Hydrology Data and Dams

Date: 
Monday, October 22, 2007

Excerpt from Silenced Rivers: The Ecology and Politics of Large Dams,
by Patrick McCully, Zed Books, London, 1996

Just as dam builders often skimp on geological surveys, so they have shown themselves willing to build on the basis of seriously inadequate hydrological data. When there is not enough water to turn a dam’s turbines or fill its canals, or so much water that the dam is threatened with breaching, an "Act of God" – drought or flood – will invariably be blamed for the ensuing electricity shortages or inundation. However, an act of dam builder – construction without sufficient data to predict how much water is likely to be available or deliberate disregard of unfavourable data – is more likely to be where the blame should be laid.

Hydrologists cannot predict exactly how much water will flow into a planned reservoir. To make a "best guess" of what river discharge will be like during the life of a dam they project past streamflow data into the future. At least several decades of streamflow records are usually necessary to make reasonably reliable predictions taking into account annual cycles of variations in rainfall – and even then there is no guarantee that rainfall patterns over the next say 50 years will be the same as those of the last 50 years, especially considering the hydrological wild card of global warming. Furthermore, to be able to estimate the maximum flood which a dam might have to discharge and to plan its pattern of releases it is necessary not only to know annual flow variations, but also seasonal, monthly and even daily peaks and troughs.

Collecting reliable streamflow data is relatively expensive and difficult and for most of the world’s rivers, very little good data exist. If the data do not exist, then a dam builder might have to wait many years for a time series long enough to be meaningful. Because of the lack of discharge data, hydrologists often extrapolate streamflow from rainfall statistics, which are generally of better quality and more common. However this introduces even more uncertainties into the hydrologists’ equations as it requires a number of assumptions to be made on the relationship between rainfall and run–off taking into account numerous factors including rainfall intensity, evaporation and ground cover.

Especially in semi–arid areas, rainfall and river discharge can vary so much over time that even "averages" based on many decades of reliable data may have little relevance in predicting future flows. Geographer William Adams explains that:

To someone judging semi–arid Africa by the standards of what is normal in a temperate environment, the variability of climate is both bizarre and incomprehensible. Ideas of "normal" rainfall, based upon notions of "average" rainfall conditions, might be appropriate to temperate environments, but have tended to prove a poor basis for planning in Africa... The use of computer simulation and statistical analysis serves to give confidence to the modern development planning process, but it does not always succeed in comprehending the variability of the African environment...

A lack of reliable hydrological data, however, frequently does not stop dam builders, who have shown themselves prepared to go ahead with as little as a couple of years of discharge data. Dam builders often build and hope for the hydrological best: there is therefore a pattern of overestimating annual flows and underestimating peak floods.

Overestimates of average flows mean that many dams fail to yield as much power and water as predicted. The huge Buendía–Entrepeñas reservoir in central Spain is formed by two dams built on the Guadiela and Tagus rivers in the late 1950s during General Franco’s decade–and–a–half dam–building binge. It has never been able to supply more than half the capacity of the aqueduct built to take its water to the Mediterranean coast. In early 1994 the reservoir contained just 17 per cent of its capacity.

Lower than expected rainfall and higher than expected leakage through its limestone bed have meant that Thailand’s largest volume reservoir, Srinakharin, completed in 1977, has never filled. During 1991, Thailand’s 25 largest dams contained a total of just under half of their combined usable capacity; the following year this figure fell to just over one third. Bhumibhol and Sirikit, World Bank–funded dams which impound the second and third largest reservoirs in Thailand, together contained only seven per cent of their total usable volume in March 1994. In their authoritative 1973 history of the World Bank, Edward Mason and Robert Asher of the Brookings Institution state:

The electricity–generating capabilities of the Ping River, on which the Bhumiphol Dam was built, were substantially overestimated... Taking into account the shortfall in the generating capacities of the Bhumiphol Dam and the... disappointments in the project’s contribution to agricultural output, a current reassessment of this sizeable multipurpose development effort would probably call into question its economic justification.

The 1922 Colorado River Compact laid the legal groundwork for the development of Hoover and the many other dams on the river. The writers of the Compact assumed on the basis of 18 years of streamflow measurements that the annual flow of the river averaged 17.5 million acre–feet (one acre–foot equals the amount of water which would flood one acre to a depth of one foot – 21.6 km³). By the 1950s, however, it was apparent that the first two decades of this century had been exceptionally wet in the US Southwest. From 1930 to 1952, the Colorado’s flow averaged only 11.7 maf. Acceptance of this low flow figure would have brought to a shuddering halt the grandiose plans of the Bureau of Reclamation. In early 1953 the Bureau was lobbying Congress for funds to build its multi–dam Colorado River Storage Project, which was predicated on the 17.5 maf figure being conservative. Rather than scale back its ambitions, however, BuRec simply ignored the new data. BuRec did not admit to the Colorado’s "deficit" until 1965, when it conceded that the flow was likely to be around 15 million acre feet.

BuRec’s inability to accept inconvenient streamflow data is paralleled by the refusal of the authorities building Sardar Sarovar to accept the overwhelming evidence that much less water is likely to be available than was assumed when the project was planned. SSP was designed in the 1970s on the assumption that over 27 maf of water flowed down the Narmada three out of four years. Yet in 1990 the 42 years of flow data then available gave a three out of four years’ discharge past the dam site of just 22.7 maf. More recent figures indicate the flow may be even lower.