Reservoir Surfaces as Sources of Greenhouse Gases to the Atmosphere: A Global Estimate

Following flooding of landscapes to create any kind of reservoir, terrestrial plants die and no longer assimilate carbon dioxide (CO2) by photosynthesis (Figure 1), resulting in the loss of a sink for atmospheric CO2. In addition, bacteria decompose the organic carbon that was stored in plants and soils, converting it to CO2 and methane (CH4), which are then released to the atmosphere. All of the reservoirs examined to date emit CO2 and CH4 to the atmosphere (Table 1), but different landscapes contain different amounts of stored organic carbon in soils and vegetation (Schlesinger 1997), and so the potential for gas production and loss varies from site to site. For example, in the boreal region of Canada, a worst-case scenario is flooded peatlands because they possess a large store of organic carbon held in peat, which can decompose and be returned to the atmosphere as greenhouse gases over a long period (Kelly et al. 1997). Reservoirs that flood peatlands probably emit more greenhouse gases in the long term than reservoirs created over upland boreal forests, which have thin soil layers and no peat deposits. The first studies of greenhouse gas fluxes from reservoirs focused on hydroelectric generation (Rudd et al. 1993, Kelly et al. 1994, Duchemin et al. 1995) because it was, and still is, widely viewed as a carbon-free source of energy (Hoffert et al. 1998, Victor 1998). This view likely originated because before 1994, there were no data available on CO2 and CH4 emissions from reservoirs, even though it was well known that oxygen depletion resulting from active decomposition of flooded organic matter was common in waters of newly constructed reservoirs (Baxter and Glaude 1980). The first discussion of greenhouse gas emissions from reservoirs (Rudd et al. 1993) pointed out that greenhouse gas production per unit of power generated (e.g., in kWh) is not zero and should depend on the amount of organic carbon flooded to create the electricity. For example, reservoirs that flood large areas to produce few kWh, such as those built in areas with low topographical relief, would produce more greenhouse gases per kWh than reservoirs built in canyons where little area is flooded and large amounts of electricity are produced. A more recent concern is the global impact of reservoir construction on greenhouse gas emissions. Most reservoirs are developed not for hydroelectric production but rather for other purposes, including flood control, water supply, irrigation, navigation, recreation, and aquaculture (ICOLD 1998). The determination of the global effect of all types of reservoirs on the atmosphere requires two general pieces of information: flux measurements from reservoirs that vary in amount of organic carbon flooded, age, and global distribution; and the surface area of all reservoirs around the world. In this article, we examine the range of greenhouse gas fluxes available to date from both temperate and tropical reservoirs and assess the quantity and quality of data available for the global surface area of reservoirs. Although there are uncertainties in both flux and surface area information, it is important to ask the question: Could these emissions be significant on a global basis, and should we be improving our knowledge of this aspect of reservoir development? Initial calculations indicate that, globally, these emissions may be equivalent to 7% of the global warming potential of other documented anthropogenic emissions of these gases. This percentage is similar to contributions from other currently inventoried sources. As a result, we argue that these fluxes should be included in greenhouse gas inventories by country and in models of global carbon cycling. Fluxes of greenhouse gases from water surfaces can be quantified using a number of techniques (Kelly et al. 1997, Cole and Caraco 1998, Duchemin et al. 1999). For example, floating static chambers have been used to estimate the diffusive flux of CO2 and CH4 from the surface of reservoirs by calculating the linear rate of gas accumulation in the chambers over time (Figure 2). Diffusive flux of CO2 and CH4 from reservoir surfaces has also been estimated using the thin boundary layer method (Liss and Slater 1974). This calculated flux requires knowledge of the concentration gradient between the water and the air of either CO2 or CH4 and the gas exchange coefficient for the given gas at a given temperature. The concentration gradient is expressed as the difference between the measured partial pressure of dissolved gas in the water and the calculated partial pressure of the gas in the water if it were in equilibrium with the atmosphere. The gas exchange coefficient can simply be derived as a function of wind speed or of the rate of removal of gases from just above the water surface. Gases formed in decomposing organic matter at the bottom of reservoirs that ebullate directly to the reservoir surface in bubbles are measured using inverted funnel traps. Several studies have compared the chamber and thin boundary layer techniques (Kelly et al. 1997, Cole and Caraco 1998, Duchemin et al. 1999). For the wind speeds and conditions used to calculate fluxes from the surface of an experimentally created reservoir at the Experimental Lakes Area (ELA) in Ontario, Canada, agreement between the chamber and thin boundary layer techniques was very good (Kelly et al. 1997) and fluxes measured also agreed with a third