Policy Analysis pubs.acs.org/est

Addressing Biogenic Greenhouse Gas Emissions from Hydropower in LCA Edgar G. Hertwich* Industrial Ecology Programme and Department of Energy and Process Engineering, Norwegian University of Science and Technology (NTNU), 7491 Trondheim, Norway S Supporting Information *

ABSTRACT: The ability of hydropower to contribute to climate change mitigation is sometimes questioned, citing emissions of methane and carbon dioxide resulting from the degradation of biogenic carbon in hydropower reservoirs. These emissions are, however, not always addressed in life cycle assessment, leading to a bias in technology comparisons, and often misunderstood. The objective of this paper is to review and analyze the generation of greenhouse gas emissions from reservoirs for the purpose of technology assessment, relating established emission measurements to power generation. A literature review, data collection, and statistical analysis of methane and CO2 emissions are conducted. In a sample of 82 measurements, methane emissions per kWh hydropower generated are log-normally distributed, ranging from micrograms to 10s of kg. A multivariate regression analysis shows that the reservoir area per kWh electricity is the most important explanatory variable. Methane emissions flux per reservoir area are correlated with the natural net primary production of the area, the age of the power plant, and the inclusion of bubbling emissions in the measurement. Even together, these factors fail to explain most of the variation in the methane flux. The global average emissions from hydropower are estimated to be 85 gCO2/kWh and 3 gCH4/kWh, with a multiplicative uncertainty factor of 2. GHG emissions from hydropower can be largely avoided by ceasing to build hydropower plants with high land use per unit of electricity generated.



significant3 but are not considered here. In recent years, global assessments4,5 have increasingly relied on life cycle assessment (LCA) to evaluate greenhouse gas emissions and climate benefits of energy technologies. As the underlying LCA literature lacks a consistent treatment of bGHG emissions from hydropower, these emissions have been ignored in summary assessments, potentially misleading policy makers.6,7 Biogenic GHG emissions of hydropower are related to bacterial processes happening mostly in reservoirs which produce CO2, CH4, and N2O. Biogenic CO2 and CH4 are produced by the oxidation of organic carbon from biomass or detritus, organic carbon matter in soil, or sediments. The principal concern from a climate perspective is methane formation, because it is a stronger greenhouse gas. The organic carbon comes from the flooding of biomass and soil when the reservoir is filled (land use change), is transported to the reservoir by rivers, or grows in the reservoir.8 Nitrous oxide forms as part of the denitrification of nitrogen bound in organic

INTRODUCTION Hydropower remains the most important source of renewable electricity supply, providing 3288 TWh in 2009, or 6.1% of the global primary energy supply. The remaining technical potential is on the order of 10−15 thousand TWh per year; the current growth rate is about 3% per year.1 Important unutilized resources are concentrated in regions with a potential for accelerating economic development, such as Africa and South America, potentially leading to an acceleration of hydropower construction in the future. Important drivers for hydropower deployment are energy security, development, and climate change mitigation. Through clean development mechanisms (CDM), Annex B countries provide funding for hydropower projects in developing countries in order to reduce greenhouse gas emissions. International Rivers counts almost 2500 hydropower projects currently in the CDM pipeline, with a total capacity of 244GW; certified emissions reductions have been issued for 642 projects totaling 132 million tonnes (Mt) CO2.2 The climate benefit of hydropower, however, is currently poorly understood due to the role of biogenic greenhouse gas (bGHG) emissions, albedo changes, and changes in evaporation, affecting the latent heat flux to the atmosphere. Currently, CDM projects must have a power density above 4 W per m2 reservoir to limit these effects. Physical effects may be © 2013 American Chemical Society

Received: Revised: Accepted: Published: 9604

April 25, 2013 August 2, 2013 August 2, 2013 August 2, 2013 dx.doi.org/10.1021/es401820p | Environ. Sci. Technol. 2013, 47, 9604−9611

