Philip B. Williams, Ph.D., P.E.
The Yangtze is not only a river of water, it is also a river of sediment. The flow of the Yangtze carries with it the fifth-largest sediment discharge of any river in the world, equivalent to about 4 percent of all river-borne sediment discharged to all the oceans of the world.
In a river like the Yangtze, most of this sediment is conveyed during floods, and it is sediment deposited from these floodwaters that has formed the extensive fertile floodplain of the Yangtze Valley downstream of the Three Gorges. As the river flows across the valley to the sea its waters erode and redeposit the sands and silts of its bed and banks.
This dynamic equilibrium between the river’s erosion and deposition creates the slope and course of the river channel. It is on this active form of the river channel that cities have grown up and levees for flood defence have been constructed. Further downstream, the Yangtze estuary and its low-lying fertile coastal plains have been created by sediments brought down by the river, and major cities like Shanghai have been established on these fragile landforms.
When a large reservoir is filled with water, the flow of sediment through the river system is interrupted. The coarser sediment materials, the boulders, gravels and sands, are deposited at the upper end of the reservoir forming a delta. Almost all of this coarser material is carried by the flow along the river bed and is referred to as bed load. The amount of bed load moved is determined mainly by the flow velocity, with most of the bed load moved during high-velocity flood events. When the velocity of the river decreases suddenly as it merges with the slow-moving reservoir water, the bed load is deposited raising the river bed, a process known as aggradation. Because the river channel fills with sediment, floodwaters spill out more frequently onto the adjacent floodplain.
The bed load entering a large reservoir, like that of the Three Gorges, is trapped in the reservoir. The discharge, free of bed load, will erode sediment from the bed and banks downstream. This causes the lowering of the river bed, a process known as degradation. Degradation downstream of a dam the size of the Three Gorges would extend for hundreds of kilometres, threatening cities and populations that live downstream.
Usually the bed load comprises a small but important proportion of the total sediment load. Researchers typically estimate the proportion of bed load in large rivers to be in the range of 2 to 8 percent of the total sediment load. The rest of the sediment, generally consisting of fine sands, silts and mud, is referred to as the suspended load. This finer material takes longer to settle in a reservoir, and is therefore normally transported further into the reservoir than the bed load.
It is only in the last few decades that dam designers have started to recognize the need to evaluate thoroughly the processes of reservoir sedimentation and the long-term impacts on the dam and the river upstream and downstream. Now, with about 1 percent of the world’s total reservoir storage capacity lost to sedimentation each year, there is a growing recognition of the need to extend the lifetime of these reservoirs as long as possible by attempting to use new types of operations that try to maximize the amount of suspended sediment flushed through the reservoir.
Techniques to deal with reservoir sedimentation problems are still in the earliest stages of development, and there are still many problems inherent in sedimentation analysis that have not been worked out. Despite these problems, the proponents of the Three Gorges Dam have put forward a dam design and method of operation that they say will deal completely and effectively with the sedimentation problem.
A review of the sedimentation analysis and the design and operation procedures for dealing with sedimentation, contained in the CYJV feasibility study, reveals fundamental errors in the analysis done by CYJV and flawed premises on which the CYJV analysis is based.
General Problems with Sedimentation Analysis
The planners of such a large-scale dam project as the Three Gorges have an extraordinarily difficult task in analyzing and predicting the impacts of sedimentation, for the following reasons:
First, the type of operation and design proposed to minimize sedimentation has not been successfully demonstrated on such a large dam project. In effect, the Three Gorges Dam is a gigantic experiment in river management.
Second, the scale of the river and sediment discharges that the planners are attempting to manage is unprecedented. CYJV reports that total sediment discharge at Yichang is 530 million tons/yr, equivalent in volume to about 0.43 cubic kilometres (km3) per year, compared to the total normal reservoir power pool volume of 26 km3. The only previous attempt to manage sediment flows of this magnitude was at the Sanmenxia Dam on the Yellow River, which is widely acknowledged outside China to have been a costly failure due to unanticipated sedimentation problems.
Finally, the level of confidence required of the feasibility analysis in recommending flushing as a sediment control mechanism exceeds the confidence limits of the state-of-the-art of the sciences of sediment hydraulics and fluvial geomorphology. Even with the best analytic and data collection methods, sediment discharge can only be predicted within error bands of several hundred percent for a given flood event.
