Notes on the Science of Extreme Situations, Paper No. 16

A PRELIMINARY COMPARATIVE CLASSIFICATION SCHEME FOR DISASTERS

By Walter G. Green III

Copyright 2006 by Walter G. Green III.  All rights reserved.  Permission to reproduce copies for instructional use and individual copies for personal use as needed is granted to university faculty, researchers, and students.

INTRODUCTION:

(1)      In the development of the Disaster Database Project (Green 2006), assessments of individual events suggested that there are two categories of disasters:

… events that are readily classified using existing scales (Saffir-Simpson for hurricanes, Fujita for tornadoes, a variety of scales for earthquakes, the volcanic explosivity index, etc.) (see Woo 1999 for a detailed examination of these largely quantitative scales).

… events for which there are no classification scales (aviation crashes, flash floods, highway accidents, ferry sinkings, etc.).

(2)      In addition, it is possible to compare one event for which a scale exists with another event of the same type, but there is no common means of comparing the magnitude of one specific event with another event, either contemporary or separated in time.  Although there have been efforts to develop scales that cross these boundaries (see Henry Fischer’s Disaster Scale reported by the American Sociological Association, 2004), those I am familiar with are not optimized for use with a variety of data categories.

(3)      The challenge is that evaluation of events contained in the database of the Disaster Database Project is constrained by imperfect data.  Sources may not offer sufficient information to identify quantitative entries in data fields, especially in the case of historical events.  In addition, data quality is highly variable, and data is subject to being skewed by political, economic, or symbolic considerations (for an example, see the manipulation of disaster reports in both the San Francisco earthquake and the Lake Okeechobee hurricane, Hansen and Condon 1989, Mykle 2002). 

THE DISASTER INDEX:

(4)      My goal was to develop a scale that allowed easy comparison of the magnitude of disaster events on a macro level.  Such a scale must allow comparison regardless of time, place, and category for it to be useful in the context of a large database with a wide range of dates, event types, and locations.  The resulting index is based on a selection of categories of data and the assignment of numerical values to ranges of measurements (see Table 1).  The metrics, and the values assigned to them, are currently being tested, and are subject to modification.  They are admittedly subjective, and are derived from an inspection of the data currently stored in the database.

Table 1.  Metrics for the Disaster Index

(5)      Construction of the scale elements was based on an assessment of the ranges of data in the Disaster Database.  In each case, typical reported ranges of events were identified based on observed clusters of data.  I believe the core indicators in Table 1 offer a reasonable selection of metrics that (a) recur from event to event, (b) can be assessed in a gross sense, and (c) are appropriate for assessment of magnitude.

(6)      The decision process in developing these factors was essentially qualitative.  For example, I started with what is probably the most often reported measure of event magnitude, human deaths.  Human deaths seem logically related to humans injured and to animal deaths (an indictor of agricultural impact and, in the developing world, of economic impact) – all reflect impacts on living things.  However, the temptation is to consider human deaths as a more important metric that either injuries of animal deaths.  In reality, I suggest that human injuries and deaths impose different stresses on both the response system and on the impact community.  Response to injuries demands more resources and more rapid response, and may generate more immediate and short term costs.  Fatalities, on the other hand, have the potential to generate higher psychological impacts, greater disruption of social networks, and longer term opportunity costs.

(7)      How do we score, however, the event in which there is a large count of killed, but only a small number of injured?  To take the extreme case in an incident in which 1,000,001 died (a score of 7) and 6 were injured (a score of 1), the lower score, when applied in computations, would seem to incorrectly reflect the overall severity of the event.  To compensate for this disparity, I suggest a rule that injured are only applied as an event criteria if the data results in a score equal at least to the score for fatalities minus 1.  This criteria excludes events with great disparities that would not comparatively tax the response system, and still allows for lower injury totals to influence the overall scoring in events that result in injury totals that approach, but do not reach the fatality totals. It should be noted that this reverse relationship with more fatalities than injuries appears to be unusual except in those cases in which there are very large numbers of casualties relative to the exposed population with only small numbers of survivors.  Airline crashes and mine accidents offer two examples in which this relationship appears to be more common.

(8)      There is considerable fuzziness inherent in this collection of metrics – in the trustworthiness of the data, in the relationship between the various metrics, and in the availability of data at all.  How can we then assess one disaster based on data in five of the metrics, versus another with data in all or in only one?  The solution I propose is to add the available scores and determine the mean.  The mean is then compared to a table (see Table 2) to convert it to a single digit classification number that represents a range of means.  In this process, it is important to use only reported data, and to not make assumptions about that which is not reported.

Table 2. Classification of Disaster Events by Mean Score

(9)      There are significant limitations to this method.  First, it is inherently inaccurate at some level, and that level may be difficult to identify.  For those who wish hard, quantifiable data, these scores may represent a range of uncertainty that is uncomfortable.  Second, classifications are subject to fluctuation as additional information is received or discovered.  This is not a disabling limitation, and it is not uncommon for disaster events classified by other scales to be subject to reevaluation.  Third, I am by no means certain that these metrics define the actual magnitude of disasters.  They are based on what is available, commonly reported information found in historical, secondary, or contemporaneous accounts and analysis of events.  However, there may be better measures of magnitude that are not commonly reported.

RECOMMENDATIONS FOR FURTHER WORK:

(10)    This Disaster Index is a preliminary model.  I encourage other researchers to test it, compare events based on their available data, and provide feedback as to the relative accuracy of the Index as an assessment and comparative tool.

WORKS CITED:

American Sociological Association. “Sociologist proposes disaster scale to facilitate recovery and research.” Science Blog. Location http://www.scienceblog.com/community/older/2003/C/2003985.html; 2004.

Green, Walter G., III. Disaster Database Project. Location http://learning.richmond.edu/disasters/; 2006.

Hansen, Gladys and Emmet Condon. Denial of Disaster: The Untold Story and Photographs of the San Francisco Earthquake and Fire of 1906. San Francisco, California, United States of America; Cameron and Company; 1989.

Mykle, Robert. Killer ‘cane: The Deadly Hurricane of 1928. New York, New York, United States of America; Cooper Square Press; 2002.