Environmental Science & Technology

Policy Analysis

terrestrial landscape. Of these, 230 Tg are buried in sediments, 750 Tg or more are released to the atmosphere, and 900 Tg are delivered to the ocean. There is a substantial uncertainty, and ref 18 estimates an annual input of 2.9 Pg to freshwaters, a sedimentation of 0.6 Pg, and atmospheric emissions of 1.4 Pg. A further increase in the estimated role of freshwater in the carbon cycle may potentially emerge once new measurements from Asia are fully considered.19 On the other hand, models for the nutrient export of rivers20 indicate carbon flows in line with Cole et al. For comparison, the net primary production of terrestrial plants accounts for ca. 60 PgC/y, while total fossil fuel emissions account for 7.7 PgC/y. The carbon flow in rivers consists of dissolved inorganic carbon (DIC, i.e., CO2, carbonic acid and its dissociated forms, ca. 0.3 PgC/y), dissolved organic carbon (DOC, like humic acids, ca. 0.2 PgC/y), and particulate organic carbon (POC, i.e., dead plant material, ca. 0.1−0.4 PgC/y). Groundwater flow directly to estuaries delivers ca. 0.2 PgC/y.17,20 The ultimate fate of the organic carbon transported by rivers to the coastal or open ocean is not yet well investigated. Much of the organic carbon is mineralized, but some of the particulate organic carbon will sediment.21 If marine vegetation is a guide, about 25% of the organic carbon is transported from the upper layers of the ocean to the deep ocean where it stays removed from the short-term carbon cycle. In order to understand the effect of hydropower dams on the climate, we need to understand how the dams change the global carbon cycle by increasing sedimentation22 and decay in freshwater bodies and reducing the transport of organic carbon to oceans and its sedimentation or decay there. The quantity, timing, and oxidation state of carbon released to the atmosphere are of interest. Methane Balance. Global methane concentrations have increased by almost 150% since the onset of the industrial revolution and contribute 20% to increased radiative forcing from greenhouse gases. Lakes and rivers play an important role in the methane balance of the atmosphere. Methane constitutes around 4% of the carbon released from lakes.23 Global methane emissions estimates have large ranges because both freshwater area and emissions rates are uncertain. In a review of the methane balance literature, Kirschke et al.24 estimate the contribution from wetlands to be on the order of 200 TgCH4/y and specify the freshwater component as 40 TgCH4/y. Bastviken et al.23 estimate emissions from freshwater lakes and rivers as 100 TgCH4/year. By comparison, the total natural and anthropogenic emissions are on the order of 600 TgCH4/y. Methane emissions are estimated either with bottom-up methods, measuring emissions rates at selected sites and upscaling with the area, or top-down, measuring concentration gradients and relying on inverse modeling of atmospheric processes to specify emissions sources required to produce the measured concentrations. The measurement of carbon isotope ratios is used to assign emissions to source categories.24 Humans interfere with the natural methane balance in several ways. Increased populations of ruminants (cows, sheep) who produce methane through etheric fermentation, rice paddies, leakage from fossil fuel systems, increased biogenic carbon input to freshwaters through soil erosion and eutrophication are the most important causes for increased methane emissions.