Recognizing these difficulties and taking into account the importance sedimentation analysis has for critical design issues – namely, the useful life of the reservoir, navigation, and the impacts on millions of people living upstream, downstream and on the estuary – it would be reasonable to expect that CYJV’s analysis would adopt a conservative approach. For example, one would expect CYJV to take into account actual worldwide experience with reservoir sedimentation control. Also CYJV should have reviewed worldwide experience with river channel changes due to reservoirs. Furthermore, it was CYJV’s responsibility to test independently YVPO’s data, and the sensitivity to uncertainties of this crucial analysis.
Inconsistent and Incomplete CYJV Analysis
CYJV did not review YVPO’s original data nor did it analyze the long-term consequences of sedimentation in the Three Gorges reservoir. Despite this incomplete review, CYJV made a finding that is unprecedented in the field of sediment hydraulics: CYJV predicted that the Three Gorges reservoir storage can be preserved indefinitely. But, the data and analyses on which CYJV bases this extraordinary finding are inconsistent, incomplete, and fail to substantiate the claim.
CYJV states categorically in its summary document written for decision makers and potential financiers, “about 90% of [the reservoir’s] effective storage can be preserved indefinitely. Reservoir sedimentation will not limit the useful life of the project.”1 However, a closer reading of CYJV’s technical report on sedimentation qualifies and undercuts this sweeping statement.
According to CYJV’s sediment volume, “preserved indefinitely” actually means “85 to 90% of the regulating and flood control storage will be preserved after 100 years of operation, when the system reaches or is approaching equilibrium.”2* Equilibrium is further qualified as follows: “Approximate equilibrium is reached when 90 to 95% of the sediment entering the reservoir is flushed through the reservoir.”3 What this really means is that CYJV estimates that after 100 years, about 50 percent of the reservoir volume will have filled up with sediment,4 but maintains that only about 10 to 15 percent of the active power pool will have been filled because CYJV predicts most of the sediment will fill the deepest part of the reservoir. Furthermore, CYJV predicts that over the first 100 years of operation, the rate of siltation in the reservoir will decrease from about 70 percent5 of incoming sediment to only 10 percent.
Although CYJV acknowledges that sediment will continue to accumulate in the reservoir after 100 years, it does not address the inevitability of most of the reservoir storage ultimately being lost. Nor does it address the future task of decommissioning a reservoir like Three Gorges when much of its storage has been filled with sediment.
As well as being contradictory and incomplete, CYJV’s sedimentation analysis is unreliable because it is based on flawed premises.
Flawed Bases of CYJV’s Sedimentation Analysis
1. Incomplete Review of YVPO Analysis CYJV’s task was to make an independent review of YVPO’s sedimentation predictions. CYJV concluded that this work was “very satisfactory” and “of consistently high quality.”6 CYJV came to this conclusion apparently without verifying YVPO’s original data. Although CYJV lists the main sedimentation reports prepared by several Chinese agencies and research institutes, it admits that only “a number of brief summary reports and publications from these groups were made available by YVPO.”7
While China has experienced and competent hydraulic engineers, it is not clear what role they have played in selecting the key assumptions used in YVPO’s analysis. The question of feasibility of the project is highly politicized and if there was a negative conclusion to the sedimentation analysis, it would be difficult to argue that the project makes sense. In these circumstances, as an outside reviewer, CYJV has the responsibility to evaluate rigorously the key assumptions and the degree of confidence in these assumptions, and to verify the veracity of the data used.
It does not do so, but rather replicates the model developed by YVPO and uses the same assumptions and unverified sediment data. Not surprisingly it produces similar results, which it then presents as confirmation of the YVPO results.
2. Unreliable Equilibrium Slope Calculation
The premise on which CYJV appears to place most importance and which is described in the executive summary8 is a prediction that sediment would deposit in the reservoir at an equilibrium slope. Once the sediment has settled into an equilibrium slope, the same quantity of sediment would enter and be flushed out of the reservoir (i.e., no additional sediment would be trapped behind the dam).
The estimation of the equilibrium slope is extremely important because it not only affects the estimation of reservoir sedimentation but also aggradation and navigation in the river channels upstream. If the slope prediction is underestimated by only 10 percent, the river bed and flood levels upstream in Chongqing would be about 4 metres higher, flooding out hundreds of thousands of people and impeding navigation.
