The theory sounded like common sense, which ought to have been the first warning. Confine the river between levees, force all that water into a narrower channel, and the increased velocity scours the riverbed deeper. A deeper channel carries more water. The river digs its own solution.

In 1861, Captain Andrew Atkinson Humphreys and Lieutenant Henry Abbot published this argument in a report nearly six hundred pages long, with a hundred pages of gauge registers, soundings, velocities, and computed cross-sections. They claimed to have ascertained "every important fact connected with the various physical conditions of the river." After four hundred pages of data, they arrived at their conclusion: the Mississippi could be controlled with levees alone, from Cape Girardeau to the Gulf. The report was so thorough that questioning it felt like questioning arithmetic. Nobody enjoys questioning arithmetic.

There was a man who questioned it anyway. Charles Ellet, Jr., a civilian engineer, had warned nine years earlier that levees would make floods worse. His logic was plain enough: confining water that formerly spread across thousands of square miles of lowland meant it was "compelled to rise higher and flow faster." Ellet proposed reservoirs and widened bayous to intercept floodwaters upstream. Humphreys rejected this on grounds that now read as competitive rather than scientific. The Corps sided with Humphreys. By 1885, the Mississippi River Commission had adopted levees-only as official doctrine and begun sealing the natural outlets that existed, plugging the very escape valves the river had spent millennia carving for itself.

What happened next was observable to anyone standing on the riverbank, which may be why nobody at the Commission observed it. The levees did not scour the riverbed deeper. The river rose. Levees built to seven feet had to be raised to twelve, then twenty, then thirty-eight feet. Each increase in height increased the volume of water contained, which increased the pressure, which required higher levees. The Commission's response to each flood was to blame substandard construction. They never questioned the policy. Ellet's prediction accumulated in plain sight for four decades while the walls grew taller and the men who built them grew more certain.

By October 1926, the gauge at Vicksburg read over forty feet at a time of year when it usually hovered near zero. Readings across the Ohio, Missouri, and Mississippi were the highest ever recorded. That same year, the Army Corps assured the public the levees could prevent any future flooding. Through the winter and into spring of 1927, the river passed the highest stage ever known, 52.3 feet reached in 1922, and kept climbing. A foot a day. The Commission's assurance stood while the river rose past every number that was supposed to be impossible.

At Mounds Landing, twelve miles north of Greenville, Mississippi, the levee sat where the river turned ninety degrees. The current hammered the earthen wall like something trying to get out. Sand boils erupted at its base. The levee shook. Hundreds of men worked through the night of April 20, piling sandbags as water ran over the top. On the morning of April 21, sometime between 6:30 and 8:00 a.m. (sources disagree on the precise moment), the levee gave way.

The crevasse tore open roughly three-quarters of a mile wide and over a hundred feet deep. The volume of water pouring through exceeded Niagara Falls. It uprooted trees, knocked buildings flat, washed workers into the torrent. The district engineer wired Washington: "Crevasse will overflow the entire Mississippi Delta."

The Delta is flat. That is the fact that turns a breach into an erasure. Nothing stood between the broken levee and the Yazoo hills to slow or stop the water. It poured across open cotton land with nothing to resist it. A thirty-foot wall of muddy river swallowed plantation houses whole. Even Egypt Ridge, so named because no flood had ever reached it, went under. Where farms and towns had been, it looked like an ocean.

In Greenville, the fire siren blew. A resident who had been buying lumber for a boat when the alarm sounded watched six-foot currents run through the streets by nightfall. That night a phone call came: the water was over the counter in Payne's store at Winterville. "We had figured on four feet of water," the resident wrote, "but decided we had better prepare for eight feet." At eleven the siren blew again and the people went crazy. In Yazoo City, seventy-five miles away, floodwater covered rooftops. People stranded in treetops waited for boats, praying the trees held. Within ten days, nearly a million acres lay under ten feet of water. The river flowed through the crevasse for months.

That Greenville resident left behind a sentence that holds the whole history of the doctrine. Previous floods, he wrote, had been called "overflows," and they left fertile deposits on the land.

The difference between an overflow and a catastrophe was sixty-six years of engineering confidence.

The Corps never attempted to defend the doctrine again. The replacement system included the spillways and floodways Ellet had proposed seventy-five years earlier. The engineering knowledge hadn't changed. Ellet had it right in 1852. The river just finally ran out of patience with an institution that preferred its own data to the water lapping over the sandbags.

A 65-acre lake still sits where the levee broke. The river made its point, and the point is permanent. It is possible to be thorough, credentialed, and catastrophically wrong for decades, provided nobody with the authority to stop you has the nerve to say so. That particular lesson has a shelf life of about one generation before it needs learning again, and we are well past due.