matter. N2O emissions have shown to be relatively minor for boreal reservoirs.8 The aim of this paper is to inform the comparative assessment of energy technologies about bGHG emissions of hydropower per unit of electricity delivered. As a result, the wider importance of bGHG emissions of hydropower is illuminated, and recommendations for the consideration of bGHG in LCA are derived. The paper goes beyond reviewing existing LCA studies.4,9,10 The paper consists of a review of the scientific literature on biogenic GHG emissions from dams and an interpretation of insights provided by this literature from the perspective of LCA. A data set of methane emissions per kWh is assembled and analyzed, providing an insight into variables driving the emissions of power plants. Recommendations are provided for the assessment of bGHG in future hydropower LCAs. In its Special Report on Renewable Energy (SRREN), the IPCC did a thorough job in reviewing the environmental aspects of energy technologies, relying also on LCA.4,9 While biogenic GHG emissions were discussed both for bioenergy and hydropower, they were left out of the comparison charts presented in the summary. The hydropower chapter in SRREN9 presents a detailed discussion of mechanisms for biogenic emissions, but the emissions rates were provided per reservoir area, not referred to the electricity generated. The IPCC’s review of hydropower LCAs identified 27 estimates of life-cycle GHG emissions from 11 distinct references, including refs 11−15. Sixteen estimates from seven references include bGHG emissions, and only three estimates from two references include emissions from the decommissioning phase.9 The assessment in SRREN mixed LCAs that consider and ignored biogenic GHG emissions, and the number of cases is too small to be representative. Similarly, a recent review of hydropower LCAs16 did not deal systematically with biogenic emissions. Biogenic emissions were identified with land use change, which is only partially correct. Also, the concepts of gross versus net emissions used in the environmental science literature were not explained in ref 16, potentially leading to misunderstandings. The present paper is structured in the following manner: First, I present an overview over the role of rivers in the global carbon cycle and identify the mechanisms by which dams interfere with this carbon cycle and thus affect the concentration of greenhouse gases in the atmosphere. Second, I discuss the mechanisms and pattern of biogenic CH4 and CO2 emissions and review their measurement. Third, I report the results of a statistical analysis of measured emissions per kWh of power generated and develop a regression equation that can be used, in the absence of measurements, to estimate emissions from other hydropower projects. Fourth, I provide a tentative estimate of global emissions per unit electricity generated. Finally, I discuss the need for further assessments and give recommendations for how to conduct such assessments.



FRESHWATER AND THE GLOBAL CARBON CYCLE Flows of Carbon. While often neglected, rivers and lakes have an important role in the global carbon cycle (Figure S1). Freshwater is both the recipient of organic matter from soil and terrestrial biomass and the location for further biomass growth, fixing CO2 from the atmosphere.17,18 Rivers transport carbon to the ocean in organic and inorganic forms, and freshwater returns part of the carbon to the atmosphere. In an initial assessment of the carbon flows through freshwater, Cole et al.17 estimate that freshwater bodies receive 1900 TgC/y from the



CLIMATE IMPACT OF HYDROPOWER Changing the Flow of Carbon. Hydropower dams interfere with the carbon cycle in several ways, potentially changing the carbon cycle, the storage of carbon in sediments 9605

dx.doi.org/10.1021/es401820p | Environ. Sci. Technol. 2013, 47, 9604−9611

Environmental Science & Technology

Policy Analysis

Figure 1. Possible pathways for biogenic methane and carbon dioxide emissions from hydropower stations (from ref 9, Figure 5.16).

estimated based on typical emissions factors for the original land cover.8 Field measurements have been conducted for the Eastman 1 reservoir in Quebec, Canada.29,30 Note that emissions before the flooding may be positive (e.g., if a wetland area is flooded), negative (e.g., if a forest or agricultural area is flooded), or close to zero. A significant portion of the CO2 emissions from a hydropower reservoir represents carbon that also in other conditions would be oxidized and returned to the atmosphere, either from the river or the ocean. In addition, dams lead to sedimentation in the reservoir and may reduce sedimentation downstream and the ocean; they hence affect the future fate of carbon and the likelihood and timing of the return of the carbon to the atmosphere. Ideally, a complete modeling or empirical assessment of the effect of the life cycle of the dam on the carbon cycle, including the transport of carbon to the ocean, should be combined with pulse-response function modeling31 to consistently assess the climate impact of a dam. The strong spatial and temporal variation of emissions, including the importance of rare events,32 pose challenges to modeling and measurement. Further, rivers are often changed by a series of dams and the impacts of one dam interact with those of increased nutrient input and soil erosion,33−35 creating a situation where the total impacts on the carbon cycle are difficult to attribute to a single cause. There is a need to better understand the anthropogenic impacts on the freshwater− carbon cycle, including its interaction with the ocean. Without a detailed before-after investigation of the freshwater−carbon flux and deposition, including its transport to and fate in the ocean, I recommend neither to attribute reservoir CO2 emissions to hydropower nor to credit hydropower for carbon burial in reservoir sediments. Measurement of GHG Fluxes. In this section, I describe emission mechanisms and emission measurement. There is a lively debate in the literature and disagreements on important issues.7,36−38 Empirical studies need to address a number of emission sources (Figure 1):

