With an annual European average market loss of roughly €4.4 billion, extratropical cyclones constitute one of the major risks insurers face, especially when these storms come in succession. “Storm clustering” happens when two or more storms occur within a short timeframe: it characterizes the dependence between several closely-timed storms. In meteorological terms, a storm is the encounter of a low pressure system with a single jet-stream, whereas in the case of storm clustering, the clustered storms are seen as the encounter of several low pressure systems with a single jet-stream.
This phenomenon, illustrated by the infamous Lothar and Martin storms of 1999 (as-if market loss: €13 billion), plays a major role in reinsurance. Indeed, in order to take the dependence between storms into account, the "hours clause" in reinsurance treaties allows the aggregation of events occurring in a given timeframe (e.g. 72 hours). All attained losses are thus considered as losses from a single event.
In order to better apprehend storm clustering, we first defined a set of axioms by questioning the market’s clustering hypotheses (i.e. brokers, reinsurers, modelling agencies). We then used 200 years of simulated climatic data from Météo-France in order to test each of these axioms. Finally, we derived a proper paradigm for storm clustering, and we built a scoring methodology based on this paradigm for storm models.
We also investigated the genesis of storm clustering by studying the link between the subtropical jet-stream, low pressure systems and storms affecting France. The preliminary results show that the subtropical jet-stream responsible for extratropical cyclones over Western Europe is often stable over a period of a week or even longer. Moreover, evidence supports the idea that this jet-stream is the main cause behind Western European Storms. As such, we believe that the 72 hours clause is underestimated and should be re-evaluated in the light of these facts.