Winter storms, or extratropical cyclones in meteorological terms, are a major contributor to Europe's overall catastrophe risk. In December 1999, Europe was hit by three devastating windstorms - Anatol, Lothar and Martin. They were among the most destructive and costliest storms to hit Europe.
Anatol developed on 2 December in the North Atlantic northwest of Ireland.
On 3 December, the storm hit northern Europe with record wind gusts of 180-185 kph over the German Bight and southern Denmark. Anatol caused more than 20 fatalities, EUR2.9bn in economic losses and left 165,000 households without electricity.
On 25 December, Lothar formed over the north eastern Atlantic. Arriving on the Brittany coast on the morning of December 26, the storm took only nine hours to cross France. It led to the closure of Paris' two airports for several hours. Half a dozen people were seriously injured in the French capital by falling walls or collapsing roofs, and winds were so high that police at one point barred vehicles and pedestrians from the Champs Elysees because of flying roof debris. Many weather stations registered record wind speeds, including gusts of almost 180 kph at Paris-Orly airport and 215 kph atop the Eiffel Tower. As the storm left France, gale-force winds swept through Switzerland with peak gusts of 50 to 100 kph reported in the flatlands and 200 kph atop the Jungfrau peak in central Switzerland.
Overall, Lothar was responsible for 110 fatalities, EUR11.5bn in economic losses, and it left over 4 million households temporarily without electricity.
By the evening of 26 December, Lothar had left western Europe, but another storm, Martin, followed close behind. On 27 December, Martin battered the regions around Bordeaux, Biarritz and Toulouse in south western France with winds of more than 130 kph and affected areas as far as northern Spain, western parts of Switzerland and northern Italy. All flights to Bordeaux, Toulouse and Biarritz were canceled and train traffic was suspended throughout the region. Though less damaging than Lothar, Martin was, nevertheless, responsible for 30 fatalities, EUR4 billion in economic losses, and the temporary loss of electricity in over 1 million households.
Modelling extratropical cyclones using NWP
The purpose of catastrophe modeling is to anticipate the likelihood and severity of catastrophic events like these so that companies can prepare appropriately for their financial impact. Windstorms have traditionally been modelled using statistical parameterisation, where local wind speeds are estimated using simplified factors derived from statistical distributions.
While often adequate for estimating the overall losses for an event, these techniques lack the realism and high resolution required to produce accurate loss estimates for individual properties or portfolios that deviate from the industry wide distribution of exposures. Moreover, they are not optimal for estimating the frequency of events such as severe European extratropical cyclones, which are infrequent in nature and whose frequency and intensity may be affected by climate change.
Numerical weather prediction (NWP), which is used by advanced meteorological agencies around the world, replaces the simplified parametric components of statistical models with much more realistic components that embody the actual physics of the earth's atmosphere. By solving dynamically the mathematical equations that govern atmospheric flow, NWP accurately depicts the time dependent, three dimensional structure of extratropical cyclones.
Since it is vertical effects - such as gravity waves and convective downdrafts - that typically result in the most serious damage, NWP represents a major advance over the conventional, parametric approach.
NWP based probabilistic catastrophe risk models
To create a fully probabilistic model that estimates the likelihood and severity of catastrophic events, catastrophe modellers simulate thousands of potential events to cover the entire spectrum of possible storm experience.
Statistical models generate events by drawing from probability distributions developed for a small subset of storm characteristics. Any natural correlations between these storm characteristics must be externally imposed using rough approximations. Moreover, when based solely on limited historical data, a statistical model will not capture the full range of events that can occur in nature.
With a physical model, the laws of nature that govern the model enforce appropriate correlations between storm characteristics and outputs, such as local wind speeds. Since extratropical cyclones are greatly influenced by surrounding atmospheric conditions such as high-pressure ridges, low-pressure troughs and the jet stream, any changes to these 'initial' conditions will result in a different windstorm - one that can potentially be quite different in effect. This natural phenomenon can be used to generate hypothetical storms that have not yet occurred. By repeatedly perturbing the initial conditions of actual storms by small, but equally likely amounts, a very large set of potential storms can be generated that will reflect all possible combinations of intensity, spatial footprint and temporal evolution - an ensemble in meteorological terms.
In this fashion, a very large catalogue of realistic storms is produced that reflects the true likelihood of various combinations of atmospheric conditions and resulting storm characteristics. As the historical record shows, most of the simulated storms are of minimal intensity and cause little damage. However, a reasonable number are of moderate intensity, and a few represent the extreme storms that result from the low probability (but still possible) confluence of extreme conditions involving storm intensity, vertical transfer of energy, and jet stream strength and position.
Thus, the frequency of the most extreme events from NWP based models results from the probability of critical atmospheric conditions all occurring simultaneously, which is a natural output of an NWP based ensemble. In contrast, the frequency of events from traditional parameterised models is based on idealised statistical distributions, which are mathematical approximations of nature at best. Therefore, NWP based models are better suited for the creation of fully probabilistic catastrophe risk models.
Lessons for today
Although it has been only five years since these storms wreaked havoc on Europe, the insured losses from a recurrence of these windstorms today would be higher simply due to changes in property values and the costs of repair and replacement. Moreover, these storms could have easily been worse. Although Lothar's most damaging winds were quite close to Paris, AIR Worldwide's NWP based European Extratropical Cyclone Model indicates a shift of the storm just 60 km to the northeast would have produced insured losses nearly 25% higher than the actual event.
Of course, the real purpose of catastrophe models is not to estimate losses from historical storms, but to anticipate the likelihood and severity of future events. AIR's European Extratropical Cyclone Model indicates that European losses on the order of EUR7 billion (approximately the size of Lothar) have a return period of 10 to 15 years, while losses of EUR3 billion (approximately the size of Anatol and Martin) have a return period of approximately five years.
These frequencies appear quite reasonable when compared with the last 30 years of actual data. Table 1 contains the actual insured losses for the ten major European windstorms since 1976 adjusted for inflation, population growth and other factors influencing property values. There are three storms- Capella, Daria and Lothar- that would have caused European losses in excess of EUR5 billion at current, exposed property values. The occurrence of three storms in roughly 30 years indicates that a return period of approximately 10 years is reasonable for a loss of this size or greater.
Similarly, a EUR3 billion loss should be expected about once every five years since this size loss has been experienced five times since 1976.
Since 1999, Europe has experienced a period of relative calm. However, as population density continuez to increase, so does the potential for large losses. NWP based models are the best tools for estimating the likelihood and severity of extratropical cyclone risk, which is required by corporations, insurers and reinsurers to engage in the proactive decision making and strategic planning necessary for financial stability in the face of such extreme events.
Dr. Peter Dailey manages AIR Worldwide's atmospheric science team. which develops AIR's extratropical cyclone, hurricane, winter storm and severe thunderstorm models.