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

Background

For centuries the Chinese people have been building earthen dykes to prevent the Yangtze River from overtopping its banks during the flood season and inundating the adjacent floodplain. Over the past four decades, the Chinese government has mobilized the people of the Yangtze Valley to substantially upgrade the existing system of flood protection dykes and diversion works. Flood storage reservoirs and control structures have been constructed, overflow diversion areas have been established, and roughly 30,000 kilometres of dykes now line the Yangtze and its tributaries, and protect urban centres situated in flood-prone areas. The main dykes protecting the valley have been reinforced and raised to an average height of 12 metres and along the critical 180-kilometre Jinjiang reach, the dykes stand 16 metres high in places. Today the Yangtze Valley is capable of safely diverting and storing roughly half of a flood equivalent to that which occurred in 1954.

Ironically, as the dykes have been raised to increase the river’s channel capacity, the risk of a flood disaster has grown dangerously high. If the Jinjiang Dyke – the main dyke in the middle reach which Yangtze Valley Planning Office engineers view as the critical component of the existing system – were to fail, a huge flood would spread across the densely populated floodplain, killing at least 100,000 people, and inundating major urban centres. Conceivably, the river might change course entirely and rush headlong into the city of Wuhan with a population of six million people.

The Existing Flood Management System

The fertile floodplain of the middle Yangtze Valley has been formed over hundreds of thousands of years by sediments deposited from the flood waters of the Yangtze and its major tributaries. Over the last two thousand years, as the population of the valley grew and farming became more intensive, a sophisticated flood management system was constructed to reduce the risk of flooding. This system has been substantially improved in the last 40 years and now consists of a network of river dykes, overflow diversion areas, control weirs, floodplain lakes and ring dykes. The guiding philosophy in the management of this system has been to attempt to absorb the huge volumes of floodwater by temporarily storing them in emergency diversion areas on the floodplain and in floodplain lakes rather than attempt to ‘control’ all floods.

With this system, three types of flooding can occur:

• Along the river itself, between the main river dykes, there are settlements and cultivated land. Referred to as ‘beach areas’ in the CYJV report, they are susceptible to flooding from frequent high river stages. Here people know they are likely to be flooded and therefore take reasonable precautions to limit flood damage. According to CYJV, flood damages can occur in beach areas if the flood exceeds the expected 5-year flood level.

   

• In the overflow diversion areas, covering about 800 square kilometres, about 4 million people live and farm. The use of these overflow areas to store flood waters is infrequent, and according to CYJV, occurs about once in 20 years. Although the residents of the overflow diversion areas know that they live in flood overflow areas, the level of protection varies. Many of the towns are protected by ring dykes and some areas have constructed refuges protected by dykes. Other areas have little protection.

   

• Former floodplain areas protected by major river dykes have virtually eliminated flooding. Referred to as ‘primary protected areas’ in the CYJV report, these floodplains have been intensively developed. Of particular importance is the Jingbei Plain protected by the Jianjiang Dyke against flood levels expected only once every 250 years (or a 1:250 flood probability). The Jingbei Plain and major cities, such as Zhijiang, Shashi, and Wuhan, depend on maintaining the integrity of the 12 to 16-metre-high Jianjiang Dyke during extreme floods. The integrity can be threatened by seepage, erosion, and overtopping.

   

This flood management system has functioned well, significantly limiting flood damages in the worst flood of the century, the 1954 flood, which CYJV estimates to be a once in 200-year event. The flood management system was successful because of its ability to temporarily store huge volumes of water during the flood. The 1954 flood is estimated to have discharged 366 cubic kilometres of water in 60 days. At least 100 cubic kilometres were probably stored in floodplain lakes, overflow diversion areas, and in river channels.

This puts into perspective the comparatively minor role the Three Gorges Project would have on reducing flood volumes. If designed safely, and operated correctly, the dam is intended to store 31 cubic kilometres which is less then 10 percent of the 387 cubic kilometre volume of the 1000-year flood.

Review of the CYJV Flood Control Analysis

Although most of the economic benefits claimed for the Three Gorges Project are for electricity generation, the dam’s advocates have maintained that flood protection is the primary need for the project. In fact, it has been frequently stated that the dam must be built to avert catastrophic flooding affecting millions of people living along the middle reaches of the Yangtze River. For a project of this magnitude – one of the world’s largest civil engineering projects that could potentially affect hundreds of millions of people – the CYJV feasibility study should provide an objective and systematic analysis of the reduction in flood risk which would result if the dam were built, and evidence that the project would be the most cost-effective approach for providing additional flood protection. Moreover, such an analysis should be accurate, use the best available techniques and have a consistent methodology.

