When Science Goes Wrong (19 page)

It so happened that the remaining intact segment of the dam carried a gauge whose function had been to measure the depth of water in the reservoir. The depth was continuously recorded on a drum that was rotated by a clockwork mechanism. When the commissioners inspected the drum, they found that reservoir level appeared to have begun falling at about 11.30pm on the evening of the disaster and had fallen by about four inches by the time of the collapse. This suggested that there was an initial small breach in the dam or its foundations that had gradually widened itself until the overlying section of the dam lost its support and suddenly collapsed into the breach.

According to the commissioners’ final report, the disaster was not caused by any shortcomings in the design of the dam itself, but by defects in its foundations – most particularly, by the weakness of the Sespe conglomerate, especially when wet. Because some of the segments from the right (western) side of the dam had been swept a long way downstream, the commissioners concluded that the initial break had occurred on that side, probably at or near the location where the San Francisquito Fault passed under the dam. Reservoir water had eaten away at the fault gouge, or the weak conglomerate near it, until the dam’s right abutment had collapsed. The torrent of water, curving around the centre of the dam in a giant eddy, had eroded the foundations of the dam’s left abutment until it, too, collapsed, which in turn provoked the landslide on the canyon’s western slope.

Given the nature of the rock of the Sespe Formation, the commissioners concluded that failure of the dam was inevitable, unless water could have been prevented from reaching the dam’s foundation. They mentioned several steps that could have been taken to slow down the entry of water. These included the construction of a deep cut-off wall (a concrete wall built deep into the foundations at the upstream face of the dam), pressure grouting of the foundations (to fill spaces that would have permitted the entry of water), the drilling of drainage wells to remove water that did enter the foundations, and the construction of inspection galleries within the dam that would have allowed the state of its foundations to be monitored. But these steps, they believed, would only have postponed the dam’s ultimate collapse.

In spite of Mulholland’s suspicions about sabotage, he evidently did accept the notion that inherent problems with the dam or its foundations were the more likely cause of the disaster. That raised an urgent problem, because the near-twin of the St. Francis Dam was holding back the now-full Hollywood Reservoir. And the reservoir lay not in some undeveloped canyon 40 miles from Los Angeles but immediately above residential neighbourhoods within the city itself. To ward off a repeat of the disaster, Mulholland ordered a lowering of the reservoir.

The theory put forward by the commission – that the weakness of the Sespe conglomerate and its poor response to wetting were the prime reasons for the dam’s failure – was widely accepted. It was used, for example, at the coroner’s inquest on the victims, as part of an (unsuccessful) attempt to bring criminal charges against Mulholland. Even today, there are experts who agree pretty much word for word with the commission’s report. Geologist Jack Green of California State University at Long Beach, for example, has a web page devoted to the disaster in which he states that the dam broke because water eroded through the dam’s foundations in the neighbourhood of the fault, just as the commission concluded.

Even in 1928, however, there were experts who voiced disagreement with the commission’s theory of the disaster. Carl Grunsky was a well-respected consulting engineer based in San Francisco who had been retained by local ranchers during the construction of the dam, and who therefore had good opportunity to study it. Two months after the disaster Grunsky, along with his son and a Stanford geologist, presented an account of what had happened that differed markedly from that of the official commission. They argued that the dam broke because it was subjected to a kind of pincer action. When the dam was filled, the Sespe conglomerate began to swell as water invaded it, so the right abutment attempted to push the dam leftward. Meanwhile, the foundations of the left abutment responded to the pressure of the dam by starting to slide. This incipient landslide was slight enough to be unnoticeable, but it pushed the dam rightward. These forces, building gradually over the months prior to the failure, caused cracks to develop that partially separated the centre of the dam from its two abutments. These cracks had been seen, but they were attributed to irregular contraction of the concrete during drying, and they did not cause any concern.

The drop in the level of the reservoir during the 30 to 40 minutes prior to the dam’s collapse, as recorded by the gauge, was an illusion, according to Grunsky. He calculated that for the drop to be real, such an enormous amount of water would have been flowing past the lower powerhouse that it could not have escaped notice, and the workers and their families would have had plenty of time to save themselves. In fact, the operator of the upper powerhouse, a man named Henry Silvey, had spoken by telephone with the duty officer at the lower powerhouse just 10 minutes before the dam collapsed, and he did not report anything amiss.

What, then, had caused the appearance of a drop in water level? Grunsky found what he considered a decisive clue: while inspecting the remaining standing block of the dam, he noticed a ladder that had become trapped and crushed within a horizontal crack in its up-stream face. The crack must have opened by several inches to allow the ladder to enter; in other words, the dam must have tilted upward and forward. According to Grunsky, it was this upward tilting of the dam, caused by the pincer action described above, that had raised the gauge slightly out of the water, producing an illusory record of a loss of water prior to the collapse.

A further clue relating to the possibility of an incipient landslide was reported by Charles Outland, the historian mentioned earlier. Thirty-four years after the dam collapsed, a member of the staff of the upper powerhouse (probably Henry Silvey, but he requested anonymity) told Outland that he had driven up the canyon on the evening of the disaster. As he passed the left abutment of the dam, he was disturbed to notice that the roadway had sagged downward by up to 12 inches, producing a scarp of that height that he had to negotiate with great care. Given that the road was cut into bedrock just above the dam’s left abutment, the presence of the scarp suggested that a large block of the canyon slope under the dam’s left abutment had begun a downward slide well before there was any other sign of the dam’s failure.

