MPE 2013+ Workshop on Natural Disasters

May 13 - 15, 2015
Georgia Institute of Technology

Organizers:
Lora Billings, Montclair State University, billingsl at mail.montclair.edu
Carlos Castillo-Chavez, Arizona State University, ccchavez at asu.edu
Midge Cozzens, DIMACS, Rutgers University, midgec at dimacs.rutgers.edu
Eva Lee, Chair, Georgia Institute of Technology, eva.lee at gatech.edu
John Mitchell, Rensselaer Polytechnic Institute, mitchj at rpi.edu
Fred Roberts, DIMACS, Rutgers University, froberts at dimacs.rutgers.edu
William (Al) Wallace, Rensselaer Polytechnic Institute, wallaw at rpi.edu
Presented under the auspices of the DIMACS Special Program: Mathematics of Planet Earth 2013+.

Abstracts:


Carlos Castillo-Chavez, Arizona State University

Title: Mathematical Models for the Study of Single Epidemic Outbreaks with Applications to Ebola, SARS and Influenza

Over the past few decades we have seen the disruptive and fatal consequences that accompany SARS, Ebola and pandemic influenza outbreaks.

What do we know and don't know about their dynamics, intervention measures and responses? What have we learned from them?

In this talk, I will discuss the impact that the study of single disease outbreaks has had in the fields of computational and mathematical epidemiology, modeling and public health.

I will use the 2002-2003 SARS epidemic, the 2009-10 influenza pandemic and the 2014-15 Ebola outbreaks in West Africa to highlight the impact of models and mathematics in addressing public health concerns as well as on identifying challenges and opportunities for theoreticians, epidemiologists, modelers, computational scientists and mathematicians.


Cham Dallas, University of Georgia

Title: Addressing Healthcare System Security and Resilience in High Consequence Natural Disasters

In high consequence natural disaster events, it is likely that in addition to a lack of adequate numbers of properly trained personnel to treat mass casualties, it will become increasingly difficult to protect both health care workers and patients in the accompanying social upheaval. There is little doubt that high consequence events such as pandemic flu outbreaks will exceed the medical care response capacity of nearly all emergency response systems, yet the danger to the operation of these facilities is even greater from a lack of adequate security in crisis. Indeed, security has historically been one of those prime budget cutting items for healthcare systems under fiscal pressure. However justified that might be in relatively stable times, the lack of adequate security in times of crisis will be devastating to the continued operation of hospitals and health care communities. Ironically, just when health care units adopt their most crucial functions to the community, they are likely to be invalidated in those irreplaceable roles at the very peak of the time period in which they are needed by the health security gap.

Approaches to meet this critical workforce need will be presented, including the incorporation of infrastructure modifications, mobilizing properly trained security and health care personnel, and devising innovative and fiscally feasible response plans to address this new threat environment. Of particular importance will be the compilation and training of associated health professionals and security personnel not currently utilized in emergency plans in local hospitals. In addition, options for training and organizing citizens into support groups for the security needs of these and traditional emergency medical personnel will be presented.

Common operating improvements for security protection will be shown, which would include decreasing access points for the hospitals, controlling "surge" patient flow, managing intruders, incorporating transportation barriers, and setting standards for exclusion zones. Examples of the consequences of not incorporating such measures will be discussed, such as the collapse of healthcare systems during Hurricane Katrina. Effectiveness of extending this health care security training to citizen volunteers in ancillary health care and security roles and direct medical security support will be discussed.


Jeffrey W. Herrmann, University of Maryland

Title: Mitigating the Risk of an Anthrax Attack with Medical Countermeasures

The deliberate release of aerosolized anthrax spores in a large city will expose many thousands of residents to this deadly disease. Promptly distributing medical countermeasures (antibiotics) to those exposed is a key step in preventing illness and deaths. Avoiding delays in this distribution is critical, but such a response will require enormous resources. State and local health departments have developed contingency plans for points of dispensing (PODs), the primary distribution channel, and other countermeasures and strategies have been proposed and tested, including prepositioning. Prepositioned medications include forward - deployed (local) stockpiles, workplace and hospital caches, and predispensed medical countermeasures that are stored by heads of households.

