SELF ORGANIZATION IN DISASTER RESPONSE:
THE GREAT HANSHIN, JAPAN EARTHQUAKE OF JANUARY 17, 1995
1996
Return to Hazards Center Home Page
Return to Quick Response Paper Index
This material is based upon work supported by the National Science Foundation under Grant No. CMS-9632458. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.
I acknowledge, with warm thanks and sincere appreciation, the assistance and support of Japanese and American colleagues who have contributed to the conduct of this study. There are too many to name, but several deserve special mention. In Japan, I received gracious and unfailing support from Shunji Fujii, Taisei Engineering Corporation, Tokyo; Kunio Funahashi, Department of Architectural Engineering, Osaka University, Osaka; Muneo Ohta, M.D., Director, Senri Critical Care Medical Center, Osaka; Tatsuro Kai, M.D., Deputy Director, Senri Critical Care Medical Center, Osaka; Kentaro Serita, Dean, Graduate School of International Cooperation Studies, Kobe University; Juntaro Ashikari, Nishinomiya; and Laurie Lofgren, Northwest Airlines, Narita. In the U.S., I benefited greatly from the generous assistance of Masakane Kawaii, AICP, San Francisco; Takehiko Serai, Takako Lowenstein, Yoshiko Koda, Elizabeth Robedeau, Tamara Taylor, Susan Wade, and Ernesto Pretto, M.D., all of the University of Pittsburgh. I also thank my colleagues, Sharlene Adamson, M.D., Center for Emergency Medicine, and Nancy Bowen, Graduate School of Public Health, University of Pittsburgh, for their good will, effort, and perseverance in the conduct of field research.
This research was supported in part by a National Science Foundation Quick Response Grant administered through the Natural Hazards Center, University of Colorado, Boulder. Additional funds were received from the University Center for International Studies, the Office of Research, and the Graduate School of Public and International Affairs, all at the University of Pittsburgh. The Institute of Governmental Studies, University of California, Berkeley, provided a supportive scholarly environment for research and writing during the summer months, 1995.
Building a community capacity for rapid transition in event of disaster is fundamental to effective disaster response. In every disaster environment, evidence exists of spontaneous reallocation of resources and reorganization of action to meet urgent human needs. However, such systems vary significantly in timing, efficiency, effectiveness, and reliability in practice, with associated costs in lives and property. This study explores the characteristics of self organization in order to characterize more accurately the dynamics of this process in a complex, interdependent, densely populated, metropolitan environment.
The conditions of self organization require a sociotechnical system that combines sufficient structure to facilitate exchange of information among its multiple components with sufficient flexibility to adapt action to the demands of a dynamic environment. Self organization depends upon ready access to timely, accurate information through an information infrastructure that supports systematic monitoring of critical conditions, feedback to responsible participants, and revision of actions taken in the light of new information.
Inquiry into actual processes of self organization in disaster environments requires appropriate measurement. While we can devise theoretic measures of assessment, most require sophisticated computer technology that has not yet been incorporated into practicing disaster operations centers. Building upon the concepts outlined by Stuart Kauffman (1993: 174, 208-227) in his discussion of the interaction among actors in a complex, dynamic system, I outline a preliminary methodology that reveals the major characteristics of rapidly evolving disaster response systems and their inherent processes of self organization.
The study was initially designed to address four basic questions regarding emergency medical services following the Hanshin Earthquake of January 17, 1995. They were:
This study focuses on the emergence of a disaster response system in a technically advanced, densely populated, metropolitan region. The size, shape, and timing with which the system evolved are critical to defining not only the losses endured in the given event, but also the scope of interorganizational action required for reconstruction and recovery. Specifically, I examine the process of self organization in the Hanshin-Awaji, Japan, earthquake on January 17, 1995. In societies vulnerable to seismic risk, severe earthquakes represent the civilian equivalent of war, a massive shock to the entire technical, organizational, economic, political, and social system. How a complex, sociotechnical system responds to this shock provides valuable insight into its likely evolution to the next phase in its performance, and its likely actions to prevent recurrence.
This study is significant for at least four reasons. First, it explores means of facilitating the emergence of self organization in disaster environments in order to reduce losses in lives and property in disaster-afflicted communities. Second, it presents a preliminary methodology for assessing rapidly evolving response systems following disaster. Third, it offers suggestions for continuing research on complex, adaptive systems (CAS) in disaster environments. Finally, the findings contribute to our theoretical understanding of the dynamics of CAS and institutional change.
