The objective of this research is to model and analyze partnership creation, length, and conclusion in networks of agencies responding to extreme events. The researchers develop models to explore how characteristics of partnerships could be used to predict dynamics in agency investment, commitment length, partnership selection, and exit timing. The research collects and validates data by interviewing agencies active in responding to extreme events. This project compares agency behavior in two separate disaster response scenarios, mathematically models the life-cycle of agency partnerships during disaster operations, and conducts controlled experiments to analyze agency decision making.

Extreme events have had a significant impact on the world over the last several years, including earthquakes, tsunamis, hurricanes, and tropical storms. The scale and scope of events like Hurricane Sandy, the earthquake and tsunami in Japan, and the tornados around the United States make it imperative to increase our understanding of how government, non-governmental, and business agencies interact with one another in the aftermath of such extreme events. This study builds on previous work in emergency management to provide an analysis of the partnership selection and resource sharing processes that occur following an extreme event. By documenting the dynamical change in roles and flow of resources following extreme events, this work tests hypotheses regarding agencies and how they are impacted by partnerships, goals, roles, and prior involvement. With the support of a host of agencies that are actively involved in relief operations in the U.S. and around the world, the model results are checked for relevance, accuracy, and correct representation of the agencies and individuals responding to extreme events.

Project Report

The goal of this project was to model and analyze partnership creation, length, and conclusion in networks of agencies responding to disasters. By documenting the dynamical change in roles and flow of resources following extreme events, this project was designed to provide a testing ground for how agency decisions are impacted by partnerships, goals, roles, and prior involvement. The work presented here explores several different approaches to the problems faced by agencies involved in disaster relief operations. While many of the techniques presented could be used to develop better plans and mitigate the impact of future disasters, the initial motivation was to develop a set of models and tools that could be used in real-time to understand how to respond well to a current disaster situation. Thus, each of the tools developed in this project was built to support heuristics for real-time application, or to be incorporated into agency planning and preparation to improve response to future disasters. During 2013, the research team has traveled to multiple regions impacted by disasters throughout the United States to conduct interviews with organizations involved in the relief effort, and also to continue to develop a network of research partners to conduct model testing and evaluation. Agencies were interviewed through random sampling (i.e., participants identified through interactions in the field or web searches) and through snowball sampling (i.e., building a network of new participants based on the recommendation of previous participants). We interviewed over 75 organizations involved in emergency management operations around the United States, collecting data and information about the different types of partnerships that have occurred, the types of relief activities that agencies have participated in, and the length of the relief operation. This group of participants was interviewed in relation to two major disasters in the United States: the tornado in Joplin, Missouri, and Hurricane Sandy. We have also developed a set of experiments for students to participate in to collect data about decision making and resource allocation. In each experiment participants were only asked to make a single decision (e.g., responding to a request, ending a partnership). In each experiment there were several variables which made the trial unique. The upper and lower bounds for each variable was derived from the results of the interviews or simplified to provide a clear decision-making environment for participants. Each experiment provided a unique decision environment for the research participants. In this section we review each of the experiments and discuss the design and goals with supporting visuals. The experiments were built in Excel using Visual Basic for Applications®. Each trial of the six experiments had randomly generated parameters, and a logic component for how other agencies in the disaster environment acted. Many of the most significant results during this study were found during the development, or from the output, of a simulation framework called the Disaster Relief Agent-based network Management and Adaptation System (DRAMAS) for exploring the life-cycle of an agency partnership responding to disasters, and was developed to support hypothesis selection/testing regarding agency behavior. DRAMAS is an agent-based simulation that serves as a descriptive model of agency behavior in a relief environment. DRAMAS provides a quantitative approach to understanding disaster relief, which can help accelerate future studies in longitudinal agency behavior after a disaster happens. The DRAMAS model was built in AnyLogic®, a simulation program written in Java, and built on top of Eclipse to manage and compile the code. The visual representation of the network dynamics provides an interface for decision makers to interact with the changing disaster environment and observe the long-term consequence of different decisions. DRAMAS was validated by comparing the model output with the results of interviews with relief agencies, as well as current literature on humanitarian aid and disaster relief. Finally, the models were calibrated using data from experimental work with students at the University at Buffalo. The results from the DRAMAS agent-based model were disseminated to all of the communities that participated in the study or demonstrated interest in the results via the press release documents. Additionally, the detailed results and additional research avenues are detailed in the full dissertation and review of the product. Finally, many of the mechanics and results from DRAMAS and the study are currently under review for publication or in preparation for submission to peer-reviewed journals.

Agency
National Science Foundation (NSF)
Institute
Division of Social and Economic Sciences (SES)
Type
Standard Grant (Standard)
Application #
1261058
Program Officer
Robert O'Connor
Project Start
Project End
Budget Start
2013-02-15
Budget End
2015-01-31
Support Year
Fiscal Year
2012
Total Cost
$12,500
Indirect Cost
Name
Suny at Buffalo
Department
Type
DUNS #
City
Buffalo
State
NY
Country
United States
Zip Code
14228