This talk will present operations research models for a risk management problem in the context of preparing for a bioterrorism attack. The first model is a predictive model that estimates the number of deaths based on information about the attack scenario and the use of prepositioned medical countermeasures. The second model is an optimization model that can determine the optimal allocation of medical countermeasures.

This talk will also discuss the analysis of social media data to describe the spread of information about an event and how those results can be used in these models.


James H. Lambert, University of Virginia

Title: Resilience Analytics: Population Behaviors Influencing Preparedness and Risk Management for a Dirty Bomb

Important questions arise with sheltering populations in regional emergencies such as a terror attacks and natural disasters. The topic engages experts across domains including emergency management, infrastructure engineering, health care, risk communication, water and food supply, information sharing, freight logistics, security, business continuity, and others. Knowledge of population behaviors should influence the many dimensions of resilience, including protection, prevention, response, and recovery. Of particular concern are the behaviors and needs of the resident and non-resident populations, including those at home, at work, and traveling. This talk describes an engineering study that builds on a telephone survey of several thousand residents of the national capital region. We characterize detailed evidence of prospective behaviors in the event of a variety of dirty-bomb scenarios. We suggest the assumptions of population behaviors that should most influence agency priorities. We describe theory of resilience of cyber-physical systems in light of evolving understanding of population and workforce behaviors, particularly for resilience analytics and the value of new information.

Biography:

The presenter is the President-Elect of the Society for Risk Analysis. He is a Fellow of the American Society of Civil Engineers and a Fellow of the Society for Risk Analysis. He is a Senior Member of the IEEE and a registered Professional Engineer. He is a Diplomate of the American Academy of Water Resources Engineers. He is a Research Professor of Systems and Information Engineering at the University of Virginia. He holds a joint appointment in the Department of Engineering and Society. He serves on panels of the National Academies on the topics of engineering systems and risk analysis. He is a co-organizer of several international meetings for NATO. He gave recent invited talks in Japan, Italy, Germany, Croatia, Denmark, Canada, Mexico, China, Singapore, United Arab Emirates, Turkey, France, Greece, Spain, Iceland, Chile, and the USA.


Earl "Rusty" Lee, University of Delaware

Title: Models to Assess and Improve Community Resilience

A generally accepted definition of community resilience is "a measure of the sustained ability of a community to utilize available resources to respond to, withstand, and recover from adverse situations". But which resources and what situations? And how do we best invest to respond to, withstand and recover? This work has developed a framework for community's to measure and improve their resilience. First, an Analytic Hierarchal Process (AHP) is provided for a community to determine the relative value and importance of community infrastructure, with respect to a specific disaster. This AHP process is repeated for a broad scenario of adverse, but plausible situations. This set of prioritized community assets is then combined with an interdependent, critical infrastructure model to assess how failures in public and private infrastructure components affects the community's ability to respond to an adverse situation. These failures and impacts can then guide discussion and decision making for planning, mitigation or investment to improve community resilience.


YiFan Liu, Georgia Institute of Technology

Title: Strategies for Ebola Containment: A Biological-Behavioral-Logistics Computation Decision Framework

The 2014 West African Ebola virus outbreak is the largest ever to occur. It started in December 2013 and has since infected over 25,000 individuals and resulted in over 10,000 deaths. Mathematical epidemiology models can be useful for understanding the biology of disease and its spread characteristics. They can also provide an analytic framework for public health responders and policy makers to derive strategies for effective containment. In this study, we proposed a general-purpose epidemiology model for Ebola virus based on our 6-stage compartmental disease model. Population is divided into subgroups based on their risk factors and are assigned their own compartments and disease and risk parameters. Specifically these parameters reflect the difference in protection and infection contact media among friends, families, passer-bys, and healthcare workers. Further, a distinct characteristic of Ebola involves infection during the burial process. Our model includes a special state that describes the burial process (unsafe versus safe burial). We use the data from Guinea, Sierra Leone and Liberia for training of these model parameters. Using data from December 2013 to September 2014 for parameters estimation, we report the estimated number of beds required and the expected containment results for October 2014 until April 30, 2015. The model is generalizable; other human behavior and community patterns at different time horizon can be incorporated using time-variant parameters. This work is joined with the Centers for Disease Control and Prevention. It is partially supported by a grant from the National Science Foundation.


