The razor in the waterfall: using longitudinal data to sharpen the analysis of cascading disaster risk

Abstract

Cascading disasters progress from a triggering disaster event to a diverse range of consequent disasters. Disasters following the Great East Japan Earthquake of 2011 highlight how these cascades also progress to multiple geographical locations. However, the very low frequency of these events means their analysis has usually excluded base-rate data. This common practice risks overestimating future likelihoods. A simplified approach to base-rates for less catastrophic cascades, following the rule of Occam’s razor, may help develop more accurate predictions of future likelihoods. The current research hypothesized that an intuitively relevant (0.05) probability of cascading flood-related disasters could be derived from a large, generic register of disaster events. A threshold-based analysis of transitions between phases of a hydrologic flood-related cascade was performed using ten years of data from the USA state of Florida. This analysis identified a 0.05 probability of flood-related cascading disasters. The same analytical methods were reliable when applied to subsequent data, from the year 2000. Similar approaches to information extraction and probability analysis can be applied to climatic data collected at more regular intervals. This will improve the usefulness of analytical results, which can then be added to expert analyses of more contemporary events and scenarios.

Infrastructural Aspects of Rain-Related Cascading Disasters: A Systematic Literature Review

“…Where permitted by an expert rating protocol, experts could be prompted to consider both sets of values. This would help mitigate a perceptual bias called the base-rate fallacy , where individuals tend to inflate the likelihood of recent disaster linkages, by ensuring that each expert considers how relatively infrequently those linkages occur [ 12 ].…”

“…Huggins et al [ 12 ] highlighted the potential for using localized, longitudinal data to study transitions from one disaster state to another. However, large and well-structured sets of relevant data are often not available for analysis.…”

“…However, large and well-structured sets of relevant data are often not available for analysis. Kar-Purkayastha, Clarke, and Murray [ 13 ], and Huggins et al [ 12 ] have outlined how open-access disaster impact databases typically lack important chronological, geographic, and other details. Associated challenges can be worsened by government agencies who are reluctant to allow researchers to access more detailed disaster impact data at a national scale [ 14 ].…”