Unfortunately, the study does not do this, for the following reasons:

Purpose of the Project Is Misrepresented

Throughout the feasibility study there appears to be confusion over the main flood control goal. The study summary, which is written for potential financiers and key policy makers, states that the project’s main flood control function is to protect 20,000 square kilometres of downstream floodplain inhabited by 10 million people. Design Volume 4, on the other hand, states that the project “can provide flood protection to approximately 1 million square kilometres.”¹ And then Volume 7 on flood control states that the “primary flood control objective of the project and its operation must be to guarantee the integrity of the major dykes protecting the Jingbei Plain and the Jingnan Region” – an area of 6700 square kilo-metres.²

The computed flood benefits* described in the detailed technical analysis do not substantiate even the least grandiose of these statements. According to CYJV, in the event of a 1000-year flood, the Three Gorges Dam would protect 4000 square kilometres from floods, which is only 30 percent of the total area that would be inundated without the project. But in the benefit analysis, even these figures are misleading. About three-quarters of the benefits claimed for the project are attributable to the reduction in the frequency, and not the elimination, of flooding of just three areas – two flood overflow diversion areas and the beach areas on the river side of the dykes. Only about 15 percent of the total economic benefits are attributable to protecting the highly populated Jingbei Plain and Jingnan Region, but the flood control storage and operation at Three Gorges is designed specifically for this 15 percent.

Only a detailed reading of the study reveals that the goals and operational design of the project are to upgrade protection for the Jingbei and Jingnan regions from the current 250-year flood level to 1000-year flood level. This means that the flood control operational design is not based on an objective cost-benefit analysis, but on a subjective judgement that these areas require levels of protection higher than the 250-year flood.

Unrealistic Operational Criteria

CYJV has recommended an operational plan for the dam – CYJV’s Flood Control Operation – to release or store floodwater in the reservoir depending on water levels in the river downstream:

• During flood season, reservoir levels would be lowered from the normal pool level (NPL) of 160 metres to the flood control level (FCL) of 140 metres above sea level. Flood storage would be provided above 140 metres and moderate floods up to the 50-year flood level could be stored without the reservoir exceeding the NPL. For greater floods, the reservoir would rise higher depending on the magnitude of the inflows and the water level in the main channel downstream.

   

• In the event of a 100-year flood, water would be stored in the reservoir so that the water levels in the middle reach (at Shashi) would not exceed 44.5 metres. As such, diversion into the Jinjiang diversion area would be avoided – a practice which, without alternative flood protection measures, is becoming increasingly impractical due to rapid urban expansion and industrial development in the area.

   

• In the event of a 1000-year flood, water would have to be diverted into the Jinjiang diversion area, but the water level at Shashi would not exceed 45 metres, thereby avoiding a breach in the Jinjiang Dyke.

   

• To achieve the flood benefits claimed for the project means that this operational schedule would have to be rigorously followed – particularly during large floods. Actual operating experience of large multipurpose reservoirs during large floods indicates that there can be substantial deviations from prescribed operating procedures which result in greatly reduced flood benefits.

   

There is no reason to believe that the Three Gorges Project operational design would not share these problems, specifically, because of the following factors:

• Conflict with people living in the reservoir area

  CYJV fails to recognize the most significant operational problem which would negate many of the flood benefits claimed. Roughly half a million people are currently living in what would be the designated flood storage area between 162 metres and 182 metres above sea level. These people would be inundated when floodwater was stored in the reservoir area above 162 metres.* In the event of a flood requiring storage above 162 metres, the dam operators would have to choose between flooding out large numbers of people living in the reservoir area or large numbers of people living downstream adjacent to the river and in overflow diversion areas. For such a flood, the CYJV flood analysis fails to demonstrate that more people downstream would be protected by the proposed flood control operation at the dam, than would be flooded in the reservoir area.

   

• Conflict with power generation

  Generally speaking, the higher the level of water in the reservoir, the more power can be generated. Because power generation provides cash revenue there would be strong institutional pressure to operate the dam for maximum power generation. This could be done either by delaying emptying of the reservoir, which has to be done prior to the flood season, or by keeping the reservoir higher than required for flood control. In fact, CYJV acknowledges that if its 160-metre recommended scheme is built, the reservoir may, in future, be operated at higher levels than presently stated, thereby sacrificing flood control in order to increase power generation. Such potential changes in operation are not reflected in the analysis of benefits.

