by Philip B. Williams, Ph.D., P.E.

The consequence of failure at the Three Gorges Dam would rank as history’s worst man-made disaster. More than 75 million people live downstream on an intensively cultivated floodplain that provides much of China’s food. It is therefore reasonable to expect that a key design criterion for the project is ensuring that the risk of failure is kept extremely low.

Because of the limited operating experience with large dam projects of this type, and the disquieting number of safety incidents that have threatened the integrity of large dams in the last two decades, it is reasonable to expect that CYJV would use the best state-of-the-art techniques to demonstrate that the design, construction, operation and decommissioning of the project would keep the risk of failure acceptably low.

Unfortunately, CYJV does not address the safety issue either systematically or coherently. It provides no acceptable risk criteria, no mapping of the area and population at risk, no comprehensive risk assessment which identifies all the potential failure modes, and no identification of fail-safe measures. Because safety is not analyzed as a discrete topic, major failure mechanisms and combinations of failure mechanisms are ignored. There are many such possibilities; for example, a reservoir-induced earthquake that initiates new landslides close to the dam; sabotage or military action that disables spillway gates immediately before the flood season; unanticipated delays in construction leading to the overtopping and washing out of one of the cofferdams.*

Although CYJV discusses some safety issues, it makes many major assumptions and gross underestimations about the dam’s design which effectively put the probable risk of dam failure greater than the risk of a 1000-year flood for which the project is designed. Examples of the flaws in CYJV’s analysis are as follows:

• Underestimation of earthquake ground accelerations

  One of the most important structural design criteria for a dam is the estimation of ground acceleration in the event of what is termed the maximum credible earthquake (MCE). For the Three Gorges design, a 6.5 magnitude earthquake occurring on a fault 17 kilometres away is used for structural analysis. There are substantial uncertainties in the selection of the MCE and also in the prediction of ground accelerations at various distances from the fault. CYJV uses a ground acceleration factor only one-third the value that would be used in a reasonably prudent design.¹ CYJV’s use of these values effectively results in the most optimistic interpretation possible of likely ground accelerations due to earthquakes.

• Inadequate analysis of reservoir-induced seismicity

  CYJV recognizes that the weight of the water in a large reservoir can initiate earthquakes. However, in developing ground acceleration design criteria, CYJV uses only historical records of earthquakes, which means that the design accelerations selected are likely to be too low and/or would occur more frequently than expected. In addition, there appears to be substantial uncertainty about the movement of the most important of these faults, since CYJV stresses the “need for careful assessment”² – an assessment that presumably has not yet been undertaken.

  The treatment of reservoir-induced seismicity (RIS)³ is cursory and does not acknowledge the serious potential for structural damage, property damage and loss of life downstream that could occur. CYJV assumes that RIS occurs only on faults that are presently proven active, and implies that only short lengths of long faults close to the dam site would be activated. The length of faults passing under the dam itself and the displacement that would occur if these were activated are not identified. Therefore, it appears that the dam design is based on the optimistic assumption that no movement would occur on these faults, despite the experience with RIS elsewhere. For example, the Koyna Dam in India initiated an earthquake (approximately 6.0 in magnitude) that seriously damaged the dam and killed 200 people in an area that had not previously been seismically active.

• Inadequate analysis of structural stability

  Apart from the optimistic estimates of ground acceleration during earthquakes and the fact that potential RIS is downplayed, it is clear that there are substantial unresolved problems related to the structural design of the dam which, if satisfactorily resolved, could add hundreds of millions of dollars to the project cost.

  For example, with higher, more realistic assumptions for ground acceleration, the upstream face of the dam would be subject to stresses which would almost certainly cause cracking. And while CYJV recognizes that this would occur, it did not conduct the necessary analysis of the dam to identify where cracking could occur and what design modifications are needed.

  Furthermore, CYJV failed to conduct a comprehensive assessment of project operation management to analyze possible failure modes such as rupture of the dam due to fault movement underneath it, and the performance of the dam during an earthquake with prior cracking.

Other examples of how CYJV has systematically downplayed the risk of dam failure are as follows:

• Underestimation of the risks caused by catastrophic landslides

  In 1963, at the Vaiont Dam in Italy, a landslide in the reservoir generated a flood wave that killed 4000 people. In the Three Gorges region, major landslides occur every few years, disrupting navigation and causing property damage and loss of life. CYJV states that the Three Gorges Project would result in “no significant change in slope stability,”⁴ which is highly questionable considering that wide fluctuations in reservoir levels in the Three Gorges region are highly likely to have a destabilizing effect on potential slide areas.

