A Review of Hydrological Aspects of the Proposed Epupa Dam and Reservoir, Cunene River, Namibia

Peter Willing
Thursday, January 8, 1998
1. Executive Summary

From a hydrologist’s point of view, the Feasibility Study of proposed hydroelectric power dams on the Cunene River has some serious deficiencies. In order of importance, they are: 1) The study is organised so as to be virtually inaccessible to even a careful reader. Separate pieces of the same subject matter are scattered in illogical places throughout the voluminous corpus of the study. 2) Flow data, and estimations in the absence of data, are of low reliability. The entire hydrological analysis is based on the premise that a meagre 12 year streamflow record from a location 200 km upstream of the dam site can be linked with a longer record from another river basin to synthesise a reliable theoretical streamflow data base with which to assess the viability of a 1.9 billion dollar hydropower project. This proposition is shaky at best. 3) The analysis of the effects of changes in the river’s sediment regime as a result of building the dam are incomplete and ignore the most important issues, such as increased degradation downstream and upset of the sediment regime in the delta. 4) The study lacks a definitive appraisal of the thermal and nutrient regime in the reservoir and river. The potential of drastic changes in the temperature, oxygen and nutrient regimes of the reservoir and lower river to upset existing ecological and cultural relationships have been naively understated. These points are elaborated below.

The conclusion is that the Feasibility Study is not a sound basis for evaluating either the costs or the benefits of the project, nor for proceeding with major financial or other irreversible commitments to the project.

2. Basis of Review

This review is based on the electronic version of the Feasibility Study, which was posted on the Burmeister Partners web site. This version of the Feasibility Study is incomplete in several respects, the most important of which is the missing figures from the hydrology chapters (Baynes Main Report, Chapter 7; Epupa Main Report, Chapter 7) and the reservoir operation (Epupa Main Report, Chapter 13). Appendices with supporting data and references are also missing. The hydrology sections are necessarily quantitative, and much of the analysis is best portrayed graphically. If the figures are not available, this severely limits the scope of review that can be carried out on the study.

I have concentrated my attention on the hydrology and sediment, and have referred to the chapters on reservoir and dam engineering design and construction. The chapters on the hydrology of the Epupa dam site and the Baynes Mountain site appear to be virtually identical with the exception of the names and physical characteristics of the sites. I have not read the entire electronic version of the Feasibility Study.

3. Organisation

The Feasibility Study was not written to assist the reader in understanding the project and its impacts. The organisation of the study is convoluted and redundant. The duplicate chapters on Baynes and Epupa sites with nothing but the names changed is an example. Material relating to hydrology and reservoir management is not presented or even referred to in the appropriate sections, but is referenced in general sections elsewhere. A reservoir stratification model was considered (Chapter 2, Epupa Summary Report), but it was not mentioned in the report section on reservoir operation (Epupa Main Report, Chapter 13). Hydrologic simulations of downstream effects on the river were included with the economic simulations, but were not mentioned in the chapters on hydrology. This lack of coherent organisation makes a technical appreciation of the project nearly impossible.

4. Flow Data and Estimations (Chapter 7)

The central proposition of the hydrologic analysis is as follows: There is no flow record at either proposed dam site (Epupa or Baynes). There is only a short record at an upstream site (Ruacana), that we hope will tell us something about the dam sites. There is a long record from one site on another entirely different river basin (Rundu). So we can synthesise an artificial hydrological picture for Ruacana based on the twelve years for which Rundu and Ruacana have an overlapping record, and then translate that picture 200 km downstream to the proposed damsites. The correlation between the two gauges on the Cunene and Okavango Rivers depends on the assumption that the two watersheds are hydrologically similar. There is no detailed description of those two watersheds that would convince us of this similarity. The whole integrity of the analysis depends on our confidence in the correlations among all the sites. While the consultant should be commended for developing a creative approach to the problem of inadequate data, confidence in the synthetic streamflow record as a basis for evaluating the Epupa hydropower project is not warranted. The limitations of the data have been conveniently forgotten: Chapter 11 says, "In this simulation, the last 49 years of stream flow for the Cunene river are used." 49 years of stream flow for the Cunene river do not exist. Only 12 years of stream flow for the Cunene exist, and the rest are based on an assumed similarity to a distant location on the Okavango. The simulation in question is the core of all the predictions of economic behaviour of the Baynes and Epupa schemes.