   

• Conflict with downstream development

  Most of the expected flood benefits are based on the assumption that population and development would increase along the beach areas and overflow diversion areas. In general, when the area downstream of a dam becomes increasingly populated and developed, dam operators frequently disregard operating rules and fail to make required large flow releases for flood control during major floods.

  At the Three Gorges Dam, operators would be faced with the difficult choice of opening the spillway gates and flooding downstream property, as required by the design, versus allowing water levels in the reservoir to rise and waiting until the last minute to open the gates while hoping that the flood would abate. If the flood did not abate and the dam operators had delayed opening the spillway gates, they would then be forced to make large emergency releases from the reservoir which could cause a catastrophic flood downstream – negating expected flood benefits. This was not considered in the probability analysis.

   

• Unrealistic assumption of perfect spillway operation

  Actual operating experience with large flood control dams show that when large floods occur there is a substantial risk that designated overflow spillway capacity is not achieved due to operator errors or malfunctions of the spillway (e.g., the experience with Tarbela, Pakistan, or Glen Canyon, U.S.).

   

• Inadequate analysis of flood levels

  In order to design the dam for flood control operation and to calculate flood benefits associated with the Three Gorges Project, it is necessary first to estimate flood risk along the middle reach by pinpointing when and where peak flows would first overtop the dykes. For large rivers like the Yangtze, with its complex flow network, diversion areas, floodplain lakes and large tributary flows, it is a complicated task to determine how water levels during a flood would vary with space and time.

  The most accurate tool for this kind of analysis is by computer simulation using a hydrodynamic model, which is capable of simulating downstream flows and water levels for various flood scenarios throughout the river system at different times. CYJV recognizes that analysis using the hydrodynamic model would provide the most accurate results but rejects its use in favour of a more analytically crude model because the “accuracy of the input data is much beyond the scope of the present study.”³ (Presumably, the input data referred to is a detailed mapping of downstream channel dimensions.) This flood routing model predicts water levels at only 12 locations along roughly 740 kilometres of river in order to identify the times and locations where the flood flow would first overtop the dykes. Use of this simplistic model in a developed country would not be acceptable for such an important flood benefit analysis.

  Generally, to validate the results of such an analysis, flows computed using the flood routing model are compared with actual observed flows along the river. But CYJV’s validation is not convincing because there is a discrepancy between observed and computed flows during the peak flow period of a magnitude about the same as the total flood storage capacity of the Three Gorges reservoir. CYJV acknowledges that the main reasons for this discrepancy are hydrodynamic effects, such as backwater effects and channel conditions, which it has declined to model, but nevertheless concludes that “the model represents reasonably well” the river flows.⁴

  CYJV emphasizes the importance of accurate water level predictions by noting that a 20 centimetre change in predicted water level against a dyke could mean a 20 percent change in flood benefits. But the use of a flood routing model instead of a hydrodynamic model means that there are likely to be substantial errors in predictions as well as in flood control operational design. In addition, because key hydrodynamic factors affecting flood risk have been ignored, expected flood benefits achieved by upstream storage would not be as large if all these other factors had been considered. It is also possible that because the model does not predict where the peak flows are likely to first overtop the dykes, the operational design could increase flood risk in some areas.

Flood Benefits Exaggerated

The inadequate flood analysis and unrealistic operating assumptions together tend to overstate the flood benefits of the project. In addition, CYJV makes two other major assumptions that greatly exaggerate the value of the flood benefits:

• Economic growth rates that increase property value between 10 and 30 times present values

  Some flood protection agencies in developed countries consider the use of any predictions of future growth in flood benefit analysis as conjectural and invalid. CYJV’s assumption that such a drastic increase in investment would occur over the next 60 years for land that would remain flood prone is highly questionable.

   

• Use of the same economic growth rates with and without the dam

  One of the basic tenets of flood management planning is that the perception of flood damage potential and the type and value of land use are closely related. Obviously people prefer not to build in flood-prone areas and, even if they do, they would want to flood-proof their property. These actions would greatly reduce the flood damage potential in the “no dam” scenario and hence the net benefits claimed for the project. If, for example, the growth rate without the project was 40% of that with the project, total flood benefits would be zero.

