The Dyke That Moved Backwards

On 30 January 1995 the province of Gelderland started telling people to leave. Over the following days roughly a quarter of a million residents of the Betuwe, the Land van Maas en Waal and the Bommelerwaard were evacuated, along with something close to a million cattle, pigs and chickens. The Rhine and the Meuse were both running at levels nobody had planned for, the dykes between them were saturated peat and clay, and no engineer in the water boards was willing to stand up and say they would hold.

They held. Nothing breached, nobody drowned, and the country was left with a week of televised anxiety, a bill for the livestock and a second warning in thirteen months — the Meuse had already flooded Limburg in December 1993. The immediate response was conventional: the Delta Act for the Large Rivers, passed the same year, drove dyke reinforcement through the planning system at emergency speed. The response that followed over the next decade was not conventional at all. The Dutch state concluded that the next answer could not be a higher wall.

Why the wall stopped working

A dyke is a device for narrowing a river. Water that would once have spread sideways across a floodplain is held in a channel instead, so it stands higher, moves faster, and arrives downstream sooner and in greater quantity. Every metre added in one place raises the level in the next, which is why river engineering between neighbours is a diplomatic subject and why the Rhine has had an international commission since the 1950s.

Two other things were moving the wrong way at the same time. The land behind the dykes keeps sinking — the western peat districts by several millimetres a year, because drained peat oxidises and leaves as carbon dioxide — while the river beds between their walls silt up. The gap between the water and the ground being protected from it widens every decade whether or not anyone touches anything. And the deeper the polder, the worse a breach becomes: about a quarter of the country lies below sea level and something like 60 per cent of it would flood if the defences were removed. The storm of 1 February 1953, which killed 1,836 people across Zeeland, South Holland and western North Brabant in a single night, is the standing demonstration of what the failure mode looks like.

Buying the floodplain back

The programme was called Ruimte voor de Rivier — Room for the River. Parliament approved it in 2006 and it ran from 2007 until it was declared complete in 2019: more than thirty separate interventions on the Waal, the Nederrijn, the IJssel and the Meuse, for a budget of around €2.3 billion. Dykes were moved back from the water. Side channels were dug, groynes lowered, floodplains excavated, summer dykes taken out and entire polders opened. The engineering target was a single number — the Rhine branches had to be able to carry 16,000 cubic metres a second at Lobith, where the river crosses the German border, instead of 15,000. The second objective, written into the decision alongside safety, was spatial quality: the works had to leave the places they happened in better than they found them. That is an unusual clause to put in a flood defence law, and it is most of the reason the programme is remembered fondly rather than resented.

A higher dyke does not remove the water. It moves the problem downstream and raises the price of being wrong.

The most conspicuous of the thirty is at Nijmegen, where the Waal turns hard and narrows, and where the water backed up in 1995 and 1993 before it. The fix was to shift the dyke on the north bank 350 metres inland, into the suburb of Lent, and cut a channel through the ground behind it. About fifty houses were bought out and demolished. What stands there now is a four-kilometre secondary channel, an island in the middle of the river called Veur-Lent, three new bridges and a long park along a waterfront the city did not previously have. It opened in 2016 and people swim in it. The households that were expropriated are a smaller and quieter constituency than the ones who got the park, which is the political arithmetic of the entire programme visible in one place.

The farms on the mounds

Sixty kilometres west, in the Overdiepse Polder on the Bergsche Maas, the answer was stranger. Seventeen farms worked the polder. The plan the government settled on was to lower the dyke on the river side so the land floods in an extreme year, perhaps once in a generation, and to raise eight artificial mounds inside it, each broad enough to carry a farmhouse, its yard and its animals. The mound is the oldest building type in the northern Netherlands, a thousand years older than dykes, and here it was reintroduced as current policy; the farmers put the idea forward themselves. Downstream at the Noordwaard, near Werkendam, roughly 4,500 hectares were simply given back. The word for that — ontpoldering, un-poldering — turned radioactive in the same years. The Hedwigepolder in Zeeland, promised to Flanders under a treaty signed in 2005 as compensation for deepening the Westerschelde, took some seventeen years of litigation and parliamentary revolt before the sea was finally let into it in 2022.

What it did not buy

The larger change was legal. On 1 January 2017 the Netherlands replaced the standard it had used since the Delta Works — a required height for each dyke ring, expressed as the water level it must survive — with one expressed as risk to people. Everyone living behind a primary defence is now entitled to a probability of dying in a flood no greater than one in 100,000 a year, with stricter figures where a breach would be catastrophic or where vital infrastructure sits. That reframing is what allows a lowered dyke and a deliberately flooded polder to count as an improvement: the thing being measured is no longer the wall, it is the consequence. A Delta Commissioner, a rolling Delta Programme and a ring-fenced Delta Fund of well over a billion euros a year keep the work funded past the electoral cycle, which is the other thing the Dutch have that most countries do not.

What none of it addresses is the ground. The peat districts keep sinking, fastest where they are drained hardest, and the damage arrives as cracked walls and rotting timber pile foundations rather than as floods. Drought turned out to be a failure mode of its own: on 26 August 2003, after a dry summer, a peat dyke at Wilnis south of Amsterdam slid sideways because it had lost the weight of its own water, and 1,500 people were evacuated from beside a canal their village had lived next to for centuries. Room for the River bought the rivers a few decimetres of headroom — thirty-four centimetres at Nijmegen, where it did the most — and it bought them well. It did not buy the century. The Delta Programme now plans against sea level scenarios that make the Rhine delta a harder problem than the one the Delta Works were built for, and the honest position in Dutch water policy — stated more plainly here than almost anywhere else — is that some of the answers are not known yet.