The analysis also assumes that future behaviour will be the same as the past. There is an explicit assumption of "stationarity" in the record; i.e. that there is no long-term trend in the data, and that past variability of hydrologic behaviour is considered a reliable guide to future variability. The data series being used does not take account of recent extreme low runoff conditions, nor does it take into account global warming possibilities. The essence of global warming is a long-term upward trend in temperature data, which would drive trends in other characteristics such as evaporation, rainfall, and runoff. The assumption of a stationary record, in the face of evidence to the contrary, is a risky basis for a very long-lived and highly capital intensive investment that depends on streamflow.

The feasibility study has a prolonged discussion (Section 7.5) of speed gauging data. What this refers to is the point velocity measurements that are supposedly taken on a regular basis, in order to establish the relationship between depth of water, i.e. of gauge height, and volume of discharge; which must be calculated. This relationship is depicted graphically in a rating curve that is unique to the gauging site and the watershed’s hydrologic response. Various manipulations were carried out on the rating curves at Rundu on the Okavango River and at Ruacana on the Cunene River. These rating curves are not available for inspection, the explanation is not clear, and the work is not shown. This makes it impossible to evaluate the hydrologic conclusions. It appears that certain river discharge measurements were rejected because they "were clearly identified as outliers" and because of "documentation discrepancies." The study does not say which measurements were rejected, or how many, or how many were left as a basis for the correlation. The shape of a rating curve is extremely sensitive to the speed gaugings at both high and low flows. It is curious that one would hand-pick one set of data that is considered reliable and then say that the correlation with another set of data is good. Would it get better if more data were thrown out? There must have been a problem with low velocity discharge measurements, where "current meters are at their limit of applicability" ( There is no reason why a standard current meter should not function acceptably at velocities well below 0.1 m/sec., which would give adequate discharge measurements in suitable river cross sections.

The study says "all authors agree" on the good correlation between Rundu and Ruacana, yet the seasonal lack of correlation between the two is glossed over (Section 7.4.4). The study allows for possible error in the rating curve by introducing two error bands, the lower of which is 15% below the best fit rating. Yet they use the best fit rating curve instead of the more conservative "minimum" or "worst case" ratings as a basis for the estimate of mean annual runoff. Given the weaknesses of the data, the estimate based on best fit is not entitled to great confidence. The difference could have severe impacts on the economic performance of the hydropower schemes.

The analysis (7.6.1) treats the data as though there is no serial correlation between annual runoff volumes, in other words the study assumes dry years do not necessarily follow other dry years. Elsewhere, the study says (7.9.3) that "dry periods are considerably longer than one year." It says (11.6.5) that Baynes could run dry at the end of the first of a series of dry years. A conclusion of no annual serial correlation is also at odds with the findings of other investigators (Kerr, 1985), and with emerging understandings of cyclical El Niño - Southern Oscillation effects (Philander, 1989). It is a generous assumption that low flow years do not occur consecutively. If in fact they do occur back to back, then the estimates of the duration of a drought or the runoff produced during a drought would be over-optimistic and the economic performance of the hydroelectric scheme could be considerably worse than assumed.

Evaporation data for the Epupa site was used directly for the Baynes site, which is about 35 km farther west and 180 m lower in elevation. The evaporation at this site would be higher, but the magnitude is not known. Evaporation rates higher than estimated would degrade the economic performance of the reservoir scheme.

