T. Ohsumi on January 25, 2019)

166

2) Tsunami Fuuishi (Ohama Elementary School): Fig.5-2 to 5-4

"Tsunami Fuuishi" at the Sakihara Park at Ohhama, which is said to have been moved by the

tsunami, has a maximum diameter of ~11.5 m, a depth of ~ 9.6 m, and a height of ~ 6 m. It is

estimated that it was launched by the Skishima Great Tsunami about 2000 years ago from the

dating result of the corals adhering to the surface.

The height of "Tsunami Ufuishi" is of 10.0 m elevation (data source: DEM 5 B), and it is ~150

m distance from the coast of the Miyara Bay.

（GSI Web Map used）

Fig. 5-2 "Tsunami Ufuishi" at the Sakihara Park of Ohhama.

Fig. 5-3 Close up of "Tsunami Ufuishi". There are fossil of corals. (photo taken by T. Ohsumi on January 22, 2019)

167

Fig. 5-4 Natural Monument (Nationally Designated).

Information Plate; The name "Tsunami Ufuishi" (Tsunami boulder) was applied by the late local history researcher Mr. Kiyoshi Makino. Originally, Makino seemed to believe that it was the tsunami boulder in the 1771 Great Yaeyama/Miwa Tsunami which caused great damage mainly in Ishigaki Island in 1771. However, from the results of the carbon dating of the corals adhering to the surface, it became clear that it was launched to its present place by the Sakishima tsunami about 2,000 years ago. Also, in a study that specializes in paleomagnetism, although this stone did not move greatly at the time of the 1771 Great Yaeyama/Miwa Tsunami, the possibility that the terrestrial magnetism was moved due to rotation by the power of the wave has been pointed out. In this way, as a result of repeated scientific verifications, there is no doubt that it is a tsunami boulder, and there is also a moral element that the devastating tsunami hit the area before the 1771 Great Yaeyama/Miwa Tsunami. I understood that it was a tsunami boulder. In addition to this "Tsunami Ufuishi", "Takakoruseishi" (Ohhama), which is said to have been launched by a tsunami or located in the eastern cast of Isjgaki Island, Amataria-Suuari (Touri), Yasura-Ufu-Kane (Hirakubo) and Bari- Stone (Ibaruma) are listed as the "eastern cast of Isjgaki Island's tsunami boulders". It is specified as a Natural Monument (Nationally Designated). The four megaliths other than "Tsunami Ufuishi" were tsunami boulders from the 1771 Great Yaeyama Tsunami, according to the historical description “Ohnaminotoki-Kakumurano-Nariyukisyo” and the results of scientific inspections by carbon dating and so on. June 2016, Ishigaki City Board of Education Cultural assets section.

168

3) Coast of Miyara Bay: Fig. 5-5

In coast of Miyara Bay, Ryukyu limestone which seems to be tsunami boulders is scattered in

Miyara Bay. These tsunami boulders are shown by aerial photograph.

（GSI Web Map used）

Fig. 5-5a Ryukyu limestone in Miyara Bay.

Fig. 5-5b Ryukyu limestone in Miyara Bay.

Fig. 5-5c Ryukyu limestone in Miyara Bay. (photo taken by T. Ohsumi on January 24, 2019)

169

4) Taka Koruseishi: Fig. 5-6

"Taka Koruseishi" at Aza Ohham in Ishigaki City, Okinawa Prefecture, is also a Natural

Monument (Nationally Designated) the same as "Tsunami Ufuishi". The height of "Taka

Koruseishi" is 2.4 m (data source: DEM 5 B), and it is at a distance of ~ 100 m from the coastline.

It is thought that the stone, which was carried to "Koruse Utaki" by the Sakishima Great Tsunami

2,000 years ago moved to the north again ~ 600 m in the 1771 Yaeyama Tsunami.

（GSI Web Map used）

Fig. 5-6 "Taka Koruseishi " in Ohham in Ishigaki City height: 2.4 m elevation (data source: DEM 5 B) (photo taken by T. Ohsumi on January 22, 2019)

5) Matsumuto-House: Ryukyu Limestone: Fig. 5-7 Matsumuto-House is located in "Matsumuto-House" at Tonoshiro, inland 615 m from

Yashimacho port facility, where the 1771 Yaeyama/Miwa tsunami's tsunami boulder is placed. The

height is 7.7 m elevation (data source: DEM 5 B).

170

（GSI Web Map used）

Fig. 5-7 Tsunami boulder at "Matsutou-House" height: 7.7 m elevation (data source: DEM 5 B).

Information plate: a In the 1771 Yaeyama earthquake the boulders in the sea were launched onto the land by the great tsunami of the Miwaji era. At the moment, it is this tsunami boulder at "Matsutou-House".

171

Information plate: b Ishigaki Island and Miyako Island are related to the Yaeyama Earthquake which occurred in south-southeast of Ishigaki Island around 8 am on April 24, 1771. It was a devastating tsunami with about 12,000 deaths, and more than 2,000 houses swept away by overwhelming large waves.

In Ishigaki Island near the epicenter, the tsunami landed from the southeast of the island, and moved immediately along the river into the deep part of the island, instantly flooding the fields, houses and livestock towards Nagura Bay.

According to the historical documents, Miyara, in the village tsunami runup height reached as much as 85 m. With this tsunami, coastal rocks were lifted up onto the land, and in remote islands such as Kuroshima and Shinjyo Island, the aftermath of the tsunami swept away the whole island.

In Yaeyama, of the total population of 28,992 people, 9,313 died, that is 32% of the total population was lost. There was a total collapse of the eight villages of Maesato, Ohhama, Miyara, Shiraho, Nakayumi, Ibaruma, Yasura and Yarabu. Partial collapse took the seven villages of Arakawa, Ishigaki, Ohkawa, Tonoshiro, Hirae, Kuroshimaand Shinjyo. Especially great was the damage to Shiraho Village, 98% of the village population of 1,574 people were exposed to the wave. On the other hand, Miyaki Island was a bit far from epicenter, but still 12 villages were damaged and there were 2,548 victims. In addition, deaths due to crop damage, famine, pestilence, and other events that followed the tsunami continued, and the population of Yaeyama Islands dramatically declined. For this reason, forcibly relocated from the village where damage was little to village which was destroyed, village rebuilding was done. However, deaths due to starvation and infectious diseases continued after that, and the population of Yaeyama sharply declined. For this reason, in the village which has been destroyed, the village was forcibly resettled to the village with little damage, and the village is being rebuilt. However, the deaths due to starvation and infectious diseases, etc. continued afterwards, and in the Meiji period about 100 years later, the population of Yaeyama Islands has decreased to one-third before the earthquake. The severe head tax by the Shuri Kingdom also had a big influence on this.

Regarding the Meiwa Yaeyama Tsunami, it is summarized in detail in the report of the warehouse overseer "Kuramoto", the shogunate government's branch office.

(photo taken by T. Ohsumi on January 22, 2019)

172

6) Meiwa-Ohtsunami Victim Memorial Monument: Fig.5-8 to 5-15 Located in Ishigaki Cityi Aza Miyara, according to Hatori (1988) [19] it was said to be 30 m high at the 1771 Yaeyama Tsunami. This point is ~1 km from the Miyaragawa River and the height of this point is a 63.2 m elevation (data source: DEM5 B). According to "Ohnaminotoki -

Kakumurano - Nariyukisyo", the tsunami run-up was 84.8 m high at Ishigaki Island. The tsunami hit with three waves and the second wave was said to have been the biggest.

（GSI Web Map used）

Fig.5-8 Meiwa-Ohtsunami Victim Fig.5-9 Distance from the Monument to the Memorial Monument. Miyaragawa River ~1 km, from the coastline 1.2 km and altitude: 63.2 m elevation (data source: DEM5 B).

(photo taken by T. Ohsumi on January 23, 2019)

Fig.5-10 Contents of Inscription panels are shown below;

173

た。関、浄財と石垣市、団体のに語りつり、全遭難死亡者の御協力を仰いあわせずるこが命した人人を思うと竹富町、れていてことた。がでこくこのものできない与那国町並びにみ未曾有の人頭税制下のたまを合祀してそのとことし、続発出して、を念願し、に作、こ。き、垣島は災害ののそのこ飢餓、塔を建立しの八重山社会のい諸機半潰し、影響はまことに計り難い潰壊滅的打撃をうけ、たび有志相謀り、島内外各面まなお断腸の歴史が人口は年年減少の半潰あわせ伝染病など冥福を祈津波で永く後世遭難死亡者は度もくりか歩みて十三村、にが念を禁あるを一層困難なよる飢餓者、えり立ちたちまち島島村村を襲っ干き、群島一途をたどような音が。九三一三人にえした。加えてその地震が東北・ほかもの日（ばあり、に病死者もがあっり、とどろき、一七七一年四月二十四日）、八重山の東南海上に史上有名な八重山の黒島、乾隆後の達した。それ新城二村が凶が間もなく外の大波が止むと黒雲のた。石垣島の明和大波は三瀬まで潮がように午前八時ご東方にひる雷鳴のろ大一九八三年（昭和五八）こ四月二四日の天災から二一二年、こうして群島の狂瀾怒涛の津波は政治・なかで落経済・文化の中心地石

明和大津波遭難者慰霊碑建立期成会石垣島東岸と三十六年（古記禄大波之時各村之形行書に日本年号明和八年）三月十よれ

碑

南岸

で激甚を文

きわめ

、

の全

Fig. 5-11 Inscription：According to the "Ohnaminotoki - Kakumurano - Nariyukisyo", (April 24, 1771) Information plate Yaeyama, there was the Great Earthquake in Yaeyama Islands around 8 o'clock in the morning. When the earthquake had settled, a thunder-like sound roared in the eastern area of Ishigaki Island, soon the tide ebbed to the outer shores. The sea tides drifted soon toward northeast and southeast. Large waves from the ocean struck the streets like black clouds and also the town island of Shimamura village. Immediately, the wave was out of the Straits. A big wave struck like a black cloud from the northeastern and southeastern seas, and a storm covered the whole village. The waves repeatedly hit three times. This is the historically famous Meiwa-Ohtsunami of Yaeyama. The devastating tsunami fiercely struck the eastern and southern coast of Ishigaki Island. The total number of collapsed villages were 13. 2 villages that were half collapsed were on Kuroshima Island and Aragusuku Island. The number of deaths climbed up to over 933. Thus, Ishigaki Island, the political, economic and cultural center of the archipelago, was hit by a devastating tsunami. There was also hunger, disease and death due to subsequent harvest, epidemic diseases and so on. The population decreased year by year. Following this, the "Jinto" taxable trading of the Yaeyama society became more difficult, and its influence was truly difficult to measure. Even after 212 years, there is still an overwhelming sense of grief when this study think of the people who suffered this disaster. This monument was built in memory of the people who died on all of the islands. This study hope that the history of this unprecedented disaster is handed down for generations. This monument was built with the cooperation of Ishigaki City, Taketomi Cho, Yonakuni Cho, organizations, and organizations from inside as well as outside of the island. April 24, 1983 Meiwa - Ohtsunami Victim Memorial Monument Proposer

174

桃林寺及び所一三棟、畝一０歩、二歩、同寺の戸、会所四棟、住家の作物被害、田畑のその二０仁王像二、人を含む）全潰、五人（他の元の流失、御嶽一四棟、四・田畑総計一七九五町二反六流潰流失、公用で離島か総計二一七六戸、二・二・権現堂、総計一六四二町四反五畝一七％、遭難死亡者、二二％に平得、黒島二九三人（二％）潰した村、蔵元庁舎、在橋梁六座、貢納米等保らきて遭難死亡した三七六離島の当る）全潰した村浸水一００三書引流」黒島、石垣島の村番内、三・ートル五．石垣島八八一五人（一％）新城のと或弐拾丈（大川、ある災害の）七～七八・石垣島で「、、計七村津波の沖ノ石垣、新城島定され録（六０・石陸へ状況（七メ四・潮揚高弐拾八丈（新川、る九石垣島白保崎南南東四〇キロる））ー六メ寄揚、三度大波之時各村之形行地震の大波之時各村之形行書にトル登野城、明和大津波災害関ー）陸ノトル規模と或弐、石並大木根乍被）八四・位置（或弐拾五・参丈（係諸八メ東京天文台編理科年表に六～九メメ記よるー六丈（ートルトル））七と、測よ

仲与銘御手形寫御問合控等による）

M ( マグニチュード

北緯二四度

番、頭職等の公職者八八人及び蔵伊原間総計九三一三人（石垣島の群島人口の真栄里三

安良 )

七・四「八重山地震津波」と記

録抜粋屋良部の計八村、半

震源地東経一二大浜

宮良

白

Fig. 5-12 Excerpts from records on the Meiwa Great Tsunami disaster Information plate According to Tokyo astronomical observations, science chronology edited, “The Yaeyama Tsunami Jishinn” with a magnitude of 7.4, the epicenter was at the north latitude 124.3 and east longitude of 24. The epicenter was Ishigaki Island Shiraho Saki south southeast 40 km. According to "Ohnaminotoki - Kakumurano - Nariyukisyo", the tsunami situation occurring at Ishigaki Island tsunami run-up was 84.8 m, 60.6 m, 75.7 - 78.7 m, 6 - 7 m, offshore stones were launched, the stones on the land and the roots of big trees were blown away. Totally collapsed villages: Maesato on Ishigaki Island, Ohhama Miyara, Shiraho Nakayomi, Ibaruma, Yasura, and Tarumachi on Yarabu Partially collapsed villages: Seven villages of Kuroshima Island, Aragusuku Island of Ohkawa, Ishigaki, Arakawa, Tonoshiro, and Hirae, remote island of Ishigaki Island the condition of being overtaxed as “Jintouzei”. With a total of 9,313 (which is equivalent to 32.22% of the archipelago population), Ishigaki Island suffered 8,815 deaths (94.7%), 299 deaths (3.1%) for Kuroshima Island, 205 deaths (2.2%) for Aragusuku Island. Housing totally collapse, total 2,176 houses, inundation of 1,003 units, farmland swept away: 1,626 hectare, crop damage 1,777 hectarea, the public building of warehouse overseer "Kuramoto", 13 village guard stations at "Murabansyo", 4 meeting places at "Kaisyo", 14 shrines at "Utaki", bridge 6 shoe, the Tourinji temple and 2 statues of Nio Guardians of the same temple, Gongen, tribute rice and so on, were damaged.

