How can we tell if things are getting worse?

Blog - 31.05.2018

May 21st, 2018 | Richard Dixon

Last year saw quite a few proclamations around the role of climate change in the ferocity of the hurricane season. It was with this at the back of my mind that I finally got round to reading Roger Pielke Jnr's book The Rightful Place of Science: Disasters and Climate Change. A fair part of the book is an illuminating read on climate change and its relationship to the perceived increase of extreme events - and also, disappointingly, how Pielke has been treated by fellow scientists merely for thinking differently from the crowd. 

Prompted partly by the work of Pielke, a slightly frosty exchange on Twitter between two modelling firms around historical evidence for the use of near-term views of hurricane risk and also the forthcoming hurricane season, I wanted to take a look at how easy it is to tell whether there is an increasing trend in a particular hazard, driven by climate change or otherwise. This also ties in with some work I'll be presenting on European Windstorms at the Oasis conference around whether we are seeing any changes in windstorm frequency in recent decades by using multiple simulations of historical conditions.

So, I’ve done a very, very simple exercise. Imagine an extreme event of your choosing - be it flood, hurricane, windstorm or plague of locusts - whose underlying risk of occurrence is becoming more frequent, such that it increases in yearly probability from 5% (20 year RP) in 1940 to 10% (10 year RP) in 2020, as shown in the chart below. As you might imagine, this amounts to a fairly noticeable doubling of the frequency of this event over this period.

RD Blog image 1

The thing I'm wondering is: "what are the chances that a change in this low-frequency event becomes evident from the historical record over time?”. For the purposes of this study, I’ve decided to compare the first half of the period (1940-1980) to the second half (1980-2020) to look for evidence that our event becomes more frequent between these two periods.

So, I've effectively simulated 1000 "histories" of 1940-2020 and tallied the occurrence of the event from 1940-1980 and 1980-2020 in these 1000 histories, to see how "detectable" a change in the event frequency is across all the simulations.

To keep it simple, for each “history” from 1940-2020, the results have been split into three categories:

• More events are evident from 1980-2020 (Increase)
• Fewer events are evident from 1940-1980 (Decrease)
• No change in the number of events between 1940-1980 and 1980-2020 (No change)
My gut feel was that this change would be detectable pretty much most of the time, given the size of the change in the frequency from a 20 year RP to a 10 year RP:
RD Blog image 2

So yes: in the majority of 1000 simulations of 1940-2020 (591 of 1000, so around 60%), the increase in underlying risk results in an increase in the event occurrence in the latter (1980-2020) period.

However, what surprised me the most in this example is that in 255+154 = 409 of the 1000 80-year simulations (around 40%), no change or even a decrease in the frequency of events was detected: despite a doubling in the underlying risk.

Even though the underlying risk was increasing notably over time, you effectively only have a 60% chance of noticing it in the historical data: at least in the way we’ve binned the data in this simple example. Essentially: if you’re lucky, the “history” that you see in the face of an increasing (or indeed decreasing) underlying hazard will represent the changes in the underlying risk.

It’s admittedly a very crude example and I could look into generalising this by way of the importance of length of history for detecting events, for example, but this example certainly helps me to think a little more clearly – or maybe a little more inquisitively - on how we use historical data before shouting a “yes” or “no” to thing such as climate change influences.

It also highlights how short historical records may not show up underlying trends in the data.

What’s the next step? Well I think academia are already starting to tackle it. Re-simulating our atmosphere’s recent history tens or hundreds of times in climate models may help provide us with additional “histories” that could provide more clarity around the role of climate change in the changes of frequency in tail risk events. These additional historical simulations could help us from relying solely on the one history that we have to try and spot trends. Of course, let’s not fool ourselves that this is the golden bullet: climate modelling isn’t without its problems. Many climate models have biases that need correcting; they also cannot properly resolve the more intense events.

However, this methodology is certainly a step in the right direction to better getting to grips with the thorny issue of understanding whether underlying forcings such as climate change may already be affecting the underlying risk in our catastrophe models.

About Richard Dixon:

Richard Dixon-437162-editedRichard has spent the last 17 years in the insurance industry building and researching catastrophe models at a model vendor and then evaluating them whilst working for brokers and reinsurers. He now is a consultant to the insurance industry at CatInsight, specialising in model evaluation. Most recently he was Head of Catastrophe Research at Hiscox, being responsible for their internal "View of Risk". Prior to working in the insurance industry, he obtained a PhD in meteorology at Reading University, specialising in extra-tropical cyclones. For more information, visit Richard Dixon's blog:

About ModExTM

ModEx is the cat risk modelling platform for the (re)insurance industry, delivered by Simplitium and powered by Oasis LMF.

ModEx delivers flexible, scalable and secure cat risk modelling services via a hosted and fully managed environment. The platform provides the (re)insurance industry with a cost-effective and reliable way to meet their cat risk modelling requirements. The ModEx solution caters for the full suite of perils and allows model providers and other (re)insurance systems to meet the individual needs of customers through a robust and secure shared services platform.

The platform is fully independent and transparent, and provides (re)insurance firms with the tools to gain a deeper understanding of their risk by accessing the best models from large and niche model developers from around the world. For further information, please visit