Mobile homes are flooded in L'Aiguillon sur Mer in southwestern France, after Windstorm Xynthia hit the region. Few flood models are available in Europe and a pan-European model is years away, experts say.
Flood models have improved since catastrophic losses struck Europe in 2002 and 2007, but detailed models are available for only a handful of countries, and a pan-European flood model is still several years away.
Flood models still are relatively new, and experts say recent changes have yielded improvements. Nevertheless, a lack of data has hindered efforts to create broad, robust models.
Updated flood models offered by Risk Management Solutions Inc., which released its first European flood model in 2002, have improved so exposures can be modeled for a specific street or single building, said Clare Souch, vp of natural catastrophe and portfolio solutions in London.
“Updates in 2006 and 2008 were a "step change,' benefiting from the experience of the 2002 and 2007 floods, advancements in computing, the availability of higher-resolution data and improved methodologies for rainfall simulation,” Ms. Souch said.
But with no major European floods since 2007, developmental activity has slowed, said Jens Mehlhorn, head of the flood group at Swiss Reinsurance Co. in Zurich.
“Catastrophe modeling firms are planning pan-European flood models, but this appears to have dropped down their list of priorities. Activity has slowed, but if modeling companies do not develop their flood models now, it will be at least three years before any new models are released,” Mr. Mehlhorn said.
Only national flood models are available in Europe, said Mr. Mehlhorn. “But it is important that we have pan-European models, because flooding does not stop at national borders.”
For example, August 2002 floods in central Europe occurred on the Elbe and Danube river systems and affected Austria, the Czech Republic and Germany, Mr. Mehlhorn said. Those floods generated an estimated $3.4 billion in insured losses.
Indeed, modeling companies' ability to gauge flooding losses is limited to just a few countries—the United Kingdom, Germany, Belgium and parts of Austria, said Tobias Ellenrieder of Munich Reinsurance Co. in Munich.
Munich Re has its own models to help fill gaps in outside models' quality and geographical spread, but the reinsurer would like to see modeling companies develop a flood model for all of Europe, noting that many reinsurance treaties are written on a cross-border basis.
Munich Re also would like to see a multiperil model developed for Europe, Mr. Ellenrieder said. “Hydrological and meteorological correlations exist—such as the correlation between a windstorm and coastal flooding or convection storms and flash flooding—but models currently only model just one peril,” he said.
AIR Worldwide Corp. in Boston is working toward a pan-European flood model, said Milan Simic, managing director at AIR Worldwide Ltd. in London. It will simulate flooding events across Europe. So far, AIR has a detailed flood model only for the United Kingdom.
Mr. Simic acknowledged that reinsurers are disappointed they do not yet have a pan-European flood model, “but there is a lack of detailed and affordable data—such as digital terrain data—in all European countries. And each country produces its own data, and these do not always mesh well together,” he said.
AIR plans to add countries to the flood model where there is greatest demand, he said.
Oakland, Calif.-based catastrophe risk modeling company EQECAT Inc. released its EuroFlood model for Germany in 2006 and added Danube basin river flooding in Austria in 2007, said James Webb, product marketing lead for European products in London. A 2009 update added agricultural and municipal exposures and faster data processing capabilities.
EQECAT already has a pan-European precipitation event set that would aid developing a flood model, Mr. Webb said.
RMS expects to release a unified European flood model in the next two to three years, said Ms. Souch. “We will build the European flood model in one go and have one release across Europe incorporating the three existing models into the single-event set.”
Flood models have improved in recent years but shortcomings remain, said Adam Podlaha, head of impact forecasting-international at Aon Benfield Inc. in London. For example, methodologies that model the link between rainfall and runoff are relatively “weak,” he said.
In general, the data quality in flood models needs to improve, said Mr. Podlaha. Models are capable of street- or even building-level resolution, but many insurers still use only postal codes.
The latest flood models have benefited from developments in earthquake and hurricane models, Mr. Mehlhorn said. But in many cases, model resolution quality is not matched by exposure data quality, he said. “The most accurate flood model results can be expected with exposure data on geographical coordinate resolution since the risk of flooding changes within short distances. However, this high (level of) data resolution is not always provided by our clients.”
In addition, different flood models provide different results, said Mr. Ellenrieder. “Unfortunately, model results differ because (the providers) use different methodologies, granularity, data and assumptions. And this makes comparison of the results from our flood models with those of brokers and vendors difficult,” he said.
As models become more detailed with greater data resolution, “this should translate into greater accuracy and reduce model uncertainty,” said EQECAT's Mr. Webb.
Future models likely will include areas that fall outside of flood plain exposures, said AIR Worldwide's Mr. Simic. Like flash flooding caused by heavy rain and inadequate storm drainage, areas that are not flood plains can account for 30% to 50% of overall losses in densely populated countries, he added.
AIR began modeling off-flood plain flooding several years ago, but improvements are expected as more information on groundwater flow, drainage and defenses are incorporated into models in the next four to five years, Mr. Simic said.
“The goal is to model the whole atmosphere in a single model, with better definition of what constitutes a rainstorm or a windstorm,” Mr. Simic said. “We are testing the methodology, but development will take another couple of years.”







Loading comments...
