In particular the nature of loading processes during evacuations is shown to result in hysteretic phenomena, which considerably reduce the efficiency of the network

Ⅰ.Keynote Speech: Prof. Hani S. Mahmassani, Northwestern University On the Traffic Dynamics of Urban Network Evacuations

Prof. Hani S. Mahmassani

Civil and Environmental Engineering, Northwestern University, USA

Network evacuations are complex events influenced by a myriad factors associated with the behavior of individuals, households and official entities. In this presentation, we take a macroscopic perspective aimed at estimating overall evacuation times and determining effective staging strategies of large-scale evacuations. The approach is rooted in a fundamental understanding and representation of network traffic dynamics, reflected in the properties of networkwide fundamental diagrams (NFD). In particular the nature of loading processes during evacuations is shown to result in hysteretic phenomena, which considerably reduce the efficiency of the network. Evacuation strategies that recognize these phenomena can reduce overall evacuation times and help move more people out of harm’s way.

Ⅱ.Session: Network and Resilience Modeling for Evacuation

Highway Network Protection Strategy Considering Connectivity Reliability

James C. Chu^a*, Shih-Chi Chen^b

aNational Taiwan University, Taiwan; ^bNational Central University, Taiwan

In this study, we consider a situation where some links of the network can be protected against the disaster to maintain the connectivity of the network. Because the damage to a highway network caused by a disaster is probabilistic, we develop a stochastic programming model to find the optimal protection strategy for a highway network system that maximizes connectivity reliability. Specifically, the objective function we adopted is the expected number of origin-destination (OD) pairs that can be served over all scenarios of link failure due to a disaster.

The proposed model is demonstrated with the highway bridge systems of Miaoli City, Taiwan. The natural disasters we consider are earthquakes that were simulation by the Taiwan Earthquake Loss Estimation System (TELES). The results show that the proposed model has potential for improving highway network connectivity reliability.

Keywords: highway protection, network reliability, connectivity reliability, stochastic programming

Nuclear Bunker Deployment Problem for Mass Evacuation in Response to Nuclear Power Plant Accidents

Yu-Ting Hsu^a*, Srinivas Peeta^b

aNational Taiwan Unversity, Taiwan; ^bPurdue University, USA

An evacuation problem has been generally modelled to remove the population from the areas affected or potentially threatened by a disaster to safe locations, thereby meeting some operation performance objectives to avoid or mitigate possible life loss. Pre-disaster planning and on-line operation in terms of evacuation traffic management have been the focus in the literature to facilitate mass evacuation. However, for some disasters which occur with no notice and allow comparatively limited evacuation time, existing transportation systems may hardly accommodate the evacuation demand surge generating in a short time period, especially if such disasters

- 1 -

the described situation; for example, the Fukushima Nuclear Power Plant accident, occurring after 3/11 Earthquake in Japan, which called for the clearance of the areas within a 50-mile radius from the nuclear power plant (recommended by the U.S. Nuclear Regulatory Commission) and involved the relocation of around 2 million people. Although controversial, nuclear power still accounts for over 12% of the world's electricity provision but also potentially threatens a great population, either literally or mentally. For instance, if the same recommended evacuation scope is applied to the existing and prospective nuclear power plants in the northern Taiwan, it will entail the evacuation of nearly 10 million people. The mass evacuation of such a scale requires sophisticated coordination across multiple administrative agencies and represents extensive operational challenges, peculiarly when the atmosphere of fear and panic toward nuclear disasters is present.

The deployment of nuclear bunkers may be another thread of research to address the evacuation problem of nuclear disasters, which was implemented during the Cold War as a civil defense measure in response to nuclear attacks but rarely discussed in the context of nuclear disasters. This study seeks to develop the location problem of nuclear bunkers from the perspective of relieving traffic loading over the existing transportation system, thereby avoiding severe traffic congestion and the associated danger to evacuees. The provision of nuclear bunkers within the potentially radiation-affected region can significantly shorten travel distances for some evacuees to arrive at safe locations and thereby reduce overall traffic flow volume. Hence, the problem is constructed in terms of spatially re-shaping demand pattern over origin-destination pairs to improve the performance of the evacuation network. The effectiveness of deploying nuclear bunkers on evacuation traffic management is illustrated using experimental scenarios of nuclear power plant accidents. The experimental results are analyzed in terms of evacuation network performance and risk-related measures under different disaster impact patterns, deployment scenarios, and the associated resource allocation or limitation.

