On the morning of 5 June 1976, the Teton Dam — a brand-new 305-foot (93-metre) earthfill embankment built by the U.S. Bureau of Reclamation on the Teton River in eastern Idaho — eroded itself open from the inside and released roughly 250,000 acre-feet of water onto the towns of Sugar City, Rexburg and Wilford below. Eleven people died, some 25,000 were left homeless, around 16,000 head of livestock drowned, and property losses ran from hundreds of millions of dollars into the low billions. The reservoir had never been full before; the dam was failing the first time it was asked to hold its design load. The cause was not overtopping and not an earthquake. It was internal erosion — piping — driven by water that fractured the dam’s silt core and tunnelled through open joints in the rock the core was keyed into.
The Teton was a conventional zoned embankment with an impervious central core of wind-blown silt — loess-derived material that compacts well and seals beautifully, but that erodes readily once water moves through it and holds an open pipe without collapsing. That core was trenched down into the canyon’s foundation rock: a fractured, highly permeable volcanic rhyolite riddled with open joints. The job of sealing those joints fell to a grout curtain and to slush grouting along the key trench. That sealing was incomplete. The most heavily loaded interface on the site — the contact between an erodible silt core and a jointed rock that could carry concentrated seepage — was left as a path waiting for water.
The reservoir filled fast. Impounding effectively began in October 1975 behind the still-finishing dam, and through the spring of 1976 the level rose at roughly a foot a day, accelerating to about four feet per day by June as snowmelt poured in. Springs and small seeps appeared in the right abutment in the first days of June. On the morning of 5 June a clear leak turned muddy, then grew. By mid-morning a wet spot on the downstream face was discharging twenty to thirty cubic feet per second and a whirlpool was visible upstream. Crews drove bulldozers into the widening hole in a last attempt to plug it; the machines were swallowed. At about 11:57 the crest gave way and the reservoir emptied through the breach in a matter of hours.
Two federal inquiries — the Independent Panel impanelled by the Secretary of the Interior and the Governor of Idaho, and the Interior Department’s own Teton Dam Failure Review Group — reached the same family of conclusions. The dam failed by internal erosion of the silt core, most probably initiated by hydraulic fracturing of the key-trench fill and by seepage through unsealed joints in the rhyolite beneath the grout cap. The Teton became the United States’ canonical first-filling failure, the case that forced filters, controlled filling and independent review into the heart of American dam safety, and the disaster that ended Reclamation’s era of unquestioned authority over its own designs.
—
In the early hours of 8 August 1975, the Banqiao Dam — a clay-core earthfill embankment about 24 metres high on the Ru River in Zhumadian Prefecture, Henan Province, China — was overtopped and washed away after Typhoon Nina stalled over the catchment and dropped more than a year’s rainfall in a single day. The official Chinese count of those killed directly by the flood wave was around 26,000; estimates that add the subsequent epidemics and famine across the inundated plain range up to roughly 145,000 more, for a total commonly cited between 170,000 and 230,000. The structure did not fail because the embankment was poorly built. It failed because its spillway and sluice gates could not pass the flood, and the water simply rose over the crest and cut the dam to pieces.
Banqiao had been the showpiece of a flood-control system thrown up across the Huai River basin in the early 1950s with Soviet engineering assistance. After cracking and repairs in 1955–56 it was reinforced and nicknamed the “Iron Dam” — a name that came to stand for misplaced confidence. The dam’s discharge works comprised five sluice gates and an undersized secondary spillway, together rated to pass a flood far smaller than the one that arrived. The engineer Chen Xing had argued during planning for twelve outlet gates; his recommendation was judged excessively conservative and cut to five. The single decision that governed the disaster was made on paper, years before the rain fell.
The rain, when it came, was without precedent. Typhoon Nina collided with a cold front and parked over southern Henan from 5 to 7 August 1975. More than 1,000 millimetres of rain fell in twenty-four hours near the storm centre — more than the region’s entire annual average — and three-day totals exceeded 1,600 millimetres in places. The dam had been designed for a “thousand-year” flood of roughly 300 millimetres per day. Nina delivered something closer to a two-thousand-year event, more than twice the design level. With the gates and spillway swamped and partly blocked by sediment, the reservoir crested above the dam at about 117.94 metres above sea level and overtopped. Around 01:00 the embankment breached, and some 600 million cubic metres of stored water emptied in roughly six hours.
What made Banqiao the deadliest dam disaster in recorded history was not the single breach but the cascade. The same storm overwhelmed dozens of other reservoirs in the same basin, including the Shimantan Dam on the Hong River, the second-largest in the system, which failed about half an hour before Banqiao. In total some sixty-two dams collapsed, releasing a combined flood across roughly 12,000 square kilometres of densely populated plain inhabited by more than ten million people. The investigation, conducted internally and kept secret for thirty years until the files were declassified in 2005, found what the engineering already showed: the dams were hydrologically under-designed, their discharge capacity grossly inadequate, and the basin had been packed with reservoirs whose failures fed one another.
—
On the afternoon of 31 May 1889, the South Fork Dam — an earthfill embankment some 22 metres high and 284 metres long on the Little Conemaugh River, 14 miles above Johnstown, Pennsylvania — was overtopped by floodwater and eroded away in roughly an hour, releasing Lake Conemaugh and its 14.55 million cubic metres of water down the valley. The flood reached Johnstown about an hour later as a churning mass of water and debris and killed 2,209 people. It remains the deadliest dam failure in United States history. The embankment itself did not fail by piping, sliding or foundation defect; it failed because it had no margin left to pass a flood. A succession of owners had lowered its crest, removed its low-level discharge pipes, and screened its single spillway to keep fish in the lake. When extraordinary rain raised the reservoir, the water ran over the top and cut the dam open.
The dam was not new and was not the work of careless hands at the moment it broke. It had been built by the Commonwealth of Pennsylvania between roughly 1838 and 1853 as the Western Reservoir, a feeder for the state’s canal system, and was for its day a competent embankment with a cut-stone spillway and five cast-iron discharge pipes that let an operator draw the lake down at will. The railroad made the canal obsolete; the reservoir was abandoned, partially breached in 1862, and left derelict for nearly two decades. In 1879 the site was acquired by the South Fork Fishing and Hunting Club — a private retreat for Pittsburgh’s industrial elite, including Henry Clay Frick and, later, Andrew Carnegie — which rebuilt the embankment as a resort lake without a civil engineer in charge. Every modification the club made to suit a pleasure lake subtracted from the dam’s ability to survive a flood: the discharge pipes were never replaced, the crest was cut down by about a metre for a carriage road, and iron fish screens were fixed across the spillway. None of these acts was individually unthinkable; together they converted a sound embankment into one that could not pass its design storm.
The storm came on 30–31 May 1889, when six to ten inches of rain fell on the catchment in 24 hours. The lake rose toward the lowered crest, the debris-clogged spillway could not pass the inflow, and by early afternoon water was running over the embankment. Earthfill has almost no resistance to overtopping: the overflow cut a notch in the downstream face, the notch deepened and widened by headcut erosion, and the dam unzipped. The 1891 investigation by the American Society of Civil Engineers concluded the dam would have failed regardless of the club’s changes — a finding that protected the club from liability and that modern hydraulic analysis has since overturned. South Fork is the canonical American lesson that a dam is only as safe as its spillway and its freeboard, and that an embankment with no way to release water is a flood waiting for a date.
—