Rick Luettich, University of North Carolina at Chapel Hill

Title: Using Models to Predict and Mitigate Coastal Flooding

Coastal regions are highly desirable places to live, work, recreate and retire. In fact, population growth in southeastern Atlantic and Gulf of Mexico coastal counties is nearly twice that of the national average. However, these same areas are subject to impact from hurricanes, one of the most powerful storms on Earth whose destructive potential is increasing due to climate change and relative sea-level rise.

Rapid developments in mathematical models over the past decade have provided powerful tools for predicting the severity of coastal flooding which is widely recognized as the leading cause of storm related damage and loss of life along many coasts. However, applying these models to complex coastal areas can be challenging and computationally demanding. Thus they have most often been used in probabilistic hazard assessments (e.g., to delineate areas within the "100-year" flood zone) and in hazard mitigation design studies (e.g., designing engineered features to reduce flooding). A particularly challenging application is using these models to help inform emergency response by forecasting conditions in advance of an actual event.

I will provide an overview of the widely used ADCIRC coastal hazard modeling system and describe the challenges of using ADCIRC to forecast coastal flooding associated with severe storms.


Martin Meltzer, CDC

Title: What does the Public Health Community Expect from Modelers during an Emergency Response? CDC Perspective

CDC and other public health leaders need data in order to be able to make informed decisions when preparing for and responding to Public Health Emergencies such as epidemics or pandemics of infectious diseases, or possible bioterrorist threats. Cortical decision often need to be made ahead of when relevant data from the field are available, and thus public health decision makers increasingly turn to modelers to provide estimates of the needed data. Simple spreadsheet models that can be rapidly built, are straight-forward and relatively easy to understand with clear data inputs and outputs are often the preferred method of producing data that meet the requirements set by the decision makers. Such spreadsheet-based models allow decision-makers, who are not modeling experts, to see how outputs change with different variables. Models can be updated to include hypothetical scenarios or future data that are not currently included. These models enable leaders to understand possible consequences of various decisions.


John Mitchell, RPI

Title: Using Reinforcement Learning to Improve Infrastructure Resilience

Our goal is to optimize the efficacy of reinforcing an existing infrastructure network to prevent unmet demand from imminent disruptions. The probabilities of failures for edges in the network are adjusted as the event gets closer, with the uncertainty information becoming more reliable. This leads to a sequential decision-making process, with actions taken at distinct stages in the lead up to the event. To avoid the "curses of dimensionality", we formulate and solve an approximate dynamic program.


David Ormes, US Coast Guard

Title: Geographic Response Planning in the Chesapeake Bay

Introduction
The Chesapeake Bay contains over 11,684 miles of shoreline and 4,479 square miles of surface area. It is a vibrant natural resource providing crucial habitat for fish, shellfish, and wildlife, and abundant economic benefit for the region.

Since 2011, several Area Committees within the Fifth Coast Guard District (Virginia and Coastal Maryland Area Committee and Upper Chesapeake Estuary Area Committee) have developed Geographic Response Plans for their respective Area Contingency Plans (ACPs) encompassing the bay and its tributaries.

Geographic Response Plan (GRP) Overview
GRPs provide tactical guidance to first responders to ensure that sensitive areas and resources are protected in the immediate aftermath of an oil spill. GRPs contain maps and descriptions of areas and resources, outline strategies to protect those resources, incorporate pre-determined booming and equipment deployment strategies, and set priorities for various spill scenarios. Developed using ESRI's ArcGIS software, the GRPs contain interactive geographic maps illustrating data of importance to planners and oil spill responders (sensitive habitat, threatened or endangered species, archaeological sites, staging areas, optimum sites for placement of spill response equipment, etc.) Data is presented as GIS overlays which can be manipulated by the user to add or remove items of interest.

Interagency Process
Development of the GRPs included extensive coordination with federal, state and local agencies, non-governmental organizations, and the private sector. Planning input was obtained, in part, through a series of workshops to identify sensitive resources at risk and develop booming and other spill response strategies to aid in their protection. The collaborative, consensus-driven approach taken was critical in ensuring that all stakeholder concerns were considered, and that the initial protective strategies represented in the GRPs were fully supported.