Keywords: Nuclear Bunker Deployment; Nuclear Power Plant Accident; Location Problem

Sampling-based Method for Road Capacity Uncertainty Analysis in Evacuation Planning

Jian Li^a*, Kaan Ozbay^b

aTongji University, China; ^bNew York University, USA

Evaluation efficiency of evacuation planning is important for public officials to avoid unnecessary economic cost and casualties. However, planning for a catastrophic event presents numerous additional challenges for capturing the impacts of uncertain events, such as potential loss or partial capacity reduction of vital transportation links. This study proposed an analytical methodology and efficient solution procedure to evaluate evacuation planning considering uncertain events impacts (e.g. incident or extreme weather) in network-scale. The road capacity was considered as variable with certain distribution in the form of day-to-day roadway traffic conditions. For solution procedure, Sample-Average Approximation (SAA) via Latin Hypercube Sampling (LHS) was employed to generate plausible realizations of link capacity values from a multi-dimensional distribution and solve the stochastic programming. The results were analyzed from expectation or probability distribution of evacuation time under empirical capacity distribution. At last a case study of no-notice evacuation planning considering accident and extreme weather was presented. The results may benefit to evaluate the feasibility of a given evacuation plan for different potential degradable roadway conditions and provide alternative more reliable evacuation choices to decision makers.

Keywords: Evacuation; Capacity Uncertainty; Sampling-based Method

- 2 -

A Resilience Optimization Model for Transportation Networks under Disasters

Ta-Yin Hu^*, Yi-No Ko, You-Shiuan Wu National Cheng Kung University, Taiwan

Transportation infrastructures support a wide range of human activities, but they are vulnerable when facing extreme conditions. Since extreme weather events are becoming more intense and frequent, the issue of enhancing transportation resilience under disasters is bringing more attention to the society in recent years. Resilience refers to the ability of bouncing back to the original functionality level after external shocks. In order to ensure the stability of travel and prevent serious delays due to the unexpected and extreme events, resilience assessment for transportation systems are required to be evaluated as well as the development of resilience improvement strategies.

Since the discussions on transportation resilience are still limited, this paper aims to measure network resilience under disasters. A mathematical model for resilience assessment with the constraints of budget and traversal time, and the determination of the resource allocation of pre- and post-disaster actions is proposed. The concept of improving resilience through preparedness and recovery activities is considered in the model. Two measure methods are developed for resilience assessment. Resilience under disaster scenarios are evaluated by a simulation-assignment model, DynaTAIWAN. Numerical experiments are conducted based on a realistic network of Sanmin district in Kaohsiung City to illustrate the application. Results of the application offer insights into the importance of resilience assessment and resource allocation of system performance improvement strategies under disasters.

Keywords: Resilience; Climate change; Transportation network

Ⅲ. Session: Measuring and Modeling Evacuation Behavior I

Modeling Route Choice Behavior of Evacuees in Highly Urbanized Area:

a Case Study of Bagong Silangan, Quezon City, Philippines

Hector Lim, Jr.^*, Ma. Bernadeth Lim, Mongkut Piantanakulchai Thammasat University, Thailand

One of the essential aspects necessary for evacuation modeling and management is taking into account the travel behavior of evacuees. This study was conducted to understand the evacuation route choice behavior of households affected by the flood in Mid-August of 2013 at Bagong Silangan, Quezon City, Philippines. Binary logit route choice models with hazard-related and socio-demographic information, and evacuation household decisions as exogenous factors, were estimated and validated. Results show that mode, evacuation destination, vehicle ownership and flood level are strong predictors. Expanding this study to other flood affected areas can be explored in the future.

Keywords: Flood; Evacuation; Route Choice; Travel Behavior; Evacuation Modeling

- 3 -

Effects of Information Provision on Going Home Behavior and Traffic Congestion at Large-Scale Disaster: Case Study of Nagoya Metropolitan Area

Toshiyuki Yamamoto^a*, Yuichi Sugiyama^b, Ryo Kanamori^a, Yu Hiroi^a

aNagoya University, Japan; ^bCentral Nippon Expressway Company Limited, Japan

The Great East Japan Earthquake in 2011 generated a lot of stranded commuters in the Tokyo metropolitan area, Japan. In Nagoya metropolitan area, Japan, large earthquakes occurred periodically, and tsunami caused by the quakes also hit the region, resulting the disasters. It is anticipated that a large-scale earthquake will also generate the same problem as occurred in the Tokyo metropolitan area at the Great East Japan Earthquake, which is the heavy traffic congestion of going home trips, and that those who go home will be complicated with tsunami refugee.

However, it is suggested that such travel demand can be reduced by providing appropriate information.

In this study, a discrete choice model of going home behavior is developed based on the questionnaire survey at the Great East Japan Earthquake in order to quantitatively analyze the effect of information provision on the going home behavior. The alternatives include car as a driver, car as a passenger, bicycle, walk, and not going home. Train and bus are excluded from the alternatives because they were out of service in reality. It is confirmed that the information about the family safety and that on the current situation affect the decision of not going home significantly from the estimation results. By using the estimation results, the number of stranded commuters in Nagoya metropolitan area at the large-scale disaster, and the emergent travel demand by those who go home by car and tsunami refugee are forecasted. Conventional mesoscopic dynamic traffic simulator is also developed to simulate the traffic congestion resulting from the emergent travel demand considering both returning home trips and evacuation from tsunami hitting area. In the traffic simulation, four scenarios are compared: base case where stochastic route choice under current situation is used, normalcy bias case where stochastic route choice under ordinary traffic situation (false recognition), real-time traffic information case where shortest path choice under current situation is used, and information access case where reduced trips by information provision. The results suggest that information provision to stranded commuters and their family members makes traffic congestion relieved.