micrometeorological flux-gradient technique (e.g., Chan et al. 1998). However, other studies found that the thin boundary layer method underestimated the flux measured by chambers (Duchemin et al. 1999) and sulfur hexaflouride (SF6) loss, especially at low wind speeds (less than 2 m/sec; Cole and Caraco 1998). In practice, the methods chosen to determine fluxes from reservoirs to the atmosphere are dictated by local conditions at the sampling site. For example, flooded tree snags or backwater bays are sheltered conditions, favoring the use of chambers because of their low and variable wind speeds. Conversely, chambers are often difficult to deploy in open stretches of windy, wavy water, and so the thin boundary layer method is appropriate. Therefore, flux measurements from reservoirs have relied on a combination of both techniques. The flux data summarized in this article (Table 1) were obtained primarily with floating chambers and/or the thin boundary layer method. Most of the reservoirs were of such a size that low wind speeds, such as in the Cole and Caraco (1998) study, were unusual, and so inclusion of data produced by both methods is reasonable. Fluxes calculated using the thin boundary layer techniques make our global greenhouse gas flux estimates somewhat conservative because bubble ebullition is often not measured in addition to diffusive fluxes (Table 1). Lack of bubble data affects estimates of CH4 flux the most because bubbles are usually composed mainly of CH4. Measurement of bubble ebullition is less likely to lead to an underestimation of overall CO2 flux because bubble fluxes of CO2 are relatively small compared with diffusive fluxes (Kelly et al. 1997, Duchemin 2000). In direct response to Rudd et al's (1993) hypothesis that greenhouse gas production in reservoirs is not zero and may depend on the amount of organic carbon flooded to create the reservoir, a unique whole-ecosystem experiment—the Experimental Lakes Area Reservoir Project (ELARP)—was initiated at the ELA. The primary goals of the ELARP were to experimentally create a reservoir to quantify in a controlled manner the net change in greenhouse gas fluxes to the atmosphere as a result of flooding and to understand the mechanisms causing these changes. The ELARP experimentally flooded a wetland, which hypothetically provided a worst-case scenario for long-term decomposition and greenhouse gas production because of the large stores of organic carbon held in peat deposits (Figure 3). Before flooding, the ELARP site consisted of both a pond surface and a peatland surface, each with its own natural characteristic greenhouse gas flux. A mixture of water and terrestrial surfaces is common in sites before reservoir construction. The pond surface emitted both CO2 and CH4 to the atmosphere, whereas the peatland surface took up CO2 and emitted a small amount of CH4. In general, lakes tend to emit both CO2 and CH4, whereas forests tend to take up both CH4 and CO2(Table 2). The whole-ecosystem flooding experiment resulted in conversion of the wetland from a small greenhouse gas sink to a relatively large source of greenhouse gases to the atmosphere. Before flooding, the wetland was on average a carbon sink of 6.6 g m2 yr of carbon. After flooding, it was a large carbon source (130 g m2 yr of carbon) in the form of CO2 (120 g m2 yr of carbon) and CH4 (9 g m2 yr of carbon; Kelly et al. 1997). Litterbags containing leaves from different trees, shrubs, and herbaceous plants, which were placed in the experimental reservoir before flooding, showed that most of this vegetation decomposed in the first 3 years of flooding (Tim R. Moore, McGill University, Montréal, Québec, personal communication). However, 7 years after flooding, gas fluxes from the flooded pond were the highest recorded over the entire course of the experiment (Carol A. Kelly and Vincent L. St. Louis, unpublished data), demonstrating that peat decomposition continued. The peat deposits before flooding contained 1 × 105 g/m2 of carbon, so if present rates of decomposition continue, there is enough organic carbon to support the current greenhouse gas flux for 2000 years (Kelly et al. 1997). The formation of floating peat islands resulting from gas buildup in the decomposing peat is another consequence of flooding peatlands (Figure 3). These floating peat islands have especially high rates of CH4 emissions, primarily because of low rates of CH4 oxidation (Scott et al. 1999). On a basis, CH4 has a global warming potential than CO2 over a time et al. In the 7 studies of reservoir sites have that of CO2 and CH4 from the flooded surfaces in (Table 1). we that in general reservoirs are of greenhouse gases to the atmosphere. A more difficult to is the of reservoir fluxes on a global are difficult for For example, in the ELARP greenhouse gas flux was measured both before and after flooding to an overall estimate of the net change of greenhouse gas flux by flooding, but in all other studies greenhouse gas flux was measured after flooding (Table 1). the loss of the carbon sink is not and the fluxes the effect on the atmosphere of reservoir Most flux estimates have been on reservoirs in temperate and that widely in size and 1). For these temperate reservoirs, we calculated average fluxes of m2 of CO2 and m2 of 1). data are available for reservoirs in tropical (Table 1). However, there to be an difference between temperate and tropical reservoirs. CH4 fluxes to be to CO2 fluxes in tropical reservoirs than they are in temperate the