Saturday, 5 August 2017

In Europe we're all going to die in disasters - or are we?

The top news on the BBC website this morning was that "deaths in Europe from extreme weather events could increase 50-fold by 2100". In my opinion, there are two lessons to be drawn from this.

The first is that the authors of the study (Forzieri et al. 2017) were very clever to release it at the time of maximum impact. As I write, the temperature outside my room is in the 40s Centigrade. The article was embargoed until 11.30 last night and pre-distributed to the mass media. Small wonder that today it got maximum exposure.

The second is that the research is pretty much worthless. It is misleading and highly unlikely to offer an accurate forecast. It is a hazards-driven study that effectively uses exposure as a surrogate for vulnerability, about which the authors have remarkably little to say (see my comments in Davis 2017). And yet it has been demonstrated all over the world that vulnerability defines death tolls - i.e., people can live in highly hazardous zones and not die if they are not vulnerable (Wisner 1993). Various African countries, India and Bangladesh have all had some notable successes in reducing disaster mortality in areas of high population growth (e.g. Paul et al. 2010). Moreover, one of the effects of the International Decade for Natural Disaster Reduction was to hold the line on death tolls (it would have been nicer if they had gone down, but anyway, it was an achievement of sorts).

By way of illustration, the current heat wave is probably going to be comparable to that of 2003, during which it is estimated that there were 70,000 excess and premature deaths (Lagadec 2004). The figure is highly contentious, but, leaving that aside, since then measures have been put in place to avoid a repetition (Boyson et al. 2014, Pascal et al. 2012). These are mainly early warning systems to detect and assist vulnerable people. In Tuscany, where I am writing this, they have been highly effective, and I believe they have in France and Spain, too.  In the United States, as population rose, heat-related mortality declined (Sheridan et al. 2009). In contrast, Forzieri et al. (2017, p. e206) forecast that heatwave deaths in southern Europe will go up by 7,000 per cent in a century. If that were so, perhaps our work in disaster risk reduction would be a waste of time.

People put faith in figures because they seem precise and scientific, even when the reasoning that supports the figures is a hollow shell. The good side of the article is that it draws attention to the problem - or to part of it (and what a pity it does not draw enough attention to the extreme dynamism of vulnerability!). The bad side is that policy may end up being based on projections that are largely fantasy. There may indeed be massive increases in mortality in weather disasters in Europe, but that would be a function of many other factors - whether there is conflict, the impact of cascades, the functionality of antibiotics, emerging threats and hazards, dependency on critical infrastructure, the status of emergency preparedness, exotic diseases, the wealth differential, etc...


Boyson, C., S. Taylor and L. Page 2014. The National Heatwave Plan: a brief evaluation of issues for frontline health staff. PLoS Currents Disasters 13 January 2014.

Davis, N. 2017. Extreme weather deaths in Europe 'could increase 50-fold by next century'. The Guardian 5 August 2017.

Forzieri, G., A. Cescatti, F. Batista e Silva and L. Feyen 2017. Increasing risk over time of weather-related hazards to the European population: a data-driven prognostic study. Lancet Planetary Health.

Lagadec, P. 2004. Understanding the French 2003 heat wave experience: beyond the heat, a multi-layered challenge. Journal of Contingencies and Crisis Management 12(4): 160-169.

Pascal, M.,  K. Laaidi, V. Wagner, A.B. Ung, S. Smaili, A. Fouillet. C. Caserio-Schönemann and P. Beaudeau 2012. How to use near real-time health indicators to support decision-making during a heatwave: the example of the French heatwave warning system. PLoS Currents Disasters 16 July 2012.

Paul, B.K., H. Rashid, M.S. Islam and L.M. Hunt 2010. Cyclone evacuation in Bangladesh: tropical cyclones Gorky (1991) vs. Sidr (2007). Environmental Hazards 9(1): 89-101.

Sheridan, S.C., A.J. Kalkstein and L.S. Kalkstein 2009. Trends in heat-related mortality in the United States, 1975-2004. Natural Hazards 50(1): 145-160.

Wisner, B. 1993. Disaster vulnerability: scale, power and daily life. GeoJournal 30(2): 127-140.

Tuesday, 1 August 2017

Seven Rules for the Application of Operations Research to Disaster Management

It is currently very fashionable to apply the methodologies of operations research to disaster mitigation, management and response. Is this a fashion or a fad? Will the algorithms be used and appreciated, or are they merely wasted effort? Do the algorithm makers understand what conditions are like in a disaster, and what the real needs of managers and responders are?

In disaster management there is a well-founded hostility towards over-sophisticated routines and equipment. Managing emergencies will always be a rough-and-ready process, in which most of what is done is a kind of approximation. Such is the nature of uncertainty and rapid change in the field that it could never be otherwise.

If operations research is to make a useful contribution to disaster management, it will have to take account of these principles:-

1.    In emergencies, 'optimisation' is a very relative term. Pre-planned activities require considerable scenario modelling in order to take account of the real needs that will be generated during a future emergency.

2.    Optimisation based on an assessment of pre-disaster conditions is unlikely to be relevant to the post-disaster situation. Infrastructure will be damaged, inefficient and probably partly non-functional.

3.    Optimisation that assumes perfect knowledge of the situation is bound to fail. During major emergencies, the common operating picture is constructed slowly and with difficulty. One cannot optimise a situation that is partially unknown.

4.    Algorithms that are designed to be used in emergency situations should be capable of deployment during emergencies. This means that at the height of a crisis time cannot be expended on collecting data or running lengthy analyses.

5.    To make an algorithm credible, evidence should be provided that it is acceptable to field commanders, who would use it or act upon the results that it provides. Optimisation is not an objective held by most emergency managers and field commanders. An algorithm that does not take account of their needs and ways of thinking is highly unlikely to be appreciated or utilised by them.

6.    Decision support systems are welcomed if they really do support decision making. No sensible emergency manager would put blind faith in an algorithm unless the results clearly demonstrate that it works and visibly improves the situation.

7.    Flexibility is an essential ingredient of any algorithm. In disasters, conditions on the ground can change abruptly and without warning. Algorithm makers need to understand the difference between 'agent-generated demands' and 'response-generated demands', as described in the classical literature on the sociology of disasters.