The results also suggest that the dynamic traffic information reduces the congestion thus helps smoother going home.

Keywords: Evacuation; Mode Choice; the Great East Japan Earthquake

An Analysis of Crisis Communication Strategies in Japan’s Nuclear Crisis

Kevin P. Hwang^*, Maria C. Lanting, Elaine Elumba National Cheng Kung University, Taiwan

Crisis communication plays a pertinent role in coordination and management of unexpected events such as natural disasters and calamities. In the light of Japan’s nuclear crisis, this study evaluates the crisis communication strategies used by Tokyo Electric Power Company (TEPCO) to its stakeholders. Using Coomb’s (1998) Situational Crisis Communication framework and guided by Freeman’s Stakeholder framework, this study identifies how TEPCO relate to every stakeholder and what crisis communication strategy it used. By means of quantitative content analysis, findings show that it mainly used excuse strategy to communicate to its stakeholders. The industry and general social, cultural, and political environment appeared to predominantly influence its communication strategies. A crisis communications plan and proactive approaches are included in the conclusion.

Keywords: Crisis communication; TEPCO; Nuclear crisis; Risk management

- 4 -

Behavioural Data Collection and Analysis for Disaster Risk Evaluation

Yasuo ASAKURA^*, Takahiko KUSAKABE, Mayumi YOSHIDA Tokyo Institute of Technology, Japan

Damage of a natural disaster on our society depends on our activity patterns as well as the magnitude of the disaster itself. When we discuss the disaster mitigation program such as evacuation plan from tsunami, it is important to know the spatio-temporal activity patterns of the people in the objective area and to estimate the potential risk in their daily activities. This study aims to develop a survey framework for observing individual’s perception and behaviour during the anticipated disaster. We will apply the Stated Preference (SP) survey after the Probe Person (PP) survey using mobile phones and WEB diary. The collected data are utilized to show the relationships between the risk and the individual behaviour. Through a pilot survey in Hamamatsu in Japan, we will present the people’s recognition on their evacuation places against tsunami and visualize the potential risk of those places on topographic map.

Keywords: Disaster Risk; Behavioural Data Collection; Probe Person Survey

Ⅳ. Special Session: Models and Algorithms for Disaster Operations Management Schedule Optimization for Highway Emergency Repairs

under Large-Scale Supply–Demand Perturbations

Shangyao Yan^a, James C. Chu^b*, Yu-Lin Shih^a

aNational Central University, Taiwan; ^bNational Taiwan University, Taiwan

A model for emergency repair problems under large-scale supply-demand perturbations is developed. The model has the following key features. First, a novel time-space network flow technique is adopted to generate detailed schedules for repair teams and allow dynamic updates of the network due to perturbations. Second, the original schedules prior to the perturbations are considered by controlling the total difference between the original schedule and the adjusted schedule. Third, to reduce computational complexity, the model is formulated with different levels of detail (individual teams versus a team group). The model is also formulated as a special mixed-integer network flow problem with side constraints, which is characterized as NP-hard. An ant-colony-system-based hybrid global search algorithm is developed to efficiently solve large-scale problems. To test how well the model formulation and the heuristic algorithm may perform in actual operations, a case study is conducted using actual data from the 1999 Chi-Chi earthquake in Taiwan. The results show that the proposed model and solution algorithm perform very well and thus have great potential for assisting with the making of emergency repair decisions in the event of disasters given large-scale perturbations in supply and demand.

Keywords: emergency repair; large-scale supply-demand perturbations; ant colony system;

threshold accepting algorithm; time-space network

- 5 -

On Scheduling Recovery Teams for Reaching Shelters in Minimum Time in Disaster Management

I-Lin Wang^*, Tsung-Min Huang National Cheng Kung University, Taiwan

How to evacuate people from dangerous regions affected by a disaster in minimum time is usually a top-priority task in humanitarian logistics. We consider a network restoration problem where the time and resources for restoring broken roads (arcs) have been estimated. We aim to schedule the recovery teams of different capabilities to restore roads such that all the people in shelters (nodes) can be evacuated with minimum time. We first discuss two special cases of our problem: (1) with only one recovery team, and (2) with infinite recovery teams, and show how to solve these two cases by efficient solution methods related with minimum spanning trees and shortest paths. Then, we propose integer programming models to deal with more general cases that take team collaboration and different modes of team resources into consideration. Our proposed solution methods can also be used for restoring the telecommunication connection to all shelters with minimum time.

