Vajont Dam — the Dam Survived, the Mountain Didn’t: a Slide That Killed ~2,000
Summary
On the night of 9 October 1963, the Vajont Dam — a 262-metre concrete double-curvature arch in the Vajont gorge above Longarone, in the Dolomite region of northern Italy — did not break. It stands today, almost undamaged. What failed was the mountain on the reservoir's left flank: roughly 270 million cubic metres of Monte Toc detached along an ancient slide plane and dropped into the lake in some forty-five seconds, displacing water that rose more than 250 metres up the opposite wall and overtopped the dam crest by about 250 metres. The resulting wave poured down the gorge and obliterated Longarone and the villages around it, killing close to two thousand people. The structure was sound. The reservoir slope was not, and no one had treated the slope as part of the structure.
The dam was the work of Carlo Semenza, among Italy's foremost dam engineers, and it was a feat of arch design — at completion in 1959 one of the tallest dams in the world, a thin double-curved shell anchored into the limestone walls of a deep gorge. The arch behaved exactly as designed; the post-event survey found damage confined to the topmost metre or so of the crest. The fatal element lay outside the engineers' frame of attention entirely. The left abutment of any arch dam carries reservoir thrust into rock, but at Vajont the rock above the waterline on Monte Toc was a stack of limestone interbedded with thin clay seams, dipping toward the gorge, and crossed by the buried floor of a prehistoric landslide. The reservoir was about to wet exactly the surface that had failed once before.
Filling proceeded in stages, and the mountain answered each rise in level. In November 1960 a minor slide of roughly 700,000 cubic metres slipped on the south bank, and an M-shaped tension crack opened along the slope, tracing the head of a far larger moving mass. From then on the operators tried to control the slide by manipulating the reservoir — raising and lowering the level to coax the rock to creep slowly and safely. In the final weeks the creep accelerated instead, from about a centimetre a week to nearly a metre a day, and on 9 October the mass went all at once.
Italy's criminal courts pursued the case for years. The L'Aquila Court of Appeal in 1970 and the Court of Cassation in 1971 confirmed that the disaster had been foreseeable and that the dam's managers were liable for what they knew, concealed and failed to act on. Vajont remains the textbook proof that a dam's reservoir is part of the dam, that an old landslide will move again when its failure plane is flooded, and that a reservoir slope must be investigated to the same standard as the foundation. It changed how dam sites are surveyed and made slope stability around the rim a mandatory subject of dam safety.
---
Timeline
The Build: A Flawless Arch Above a Flawed Reservoir
Vajont was conceived as a thin double-curvature arch — curved in both plan and section — wedged into the steep limestone walls of a deep gorge to store water for hydroelectric power. Its designer, Carlo Semenza, brought it to a final height of 262 metres, with a crest chord of only about 160 metres and a reservoir of roughly 169 million cubic metres. At completion in 1959 it was one of the tallest dams in the world and, as a piece of structural engineering, a near-perfect object. When the catastrophe came, the arch took forces vastly beyond any design load and survived with damage limited to the topmost metre of the crest. By the only measure engineers of the day applied to a dam, Vajont passed.
That is precisely the trap. An arch dam is designed against the water it impounds and the rock that receives its thrust. The slope rising above the waterline, beyond the abutments, was treated as scenery rather than structure. At Vajont the slope was Monte Toc, and its rock was a layered sequence of limestone interbedded with thin, weak clay horizons that dipped down toward the axis of the gorge. Worse, the mountainside contained the buried floor of a prehistoric landslide — a vast surface that had already failed in geological time and lay waiting to be reactivated. Carlo Semenza's son, the geologist Edoardo Semenza, working with the engineer Leopold Müller, identified this ancient slide before the reservoir was full. The recognition existed. It did not change the decision to fill.
The mechanism, in hindsight, was simple and known. A clay-lined bedding plane already at the edge of stability holds only by friction and by whatever resistance its toe provides. Fill a reservoir against the toe of that slope and two things happen: the water pressure in the clay seams rises, cutting the friction that held the mass, and the buoyant uplift of the rising lake reduces the effective weight pinning the slide in place. The reservoir, in other words, was a machine purpose-built to lubricate and unload the exact surface that had failed before. The arch was magnificent. The lake behind it was a loaded trigger.
The Failure Sequence: A Mountain Coaxed, Then Lost
The slope began answering the reservoir almost immediately. On 4 November 1960, with the water near 170 metres deep, about 700,000 cubic metres slipped on the south bank, and a continuous tension crack roughly two kilometres long opened across the hillside in an M-shaped plan. That crack was not a local cosmetic feature; it was the head scarp of a single coherent mass, perhaps a thousand times larger, beginning to detach. From that point the operators understood they were managing a moving mountain.
