The Czech floods of 2002 may have provided the trigger for a giant step forward in modeling flood exposures

The 2002 Czech floods were truly historic. Not only was the cost in terms of human suffering immense, but also the event holds, at least for the present, the somewhat questionable distinction of being the largest ever insured flood event.

The floods were all the more extraordinary for following so hot on the heels of the 1997 Morava inundation, which generated insured losses of $0.4bn and economic losses of $2.4bn - in itself an event which few would have expected to be repeated, let alone exceeded, within just five years.

As figures now stand, the overall insured cost of the 2002 floods, including the impact on Austria and Germany, comes to $2.95bn. Within that, the gross 2002 flood loss for the Czech market was $1.42bn, a figure that is put into sharp relief when one considers that in 2001, total Czech property premiums, domestic and commercial, came to $450m.

The insured Czech loss equated to 53% of the overall economic loss for the country of $3bn and to 1.67% of its gross domestic product (GDP).

By contrast, the Austrian and German insured losses came to 0.15% and 0.06% respectively of those countries' GDPs, and represented 12% and 15% of their total economic losses.

Equally interesting is the fact that only 3% of the Czech market's insured flood losses was paid by insurers; 97% was picked up by their reinsurers.

Accordingly, in September 2002, as the reinsurance world started to gear up for the January 2003 renewals, it hardly came as a surprise that Czech cedants found little appetite among their historic reinsurance partners for continuing to support their traditional proportional programmes.

As one leading figure in the Czech insurance market wryly commented, after the 1997 floods Czech insurers received lots of letters of condolence from their reinsurers proffering unsolicited pledges of help and unrequested offers of cash advances. After the 2002 floods, the reinsurers sent loss adjusters and asked how it was possible it had happened again.

As the shutters came down on the market's traditional sources of reinsurance, cedants had to turn to the excess of loss (XL) market for outwards protections.

But in order for this sector to be able to offer appropriately priced cover, cedants need to be able to provide credible estimations for their probable maximum losses (PML) and return periods.

In the Czech case this was undoubtedly something easier said than done given that prior to 2002 meaningful probabilistic flood models (i.e. those based on detailed historic events and giving flood loss estimates associated to various return periods as compared to indicative models which calculate a maximum flood plain and an estimation of PML) had yet to be developed for any Central European territory.

The reason for this is that compared to the other two major natural perils - earthquake and windstorm - floods are a much trickier proposition to accurately model. An earthquake will occur in a specific location, have a specific intensity and last for a specific time period. From this can be calculated the potential spread of damage and its associated cost.

Similarly, windstorms are created from specific meteorological conditions and follow a path and speed that are governed by those factors with little or no ability for events to be significantly impacted by other influences.

Flood factors

By contrast, floods can be affected by myriad factors ranging from a small difference in location to the effectiveness or otherwise of human intervention.

For instance, the flow of floodwaters will be a result of factors such as riverbed terrain, the current height of the groundwater, the location and duration of the rainfall event triggering the flood, and issues such as deforestation and urbanisation.

In addition, the degree of damage will be a function of how long the floodwaters remain, the height they reach in properties and the type of buildings affected.

This means that equally crucial is not just whether flood defences and emergency alert systems exist, but also the effectiveness with which people apply them.

This introduces a significant human element difficult to account for in the calculations. For instance, in August 2002 the activation of temporary 'demountable' flood defences played a major role in protecting Prague's Old Town Square region.

That said, with flood modeling the next significant challenge to crack, it just so happened that in June 2002, Benfield's ReMetrics Natural Hazards Team had initiated a project to create a probabilistic Czech flood model using the loss data generated by the 1997 floods. The subsequent floods allowed the team to correlate results and validate the model in time for it to be launched in November, giving placing brokers the necessary tools to better negotiate new XL-based outwards programmes.

Not only were Czech cedants able to obtain much needed cover, but in many cases they were able to do so at a far more reasonable price than would otherwise have been the case.

The Czech flood models are currently evolving. The 2002 version went to postcode (zip code) level and used generic flood vulnerability curves.

Last year, the vulnerability model data was updated using the claims information generated by the 2002 event to derive Czech-specific vulnerability curves.

This year the model is being refined to better model events in the one to 25 year range.

In addition, the Charles University in Prague is conducting a Benfield-backed project to go to selected properties and collect specific physical data on the impact of the floods, enabling more detailed vulnerability curves to be created based on the analysis of single risks.

At the same time, the ReMetrics Natural Hazards Team is working in conjunction with insurers to move the model beyond postcode level and have geocoding of all risks - in other words, create a model in which any one of the country's 2.4m actual addresses can be mapped on the flood plain and its risk assessed individually.

In part, this latter initiative has been assisted by the fact that since August 2002 Czech insurers have had to completely rethink their underwriting philosophy because the old system simply did not work any more.

Accordingly, between August 2002 and November 2002, the risk carriers were faced with - and delivered upon - the mammoth challenge of repositioning their books of business through the introduction of rate increases, limitations of cover and far greater underwriting selection.

Again, this initiative was in part assisted by new modeling-driven underwriting tools, in this case MaGIS, a geographic information system developed by local technology firm MultiMedia Computer working with Benfield. The applications of MaGIS allow underwriters to enter individual risk data at the geocoding level. Two main versions have been created - Swiss Re's FRAT (Flood Risk Assessment Tool) and a system for the Czech Association of Insurers (CAP), although some risk carriers have chosen to use MaGIS in conjunction with flood zones that they have designed themselves.

All in all, it is probably fair to say that the Czech market is now the most intensely modeled flood risk in the world. But far more significant, the Czech experience has also paved the way for similar initiatives across Europe. For instance, drawing on the flood modeling experience gained from the Czech model, Benfield has been able to also develop a probability model for Slovakia. In addition it has produced indicative models for Austria and Poland, the latter of which has just been superseded by the introduction of a probabilistic model.

Separately, EQECAT is working on models for the German market while in France the rising cost of natural perils reinsurance for government-backed, state-owned reinsurer Caisse Centrale de Reassurance (CCR), has led to the creation of a joint government and industry project aimed at improving the sector's flood assessment capabilities through data sharing.

It is also possible that risk carriers in this sector may turn to the private sector for assistance if rising premium costs force them to seek reinsurance protection away from the CCR, which in May reported 2003 net profits were down by 24% due to the impact of recent natural perils such as last summer's major drought and the massive floods that hit south-eastern and central parts of France in December

As such, therefore, it may not be too fanciful to predict that in decades to come, when the market looks back at the Czech floods, they will not just be regarded as one of the major natural catastrophes of the first decade of the 21st century. They may equally be remembered as the fundamental factor in the creation of the world's first real flood models, so closing a key development gap in modeling terms.

With this gap closed, it becomes also possible to imagine the day when insurers will be able to have outwards programmes based on pan-European models giving a combined picture of their exposure to the 'big three' pan-European natural perils - earthquake, windstorm and flood - a truly significant step forward for the industry as a whole.

- Bruce Selby Bennett is Head of Central and Eastern Europe at reinsurance intermediary Benfield and San Chhea is a member of the ReMetrics Natural Hazards Team at Benfield.