175

There is a Tacolasser Stone on the north side of this area. "Takorasa stone" means a high dark

coast stone. "Takorasa Stone" has a legend that those surviving persons the 1771 Meiwa era

Yaeyama Tsunami gathered to rely on torchlight. This stone after the tsunami and had burned a torch

after the sunset, survivors gathered to rely on that lighting, as they rescued survivors and procured

foods.

Fig.5-13 Tacolasser Stones located on the north side of this area. This stones which was supposed to have gathered people who survived the Yaeyama/Meiwa Tsunami. A legend has it that "Survivors of the 1771 Meiwa era Yaeyama/Meiwa Tsunami gathered to rely on torchlight. (photo taken by T. Ohsumi on January 23, 2019)

According to Goto and Shimabukuro (2012) [22], the description of the 85 m tsunami run up

described in the Inscription says that the description of the survey result needs to be evaluated taking

the error into consideration. Actually, the height value is not so high at the point set at 85 m. As the

result of numerical calculation does not arrive at 85 m, it seems that the Tredecesor of the ancient

document contains something that is overestimated significantly.

The flow path of the tsunami that went from Miyara Bay to Nagura Bay listed in the ancient

document is indicated by the data source: DEM 5 B, the height is at most ~ 30 m elevation. Also,

from the description of survivor's rescue and procurement of food in front of the ~ 60 m high or

more Tacolasser the 85 m record of tsunami run up is of low reliability. For the tsunami run-up, 30

176

m of tsunami run-up height indicated by Hatori (1988) [21] is reliable.

According to Banzai (2015) [23], as to talk with the local people, the tsunami rushed to Panna

Dake, crossed the small mountain in the north side and came to the sea on the side of Nakura Bay,

when the stone crossed the mountain. Fig.5-15 shows a full view of Nagura Bay and the low land

from the north side of Mt. Panna.

Fig.5-14 The 1771 Yaeyama/Meiwa tsunami's historical tsunami trace and run-up route in the Ishigaki Island (after Goto and Shimabukuro (2012) [22]) （GSI Web Map used）

Nagura Bay Tsunami run-up Tsunami run-up

Mt. Panna

Fig.5-15 A full view of Nagura Bay and the north side of Mt. Panna. (photo taken by T. Ohsumi on January 25, 2019)

177

7) Amataria-Suari: Fig.5-16 to 5-17 “Amatariya-Suari” is inland of 160 m from the coastline, and the altitude is 6.2 m elevation (data source: DEM 5 B). It is described in historical documents that Lime stone which was on the beach

“Amateriya” was launched inland. Coral fossils are scattered around.

（GSI Web Map used）

Fig.5-16 View of “Amatariya-Suari ”.

Fig.5-17 Coral fossils are scattered around in this point. (photo taken by T. Ohsumi on January 23, 2019)

178

8) Other historical tsunami trace points: Fig.5-18

Other historical tsunami trace points and current picture shows Fig.5-18. Okinawa limestones are scattered, but it is unknown whether these are washed away by the tsunamis or the storm waves.

However, it seems that the events that were shown by the tsunami are repeated. Goto et al. (2010)

[24] classified the tsunamis or the storm waves. Storm wave boulders are distributed on the flat reef within 300 m landward of the reef edge in the Ryukyu Islands. However, tsunami boulders on the

Ryukyu Islands are deposited far landward of this limit.

Fig.5-18 Other historical tsunami trace points and current pictures. (photo taken by T. Ohsumi on January 22-24, 2019)

179

9) Villages moved to uphill area: Fig.5-19 Shiraho Village is located in the southeastern part of Ishigaki Island. The 1651 population totals

248 people Miyara and Shiraho. Shiraho Village became independent in 1713 by moving the over

300 farmers from Hateruma Island by a political orientation.

The population of Shiraho Village was 1,574, but in the 1771 Yaeyama/Meiwa tsunami,

1,546 were drowned, and the village was totally collapsed. 234 houses, 203 cows/horses were swept away, and upland field 373 hectare and rice field 1 hectare became uncultivable. Therefore,

418 farmers relocated from Hateruma Island again. The new village was rebuilt in the Uenoji of the northern plateau about 1 km northwest of the current village. However, this site was inconvenient and eventually returned to around 1793 where the previous village.

Miyara Village is located in Miyara Bay, southeast of Ishigaki Island. The village was under the cliff on the left bank of the Miyara River estuary. The population of Miyara Village was 1,221, but in the 1771 Yaeyama/Meiwa tsunami, 1,050 were drowned, and the village was totally collapsed. 149 housings and 107 cows/ horses were swept away. 214 hectare of upland field and

1 hectare of rice field became uncultivable. Number of 320 farmers were relocated from Kohana

Island. In addition, the village was built in Kanda, ~2 km northwest of the old village on the left bank of the Miyara River, and then relocated to the present location.

Ohama Village is a southern area of Ishigaki Island, and Ohhama village is the political and economic center of old Ohhama village along the present Route 390. The population of Ohama

Village was 1,402, but in the 1771 Yaeyama/Meiwa tsunami, 1,287 were drowned, and the village was totally collapsed. 210 houses swept away, and 100 cows/horses drowned. The 536 hectare of the fields washed away. 373 hectare of upland field and 9 hectare of rice field became landless and became uncultivable. 419 people moved from Hateruma Island by a political orientation.

180

However, reconstruction after the tsunami was extremely difficult. It was not clear before that, but the village has returned from uphill to previous location.

The population of Hirae Village was 1,178, but 560 people were drowned by the 1771

Yaeyama/Meiwa tsunami. The south side of the line from the east to the west of the village was collapsed, but 618 survived. The population of Maezato Village was 1,173, but in the 1771

Yaeyama/Meiwa tsunami, 908 were drowned. The village was totally collapsed and the surviving 265 people would not be able to rebuild the village. Thus, 313 people from Iriomote

Village was relocated to the new village, and the total number of people left was 558, the original village was built in the area called Kajyauchibaru, north-northeast ~ 1.5km from Maezato Village.

The population of Ohkawa Village was 1,304 in 1760, but in the 1771 Yaeyama/Meiwa tsunami,

412 were drowned. 174 housing was swept away but there were no damage to the fame fields. In order to avoid tsunami damage, the village started relocating to Bunni in uphill 3 km north, but returned to its previous location in 1775 due to inconvenience.

181

Fig.5-19 Villages moved to uphill area. [34, 35]

182

5.2 Miyako Island

According to the ancient document "Kyuyou" [27], the number of casualties and missing people

were 2,548 in Miyako Island. According to this document, the tsunami attacked about three times

after the earthquake in Miyako Island. At Yonaha-Maehama, there is a stone monument processed

in coral limestone (Fig.5-20), height 64 cm, width 29 cm, thickness 14 cm on the hill behind. In the

inscription, the year of Tsunami occurrence is indicated by the lunar calendar, the center can be read

as "Miyakuni · Shinzato · Uruka · Tomori". Many of the victims in this area, which had been

particularly heavily damaged, flowed to the front of Yonaha Beach, and legend history has that the

bodies were buried in Maeyama. This monument is what shows this history.

Fig. 5-20 Tsunami monument at Yonaha-Maehama. Fig.5-21 Otoiai-Gaki (Courtesy of Miyakojima-City General Museum) The damage caused by the tsunami in the Miyako Islands at that time was recorded in detail in "Otoiai-Gaki" (Fig.5-21). In this historical record, the most damaged in the Miyako

Islands were the villages of Miyakuni, Uruka, Tomori and Shinzato-Motojima located on the south coast of Miyako Island. Housing was collapsed 2,176 houses and ~2,000 farmers.

According to "Kyuyou" [27], 2,548 people were drowned by the 1771 Yaeyama/Meiwa tsunami.

In the vicinity of the airport in the Ohgami Island, Ryukyu limestone thought to be tsunami boulders is scattered in the Maja fisherman’s port.

183

(GSI Web Map used)

a: Ohgami Island

b: Maja fisherman’s port

Fig. 5-22 Ryukyu limestones in the northern eastern Miyako Island with the Ohgami Island (a) and the Maja fisherman’s port (b). (Aerial photo taken by T. Ohsumi on February 21, 2019)

2) Sawada-no-Hama Beach: Fig. 5-23

Sawada-no-Hama Coral Reef is located on the east side of the Shimoji Air Port, Ryukyu

limestones which are thought to be tsunami boulders, which are dotted on the coast. It seems that

many Ryukyu limestones were used at the time of airport construction, but the more than 300

scenic stones washed up by the 1771 Yaeyama/Meiwa Tsunami.

Watari and Omoto (2012) [28] assumed rocks of Fossil Porites sp. with diameters of ~10 cm to

3 m at Sawada-no-Hama Beach. And the rocks separated and moved from the foundation rocks.

The surrounding situation that the rocks were thought to be beached tsunami boulders carried from

the Reef edge. Watari and Omoto (2012) [28] picked out three samples of the Porites sp., which

scattered rocks suggest in the Sawada-no-Hama Beach and performed the sampling sites for 14C

ages and calibrated ages of three fossil Porites samples with their outermost shell as the dating 184

samples. These resulted got AD 1571 to 1726 datings. This age coincided with in the 1771

Yaeyama/Meiwa tsunami in the error range of 1 σ, revealed that there is a strong possibility that

in the 1771 Yaeyama/Meiwa tsunami reached northwest beach of Shimoji Island.

According to Kawana and Nakada (1994) [21], the possibility of being launched by the tsunami

that struck from Okinawa trough point is pointed out from the characteristics of the rock mass

distribution and tsunami boulders showing the age of ca. 550 to 650 years ago.

Watari and Omoto (2012) [28] objected no doubt be raised that, one of the samples Porites sp.

was inhabited at Reef edge, peeled from rock by in the 1771 Yaeyama/Meiwa tsunami. It was

transported to northwest beach of Shimoji Island and estimated that it was died as a result of being

launched.

(GSI Web Map used)

Fig. 5-23 Sawada-no-Hama Coral Reef.

185

Information plate Sawada-no-Hama coral reef is inside the arc-shaped shore reef ("Inaukatabaru" in Okinawan dialect) located northwest of lrabu Island and Shimoji Island. The Beautiful Sawada-no-Hama, selected as one of Japan's 100 best beaches, the more than 300 scenic rocks washed up by the 1771 Yaeyama/Meiwa Tsunami, and the gently sloping beach make a natural work of art created by subtropical nature, coral reefs, and the Kuroshio Current. The atoll is one of the best in the prefecture, and from olden times it has supported the lives of the islanders as an excellent fishing ground. The changing colors of the sea due to the tide and weather, and the tropical fish that inhabit the lagoon present a unique natural sight. There are fish traps in the area, and the sun setting on the horizon is spectacular. The scattered rocks suggest that in addition to the tsunami in 1771, the island has been struck by a major tsunami three times in the past 2,000 years. Sawada-no-Hama coral reef was designated a monument of scenic beauty of Miyakojima City on June 3, 1994

(photo taken by T. Ohsumi on February 22, 2019) 3) Shimoji Island Monolith (Obi-Iwa): Fig. 5-24

Shimoji Island Monolith is history tsunami trace points, tsunami run-up was 10 m. This boulder

is 12.5 m high. This ground level is 12 m (data source from DEM10B). There is a notch at the

height of ~10 m, this boulder was the position of the former seawater level.