The capacity to make 'transitions' between different states, an essential feature of complex behavior (Prigogine and Nikolis, 1987: 36), distinguishes nonlinear from linear social systems. With linear models of policy analysis, econometrics, or trend analysis, we are able to chart the performance of established systems within prescribed parameters over time (McKibbin and Sachs, 1991). However, these models do not allow us to anticipate future states in dynamic systems or to predict with any degree of certainty what outcomes would follow from which alternative courses of action. Linear models assume that existing conditions are likely to remain stable over time, and that conditions operating in the future will function very much as they did in the past. In the rapidly changing, complex environments of disaster, such models are often invalid or misleading.
In nonlinear systems, differences in initial conditions precipitate variations in performance that increase markedly in processes iterated over time (Prigogine and Stengers, 1984; Ruelle, 1991; Prigogine and Nikolis, 1987). Nonlinear systems also exhibit primary characteristics of stochasticity and irreversibility in time (Gell-Mann, 1994). That is, random events set in motion sequences of reasoning and action that differ from previous behavior in the system, generate different dynamics of selection and evolution in performance (Kauffman, 1993), and create different memories and interpretations of that experience. Disaster events, therefore, produce unique combinations of choices, actions, and reasoning that could not be predicted. Once generated and instantiated in experience and practice, however, these reasoning and action processes cannot be reversed. We need a set of nonlinear measures and supporting concepts to capture this dynamic exchange of information, attention, and action both within the system and between the evolving system and its environment.
Borrowing from biology, the concept of an N-K system (Kauffman, 1993: 175-209) permits identification of basic characteristics of self organizing systems that reallocate their energy and action to serve changing system needs. In Kauffman's original model, N equals the number of actors in the system, K equals the number of interactions among these actors, and P equals the 'bias for choice' among the actors, or the goal of the system that drives action. These three measures allow the identification of a fourth measure - the boundaries of the system - operating in response to specific events, times, conditions, and locations in the wider environment. Defining the boundaries helps to identify the relationships between types of complex systems and especially to distinguish subsystems within larger systems.
The measure, D, represents the duration of the interactions among actors in the system, acknowledging that some interactions may be intense but brief, while others may continue at lower levels of effort over longer periods of time. A final measure, T, denotes the types of transactions that are carried out by organizations operating as participants in the system. Other characteristics regarding the sources of support and conditions of the environment may be identified and mapped, but this set of measures provides an initial assessment of the operating characteristics of a complex, adaptive system.
This set of characteristics may be summarized as follows (Comfort, 1994: 306-307):
Until such technology is fully incorporated into disaster management operations, cruder measures may be used to distinguish evolving systems in a complex society. We can identify sets of interacting organizations, which, in turn, contain subsystems that perform separate functions in related contexts. Such measures document patterns of interaction within and between interdependent organizational systems in a first step toward charting the dynamics of transition in disaster response. They also identify gaps in the exchange of information among these interdependent systems that inhibit acceptance of alternative strategies, reinforcing static behavior and resistance to change. These methods use standard data collection methods as interim procedures, but follow the logic of nonlinear reasoning.
I used the N-K model to guide data collection and analysis in the field study of the disaster response system that evolved following the 1995 Hanshin, Japan, earthquake. Computerized information systems were not fully in place to provide measures of exchange of information among participating organizations in this case of disaster operations, although partial records of some activity were available. I present data from the field research, following the logic of the N-K model and noting gaps and inadequacies in this preliminary application.
Three types of data were used in the conduct of this field study: 1) semistructured interviews with responsible management personnel at the municipal, prefectural, and national levels; 2) operations records and reports prepared by the participating agencies as well as professional reports prepared by external organizations; and 3) content analyses of newspaper reports of disaster response operations in local and national newspapers following the earthquake. Interviews were conducted in the field with the assistance of local translators. A detailed description of the sample design and questionnaire used for the survey of practicing managers and procedures for conducting the organizational analyses are available from the author.
