Guy Carpenter has developed a state-of-the-art probabilistic flood model for mainland France in collaboration with hydrological and hydraulic modeling experts, JBA Consulting (JBA) and Intermap Technologies (Intermap), a global provider of high-quality 3D digital elevation models.
By combining a high resolution digital terrain model of mainland France with a unique 2D hydrodynamic modeling approach, Guy Carpenter’s France flood model contains the most detailed and uniformly accurate set of countrywide flood hazard maps for the French territory currently available.
Guy Carpenter’s new flood model integrates these state-of-the-art flood hazard maps into the next generation of our probabilistic model platform, providing client companies with France’s first countrywide probabilistic flood model tailored for reinsurance and risk-management purposes.
Flood plain boundary maps have been developed in collaboration with JBA, who have carried out riverflow modeling for over 80,000 km of the French river network using their cutting-edge 2D hydraulic flow model, JFLOW-GPU. Hydraulic modeling was carried out using a uniformly accurate 5 meter horizontal resolution digital elevation model for mainland France, developed by Intermap as part of its NEXTMap® Europe program. Together, these technologies have enabled the production of highly accurate flood plain boundary maps for the French river network, inclusive of water depths, for six return periods (10, 25, 50, 100, 250 and 1000 years).
Flood maps taking river defenses into account have been produced for all urban areas, representing 12,000 km of the modeled river network.
The stochastic event set includes 4,000 synthetic events extrapolated from 50 years of observed river discharge data using the theory of extreme value statistics. The event set has been developed to reflect the nature of the temporal and spatial correlation of discharge between river basins and represents the equivalent of 1,000 years of observed river discharge data.
The model also contains the flood boundaries of 14 historic events that have occurred in France over the past 150 years. The historic events are independent of the stochastic event set, but can be used to evaluate flood losses based on individual scenarios. The historic events can also be plotted on the final exceedance probability curve to act as benchmarks against which to validate losses generated from the stochastic event set.
Modeling of other flood types has also been undertaken, including flooding occurrence both on- and off-plain due to groundwater fluctuations and pluvial flood. Modeling of the groundwater and pluvial flooding hazard has been carried out for 43 major urban areas (representing approximately 50 percent of the French population), by studying the physical interactions between the components of climate, soil and river water level. The approach has enabled the refinement of current estimations of the flood hazard based solely upon the on-plain river flooding component.
The model’s “Built Environment” is a high-resolution building stock database, and forms an essential component of the France flood model. This database allows the spatial re-distribution of exposures provided at a coarse resolution (5-digit postcode or municipality), in addition to providing an accurate representation of structural characteristics that may be missing in the original exposure data.
Vulnerability functions that translate specific flood hazard intensity into an estimation of damage to individual risk types have been developed for residential, commercial, municipal, agricultural and industrial business lines. Functions have been derived from information gathered from both international and French data sources, including claims data from a number of French insurers. The functions take account of the building stock heterogeneity present within each risk classification and include secondary modifiers for certain lines (e.g. floor level situation/number of stories).
Figure 2 Relief map of France featuring the undefended extent for an extreme flood (1000 year period). Sources: JBA Consulting
Building vulnerabilities reflect two components of probability; first, the likelihood that a risk will be damaged given a specific hazard intensity and second, a probability distribution that quantifies the amount of damage to a risk for the specified hazard intensity.
The vulnerability functions have been based on the hazard characteristics of site flood depth and flood duration, in addition to taking account of demand surge effects due to extreme events. A novel aspect of the curves with respect to this new model is that the functions now capture the importance of the contribution of relatively small claims to the overall loss profile from a flood. Guy Carpenter is currently working with an external consultant to further refine the approach with respect to the parameters that most influence building damage caused by flood.
Model Inputs and Outputs
The France flood model accepts portfolio data at all levels of resolution. The high-resolution disaggregation module allows the import and treatment of low resolution data, but the model equally allows coordinate import for portfolio data geocoded at the address level.
The model produces a broad range of outputs, including exceedance probability curves and detailed event-by-event loss tables.
The newly designed portfolio loss calculation software of Guy Carpenter’s model platform (G-CAT Portfolio Analyst) allows the application of insurance terms from coverage to policy, thus reflecting the reality of an individual set of claims. In this way, the model can quantify the impact of elaborate limit and deductible combinations.
Figure 3 Flood boundary of the 1910 flood event of the river Seine included in the model’s historic events set. Source: Direction Régionale de l’Environnement d’Ille-de-France, 2007 © Digital Globe
Guy Carpenter & Company, LLC, Intermap Technologies Inc. and Jeremy Benn Associates Ltd (the “Authors”) provide this report for general information only. The information contained herein is based on sources we believe reliable, but we do not guarantee its accuracy, and it should be understood to be general insurance/reinsurance information only. The Authors make no representations or warranties, express or implied. The information is not intended to be taken as advice with respect to any individual situation and cannot be relied upon as such. Readers are cautioned not to place undue reliance on any historical, current or forward-looking statements.The Authors undertake no obligation to update or revise publicly any historical, current or forward-looking statements, whether as a result of new information, research,future events or otherwise.