186

(GSI Web Map used)

a b

d 下地島巨岩（史跡）

c この巨岩は、１７７１（明和８）年３月１０日大津波で打ち上げられたと伝えられている。岩の中央がやや引っ込んでおり、人間が帯を締めている姿に見える事から、帯岩（通称：オコスコビジー）とも呼ばれている。岩の高さ１２．５メートル、周り５９．９メートルもある巨岩で、重量は詳かでない。この木泊部落は津波襲来により全滅したといわれ、打ち上げられた大小様々な岩塊は、津波の置き土産としてその威力を推して知るべしだろう。その後、無数にあった岩塊は飛行場建設時に使用されたが、この巨石は当時の伊良部町の要請で残され下地島牧中に威容を留めている。そしていつ頃か民間信仰が広まり、大漁祈願祭や航海安全、家内安全の祈願が行われており、島建ての岩守護神として定着しつつある。宮古島市指定史跡（昭和 54 年 6 月 1 日）

Fig. 5-24 Shimojijima Monolith (Obi-Iwa), a: Front view of the Shimojijima Monolith, b: Behind view of the Shimojijima Monolith, c: Distant view of the Shimojijima Monolith from “Toori-Ike”,

187

Information plate This massive rock was reportedly washed up by the tsunami of March 10, 1771. The middle of the rock is slightly indented, and since it looks a bit like a person wearing a belt (obi), it's called the 'Obi-Iwa, meaning belt rock' ("Okosukobijee" in Okinawan dialect). This giant boulder is 12.5 m high and 59.9 m around. Its weight is not known. Kidomari village which was located near here at that time is said to have been wiped out by the onslaught of the tsunami, and the various sized rocks that it carried to the shore are testament to its power. Afterwards, countless rocks were used to construct the airfield but this massive rock was left at the behest of what was then lrabu, and it remains in its magnificence at Makinaka on ShiMoji Island. It ultimately became the focus of folk religion when people prayed here for bountiful catches of fish, or safety at sea and in the home, becoming a guardian deity of the island.

(from Information plate of Designated a historic site of Miyakojima City on June 1, 1979, photo taken by T. Ohsumi on February 23, 2019)

4) Eastern Henna Cliff: Fig. 5-25

The elevation of the cape showing the position of Eastern Henna Cliff and tsunami boulders is 10 m to 20 m. Tsunami boulders concentrate on the northeast side of the cape. According to Nakaza et al. (2015) [29], it is estimated that the cliffs collapsed and flowed out from the south side by the tsunami over the cape and Tsunami boulders on the cape was stripped off the southern side of the cape and was launched.

Accoding to Kawana and Nakada (1994) [21], dated samples were collected from the uppermost parts of these Holocene coralline boulders and fragments. Kawana and Nakada (1994) [21] restored a tsunami history in the area during the past several thousand years. Kawana and Nakada (1994) [21] dated the coralline boulders were much older than the age of the Yaeyama/Meiwa tsunami about 200 yr BP. However, Nakaza et al. (2015) [29] objected no doubt be raised that the existing explanations on pre-historical tsunami was in the Ryukyu Islands, that was research discloses that there were no profound or consistent traces of other tsunamis occurring in the region other than the Yaeyama/Meiwa

Tsunami.

188

(GSI Web Map used)

a b

Fig. 5-25 Eastern-Cliff, a: Southeast ward overview of the boulders deposited on the reef from foot of the Henna-cliff., b: Northwestward overview of the boulders deposited on the reef from tip of the Henna-cliff. (photo taken by T. Ohsumi on February 22, 2019)

189

5) Other historical tsunami trace points: Fig. 5-26 Other history tsunami trace points and current picture shows Fig. 5-26.

Fig. 5-26 Other history tsunami trace points and current pictures. (photo taken by T. Ohsumi on February 20-22, 2019) 6) Villages moved to uphill area

According to the ancient document "Kyuyou" [27], Miyakoku, Shinri, Sunagawa, Yoshimoto

Motohima in the Motoshima area was considered a village site destroyed by the Yaeyama/Meiwa

tsunami. Fig. 5-27 shows the bird view of the survey area of the Motojima Ruins. Amai Well

(24.73 N, 125.35 E; Elevation 16 m, Fig. 5-28) was for Tomori, Uruka and Shinzato residents,

before the water supply has spread in Gusukube in 1965, this well begins drinking water. This well

was used a precious water resource for a long time even after moving to the current colony, before

the 1771 Yaeyama/Meyama tsunami.

190

Uipya-Yama Ruins

A village trace of Uruka, Tomori and Shinzato-Motojima (Elevation ~25 m below) near the remains of Uipya-Yama of Miyakojima southeast shore hilly areas (Elevation ~50 m to 60 m: Fig.

5-29) is almost distributed over the East-West directions. According to "Kyouyou" of ancient documents, as a result of the Yaeyama/Meiwa tsunami, the lower‐lying districts “Motojima” of

Miyakojima Island were flooded. Moved to current Uruka, Tomori and Shinzato villages from the lower-lying districts.

However, this description differs from the scientific investigation (e.g. Kawana and Nakata

(1994) [21]). Kawana and Nakata (1994) [21] restored a tsunami history in the area during the past several thousand years. Accroding to Kawana and Nakata (1994) [21], most of the coralline boulders were dated are much older than the age of the Yaeyama/Meiwa tsunami about 200 yr. BP.

These area had been attacked by huge tsunamis around 600, 1,100, 2,000 and 2,400 yr. BP during the last 3,000 years.

Archaeological excavation

The scientific investigation is based on coralline boulders mass which is analogized as being launched by the tsunami and 14 C of coral fossils attached. Nakaza et al. (2015) [29] recommended archaeological excavation as a direct demonstration on the assumption that dating is not necessarily directly related to occurrence in the tsunami.

Motojima Ruins

In the southern part of Miyako Island, Motojima Ruins are being investigated. A survey report of the Shinzato-Motojima-Uechi plateaus Ruins and Shinzato-Higashi-Motojima Ruins [30] accompanying the road construction of Tomori-Uechi has been issued from the Okinawa

Prefectural Center for Archaeological Operations. This report confirms that the masonry found at the Shinzato-Higashi-Motojima remains continues to the masonry of the Uipya Yama Ruins (Fig. 191

5-29) located northwest.

Miyajkuni-Motoyama-Uechi Ruins

The survey report of the Miyajkuni-Motoyama-Uechi Ruins mounds accompanying the road

construction of Bora-Uechi [31] has been issued from the Okinawa Prefectural Center for

Archaeological Operations.

Shinzato-Higashi-Motojima Ruins

Shinzato-Higashi-Motojima Ruins has been confirmed that the masonry found at the Shinzato-

Higashi-Motojima remains continues to the masonry of the Uipya Yama Ruins located northwest.

In this report the dating of ruins was written as ca.14-15 century.

Shimoji (2007) [32] reported the Museum Bulletin of "Amare village and legendary tsunami"

and summarized the archaeological excavation (Table 5-1, Fig. 5-27). According to some reports,

dating were assumed to be the remains of ca .14 centurie, as for the tsunami, the tsunami before

1771 the Yaeyama/Meiwa tsunami was assumed at the about ca .15th century.

Table 5-1 Motojima villages moved to uphill area. Date Elevation Moved to Elevation Miyakuni-Motojima Ruins 14-18th century 10 m Miyakuni 40 m Shinzato-Motojima Ruins 14-15th century 15 m Shinzato 40 m Uruka-Motojima Ruins 14-18th century 10 m Sunagawa 40 m Tomori-Motojima Ruins 14-18th century 15 m Tomori 50 m

192

Fig. 5-27 Brid view of the survey area of the Motojima Ruins from Miyakuni. (Courtesy of Okinawa Prefectural Center for Archaeological Operations)

Fig. 5-28 Motojima villages moved to uphill area. (GSI Web Map used)

Elevation 51m（data source：DEM10B）

193

県指定史跡「上比屋山遺跡」指定年月曰昭和 31(1956) 年 2 月 22 日砂川集落南、高さ 40m の琉球石灰岩丘陵にある 14－15 世紀の遺構で、南側の砂川元島遺跨とあわせて広い集落跡を形成している。遺構内からは、宮古式や八重山式と呼ばれる土器や青磁、沖縄製陶器、南蛮陶器などか見つかっている。特に責磁が多く．もその解釈をめくって「倭寇の根拠地」説、「貿易で栄えた港町」説、「貿男の中継某地」説なとかある。また、遺跡内には、1 0 ヶ所余の御嶽があって、うち 3 ヵ所は昔のまま石垣の上に茅の屋根が置かれている。丘陵上にはトゥーンカイフツイス(遠見台）もある。

Fig. 5-29 Uipya-Yama Ruins with information plate. Information plate Designated a Prefectural Historical Site on February 22nd, 1956 These ruins, dating from the 14th to 15th centuries, are on a 40·m high Ryukyu limestone hill to the south of south of Sunagawa, kown as Uruka, village in the Gukukube area. Here lie the remains of a large village, called Motojima. Miyaiko and Yaeyama indigenous pottery, celadon porcelain from China, and ceramics from: Okinawa and Nanban can be found at this site. Since celadon porcelain has been found in significantly larger quantities, several theories have developed. This site may have been a base for Japanese pirates in the middle Ages, apart town that flourished through trade, or a trade transit point. There are more than 10 utaki, holy sites among the ruins. Three of them have been espinally well-preserved and still have thatched roofs with stone walls just as they had in ancient times. There is also a “Tuhokaiutsuisu”, observatory or look-out point, on the top of the hill.

(photo taken by T. Ohsumi on February 22, 2019)

194

6. Seabed topographic data

6.1 Island shelf points

As part of the “Project for the Comprehensive Analysis and Evaluation of Offshore Fault

Information” (The Project), initiated in 2013 by the Japanese Ministry of Education, Culture, Sports,

Science, and Technology, this study used the Red Relief Image Map (RRIM) method to visualize

the seabed topography of the area around the southern Nansei Islands [33]. The RRIM method is

based on multilayered topographic information computed from gridded three-dimensional digital

elevation model data. An RRIM can be used to visualize the topographic slope, concavities, and

convexities without requirement for any special geomatics or mapping information. From the

topographic data of the seabed in the generated RRIM (Fig. 6-1), it was possible to identify

submarine landslide points in the Yaeyama area, the insular slopes of the forearc basin, sliding earth

masses deposited by landslides, and points where seabed alluvial fans had formed. The lip–surface

conduits in the direction of the insular slopes near Yaeyama Islands are shown in Fig. 6-2.

The solid white line in the figure shows the submarine landslide occurrence in region A, where

the spread of the seafloor alluvial fan is observed, and the yellow broken line shows the submarine

landslide occurrence in region B, which also appears continuous.

195

Alluvial fan

Fig. 6-1 Seabed topography of the area around the Nansei Islands produced by the Red Relief Image

Map visualization method

Solid white line shown in the figure, the submarine landslide occurrence A region. Yellow broken line shows the submarine landslide occurrence B region. (Courtesy of JAMSTEC)

Fig. 6-2 Lip–surface conduits in the slope direction produced by the Red Relief Image Map visualization method. (Courtesy of JAMSTEC)

196

6.2 Analysis section

The seabed topographic data and slide points in the RRIM represent the current situation, i.e., after the occurrence of landslides. As stability analyses are performed on seabed topographic data before the occurrence of a landslide, we had to infer the seabed topographic data before the occurrence of a landslide in this study. A cross section of the current seabed topography is shown in

Fig. 6-3.

Ohyagi (2004) [16] considered the submarine landslide a turbidity current, but considering the simplification of the method, it is set with a circular slip shape, and the slope failure volume and sedimentation volume are set to be equivalent.

Before the occurrence of landslide After the occurrence of landslide

Fig. 6-3 Topographic cross section of the current seabed.

197

6.3 Submarine landslide points

Okamura et al. (2018) [2] represents the latest development in relevant knowledge but the detailed description has yet to be published.

A study by Matsumito and Kimura (1993) [3] considered a large area around Yaeyama Islands based on surveys conducted by R/V Kaiyo and R/V Yokosuka. Following a precise topographic investigation, the following factors were considered relevant to the occurrence of the

Yaeyama/Meiwa earthquake:

1) Submarine landslides have accumulated soil and sediment forming seabed alluvial fans.

2) Seabed mud sampling indicates the alluvial fans were formed by multiple submarine landslides.

3) With respect to the newest collapses and submarine landslides, the possibility of them being related to a tsunami is high because the points of seabed mud sampling were coincident with the epicenter of the Yaeyama/Meiwa earthquake.

4) Submarine landslides suggest that seabed topographic points were formed in association with an earthquake because there is evidence of a devastating historical earthquake that occurred off the coast of Yaeyama Islands.

Based on the above findings from the surveys, this study considered the region of landslide occurrence to be aligned in the east–west direction, given the spread of seabed alluvial fans and the cross-sectional position corresponding to the seabed topographic points.

