The water footprint of electricity from hydropower

Introduction Conclusions References


Introduction
The need to supply a growing population with sufficient fresh water in the context of increasing water scarcity and declining water quality has brought sustainable water resources management to the forefront of the global development agenda.For centuries, dams have played a key role in human development, bringing about significant social and economic improvements, but also having significant impacts on local communities and ecosystems.About 30-40 % of irrigated land worldwide relies on water stored behind dams (World Commission on Dams, 2000) and hydropower accounted for 16 % of world electricity in 2008 (IEA, 2010).
Large hydropower dams have both positive and negative effects (Sternberg, 2008(Sternberg, , 2010)).Dams have been built to regulate river flows, store water to guarantee adequate supply of water in dry periods, control floods, irrigate agricultural lands, provide for navigation and to generate electricity.Negative impacts associated with the building of large dams include displacement of people, loss of land and alteration of river flows and water quality affecting downstream people and ecosystems (Gleick, 1993;Rosenberg et al., 1995;Poff et al., 1997;Scudder, 1997;Lerer and Scudder, 1999;Tilt et al., 2009).Worldwide, many countries are likely to continue depending on hydroelectric dams as their source of electricity.But such development should be in a manner which addresses environmental concerns and the question how water resources can best be allocated.
It has been debated whether hydroelectric generation is merely an in-stream water user or whether it also consumes water, in the sense of effectively taking away water from the river (Cooley et al., 2011).In the World Congress organized by the International Hydropower Association, 14-17 June 2011 in Brazil, a special session was devoted to the question: does hydropower consume water?The session explored different interpretations of water "consumption" in an attempt to recognize the energy impacts on water (Aguilar et al., 2011).In this paper we provide scientific support for the argument that the production of hydroelectricity is in most cases a significant water consumer.
As an indicator of water consumption of hydroelectricity we use the concept of the water footprint, which measures the volume of freshwater consumed and polluted to produce the product along its supply chain.The water footprint of a Published by Copernicus Publications on behalf of the European Geosciences Union.
product is equal to the sum of freshwater consumed or polluted divided by the quantity of production of the product (Hoekstra and Chapagain, 2008;Hoekstra et al., 2011).The water footprint consists of three components: the green water footprint (consumptive use of rainwater), the blue water footprint (consumptive use of ground or surface water) and the grey water footprint (the volume of water polluted).The analysis in this paper is restricted to the quantification of the blue water footprint of hydroelectricity and focuses on the consumptive use of water that relates to the evaporation from the artificial reservoirs that are created behind hydroelectric dams.
Storage of water behind large hydropower dams leads to consumptive water use through evaporation from the open water surface of the artificial lake.Gleick (1993) has shown that between 0.01 and 56 m 3 GJ −1 , or on average 1.5 m 3 of water per GJ of electricity produced is evaporated from hydroelectric facilities in California.In a recent study for New Zealand, Herath et al. (2011) estimated the water footprint of hydro-electric generation for eight plants and found values between 0.8 and 32 m 3 GJ −1 .In another recent study, Pfister et al. (2011) report values between 0.3 and 170 m 3 GJ −1 based on a few cases from the USA, Switzerland and Tanzania.By combining the estimate of global evaporation from artificial water reservoirs in the world from Shiklomanov (2000) with data on global hydroelectric generation from Gleick (1993), Gerbens-Leenes et al. (2009a) estimated that the global average blue water footprint of electricity from hydropower is 22 m 3 GJ −1 .
The objective of the current study is to estimate the blue water footprint of hydroelectricity for 35 selected reservoirs.First we estimate the evaporation throughout the year for the selected reservoirs.Next, we calculate the water footprint of hydropower based on the annual evaporation rate and energy generated.We have considered both the theoretical maximum and the actual hydroelectric generation of the plant.The theoretical maximum hydroelectric generation refers to the energy that could be generated with 100 % hydropower availability.Since this theoretical maximum is not realistically attainable, comparisons among the hydropower plants and further discussion of the water footprint will be based on the actual energy generation.
The selection of the hydropower plants has been largely arbitrary and mostly based on the availability of data.The selected plants are shown in Fig. 1.All plants selected have been primarily built for the purpose of hydroelectric generation, although some serve other purposes as well.With the exception of the largest hydropower plants such as Itaipu, Tucurui, Sayano Shushenskaya, Robert-Bourossa, Yacyreta and Cahora Bassa all hydropower plants selected are the ones included in World Bank (1996).The 35 hydropower plants have a total capacity of about 73 GW and represent 8 % of the global installed hydroelectric capacity of 924 GW in 2007(IEA, 2010).