Organizationally, the area was not well prepared for seismic risk. Although the islands of Japan are located at the juncture of three tectonic plates and seismic risk is well known in the nation, residents had generally believed the Hanshin region, which last experienced a moderate earthquake (6.1 Richter scale) in 1916, to be relatively stable in contrast to the Tokyo Region, which had suffered a major earthquake with heavy losses in 1923. Relatively little investment had been made in earthquake preparedness, either by public organizations or residents. While cities in the region had emergency plans, their preparation had been primarily oriented toward small, local disasters of fires and floods.
Private utility companies, such as Kansai Electric Co. and Osaka Gas Co., demonstrated substantial investment in seismic mitigation efforts to protect their interests, but were not directly linked to the public agencies. Socially, there existed little tradition of voluntary organizations or community self help associations. Most people focused their lives on their work associations and their families. Although the initial technical systems were strong, there was little interorganizational capacity to reallocate resources and action in timely response when these interdependent systems failed under the severe shock of the unanticipated earthquake.
The damage was extensive. The National Land Agency reported 5,502 dead (1), as of April 23, 1995, and an additional 800 persons died of earthquake-related consequences over the next several months, for a total death toll of 6,302 from the event. The total number of injured reported in the April 1995 statistics reached 41,648. Of that number, 1,961 were seriously wounded. A total of 390,692 houses were reported destroyed, half-destroyed, or partially destroyed, while 3,669 buildings, public and private, were destroyed or damaged.
The dynamics of the destruction were sobering. The strong vertical ground motion ruptured underground gas and water mains, causing leaks and disrupting service throughout the region. The earthquake damaged an estimated 4,500 km. of gas lines and left 1,200,000 houses without water. Electrical facilities were also damaged, cutting off sources of electrical power to 850,000 city departments, businesses, and households. The total cost of the disaster is estimated at US$200 billion.
As the gas mains ruptured, fires broke out. With no water available for fire suppression, the fires raged largely unchecked through seriously damaged sections of the city. In Kobe, 60 fires broke out before 6:00 a.m. on January 17, 1995, and burned simultaneously. Before 9:00 a.m., the number of fires burning simultaneously had increased to 85, with a total of 109 fires reported for the city of Kobe, and a total of 294 fires for the entire earthquake-affected area. Debris from collapsed buildings blocked the streets, preventing fire trucks from getting through. Over 9,403 blockages in roads were reported for the area.
Under these conditions, communications capability was critical. The Kobe Fire Department had just installed an advanced computerized dispatch system with video monitors in December 1994. But on January 17, 1995, it was not yet operational and was not used during the disaster operations. Telephone lines were out of order during the first day in a large areas of the region, while others were overloaded. Over 1800 emergency calls made on 118 emergency circuits were recorded on the 119 dispatch logs on January 17, 1995, at roughly 100 calls per hour or 1.7 calls per minute (2). These, however, were only the calls that could get through. The number of calls attempted, but not completed, cannot be estimated. Fire departments had their own radio systems, but could not communicate with other departments. Communications capability proved very limited in the first critical hours following the earthquake.
These conditions proved overwhelming for the Kobe Fire Department, which had primary responsibility for emergency response, with a total of 11 fire stations in the city, 176 engines, and 305 personnel on duty when the earthquake occurred. Three of the 11 stations were damaged in the earthquake, and even with emergency call-out procedures, only 663 personnel were able to report for duty within the first two hours (3). The actual destruction was beyond any training scenario for municipal emergency response.
For example, the fire department operates on a "doctor-car" system for emergency medical services. When a citizen calls the emergency number, 119, the fire department responds by dispatching a team of paramedics that can administer first aid. If an injury is involved, the fire department also calls the local hospital, which dispatches a physician trained in emergency medicine. The firemen, even though they may be "first on scene," are legally prevented from administering any emergency medical procedure unless it is supervised by a licensed physician. While this procedure works well for traffic accidents or small fires, it hinders delivery of emergency medical services in large-scale disasters (4). While the toll in lives of this practice cannot be calculated by any exact measure, informed paramedics and physicians were sobered by its consequences and concur that it must be changed (5).
In this urgent context, the operations logs of the Kobe Fire Department document actions taken during the first critical hours following the earthquake. These records correspond with reports from governmental agencies at the prefectural and national levels (6). (A record of operations for the three levels of government during thefirst hours of response on January 17, 1995, is available from the author.)