6.4 External force from tsunami propagation analysis

To establish the external force, both earthquake motion and tsunami wave force can be considered.

However, according to Dainippon Earthquake History [34], the estimated seismic intensity of

Ishigaki Island is about 4, which is said to generate minimal damage through earthquake motion.

198

Even in The Project in 2015 [33], the estimated instrumental seismic intensity of Ishigaki Island was about IV. In addition, the duration of earthquake ground motion is limited in comparison with the duration of a tsunami; therefore, the external force was set as the wave force of the tsunami. The flow rate of the horizontal component was converted to the acceleration to make it an external force of the dynamic load．

This study set the tsunami source model from the following two cases and we calculated the waveform of the maximum tsunami height at the insular slope point. Based on this result, this was taken as the external force. A two-dimensional coordinate system and the Universal Transverse

Mercator UTM 53 projection were used. The data were interpolated using inverse distance weighting, and topographic data with mesh sizes of 1350 and 450 m (150 and 50 m) around Yaeyama

Islands (Hateruma and Iriomote islands) were prepared (Fig. 6-4).

Fig. 6-4 Model domain areas around Yaeyama Islands.

6.4.1 Tsunami source model

Based on comparison of traces of the 1771 Yaeyama/Meiwa earthquake in coastal regions of

Miyako and Yaeyama Islands among the source fault models used in the Tsunami Hazard

Assessment Consignment Report in Okinawa Prefecture of 2012 [35], this study set Case 1 with

the P1 fault as the tsunami source model (Mw 8.8) because it is easily reproducible (Fig. 6-5 a).

As for Case 2, the closest fault (01, NI - trench - S) shown in The Project in 2015 [33] was selected

199 as the tsunami source model. Also, with reference to the Tsunami Hazard Assessment

Consignment Report in Okinawa Prefecture of 2012 [35], the slip factor (D) was adjusted such that the Mw value was 8.1 (Fig. 6-5 b).

Fig. 6-5 Tsunami rupture models (load cases).

200

6.4.2 Tsunami propagation analysis

Tsunami propagation was simulated using equations based on nonlinear longwave theory (Table

6-1). The maximum tsunami height in the insular slope point was calculated from tsunami propagation analysis. Tsunami propagation was simulated using equations based on nonlinear longwave theory, taking into account friction and advection on the seabed. This simulation used a finite difference method with a leapfrog scheme on a staggered grid. The computational time step for each grid size in the finite difference method was set according to Courant–Friedrickson–Lewy conditions to ensure stability of the calculation.

Table 6-1 Details of tsunami propagation simulation.

201

6.4.3 Waveform of water level fluctuation amount

The load condition of Case 1 was the waveform from the flow rate. The upper, middle, and lower panels of Fig. 6-6 a show the water level fluctuation amount (NS), flow rate (NS), and

Fourier spectrum from the flow rate, respectively. As the data of water level fluctuation amount had a time step (Δt) of 5 s, the Nyquist frequency was 0.1 Hz, which shows the frequency range was 0.1 Hz or less. The dominant frequency was 0.00165 Hz (10.1 min). For Case 2, the upper, middle, and lower panels of Fig. 6-6 b show the waveform from the water level fluctuation amount

(NS), flow rate (NS), and Fourier spectrum from the water level fluctuation amount (NS), respectively. The dominant frequency of 0.000781 Hz (21 min) was longer than Case 1. In both cases, the time history of the north–south flow rate at the depth near the center of gravity of a sliding earth mass was used as a load point.

a b

Water level fluctuation amount (NS)

Flow (NS) Rate

Fourier spectrum from Water level fluctuation amount (NS)

Fig. 6-6 Waveforms and Fourier spectra from flow rate.

202

6.5 Seabed topographic data 6.5.1 Crustal structural data based

In The Project in 2015 [33], a velocity model was built for the area of the southern Nansei

Islands. In creating the velocity model, we used the polymerization velocity obtained in the process

of seismic profiling, together with velocity data obtained from analysis of Ocean Bottom

Seismograph (OBS) seismic crustal structure surveys conducted by the Japan Coast Guard.

In a horizon interpretation of crustal structural data, interpretation of both the acoustic basement

and the surface of the unconformity of the Tertiary period were applied to the area of the southern

Nansei Islands. Construction of a three-dimensional velocity structure was undertaken using

surface data of the seabed, A-horizon, and B-horizon in the area of the southern Nansei Islands to

construct a layered structure. We used OBS velocity information and well data to create a basic

speed cube. Analysis of the physical properties provided data and a geophysical database of

submarine earthquakes, from which a velocity structure model was constructed. The gridded

velocity structure model was created with 1 km horizontal resolution and 100 m vertical resolution.

Using this velocity structure model, depth conversion was performed based on analysis of the

physical properties and the fault plane. Work was conducted on a temporal cross section using

various geophysical databases with different survey execution dates and survey specifications,

together with depth conversion using a unified three-dimensional velocity structure model, to

confirm the shape of the fault. The tsunami source model was set and the waveform of the

maximum tsunami height in the insular slope point was calculated, which was taken as the external

force.

6.5.2 Soil modulus The physical properties of the seabed used in the analysis are summarized in Table 6-2, together

with the properties used for modeling seawater as a liquid element. The velocity structure model

was constructed for the area of the southern Nansei Islands areas as part of the work of The Project

in 2015 [33]. This study used the polymerization velocity obtained in the process of a multichannel 203

reflection seismic data, together with OBS velocity data derived from analysis conducted by the

Japan Coast Guard. Thus, the shear speed of the seabed was established as 550 m/s.

Table 6-2 Soil modulus. Seabeded Soil modulus value Source Shinajiri-mudstone Density: ρ(g/cm3) 2.5 (surface) Shear velocity: Vs (m/s) 550 Minimum data of OBS Poisson ratio: ν 0.45 Shinajiri-mudstone Damping rastio: h (%) 2.0 Internal frictional angle 30 Nakamura et al. (2011)， φ(deg) Chen, et al.（2006） Cohesion: c (kN/m2) 25.0 2 Tensile strengtσt (kN/m ) 0 Sea water Soil modulus value Source Density: ρ(g/cm3) 1.03 Sea water Shear velocity: Vs(m/s) ※1 Minimum data of Vs Poisson ratio: ν 0.49 Closed to 0.5 Same data with Shinajiri- Damping ratio: h(%) 2.0 mudstone 10 ※1: associated to 0.001kN/m2

6.5.3 Adhesive strength test

Analyses by both Nakamura et al. (2011) [36] and Chen et al. (2006) [37] were applied depending on the magnitude of displacement and slip-surface conditions of landslides, and the average shear strength parameters of landslides were calculated. Stable analysis was performed by appropriate application of the shear strength obtained following triaxial compression and ring shear tests, assuming appropriate perception of the actual slip surface. Consequently, the coefficient and internal frictional angle of the Shimajiri-sandy mudstone were determined as c = 23.2–26.7 kN/m 2 and φ =

28.6–36.6°, respectively. Based on the priority study from Nakamura et al. (2011) [37] and Chen et al. (2006) [38] of laboratory tests, the strength of a test sample for the submarine landslide point at

204 a depth of about 2,000 m was taken as an intermediate value of the strength of land-based mudstone

(Shimajiri-mudstone: c = 25 kN/m2, φ = 30°). However, the initial value of φ was reduced to 20° to ensure subsidence would occur up to a depth of 2400 m. The triaxial compression test determined the peak strength of an undisturbed normal consolidated sample collected using a triple-pipe core tube. In the ring shear for constant stress, the peak strength of the embankment soil was taken as the consolidated strength, from which the corresponding residual strength of the weak plane was obtained.

205

6.6 Application of submarine landslide duration using the Newmark method

Given that the duration of a submarine landslide is unknown, Hiraishi et al. (2001) [3] performed numerical calculations by changing the duration from 30–90 s. We applied the same process for submarine landslides in this study using the Newmark method [39] as dynamic load.

6.6.1 Newmark sliding block method

The modified Newmark sliding block method [39] calculates the response acceleration using

the dynamic finite element method. Horii et al. (1997) [40] developed the method to obtain the

rotational displacement from the equilibrium moments of the surface of rupture of circular slip.

The method has been used for seismic performance evaluation (deformation capacity evaluation)

[41] of railway embankment renovations. As the input parameters required for execution of the

stability analysis are the same as for circular slip analysis, it is considered a practical method. The

method assumes that the fracture form of an embankment is an arc, and it requires that the

earthquake waveform be entered for the surface of the rupture. Then, the equation of motion is

integrated to determine the amount of settlement of the embankment. At that time, the time history

for which the slip safety factor is <1.0 can be calculated. As the sliding force decreases and the

slip safety factor increases to become ≥1.0, the stable time can be taken as the duration of a

submarine landslide.

206

6.6.2 Two-dimensional dynamic interaction analysis code: Super-FLUSH/2D

The Super-FLUSH/2D Ver. 6.1 [42] dynamic interaction analysis code is a coupled soil– structure system capable of modeling liquid elements, e.g., seawater. The analysis code of Super-

FLUSH/2D used in this study comprised a frequency-domain analysis code for which the finite element mesh size was set based on the following equation such that the vibration component propagated appropriately:

H Vs : shear wave velocity (1) 𝑉𝑉𝑉𝑉 N : constant (5 or more) ≦ 𝑛𝑛∙𝐹𝐹𝐹𝐹𝐹𝐹𝐹𝐹 Fmax : upper limit of analysis frequency

As the external force, the flow rates of Case 1 (NS direction) and Case 2 (NS direction) shown in Figs. 6-7 and 6-8 were converted to acceleration to produce an external force of the dynamic load in Super-FLUSH/2D. In both cases, the load exceeds 1 G; therefore, the duration of the flow caused by the tsunami acting on the slope is much longer than the dura-tion of earthquake motion.

As the data of water level fluctuation amount had a time step (Δt) of 5 s, the Nyquist frequency was 0.1 Hz. However, the shear wave velocity (Vs) of the target seabed was 550 m/s; thus, element size H was <1100 m. As the element size was assumed as 100 m, it is considered that an element size of 100 m could correspond to the frequency component of the time history of sea level variation used in this study. The analytical mesh of seabed topography before the occurrence of landslide, analogized from the shape of the bottom of the sea, is shown in Fig. 6-9. The cross section is the submarine landslide occurrence region A, shown by the white solid line in Fig. 6-1.

207

Time [min.]

Fig. 6-7 Acceleration to make an external force of the dynamic load: (Case1)

Fig. 6-8 Acceleration to make an external force of the dynamic load: (Case2)

ca.15,000 m

ca.70,000 m

Close up

Close up area ※The red line shows a slip line Fig. 6-9 Analysis model. 208

6.7 Calculation of duration of submarine landslides using the Newmark method

1) Safety factor of landslides

The safety factor of submarine landslides of waveforms is shown in the upper panels of Fig. 6-

10. 2) Sliding residual displacement The sliding residual displacement of the waveforms is shown in the lower panels of Fig. 6-10.

The sliding residual displacement was eventually 2,400 m in both cases. The occurrence time for

Case 1 was 45 s (elapsed time: 660–705 s), while that for Case 2 was 276 s (elapsed time: 1,356–

1,632 s), which rounds with 280 s.

Case1 Case2 5.0 5.0

4.0 4.0

3.0 3.0

2.0 2.0

1.0 1.0 Safety Factor : Fs : Fs Factor Safety Safety Factor : Fs : Fs Factor Safety

0.0 0.0 0 10 20 30 40 50 60 0 10 20 30 40 50 60 Time [min.] Time [min.] Time history of FS : Case1 Time history of FS: Case2

-500

-1000 276 s（1,356-1,632 s） 45 s（660-705 s） 22.60-27.20 min -1500 11.00- 11.75 min m] [ -2000

-2500

-3000 0 10 20 30 40 50 60 Sliding displacement residual Time [min.] Time [min.] Time history of Sliding residual Time history of Sliding residual displacement [m] : Case1 displacement [m] : Case2

Fig. 6-10 Application of submarine landslide duration determined using the Newmark method.

209

6.8 Analysis of tsunami propagation by submarine landslides Using the calculated external forces and slope failure durations, tsunami propagation analyses were performed for the slope failures (two cases) using seabed topographic data before and after slope failure. In this study, tsunami heights were calculated for the Yaeyama/Meiwa tsunami in A region. However, the amount of seabed topographic variation did not follow the formula of Okada

(1992) [36]; instead, it was entered directly.

Representative points of historical tsunami traces, which were arranged by Hatori (1988) [19] based on Kato (1987) [20] and Kawana and Nakada (1987) [21] and so on shown, were compared with calculated values for Ishigaki Island (ten points) and Miyako Island (three points) (Fig. 6-11).

Quantitative comparison was performed using the geometric mean K and geometric standard deviation κ, in accordance with Ida (1977), as indices showing the spatial conformity of the trace and calculated values. Table 6-3 shows the historical tsunami traces and the maximum tsunami heights.