  Although CYJV discusses the risk of landslides in the reservoir, it did not investigate the effect of earthquakes, including those induced by the reservoir itself, on activating landslides in areas it has rated as stable. Nor did it evaluate the impact of landslide waves on spillway gates at the time of rapid drawdown* in the reservoir immediately prior to the flood season.

  Finally, CYJV did not conduct a systematic analysis of zones at risk from waves 20 to 50 metres high that could result from individual landslides and could conceivably kill tens of thousands of people living near the reservoir. Therefore, it appears that the threat to people living around the reservoir and dowstream, and the threat to safe operation of the dam, have been greatly underestimated.

  Apart from other optimistic assumptions, CYJV’s discussion of impacts due to landslide-generated flood waves assumes that all people living in the reservoir area would be relocated above the 182-metre elevation mark, contrary to the resettlement plans stating that only people living below the 162-metre elevation mark would be relocated.

• Underestimation of risk of spillway failure

  As the world’s largest hydroelectric dam on the world’s third longest river, the Three Gorges Project incorporates many experimental technological innovations. One such experiment is the construction of the world’s largest submerged spillway bays. Each of the 27 spillway units has a capacity equivalent to the average flow of the Missouri River in the United States. CYJV confidently asserts “there is no reason to believe that these structures could not be successfully designed, constructed and operated,” even though the discharge per unit width is “well beyond proven world experience.” CYJV’s confidence in the spillways is further undercut in the same paragraph with the statement: “The feasibility of such a high unit discharge should be reviewed during final design.”⁵

  In fact, operating experience with extremely large flows through such spillways has not been good. At the Tarbela Dam (Pakistan), and the Glen Canyon and Hoover Dams (U.S.), extremely high velocities and pressures caused cavitation* and erosion which threatened the structural integrity of the dam and necessitated serious and costly repairs. Similarly, at the Three Gorges Dam there would be a high possibility of failure.

  Another questionable assumption is that the “good and homogeneous quality” of the rock immediately downstream of the dam would minimize scouring (erosion of the channel caused by the river’s flow).⁶ Actual operating experience with this is very limited, but there is a significant possibility that scouring could threaten the structural integrity of the dam, as nearly occurred at the Tarbela Dam. Once scouring begins it is very difficult to correct and requires continual remedial measures which can add significantly to operating costs.

• Failure to consider downstream effects of cofferdam failure

  During the construction of the project a series of temporary cofferdams would be constructed across the river in order to divert its flow. The second and third phase cofferdams would, at best, be capable of withstanding a 100-year flood and a 200-year flood, respectively. If larger floods occur, these cofferdams could quickly wash out,* releasing a flood wave that would overwhelm the Gezhouba Dam** and continue downstream to overtop the Jingjiang Dyke, drowning hundreds of thousands of people. CYJV estimates the probability of such a catastrophe to be about 1 in 20, which should be considered an unacceptably high risk.

• No provision for decommissioning of the dam

  The risk of dam failure increases with its age as construction materials deteriorate, mechanical systems such as spillway outlet gates fail, and the effects of a series of problems, such as corrosion, abrasion, sedimentation, and downstream scouring, become intractable.

  CYJV has calculated the costs and benefits of the project over a 50-year period (for comparison’s sake, Chinese culture has developed alongside the Yangtze River over some 4000 years). Regardless of the dam’s economic lifespan, CYJV should have made provision in the feasibility study for decommissioning the project in a way that would ensure the safety of those living downstream. The costs of decommissioning should have been included in the cost-benefit analsyis.

Sources and Further Commentary

*A cofferdam is a temporary dam built across the river to divert the river’s flow around the dam site while the permanent dam structure is under construction. Plans include construction of three cofferdams during three stages of the Three Gorges Project construction.

*Drawdown refers to the release of water to lower the level of water in the reservoir.

*Cavitation: extremely high velocities cause negative pressure which can break off pieces of the spillway’s concrete surface.

*On a much smaller scale, such a cofferdam failure occurred in 1986 at the Auburn Dam site on the American River (U.S.). The cofferdam was washed out by a flood flow, only one-fifth of the volume that would be released from the Three Gorges cofferdam, but fortunately a disaster was averted by the Folsom reservoir downstream which was able to contain the flood.

**The Gezhouba reservoir capacity is less than one-half of a cubic kilometre (1/2 km3), a mere fraction of the Three Gorges reservoir capacity.