Method and data
The water footprint of electricity (WF, m 3 GJ −1 ) generated from hydropower is calculated by dividing the amount of water evaporated from the reservoir annually (WE, m 3 yr −1 ) by the amount of energy generated (EG, GJ yr −1 ): The total volume of evaporated water (WE, m 3 yr −1 ) from the hydropower reservoir over the year is: where E is the daily evaporation (mm day −1 ) and A the area of the reservoir (ha).Data on installed hydroelectric capacity, actual hydroelectric generation and reservoir area were obtained from the World Bank (1996).For some hydropower plants data were obtained from Goodland (1997) and other sources.Data on reservoir water holding capacity were obtained mainly from Chao et al. (2008).
There are a number of methods for the measurement or estimation of evaporation.These methods can be grouped into several categories including (Singh and Xu, 1997): (i) empirical, (ii) water budget, (iii) energy budget, (iv) mass transfer and (v) a combination of the previous methods.
Empirical methods relate pan evaporation, actual lake evaporation or lysimeter measurements to meteorological factors using regression analyses.The weakness of these empirical methods is that they have a limited range of applicability.The water budget methods are simple and can potentially provide a more reliable estimate of evaporation, as long as each water budget component is accurately measured.However, owing to difficulties in measuring some of the variables such as the seepage rate in a water system the water budget methods rarely produce reliable results in practice (Lenters et al., 2005;Singh and Xu, 1997).In the energy budget method, the evaporation from a water body is estimated as the difference between energy inputs and outputs measured at a site.Energy budget methods are considered to be the most reliable in theory (Lenters et al., 2005;Singh and Xu, 1997), but require costly instrumentation and a large commitment of personnel for field work and data processing (Winter et al., 1995).The mass-transfer (aerodynamic) based methods utilize the concept of eddy motion transfer of water vapour from an evaporating surface to the atmosphere.The mass-transfer methods normally use easily measurable variables and give satisfactory results in many cases.However, measurement of wind speed and air temperature at inconsistent heights, have resulted in a large number of equations with similar or identical structure (Singh and Xu, 1997).The combination methods combine the mass transfer and energy Owing to its limited empirical basis, the Penman-Monteith equation is more readily applicable to a variety of water bodies.In addition, the model takes into account heat storage within water bodies.Therefore, for the purpose of the current study the Penman-Monteith equation with heat storage is considered suitable for the estimation of evaporation from the selected hydropower reservoirs.
The evaporation from the water surface (E, mm day −1 ) is estimated using the Penman-Monteith equation with an inclusion of water body heat storage.This equation is written as (McJannet et al., 2008): where E is open water evaporation (mm day −1 ); λ the latent heat of vaporization (MJ kg −1 ); w the slope of the temperature saturation water vapour curve at water temperature (kPa • C −1 ); R n net radiation (MJ m −2 day −1 ); G the change in heat storage in the water body (MJ m −2 day −1 ); f (u) the wind function (MJ m −2 day −1 kPa −1 ); e w the saturated vapour pressure at water temperature (kPa); e a the vapour pressure at air temperature (kPa); and γ the psychrometric constant (kPa • C −1 ).The full description of the method used is presented in the Supplement.
Daily values of mean air temperature, dew point temperature and wind speed for the selected meteorological stations were obtained from NCDC (2009).The daily data for the years 1996-2005 were averaged in order to fill missing values and smooth out some inconsistencies in the data.Monthly values of cloud cover and percentage of maximum possible sunshine with a spatial resolution of 10 arc minute were obtained from the CRU CL-2.0 database (New et al., 2002).The cloud cover and sunshine duration were available only as monthly averages for the period 1961-1990.Therefore the monthly average values were used as daily values for each month of the year.
The water footprint of electricity from hydropower is compared with the water footprint of electricity from combustion of primary crops.The latter has been calculated per type of crop by first multiplying the water footprint of the primary crop in m 3 ton −1 from Mekonnen and Hoekstra (2011) by the harvest index for that crop to get the water footprint in m 3 per ton of total biomass harvested.Harvest indices were taken from Gerbens- Leenes et al. (2009a, b).Next, the water footprint of total biomass was divided by the bio-electricity output per unit of crop (GJ ton −1 ) as reported by Gerbens-Leenes et al. (2008).

