With almost no windstorm and earthquake risk, flood is the number one peril in many central and eastern European countries

This was dramatically brought home in 2002 and 1997 when devastating floods hit the region, causing insured losses throughout Europe of EUR3.2 billion and EUR1.8 billion respectively. As a result of these losses, risk awareness of flood perils has increased as has the need for quality flood risk assessment tools in the region. Flood risk assessment tools are now available to the Czech insurance market with other markets expected to follow shortly.

Wake up call to the flood peril

In 2002, two depressions brought massive flooding to Europe, the like of which had not been seen for at least a century. The floods affected an extensive area across Europe from the UK to Spain and as far east as the Black Sea coast. While initial heavy precipitation soon caused small rivers to burst their banks, persistent rainfall in Central Europe eventually gave rise to extremely high water levels along large rivers such as the Danube, the Elbe, the Vltava and the Mulde. Total economic damage in Europe reached more than EUR15 billion. During this event, the Moldau River in the Czech Republic exceeded the water level of the major 1890 floods in Prague. Some 200,000 Czech residents were evacuated during the flooding.

Insured losses in the Czech Republic amounted to EUR1.2 billion, compared with an economic loss of EUR2.3 billion.

The 2002 floods were not an isolated case. Only five years earlier, in July 1997, the two depressions Xoslka and Zoe brought periods of exceptionally heavy rain to the Czech Republic and Poland. The subsequent flood flows inundated about 2% and 6% of the land surface of Poland and the Czech Republic, respectively. The Odra River in Poland and Morava River in the Czech Republic both exceeded 100-year water levels in many places by more than two meters. The devastating result of the flooding was 100 dead, 300,000 evacuated. Economic damage totalled around EUR3.5 billion (at 1997 prices). About one half of the losses (EUR1.6 billion) were in the Czech Republic out of which EUR0.3 billion were insured.

Industry rethink

The floods of 2002 and 1997 threatened or even ruined many people's livelihoods.

In some countries, such as Germany, the situation was made worse by the fact that much of the material damage and economic losses was uninsured.

The insurance and reinsurance industry has since responded with a number of initiatives as part of its commitment to providing traditional insurance protection for flood perils. In Germany, the industry has been working with the government to finds ways of developing the market for flood coverage.

While in the Czech Republic, Poland and Slovakia, where flood insurance penetration is not an issue, the focus has been on technical developments, such as flood risk assessment tools, to improve risk assessment and aid pricing.

Flood tools in Czech market

Shortly after the 1997 flood, the Czech Association of Insurers (xC AP) started an initiative to improve flood risk assessment in the Czech Republic. However, the project evolved only slowly due to high costs, and it took the 2002 flood event before a tool could be launched in 2003.

The basis of the tool is four flood risk zones covering part of the river network in the Czech Republic (16,000 km) and a database of 2.5 million street addresses. Insurers can use the tool for selective underwriting and to establish risk adequate premiums.

In addition to the flood risk assessment tool developed by xC AP, reinsurance broker Benfield Group has developed an event based model which is primarily used for peak loss studies and structuring and pricing of catastrophe excess of loss reinsurance programmes.

Swiss Re has also invested in developing general risk analysis techniques for flooding. The fruits of this labour include a method which allows a country to be divided into risk zones, i.e. different flood risk categories, and a simulation tool for floods with emphasis on pricing non-proportional reinsurance covers. In a joint venture with the Czech Multi Media Company (MMC), Swiss Re made its knowledge of flood risk delineation available to the Czech insurance market in the form of the Flood Risk Assessment Tool, FRAT 1.0.

Solution for the Czech market

FRAT 1.0 is designed as a stand-alone software solution (CD-ROM) for use by insurers when assessing and pricing primary property business.

It is based on Swiss Re's flood zone delineation technique - see box text - and it offers two basic functionalities.

First the user, for instance a risk manager or insurance agent, enters information about the location of a premise by full address (street, house number, city). The address, or part thereof, is located and transformed into geographic co-ordinates, which are used for flood risk zoning analysis.