Incident Command System (ICS) Integration
A distinguishing feature of the Fifth District GRPs is the inclusion of pre-scripted ICS forms, notably the Resources at Risk Summary form (ICS-232) and Assignment List (ICS-204) form to supplement the GRP response strategies and facilitate the rapid development of ICS Incident Action Plans. The ICS-232 summarizes environmental, historic, and cultural resources at risk within a given area, while the ICS-204 highlights personnel and equipment required to implement selected response strategies.

Conclusion
GRPs can help reduce the environmental and economic consequences of an oil spill. Because GRP initial response strategies and tactics are pre-approved by the Area Committee, they can be implemented without delay during the initial critical hours of response.


Srinivas Peeta, Purdue University

Title: Integrating Planning and Operations: Insights from Disaster Response Modeling

This talk briefly illustrates some insights from a modeling perspective related to the need, benefits and challenges associated with integrating planning and operations in the context of the transportation response to disaster events. The modeling characteristics and dimensions are identified. Insights and implementation issues are illustrated using earthquake and no-notice evacuation contexts.


Richard Retz, Mayor's Office, Houston, Texas

Title: How Much Is Enough When It Comes To Disaster Response

Predicting the immediate needs after a disaster is difficult at best, but maintaining that fine balance between under or over responding to a disaster is often impossible. Lives matter, but so does the allocation of precious resources. How can we better predict the needs of disaster victims and meet those needs without squandering limited personnel, supplies and response budgets?

Highlighting real world examples from Hurricanes that have affected the Texas Gulf Coast learn what tools have been develop to help Emergency Managers and where the shortfalls still remain. The development of new emergency preparedness and response models/tools is needed to save lives and ensure that the right help is delivered to those that need it the most.


Elaine Spiller, Marquette University

Title: Combining Deterministic and Stochastic Models For Hazard Forecasting and Uncertainty Quantification

Large granular volcanic events - pyroclastic flows - are rare yet potentially devastating for communities situated near volcanoes. It is human nature to suppose that volcanoes (and other geophysical phenomenon) will behave as they have previously, but proper assessment of natural hazards must include scenarios that are "worse" than any previously recorded. This task inherently requires a combination of physical and statistical modeling as well as large scale computations. Furthermore, one must carefully handle the rare nature of the most dangerous events to keep probability calculations computationally feasible. To this end, model surrogates prove a powerful tool in developing probabilistic hazard maps. In this talk, we will describe an efficient strategy for both assessing hazards and for quantifying multiple sources of uncertainty that arise in the modeling process.

Although developed in the context granular volcanic flows, our process for efficient and flexible probabilistic hazard mapping is not specific to a particular hazard, physical model, or scenario model. Thus we believe our approach could aid in assessing the impact of a wide variety of potential hazards under different environmental scenarios.


Contributed talks and students:


Jan M. Baetens, Ghent University, Ghent, Belgium

Title: Spatially Explicit Modelling of Wildfires in Europe

Since wildfires are causing substantial economic, ecological and social losses in many parts of the world, several fire-prone countries or regions have deployed so-called forest information systems. Yet, up to this day most of them still lack a simulation module that allows for a quasi real-time simulation of active wildfires given their extent and weather predictions. Such a module could support fire suppression teams by enabling quasi real-time scenario analyses, and as such support the emergency decision making process, though its development in a European context is an intricate task due to the heterogeneity and fragmentation of the European landscape, as well as the intense exploitation of most part of the continent.

As a first step towards the further advancement of the European Forest Fire Information System, the Joint Research Centre of the European Commission has selected ten well-documented fires from several member states for testing, calibrating and validating prospective European forest fire simulators. As a next step, we show in this work how a spatially explicit model can be used to realistically simulate the propagation of wildfires in Europe. Our model constitutes an extension to the one proposed by Alexandridis et al. (2011) to describe the wildfire that swept through Spetses Island, Greece, in 1990.


Martha Bauver, Montclair State University

Title: Computing the Optimal Path in Stochastic Dynamical Systems

In stochastic systems, we are often interested in quantifying the optimal path that maximizes the probability of switching between metastable states. However, in high-dimensional systems, the optimal path is often extremely difficult to approximate. We consider problems from population biology and demonstrate a constructive methodology to quantify the optimal path using a combination of finite-time Lyaponuv exponents, statistical selection criteria, and a Newton-based iterative minimizing scheme.