5. Sediment (Chapter 8)

Discussion in the Feasibility Study: The scope of the sediment discussion appears to be confined to a very narrow set of questions to be examined, enumerated in the beginning of Chapter 8: "the long term effect on reservoir storage volume . . . the likely deposition pattern . . . possible effects on the intake structure to the powerhouse." These are important questions, but they are hardly a complete treatment of the issue. The study could reasonably be expected to offer a reach-specific discussion of the characteristics of the bed load of the river above and below the proposed reservoirs. Section 8.4.1 outlines an approach that includes "landscape analysis and the application of sediment yield figures based on empirical data from similar environments," yet the analysis does not make use of experiences with sediment erosion and deposition at other reservoirs. While observing that the "sediment data is insufficient for a conventional analysis," i.e. frequently sampled data over five years, the study ignores river characteristics that would be very useful and would not take much time or money to find out: reach gradient, channel confinement, sediment storage volume estimates, sediment particle size distribution, lithology and weathering characteristics -- in short, a geomorphologic analysis of the river. While observing that the river has several gradient and sediment regimes over its length, the generalised description is never interrupted by an application of these differences to the two proposed dam sites. The sediment analysis is based on "weak data . . . no measurements" (Section 8.5). No measurements of bed load have been made. Nonetheless, the study says "the reach between Ruacana and Epupa shows clear evidence of substantial bed load transport," and "large quantities of sediment in channel storage, especially in the ephemeral tributary streams." This sediment is now in an equilibrium with the rest of the river; the study does not acknowledge that the equilibrium will be upset by dam construction.

The sediment chapter has an unwarranted preoccupation with sediment rating curves, which even the Feasibility Study admits are of little use in the Cunene situation. As an illustration, Leopold et al. (1964) show a sediment rating curve for the Rio Grande River in New Mexico that exhibits two to three orders of magnitude of variability in the sediment transport rate for the same flow. Authority for the "normally used sediment discharge rating curves" (section 8.5.1) is lacking. No details are offered of the techniques used for the described sediment samples; sediment sample results are highly sensitive to method (Leopold et al., 1964).

The Ruacana reservoir sediment survey (section 8.7) is inconclusive. The study refers to modelling effort applied to the case, but the model and its inputs are not described. Apparently this modelling effort was never finished, because the study points out the necessity of completing it (Section 8.9).This section has several unsupported assertions about the sediment at Ruacana: "the impact [of extreme events] is generally balanced out within less than 10 years" and "it is found that . . . after 10 years the accumulated volume provides a reliable measure of the long term average load, irrespective of hydrological patterns during the period."

The loss of capacity due to sedimentation at Baynes or Epupa is probably, as the study concludes, not likely to be of major short-term consequence in reservoirs with storage capacities that are a large multiple of the likely sediment accretion rate. The sediment analysis is not a strong one, however. A paraphrase of the analysis would be "We don’t know what the sediment yield would be, all we can do is make a couple of educated guesses. But even if the sediment yield is twice as much as our biggest guess, the reservoir will still hold it."

Sediment effects not considered in the Feasibility Study: The bed load will be interrupted by the reservoir; there will be no bed load moving beyond the still water at the head of the pool, and none moving through the reservoir to the downstream reaches. Energy now utilised in transporting upstream sediment will be available to move other sediment, in unforeseen and possibly undesirable ways. This difference in the dynamics of the river can produce far-reaching effects downstream. The dams will impose a complete change on every physical characteristic of the river. Each perturbation of one characteristic will have a reaction on the other characteristics. The likely direction and magnitude of these adaptations is an essential piece of the analysis. No discussion is offered at all on the downstream effects of sediment trapping. This can consist of bed armouring, channel bed and bank scouring, destabilisation of existing gravel bars, desiccation of riparian habitat, and perturbation of the sediment budget of the estuary at the mouth of the river. The riparian vegetation, the dynamics of the estuary, the littoral drift to the north of the river mouth, the lagoon, and the biological systems that depend on them are delicately adjusted to the accretion of sediment in the river. This sediment supply will be interrupted permanently. Some combination of bed erosion and bank erosion downstream of the reservoir can be expected to compensate for this loss of sediment supply. The stream reaches in question are high gradient reaches with relatively high velocities. At velocities of 4 to 5 km/h (Simmons et al., 1993), this clear water will have considerable erosive power.