Secondly, the system returns information about the flood risk of the selected location and highlights it on a visualisation screen. The tool distinguishes six different flood risk zones (zones ranging from 1, very low, to 6, very high risk) and the historically observed maximum flood boundary. The result is also translated into xC AP's tariff zone naming convention.

FRAT 1.0 flood risk zoning is based on the best digital terrain model (DTM) that is available in the Czech Republic. The DTM features a horizontal resolution of 10 m, which means that there is a reading of the elevation every 10 m.

The address database used in FRAT 1.0 is the most up-to-date geo database currently available in the Czech Republic and covers about 70% of addresses in insurance portfolios (CEDA database). This makes it possible to locate the co-ordinates of individual doorsteps in the 75 largest cities of the Czech Republic (Fig 2 and 3).

The position of the premises in question is visualised on a map. The system uses Atlas CR 150 and city maps (scale 1:10,000) for approximately 75 cities of the Czech Republic.

Further developments

FRAT 1.0 is now used by almost all property insurers in the Czech Republic, allowing them to identify high exposed risks and more accurately price flood risks. The Insurance Association and major insurers in Slovakia have expressed their wish to have a similar tool. Also, the Polish Insurance Chamber started a working group which evaluates possible developments of flood risk assessment tools. A FRAT like tool is high on the list.

Swiss Re and MMC have close contact to the insurance associations in both markets and are heavily involved in both developments.

While the first and most important step is to have proper flood risk assessment tools based on detailed flood risk zones in place, there is a wider trend towards risk dependent pricing functionalities, accumulation control and, finally, also event based models for peak loss estimations.

There are approaches available; it is only a matter of willingness to have such tools in eastern and central European markets.

Jens Melhorn is head of flood group, Swiss Re, and Joseph Breitsameter is a natural perils specialist at Swiss Re, Germany.


Information on areas threatened by flood is a prerequisite for flood risk assessment. However, thematic maps which depict the flood hazard on a national level are available for very few countries in the world.

Therefore, we have developed a geo-statistical model which yields nation-wide 50 to 500 year flood extents.

The model is called geomorphologic regression since it uses a novel method of multiple non-linear regression analysis and geomorphologic catchment attributes as parameters.

The underlying assumptions of the geomorphologic regression model are that, first, naturally flowing rivers shape their channel and flood plain according to basin inherent forces and characteristics. Second, the flood water extent strongly depends on the shape of the flood plain. For example, in V-shaped valleys one would expect high water levels but a small flood extent. In contrast, the flood extent in low land flood plains might be large but water levels are low. In this case, only small differences in elevation may be sufficient for property to be safe from flooding.

The inherent forces of a river can be described by flood water volume and basin characteristics like catchment area, slope etc. which can be derived from a digital terrain model. Flood water volume can also be described by catchment attributes, including climate or any other rainfall information.

The characteristics of a location with respect to its situation within a flood plain can be defined by the vertical and horizontal distance to the relevant river.

In the model, the probability of any location being inside a flood risk zone is dependent on the three parameters: horizontal and vertical distance to the next river and the catchment area of the next river at that point.

The regression model was calibrated and validated at known flood risk zones in the US. In a second validation study with independent flood zone data in the UK, the ability of the model to provide reasonable flood risk information in a different environment to the US could be approved (Fig 1 above). The prediction success of the methodology was such that Swiss Re decided to apply for the patent.


In the insurance and reinsurance industry, two different groups of flood risk assessment models are used. The first group of models is based on flood zones which describe the flood risk along rivers. The risk zones depict the probability (for example, on average once in 100 years) of each location in a country being flooded. Zone based tools are generally used by primary insurance companies for risk selection and premium calculation.

Additionally, zoning models help to determine the flood exposure of entire portfolios by providing the proportions of risks located in the different flood risk zones. This information can then be used to calculate the annual expected loss of a portfolio.

However, zone based tools do not provide peak losses caused by single flood events, since not all rivers in a country are flooded at the same time and at the same intensity. This is taken into account by the second group of flood risk assessment tools. The so-called, event based tools calculate flood losses for a series of historic and/or probabilistic events and are mainly used by reinsurance companies for structuring and pricing non-proportional treaty business.