From earliest history, flood has been the major natural catastrophe in northern Europe.
Dr Robert Muir-Wood looks at the historical record and how the risks have changed.

Twenty thousand people were claimed drowned on the coasts between the Rhine and Elbe in 1164, thirty-six thousand in a flood in 1219 and fifty thousand in the Netherlands on the 14 December 1287. On 16 January 1362, the most damaging windstorm known to have affected central England, on the continent caused the Grosse Mandranke – the great man drowning – in which tens of thousands of lives were lost and about 40 parishes were permanently removed from the map of the west coast of Schleswig-Holstein (between Denmark and Germany).

Even greater was the loss of life in the “All Saint's Day flood” between 1and 2 November 1570 when the cities of Amsterdam, Muyden, Rotterdam andDordrecht were flooded, as well as all the islands of Zeeland. Without acensus, estimates of loss of life in all these early events are prone to exaggeration (figures given for the 1570 flood range from 100,000-400,000).However, these were undoubtedly cataclysmic events, the largest catastrophes in the past millennium of Europe's history.

In the most exposed areas around the southern North Sea, the first medieval flood defences encouraged increased concentrations of people and property in the fertile lands close to the sea. As with all flood protection schemes, while losses at short return periods were effectively removed, there was a dramatic “cliff-edge” increase in the potential for catastrophic losses. Although at lower probabilities and longer return periods, these catastrophic losses far exceed what would have occurred if the defences had never been built because of the accompanying population increase.

This is as true today as it was when the first earth embankments wereconstructed. All that has changed is the level of risk associated with the “catastrophe threshold”. In late medieval times, embankments protected against the 10 or 20 year flood. By the eighteenth century, following the loss of 11,500 lives in a great surge of 1717, there were big improvements in defences and only 800 drowned in the largest nineteenth century surge in 1825. Yet in the twentieth century, these reductions in risk appeared to have been checked: more than 2100 drowned ˆ of which around 90% were in Holland - in the 1953 surge, and 312 people died in north-west Germany in 1962, when 20% of the city of Hamburg, including 120,000 houses and sections of the U-bahn tunnels were flooded.

The risk today
So what is the risk to life and property today from flood? The answer is: very variable. The experience of the 1953 floods in Holland was so traumatic that it became government policy to reduce flood risk below designated levels of “one in 10,000 per year” for much of the country and “one in 1250 per year” close to the principal rivers and canals. This national focus on risk mitigation is not surprising for a country that is25% (and sinking) below mean sea-level and which, without protection, would suffer floods to 65% of its land area.

In Germany, too, the 1962 floods that reached 3.9m above the expectedlevel, inspired an enlightened programme to raise the height ofembankments, in time for the 3 January 1976 surge to arrive a full 0.4mhigher than 1962. Although 10,000 people were evacuated, there were nofatalities. The 1976 surge, in turn, spurred the Danish government toinvest in improved defences that were successfully tested five years laterwhen a 5m surge accompanied the most damaging windstorm in 100 years innorthern Denmark.

These co-ordinated national responses contrast with the situation along the east coast of the UK where sea-defences are a patchwork of local decisions about risk: with defences built to withstand surges with varying inexceedence probabilities from one in 50 to one in 10,000 per year. Althoughthis approach is often criticised, the floodable areas are relatively independent, and in the 1990s attention has focussed on improving and upgrading defences where there is the greatest risk.

Since the construction of the Thames Barrier, there is not such aconcentration of flood exposure in the southern UK as in the Netherlands,where a single major breach could imperil the sub sea-level economicheartland of the nation, and potentially cause a catastrophic loss comparable to a major Californian city earthquake.

Coastal protection
The coasts around the southern North Sea are sinking, while the thick Rhinedelta sediments of Holland are compacting and subsiding. The globalsea-level is also rising, and so every 100 years, the height of defences needs to be raised by around 0.2-0.4m just to keep the risk of flooding constant. Leave the defences alone, and the risk of failure rises: more than doubling every century. The great inlet of the Zuider-Zee was initially created by the St Luciastorm-surge of 1287, when the sea first broke through the protective sand-dunes. Repeated medieval storm surges expanded the permanent flooding to occupy more than one-fifth of the original land area of the Netherlands.