Sediment problems other than loss of reservoir storage capacity are well known to the dam building industry (Hu, 1995). These problems have been known to have drastic effects on the operation and economics of reservoirs. For numerous projects, downstream starvation of river sediment load threatens to exceed reservoir siltation as a cost of dam construction (Kazmann, 1972). The Feasibility Study would be strengthened by a discussion of relevant African cases, and an explanation of why the Cunene River should be considered exempt from these kinds of problems.

6. Temperature

It is virtually certain that either the Baynes or Epupa reservoirs would develop a stable thermal stratification early in the life of the reservoir. Although the information in the study is inconsistent, it appears that Epupa is planned for a maximum depth of 157 M and volume of 11.5 billion M3, and Baynes is planned for 200 M depth and volume of 2.6 billion M3. The stratification phenomenon does not draw comment in the reservoir operation chapter (13) of the Feasibility Study, although references to it are scattered in other sections. A stratification model is mentioned (2.5.1) but neither inputs nor outputs to the model are supplied.

Stratification means that the water body segregates into two distinct layers of water that rarely if ever mix. These layers are different in temperature, oxygenation, nutrients, and density. In deep tropical water bodies, a small temperature difference between surface and bottom can be sufficient to maintain a stable stratification (Hutchinson, 1957). An initially weak thermal stratification can become stronger because of differences in chemical constituents. It is known that the temperatures at the mouth of the free-flowing Cunene River are warm, ranging from about 19 C. to 26 C. (Simmons et al., 1993).

The dam design makes allowance for variable depth releases, so that the chemistry and
temperature of release water can be managed (20.2.2). What the Cunene reservoirs will be like is unknown; but typically, reservoir releases from the epilimnion (upper layer) tend to be warm, partially oxygenated, and relatively lean in nutrients. Releases from the hypolimnion (lower layer) tend to be colder, anoxic, rich in nutrients, and rich in anaerobic decomposition products.. Releasing cold water that may be completely devoid of oxygen will have far-reaching effects on the downstream ecosystem. On the other hand, releases from the upper layer could isolate the hypolimnion and cause its water quality to degenerate as it accumulates nutrients and organic matter. Even with variable depth releases, it will be impossible to mimic the characteristics of the free-flowing river.

The Feasibility Study deals with these issues by saying, in essence, "We do not know what will happen as a result of impounding the river. We do not know what kind of aquatic habitat is there now, nor what will become of it, nor what the needs of the surviving species will be. But we will monitor it, and adapt mitigation after the fact, including operational constraints, to suit the situation." It is a severe defect of the proposals that no one knows what the possible effects are; much less what the mitigation measures will cost directly, nor how they would affect the economic benefits of the project.

7. Reservoir dynamics

Damming the river will bring about dramatic changes in its sediment, temperature, and nutrient economy. These changes will have equally dramatic, but unknown, consequences on stream morphology and ecology in downstream reaches of the river. The Feasibility Report makes no acknowledgement of the complete change of state of the river basin attendant on construction of a reservoir. It does assume unrestricted operation of the reservoir, presumably to meet hydropower objectives at all times (Section 13.3). The Baynes reservoir pool level will apparently be highly variable: the standard deviation of the synthesised annual flows is 45%, and the active reservoir storage is 1.5 times the mean annual runoff. This raises doubts that the proposed 50 m drawdown in the simulation chapter will be observed. The study acknowledges the possibility of running the reservoir empty (11.6.5).