Elizabeth B. Connelly, University of Virginia

Title: Multicriteria Decision Analytics to Inform Emergency Management and Planning Subject to Uncertain Disaster Scenarios

An urgency of community resilience through adequate disaster preparedness and response operations has increased in the aftermaths of the Boston Marathon bombings, typhoon Haiyan, Japan earthquake and tsunami, hurricane Katrina, and other high profile disaster events. High-level coordination among large, independent organizations including police, military, and transportation agencies is critical for evacuation and relief efforts following an emergency. As there are a growing number of complexities in the disaster response operations, there is opportunity to advance research to improve emergency response efforts. In particular, there is need for strategic prioritization of investment alternatives for disaster response and recovery supply chain operations, with recognition of the diverse uncertainties and criteria influencing the decisions of emergency response agencies. Emergency plans can be rendered ineffective or become unrealistic when they fail to address associated deep uncertainties stemming from emergent and future disaster scenarios, particularly when disparate elements are contributing. For example, distribution of critical aid resources is vulnerable to uncertain population behaviors, climate factors, integrity of the transportation network, public perception concerns, and other factors. It is also critical for emergency response agencies to analyze their investments with consideration of multiple and possibly conflicting criteria. For example, consideration of health, safety, cost, environmental impacts, and other non-commensurate criteria is necessary to reflect emergency management objectives.

In particular, there is critical need for scenario-based prioritization of emergency response efforts in world capitals such as Rio de Janeiro, Brazil. In 2012, the United Nations International Strategy for Disaster Reduction announced the opening of a Centre of Excellence for Disaster Risk Reduction. The decision to locate the Centre in Rio de Janeiro, Brazil was in part due to recent landslides that resulted in 900 deaths and economic losses exceeding one billion dollars from 2010-2011. Losses due to floods in recent years have approached 10% of the GDP of the entire nation. The establishment of the Centre is harmonious with Brazil's participation in the Hyogo Framework for Action, an international agreement and plan for coordination of disaster and related risk reduction. Further, Brazil offers an interesting case study because of increased press and international attention surrounding both the 2014 FIFA World Cup and the 2016 Summer Olympics. This paper will demonstrate an integration of scenario planning with multi-criteria decision analysis for the prioritization of emergency disaster planning initiatives for Rio de Janeiro, Brazil. This methodology will enable emergency planners to identify investment alternatives with the highest priority as well as which initiatives are the most robust across a variety of emergency scenarios. The conflicting criteria considered for this analysis include health and safety, cost, environmental considerations, and coordination and planning of the government and private sector, to address the plurality of viewpoints involved in disaster preparedness. The criteria also address the need for innovative and adaptive emergency plans in the face of unforeseeable disaster scenarios. The investment alternatives include increasing the supply of food, water, sheltering, and medical supplies, improving coordination among responders, and building infrastructure for distributing information prior to (for educational purposes) and following an emergency, among others. Emergent conditions, including fluctuations in tourist or favela populations, population behaviors, and damage to infrastructure, etc., are combined to form disaster scenarios. The scenarios include three natural disaster scenarios, namely landslide, heavy rainfall, and drought, as well as one related to a radiological disaster. Additionally, two scenarios consider potential terrorist attacks or other emergencies during the 2014 World Cup and 2016 Summer Olympics, when increased tourism is likely to strain agency resources.

The goal of this methodology is to guide emergency management agencies in addressing key deep uncertainties and to close existing knowledge gaps to systematically prioritize investment alternatives under several disaster scenarios. The analysis specifically considers the viewpoints of the agencies responsible for protective investment for disaster preparedness within the Brazilian government. The supporting agencies include the key military branches, other national and regional agencies and organizations responsible for rescue efforts, and various agencies involved in planning for emergencies affecting the transportation infrastructure, environment, and energy infrastructure. Qualitative ratings are assigned to investment alternatives based on how well they address categories of criteria. Each criterion is weighted to determine the relative importance of the criteria, where the relative importance of each criterion may change during according to scenario. The results give the rank order of initiatives based on value scores of each initiative for the scenarios, revealing how priorities change as a result of various scenarios, identifying scenarios that are most and least impactful to priorities, and identify investment alternatives that are most and least robust to emergent conditions. The results will provide decision support to these emergency planners to preserve supply chain operations in response to combinations of emergent conditions that constitute a variety of disaster scenarios. The analytical methods are suitable for humanitarian relief organization investment, regional disaster analysis, and private sector supply chains.