Keywords: network restoration; humanitarian logistics; disaster management

Modeling Multimodal Freight Transportation Network Performance under Disruptions

Jing Dong^*, Navid Shafieirad, Micah Makaiwi Iowa State University, USA

To facilitate a region’s freight transportation systems planning and operations and minimize risk associated with increasing multimodal freight movements, this paper presents a modeling framework for evaluating and optimizing freight flows on a multimodal transportation network under disruption. Unexpected events such as earthquakes, floods and other manmade or natural disasters would cause significant economic losses. When parts of the transportation network are closed or operated at a reduced capacity, the delay of commodity movements would further increase such losses. Shifting to an alternative route or mode might help to mitigate the negative impacts. In this paper, a multimodal freight transportation network model is developed to simulate commodity movements, evaluate the impacts of disruptions, and develop effective emergency operation plans. Using the Federal Highway Administration’s (FHWA) Freight Analysis Framework Version 3 (FAF3) database, a case study was constructed to model the transportation of cereal grains from Iowa to other states. A hypothetical disruption scenario considering closing part of the railways in central Iowa was evaluated using the proposed network model. In particular, when the shipments are diverted to the Mississippi River on the east side the Iowa, the delays of barges at the locks are estimated using the dynamic queuing model and compared with the static model.

Keywords: Intermodal freight transportation; emergency management; performance evaluation

- 6 -

Heuristics for the Life Detector Routing Problem over a Street Network:

the Search-and-Rescue after Earthquakes

Chung-Cheng Lu^*, Shu-Hao Hu National Taipei University of Technology, Taiwan

Life detectors/finders are seeing a wider use for effectively detecting and locating buried victims in various disaster situations, such as earthquakes, landslides, and avalanches. To efficiently perform search-and-rescue using life detectors after massive earthquakes in urban areas, emergence response agencies need to design shortest routes for life detectors to visit collapsed buildings and find trapped people. The life detector routing problem (LDRP) aims to find a shortest route for the life detector to scan a set of collapsed buildings in an urban street network. Each collapsed building must be scanned at least once. Unlike the traditional traveling salesman problem, a LDRP tour does not physically visit each collapsed building because the life detector can remotely read the signs of life. With this remote detection feature, the life detector only needs to traverse a set of selected streets to scan all of the collapsed buildings in a district, without physically visiting each collapsed building. This paper presents two LDRP algorithms that exploit this feature and realize substantial savings over traditional solutions. In addition to a simple two-stage heuristic, a multi-start simulated annealing heuristic is developed for the LDRP.

The two algorithms are evaluated on several real problem instances generated based on three street networks in Taiwan. We compare the algorithmic performances of the algorithms in solution quality and computational efficiency on the three networks. The effect of the length of detection radius is also examined.

Keywords: disaster operation management; emergency management; life detector; heuristics;

simulated annealing.

Ⅴ. Session: Measuring and Modeling Evacuation Behavior II

Evacuation Behaviors During 2011 Great East Japan Earthquake and Tsunami

Nam Yi Yun^*, Masanori Hamada Waseda University, Japan

Evacuation of the 2011 Great East Japan Earthquake and Tsunami was a large-scale evacuation of over thousands of people escaping from the earthquake-induced tsunami. The survivors’ evacuation experiences provided valuable insights into the factors that helped with survival and some very important practical issues regarding tsunami evacuation. It indicates that the disaster could be the product of other factors such as behavioral or environmental factor. In order to study the human impact in the 2011 Great East Japan earthquake and tsunami, it investigates the relationships among evacuation behaviors (i.e., evacuation starting time), preparedness before the disaster, and evacuee’s characteristics and survival rate of the 2011 disaster. Results using 1,153 data show that behaviors during the disaster differentiated for the survivors and the dead and missing. Moreover, a model for the prediction of the death is developed based on the analysis of each evacuation behavior factors on the fatalities; integrated strategies are proposed and discussed for the reduction of casualties in the future large-scaled natural disasters. Additionally, the analysis from descriptive comments on the survivors and the dead and missing shows differences in behaviors taken by the survivors or by the non-survivors under the disaster, suggesting how to provide a more practical manual.

Keywords: Evacuation behaviors; tsunami; 2011 Great East Japan earthquake and tsunami;

comparative analysis using survivors and non-survivors

- 7 -

Flood Evacuation Decision Modeling for High Risk Urban Area in the Philippines

Ma. Bernadeth Lim^a*, Hector Lim, Jr.^a, Mongkut Piantanakulchai^a, Francis Aldrine Uy^b

aThammasat University, Thailand; ^bSchool of Civil, Environmental and Geological Engineering, Philippines Analysis of influential factors to evacuation decision, a key input to evacuation planning, is important for better management in future evacuations. This study aimed to develop a model of flood evacuation decision using discrete choice modeling framework. Results show that age, income, house ownership, number of house floor levels, and flood level are strong influential factors to flood evacuation decision. An internal validation using bootstrap technique shows consistent results. Whether findings in this study could be generalized to urban cities in the Philippines is a subject for further investigation.

