A river confined between two levees rarely overflows due to additional rain falling upstream or unusual snowmelt. What increases downstream is the river itself: the same mass of water, compressed into a channel too narrow for it, and returned downstream faster and higher than it would have been without protection. This idea was first articulated by an American engineer, Charles Ellet Jr., in a report to the federal government in 1852. He warned that confining the Mississippi River in a narrow channel caused the water to rise higher and flow faster. A century and a half later, the mechanics behind this phenomenon remain unchanged, though the scale has expanded to cover entire continents. The reasoning of 19th-century engineers was summarized in one sentence: confining a river forces it to dig its own bed and evacuate floods more effectively. However, this acceleration does not stop at the boundary of the next county. When rivers rise, they can no longer spread naturally in the floodplain as they did before the protective works; they flow more strongly and faster, sending more water downstream. A study published in 2018 attempted to precisely measure this effect by cross-referencing river sediments and tree rings over five centuries. The result surprised historical defenders of the levee system: approximately 75% of the increase in the magnitude of century-scale floods on the lower Mississippi over the past five centuries would be attributable to the river engineering itself. This phenomenon has been fueled by spectacular engineering decisions. In the 1930s and 1940s, dredges corrected the river's course, shortening it by 150 miles, so that floodwaters could flow faster downstream. A shorter river means a steeper slope, hence a faster current, hence a flood wave that arrives more abruptly for downstream residents. The great flood of 1927 forced a first reconsideration. The American Congress then imposed a response at the scale of the entire basin. After the great flood of 1927, the Congress required that the Army Corps of Engineers build a massive system of levees and dams on the lower Mississippi. A century later, these structures still shape the agricultural territory of the Midwest, and their logic remains the same: protect a section by accelerating the flow toward the next. Today, levees exist in one out of every five American counties. As many segments that, at their own scale, reproduce the same transfer of risk downstream. On the other side of the Atlantic, an entire country eventually drew the opposite conclusion from the same observation. The Netherlands have spent a thousand years raising their levees along the Rhine, Meuse, Waal, and IJssel. Then two floods in succession changed the course. In 1993 and 1995, floods threatened to devastate the regions around the delta, prompting the evacuation of more than 200,000 people in the surrounding areas, without any levee failing. The country had narrowly avoided catastrophe without even experiencing a breach of structures, a signal considered strong enough to question the all-concrete strategy. The Dutch government then approved a change in doctrine. In 2006, it validated the Room for the River program, involving the state, provinces, water authorities, and municipalities; the program took place between 2007 and 2018 and resulted in 34 projects spread across the country's four main rivers. The total budget exceeded 2.2 billion euros. The main idea broke with a century of technical reflexes: instead of raising levees further, they were moved back to give space back to the main riverbed. The reasoning of Dutch engineers was based on an evidence long overlooked elsewhere. Continuing to raise levees was no longer feasible, as higher levees exposed disproportionate risks in the event of a breach. Retreating a levee, moving a village. In Nijmegen, on the Waal, the theory met concrete challenges. The most controversial project in the program involved moving the levee inland on a bank already inhabited. The relocation of the levee in Nijmegen led to the demolition of about fifty houses and the disappearance of several businesses in the village of Lent. The construction site redesigned 120 hectares of floodplain for a cost of 126 million euros. An undertaking of this magnitude, decided to protect a city, disrupted another just a few hundred meters away. The scientists who evaluated the device agree on a rare point in the field of river engineering: the gamble worked rather well. The intervention ultimately reduced maximum flood levels by 35 centimeters, 8 centimeters more than anticipated. Translated to the scale of the delta, the benefit protects a considerable population. Thanks to the various Room for the River projects, four million people living in the Dutch delta are now better protected. The contrast between the two sides of the Atlantic ultimately rests on a question of timing rather than climate. Americans built their levee system on a confinement doctrine thought up in the 19th century and never really abandoned since, despite the early warnings from the engineers themselves. The Dutch, faced with the same dilemma a century later, had the time to observe the limits of the model in their neighbors before applying it to themselves. The program now continues under an expanded version, called Room for the River 2.0, which incorporates freshwater safety and the quality of the drinking water supply alongside the sole protection against floods.