Both these assumptions illustrate a fundamental planning error that permeates CYJV’s flood control analysis. It focuses on flood control storage as a goal in itself, whereas it is merely one tool for managing floods. Rather, CYJV should have recognized the primary goal as flood damage reduction, which is achieved through an integrated strategy that includes dykes, diversion, zoning, and flood proofing.

Because of this, the project designers have built a methodological flaw into their analysis – they fail to analyze flood management of the Yangtze River as a complete system, which would require incorporating not only “plumbing” but also hydrologic, geomorphic, social and economic factors.

Project Costs Ignored

• Relocating people out of the reservoir area

  If any realistic flood control operation were to be carried out, the half-million people living between 162 and 182 metres elevation in the reservoir region would have to be relocated, which could increase the total project budget by at least 20%.

   

• Backwater flood damages upstream

   

  CYJV predicts that the 100-year flood level would rise at least one to two metres within 30 years due to sedimentation in the backwater near the city of Chongqing. Although the increased level of flooding would affect hundreds of thousands of people, CYJV failed to consider this in its analysis of flood damages. Rather, it defines a “critical” flood at a level of 198 metres or higher, which corresponds to the elevation of the Chongqing-Chengdu railway. This “critical” level is inconsistent with CYJV’s own figures, which show damaging flows to occur at much lower flood levels.*

   

  CYJV circumvents the problem of sediment deposition by claiming that “future upstream reservoirs and conservation measures” would reduce the high sediment load carried by the river during floods.⁵ Neither the cost of reducing sediment input to the reservoir nor the cost of increased flood damages in Chongqing are included in project costs.

   

• Repairing downstream dykes

   

  CYJV acknowledges that clearwater flows* would erode the downstream dykes in the middle reaches and claims this problem would be managed by river training works and dredging. Neither of these measures would prevent undercutting of the dykes downstream and it is worth noting that elsewhere in the report, when discussing alternative flood protection improvements, CYJV lists dredging as “not very promising because it interferes with the natural equilibrium of the river, and would require frequent and costly maintenance.”⁶ Such costs, which CYJV recognizes can be exorbitant, are not included in the cost-benefit analysis.**

   

• Flood damages due to coastal erosion

   

  Adjacent to the Yangtze River mouth, extensive offshore mud flats are formed by the sediment discharged from the river in a cycle of sediment deposition and erosion. Several hundred kilometres of low-lying shoreline depend on the mud flats for protection against coastal flooding. CYJV failed to analyze the shoreline erosion and coastal flooding that will result from both the reduction in sediment delivery due to the effect of reduced flood peaks and the capture of sediment in the reservoir – a serious omission considering that the Yangtze River ranks fifth in the world in terms of its sediment discharge to the ocean. The costs of either increased coastal flooding or additional shoreline protection are not included in the cost-benefit analysis.

Failure to Consider Cost-Effective Alternatives

Because the primary purpose of the project has been defined as flood control rather than flood damage reduction, the importance of other components of the flood management system have been downplayed. For example, the existing system of dykes and diversion areas provides substantially greater flood benefits than flood control storage upstream on Yangtze tributaries. Combined with flood retention in the floodplain lakes, the total flood storage capacity downstream of the Three Gorges appears to be about 200 cubic kilometres, compared to the Three Gorges reservoir’s maximum flood storage capacity of 31 cubic kilometres.

CYJV states that a long-term program of flood control improvements is already under way but does not present this as an alternative nor discuss how the benefits of small improvements to the dykes would affect the cost-benefit ratio. On the one hand, CYJV states without substantiation that such improvements “would be uneconomic and impracticable,”⁷ and, on the other, its analysis shows that a 20-centimetre increase in the height of dykes could increase flood benefits by 20%.⁸

One important alternative component of an improved flood management system (which was relegated to an appendix of the study and not even mentioned in the study’s summary of alternatives) is the provision of ring dykes and refuge centres to protect populated areas within the diversion and beach areas. Existing ring dykes average eight metres high and encompass some 21 towns, protecting 93,000 people in the Jinjiang Diversion Area. Also, there are a number of refuge centres protected by ring dykes in this area that would serve 150,000 people in the event of flooding. CYJV concludes that “protective dykes could be justified if the annual probability of flooding is more than 1.5%,” which is equivalent to a 75-year flood.⁹ Since this is the case for the three diversion and beach areas which already provide most of the project’s claimed flood benefits, then this would be a more cost-effective alternative to building the Three Gorges Project.