Results: the water footprint of hydroelectricity
The aggregated blue water footprint of the 35 selected hydropower plants is 90 Gm 3 yr −1 , which is equivalent to 10 % of the blue water footprint of global crop production in the year 2000 (Mekonnen and Hoekstra, 2011;Fader et al., 2011).The total blue water footprint of hydroelectric generation in the world must be considerably larger if one considers the fact that this study covers only 8 % of the global installed hydroelectric capacity.The annual evaporation from hydropower reservoirs depends on both climate (which determines the evaporation from the water surface in mm yr −1 ) and reservoir area.
The water footprint of electricity from hydropower for the 35 selected hydropower plants is presented in Table 1.The average water footprint of electricity from hydropower for the selected plants is 68 m 3 GJ −1 .There is a large variation in water footprint among the different power plants, ranging from 0.3 m 3 GJ −1 for San Carlos in Colombia to 846 m 3 GJ −1 for Akosombo-Kpong in Ghana.
Most of the reservoirs show an evaporation rate between 2000 and 3000 mm yr −1 .Reservoirs in the (sub)tropics have generally a higher evaporation rate than reservoirs in temperate regions.The surface water evaporation varies from no more than 486 mm yr −1 from the Sayano Shushenskaya reservoir in Russia to 3059 mm yr −1 from the Cahora Bassa reservoir in the Zambezi River in Mozambique (Table 1).Minimum and maximum evaporation rates thus differ by a factor of six, which partially explains the differences between the water footprints of different hydropower reservoirs.The size of the reservoir surface in relation to the installed hydroelectric capacity, however, has a much bigger impact on the ultimate water footprint of hydroelectricity.While the average reservoir area per unit of installed capacity in the reservoirs studied is 83 ha MW −1 , the minimum is 0.26 ha MW −1 (San Carlos reservoir, Colombia) and the maximum 720 ha MW −1 (Akosombo-Kpong in the Volta River, Ghana).The total evaporation from a hydropower reservoir thus depends more on its size than on climate.This is illustrated in Fig. 2, which shows a more or less linear relationship between the water footprint of the power plants and ha MW −1 .Hydropower plants that inundate a large area per unit of installed capacity have in general a larger water footprint per unit of electricity generated than those that flood a small area per unit of installed capacity.
The largest hydropower plant in terms of installed hydroelectric capacity in this study, the Itaipu dam in the Paraná River at the border of Brazil and Paraguay, has a water footprint of 7.6 m 3 GJ −1 .The second-largest studied hydropower plant in terms of MW, the Guri reservoir in Brazil, has a water footprint that is close to the global average of 68 m 3 GJ −1 found in this study.The largest plant in terms of MW that has a water footprint far beyond the average found in this study is the Cahora Bassa dam in the Zambezi River in Mozambique, with a water footprint of 186 m 3 GJ −1 .