Around 1990, a new analysis of the disaster was undertaken by J David Rogers, then an engineering geologist in private practice in the Bay Area and Los Angeles. (Rogers is now an associate professor at the University of Missouri.) First, Rogers pointed out that the east wall of the canyon, at the point where the dam was built, had been the site of an enormous natural landslide in prehistoric times, in which a giant sector of the mica schist had slid and tilted down, ending up against the Sespe conglomerate and completely blocking the canyon. It was this episode that created the ancient lake that once filled the upper canyon. Then, at some later time, the rising waters breached this natural dam: the lake emptied, leaving the broad, sediment-filled valley that struck Mulholland as so suitable for a reservoir. Thus, the dam’s left abutment was built on rock that was inherently unstable and liable to experience another slide, especially when subjected to the pressure of the dam and the lubricating effect of water that entered the schist. Rogers agreed with Grunsky that an incipient slide in the rocky foundations of the left abutment was a key causative factor in the dam’s collapse.

Another factor on which Rogers placed a great deal of emphasis was what is called ‘hydrostatic uplift’. This has to do with that staple of junior school physics, Archimedes’ principle. The principle can be stated thus: a body immersed in a fluid experiences a buoyant force equal to the weight of the fluid displaced. This is why bodies that are less dense than water float at the water’s surface. But even bodies that are denser than water, such as those made of concrete, will experience a reduction in their apparent weight when they are immersed in water.

As long as a gravity dam rests on dry foundations, the full weight of the dam thrusts downward. If reservoir water enters the dam or its foundations, however, the dam is effectively immersed in the water. The dam then experiences a buoyant force equal to the hydrostatic pressure of the reservoir’s water column, and this upward force nullifies a significant portion of the dam’s weight. According to Rogers’s calculations, if the St. Francis Dam experienced full hydrostatic uplift, its apparent weight would have decreased by about 45 per cent. As a result, the thrust from the combination of the dam’s weight and the reservoir’s horizontal pressure would no longer have been directed into the bedrock within the middle third of the dam’s front-to-back extent, as called for in the dam’s design. Rather, it would be shifted 240ft downstream – well beyond the toe of the dam.

The theory of hydrostatic uplift, as applied to dam construction, was poorly understood in the 1920s – the textbooks on which Mulholland’s design engineers relied barely mentioned the topic. That is probably why Mulholland did not take effective precautions to prevent reservoir water from entering the dam’s foundations: he was concerned about such percolation only insofar as it might erode the foundations, not as a factor tending to lift the dam.

Mulholland did install a few relief wells in the dam to drain off any water that seeped under it. These wells were placed only under the relatively narrow central section of the dam, however – the portion that rested on the canyon floor. There were no relief wells in the dam’s abutments, which together formed the bulk of the dam’s width. Yet abutments are readily affected by uplift, because as they climb the sloping sides of the canyon, their height and weight decrease so that the stabilising downward thrust due to gravity becomes less.

Rogers
found a photograph of the dam, taken some months before the filling of the reservoir was complete, in which the line of contact between the left abutment and the Pelona Schist, on the downstream face of the dam, was clearly wet. This observation, combined with the leakage of water in that area noted by Mulholland on the morning before the failure, is evidence that reservoir water had not merely entered the mica schist but had percolated all the way through to the downstream face of the left abutment. This water was not only facilitating slippage of the ancient landslide but also was exerting a gradually increasing uplift on the abutment. This tended to separate the abutment from the more stable central section of the dam.

As the mica schist crept progressively downward during the days and hours before the collapse, the incipient landslide pressed against the dam. Because of the dam’s curvature, the abutments were oriented quite obliquely to the canyon walls. The pressure of the slide was therefore exerted primarily on the upstream face of the left abutment, putting the entire upstream face of the dam into tension. A short while before the final collapse, in Rogers’s interpretation, a fairly small block of concrete near the base of the left abutment fell out of the dam. (This block was later found farther downstream than any other block, suggesting that it was the first to yield.) Water poured through this orifice: it undercut the mica schist, and it also entered transverse cracks in the dam, causing the central portion of the dam to experience full hydrostatic uplift and therefore to tilt upward by a few inches. This was what caused the water gauge to record an apparent fall in the reservoir level.

Then, at 11.57:30pm – the time defined by the power outage – the schist was undercut to the point that a giant landslide occurred, involving about 900,000 tons of rock. The landslide destroyed the remaining part of the left abutment, but the material in the slide plugged the gap in the dam for a short time. Then water tore through the slide and the catastrophic emptying of the reservoir began.

Rogers
pointed to several observations that supported the idea of a landslide-induced breach in the left abutment as being the first event in the dam’s collapse, rather than a failure of the Sespe conglomerate under the
right
abutment as the governor’s commission had concluded. For one thing, after the disaster a line of debris was found on the western shore of the reservoir near the dam, and this debris extended for several feet above what had been the reservoir’s high-water level. Rogers concluded that this line of debris was thrown up by a large wave – a tsunami, essentially – generated by the landslide. For the debris to have been left above the high-water level, the reservoir must have been full, or very nearly so, at the time the landslide occurred. Furthermore, although both abutments eventually failed, the rocky foundations under the right abutment were scoured away much less than under the left foundations, even though the Sespe conglomerate was relatively weak. This indicated that the reservoir level was already relatively low when the right abutment breached. As a further sign that the right abutment failed late in the proceedings, the scour level along the west side of the canyon just downstream of the dam was much lower than it should have been if water from a full reservoir had been pouring through the right abutment. A final clue was offered by a 30ft-long pipe, which was attached to the underside of the water gauge and which ran down the upstream face of the dam. After the collapse, the pipe was visibly bent in the direction of the left abutment. This happened, according to Rogers, because while the reservoir level was nearly full, water was rushing toward the breach in the left abutment caused by the landslide.