The record reveals the gaps in information that seriously affect the coordination within and between jurisdictions. For more than four hours, neither the Hyogo Prefectural Government nor the National Fire Defense Agency in Tokyo had a clear picture of the degree of destruction and damage in Kobe. Constrained from action by the existing law until a request for assistance had been received from the City of Kobe, these governmental agencies did not enter the response system until more than four hours after the initial shock. In further irony, existing law kept the dogs from the French Search and Rescue team in quarantine, not allowing them to enter response operations until the fourth day after the earthquake when they extricated dead bodies, instead of living people (7). These conditions indicate the perverse effects of law, intended to protect citizens, in restricting the capacity of public managers to carry out their mission under dynamic conditions (Smart and Vertinsky, 1977).
Especially damaging was the lack of communication between the public agencies and the private utility companies. Osaka Gas Company, a private company and owner of the ruptured gas lines that were fueling the fires, was apparently not in communication with the city departments. Only after reviewing the data from their own seismic monitoring devices at approximately 11:00 a.m. on January 17, 1995, did the company's Policy Group decide to close off five of the fifty-five blocks of their 50,000 km. distribution system. The company began to shut off the gas in critically damaged areas at 11:30 a.m., and completed the task about 11:50 a.m., six hours after the main shock occurred (8). The gap in information and communication regarding emergency medical services was even more severe. For the first 12 hours following the earthquake on January 17, 1995, the Emergency Medical Division, Ministry of Health in Tokyo had little to no information regarding the conditions and need for medical services in the disaster area. On January 23, 1995, six days after the earthquake, an Emergency Operations Center was established in Kobe with the assistance of the national Ministry of Health (9).
As the immediate shock of the earthquake gave way to action and as information about the damage became clearer, a disaster response system began to evolve across organizational and jurisdictional lines that eventually engaged assistance from all of Japan's 47 prefectures, 44 nations and regional organizations, and the United Nations. Figure 1 presents a profile of this response system, ordered by first mention, funding source, and day for the first five days of disaster response, January 18 - 22, 1995. First mention indicates the organization's entry into the disaster response system as reported in the Japan Times. On the first three days following the earthquake, public organizations received the largest proportion of first mentions, with a small number of first mentions for private organizations and no mention of nonprofit organizations until Saturday, January 21, 1995, four days after the earthquake. By Sunday, January 22, 1995, nonprofit organizations received the largest number of first mentions.
Figure 2 presents the breakdown for public organizations by first mention, day and jurisdictional level. Figure 3 presents the profile of the response system by type of funding source and week for the first month following the earthquake, identifying 392 organizations as participants in disaster operations from news reported in the Japan Times. Figure 4 presents the profile of organizations engaged in disaster response by frequency of mention, funding source, and day for the first five days. This profile shows the clear predominance of public organizations in reported response operations for the first four days, limited mention of private organizations during this time, and no mention of nonprofit organizations until day 4, with an increasing number of mentions for nonprofit organizations on day 5.
From these profiles, it is clear that different types of organizations entered the response process at different times and engaged in disaster and relief operations for different periods of time. Participation of both private and nonprofit organizations peaked in the third week of operations, as reported in the Japan Times. Interesting also is the active role of the yamaguchigumi, the Japanese mafia, in the organization and distribution of disaster relief, in apparent cooperation with local governmental officials. (A list of all organizations identified as part of the interorganizational response system is available from the author.) Other data show the extraordinary mobilization of a national response through fire departments, medical volunteers, and water teams from 43 of Japan's 47 prefectures. These activities all required communication and coordination, central to cooperative action.
The active N-K system, in Kaufmann's terms, is the product of interactions among the identified actors in the system. (The author has prepared and can supply a table and graph that presents the number of interactions among types of organizations reported in news articles on disaster response in the Japan Times in the first month following the earthquake, January 18 - February 17, 1995.) The data represent only one source of public documentation for interactions among organizations participating in the response system, but they do indicate some important characteristics of the response. The largest number of interactions, 72.84%, occurred among governmental organizations, with private organizations ranking a distant second place at 8.64%. Nonprofit organizations ranked third, with 6.17% of the reported actions.
The author also examined (table and graph available upon request) the number of interactions performed in disaster response operations that were reported in news articles in the Japan Times during the first month following the earthquake. These data indicate that the largest number of reported transactions, 25.0%, involved disaster relief activities, while the second largest number, 15.91%, occurred in two different areas, medical/health care and communication and coordination functions, and the third largest number, 14.77%, were activities related to financial assistance.