In Case 1, the calculated maximum tsunami height at Iwasaki, Miyarawan, and Shirahozaki was

20.86, 19.14, and 21.02 m, respectively, whereas the height of the tsunami trace at each of these locations was 30 m ( = 0.944, κ = 1.519) (Fig. 6-12). In Case 2, the calculated maximum tsunami height at Iwasaki, Miyarawan,𝐾𝐾 and Shirahozaki was 5.01, 6.65, and 8.69 m, respectively ( = 2.217,

κ = 1.677). Thus, reasonable agreement was obtained for Case 1 with short landslide duration𝐾𝐾 (Fig.

6-13). However, despite the conformity, Case 1 is considered to represent a magnitude less than the level of Mw 8.8 of the 1771 Yaeyama/Meiwa earthquake. Considering the submarine landslide also occurred at Mw 8.1, debris at the site of the landslide extended to the east in this seabed area. In

Case 2+, a doubly extended area, which involved simultaneous failure of region A and region B, was added in the eastward direction was as shown in Fig. 6-14. In this scenario, the calculated maximum tsunami height at Inoda was 15.57 m ( = 1.418, κ = 1.655) (Table 6-4, Fig. 6-15).

The occurrence time of sliding residual substantial𝐾𝐾 for Case 2 was 198 s (elapsed time: 1,380–

1,578 s) (Fig. 6-16). Here, the calculation of duration of submarine landslides was made Case 2 ++

210 using sliding residual substantial for Case 2 was 198 s (elapsed time: 1,380–1,578 s), which rounds with 200 s, using a doubly extended area of Case 2+. In this scenario, the calculated maximum tsunami height at Inoda was 20.13 m ( = 1.131, κ = 1.559) (Table 6-5).

𝐾𝐾

Fig. 6-11 Historical tsunami trace points.

211

Fig. 6-12 Numerical results for propagation of the tsunami at the Miyako–Yaeyama Islands at 60 s, 120 s, 180 s, 240s, 300 s, 600 s, 900 s, 1200 s, 1500 s, 1800 s, 2700 s. Positive (red) and negative (blue) values indicate the sea level above and below the still water level. (Case1: Slope failure duration: 45 s).

Fig. 6-13 Numerical results for propagation of the tsunami at the Miyako–Yaeyama Islands at 60 s, 120 s, 180 s, 240s, 300 s, 600 s, 900 s, 1200 s, 1500 s, 1800 s, 2700 s. Positive (red) and negative (blue) values indicate the sea level above and below the still water level. (Case2: Slope failure duration: 280 s).

212

Table 6-3 Historical tsunami traces and presentation results Case1: Slope failure duration : 45 s) 地名 Place name Historical tsunami Maximum tsunami height trace (m) (m) 平久保 Hirakubo 4.0 6.22 伊土名 Itona 4.0 7.88 吉原 Yoshihara 4.5 8.08 名蔵 Nagura 5.0 3.52 四箇 Shika 10.0 9.91 登野城 Tonoshiro 10.0 15.35 宮良湾 Miyarawan 30.0 19.14 白保崎 Shirahozaki 30.0 21.02 伊野田 Inoda 15.0 21.45 岩崎 Iwasaki 30.0 20.86 下地島 Shimojijima 10.0 15.70 平良湾 Hirarawan 4.5 3.21 宮国 Miyaguni 18.0 12.99

Case2: Slope failure duration : 280 s) 地名 Place name Historical tsunami Maximum tsunami height trace (m) (m) 平久保 Hirakubo 4.0 3.08 伊土名 Itona 4.0 3.59 吉原 Yoshihara 4.5 2.65 名蔵 Nagura 5.0 2.87 四箇 Shika 10.0 3.72 登野城 Tonoshiro 10.0 7.87 宮良湾 Miyarawan 30.0 6.65 白保崎 Shirahozaki 30.0 8.69 伊野田 Inoda 15.0 9.81 岩崎 Iwasaki 30.0 5.01 下地島 Shimojijima 10.0 3.99 平良湾 Hirarawan 4.5 2.84 宮国 Miyaguni 18.0 4.76

Fig. 6-14 Numerical results for propagation of the tsunami at the Miyako–Yaeyama Islands at 60 s, 120 s, 180 s, 240s, 300 s, 600 s, 900 s, 1200 s, 1500 s, 1800 s, 2700 s. Positive (red) and negative (blue) values indicate the sea level above and below the still water level. (Case2+: Doubly extended area in the eastward direction and slope failure duration: 280 s).

213

Table 6-4 Historical tsunami trace and presentation results - Doubly extended area in the eastward direction -

Case2+: Slope failure duration: 280 s

Table 6-5 Historical tsunami trace and presentation results – Slope failure duration: 200 s – Case2++: Slope failure duration: 200 s

214

[m] 23.00 smin 0 (1,380 s)

-500

-1000

-1500 26.30 min (1,578 s) -2000

-2500 Sliding residualdisplacement 0 10 20 30 40 50 60 Time [min.]

Fig. 6-15 Close up of time history of sliding residual displacement [m]: Case2.

● Historical Tsunami

🔶🔶 Case1 ■ Case2 ▲ Case2+ ▲ Case2++

Case1: Slope failure duration: 45 s, = 0.944, κ = 1.519 Case2: Slope failure duration: 280 s, = 2.217, κ = 1.677

Case2+: Slope failure duration: 280 s, 𝐾𝐾 = 1.418, κ = 1.655 Case2++: Slope failure duration: 200 s, 𝐾𝐾 = 1.131, = 1.559 𝐾𝐾 𝐾𝐾 𝜅𝜅

Fig.6-16 Comparison of historical tsunami traces and presentation results

215

6.9 Discussion

The NIED has been creating tsunami source models for all possible earthquakes based on comprehensive analysis and evaluation of offshore fault informatics within the JAMSTEC project.

In such work, there are primarily two types of parameter: those based on survey data and uncertain parameters. In addition, some parameters cannot be obtained, and they must be set based on previous research and the setting policies of existing fault models. There are two types of uncertainty associated with the parameters: epistemological uncertainty and random variation. In the former, the uncertainty is caused by differences such as in the interpretation of ideas and data among experts, although it can be reduced by accumulating data. For parameters with such uncertainty, multiple cases are assumed. In the latter case, variation is caused by randomness in phenomena and in the distribution of physical properties. Even if data are accumulated, the effects of random variation cannot be reduced. Regarding parameters with such uncertainty, this study set the range of variation and then this study evaluate the effect that it brings to the result.

In this study, tsunami source modeling was attempted to estimate the duration of the submarine landslides associated with the 1771 Yaeyama/Meiwa earthquake, based on the assumed activity of the target faults and by setting parameters with consideration of the epistemological uncertainty.

Based on the source fault of the 1771 Yaeyama/Meiwa earthquake, tsunami propagation analyses were conducted. Dynamic stability was evaluated using the Newmark method, based on the stress state by applying the equivalent wave force as the load, for a cross section of the Yaeyama/Meiwa

Islands shelf. This study involved many parameters with epistemological uncertainty and this study proceeded by adopting the following assumptions (Fig. 6-17):

1) The tsunami was generated by insular slope failure.

2) Slope failure was assumed caused not by acceleration of the earthquake but by wave force due to the tsunami

3) Seabed structure was set based on the geological structure of JAMSTEC.

4) Tsunami wave force was set using nonlinear longwave theory for two cases based on the 216 seawater level fluctuation amount and flow rate time history data at the insular slope point.

5) Island shelf failure was assumed a two-dimensional slip plane from the current cross section.

6) Island shelf failure was considered for two cases with a two-dimensional sliding surface and changed width based on alluvial fan points on the seabed.

7) Secondary dynamic finite element analysis (Super–FLUSH/2D) considering the liquid element was applied to slope failure.

8) Slope failure duration was calculated based on sliding residual displacement analysis using the

Newmark method.

9) Residual strength was set based on triaxle compression and ring shear tests of land-based samples, following Nakamura et al. (2011) [37], assuming that Shimajiri-mudstone is distributed on the slope.

10) The alluvial fan region of the submarine landslide occurrence is assumed for two cases (Case

1: region A, Case 2: region A and region B as simultaneous failure).

11) Duration of slope failure in Case 1 was 45 s; in Case 2, the duration of slope failure was 280 s.

12) Comparison of coastal maximum tsunami heights and historical tsunami traces (K, κ) was improved by carrying out tsunami propagation analyses on extended area, in which simultaneous failure of region A and region B was added in the eastward direction. Additionally, by reducing duration of slope failure, K and κ value was improved.

It was found that as the value of the set safety factor fell below 1.0, a residual deformation amount occurred. In Case 1, the sliding residual displacement finally occurred at 746 m and the duration of slope failure was 45 s. In Case 2, the duration of slope failure was 280 s. The insular slope can be regarded as a bank and the modified Tanimoto formula, discussed by the Ministry of

Land, Infrastructure, and Transport’s Port Authority (September 20, 2013) [43], was used to determine the wave force acting at the time of sliding. This study considered the load but this study judged it was beyond the applicable range. Next, this study considered application of the Goda 217 formula [44]; however, it transpired that the external force was about one-fifth the maximum value of the dynamical pressure.

The amount of seabed topographic variation did not follow the formula of Okada (1992) [37];

instead, it was entered directly and comparison was undertaken between the historical tsunami

traces and the maximum coastal tsunami heights of each case. Thus, the tendency of the 1771

Yaeyama/Meiwa earthquake tsunami was reproduced. In addition, using the difference of

landslide duration, this study performed analyses of tsunami propagation caused by landslides

and this study found that the influence of duration has considerable effect on maximum coastal

tsunami height.

A region and B region where the spread of where seabed alluvial fans had formed assumed

landslides occur at the same time. For the submarine landslides caused by the tsunami associated

with an event of Mw 8.1, based on seismic reflection data and the fault, the execution of tsunami

propagation analyses produced coastal tsunami heights of 〜20 m. At Inoda point on the eastern

coast of Ishigaki Island, result was obtained good reproducibility of 15 m of historical tsunami

trace and ~15 m of maximum tsunami height calculated in Case 2 +. Here, the calculation of

duration of submarine landslides was made Case 2 ++ using sliding residual substantial for Case

2 was 200 s, the calculated maximum tsunami height at Inoda was 20.13 m ( = 1.131, κ = 1.559)

𝐾𝐾

218

The historical tsunami traces of 30 m are run-up heights. The points at Iwasaki and Shirahozaki are over 1 km from the coast-line; therefore, it can be considered that the scale reproduction could also be evaluated in this analysis.

Fig. 6-17 Selection of parameters with epistemological uncertainty.

219

References [1] Imamura, F., Yoshida, K. and Andrew, M. (2001), The 1771 Meiwa earthquake in Ishigaki Island in Okinawa Prefecture Numerical Method of Tsunami Stone Movement, Proceeding of Coastal Engineering,, Coastal Engineering Committee, JSCE, Vol. 48, pp.346-350. [in Japanese] [2] Okamura, Y., Nishizawa, A. and Fujii, Y. (2018), Source of the 1771 Yaeyama tsunami, Japan Geoscience Union Meeting 2018，HDS10-03. [in Japanese] [3] Hiraishi, T, Shibaki, H., and Hara, N. (2001), Numerical Simulation of Meiwa-Yaeyama Earthquake Tsunami in Landslide Model with Circular Rupture, Proceeding of Coastal Engineering, Vol.48,pp.351-355. [in Japanese] [4] Matsumoto, T. and Kimura, M. (1993), Detailed Bathymetric Survey in the Sea Region of the Estimated Source Area of the 1771 Yaeyama Earthquake Tsunami and Consideration of the Mechanism of Its Occurrence, Vo. 45, Second volume of “JISHIN”，pp.417-426. [in Japanese with English abstract] [5] Aylsworth, J.M. (2012), New Canadian Teaching Resources from Natural Resources Canada for Earthquakes, Landslides and Tsunamis (lesson plans, maps and event timelines) (January 2012); Geological Survey of Canada, Open File 7073, 7 p. + 48 PDF files, doi:10.4095/289872 [6] Tsuji, K. and Ueda, K. (1996), Tsunami in Kagoshima Gulf with the Anei Sakurajima eruption， Proceeding of Volcanological Soc. of Japan, 186. [in Japanese] [7] Normanbhoy N. and K. Satake (1995), Generation mechanism of tsunamis from the 1883 Krakatau eruption. Geophys. Res. Lett., 22, 4, pp.509-512. [7] Lander, J. F. and P. A. Lockridege (1989), United States Tsunamis 1690-1988, Pub. 41-2, National Geophysical Data Center, 265 pp. [8] Lander, J. F. and Lockridege, P. A. (1989), Uses of a tsunami data base for research and operations (abstract), EOS Trans., Amer. Geophys. Union 67, 1003. [9] Yoshimoto, M., Takarada, S. and Takahashi, R. (2007), Eruptive history of Hokkaido-Komagatake volcano, northern , Japan, The Journal of the Geological Society of Japan, 113, new and revised edition, pp.81-92． [10] Satake (2007), Volcanic origin of the 1741 Oshima-Oshima tsunami in the Japan Sea. Earth Planets Space, 59, pp.381–390. [11] Tsuji, Y. and Hino, T. (1993), Damage and Inundation Height of the 1792 Shimabara Landslide Tsunami along the Coast of Kumamoto Prefecture, Bulletin of the Earthquake Research Institute, University of Tokyo,Vol. 68, No. 2, pp.91-176.Toshiji and Hino, 1993 [12] Baba,T., H. Matsumoto, K. Kashiwase, T. Hyakudome, Y. Kaneda, and M. Sano (2012), Micro- bathymetric evidence for the effect of submarine mass movement on tsunami generation during the 2009 Suruga Bay earthquake, Japan., in (ed) Yamada et al., Submarine Mass Movements and Their Consequences. Advances in Natural and Technological Hazards Research. 31, 485-495, Springer. [13] Friz H. M., F. Mohammed and J. Yoo (2009), Lituya Bay Landslide Impact Generated Mega- Tsunami 50th Anniversary. Pure appl. geophys. 166, 153–175. doi:10.1007/s00024-008-0435-4 [14] Weiss R., H. M. Fritz, and K. Wunnemann (2009). Hybrid modeling of the mega-tsunami in