Barry Dewitt, Carnegie Mellon University

Title: Tornado Risk Perception from Visual Cues

Lay judgments of environmental risks are central to both immediate decisions (e.g., taking shelter from a storm) and long-term ones (e.g., building in locations subject to storm surges). Using methods from mathematical psychology, we provide a general approach to studying lay perceptions of environmental risks. As a rst application of these methods, we investigate a setting where lay decisions have not taken full advantage of advances in natural science understanding: tornado forecasts in the US and Canada. Due to the challenges of tornado forecasting, members of the public must often evaluate the risks on their own. We study participants' ( N = 400) perceptions of cloud formations, the most prominent physical environmental cue indicating tornado threat. We use signal detection theory to analyze how well people can distinguish tornadic from non-tornadic clouds, and multidimensional scaling to determine how people make these judgments. We nd that participants have good heuristics, but with predictable biases, leading to misidentifying the tornado risk of certain cloud types. Participants viewed clouds with so-called upper- and mid-level tornadic features as less tornadic than they are, focusing on the darkness of the weather scene and the ease of discerning its features to make their judgments, with darker and harder-to-discern cloud formations rated as more tornadic. We recommend training (e.g., for new storm spotters) and communications (e.g., tornado warnings) account for the conditions leading to systematic misclassi cations of tornadicity, which could result in both better detection and more e ective risk communications.


Eric Forgoston, Montclair State University

Title: Collaborative Tracking of Geophysical Fluid Dynamics

There has been a steady increase in the deployment of autonomous underwater and surface vehicles for applications such as ocean monitoring, tracking of marine processes, and forecasting contaminant transport. The underwater environment poses unique challenges since robots must operate in a communication and localization-limited environment where their dynamics are tightly coupled with the environmental dynamics. This work presents current efforts in understanding the impact of geophysical uid dynamics on underwater vehicle control and autonomy. The focus of the talk is on the use of collaborative vehicles to track Lagrangian Coherent Structures (LCS). A control strategy is formulated that utilizes knowledge of the LCS and enables mobile sensors to autonomously maintain a desired distribution in the environment.

Research supported by the National Science Foundation and the Office of Naval Research.


Andres D. Gonzalez, Graduate Student, Dept. of Civil & Environmental Engineering, Rice University

Title: Resilience Optimization as an Interdependent Network Design Problem

The Interdependent Network Design Problem (INDP) is associated with finding the least-cost recovery strategy of a partially damaged system of interdependent networks. To solve the INDP, a Mixed Integer Programming (MIP) formulation has been successfully used to model budget, resources, and operational constraints, while also taking into account physical and geographical interdependencies. Even though the INDP MIP model has been developed and recently used to study and optimize the recovery of a system of systems, it can also provide important information to reduce its vulnerability. This work explores diverse applications of the INDP beyond defining the recovery process of an already damaged system of interdependent infrastructure, describing how it can also be successfully used to extract information of the system and its components prior the occurrence of an event. Given a particular hazard, we use the INDP to determine the criticality of components in the system, as a function of their failure and recovery rates, to optimize resource allocation and retrofitting processes. By doing so, the INDP MIP model shows to be an effective tool not only to optimizing the reconstruction of a system, but also to improving its overall resilience. To exemplify the proposed methodology, we study the system of interdependent utilities in Shelby County, TN, composed by gas, water, power, and communications networks at the transmission level, all under earthquake hazards. Results show that the INDP-based methodologies provide unique process-based insights to enhance the resilience of a system of interdependent networks, both from prevent and post-event analyses.


Xiaozheng He, Purdue University

Title: Pre-disaster Investment Decisions for Strengthening the Chinese Railway System to Earthquakes

We propose an analytical model for a pre-disaster investment problem that seeks to select and prioritize links in China Railway Network (CRN) to strengthen under a limited budg et with the objective of minimizing the post-earthquake loss of the CRN, which is represented in terms of the functionality deterioration of railway system service. A heuristic is used to solve the problem. Numerical experiments using real-world data illustrate the tractability and effectiveness of the proposed method.