The response system in the Hanshin earthquake appeared overwhelmed and tenuous at first, giving way to confidence and strength as participants gained in knowledge and commitment to response operations. The manner in which the response system evolved, however, had a critical effect on performance in the search and rescue operations, and, therefore, the number of lives saved from collapsed buildings. Table 1 presents a brief profile of the number of persons rescued from collapsed buildings who lived, by day of rescue.
Table 1 Number of Live Rescues, by Day of Rescue Date Jan. 17 Jan. 18 Jan. 19 Jan. 20 Jan. 21 Total rescued 604 452 408 238 121 Total who lived 486 129 89 14 7 Percent rescued who lived 80.5 28.5 21.8 5.9 5.8 (Source: Kobe Fire Department, "The Great Hanshin-Awaji Earthquake [Kobe City Area]: Record of Fire Fighting in Kobe. 1995:12.)The figures reveal a startling drop in the proportion of persons who lived after being rescued from collapsed buildings and underscore the critical issue of timeliness in emergency response. This same pattern of steep decline by day of rescue has been confirmed in studies of other disaster response operations (Pretto et al., 1992). The cost in timeliness, measured in loss of lives, is sobering, and appears directly related to the functioning of the information infrastructure that undergirds interorganizational decision making and action. The issue of timeliness is also illustrated by the high number of people living in shelters months after the disaster (10).
Interdependent emergency response organizations were unable to make rapid transition to an emergency response system vital to saving lives in the first hours following the earthquake. Six critical factors inhibited this transition:
The business sector had invested in information technology, and performed well within its limited range, but it did not have clear, effective communication linkages with public sector agencies responsible for life and property. Public sector investments in information technology either were not fully operational, e.g. Kobe Fire Department's GIS and computerized dispatch system, or failed, e.g. Hyogo Prefecture's satellite communication system, to support decision making in disaster operations
Self organization did occur, but later and more sporadically in the response period. Volunteer groups were organized, many for the first time, to function in this disaster and entered the response period on the fourth day. The challenge is to build upon this spontaneous base of interest and experience to foster a continuing exchange of information, knowledge, and skills in the mitigation of seismic risk in Japan and other nations.
Japan's initial conditions of a strong economy and good organizational skills became evident as the system turned from response to recovery. For example, the recovery system mobilized resources from 43 out of 47 prefectures in the nation to provide water to areas in which the distribution system had been damaged. This manual distribution system operated efficiently for about 2.5 months.
Second, the critical function of aggregating units from different levels of governmental jurisdiction into a wider system of disaster response underscores both the difficulty of this task under linear models of organization and the interdependence of these units in a massive, large-scale disaster. This function and its capacity to mobilize resources - personnel, equipment, supplies, skills, and knowledge - require a mechanism of information exchange to achieve a shared, system-wide goal: protection of life and property. Timely information exchange, guided by a clear `internal model' or goal for action and prompt feedback, appears more efficient in allowing organizations to adapt to rapidly changing disaster environments than direction by centralized command and control. The result is a nonlinear, dynamic system that is able to coordinate diverse resources, materials, and personnel across previously established organizational and jurisdictional boundaries in accordance with changing needs. Such a system uses processes of self organization in which informed participants initiate action, but adjust their action to that of others operating toward the same goal to achieve a timely, efficient response (Rochlin et al, 1987). It is essentially an organizational system operating in parallel, supported by a strong, distributed, interactive, information system.
Third, the goal of the disaster response system serves as an `internal model' or `mental compass' for self organizing processes. This goal allows participants from diverse perspectives, experience, and resources to adjust their actions and contributions to that of other participants in the system. Finally, formation of an `epistemic community' (Haas, 1990) of knowledgeable people from diverse backgrounds, experience, and organizations that focuses on the shared problem is vital to the articulation of a common goal for reduction of risk and formulation of action strategies that can be communicated to a wider set of responsible actors. This step is essential to the development of `resonance' or willingness to support shared action when necessary to sustain the goal of a humane society. There are some indications that the formation of such a community is already taking place in Japan, for example, through the Pan Pacific Forum in Kobe that is meeting regularly to form policy recommendations for the reconstruction process. The group includes about 40 responsible community leaders from education, business, medical care, publishing, and voluntary organizations (11).
Return to Hazards Center Home Page