Keywords: Flood; Evacuation Decision; Travel Behavior; Evacuation Modeling; Discrete Choice;

Validation

Directions and Purposes of Trips in a Disaster – an Empirical Analysis of 2011 Tohoku Earthquake Case

Takamasa Iryo^*, Kazunobu Amano Kobe University, Japan

Evacuation is a special type of transport activity in which everyone in a hazardous area need to leave to survive. In the context of transport network analysis, such movements can be modelled as a network in which only a unique destination (i.e. a safe area) exists. Such a network referred to as a single-destination network. As evacuation is a dynamic phenomenon, network performance should be analysed by a methodology of dynamic traffic assignment (DTA). Although general DTA problems are mathematically complicated, it is known that the problems will be much easier for a single-origin or single-destination case. Note that we can use a dummy node to represent a safe area that is wide and hence cannot be physically represented as a one node. The above consideration would imply that analysing evacuation traffic in a network is much easier by adopting the single-destination network assumption, which has also been employed in several existing studies.

However, the single-destination-network assumption for analysing evacuation traffic may be risky if it is not correct. Vehicles going to the same direction can merge into one stream at intersection in the network, while vehicles going to different directions should impede each other and may make traffic in the network stuck. Although the above idea may or may not be collect, any chance of network stuck must be eliminated to avoid the worst case in which anyone cannot move on gridlocked roads, which was actually observed in the 2011 Tohoku Earthquake.

Analysing empirical data upon a disaster is perhaps the strongest approach to understand how far the single-destination assumption is correct. Many efforts have been made to take data of people’s motion in disasters of the 2011 Tohoku Earthquake. There are two approaches. One approach is utilising so-called ‘big data’. The other approach is a questionnaire-based survey like the person-trip survey. The Japanese government (MLIT, ministry of land, infrastructure, transport, and tourism) conducted a person-trip survey for several hours just after the earthquake in the cities suffered by Tsunami. The data is disclosed to any scholar. The advantage of the questionnaire-based survey over the big data is that the questionnaire can detect the reason of each trip. This implies that we can know whether each trip aimed to evacuate or had another purpose such as meeting family.

This study analyses the MLIT data to discover the major direction of motions and purposes of trips after the earthquake occurs. The purposes of trips were classified into ‘evacuation’, ‘meeting or searching family’, and ‘others’. The major results of the analysis are:

1. Although evacuation trips were dominant in most hours after the earthquake, family-related trips also significantly existed for around half hour just after the earthquake.

2. Directions of motions were strongly biased to the mountain side for evacuation trips, while there was no significant bias for family-related trips.

- 8 -

the earlier stage of the disaster to avoid a network begin stuck before people complete evacuation.

Keywords: Evacuation from tsunami, Empirical data analysis, 2011 Tohoku Earthquake

Verification and Validation of Numerical Analysis Method for Tsunami Mass Evacuation

Muneo Hori ^a*, Takashi Seumatsu^b, Hideaki Araki^c, Yoshihiro Okumura^d

aThe University of Tokyo, Japan; ^bVector Research Institute, Inc., Japan;

cKozo Keikaku Engineering, Inc., Japan; ^dKyoto University, , Japan

In recent years, numerical analysis of tsunami mass evacuation is more often used in order to make an emergency evacuation plan. However, there remain questions about the quality of the analysis method that is related to the accuracy and reliability of the results. This paper clarifies a basic concept of the quality of the analysis method for mass evacuation, and proposes a method of assuring the quality according to the concept. The method is standard in view of software engineering, as it takes the two steps of verification and validation; verification is to examine whether an action of an evacuee model is correctly computed and validation is to examine whether characteristics of mass behavior is computed to desired accuracy and reliability. The method of verification and validation is applied to four numerical analysis of tsunami mass evacuation; data obtained from 2011 Great East Japan Earthquake Disaster are used for validation.

The applicability of the proposed method as the quality assurance is examined.

Keywords: mass evacuation simulation; quality assurance; verification and validation; tsunami

- 9 -

Ⅵ. Keynote Speech: Prof. Masao, KUWAHARA, Tohoku University

Evacuation behavior and traffic congestion just after the Great East Japan Earthquake

Prof. Masao KUWAHARA

Graduate School of Information Sciences, Tohoku University, Japan

This presentation explain our recent study on monitoring how people moved just after the zreat East Japan earthquake using probe and mobile data. At 14:46 on Friday, March 11, 2011, one of the biggest earthquakes (magnitude 9.0) our country ever experienced occurred approximately 70km east of Sanriku coast in Tohoku. Then, about 40-50 minutes later, a mountainous tsunami caused by the earthquake hit the 500km long coast facing the Pacific Ocean. The tsunami with the height more than 20m, brought catastrophic damage along the coastal area and also caused serious nuclear power problems in Fukushima.