220

Lituya Bay after half a century. Geophys. Res. Lett., 36, L09602; doi:10.1029/2009GL037814. [15] Chicxulub - Earth Impact Database (2011) http://www.passc.net/EarthImpactDatabase/chicxulub.html (2020.01.01, access) [16] Ohyagi, N. (2004), Landslides at the bottom of the sea and lake，Geographical geology recognition and terminology, Geological Geology Terminology Committee, Japan landslide society,Ⅱ.Chapter 9, pp.187-2014，ISBN4902628007. [in Japanese] [17] Heezen, B. C. and Ewing, M. (1952), Turbidity currents and submarine slump, and 1929 Grand Banks earthquake, Am. Jour, Science, Vol. 250, pp. 849 -873. [18] Gibo, S.,Nakamura, M., Higa, Y. and Yoshizawa, M. (2003), The Shear Strength of Strongly Weathered and Fractured Mustones for Slope Stabbility Analysis – Stability of slopes in the mudstone area of the Shinmajiiri-group, Okinawa - , Trans. of J S I D R E, No.227, pp.113-118. [in Japanese with English abstract] [19] Hatori, T. (1988), Tsunami Magnitudes and Source Areas along the Ryukyu Islands, ZISIN2, Vo l. 41, pp.541-547. [in Japanese with English abstract] [20] Kato, Y. (1987), Run-up Height of Yaeyama Seismic Tsunami (1771), ZISIN2, Vol. 40, pp.377- 381. [in Japanese with English abstract] [21] Kawana, T. and Nakada, T. (1994), Timing of Late Holocene Tsunamis Originated around the Southern Ryukyu Islands , Japan , Deduced from Coralline Tsunami Deposits, Journal of Geography, Vol.103, No.4, pp.352-376. [in Japanese with English abstract] [22] Goto, K. and Shimabukuro, A. (2012), Interdisciplinary study of the 1771 Meiwa Tsunami, KAGAKU, Vol.82 No.2, pp.208-214. [in Japanese] [23] Banzai, T. (2015), Research on the historical change of Ishigaki Island, and the activities of youth, which enable to inherit and reproduce the traditional Ryukyu culture, Educational Psychology Course, Faculty of Education, Journal of Saitama University (Faculty of Education)，64（2）, pp.85-119. [in Japanese with English abstract] [24] Goto, K., Kawana, K. and Imamura, F. (2010), Historical and geological evidence of boulders deposited by tsunamis, southern Ryukyu Islands, Japan, Earth-Science Reviews 102 (1-2), pp.77-99. [25] Kurayoshi, T., Jyunichi, Y., Kazuyuki, T., Fusaaki, M., Masanobu, A. and Shigehisa, K. (2008), Grants-in-Aid for scientific research expenses (Basic research program (B)) research report in 2005- 2007. [26] The Historical Atlas of Okinawa, KASHIWASHOBO Publishing Co., Ltd, 1983. [27] Kyuyou, Okinawa culture historical materials collection 5, Kadokawa Sophia Bunko, 793p., 2011. [in Japanese] [28] Watari, S. and Omoto, K. (2012), Calibrated Radiocarbon Ages of Porites Boulders Collected from Northwest Beach of Shimoji Island, West of Miyako Island, SW Japan, Annals of the geography, Vo l . 54, No.1, pp.1-6. [in Japanese] [29] Nakaza, E., Iribe, T., Tokuhisa, U., Miyazato, N.,Inagaki, K. and Savou, R. (2013), Prehistrical and Historical Tsunami of Ryukyu Islands Estimated Thgrough Tsunami Deposits, Journal of Japan

221

Society of Civil Engineers, Ser. B3 (Ocean Engineering), Vol.69, No.2, pp. I_515-I_520. [in Japanese with English abstract] [30] Shinzato-Motojima-Uechi plateaus ruins and Shinzato-Higashi-Motojima ruins: Survey report for road construction of Tomori-Uechi, Okinawa Prefectural Center for Archaeological Operations, 2002. [in Japanese] [31] Miyajkuni-Motoyama-Uechi-Kofun mounds: Survey report for road construction of Bora-Uechi, Okinawa Prefectural Center for Archaeological Operations, 2013. [in Japanese] [32] Shimoji, K. (2007), Amare village and legendary tsunami, Miyakojima-City General Museum Bulletin, No.11, pp.1-12. [33] Project for the Comprehensive Analysis and Evaluation of Offshore Fault Informatics (2015)， Headquarters of Earthquake Res. Promotion，Ministry of Education, Culture, Sports, Science and Technology. [in Japanese] [34] Dainippon Earthquake History (1943), new and revised edition. Vol. 2, pp.1621 - 1783， Association of earthquake disaster prevention (Shinsaiyoboukyoukai). [in Japanese] [35] Tsunami Hazard Assessment Consignment Report in Okinawa Prefecture of 2012, Okinawa Prefecture. [in Japanese] [36] Okada, Y. (1992), Internal deformation due to shear and tensile in a half-space, Bull. Seismol. Soc. Am., Vol. 85, pp. 1018-1040. [37] Nakamura, S., Gibo S., Kimura, S. and Shriwantha, B. (2011), Average shear strength parameters along the slip surface of various of landslides –Shimajiri mudstone landslides, Okinawa-, Landslides, Vol.48, No.5,1-262，pp. 251-262. [in Japanese with English abstract] [38] Chen, C., Gibo, S., Sasaki, K. and Nakamura, S. (2007), Classification of landslide types observed in the area of Shimajiri-madstone., Okinawa Island - For the risk evaluation of Landslide -, Landslides, Vol.43, No.6, pp.339-350. [in Japanese with English abstract] [39] Newmark, N.M. (1965), Effects of Earthquake on Dams and Embankments, Geotechnique, Vol.15, No.2, pp.137-160. [40] Horii, K., Tateyama, K, Uchida, Y., Koseki, J. and Tatsuoka , F. (1997), Seismic failure deforfmation prediction of railway embankment by Newmark method，32th Geotechnical research presentation，pp.1895-1896. [in Japanese] [41] Design standards for railway structures/ Manual for earthquake‐resistant design (1999), RTRI; Railway Technical Res. [in Japanese] [42] SuperFLUSH/2D Ver.6.1 (2014), Jishin Kogaku Kenkyusyo, Inc. [in Japanese] [43] Guidelines for Tsunami Seawall Design with the Breakwater-Resistant, Ministry of Land, Infrastructure and Transport Ports and Harbors Bureau, 2013. [in Japanese] [44] Moto, K., Minami, K. and Sato, H. (1977) A comparative examination of the wave pressure formulas applied to the design of composite breakwaters, Technical Note of The Port and Harbor research Institute Ministry of Transport, Japan, No.270, 60 p. [in Japanese with English abstract]

222

Chapter 6 Seismicity in Mediterranean Sea and Evaluation for the Tsunami for the AD365 Crete Earthquake

1. Motivation

The West Asian region is a tectonically active area due to crustal deformation; the associated earthquakes occur on a large scale and have been recorded from the historical period to the present.

Investigating the most suitable solution for this crustal movement will contribute to this region’s earthquake and tsunami disaster mitigation. The most reliable parameters were defined by researchers and applied with a non-uniform distribution in the fault plane based on Papadimitriou et al. (2008)

[1]. The calculated AD365 earthquake waveform provides an indication of maximum acceleration using the stochastic Green’s function method with the selected parameters. Using this estimation, damage to masonry structures can be calculated by Ohsumi et al. (2016) [2]. This earthquake was followed by a tsunami that devastated the southern and eastern coasts of the Mediterranean. Based on these results, risk mitigation from seismic and tsunami events should focus on high densely populated areas with thick sedimentary layers in the Mediterranean.

Keywords: AD365 Crete earthquake, upheaval, stochastic Green’s function method, Mediterranean

223

2. Back ground

West Asia is an area of active crustal deformation with a history large magnitude earthquakes.

Because crustal movement is ongoing, investigating the seismicity in this region may contribute to understanding and protecting against earthquake and tsunami disasters.

According to Hori and Kaneda (2013) [3], the relative plate motion of tectonics from the area around the Mediterranean Sea and west to India is 2-4 cm/ yr. This movement is smaller than other convergent plate boundaries that are similar to the collision zone in the Himalayas–Tibet mountain belt. However, this region is characterized by a subduction zone and a predominance of strike-slip faults that partly exist in the north Anatolia dislocation. These subduction zones cause earthquakes and tsunamis.

In West Asia, the African plate is sub-ducting beneath the Anatolian plate at a rate of 1 to 3.5 cm/yr, which results in frequent large magnitude earthquakes along this subduction zone. In AD 365 (Fig.2-

1), a large magnitude (M8.5) earthquake occurred near Crete (e.g., Fischer (2007) [4], Shaw et al.

(2008) [5], Stiros (2010) [6], Papadimitriou and Karakostas (2008) [1]). The AD 365 earthquake is one of the best known ancient earthquakes in the eastern Mediterranean. It caused a tsunami that resulted in great damage to Syria, northern Egypt, and the Greek coast.

Crete, located 160 km south of the Greek mainland, is the largest among approximately 3,000

Islands in the Aegean, with an area of 8,336 km2. Ancient earthquakes in Crete have been re-ported in various books by Ambrasseys (e.g., 1994) [7]. In the 4th century, Ammiaus, a historian and military service member, wrote a historical document consisting of 31 volumes. Due to the Christian propagation, which began in the age of the Roman Empire, historical documentation was common in the 4th–5th centuries.

Secondary tsunami damage caused by the AD 365 earthquake was greater than that from the primary earthquake in the Peloponnese peninsula. Because the magnitude was higher than M8, the tremor propagated across a large area surrounding the Mediterranean. More recent smaller magnitude earthquakes in the Greek Islands have also affected a wide area around the Mediterranean. John 224

Cassian, a theologian in the 4th–5th century, and Sozomenes, a Byzantine historian in the 5th century, described evidence for widespread flooding from the tsunami by tracing damage on the roofs of buildings and subsequent retreat of the coastline. A Byzantine historian, George the Monk, mentioned the tsunami in a 9th century chronicle. The tsunami caused by the AD 365 earthquake was also chronicled by Theopanos in the 8th–9th centuries, Cedrenus in the 11th century, and Glycas in the

12th century. According to the literature, the tsunami destroyed 50,000 houses and caused 5,000 casualties in Alexandria, Egypt.

Because large magnitude earthquakes and associated disasters will likely occur in the future, this study investigates and describes the characteristics of the AD 365 Crete earthquake.

Fig. 2-1 Epicenter of the AD 365 Crete Earthquake.

225

3. Crustal movement

Flemming (1978) [8] evaluated land subsidence and upheaval and estimated the relative rate of annual sea level rise (1.05 mm/yr.) based on research on the south-west coast of Turkey and at about

175 points in Cyprus.

Pitazzoli et al. revised his previous interpretation (1982, 1986) [9] (Fig. 3-1) after detailed survey and radio-carbon dating of the samples obtained from Antikyhira Island. Significant co-seismic uplift that took place during a short period was demonstrated by over 30 radiocarbon dates from 12 regions in Greece and precise sea-level indicators in the eastern Mediterranean. Therefore, it is assumed that the scale of uplift in Crete was 0.5 to 1.0 m in general but gradually increased toward the south-west and reached approximately 9 m. Radiocarbon dates show that the largest change occurred between

261 and 425.