Emily Heath, Rensselaer Polytechnic Institute

Title: Applying Ranking and Selection Procedures to Long-Term Mitigation for Improved Network Restoration

In this paper we consider methods to determine the best single arc mitigation plan for improving rapid recovery of a network with a given level of statistical certainty. This problem is motivated by infrastructure managers interested in increasing the resilience of their systems through costly long-term mitigation proce dures. Our problem is two stage, where we consider a small number of prevent decisions for mitigation, with a large second-stage integer programming problem to capture the restoration process for each damage scenario and each mitigation plan. We consider a ranking and selection (R&S) procedure and compare its performance against a brute force method using standard statistical testing on problems with low, medium, and high damage levels. These comparisons are made by using the same number of integer progra ms for each method, and comparing the level of confidence achieved to determi ne a best single arc mitigation plan. We find that R&S procedures perform as well or better than brute force procedures in all cases, and significantly outperform the brute force procedure in almost all cases (five out of six). Having developed a general framework for determining the best single arc mitigation plan for any network, we conclude with thoughts and challenges on how this framework can be expanded and applied to different problems.


Timothy Lant, Marquette University

Title: The Role of Modeling in Science Preparedness for Decision-making during Pandemics, Disasters, and Extreme Events

Recent events including the 2014 Ebola outbreak in West Africa have tested the United States' and the World's ability to rapidly assemble and deploy scientific resources to support decision-making at the federal and international level. Computational and Mathematical modeling are now expected to be part of public health decision-making and at the forefront of public dialogue and media reporting. Disasters pose unique challenges and opportunities to the scientific knowledge production enterprise to provide timely, critical information to health officials and decision-makers to forecast, mitigate, and recover. The ability of scientists and health officials to work together has now become part of the national infrastructure for preparedness. Dr. Lant will discuss how these dynamics support decision-making for medical countermeasures and medical response, and what might be done to improve our overall science preparedness in the future.


Garrett Nieddu, Montclair State University

Title: Analysis and Control of Pre-extinction Dynamics in Stochastic Populations

We consider a stochastic population model where the intrinsic noise causes random switching between metastable states before the population goes extinct. Switching and extinction times are found using a master equation approach and a WKB approximation. In addition, a probabilistic argument is used to understand the pre-extinction cycling dynamics. We also implement a control method to decrease the mean time to extinction. Analytical results agree well with numerical Monte Carlo simulations.


Christa Pettie, University of Maryland-College Park

Title: Information Diffusion: A Study of Twitter During Large Scale Events

The diffusion of information through a population about a natural disaster or other emergency affects how and when the public reacts to the situation, including evacuation and the demand for assistance. Thus, it is important to understand how and at what speed important information spreads. Social media are an important part of this diffusion and provide a convenient and effective way to measure it. The events studied in for this effort were Hurricane Sandy, Hurricane Irene, the 2012 Presidential Election, and the capture and death of Osama Bin Laden. This study used data about a social network of 15,000 Twitter users and their tweets. Information such as the time of a tweet, the user name, the tweet content, and the tweet ID was analyzed to measure the diffusion of information and track the trajectory of retweets. The spread of information was visualized and analyzed to determine how far and how fast the information spread. The results showed how different types of information spread indicated the importance of different topics to these users. The network of popular users also demonstrated how effectively a message can pass through several users. Understanding how information spreads will benefit policy makers and emergency managers who want to get the right information to the right people, so that they can respond optimally to an emergency.


Thomas C. Sharkey, Rensselaer Polytechnic Institute

Title: On the Value of Information-Sharing in Interdependent Infrastructure Restoration

We examine the new concept of restoration interdependencies that arise in interdependent infrastructure restoration (IIR) and how information-sharing can help mitigate their impact on the effectiveness of IIR efforts. Restoration interdependencies occur whenever a restoration task in one infrastructure is impacted by a restoration task, or lack thereof, in another infrastructure. These interdependencies can impact the timeline of the restoration efforts an infrastructure due to a lack of information about other infrastructures' restoration efforts. We present new scheduling models to capture restoration interdependencies and the role of decentralized decision-making in IIR efforts. Computational testing, on damage instances motivated by Hurricane Sandy, indicate that information-sharing between infrastructures can greatly reduce the "price" of decentralized decision-making in IIR efforts.


