Creating a multi-peril model of The French Antilles



The French overseas territories of Guadeloupe, Martinique, Saint-Martin and Saint-Barthélemy are collectively referred to as the French Antilles. The islands have a subtropical climate tempered by trade winds, with moderately high humidity and a rainy season from June to October. Periodically, the islands are affected by tropical storms; each is subjected to winds exceeding hurricane strength approximately once every ten years on average. 


Model Development

JBA was commissioned by Caisse Centrale de Reassurance (CCR), the French state-backed reinsurance company, to develop a probabilistic catastrophe loss model for hurricanes. The resultant model is multi-peril; it includes property damage quantification for the full range of plausible wind, rainfall-induced flooding and coastal storm-surge events. 

The model is driven by many thousands of stochastic hurricane tracks in the North Atlantic. A built environment dataset, which can augment the structure-type information provided in insurance portfolios, and a suite of region-specific vulnerability functions complete the model. 


The work presented a number of interesting challenges, for example:

(a) The need to generate stochastic hurricane tracks and associated attributes that extended beyond the historical ranges with respect to location and intensity, but which were nonetheless physically realistic.

(b) The need to model the wave setup component of extreme sea-levels associated with hurricane passage (in addition to the wind and pressure driven surge); this was particularly important due to the shallow bathymetry of the region.  The inundation of nearshore areas by the sea can cause considerable property damage. During hurricane passage, low-pressures and high wind speeds elevate the sea surface (storm-surge). The additional hazard component of wind-driven waves, known as wave setup, is often neglected, but can contribute significantly to coastal flood risk. 

How we resolved (a)

A new approach to stochastic hurricane track development, which combined the best elements of previously published approaches and the results of R&D by JBA, was developed. The final database for the entire North-Atlantic region contains 153,882 hurricanes, and spans a period of 10,000 years. 

How we resolved (b)

Here, we produced coastal flood footprints for synthetic flood events that accounted for both wave setup and pressure-driven storm surge. When forced by historical meteorological observations, the models were able to reproduce historically observed sea-levels extremely well. They were then applied to simulate surge associated with plausible synthetic events (based on estimated extreme meteorological boundary conditions). 

Hurricane Luis