Comparison with the water footprints of other forms of energy
When we compare the water footprint of electricity from hydropower with the water footprint of electricity from other renewable sources, it appears that hydroelectricity has a relatively large water footprint per GJ.The blue water footprint of electricity from wind and solar energy is estimated to be well below 1 m 3 GJ −1 (Gerbens- Leenes et al., 2009a).
The blue water footprint of bio-electricity -when derived from combustion of the full biomass of primary cropsranges from zero to 40 m 3 GJ −1 , depending on which crop is used for comparison and to which extent it is irrigated.The 40 m 3 GJ −1 refers to bio-electricity from combustion of cotton, which is a rather theoretical example, because cotton is in practice not used for the purpose of electricity generation.Also other crops are rarely used for that purpose.More common feedstock for the generation of bio-electricity are crop residues, animal manure, wood wastes from forestry and industry, residues from food and paper industries, municipal green wastes and sewage sludge.In all those cases, the water footprint of bio-electricity will be much lower than the water footprint of bio-electricity from combustion of primary crops, because the water footprint of biomass is largely attributed to the primary product and not to the residues and waste (Hoekstra et al., 2011).Figure 3 compares the blue water footprint of electricity from hydropower with the total (green + blue + grey) water footprint of electricity from combustion of primary crops.For a fair comparison one should compare the blue water footprints.But even when comparing the total water footprints, bioelectricity from a number of crops -including sugar beet, sugar cane and maize -will have a smaller water footprint than hydroelectricity.In other words, one drop of blue water allocated for consumption for hydroelectric generation generally yields much less energy than one drop of blue water allocated for consumption in crop production for generating feedstock for bioelectricity.This is not to suggest that in general it is advisable to allocate water to grow crops for producing bioelectricity rather than to generate a much lower amount of hydroelectricity at the cost of the same volume of water.In many cases this alternative allocation is not a reasonable choice, or even impossible (e.g.due to the unavailability of suitable land).Besides, for such broad decisions as investing in different sectors, one needs to take into account all relevant economic, social and environmental factors, not the factor of water productivity alone.Also one should account for the fact that many hydroelectric dams are designed to serve other purposes as well.What we do want to argue, however, is that hydroelectric generation is generally a large water consumer and that in allocating water to hydroelectric generation it is advisable to explore the foregone costs by not allocating the water to alternative uses, either upstream or downstream of the location of a planned hydropower reservoir.Alternative uses include crop growing for bioelectricity, but more common alternatives are to allocate the blue water to grow crops for food, feed, fibers or biofuel or to let the blue water in the river to maintain environmental flows.

Sensitivity analysis
The sensitivity of the calculated lake evaporation to errors in input data was tested by varying the following parameters: air temperature, wind speed, water depth and lake area.On average, for the lakes studied, a variation of the water depth by ±50 % has little effects on the evaporation (±1.6 %).Variation in wind speed has even lower effects on the annual lake evaporation.This is in agreement with the finding of Xu and Singh (1998).Therefore, possible data errors in these two parameters have very little effect on our final result.However, the effects of changes in air temperature and lake area are quite significant.On average, for the lakes studied, variation of the air temperature by ±10 % results in an increase or decrease of the evaporation by almost an equal percentage.The total evaporation amount per year is equal to the product of the evaporation rate (mm) and surface area.Thus, an error in the surface area of the lake by 10 % leads to an error in total evaporation of 10 % as well.
Besides, there could be errors associated with the use of the climatic data measured over the land surface instead of over the water surface.Although the model applied is shown to estimate the water temperature reasonably well when verified against measured data (Keijman, 1974;De Bruin, 1982;Finch, 2001), we may not fully capture the actual water temperature, particularly at a daily time step, as there will be some time lag between the equilibrium temperature (the temperature at which the net rate of heat exchange equals zero) and the water temperature.This may lead to some error in the annual evaporation estimate.