The Dutch attention to regaining their territorial integrity, whatever its sub-sea elevation (and whatever the associated cost), contrasts with the situation around much of the rest of the southern North Sea where the former principal town of East Anglia in England, Dunwich, on the coast of Suffolk, has been completely eroded away and the island of Heligoland, 60 km across in 800 AD, is today only 1.5 km wide.

The coastal flood protection systems comprise both passive and active components, and it is the active systems that can be the most vulnerable. On 27 October 1997, a dredger struck the Thames Barrier, jamming a gate in an open position for more than a month, fortunately surge-free, as well as clogging the gate mechanism with 3000 tons of gravel dropped by the sunkendredger.

In catastrophes, systemic failures proliferate. The catastrophic windstormthat causes a major storm surge may already have taken out the power supply(as did the 1987J storm in Sussex and Kent) and mobile-phone and television transmitters. Even though the 1953 surge wave took more than six hours to move down the UK coast, the loss of the telephone lines meant no warnings were passed along. Modelling flood catastrophe riskThe link between the windstorm and the associated surge is not only important for considering the risk for a specific facility but also for any insurer or reinsurer concerned with its total potential losses. In 1998, RMS launched its east coast UK storm surge flood catastrophe analysis model comprising 3000 combinations of surges with different states of the tide. Each surge is generated from the stochastic windstorms in the multi-territory European windstorm model. Within the surge model, the 900 individual defences have each been modelled for their potential to be overtopped or breached in a surge event, and the volume has been time-stepped through the defences to allow the flood to be propagated inland.

Previous models employed to determine how much surge-flood reinsurance to purchase were highly over-conservative, discounting the role of the sea-defences and “bath-tubbing” the flood at constant elevation inland. However, as actual surge events on the UK east coast show, the duration ofhigh-water is typically only two to four hours, during which time there is no chance for a flood to equilibrate its level across a wide flood plain.What do the expected levels of annual flood risk of one in 1250 and one in 10,000 in Holland mean for a specific facility? The actual risk is certainly higher than these designations. Flooding could be from a direct failure of part of the sea-defence system in a catastrophic windstorm and storm surge, from intense rainfall overwhelming the local pumping or storm sewerage capacity, or even a breach in the side of a neighbouring canal.

For some areas of the country the principal risk is the rivers: 200,000people were evacuated in January 1995 when the region between the Waal andMaas rivers came within a whisker of being flooded in a Rhine flood with a return period of only “55 years”. Chastened by this experience, the Dutch have improved the river defences, but they have admitted that, for ecological and amenity considerations, continued raising of the river defences is not viable, and a proposal is now under consideration to dig an additional channel for the Rhine to take more of the peak discharges to the north.

River flood is a major problem in a number of European towns and cities. In 1910, four km3 of water flowed through the streets of Paris, requiring the evacuation of 150,000 people and causing a loss estimated at today's valuesin excess of euro 1 billion. In 1930, catastrophic floods on the river Tarn flooded Montauban to the depth of 6-7m, drowned 200 people, destroyed 11major bridges and put 500 factories out of action.

Remedial measures (principally dams) on both the Seine and the Tarn have reduced the risk of moderate floods but, like the sea-defences, there remains a cliff-edge of probabilities beyond which there is a prospect of major catastrophic losses. In the UK the floods of Easter 1997, showed how a relatively local event affecting the headwaters of a number of catchments could generate losses in excess of euros 200 million. On the major rivers draining to the north of the mountains of central Europe, including the Rhine (1993 and 1995) and the Oder (1997), economic losses from major floods have exceeded euros 1 billion. Passing to the south of Europe, higher temperatures lead to the prospect of more intense rainfall events, in particular for autumn storms replenished from the warm Mediterranean and depositing their precipitation on the first topography inland. Catastrophic floods caused hundreds of millions of euros damage in France at Nimes in 1988 and Vaison la Romaine in 1992, and on the Arno through Florence in 1966.