Recently, advanced communication technologies provide information generated from individuals. For instance, a probe vehicle carrying GPS and a communication equipment sends its position at a certain time interval and a mobile phone also gives us its position under a certain circumstance. Under the chaotic condition just after the earthquake and tsunami, most of infrastructure sensors, such as traffic detectors, were completely damaged at the coastal area and only those probe and mobile data were available to monitor how people behaved under the serious disaster.

We focused on Ishinomaki, the worst disaster area in northern Miyagi Prefecture, gathering probe and mobile phone data. After the data processing of interpolating raw data, map-matching onto a digital map, and clustering them for modes, the following surprising, even shocking facts were found: (1) Pick-up behavior: A number of people did not immediately evacuate, but moved toward the tsunami area – probably to observe their family members and relatives. (2) Sudden Grid-Lock: All vehicles on roads in the city stopped under the violent earthquake, and then a considerable number of vehicles came onto the roads at once. This brought very sudden traffic congestion and caused Grid-Lock phenomena

Ⅶ. Session: Network Modeling for Emergency Evacuation

Heterogeneous Multi-modal Evacuation Modelling Considering Detailed Interactions and Mitigation Measures

Leonel Aguilar^a, Lalith Wijerathne^a*, Stephen Jacob^b, Muneo Hori^a, Tsuyoshi Ichimura^a, Seizo Tanaka^a

aThe University of Tokyo, Japan; ^bJGC Corporation, Japan

This paper presents a scalable multi-agent based evacuation simulation software capable of simulating mixed mode evacuations of large areas including sub meter details of the environment.

A mathematical framework is introduced in order to provide a framework for the specification and description of the software, enabling to compare the developed mutli agent system with other existing evacuation simulation tools and identify underlying differences and areas of improvement. Details of the high resolutions environment, including the modeling of dynamic changes like earthquake induced damages, etc. are briefly explained. Further, the basic constituent functions of autonomous agents, which enable them to perceive their visible surrounding and interact with it and neighboring agents, are also briefly explained. Several demonstrative large urban area evacuation scenarios are presented in order to highlight the need of incorporating detailed models of environment and the complex agent functions. Most of these scenarios, like night time evacuation in damaged environment, cannot be simulated with the simplified models, without sacrificing accuracy. These scenarios range from evacuations with different lighting conditions and environmental damages, to mixed mode evacuations using cars and pedestrians.

Keywords: Type your keywords here, separated by semicolons

- 10 -

Network Based Clustering in Disasters for the Emergency Medical Service

Albert Y. Chen^*, Ting-Yi Yu National Taiwan University, Taiwan

The Emergency Medical Service (EMS) is the function for the provision of immediate and appropriate aid to a patient. As part of the traditional triad first responders, EMS plays an important role in disaster response. In this work, the dependency of the EMS on the service provided by the transportation infrastructure is considered. The objective of this research is to improve the effectiveness of EMS after the disaster. In this research, the Geographic Information System (GIS) serves as a tool for the service area analysis of EMS before and after of the disaster.

Cluster analysis is applied to determine the on-post EMS facilities for temporary medical services.

Keywords: EMS; disaster; GIS; clustering; R-tree

A Link-based Evacuation Model for Emergency Evacuation under Typhoons

Tsai-Yun Liao^*a, Yi-Wei Chen^b, and Ya-Han Chang^b

aNational Chiayi University, Taiwan; ^bNational Cheng Kung University, Taiwan

Climate change has become a global and environmental issue in the 21st century. All natural, human, and transportation infrastructure systems are affected by the climate change. Emergency management has become an important issue around the world. Emergency management aims at decreasing vulnerability and coping with disasters. Emergency evacuation, one important issue in emergency management, is defined as the rapid mass removal of people from dangerous places to safer locations. Transportation infrastructure is affected by natural disasters and the issue of network reliability has become an important issue in recent years.

Taiwan is often hit by typhoon. Typhoons always bring strong wind and heavy rainfall that result in serious disasters. Especially the precipitation is the major reason which causes the mudslide and roads damage. Until now, few researches were considered about “network reliability” issue and most of them assumed that all roadway links can function properly in disaster condition. Therefore, the precipitation is one of the important measurements to evaluate the network reliability.

This research aims to study evacuation in heavy precipitation scenarios and minimize the total evacuation cost including shortest distance and minimum delay. A link-based evacuation mathematical model is constructed to maximize the evacuation performance with the consideration of evacuation routing and network reliability. In this research, network reliability is evaluated based on precipitation. The evacuation routing problem is solved through a dynamic evacuation model with minimum evacuation cost. The interrelationship between the precipitation and the road damage is established through support vector regression (SVR) model.