Discussion
The water footprints of the artificial reservoirs analyzed in this study were fully attributed to hydroelectric generation, even though some of the reservoirs serve other purposes as well, such as flood control and irrigation.We justify this choice by the fact that all selected hydropower dams and associated reservoirs were primarily created for hydroelectric generation.Future research could be directed towards the analysis of the water footprint of reservoirs created for storing water for irrigation or other purposes and on tackling the water footprint attribution issue when reservoirs are used for multiple purposes.
The model output is sensitive to a number of input parameters such as air temperature, wind speed and water body depth as shown in the previous section.Since climatic data at the dam site are available only for a few plants, data from the most nearby climatic stations have been used, some of which are a bit far from the reservoir.Due to the uncertainties in the input data, the data presented should be seen as indicative.The order of magnitude of the results, however, will not change with better data, so that the results are good enough to compare the water footprint of hydroelectricity with the water footprint of other forms of electricity and to make rough comparisons between the water footprints of different hydropower sites.
Most reservoirs have a varying water surface area over time, as a result of changes in water volume during the year and between years.The difference between minimum and maximum area relative to the maximum area over a multiyear period differs greatly across reservoirs.In this study we have used a fixed reservoir area as provided by World Bank (1996) and Goodland (1997).Since reported areas generally refer to the maximum, this can lead to some overestimation of evaporation over the year.It is very difficult to find good data of area changes over the year; future studies devoted to particular sites could improve this.
We have estimated the water footprint per reservoir by considering the total evaporation from the reservoir, whereas one could argue that before the reservoir was created there was evaporation from the area as well, probably not so much from the original flowing river (since in most cases the reservoir area is much larger than the original river water area) but possibly significant from the inundated land.However, here it is relevant to recall the definition and meaning of the water footprint.The water footprint is not meant to refer to additional evaporation (compared to some reference situation), but for quantifying the volume of water consumption that can be associated with a specific human purpose (Hoekstra et al., 2011).From this perspective, the full reservoir evaporation can be attributed to the purpose of the reservoir.
The consumptive water use of a reservoir has been quantified by considering total evaporation, even though one could hypothesize that a fraction of the evaporated water will return to the reservoir or the catchment of the reservoir.In this study, it has been assumed that evaporated water from a reservoir will not return to the reservoir or even the catchment in a significant way.Although land-use changes such as building a reservoir can influence climate at a regional or continental scale through influencing moisture recycling (Van der Ent, 2010;Eltahir et al., 1996), this process is relevant on a larger scale than the catchment, so that most of the water evaporated from a catchment can generally be considered "lost" for reuse in the same catchment.
The study has been limited to the estimation of the evaporation from reservoirs, i.e. the so-called operational water footprint of hydroelectric generation.The study does not include an assessment of the supply-chain water footprint of hydroelectric generation, which is expected to be much smaller than the operational water footprint (Inhaber, 2004;Fthenakis and Kim, 2010).The supply-chain water footprint refers to the water footprint of producing the materials used in the construction and the operation and maintenance of the site.
The water footprint is a resource use indicator, not an ecological or social impact indicator.Dams are often associated with all sorts of ecological impacts (river fragmentation, effects on water quality and biodiversity) and social impacts (displacement of people).It is to be appreciated that the water footprint of hydroelectric generation refers to freshwater consumption related to hydroelectric generation; the water footprint is not an all-inclusive indicator that reflects all environmental and social impacts of a dam and needs to be complemented with other relevant resource use and impact indicators in order to provide a full understanding of all relevant issues.
Some authors have suggested to redefine the water footprint from a volumetric measure to a local environmental impact index, by multiplying volumes by impact factors, whereby impact factors are defined based on local water scarcity.In this way, one would obtain weighted volumes of water consumption (Pfister and Hellweg, 2009;Berger and Finkbeiner, 2010).However, highly relevant information is lost in this way, because knowing the volumes of water consumed for hydroelectric generation is important in the discussion about water resources allocation.Besides, it is doubtful whether weighting volumes based on water scarcity to obtain one simple impact index can do justice to the variety of local factors that determine the various sorts of environmental and social impacts that can occur as a result of water consumption from a reservoir or as a result of a dam in a broader sense.Therefore, we recommend to consider environmental and social impacts of a dam separately and in addition to the water footprint of a reservoir, acknowledging that the latter reflects water consumption only.The main consideration from Pfister and Hellweg (2009) and Berger and Finkbeiner (2010) behind the proposal to define the water footprint as an impact indicator rather than as a resource use indicator is that this would be in line with how the carbon footprint is defined.In the research field of Life Cycle Assessment, carbon footprint is indeed interpreted as an impact indicator, but wrongly so, as argued before (Hoekstra et al., 2011).The carbon footprint is a measure of the amount of greenhouse gases emitted to the environment from human activities and does not describe environmental impacts associated with the emission of these greenhouse gases.The water footprint is consistent with the ecological and carbon footprint; they all show pressures on natural resources or on the earth's assimilation capacity, not impacts (Hoekstra et al., 2009(Hoekstra et al., , 2011;;Hoekstra, 2009).