Klementyna Szwaykowska and Ira B. Schwartz, U.S. Naval Research Laboratory

Title: Motions Patterns for Cooperative Sensing

Swarm and modular robotics are an emerging area in control of autonomous systems. Groups of small, inexpensive robotic agents can be networked together, creating aggregates that are able to achieve tasks beyond the capability of any individual agent. For example, aggregates of locally interacting agents have been proposed as a means to create scalable sensor arrays for surveillance and exploration; distributed sensing; cooperative construction; and the formation of reconfigurable modal systems.

We are motivated by the idea of using a swarm of interacting autonomous aerial robots to map a disaster area, locate survivors, and/or identify the source of a spill. The swarm should be easily deployed and controlled by an human operator, and should perform its task with minimal human supervision. In this scenario, the operator uses a high-level control to guide the swarm as a whole (e.g., to set the monitoring region), while individual agent trajectories are governed by swarm interactions. The idea is similar to the reduced-order swarm control described in and references therein. We rigorously characterize the swarm motion patterns under a simple but general swarming model, as a function of model parameters, and demonstrate control of the overall swarm behavior by tuning a small number of control parameters.

We model the robotic agents as point particles with forces corresponding to self-propulsion and inter-agent attraction. This choice can be justified in the case of swarming robots with very fast relaxation times, such as quadcopters, which can be treated as holonomic vehicles over the spatio-temporal scales required for a mapping/search mission. Our model incorporates two key components that must be considered for real-world applicability: time delay and restricted communication bandwidth.

Systems of interacting individual agents, whether natural and engineered, involve some degree of communication delay. Time delay can have significant impact on system dynamics, and can lead to instability in swarm systems. As shown in our earlier work with globally delay-coupled swarms of homogeneous and heterogeneous agents, communication delay can cause emergence of new collective motion patterns and, in the presence of noise, lead to switching between bistable patterns. At the same time, bandwidth restrictions can limit the number of neighbors that each agent in the swarm can communicate with.

We use mean-field dynamics to analytically predict transitions between regimes of different collective swarm motions as a function of model parameters for swarms consisting of heterogeneous delay-coupled agents with a random communication network. We demonstrate use of tuning parameters to achieve desired swarming behavior for cooperative sensing and verify our results through numerical simulation and lab experiments.


Heimir Thorisson, University of Virginia

Title: Risk Mitigating Resource Allocation for Waterway Infrastructure Systems

The U.S. Army Corps of Engineers (USACE) is responsible for developing and maintaining much of the U.S. public water resources. Natural disasters, especially Hurricane Katrina which in 2005 resulted in catastrophic levee failures, have encouraged the USACE to move towards a risk centric approach to resource management, balancing risks to economic, environmental and social objectives. Multicriteria decision making, incorporated with value-focused thinking and project portfolio management has been used as the basis for developing a portfolio analytic tool for the agency, known as Asset Management Portfolio Analytics (AMPA). AMPA aims to aid decision makers by providing a rational and transparent evaluation of investment alternatives.

AMPA employs a multiattribute value model to provide decision support for allocating funds for USACE work packages. Stakeholder elicitation was used to identify attributes as they relate to the agency's mission and specific business line objectives. A case study is performed with a focus on three business lines: Hydropower, Navigation and Flood Risk Management. Considering the attributes as categories of consequences, the value function quantifies consequence of the failure. The consequence of failure, as well as the probability of failure, is evaluated separately depending on whether a proposed work package receives funding, thus enabling comparison and ensuring that funded work packages reduce adverse consequences. After filtering out infeasible alternatives, work packages are selected to provide the most value at a given budget level.

In a process heavily dependent on stakeholder elicitation and assessment, various types of uncertainty can be introduced. These sources are identified and appropriate mitigation measures recommended. Lessons learned from th e effort include how water resource management can benefit from risk - informed decision making to meet multiple diverse objectives subject to evolution over time. The method can be extended to other types of infrastructure assets exposed to natural hazards, such as in transportation, electrical systems, and telecommunication.


Previous: Program
Workshop Index
DIMACS Homepage
Contacting the Center
Document last modified on May 11, 2015.