Wetzel and Likens (1991) offer a well rounded but general description of reservoir dynamics:

Reservoirs receive runoff water mainly via large streams, which have high energy for erosion, large sediment-load carrying capacities, and extensive penetration of dissolved and particulate loads into the recipient lake water. Because the inflows are primarily channelized (along the prior river course before it was inundated) and often not intercepted by energy-dispersive and biologically active wetlands and littoral interface region, runoff inputs to reservoirs are larger, more directly coupled to precipitation events and extend much farther into the lake per se than is the case in most natural lakes. All of these properties result in high, but irregularly pulsed, nutrient and sediment loading to the recipient reservoir . . .

Large areas of sediments are alternately inundated and exposed. These manipulations usually prevent the establishment of productive, stabilising wetland and littoral flora. Erosion and resuspension of floodplain sediments augment loadings from elsewhere in the drainage basin. Sediments may be shifted between aerobic and anaerobic conditions, which enhances nutrient release . . .

High nutrient and organic matter loadings occur during the early "trophic surge" period following the damming of a river . . . The environmental conditions of reservoir ecosystems tend to have large, rapid, and erratic fluctuations. Often insufficient time exists for complete population growth and reproduction to occur before a succeeding major disturbance occurs, e.g., a plume intrusion and disturbance of stratification patterns following a major rainfall. These instabilities result in biota that tend to be few and well adapted with broad physiological tolerances (low diversity, less specialisation, rapid growth). As in all restrictive, stressed environments, the productivity of adapted organisms can be high, as high or greater than in more homeostatic natural lakes.

8. Conclusion

There is no reason to think that dams on the Cunene River will be different from others in their effects or operational characteristics. A complete understanding of hydrology and river morphology are fundamental building-blocks for predicting the economic performance and side-effects of a river basin development scheme. The Epupa and Baynes Feasibility Reports available for review in December 1997 do not contain such an understanding.

Author’s Biographical information

Education: B.A., University of Washington, Seattle, Washington. M.S., Ph.D., Cornell University, Ithaca, N.Y.

Professional Affiliations: American Water Resources Association; American Geophysical Union .

Dr. Willing is Principal in the Bellingham firm of Water Resources Consulting, L.L.C. Since founding the firm in 1989, he has carried out a wide variety of assignments for public and private clients seeking to solve water-related technical questions. Examples are: hydroelectric system design, flood frequency analysis on North American rivers, wellhead protection, surface water - ground water interactions, storm water management strategy, and hydrologic basis of water rights. He has participated in technical reviews of African river basin development plans. In public sector positions, he has served as general manager of a public water supply system. He also served as chief environmental officer of a large municipal electric utility. Dr. Willing holds Adjunct Faculty appointments in Geology and Huxley College at Western Washington University, Bellingham.


Hu Chunhong, 1995. Controlling reservoir sedimentation in China. Hydropower and Dams, March 1995.

Hutchinson, G.E., 1975. A Treatise on Limnology. Wiley & Sons. 2nd ed.

Kazmann, R. 1972. Modern Hydrology. New York: Harper & Row. 2nd ed.

Kerr, R.A. 1985. Fifteen Years of African Drought. Science 227: 1453-1454.

Leopold, L.B., M.G. Wolman, and N. Miller. 1964. Fluvial Processes in Geomorphology. W.H. Freeman.

Philander, G., 1989. El Niino and La Nina. American Scientist 77(5): 451-459.

Simmons, R.E., R. Braby, S.J. Braby. 1993. Ecological studies of the Cunene River mouth: avifauna, herpetofauna, water quality, flow rates, geomorphology and implications of the Epupa Dam. Madoqua 18(2): 163-180.

U.S. Bureau of Reclamation, 1960. Design of Small Dams.

Wetzel, R.G. and G.E. Likens, (1991). Limnological Analyses. 2nd ed. Springer Verlag. 391 pp.

Comments by
Peter Willing, Ph.D.
1903 Broadway
Bellingham, Washington 98225 U.S.A
Tel: 360-734-1445 Fax: 360-676-1040
email: pwilling@telcomplus.com

Further Reviews: