Today’s catastrophe models are more scientifically robust than ever before. Dr Jayanta Guin takes a looks at how our understanding of the forces driving natural catastrophes becomes ever more sophisticated.
A critical decision for catastrophe modellers is when to incorporate new scientific theory into their models – a question of particular pertinence, for example, in deciding how to incorporate new research on the relationship between climate and tropical cyclone activity. Updating a model before scientific consensus is reached may result in unnecessary instability, as well as potential bias in the loss estimates. If new findings are incorporated after the wider scientific community comes to a general agreement, the modelled losses may change – perhaps significantly.
In the lay population, a tendency exists to regard every new “discovery” or published finding as an inviolate fact. In reality, a definitive final word is rare. Rather, science is a dynamic process in which researchers not only make new discoveries but re-examine earlier knowledge, trying to improve, build upon or extend it. This can be achieved by reanalysing old data using new techniques, or by integrating new data with the old, in order to draw a more complete picture.
In the natural catastrophe realm, these processes of reanalysis and integration may give scientists a clearer picture of precisely how past events (eg hurricanes, earthquakes and windstorms) unfolded. Armed with improved estimates of the frequency and severity of historical events, catastrophe modellers are better equipped to assess the probability that catastrophes – including more extreme events or years than have happened in the past – will occur in the future.
It is the job of scientists to investigate theories to explain physical phenomena. Competing theories nourish scientific debate, but arriving at a broad consensus can be a lengthy process. Even the consensus view may change over time, and the modeller should proceed with caution while incorporating this scientific uncertainty into the models.
Changing view of the past
Examples of the dynamic nature of science abound in catastrophe modelling. In a report issued in 1996, the United States Geological Survey estimated that the New Madrid region of the central US, which experienced a devastating series of earthquakes in the winter of 1811-1812, could expect a magnitude 8.0 earthquake once every 1,000 years on average. However, a re-examination of the data that began in 2000 resulting in a report two years later concluded that the mean frequency should be halved – to about 500 years – and that the expected magnitude was not 8.0, but rather in the order of 7.7.
Not all reanalysis efforts have such a significant impact on model results. The National Hurricane Center’s 2002 recategorisation of Hurricane Andrew from a category 4 to a category 5 hurricane, for example, made little difference to modelled hurricane loss estimates. Because models incorporate continuous probability distributions of storm parameters, such as intensity, whether Andrew was a category 4 or category 5 has minimal impact on modelled losses.
Changing view of the future
Scientific debate becomes particularly lively when the risk landscape itself appears to be changing – as in the case of climate signals and the resulting impact on hurricane activity. Unable to rely solely on past experience, scientists develop forecast models. The results produced by such models are highly uncertain however.
Since 1995, tropical cyclone activity in the Atlantic basin has been elevated above the long-term average. Scientists at the National Oceanic and Atmospheric Administration (NOAA) have linked this above-average activity to elevated sea surface temperatures (SSTs). They say SSTs are in turn linked to the positive (warm) phase of the Atlantic Multidecadal Oscillation (AMO), a naturally occurring cycle that oscillates over periods of decades.
The consensus at the NOAA is that the current warm phase is likely to continue “for years to come”. Therefore, it might seem reasonable to assume that hurricane losses will be similarly elevated and that models should adjust. However, significant caveats apply to this argument. In 2000, 2001 and 2006 – all years in which SSTs have been warmer than the long-term average – no hurricanes made landfall in the US. In two other years – 1997 and 2002 – only one tropical cyclone made landfall as a hurricane, which is also below the long-term average.
In fact, a number of climate signals other than elevated SSTs affect hurricane activity and storm track, and these may dominate and even counter their impact. In 2006, the onset of El Niño conditions produced increased wind shear in the Atlantic, which had a mitigating effect on hurricane activity despite the presence of anomalously warm Atlantic sea surface temperatures.
“The issue is how most appropriately to incorporate new scientific theory in order to provide an unbiased view of risk
There are other complexities. For example, scientists also credited the Saharan Air Layer for the low rate of storm formation in 2006. Storms over Africa’s Sahara Desert can carry significant amounts of dry, dusty air westward over the Atlantic Ocean, depriving incipient tropical cyclones the moisture and heat they need to develop.
In addition to speculation about the competing impact of various climate signals and naturally occurring feedback mechanisms, another reason to view model predictions with care is that the primary focus of scientific investigation into climatological influences on tropical cyclones has, until very recently, focused on Atlantic activity. Making the leap from increased hurricane activity in the Atlantic to increased landfall activity and, ultimately, to the affect on insured losses requires significant additional research.
The climate change/storm activity link
At AIR, climate scientists have been monitoring the relationship between global warming and Atlantic hurricane frequency since the early 1990s. Over the course of the last two years, they have undertaken extensive analyses of the link between elevated SSTs in the Atlantic and regional landfall frequency. The research has also included a critical evaluation of the historical data and its quality.
The result of this research was a new near-term catalogue of stochastic storms in 2006 and an updated version in 2007. The approach used to develop the near-term catalogue – which represents potentially increased hurricane risk over the next few years – explicitly quantifies the uncertainty in the estimates of near-term risk. The near-term model is intended to be a supplement to, rather than a replacement for, the standard model, which is based on more than 100 years of historical data.
The current state of science does not support making radical and fundamental changes to the catastrophe modelling methodology that has provided the industry with reliable results for 20 years.
The near-term catalogue is conditioned not on point forecasts of SSTs, which are highly uncertain, but rather on scientists’ projections that sea-surface temperatures are likely to remain elevated for the next few years. There is more certainty in stating that SSTs will be warmer than average over the next few years than in stating they will be warmer by a specific number of degrees. One advantage of this approach is that the inclusion of one additional season of hurricane landfall experience will not significantly change estimates of near-term risk, lending stability to model results.
You can predict the risk
The issue is not whether climate factors have an influence on hurricanes; most scientists concur that they do. The issue is how most appropriately to incorporate new scientific theory in order to provide an unbiased view of risk that insurance companies rely on to manage their exposure. In the case of the impact of elevated Atlantic SSTs on hurricane activity the issue is further complicated by the fact that the wider scientific community has not focused on landfall activity.
Catastrophe model users expect rigorous science in at least equal measure to advanced technology. Sometimes, a new scientific consensus will emerge that necessitates model enhancements. These enhancements may result in significant changes in modelled losses and consequently, the models will better reflect the true risk and enable scientifically-sound risk management decisions to be made.
Sometimes though, it is appropriate to resist the temptation to embrace the latest findings, acknowledging that more research needs to be done. It is for this reason AIR is currently offering two views of US hurricane risk while continuing to research the relationship between climate signals and hurricane landfalls.
Dr Jayanta Guin is senior vice president for research and modelling at AIR Worldwide.