newly flooded tropical reservoirs and to have fluxes than reservoirs flooded on the longer and because there are not enough data for a good we first the reservoirs before an overall estimates of average flux for tropical reservoirs are m2 of CO2 and m2 of 1). The range of average fluxes from reservoirs around the m2 of CO2 and m2 of 1) was because fluxes of CO2 and CH4 depend on a number of including the amount of organic carbon flooded, of the reservoir, and temperature. The flux per unit area of greenhouse gases from reservoir surfaces should be to the amount of organic carbon that is flooded to create the The amounts of organic carbon per unit area are found in CH4 fluxes in temperate were highest in reservoirs that flooded at peatlands ELARP and 1). Fluxes of gases from reservoirs that flood such as the areas in the are to the on the and amount of carbon flooded is not After of the flooded organic carbon is not of the organic carbon of forests is in tree which are to decomposition after of is very because decomposition by which is in the terrestrial is are to estimate flux of greenhouse gases from reservoirs using estimates of the of flooded organic carbon in forests, the of tree should be from the (Rudd et al. The of reservoirs should also greenhouse gas fluxes because newly flooded carbon, such as that found in leaves and should decompose by decomposition of more organic carbon such as soil carbon and peat (Kelly et al. 1997). Fluxes from reservoirs are to over we examined fluxes from all the reservoirs in our data (Table 1), there was a in fluxes of CO2 from temperate reservoirs with (Figure However, emissions of CO2 at m2 of years after flooding of the A similar for CH4 was not because studies not bubble and bubbles a of fluxes of CH4 than of 2000). However, the use of data from a number of reservoirs to determine the is if reservoirs flood different types of such as the quantity of flooded in different types of the of reservoirs should be over time because other than the gas flux. to this using a of reservoirs in in from years to years (Table 1). These reservoirs were created for and flooded peatlands to we were to their fluxes to the ELARP which was 2 years (Table 1), to the All of the sites greenhouse gas fluxes than the newly flooded ELARP but was not the because of the reservoirs, which was the also the highest flux measured in that (Table 1). the reservoirs still greenhouse gas emissions than natural lakes in the area reservoirs in 1 compared with in m2 of CO2 and m2 of or terrestrial surfaces before flooding (Table demonstrating that fluxes of greenhouse gases from reservoirs not similar to lakes even after of of decomposition in tropical reservoirs were probably high because water are in tropical reservoirs than they are in temperate reservoirs rates of CO2 flux than temperate reservoirs and rates of CH4 mainly because of bubble ebullition (Table 1). formation is in with high CH4 production rates because the CH4 accumulation rate the rate of the which to and bubble to greenhouse gas fluxes because they are a direct of CH4 from to the atmosphere, with bubbles emit a high of gas as CH4, with the global warming potential (e.g., et al. 1997). also greenhouse gas fluxes from reservoir For example, of the derived carbon that is in reservoirs from decompose to CO2 and CH4, and the amount of vary from reservoir to primary in reservoirs may greenhouse gas emissions to the atmosphere CO2 the but most of this carbon is decomposed and returned to the atmosphere. In reservoirs with a relatively large amount of decomposed organic carbon by photosynthesis is as CH4. amounts of CH4 can be released at reservoir time (Rudd et al. of the global warming potential of CH4 as compared to this of CH4 to the atmosphere more than any effect of The and of reservoir can also rates of to pressure by water and 1994, Duchemin et al. in The global surface area of reservoirs to which average flux estimates should be is not as well known as be The most of reservoir surface area from 1998). However, the on is and is a for the global area by reservoir construction. not all large reservoirs are For example, and more than built for more than (ICOLD surface area is for of the reservoirs that are and not reservoir surface areas at These Canada, for which data that in of there are of surface area in reservoirs that have been developed primarily for hydroelectric The addition of data from the reservoir surface area to there are very few data on of the reservoir surface area was terrestrial and was natural or This is a in the change in flux flux from water surface loss of terrestrial as to simply at current Reservoirs with less than high are not usually in the of 1998). However, the surface area of reservoirs small is For example, the of of which all in the would result in loss of or significant a surface area than the of of large for the 1998). This size not be to a global area but the data on global reservoir area (ICOLD are an both the of large in and the overall of data on small reservoirs, we estimate that the global surface area of all reservoirs is or the documented area large This area is equivalent to the estimated global surface area of natural lakes of the surface area of all reservoirs in (1998) is in found between and and to have tropical estimate that the surface area of reservoirs is in the and in temperate the average fluxes (Table 1) with the estimated surface area of reservoirs in temperate and tropical global fluxes of × of CO2 and × of 3). estimate