Disastrous Return Periods

When talking about return periods it’s easy to assume that a 1 in 100 year event will occur only once in 100 years. This is may lead to misconceptions in the understanding of risk. Consequently, it can lead to poor-decision making. Stakes could be fairly high, perhaps affecting whether or not you invest in flood protection for your home, or misconceptions may influence engineering and building code regulations, when communicating with decision-makers. The reality is that a 1 in 100 year storm, can happen once in 100 years, twice in the same 100 years, three times or even not at all! At the risk of ranting, it strikes me as a hugely misleading communication tool which continues to purvey the risk management world, and communications in the general media today.

I am not alone in this. Francesco Serinaldi, an applied statistician at Newcastle university, wrote a paper in 2014 called: Dismissing Return Periods! Using an exclamation mark in a title of an academic paper gets a thumbs up from me! He goes into much more detail than I could on this subject, describing how univariate frequency analysis can be prone to misconceptions when return period terminology is used.

Serinaldi also suggests better alternatives for engineering and risk modelling applications. These are; the use of probability of excedance, and risk of failure over the life time of a project or average life expectancy of a person perhaps. These describe more objective and robust quantifications of frequency of specific events, or categories of events, defined by a parameter or index.

Perception of safety

Another example is described by Greg Holland in his blog on the Engineering For Climate Extremes Partnership (ECEP) website. When discussing Hurricane Katrina, he also suggests how misleading the description of the levee protection as being able to withstand a 1 in 100 year storm, evoking a ‘sense of safety’. He elaborates (as I mentioned above) that their 1 in 100 year storms is simply a 1% chance of such a storm happening in any given year (irrespective of climate variability).  He explains that this means that there is a 65% chance that such a storm would occur in the next 100 years. When changing the time period it also means that there is a 25-30% chance that such a storm would occur within the next 30 years. This starts to concern a much wider stakeholder groups, including small businesses and home owners.

Return periods can also vary widely depending on the spatial scale of an event. The Great storm of 1987 in the UK was reported as a 1 in 200 year event for many of the southern counties of the UK, whereas for parts of the south coast, it has been assessed to be more like 1 in 10 years! This is misleading in that the storm itself was large enough to cover a broad swath of land with severe impacts, but within this storm return period estimations vary. It depends on how the calculations are conducted, which data are used, thresholds that are assigned for definition of event.

One generalised return period statistic is not adequate to clearly describe the risks associated with a storm. The more objective methods suggested by Serinaldi are an alternative for engineering applications.

Public perception tangent…

As a bit of a tangent, but this has made me think about public communications too. I think that the use of analogues or a narrative, to try to recreate conditions in a viewer’s mind using past experiences, is very powerful in changing perceptions and behaviour, much more so than a misleading return period estimate.  I find it fascination how perception can change based on storytelling: one thing at which humans have always excelled.

An interesting paper by Lowe et al (2006) studies the effects of blockbuster movies such as the “The Day After Tomorrow” which can act to skew perception of risk, but also increase motivation to act on climate change and sensitize the viewer. The paper also notes a lack of knowledge on how to use this new found Hollywood-induced motivation. This is an interesting area of research in its own right.

Too many blog subjects, not enough time!