・Chania ・Aptra ・ Phalasarna ・ Knossos

・ Sougia ・ Paleochora

Fig. 3-1 Contours of upheaval in Crete (up to 9 m on the southwest part of the island) modified from Pitazzoli et al. (1996) [9].

226

4. Trace of Upheaval

Fig.4-1 compares the crustal displacements and upheaval generating areas from the Fischer (2007)

[4], Shaw et al. (2008) [5] and Stiros (2010) [6] models. Pirazzoli (1996) [9] suggested that there is a trace of upheaval along the Crete coast from radiocarbon dating and provided a detailed survey with evidence for Holocene coseismicity (Fig.4-1). This data was used to compare differing uplift distributions; Table 3-1 shows the parameters from each research group used in the fault model.

Shaw et al. (2008) [5] provided the seismi Fischer (2007) [4] used a dip angle of 13°, low angle along the subduction plate, and set the average slip to 42 m. According to Murotani et al. (2013)

[11], the average slip was around 10 m during the Chile earthquake in 1960 and Tohoku earthquake in 2011. A value of 42 m is four times that of the Chili or Tohoku earthquakes. Moreover, sufficient values for upheaval distribution are not obtained using the equation from Okada (1992) [12] (Fig.4-

1a), city and topography in the area of Crete with regional seismicity corresponding to the AD 365 earthquake. According to the authors, the upheaval distribution suggested that the AD365 Crete earthquake occurred not on the subduction interface beneath Crete, but on a fault dipping at about

30° within the overriding plate. The shallow branch of the subduction zone dips at low angle to and couples with the Aegean lithosphere, while the deep branch dips freely (without coupling) at a high angle beneath the south Aegean trough. In this study, the crustal upheaval modeling from Shaw et al.

(2008) [5] was combined with the equations from Okada (1992) [12] (Fig.4-1b). Shaw et al. (2008)

[5] set the dip angle to 30° with a high angle within the overriding plate, the strike to 315°, average slip to 20 m, and fault depth to 45 km. The strike was the same angle, 315°, used by Papadimitriou and Karakostas (2008) [1] in their fault plane solution. Stiros (2010) [6] set the dip angle to 40° with a high angle within overriding plate, strike angle to 292.5°, and average slip to 16 m.

Stiros (2010) [6] analyzed elastic dislocation in the coastal upheaval data and found that this earthquake was associated with a reverse fault offshore of southwestern Crete, with a minimum magnitude of 8.5. This model is consistent with the approximate seabed trace of the fault; using observed and calculated displacements from the modelled fault, the fault depth was as deep as 70 km. 227

This study applies the crustal upheaval from the Stiros (2010) [6] modeling study and equation from

Okada (1992) [12] (Fig.4-1c). Papadimitriou and Karakostas (2008) [1] set the direction to 315° based on the seismology and topography from Papazachos (2000, 2001) [13, 14] (Fig.4-2). The earthquake moment assumed 5.7 × 1028 dyne cm from the area of dislocation and an elastic coefficient.

According to Papadimitriou and⋅ Karakostas (2008) [1], seismic coupling has been correlated with the maximum earthquake sizes that occur at a subduction zone. Subduction zones that are strongly seismically coupled periodically produce great earthquakes (Mw > 8.0) (Kanamori, 1977 [15]). In contrast, those that are seismically uncoupled produce only moderate to large earthquakes (Mw < 8.0)

(Ruff and Kanamori 1980) [16]. Reverse faulting is observed on planes with a NW or NE strike, and with approximately E-W P axis; the larger of them occurring in 1982 beneath the Mediterranean ridge.

The slip distribution by Papadimitriou and Karakostas (2008) [1] had non-uniform distribution in the fault plane (Fig.4-2). This study was in the upheaval by using each of the 128-mesh division to obtain the average slip. In addition, the geometry of the fault was almost same as in Shaw et al. (2008) [5]’s setting, except the fault length was set to 160 km. Using Shaw et al. (2008) [5]’s fault length, 100 km, the Mw scale becomes an 8.0 class using scaling low formulae for earthquake size and fault area (e.g.

R. Sato: log S = M-4.07 [17]). The M8 level is the difference between the 8.3 to 8.5 classes, which each prior study defined. Thus, ground motion can be estimated using the stochastic Green’s function method and parameters from Papadimitriou and Karakostas (2008) [1].

Table 4-1 Parameters used in the fault model.

Fischer Shaw Stiros Papadimitriou (2007) (2008) (2001) (2008)

Strike ° 297 315 292.5 315 Dip ° 13 30±5 40 35 Depth km 45 70 5 to 50 Length km 145 100 105 160 Width km 130 100 80 Slip m 42 20 16 8.9 28 28 M0 dyne cm 5.04×10 5.7×10

Mw 8.5 8.3-8.5 8.5 8.3

228

a b c

Antikythera . . Antikythera . Antikythera Island Island Island

. Chania

. Chania . Chania

Fig.4-1 Crustal displacements and upheaval generating areas from the Fischer (2007) [4], Shaw et al. (2008) [5] and Stiros (2001) [6] models. (Coulomb 3.3 was used, see Toda, et al., 2011 [10])

Fig.4-2 Slip distribution (left) and crustal displacements and upheaval generating areas (right) from Papadimitriou and Karakosta (2008) [1].

229

5. Estimating Earthquake Ground Motion 5.1 Setting parameters Tabe 5-1 shows the parameters used in the stochastic Green’s function method. As shown in table in Fig.11’s, the element fault area S (10 km × 10 km = 100 km 2) and elastic coefficient μ (from Stiros

(2001) [6]: 3.0 × 10 10, N/m 2) are used in Mo = μ DS to obtain the seismic moment for each element.

Based on the sum of the seismic moments of each element, the seismic moment of the entire fault is

3.44 × 1028 dyne·cm (Mw 8.3).

According to Somerville et al. (1999) [20] and Ishii et al. (2000) [21], slip distributions of more than 17.8 m act as the asperity area and the element less than 17.8 m acts as the background area. In the former, the stress drop is 22.1 MPa and rise time is 3 s, while in the latter, the stress drop is 2.6

MPa and the rise time is assumed to be 10 s. VR was set to ~80% of the S wave velocity from Kataoka et al. (2003) [22] and the cutoff high frequency was set to 13.5 Hz from Sato et al. (1994) [17].

Table 5-1 Parameters used in the stochastic Green’s function method. L (m) 100 Hypocenter N32.54 W (m) 86 E24.08 Strike (deg.) 315 Rapture-propagating Radial direction Dip (deg.) 35.0 Cut off frequency 13.5 (Hz) Depth (km) 5 ρ (g/cm3) 2.6 M0 (N/m) 3.44×1021 Vs (km/s) 3.5 Rise time (s) 3.0/10.0 Asperity/Background

230

5.2 Estimating the AD 365 earthquake ground motion waveforms Ohsumi et al. (2008) [2] provided the synthetic waveforms in the area of Crete with regional seismicity corresponding to the AD 365 earthquake. Fig.5-2 and Fig.5-3 show the estimated AD 365 earthquake velocity synthetic waveforms and response spectrum. At Phalasarna, 10 km from the epicenter, the estimated velocity is 102 cm/s. At Antikythera Island, 50 km from the epicenter, the estimated velocity is 57 cm/s. At Chania, 50 km from the epicenter, the estimated velocity is 108 cm/s. At Aptra, 75 km from the epicenter, the estimated velocity is 55 cm/s. At Iraklio, 150 km from the epicenter, the estimated velocity is 41 cm/s. At Athene, 270 km from the epicenter, the estimated velocity is 2 cm/s. These velocity values decline with distance from the epicenter, although the duration time expands to 200 s. The velocity response spectra for the synthetic waveforms at the

Chania and Irakio sites show a dominant natural period at 0.5 s (2 Hz), which causes non-linear behavior in the sedimentary layers.

Fig. 5-1 Estimated synthetic velocity waveforms Fig. 5-2 Velocity response spectra from the AD 365 earthquake. [2] comparison for the synthetic waveforms. [2]

231

6. Estimation of Tsunami Propagation

Tsunami propagation was simulated using equations based on non-linear long-wave theory, taking into account friction and advection on the seabed. This simulation used a finite-difference method

(FDM) with a leapfrog scheme on a staggered grid. The computational time step for each grid size in the FDM was set according to Courant- Friedrickson-Lewy (CFL) conditions to ensure stability of the calculation.

6.1 Setting parameters Topographic data were created from Shuttle Radar Topography Mission (SRTM) 30 Plus [23] data.

Grid resolution for land data is 30 arc seconds. Surface altitudes of the Earth were derived from data measured with a Synthetic Aperture Radar (SAR) onboard a space shuttle. Grid resolution for the ocean data is 1 arc minute. Offshore bathymetry data were derived from multiple sources, including the Coastal Relief Model of the National Geophysical Data Center (NGDC).

6.2 Methodology

In this study, the computational domain was bounded by latitudes 34° and 49° N, and longitudes

20° and 26° E. The domain covers the Peloponnesus Peninsula, which is located near Crete Island and the fault (Fig. 6-1). It also covers Alexandria, where historical observational of the tsunami have been documented [24], [25].

Maximum tsunami heights were calculated at several evaluation points. The distance between the fault and furthest evaluation point, located at Alexandria, is less than 1,000 km. Therefore, the two- dimensional coordinate system and Universal Transverse Mercator UTM 34 was used [23]. These data were interpolated using inverse distance weighting and topographic data with mesh sizes of 1,350 m and 450 m were prepared (Fig. 6-1).

In such analyses, horizontal displacement is normally neglected. However, Tanioka and Satake

(1996) [26] showed the effect of horizontal deformation. When the tsunami source is on a steep slope,

232 the horizontal displacement is large relative to the vertical displacement and the effect becomes significant. Thus, the initial water level of the tsunami is set as the vertical component obtained from the vertical direction and horizontal deformation effects are usually neglected. When the wave source is on a steep slope and the horizontal displacement is large relative to the vertical displacement, the effect becomes significant. Thus, this study calculated the seabed variation considering the horizontal level. The Tsunami propagation was simulated using fault parameters to estimate earthquake ground motion (Table 6-1 and Fig. 6-1). Table 6-1 summarizes the details of the simulation.

Table 6-1 Details of the tsunami propagation simulation. Governing equation Non-linear long wave theory Numerical solution Finite-difference method (FDM) with a leapfrog scheme on a staggered grid Calculation area Latitudes 34 - 49 degrees north, and longitudes 20 - 26 degrees east (This area covers the fault, the Peloponnesus Peninsula, the island of Crete, and Alexandria.) Mesh resolutions 1350 m, 450 m Boundary condition Considering tsunami run up in the land area. Transmission border nonreflective in the sea side Structures Not consider Calculation time 3 hours Initial water level Sea bed movement calculated by Okada (1992) [12]. Considering the effect of horizontal deformation by Tanioka and Satake (1996) [26]. Censored water depth 10-2 m Roughness coefficient 0.025

Fig. 6-1 Area of the tsunami propagation simulation.

233

6.3 Results Fig. 6-2 and Fig. 6-3 show snapshots of the simulation. Table 6-2 summarizes the maximum tsunami heights at representative evaluation points around Crete, Peloponnesus, and Alexandria. At

Crete, which is close to the fault, the maximum tsunami height is 14.5 m. Six minutes after the earthquake, the tsunami height increases by 50 cm. At Peloponnesus Peninsula, maximum simulated tsunami height is 8.8 m. Eleven minutes after the earthquake, tsunami height increases by 50 cm. t

Alexandria, maximum simulated tsunami height is 2.4 m. One hour and thirty-five minutes after the earthquake, the water level rises by 50 cm.

Table 6-2 Maximum tsunami heights at representative evaluation points and duration between the earthquake and arrival of different wave heights at each point. Time Maximum Maximum Tsunami Tsunami Tsunami Tsunami Evaluation points tsunami tsunami Heights Heights heights heights heights heights at 50 cm at 1 m at 3 m at 10 m Crete 14.5 m 38 min. 6 min. 8 min. 30 min. 37 min.

Peloponnesus 8.8 m 46 min. 11 min. 14 min. 31 min. -

Alexandria 2.4 m 104 min. 95 min. 102 min. - -

Fig. 6-2 Snapshots of tsunami propagation simulations for the whole study area, 1,350 m mesh.

234

(a) 0 min. (b) 30 min. (c) 60 min. (d) 90 min.

(e) 120 min. (f) 150 min. (g) 180 min.

Fig. 6-3 Snapshots of tsunami propagation simulations for Alexandria, 450 m mesh.