CYJV predicts that such an equilibrium slope, extending from the bottom of the dam’s outlet structure to the upstream end of the reservoir, would be shallow enough to preserve most of the reservoir volume indefinitely. CYJV based this assumption on several methodologies – regime equations, sediment transport equations and a numerical model.9
A closer examination of these methods shows all of them to be inadequate for the purposes of predicting sedimentation rates, volumes, and equilibrium slopes, to the level of confidence required for such an important design consideration.
These methods cannot predict the equilibrium slope with any reasonable confidence for the following reasons:
Regime equations are based on empirical data from smaller rivers and irrigation canals. An authoritative review states: “Regime relations should never, of course, be applied in cases in which the flow, sediment transport, and channel characteristics differ widely,”10 as is the case for the Yangtze.
Predictions of slope based on the sediment transport relations are very dependent on the size of sediment selected. CYJV has calculated the equilibrium slope based only on sediment sizes between 0.1 and 1.0 millimetre (mm), assuming a median of 0.23 mm. If the amount of coarser material has been underestimated (see #3 below) the slope would be considerably steeper. For example, if the median sediment diameter was 0.26 mm the equilibrium slope could be 10 percent steeper, causing flooding in Chongqing.
The one-dimensional model, and the other computational techniques, assume that flow through the reservoir can be represented by a single averaged value of velocity even though the reservoir is typically about 1 kilometre wide and 60 metres deep. This ignores the complex flow structure in the reservoir, the highly variable geometry (unlike a river) and the effect of three-dimensional density currents in distributing sediment throughout the reservoir.
Equilibrium slope predictions are extremely sensitive to the hydraulic roughness selected. CYJV data shows that the observed roughness can vary by about 100 percent depending on river stage and location within the Gorges. The report points out that a 15 percent error in slope, equivalent to about an 8 percent error in roughness, would cause the bed level at Chongqing to be 6 metres higher, flooding out hundreds of thousands of people.
CYJV does not take into account the importance of cohesion in deposits of fine sediment that limits scouring. By failing to do so, CYJV has overestimated the ability of the reservoir operators to resuspend and flush out deposited material.
CYJV does not consider the extent to which the deposition of coarser bed load material will affect the equilibrium slope over time.
If CYJV’s prediction of an equilibrium slope is not correct, sediment could continue to accumulate in the reservoir, rather than being flushed through the dam. This would decrease the storage capacity of the reservoir, cause aggradation upstream that would result in flooding and threaten hundreds of thousands of people near Chongqing, and degradation downstream that would result in erosion and threaten downstream development.
3. Miscalculation of Bed Load
CYJV has seriously underestimated the bed load of the Yangtze.
CYJV accepted YVPO’s estimate that the bed load (defined as sediment larger than 1 mm) conveyed by the Yangtze is only 0.05 percent of the total sediment. This estimate is contradicted by statistics from the Yichang Hydrological Gauging Station which indicates that the Yangtze’s bed load is 1.6 percent of its total sediment load.11
Nonetheless, CYJV’s entire sedimentation analysis is based on a bed load of 0.05 percent, an amount so small that CYJV counts it as zero in the reservoir sedimentation analysis.
Even though CYJV acknowledges that “the quality and quantity of basic field data is of crucial importance to the sediment load investigation”12 it admits it did not even review YVPO’s bed load field data. Nevertheless, CYJV judged the YVPO studies based on these data “to be well done.”13
Getting an accurate measurement of bed load is extremely difficult, even for small rivers, because of practical difficulties extracting samples from the bed of a fast-flowing river. Sediment transport rates vary across the river bed and over time, even for the same flow. Apart from experience in other rivers, there is substantial evidence that the bed load of the Yangtze is significantly greater than CYJV has determined. For example:
The existence of extensive boulder and gravel bars throughout the Three Gorges that move downstream during floods.
The quantity of sand and gravel to be excavated from the river bed downstream of the dam site for concrete aggregate for the Three Gorges Dam. CYJV states “there is no concern about the availability of these materials.”14
Reports from other Chinese researchers describe gravel and boulder deposits in the Three Gorges more than 35 metres thick.