Keywords: evacuation; emergency management; routing; network reliability; support vector regression model

- 11 -

Optimizing Substitute Bus Services after a Metro Failure: A Case Study of the 9-27 Metro Accident in Shanghai

Shuyang Zhang^a, Hong K. Lo^a*, Xiaoguang Yang^b

aThe Hong Kong University of Science and Technology, Hong Kong; ^bTongji University, China

Metro system provides efficient transport services to large numbers of travelers in urban areas around the world. Any major incident will result in operation disruption, and severe train and passenger delay. Substitute buses (SB) services can be an effective option to maintain a certain level of operation order. In this paper, we propose a formulation for planning and optimizing SB operations in the aftermath of a metro accident, and a case study based on the 9-27 metro accident in Shanghai is used to illustrate the approach.

Keywords: Disruption management; Bus operations; Metro failure; Substitute buses (SB) services;

9-27 metro accident

Ⅷ. Keynote Speech : Prof. Pamela Murray-Tuite, Virginia Tech

Evacuation Demand Modeling: Where we were, where we are, where we need to go

Prof. Pamela Murray-Tuite

Department of Civil and Environmental Engineering, Virginia Tech, USA

The state of the art in evacuation demand modeling has advanced dramatically over the last 15-20 years. This talk reviews recent advances in this area and points to directions left to pursue in the future. Early demand models were based on a multiple of the normal peak period demand or professional judgement, and had few or no ties to social behavior. Recent studies advance our understanding of different demand-related travel decisions, such as whether to evacuate or stay, when to leave, in what type of accommodations to stay, which destination to choose, what modes of transportation to use, how many household vehicles to take, what routes to use, and in what activities to participate prior to or during the evacuation trip. These more complex and socially-based models allow understanding of the factors associated with the different decisions and the development of strategies, policies, and practices that address evacuation plan weaknesses related to those influential factors. Accounting for more complex social behavior also allows us to develop evacuation management strategies that better align with evacuees’ travel intentions and to more accurately predict evacuation times. As the near future brings changes in household composition, technology, median population age, urbanization, and personal values, among many others, our behavior and demand models must continue to evolve to address these changes and emerging needs.

- 12 -

Ⅸ. Session: Modelling for Emergency Evacuation

A General Activity-based Methodology for Simulating Multimodal Transportation Networks during Emergencies

Jeroen P.T. van der Gun^*, Adam J. Pel, Bart van Arem Delft University of Technology, Holland

Many possible emergency conditions, including evacuations, negatively affect the urban transportation system by substantially increasing the travel demand and/or reducing the supplied capacity. We develop a new, general and efficient methodology suitable for simulating multimodal transportation systems subject to a wide range of emergency situations, based on the integration of an activity-based choice model with both pre-trip and en-route choices, and a macroscopic or mesoscopic dynamic network loading model. The model structure first estimates the daily equilibrium and then simulates the emergency without iterations. An evacuation case study for Delft shows the feasibility of the methodology and yields practical insights.

Keywords: urban emergencies; evacuation modelling; choice modelling; activity-based modelling;

dynamic network loading; multimodal networks; multimodal evacuation.

Minimizing Evacuation Time through Optimal Inflow Control in a Tsunami Inundated Area

Junji URATA^a*, Eiji HATO^a, Takuya MARUYAMA^b

aThe University of Tokyo, Japan; ^bKumamoto University, Japan

In the 2011 Great East Japan Earthquake, many people evacuated by car even to short distance refuges and observed delays in evacuation due to traffic congestion. But all people should evacuate in a short time, e.g., less than one hour, by the tsunami. Existence of safe high grounds, like mountains, within a few kilometers, but road capacity to mountain is usually small.

The objective of this research is to minimize evacuation time through optimal dynamic inflow control. The evacuation time is the time when the evacuation of all people is finished. The control function is dynamic and continuous in order to find the first-best solution. The calculation method to optimize dynamic inflow control is the Pontryagin’s maximum principle. The maximum principle can find the minimum evacuation finishing time. But the dynamic programming which is another method to optimize a dynamic function cannot find the minimum evacuation finishing time and cannot apply to this problem. This problem is a bi-level programming. In the upper level, the transportation manager decides an inflow control rate at an intersection. In the lower level, the evacuees choose their route and behaviors. A modelling their behaviors introduces the dynamic user optimal assignment model (DUO). The DUO assignment is reactive because users are assumed to choose their routes based on present instantaneous travel times and is subjective to the First-In-First-Out (FIFO) discipline and the flow conservation, the physical queue.

The calculation in a simple network determines optimal dynamic inflow control by application of the maximum principle. The Hamiltonian is defined by the flow conservation at the intersection and capacity constraints of their roads are their constraints. The transversality condition develops the expression for the range of optimal process of inflow control. The range can make a response to changing the departure timing ratio in evacuation.

Keywords: Type your keywords here, separated by semicolons

- 13 -

Secure Communication Analysis for Emergency Evacuation during Blackouts

Kuang-Chiu Huang^*

National Cheng Kung University, Taiwan

Electric power outages often accompany major crisis such as severe weather events and fires.

In this scenarios, loss of electricity and communications make people panic, confused and most traffic lights without backup electricity would work abnormal make emergency evacuations become complicated and difficult.