235

6.4 Comparison with results from previous studies

Maximum tsunami heights determined from the simulation were compared with results from

previous studies. Guidoboni and Ebel (2009) [24] and Yamazawa et al. (2014) [25] compared the

tsunami simulations from Lorito et al. (2008) [27] and Shaw et al. (2008) [5]. Lorito et al. (2008)

[27] did not report quantitative estimates but showed that large energy was probably captured and

carried along the Egyptian coast due to the edge waves. Shaw et al. (2008) [5] showed that waves

offshore of Alexandria reached approximately 0.6 m in height. However, they reported difficulty in

estimating the run-up in ancient Alexandria because of the many non-linear effects from near-shore

bathymetry and changes in bathymetry and land surface since AD 365. According to John Cassian

(4th–5th century), boats were washed out of the water, and onto rooftops in Alexandria. Additional

historical records (e.g. Guidoboni and Ebel, 2009 [24]) indicate that the AD 365 tsunami destroyed

cities and drowned thousands of people in coastal areas in Africa, Greece, Sicily, and along the

Adriatic. Flooding in Alexandria in AD 365 following the earthquake in Crete was recorded as a

major disaster because Alexandria was the most prosperous and densely populated city in the area.

Historical disaster records for this tsunami were also collected from other Ionian coast settlements.

7. Summary and Conclusions

Using the AD 365 earthquake parameters obtained from Shaw et al. (2008) [4] and Papadimitriou and Karakostas. (2008) [1], this study calculated motion from this large earthquake. Observed phase records at seismometer sites located on Crete Island were estimated using the Green’s function method for the AD 365 earthquake. Very good results were obtained by applying three Green’s functions to the fault geometry over a range of 160 km, where the shallow branch of the subduction zone dips at low angle to, and couples with, the Aegean lithosphere. Therefore, in this stochastic simulation, we applied Papadimitriou and Karakostas (2008) [1]’s parameters. Their study used each of the 128 mesh divisions to indicate the dislocation and upheaval. These slip distributions were then used in the Green’s function calculations in each of the 128 divisions.

236

References

[1] Papadimitriou, E. and Karakostas, V. (2008), Rupture model of the great AD 365 Crete earthquake in the south-western part of the Hellenic Arc, Acta Geophysica, 56, (2), 293–312. [2] Ohsumi, T. and Yagi, Y. (2016), Earthquake Activity in West Asia: Seismicity in the Mediterranean Sea and Evaluation of the Strong Motion for the AD 365 Crete Earthquake Using the Stochastic Green’s Function, Ancient West Asian Civilization, Springer, pp. 65- 83. [3] Hori, T. and Kaneda, Y. (2013), Giant earthquakes and tsunamis in the world: Mediterranean Sea, Report of CCEP 89. [4] Fischer, K. D. (2007), Modelling the 365 AD Crete Earthquake and its Tsunami, Geophysical Research Abstracts, 9, 09458. [5] Shaw, B., Ambraseys, N., England, P. C., Floyd, M.A., Gorman, G. J., Higham, T. F. G. Jackson, J. A., Nocquet, J.-M., Pain, C. C. and Piggott, M. D. (2008), Eastern Mediterranean tectonics and tsunami hazard inferred from the AD365 earthquake, Nature Geoscience, 1, 268-276. [6] Stiros, S. C. (2001), The AD 365 Crete earthquake and possible seismic clustering during the fourth to sixth centuries AD in the Eastern Mediterranean: a review of historical and archaeological data, Journal of Structural Geology, 23, 545-562. [7] Ambraseys, N., C., Melville and R. Adams (1994), The Seismicity of Egypt, Arabia and the Red Sea, Cambridge University Press, Cambridge. [8] Flemming, N. C. (1978) Holocene eustatic changes and coastal tectonics in the northeast Mediterranean: implications for models of crustal consumption. Philos. Trans. R. Soc. London, Ser. A, 289 (1362), 405-458 + Appendix I. [9] Pitazzoli, P. A., Laborel, J. and Stiros, S. C. (1996), Earthquake clustering in the Eastern Mediterranean during historical times, Journal of Geophysical Research, 101, B3, 6083-6097, Soli [10] Toda,S., Stein, R. S., Sevilgen,V. and Lin, J. (2011), Coulomb 3.3 Graphic-Rich Deformation and Stress-Change Software for Earthquake, Tectonic, and Volcano Research and Teaching--User Guide, Revision History for USGS Open-File Report 2011-1060. [11] Murotani, S., Satake,K. and Fujii, Y. (2013), Scaling relations of seismic moment, rupture area, average slip, and as-perity size for M~9 subduction-zone earthquakes, Geo-physical Reserarch Letters, 40, Issue 19, 5070–5074. [12] Okada, Y. (1992), Internal deformation due to shear and tensile faults in a half-space, Bull.Seismol. Soc. Am. 82, 1018-1040. [13] Papazachos, B.C., B.G. Karakostas, C.B Papazachos, and E.M. Scordilis (2000), The geometry of the Benioff zone and lithospheric kinematics in the Hellenic Arc, Tectonophysics, 319, 275- 300. [14] Papazachos, B.C., D.M. Mountrakis, C.B. Papazachos, M.D. Tranos, G.F. Karakaisis, and A.S. Savvaidis, (2001), The faults that caused the known strong earthquakes in Greece and surrounding areas during 5th century B.C. up to pre-sent, 2nd Conf. Earthq. Enging. and Engin. Seism., 237

Thessaloniki 1, 17-26. [15] Kanamori, H. (1977), The energy release in great earthquakes, J. Geophys. Res., 82, 2981-2987. [16] Ruff, L. and Kanamori, H. (1980), Seismicity and the subduction process, Phys. Earth Planet.Inter. 23, 240-252. [17] Sato, R. (1989), Japanese seismic dislocation parameter handbook，390 p，Kajima Institute Publishing Co., Ltd. [18] Irikura, K. (1986), Prediction of Strong Acceleration Motion using Empirical Green’s Function, Proc. 7th Japan Earthquake Engineering Symposium, 151-156. [19] Kamae, K.and Irikura, K. (1994) Simulation of Seismic Intensity Distribution During the 1946 Nankai Earthquake Using a Stochastically Simulated Green’s Function, Proc. of 9th Japan Earthquake Engineering Symposium, 1, 559-564. [20] Somerville, P.G., K. Irikura, R. Graves, S. Sawada, D. Wald, N. Abrahamson, Y. Iwasaki, T. Kagawa, N.Smith, and A. Kowada (1999), Characterizing crustal earthquake slip models for the prediction of strongground motion, Seismological Research Letters, 70, 59-80. [21] Ishii,T., Sato, T.and Somerville, P., G. (2000), Identification of Main Rupture Areas of Heterogeneous Fault Models for Strong-Motion Estimation, J.Struct.Constr.Eng., AIJ, No.527, pp.61-70. [22] Kataoka,S., Kusakabe,T., Murakoshi, J. and Tamura,K. (2003), Study on a Procedure for Formulating Level 2 Earthquake Motion Based on Scenario Earthquakes, RESEARCH REPORT of National Institute for Land and Infrastructure Management, No.15. [23] Scripps Institution of Oceanography ： Global Topography SRTM30_PLUS ， http://topex.ucsd.edu/www_html/srtm30_plus.html， [Accessed March. 14, 2018] [24] Guidoboni, E. and Ebel, J. E. (2008) Earthquakes and Tsunamis in the Past: A Guide to Techniques in Historical Seismology, Cambridge University Press, 604 pp. [25] Yamasawa, T. (2014) The Eastern Mediterra-nean Tsunami of 21 July A.D.365: A Review of the Literary, Departmental Bulletin Paper, Nara Prefec-tural University Research Report, Vol.24， No.4, pp.27-52. • [26] Tanioka, Y. and Satake, K. (1996), Tsunami generation by horizontal displacement of ocean bottom, Geophysical Research Letters, Vol.23，No.8，pp.861-864. [27] Lorio, S., Tiberti, M. M., Basili, R.,Piatanesi, A. and Valiensise, G. (2008), Earthquake-generated tsunamisin the Mediterranean Sea: Scenarios of potential threats to Southern Italy, Journal of Geophysical Research 113, B01310.

238

Chapter 7

Final Remarks

The survey results and findings can be summarized as given below:

In Chapter 1: Damage related to the 2011 Tohoku Earthquake in the South-central Coastal Area of

Iwate Prefecture

1) In the Touni area, housing was moved to higher elevations after the Showa Sanriku tsunami

in 1933. However, businesses began to open in the low-lying areas two years later.

2) In the Yoshihama area, housing was moved to higher elevations after the Meiji Sanriku

tsunami in 1896, and after the Showa Sanriku tsunami in 1933. Yoshihama was praised as a

“Miracle Village” and a monument with an inscription to this effect was erected.

In Kamaishi City, students at junior high and elementary schools evacuated promptly and

voluntarily, in accordance with what they had learned in evacuation drills. This became known

as the “Kamaishi Miracle”. The present study attempted to verify the effect of this prompt and

voluntary evacuation.

In Chapter 2: Beyond the Tohoku Earthquake

1) The possibility of using the municipal tsunami hazard assessments, their current status, and

issues regarding tsunami measures were assessed and discussed.

2) This study determined the current implementation status and issues related to tsunami response

measures, and compiled the views and experiences (including negative ones) of municipalities

related to evacuation.

In Chapter 3: The Comprehensive Analysis and Evaluation of Offshore Fault Informatics

1) Subduction-zone earthquakes, like the 2011 Tohoku earthquake, and active fault earthquakes

both occur in the sea surrounding Japan. To address this lack of progress, a new fault evaluation 239

model was developed in the present study.

2) Simulated tsunami heights determined using the new model agreed with heights observed

historically, such as the 1940 earthquake off the coast of Kamui (Shakotan-Oki earthquake)

and the 1983 Nihonkai–Chubu earthquake. This results indicated that the model is valid and

accurate for the faults.

3) This study will be progressing to construct source models along faults to simulate seismic

motion and tsunamis, through considering the shape and mechanism of fault planes and the

possibility of combination occurrences of earthquake, and utilizing the simulating models for

disaster prevention and mitigation.

In Chapter 4: Development of a Real-Time Damage Estimation System for Embankment Using Earthquake Early Warning Information 1) A system was proposed that could be applied to potential disaster areas for selecting priority

areas for surveying, and for identifying evacuation routes immediately after an earthquake.

2) The system provides damage estimates for embankments using EEW (Earthquake Early

Warning) seismic parameter Tables, which are calculated with a high accuracy. The system

also provides information about other structures that need priority attention.

3) In future, it is intended that this system be used to contribute to, and integrate, the “Real-time

Embankment Damage Estimation” system into the Japan Real-time Information System for

earthQuake (J-RISQ).

In Chapter 5: Seabed Landslides and Their Consequences related to the 1771 Yaeyama/Meiwa

Earthquake

1) Tsunami events were calculated using geographical features indicated by sea floor data that had

been identified before and after landslides.

240

2) This calculation process verified the Yaeyama/Meiwa earthquake that occurred in 1771 by using

the calculated duration time of the landslides. Consequently, good results were obtained that

reproduced the historical tsunami records for this earthquake. 3) An analysis of this tsunami propagation method showed that it could be applied to the analysis of the tsunami caused by the landslide generated by the 2018 Sulawesi earthquake, Indonesia.

In Chapter 6: Seismicity in Mediterranean Sea and Evaluation of the Tsunami for the AD365 Crete Earthquake 1) This study used parameters defined in various papers.

2) According to historical records, the AD 365 tsunami destroyed cities and caused the drowning

of thousands of people in coastal areas in Africa, Greece, Sicily, and along the Adriatic Sea.

3) The historical records are in agreement with our estimated maximum tsunami heights. By

reproducing historical earthquakes and tsunamis, lessons have been learned that can be used

generate measures and preparedness for major earthquakes that are likely to occur in the near

future.

241

Acknowledgments

I express my indebtedness and deep sense of gratitude to my supervisor Prof. Hemanta

Hazarika, Department of Civil Engineering, Kyusyu University, for his continuous support, advices and guidance throughout the course of this study, preparing this thesis and related manuscripts. I am deeply indebted to Prof. Noriaki Hashimoto and Prof. Taiji Matsuda for helpful advice and for reading the manuscript. I also would like to express my gratitude to Dr.

Siavash Manafi Khajeh Pasha, Mr. Divyesh Rohit and Mr. Tsubasa Maeda for their kind help.

I would like to express my gratitude to Prof. Yoshiyuki Kaneda with Kagawa

University, for his kind help.

To the NIED majestic members, I will not forget their kind and sincere help. I thank

Dr. Hiroyuki Fujiwara, Dr. Shin Aoi and Dr. Narumi Takahashi for their kind and sincere help, and constructive comments for the tsunami propagation analyses by Dr. Yuji Dohi, NIED.

I pay my sincere gratitude to my wife Yuri who gives me strong encouragement. My special thanks to Emeritus Prof. Tsuneo Katayama, the former president of NIED, for inviting me to study earthquake engineering. Also, my special thanks to Prof. Seishi Okuzono with for inviting me to study for the Geo Technology.

To the spirits of my beloved father, mother and elder sister, I will never forget the dearest persons forever.

242