or the deep ocean, and the form in which carbon is returned to the atmosphere, i.e., as CO2 or methane. The following processes are relevant: 1) Dams turn land or wetland surface into reservoirs, often inundating plants and soils, resulting in the submersion of organic carbon. Labile carbon is slowly released in the first 10− 15 years of the reservoir, either as CO2 or methane, depending on climate and reservoir characteristics, as well as the amount and character of the organic carbon.25 Dams also turn river surface into reservoir surface, increasing the fraction of carbon released as methane rather than CO2 due to the frequent formation of an anoxic bottom layer. 2) Dams interfere with the river transport of organic matter to the oceans, leading to sedimentation or decay of some of this organic matter.22 Dams reduce the flow of POC to the ocean and hence potentially its removal to the deep ocean. Reservoirs often have anoxic conditions, leading to the anaerobic digestion of organic carbon to methane. Dams also lead to the build-up of organic matter, which can be beneficial if it permanently increases the carbon stored in sediments or can be detrimental if dam removal or dredging leads to methane emissions.26 3) Dams often lead to increased surface water temperature and provide an opportunity for the growth of freshwater biomass and the absorption of nutrients. These conditions can lead to a net absorption of carbon and the sedimentation in the lake or transport downstream. To understand the impact of hydropower dams on greenhouse gas concentrations, it is hence important to understand the effect of the dam on the carbon cycle both before and after the dam is built. We need to understand how much carbon is extracted from or returned to the atmosphere, in what form, and when. Measurements of emissions after the completion of the dam measure what the literature calls gross emissions, which is the total flux of carbon to or from the surface.27 Note that these gross emissions constitute a net effect of emissions and absorption of carbon. Akin to consequential analysis in LCA,28 the hydropower literature identifies as ‘net emissions’ the difference in areal emissions before and after the flooding of the reservoir.27 Preflooding emissions are often 9606

dx.doi.org/10.1021/es401820p | Environ. Sci. Technol. 2013, 47, 9604−9611

Environmental Science & Technology

Policy Analysis

the reservoir varied from 5 to 343 with an average value of 47 mg C m−2 d−1. The flux through the dam showed a similarly large variation. The measurements were integrated to determine total flows. Emissions from the reservoir amounted to 34 GgC/y, degassing at the turbine outflow to 34 GgC/y, while diffusive emissions within 30 km downstream from the plant were 5 GgC/y.46 CO2 emissions were 2.4 TgC/y from the reservoir and 0.08 TgC/y downstream.47 The measurements show that while CO2 emissions mostly occur directly from the reservoir, for methane, emissions downstream of the power plant are equally important as those upstream. With gross emissions of 100 gCH4 and 8.5 kg CO2 per kWh, Balbina is worse than coal power. Similar measurements were also conducted in for the Petit Saut power station in French Guyana, which produces 560 GhW/y8 and releases 370 Gg CO2 and 35 Gg CH4,39 equivalent to 0.65 and 1.55 kg CO2e/ kWh. A counterexample is provided by the Nam Ngum Reservoir in Laos, for which a net absorption of CO2 during the measurement period was reported, in addition to small methane fluxes and no downstream degassing.50 The data are not referred to electricity production. For the upstream Nam Leuk reservoir in the study, the emissions are on the order of 0.05−0.1 kg CO2e/kWh. The carbon budget presented points to a significant accumulation of carbon in the sediments. Downstream emissions of methane dissolved in the water that streams across the dam either through the turbine or the weir have been measured only for Balbina and Petit Saut. For the Brazilian Tucurui51 and Curua Una52 reservoirs, methane emissions have also been estimated to exceed 1 kg CO2e/kWh, even though there is some controversy about these estimates. Demarty and Bastien review a large range of measurements and estimates of methane emission from reservoirs in tropical regions, indicating that emissions can have a large range, from only 2 to 4000 g CO2e/kWh.8 Important determinants include the amount of vegetation that is inundated, the inflow of organic carbon, and the size of the reservoir in relationship to the energy generation. In general, large reservoir areas lead to high methane emissions.8,53 Currently, analysis to evaluate the methane emissions from the Three Gorges dam in China is ongoing,54,55 but a complete balance has not yet been accomplished.