Their strategy was to control the slide through the reservoir itself — raising and lowering the level to govern how fast the mass crept, on the theory that a slow, ductile creep could be tolerated and a sudden slip avoided. Survey benchmarks tracked the slope's downhill movement continuously. For a time the rock obliged, moving by millimetres and centimetres, and the readings encouraged the belief that the slide could be metered out gently. This was the central misjudgement. A reactivated paleo-slide on a clay surface does not creep its way to a safe stop; it accumulates displacement until the geometry and pore pressure reach a threshold, then releases the entire stored mass at once.
The threshold arrived in autumn 1963. As the reservoir was taken to its highest level, the slope's creep accelerated — from about a centimetre a week earlier in the program to figures approaching a metre a day in the final days. From 26 September the operators tried to draw the reservoir down to brake the slide, but the rock was now beyond their control, moving faster than the lake could be lowered. At 22:39 on 9 October 1963 the whole mass — roughly 270 million cubic metres, a slab some two kilometres across — broke loose along the ancient surface and slid into the reservoir at speeds reaching about 110 km/h, filling the lake in roughly forty-five seconds. The water had nowhere to go but up and over. It surged more than 250 metres up the opposite wall and poured over the dam crest by about 250 metres of head. The arch held. The wave that overtopped it fell into the Piave valley and struck Longarone with a force that killed roughly nine in ten of its people. Pirago, Rivalta, Villanova and Faè ceased to exist. Erto and Casso, on the lake itself, were devastated. Close to two thousand people died in minutes.
The Reckoning: A Foreseeable Catastrophe, Confirmed in Court
Investigation and prosecution ran for years and reached a verdict sharply different from the technical exoneration handed down at Malpasset. Italian courts indicted eleven men for multiple manslaughter. The first-instance court at L'Aquila in 1969 accepted only limited responsibility, convicting a few defendants for inadequate warning. But on 3 October 1970 the L'Aquila Court of Appeal overturned that reasoning and confirmed the disaster's predictability, holding the dam's managers liable for their acts and omissions both before and after the handover from SADE to ENEL. In 1971 the Court of Cassation affirmed the finding and named engineer Alberico Biadene and ministry official Francesco Sensidoni as principally responsible. The slide had been recognised in advance; the M-crack had been mapped; the accelerating creep had been measured; the warnings had been minimised and the reservoir filled anyway. One defendant, Mario Pancini, took his own life in 1968 before judgment.
The engineering conclusion was equally clear and far more durable than the sentences. Vajont demonstrated that a dam's reservoir is part of the dam. The arch had been competently designed and competently built; it failed at nothing it was asked to do. The disaster lived entirely in the reservoir slope — an ancient landslide whose clay-lined failure surface was flooded, lubricated and unloaded by the rising lake until it reactivated and moved as a single block. The attempt to manage that slide by reservoir manipulation rested on a false model of how such a mass behaves: it does not yield gradually to a safe resting state but stores displacement and releases it catastrophically. From Vajont onward, the geology of the reservoir rim — and above all the search for old slides that a lake will reawaken — became a mandatory part of dam site investigation rather than an afterthought beyond the abutments.
---
Contributing Factors
Aftermath
Vajont killed roughly 1,900 to 2,500 people — an official 1,917, some counts near 2,000 to 2,056 — and erased Longarone and the villages below the gorge while leaving the dam itself standing. The arch survives today, a 262-metre wall above an empty gorge, kept as a monument and a place of pilgrimage. The technical legacy was a permanent change in what a dam investigation must cover. Vajont established that the reservoir is part of the structure: that the entire rim of the future lake must be surveyed for slope instability, that old and prehistoric landslides must be hunted out and assumed capable of reactivation once their failure plane is flooded, and that slope-stability analysis under reservoir loading and drawdown is a core dam-safety task, not a peripheral one. It made reservoir-induced landslides and the rapid-drawdown and rising-pore-pressure mechanisms that drive them standard subjects of engineering geology and embankment design, studied through Kiersch's contemporary account and the case literature that followed. The disaster passed into the international canon through ICOLD and the dam-safety community. Today Vajont is the byword for a single hard lesson: a dam can be perfect and still kill thousands if the mountain holding its lake is not.
---
Lessons
- Investigate the reservoir slopes to the same standard as the foundation: the failure can begin in the rim of the lake, not the dam — map the whole impoundment for old slides, weak bedding and clay seams that a rising reservoir will reawaken.
- Assume any prehistoric landslide will move again once its failure surface is flooded; a slope that failed once is pre-cut along a plane that the reservoir's pore pressure and buoyant uplift will reactivate, not a fresh intact mass.
- Do not treat a large slide as something you can meter out as safe creep: a clay-lined reactivated mass stores displacement and releases it catastrophically, so reservoir-level juggling controls a mechanism that does not permit gradual control.
- Treat accelerating creep as a final alarm, not a trend to watch: when displacement climbs from centimetres a week toward a metre a day, the time for drawdown has already passed — evacuate, do not manage.
- Heed the geologist before the lake, not the lawyer after it; when a recognised hazard is minimised to keep a schedule, the warning that was filled over becomes the finding that convicts. ---