Analysis of the French insurance market exposure to floods: a stochastic model combining river overflow and surface runoff

Abstract

Abstract. The analysis of flood exposure at a national scale for the French insurance market must combine the generation of a probabilistic event set of all possible (but which have not yet occurred) flood situations with hazard and damage modeling. In this study, hazard and damage models are calibrated on a 1995-2010 historical event set, both for hazard results (river flow, flooded areas) and loss estimations. Thus, uncertainties in the deterministic estimation of a single event loss are known before simulating a probabilistic event set. To take into account at least 90 % of the insured flood losses, the probabilistic event set must combine the river overflow (small and large catchments) with the surface runoff, due to heavy rainfall, on the slopes of the watershed. Indeed, internal studies of the CCR (Caisse Centrale de Reassurance) claim database have shown that approximately 45 % of the insured flood losses are located inside the floodplains and 45 % outside. Another 10 % is due to sea surge floods and groundwater rise. In this approach, two independent probabilistic methods are combined to create a single flood loss distribution: a generation of fictive river flows based on the historical records of the river gauge network and a generation of fictive rain fields on small catchments, calibrated on the 1958-2010 Météo-France rain database SAFRAN. All the events in the probabilistic event sets are simulated with the deterministic model. This hazard and damage distribution is used to simulate the flood losses at the national scale for an insurance company (Macif) and to generate flood areas associated with hazard return periods. The flood maps concern river overflow and surface water runoff. Validation of these maps is conducted by comparison with the address located claim data on a small catchment (downstream Argens).

The challenge of forecasting impacts of flash floods: test of a simplified hydraulic approach and validation based on insurance claim data

“…Over the last 15 years, an insurance claim database has been developed by the Caisse Centrale de Réassurance (CCR) within the framework of its reinsurance contracts with its clients (Moncoulon et al, 2014). This database covers the entire French territory, and the quality of data is considered as acceptable for the period since 1997.…”

“…And it is now recognised that end users, such as emergency managers, who have little time for situation analysis and decision making during flash floods, crucially need rapid assessment of the possible field consequences and damage severity (Schroeder et al, 2016;Cole et al, 2016). Moreover, a direct forecast of possible field consequences opens the possibility for assessing the performance of flash flood forecasting systems in ungauged areas, based on reported consequences, as a surrogate for measured flood discharges (Versini et al, 2010a;Naulin et al, 2013;Javelle et al, 2014;Moncoulon et al, 2014;Saint-Martin et al, 2016;Le Bihan et al, 2016). In the near future, real-time assimilation of proxy data for flood magnitude such as information contained in reports of rescue services or social networks can be envisaged.…”

Modelling the evolution of the financial impacts of flood and storm surge between 2015 and 2050 in France

“…The flood model developed by CCR [7] is made up of two complementary models. The first is the surface run-off model.…”

“…This overflow is based on the spread of water on a digital model of ground, describing the topography of the impacted area. These two models are complementary and each has their own importance, since the CCR's loss experience data have shown that a large proportion of damage to property -up to 50% [7] -is not located in the overflow area of the main water courses, but perhaps caused by secondary water courses or by surface run-off.…”

“…CCR (Caisse Centrale de Réassurance) is a public reinsurer providing the unlimited state-guarantee to its clients for natural disaster coverage. In this purpose, CCR has developed models to evaluate insurance damages caused by climatic events such has flood [7], drought and storm surge [8].…”