1) Diffusion of CO2, CH4, and N2O across the air−water interface. The gas flux depends on a number of variables such as wind speed, rainfall, temperature, and relative gas concentrations in air and water. Diffusive emissions can be directly measured using surface floating chambers39,40 or derived from boundary layer models.41 2) Bubble emissions: Methane produced through anaerobic digestion in sediments leads to bubbling. Temperature and hydrostatic pressure affect the bubbling rate. Bubbles come as bursts and not as a steady flow, with uneven bubbling events estimated contain a significant proportion of the total amount of methane released.42,43 Gas transport can also be mediated by aquatic plants, macrophytes.9 Methane bubbles are usually measured using funnels44 or eddy covariance45 or echosounders.43 3) Downstream emissions: Water in hydropower plants is often drawn from some depth in the reservoir. At this depth, the pressure is higher and the temperature is lower, so the solubility of gases is higher than at atmospheric pressure and temperature. Water leaving the turbine can hence be supersaturated with gases. Part of the methane is released directly at the hydropower plant after the water has passed through the turbines. This is called degassing. Another part of the methane is released from supersaturated water through diffusion or bubbling some distance from the dam.40,46 Downstream emissions are often neglected or underestimated.7,8,47 It is important to understand that emissions vary in time and space due to spatial features, weather, and seasonal effects.42,48 A proper assessment has to address these variations through measurements that extend over seasons,8 ideally supplemented with modeling exercises that serve to build a mechanistic understanding of the processes involved.44,49 Emissions also change over the lifetime of the reservoir. It is well recognized that emissions are highest in early years and decrease as the initially flooded biomass has decayed. This feature is nicely illustrated by measurements at the Petit Saut reservoir in French Guyana (Figure S2), where carbon dioxide emissions decreased by a factor of 3 and methane emissions by a factor of almost 5 from the average of the first three years to measurements 10 years later.39 Further measurements indicate that after 10 years, the reservoir reaches a steady state with constant emissions.8 The decreasing emissions reflect the degradation of initially present labile biomass and soil organic carbon. The stabilization of emissions is due to the depletion of that initial reservoir; the remaining emissions are mostly from organic carbon transported to the reservoir by tributaries or from plants grown in the reservoir.



EMISSION FACTOR ESTIMATION Based on ca. 150 available measurements of the flux of CO2 and CH4 from reservoirs, Barros et al.56 estimated the total global emissions from reservoirs. The strategy was to use emissions measurements per surface area to derive a regression equation. This equation was then used to estimate average emission rates for different climate zones, which were multiplied by the reservoir area to obtain estimates for total emissions. The study neglected emissions from downstream of the reservoirs, and some measurements do not include bubbling. The regression equation was applied to the logarithm of flux estimates, an operation that is not mean-preserving and systematically underestimates the total. It is akin to taking the geometric mean instead of the arithmetic mean, resulting in a factor of 5 difference for the given data set of methane emissions. We have supplemented Barros’ data with information on electricity generation (from various Internet sources) and the potential net primary productivity of the area57 as additional variables and reanalyzed the data set (Table S1). Figure 2 shows the frequency distribution of measurements of methane