Conclusions
Hydroelectric generation has historically been considered as a non-consumptive water user; however, through the estimation of the blue water footprint of hydroelectricity at 35 sites, this study finds that hydropower is a large consumptive user of water.The amount of water lost through evaporation annually from the selected reservoirs is equivalent to 10 % of the global blue water footprint related to crop production.The 35 sites represent only 8 % of the global installed hydroelectric capacity.The study shows that the range of water footprint values for the different hydropower plants is very wide.Although local climate has an influence, the water footprint of hydroelectricity is largely influenced by the area flooded per unit of installed capacity.The water footprint linearly increases with the area flooded per unit of installed capacity.
The water evaporated from the reservoir is seldom taken into account in assessing the pros and cons of constructing dams for hydroelectric generation.This study demonstrates that accounting for water loss through evaporation is an additional consideration when evaluating the environmental, social and economic sustainability of a proposed dam or in the evaluation of hydropower as an energy source.We recommend that water footprint assessment is added as a component in evaluations of newly proposed hydropower plants as well as in the evaluation of existing hydroelectric dams, so that the consequences of the water footprint of hydroelectric generation on downstream environmental flows and other water users can be evaluated.
The water footprint of hydroelectric dams should be considered in the context of the river basin in which this water footprint occurs, because competition over water and possible alternative uses of water differ per basin.This study contributes new information that can be used in river basin planning and water allocation decisions.
Sustainable development of hydropower requires the accounting and internalization of all external costs including water consumption.Internalization means that the economic and environmental costs of the water consumed are charged to the operator of a hydropower plant and included in the price of hydroelectricity.It should thereby be acknowledged that water consumption costs vary within the year and across river basins, since the degree of water scarcity and competition over water depend on the period within the year and local circumstances.
The current study does not claim to be exhaustive in terms of showing both the beneficial and negative effects of hydropower.The study has been restricted to the estimation of the water footprint of different hydropower plants.
Environmental issues surrounding hydropower dams relate to, for example: physical, chemical, biological and geomorphological aspects of blocking a river; flooding of natural habitats and related loss of plants and animals; alteration of water flow regimes; and water quality problems due to the decay of submerged vegetation.On the other hand, hydropower is often perceived as a clean and cost-effective source of renewable energy.Many countries rely upon hydropower for a substantial portion of their electricity supply.Between 1973 and 2008, hydroelectric generation grew from 1295 TWh to 3288 TWh, which is a growth by a factor 2.5 (IEA, 2010).Further development of hydropower should take into account all the associated environmental and social costs.In this respect, the water footprint of hydroelectricity, i.e. the consumptive use of water, should be considered as one item in environmental impact assessment studies for newly proposed hydroelectric dams.

Fig. 2 .
Fig. 2. Relation between the water footprint of hydroelectricity and the flooded area per unit of installed hydroelectric capacity.

Fig. 3 .
Fig. 3. Global average water footprint of electricity from hydropower compared to the water footprint of electricity from combustion of primary crops.

Table 1 .
Water footprint of electricity for selected hydropower plants.The data are for the combined Akosombo-Kpong system.Kpong is a runoff power plant using Akosombo dam.Akosombo and Kpong generate 1020 MW and 160 MW, respectively.** Robert-Bourossa together with La Grande-2A use the Robert-Bourossa reservoir and generate 5616 MW and 2106 MW respectively.Energy generation of La Grand-2-A is assumed to be half of that of Robert-Bourossa. *