Insurance
Flood insurance in Europe is currently very patchy: almost universal in the UK; covered by flat-rate government protection schemes in countries like France and Spain; of limited availability in Italy and Germany, and unobtainable in the country with the greatest concentration of risk: the Netherlands. To help sponsor increased privatisation of the flood risk markets requires two key components: the ability to price flood risk and the ability to manage flood risk portfolios.All the series of new probabilistic models we have under development for Europe are “catastrophe models”, which comprise the totality of the losses within a catastrophe event. For storm surge, this includes the losses from the accompanying windstorm, that may be in a different region or territory to the surge. For river flood, catastrophe modelling requires an understanding of all those areas of flooding affected in the same event that may be on different sections of one river-system, or many neighbouring catchments.

River flood catastrophe modelling has required us to undertake a whole new programme research to understand the properties of different classes of precipitation events and the correlation of the event with the antecedent conditions, such as lying snow.

A major problem for flood estimation concerns stationarity in the hydrological record. Since 1980 there have been an unusual number of severefloods across northern Europe, including two 50 year return period events on the Rhine. As winters have become warmer and wetter, there is a decreasing probability of precipitation falling as snow. This has been associated with a period of unprecedently high North Atlantic oscillation index values (based on the pressure differential between Iceland and Portugal), which is itself associated with higher North Atlantic sea-surface temperatures and known to show a strong correlation with precipitation across the whole of northern Europe. The impact of increased precipitation and raised flood risk has been particularly noticeable in Scotland. These climatological linkages increase the potential for seasonal forecasting of regional flood risk.

A key feature of all flood models is to employ very high resolution topographic and exposure data. Following the UK Midland floods in April 1998, in collaboration with Cambridge University, RMS set up Project Noah, to gather information directly from the thousands of businesses and households affected through questionnaires and interviews. This research had a number of goals including: to determine in great detail the drivers of flood losses, to quantify the impact of flood warnings on contents losses and to identify how insurers could control business interruption and additional living expenses.

One of the findings of the unit level storm surge model development was thedisparities in the relative risk of commercial versus residential portfolios. In the affected postcode sectors, about 10% of the residential exposure is at risk compared with roughly 50% of the commercial exposure.

This reflects land-development and the application of planning controls andis a very widespread feature of development around floodplains.

Despite (or sometimes because of) the improvement in flood defences, portfacilities and cargo remain at considerable risk from surges, as in 1976when the insured cargo losses in Hamburg exceeded the accompanying wind storm losses in Germany. In 1987J, a marine insurer suffered a major loss whne a shipment of Japanese cars was submerged by the surge on the quay-side in Oslo. All the major bays and inlets in the seas of northern Europe are prone to storm surge flooding, including the cities of Bristol, Cardiff, Belfast, Blackpool, Gothenberg, St Petersburg, Kiel and even the Douro estuary in Lisbon.Flood risk management solutions
For commercial and industrial facilities in Europe, flood is often the biggest source of potential catastrophe loss. Corporate risk managers and flood risk underwriters need to determine the level (and price) of the risk and prepare for it. Information about this risk is uneven: 100 year flood maps are themselves relatively crude, and recent events on a number of rivers (including the Oder, the Arno and the rivers of central England) have had return periods significantly “in excess of 100 years”, retrospectively determined on the twentieth century flow records. Both corporates and insurers also need to consider how the risk isaccumulating, because of concentrations of locations (or suppliers) that could be affected in the same storm surge or river flood event, even though the facilities may be located on different coasts or rivers, or even in different countries.

Of particular concern are liability implications of flooding, associated with the escape of effluent or chemicals. On 25 April 1998, at a mine in southern Spain, a tailings dam broke releasing 5 million m3 of highly toxic lead, copper zinc tailings that caused an estimated euros 9.4 million of damage to local agriculture and narrowly missed a major nature reserve.

Some types of floods – in particular storm surges and regional river floods – may have a significant lead-time, making it possible to take action to reduce losses. However, it individual risk managers need to foresee the risk, rather than await information from government advisories, in particular when catastrophic flooding may result from the failure of flood defences.

Over the next decade, increasing liberalisation of the European riskmarkets will mean that much more of the flood risk will be transferred tofinancial markets and insurers rather than left to individuals and governments.

Dr Robert Muir-Wood is international technical director of Risk Management Solutions (RMS). Tel: +44 (0) 207 256 3800. E-mail: robert.muir-wood@riskinc.com.