HYDROPOWER DAMS IN TROPICAL REGIONS A review of measurements presented by IPCC SRREN9 indicates that the gross emissions of both methane and CO2 are temperature dependent and thus potentially much higher in tropical regions than in temperate and boreal regions. The temperature dependence is confirmed by seasonal comparisons across existing reservoirs.49 Kemenes et al.46 provide a detailed account of the Balbina reservoir in the Brazilian Amazon, which was built in 1987 and has an average area of 1770 km2 and an installed hydroelectric capacity of 250MW. Measurements were conducted of methane emissions from the reservoir, below the reservoir, and concentrations of methane flowing through the turbines and floodgates throughout most of 2005. The diffusive emissions of 9607

dx.doi.org/10.1021/es401820p | Environ. Sci. Technol. 2013, 47, 9604−9611

Environmental Science & Technology

Policy Analysis

Figure 2. Distribution of CH4 emissions rates per kWh of electricity produced across the sample of hydropower stations. Figure 3. Methane emissions per unit electricity produced from hydropower dams as related to the reservoir area, potential net primary productivity, and age of the reservoir area. The lower right-hand panel shows measured versus predicted values based on a trivariate regression. The figure indicates that land use is the most important explanatory variable.

emissions per kWh electricity distributed. The measurements have an unweighted mean of 54 gCH4/kWh, a geometric mean of 0.61 gCH4/kWh, and a geometric standard deviation of 46. The weighted mean is 3.5 gCH4/kWh. The highest emissions rate is from the Lokka power plant in Finland58 which sits partly on peat and produces only 0.5GWh/y, but it regulates the flow of a river with a chain of power stations. The water released from the reservoir produces ca. 365 GWh/y, which I used as power generation from the reservoir in the further analysis. Across our sample, power station capacity factor, latitude, NPP, land use (m2y/kWh), and the measurement of bubbling emissions show certain correlations with methane emissions (Tables S2 and S3). Given the log-normal distribution, one may desire to predict the logarithm of the emissions factor; while not preserving the mean the estimate is more likely to be closer to the actual value for the individual power plant than an estimate based on a linear regression. The best prediction of the emission factor (gCH4/kWh) is offered by including the logarithms of land use and NPP as well as age as linear variable (adjusted r2 = 0.78). Regression coefficients and statistics are shown in Table 1. Figures 3 and S3 relate

latitude and NPP; both variables, however, are strongly correlated and hence cannot be seen as independent. A multiple regression analysis of logarithms indicates that the inclusion of age and capacity factor of the power plant or the dummy variable for the inclusion of bubbling emissions improve the prediction offered by NPP/latitude as measured by the adjusted r2, from 0.19 to 0.28. The best estimate is that including the measurement of bubbling emissions doubles the methane flux and emissions factor.



A GLOBAL ESTIMATE OF CH4 EMISSIONS Given that reservoir area is strongly correlated with methane emissions, global emissions from hydropower reservoirs are best predicted based on the reservoir area. Given the lack of information on location and age of all dams, I use climate zones as a proxy. The resulting estimate of total CH4 emissions from reservoirs (Table 2) is 2.5 times as high as the estimate provided by Barros et al.56 and similar to that of Maeck et al.22 Correcting for bubbling emissions where these were ignored would raise methane emissions estimate further to 10 TgC/y. Relating the estimates in Table 1 to a total hydropower production of 3288 TWh in 2009, the average direct emissions from hydropower dams corresponds to 3 gCH4 per kWh, close to our sample average. One should emphasize that there is a large degree of uncertainty associated with the estimate presented in Table 2,

Table 1. Regression Coefficient and Statistical Parameters for the Regression of per-kWh Emissions As a Function of Land Use, Age, and Potential Net Primary Productiona const BLand use BAge BNPP r2 F p

CO2 emissions per kWh

CH4 emissions per kWh

0.8 (−1.8 to 3.8) 0.97 (0.84−1.11) −0.006 (−0.011 to −0.0009) 0.737 (−0.16 to 1.64) 069 77.5

Addressing biogenic greenhouse gas emissions from hydropower in LCA.

The ability of hydropower to contribute to climate change mitigation is sometimes questioned, citing emissions of methane and carbon dioxide resulting...
1MB Sizes 0 Downloads 0 Views