Flood reduction strategies could achieve flood benefits equivalent to or higher than those expected with the Three Gorges Project by a combination of upgrading critical dykes, modifying overflow areas, providing refuge centres and ring dykes, flood-proofing, flood-warning systems, and development zoning. Non-structural measures such as development restrictions in flood hazard areas can greatly reduce flood damages. CYJV considers development restrictions along the river proposed by the Chinese government as a strategy to minimize flood damages caused by the Three Gorges Project in the reservoir region:

This restriction would prohibit future construction of new buildings, whether public, commercial, or residential, and new industries around the reservoir perimeter below El 182 [182 metre elevation mark]. This restriction would also effectively limit future population growth below El 182.¹⁰

But CYJV did not consider development restrictions as a potential strategy downstream. Using CYJV’s own figures, a restriction on development in the beach and diversion areas to less than 40% of the growth predicted in the major protected areas (such as the city of Wuhan) would equal all flood benefits expected with the project.

Failure to Consider Costs of Potential Catastrophic Failure

CYJV implicitly claims that the Three Gorges Project would prevent major loss of life in an extreme flood such as the one-in-1000-years category. There is repeated emphasis that failure of the Jinjiang Dyke would have catastrophic consequences. So, the flood control operation of the reservoir is directed towards managing water levels to prevent failure in the downstream channel, even though the probability of flood damage is much larger in other areas – for example, in the beach areas and diversion areas.

A systematic impact analysis of the project would probably indicate that the benefits of preventing loss of life are likely to be negated because the dam itself would increase the potential for loss of life as follows:

• Increased risk of failure of downstream dykes

  CYJV acknowledges that there would probably be an increased incidence of dyke failure due to a downcutting of the river channel which would undermine the river’s banks. This vitally important impact has been dismissed without substantiation by CYJV, with their statement that “channel morphology should not change significantly because the dykes that presently border the river are in many cases protected by rip-rap.”¹¹* As well, undercutting and erosion would be significantly aggravated by the wide, daily fluctuations in flow because of power generation demands. Equally important as the physical impact of the project would be the institutional impact, which could increase the risk of dyke failure. Construction of the dam would draw on funds allocated for upgrading and maintaining dykes. As experience with other large flood control dams has shown, the mistaken perception of improved flood protection due to the dam could lead to reduced maintenance and deterioration of downstream dykes and other flood management infrastructure.

   

• Potential failure of the dam

  As is discussed in the later chapter on dam safety, the risk of catastrophic failure at the Three Gorges Project is probably of the same order of magnitude as the probability of a 1000-year flood. CYJV failed to prepare a map of the area which would be flooded if the dam were to fail, even though the potential loss of life would number in the millions. Property damage would be so extensive that even if the probability of failure were 1 in 10,000 years, any flood benefits claimed for the dam would be negated, according to CYJV’s own figures.

   

• Large increases in population induced to settle in flood-prone areas

  Although CYJV recognizes that the presence of dykes can create a false sense of security leading to more fatalities when they fail, it has not applied this same relationship to the presence of the dam itself. CYJV’s analysis of flood benefits is clearly based on the assumption that an additional population would settle in flood-prone areas. What CYJV does not discuss is that the flood risk in these areas would increase as a result of any of the operational problems described earlier. Furthermore, CYJV does not discuss how risks would increase over time as the reservoir becomes clogged with sediment and its capacity to store flood waters is reduced.

  It is reasonable to assume that, as sedimentation increases, the dam would be operated to maximize protection against smaller floods. This would lead to a scenario where cities and towns have been built up in what were thought to be protected areas and ring dykes and refuge centres have long been neglected. In the event of a large flood, the dam would no longer be able to control flood waters as originally intended, and, inevitably, the loss of life in such circumstances would be greater than if the dam had never been built.

Sources and Further Commentary

*Flood benefits are calculated as the expected reduction in economic losses due to flood damage, if the Three Gorges Project were built.

*This figure does not include the additional hundreds of thousands of people who would be affected by backwater in the upstream vicinity of Chongqing.

*In 1981, Chongqing was hit by a flood that cost $5.4 million in flood damages and affected approximately 137,000 people. The flood stage reached 194 metres elevation above sea level which is four metres below CYJV’s “critical” flood stage of 198 metres.

*Clearwater flows: when the river’s sediment is trapped by the reservoir, the river carries less sediment downstream of the dam.

**As an example, the estimated cost (as of 1980) for an ongoing program of dyke improvements is $1.8 billion.

*Rip-rap is the term used for large rock which is commonly used to protect the surface of earthen dykes, embankments, and river banks.