Underestimation of the bed load would in turn cause the underestimation of the equilibrium slope of reservoir sedimentation, the rate and extent of upstream river aggradation, the difficulty in managing the river for sediment pass through, and the rate and extent of downstream river degradation, all of which have major negative implications for the technical, economic and environmental feasibility of the project.
4. Inadequate Empirical Estimate of Reservoir Trap Efficiency
The reservoir trap efficiency method used by CYJV to calculate reservoir sedimentation, although independent of the equilibrium slope method, cannot estimate sedimentation in the active reservoir storage zone (above 140 metres) and therefore cannot be used to predict its rate of filling. Moreover, this method underestimates sedimentation in the dead storage zone (below 140 metres) for the following reasons:
It assumes constant reservoir levels during floods. Yet the purpose of flood control is to store water during floods. The larger the flood, the greater the sediment inflow and the greater the amount of water stored. In a 20-year flood, for example, the reservoir is expected to reach 160 metres, doubling its volume. This in turn, based on CYJV’s method, could increase the trap efficiency from about 10 to 30 percent, thereby substantially increasing sedimentation in the reservoir.
It ignores the considerable reservoir volume stored during large floods in the sloping water level of the reservoir. During moderate to large floods, even if the outlet level can be maintained at 140 metres, the reservoir level at the upper end may exceed 160 metres.15
There is considerable scatter in the data from which these trap efficiency curves have been derived. For example, trap efficiencies for the envelope curves can range from 60 to 85 percent. CYJV selected 60 percent, characterizing the lower envelope curve for fine sediment, but without taking into account the error band of this data.
Underestimation of Reservoir Sedimentation Rates
In addition to the inadequacies of these methodologies to calculate sedimentation rates, the total sedimentation rates have been underestimated because of:
The underestimation of bed load sedimentation. This could account for at least 1.6 percent of the total sediment load, based on estimates by the Yichang Hydrological Gauging Station, as opposed to 0.05 percent adopted by CYJV.
The effect of landsliding into the reservoir. Within a 100 years, landsliding could fill several cubic kilometres of the reservoir. More important, where massive landslides occur, erosion-resistant control points would be formed within the reservoir, preventing the scouring of accumulated sediments upstream and resulting in induced upstream flooding.
The potential increase in sediment delivery rates as watershed conditions deteriorate.
The episodic nature of sediment delivery. After a large flood considerable aggradation could occur that may take decades to erode to an equilibrium condition. This places considerable doubt on whether the concept of equilibrium slope has real meaning in managing systems of this size.
Implications of Flaws in Sedimentation Analysis
CYJV has not convincingly demonstrated that its Three Gorges sedimentation analysis can realistically predict the actual performance of the reservoir, nor has it carried out a systematic sensitivity analysis of the cumulative effect of uncertainties in its predictions. If it had done so it should have concluded:
There is a significant risk that sedimentation rates in the reservoir, similar to those observed in other major reservoirs, could substantially impair the performance of the project in its economic lifetime.
There is a significant risk that aggradation of the river bed upstream past Chongqing will be substantially higher than estimated, flooding hundreds of thousands of people.
There is a significant risk that the aggradation of the river bed upstream and the degradation of the river below the Gezhouba locks will greatly impede navigation. Because the bed load has been underestimated, the costs of dredging navigation channels have been underestimated.
There is no realistic way the reservoir can be managed, as it fills with sediment and loses its flood control storage, to protect millions of people who have been induced to move into flood-prone areas downstream.
The capture of sediment in the reservoir will cause significant degradation of the river bed for hundreds of kilometres downstream, eroding flood control embankments, undermining bridge crossings and changing the hydrologic regime of the river on which millions of people depend. The long-term costs of additional levee repair have been seriously underestimated in the study.
There is a significant risk that sediment captured in the reservoir would accelerate coastal erosion. (This is acknowledged in a technical appendix16 but ignored in the main report.)
If the Three Gorges Project is completed as planned, it is probable that within a few hundred years the reservoir will almost entirely silt up, creating an unprecedented hazard to the millions of people living downstream, whose culture has survived and prospered for the past 4,000 years with wise management of the Yangtze River.
Sources and Further Commentary
*Perfect equilibrium is achieved when the same quantity of sediment enters, as is flushed out of the reservoir.