In order to enhance communication capabilities during blackouts for evacuations, the paper takes two steps analysis. Compare current communication tools systematically to figure out which is robust against power outages and overwhelming traffic volume for the first step. Evaluate what are required confidentiality, authentication and integrity to be embedded into life-safety messages for emergency managers, security officers, first aid, rescue squad and public as the next step.

To alleviate economic losses and protect people lives, it is necessary to pay attention on realistic performance of communication systems to gather, transmit and deliver critical information for command and executions. The outcome of the research is valuable for all emergency related institutions to design evacuations with reliable and secure communication systems.

Keywords: blackout; communication systems; evacuation; emergency

A Modified Entropy-based Dynamic Gravity Model for the Trip Distribution Problem under Typhoons

Wei-Ming Ho^a, Li-Wen Chen^b, Ta-Yin Hu^*a

aNational Cheng Kung University, Taiwan; ^bChung Hua University, Taiwan

The typhoon Morakot, which formed on August 2, 2009, was the deadliest typhoon in Taiwan’s history, responsible for over 700 deaths on the island. During the typhoon evacuation process, one critical issue is how to efficiently distribute the evacuation trips to a limited number of the shelters based on both the spatial and temporal considerations. This research proposes a modified entropy-based dynamic gravity model to reflect the spatial and temporal distribution of the evacuees and the shelters. A unique feature of proposed model is that the entropy is explicitly incorporated within the travel cost constraints. The spatial and temporal relationships between evacuees and shelters can be reflected through the impedance functions and the discretized time intervals with better performances than the traditional model. A simulation-assignment model is applied to generate the zone to zone travel time. A calibration analysis based on the solution procedure is conducted for the Jiasian network, in Kaohsiung City, which was heavily affected by the Typhoon Morakot. The calibration results show that the modified entropy-based dynamic gravity model leads to better convergence patterns in the entropy values, higher travel cost coefficients and lower average generalized trip costs than the traditional model, and is suitable for use with the evacuation plan under typhoons.

Keywords: trip distribution; evacuation; entropy; typhoon

- 14 -

Ⅹ.Student Competition (National Taiwan University, Taiwan)

Evacuation Planning for Dam-Break Inundation: Perspectives of Multidisciplinary Modeling and Engineering Education

Presented by Jui-Ming Chang, Cheng-Han Kuo, Tzu-Wei Kuo, Ting-Sia Ye, Fu-Jian Shiu, Tzu-Yin Chen, Kuan-Wen Chou, Chih-Chung Wang, Yi-Ru Wu,

Hsin-Yi Yeh, Yu-Chieh Lin, Tso-An Chang, Hou-Wei Lee

Instructed by Herve Capart, Jiing-Yun You, Pai-Hui Hsu, Albert Chen, Yu-Ting Hsu

Four student teams in the capstone course “Civil Engineering Capstone Challenge — Dam-Break Inundation and Evacuation” from National Taiwan University conduct a semester-long project to study dam-break hazards associated with the failure of Shihmen Dam and propose evacuation plans for the probable downstream Inundation. Shihmen Dam is Taiwan's third largest reservoir and the major water supply for more than three million people in northern Taiwan.

Different failure scenarios are investigated, including overtopping, piping, upstream landslide and terror attack. The consequent flood discharge along Dahan River is analyzed based on open-channel hydraulics and scale model tests to determine the dynamics of flood impact over several highly-populated riverside townships. GIS techniques are further employed to create inundation maps that describe the spatiotemporal evolution of the disaster-affected regions.

Accordingly, evacuation plans are developed to relocate the affected population to the pre-selected places of safety, where a microscopic traffic simulator is used to model vehicular and pedestrian flows of the evacuation traffic and evaluate the performance of the associated transportation networks/systems under various operational strategies.

To comprehensively and practically address the evacuation problem of dam-break inundation, a dam visit and dam manager interview are arranged for students to have better understanding of expert’s concerns and the challenging aspects of the problem. Additionally, the teams collaborate to conduct field surveys as well, in terms of geographical nature and demographic/societal characterization over the affected regions, which enable a better sense to identify critical infrastructures, select sheltering facilities, and account for the behavioral realism of evacuees. The project highlights the sophistication of an evacuation problem that needs for holistic and seamless cooperation across the boundaries of jurisdiction and knowledge of different engineering domains.

This capstone course is designed for students to learn from practice and seeks to foster them the capability to tackle a multidisciplinary problem that they may encounter in their future engineering careers. As a capstone course, it attempts to summarize the four-year Civil Engineering undergraduate program. On the other hand, it is the first trial to introduce the fundamental and integrated aspects of evacuation planning as the focus of an undergraduate course. Although the course is still at an experimental level and may need significant improvement, it can also be an interesting example from the perspective of engineering education.

Keywords: Dam-break evacuation; Inundation map; Microscopic traffic simulator;

Multidisciplinary engineering problem

- 15 -