Vajont’s message for the Three Gorges

In 1963, thousands in northern Italy lost their lives as villages were washed away by huge flooding caused by the Vajont dam. Risk expert David N Petley has studied the disaster, and considers what it means for China’s Three Gorges project.

As the sun went down on the evening of October 9, 1963, Longarone was a sleepy, traditional market town located in the Dolomite region of northern Italy. The next morning, the first rays of the sun illuminated a different sight – the town had been almost completely washed away, leaving a barren wasteland of boulders. The remains of just a few of the town’s buildings poked through, and buried in the sediment or washed tens of kilometres down the river were the bodies of 2,500 of the town’s inhabitants. The town had suffered the impact of Europe’s worst dam disaster; the flood had come from a new dam located in a valley above it.

However, as the sun rose the dam was still standing intact. And indeed it is still there today, for the disaster at Vajont was not the result of the collapse of the dam, but instead of a landslide that had crashed into the lake behind it. Now, with the director of China’s State Council Three Gorges Construction Committee warning about potential landslide disasters at the Three Gorges site, there are concerns that Vajont might provide a chilling warning for the famous dam on the Yangtze River.

The roots of the Vajont disaster can be traced to the rapid development that occurred in Italy in the aftermath of the Second World War. This led, in particular, to the growth of industry in the northern Italian cities of Milan and Turin, with the attendant increase in demand for electricity. The Vajont valley had been identified in the 1930s as an excellent location for a hydroelectric dam, having a large catchment with heavy rainfall, large amounts of snowmelt and a gorge-like morphology that meant that a comparatively small dam could hold back a large amount of water. Construction of the dam began in 1956 and was completed in 1960. At the time, the dam was considered to be a marvel of modern engineering, being the highest doubly curved arch dam, extending to 265.5 metres above the valley floor; the maximum volume of its reservoir was 115 million cubic metres. In comparison with the Three Gorges Dam, the scale of the barrage at Vajont was tiny. The potential for destruction, however, was huge.

The disaster

The first documented concerns about landslides at the dam site were expressed in 1958. As landslides were a well-known hazard during and after the construction of dams, it is perhaps surprising that this problem had not been considered earlier. Contrary to some of the reports about the Three Gorges Dam, the landslides associated with dams do not occur due to the weight of the water. Instead, the lake acts to increase the amount of water in the soils and rocks on the reservoir banks, which reduces their strength. It is this weakening that allows them to sometimes collapse into the lake. The biggest danger is that if such a landslide occurs rapidly then a wave can be generated that can overtop the dam. Large dams are not designed for this, and therefore the danger of the dam collapsing is high.

In 1959, before the dam was finished, some fairly limited investigations were undertaken of the potential banks of the new lake. A number of existing and potential landslides were identified, but the analyses of them suggested that the likelihood for large-scale movements of them was limited. As a result, no action was taken to fix them, but a very limited monitoring programme was set up.

The dam was completed in February 1960, and the filling of the lake began.  As at the Three Gorges, the first landslides were comparatively small, first occurring in March, shortly after the filling commenced. This slip was not considered to be a problem or a warning of any more serious problems. But concerns started to rise in October 1960, as a two-kilometre long crack opened on the south side reservoir bank, defining a huge landslide covering an area of about 1,700 metres in length and 1,000 metres wide. The landslide was moving continually, albeit at only three to four centimetres per day.  On November 4, when the reservoir was about 180 metres deep, a 700,000 cubic metre lump fell off the front of this larger landslide over a period of about 10 minutes. This event finally caught the attention of the site managers, but their response sowed the seeds of the subsequent disaster. They correctly identified the serious threat posed by the large landslide, but instead of abandoning the site, the site managers tried to get the landslide to slowly slide down the hill and into the lake, whereupon it would no longer be a hazard. Of course the lake would be partially filled, so a tunnel was constructed around the site to allow water to travel from the remaining upper part of the lake to the dam, allowing the continued generation of power.

For most of 1961, the lake level was kept deliberately low to allow construction of the tunnel. In October of that year the site was ready once again and filling was restarted. The plan was to slowly fill the level of the lake while the movement of the landslide was monitored.  The intention was that the landslide would slowly slip into the lake.  If the movement rate became too high the lake level would be dropped to slow the movement down. Thus the movement of the landslide would be controlled by varying the lake level. The site managers clearly felt that nature was within their control.

From October 1961 to November 1962, the level of the lake was slowly increased.  Late in 1962 the movement rate became too high, so the lake was partially emptied, whereupon the landslide effectively stopped.  Filling was then restarted in April 1963. By early September, the water depth was 245 metres. The rate of movement of the landslide slowly increased, and in October the lake started to be emptied again. Sadly, this was too little, too late.

At 10:38 pm on October 9, 1963, catastrophe stuck and the landslide collapsed. The entire mass, which weighed about the same 500 million standard-sized saloon cars, slid approximately 500 metres into the lake, accelerating at about the same speed as that of a racing car at the start of a race. The landslide slammed into the lake at about 110 kilometres per hour, blocking the gorge to a depth of 400 metres. A part of it then travelled 140 metres up the opposite bank, pushing a huge mound of water ahead of it. This wave crashed into the lower part of the village of Casso, 260 metres above lake level, washing the houses away instantly. The higher houses in the town were luckier, but most were severely damaged by a blast of air that was pushed ahead of the landslide.

But the worst was yet to come. The water displaced by the landslide swept down the lake and crashed into the dam. The volume of water was immense, about 30 million cubic metres swept over the top of the barrage. Evidence from the trees that were swept way by the flood suggests that this was about 245 metres higher then the crest of the dam, meaning that it was about half a kilometre higher than the floor of the valley. The huge flood crashed down onto the communities below without any warning, most of the victims died in their beds, unaware of the impending disaster. There were few survivors in the villages of Longarone, Pirago, Villanova, Rivalta and Fae.


Inevitably, many lessons were learned in the aftermath of the disaster. Since then, dam designers have become particularly careful to investigate the valley sides above any potential reservoir to ensure that there is no likelihood of catastrophic landslides. The banks of reservoirs are carefully monitored during and after the filling of the reservoir to ensure that there are no impending landslides, and any that are identified are monitored and sometimes repaired. This monitoring is a major task in even the smallest of reservoirs.

It is in this context that the recent announcements about the potential for landslides at the Three Gorges Dam site cause such concern. The parallels between the projects are notable. Construction of both dams occurred during a time of rapid economic development; both dams were considered to be modern marvels that pushed the limits of the available technologies. In both cases, landslides started to occur in the early stages of filling and became progressively worse. In both cases there appears to have been a belief, at times, that nature can be tamed. The geology of the two sites is also quite similar.

So what are the implications for the Three Gorges site? Most geologists believe that a landslide that caused a wave to overtop the dam, as at Vajont, is unlikely. Most potential landslides are very distant from the dam itself, meaning that by the time the wave reached the dam it would not be large enough to sweep over the top. However, such a landslide would have the capability to create a wave that was as high as that at Vajont, meaning that any communities close to the lake, in the vicinity of the landslide, would be severely threatened. Such a landslide would also be very hazardous to any boats or ships on the river over a distance of tens of kilometres. Perhaps most worrying, is the possibility that the landslide could create a river blockage that would allow a new, higher lake to form behind it. When the river reached the crest this blockage would almost certainly collapse, creating a wave that would be a most serious threat to the dam. It is clear that the authorities need to act quickly and with great care, using the best possible technologies and experts, to ensure the safety of the people around the site.

Professor David N. Petley holds the Wilson Chair in Hazard & Risk at the Department of Geography, Durham University.