AIR CLASIC/2 v13
AIR’s European windstorm model underwent a major update in 2010. There were, therefore, only a few changes in the 2011release, with an expansion in modeled countries to include the Czech Republic, Estonia, Finland, Latvia, Lithuania and Poland (1). The historical storms of Janika (2001) and Xynthia (2010) were also added to the model’s catalog.
RMS RiskLink v11.0 (SP2)
RMS’s European windstorm release saw more significant changes, requiring careful review and validation on a country-bycountry basis. The new model expanded its territory reach to include the Czech Republic, Slovakia and Poland. An updated hazard module also resulted in increased hazard in lower return periods and decreased hazard in higher return periods. Depending on the portfolio chosen, these changes, when combined with updated vulnerability curves, can result in significant loss decreases for Austria, Belgium, Luxembourg, Netherlands, Norway, Sweden and Switzerland at a 100-year return period. Increases can also be observed in Denmark, France, Germany, Ireland and the United Kingdom for the same return period. Other notable changes from the model update saw increased annual aggregate losses arising from storm clustering in a single year, and there was a significant change to the industry exposure database.
Guy Carpenter Perspective on European Windstorm Model Changes
In early testing of the RMS model, Guy Carpenter has been surprised by the extent of the vulnerability change for standard classes of occupancy and construction and the effect from storm clustering. In general, users have found the unforeseen magnitude of these changes to be confusing and time-consuming. Moreover, the changes communicated by RMS are based on a totally rebuilt industry database and not always in line with other test runs. Guy Carpenter has finished a comprehensive test project that will enable our clients to fully understand the changes in model components and the implications on the computed results.
Guy Carpenter urges greater transparency in disclosing the model components for users in order to evaluate the products more efficiently. We also anticipate greater industry collaboration for vendors who do a better job in disclosing their model assumptions.
1 The research model domain has always included these countries. However, their exposure, vulnerability and loss information were not included in Version 12.0. The stochastic catalog has not changed between the versions and event IDs remain the same.