that a of and of global reservoir fluxes of CO2 and CH4, from tropical reservoirs even though these reservoirs for of the global surface On a global basis, the CO2 flux from reservoirs was equivalent to of other anthropogenic emissions of but the CH4 flux was to of other anthropogenic CH4 emissions. These large estimated CH4 fluxes from reservoirs estimated fluxes from or (Table 3). is difficult to to these fluxes because of the in which were calculated for each reservoir in 1). However, we can a range of between and for other anthropogenic CO2 emissions, and between and for other anthropogenic CH4 emissions, using the average range of fluxes and from temperate and tropical reservoirs (Table 1). rather than a for the estimate of the current greenhouse gas flux from reservoirs globally, we to make the that the range of fluxes is such that and is to understand the of reservoir to greenhouse gas compared to other of greenhouse gases. This is especially with to CH4. CO2 and CH4 fluxes are and to a flux of carbon to the atmosphere, the fluxes from reservoir surfaces are to of carbon, or of other documented anthropogenic fluxes of carbon as CO2 and CH4. On a basis, the global warming potential of CH4 to CO2 is over a period et al. the average of reservoirs. this global warming potential of CH4 greenhouse gas emissions from reservoirs were equivalent to an average of 7% of the global warming potential of other emissions for the period (Table 3). A global warming potential of a time is used CH4 emissions to CO2 emissions from other anthropogenic of greenhouse gases. However, because reservoirs may not emit CO2 and CH4 at a rate over time they may produce fluxes at first and the of a global warming potential may lead to an underestimation of the to global and may be especially sites (e.g., 1997). A number of the of the current estimated global flux of greenhouse gases from the surface of reservoirs to global These the of greenhouse gas fluxes from landscapes before flooding and the of organic carbon in the of reservoirs. The fluxes in 1 and 3 are from reservoir surfaces after flooding has a that is to the of these surfaces as they However, it is also important to ask these current fluxes are different from those that before flooding, because the effect of reservoir on the atmosphere is the net difference between fluxes of greenhouse gases before flooding and after For example, if the entire reservoir area was a net sink for both CO2 and CH4, as forests are (Table then the loss of these should be to the current reservoir However, of the is often water and was likely CO2 derived from decomposition of organic carbon. natural water surfaces take up if net primary production is than organic carbon but this is less common in lakes than net of CO2 (Table Cole et al. All surfaces emit 2). emissions before flooding should be from the current change to reservoir current flux from reservoir surface CO2 and CH4 from terrestrial CO2 and CH4 emissions from in of all the in the above is However, the ELARP included measurements of greenhouse gas fluxes in both the terrestrial and of the that was for fluxes of other reservoirs be using flux data on and (Table if knowledge of the areas of terrestrial and surfaces were measurements of areas are often not these can we make general estimates of greenhouse gas fluxes from reservoir areas that are in current reservoir is to estimate the loss of the carbon sink as if all the area flooded been terrestrial number that be the current fluxes to the net change in the above and then to estimate the flux in current reservoirs that be to decomposition of organic carbon derived from the after flooding number that be in the above data for fluxes of greenhouse gases from and tropical forests (Table and calculations similar to those in 3 for the of reservoirs, were used to make the first all the currently flooded area been we would to × of CO2 to the current rates to for the loss of this carbon if all the currently flooded area been natural an average flux of m2 of CO2(Table then we would 3 × of CO2 from the current fluxes to the net change in the above the are an addition of × of CO2 and a of 3 × of CO2 to the global estimate of × of CO2 from reservoir surfaces in for CH4 are an addition of 2 × of CH4 and a of × of CH4 to the global estimate of 7 × of CH4 from reservoir of each of these is and so the in Therefore, an important in the net change in greenhouse gas flux to reservoir is to the average of to water in reservoir areas before However, because the is currently not available on a global basis, we have not the estimates that in reservoir water to there is of organic carbon at an estimated rate of g m2 yr of carbon, even though the percentage of organic carbon in reservoir is less than in most and and 1998). this of carbon a carbon sink with to the be if the organic carbon that are decompose to a than they would have if they to the However, there have been no studies to determine if this is and surface gas fluxes no on Reservoir construction simply that the for organic carbon is in reservoir rather than Reservoir is not greenhouse gas with to production of electricity by hydroelectric as has been (Hoffert et al. 1998, Victor 1998). Although there is in our they that reservoir fluxes are of a similar to other fluxes included in to understand anthropogenic in the global carbon (e.g., CH4 from 3). have been to the global carbon because more CO2 is emitted from anthropogenic than can be for by either the in CO2 in the atmosphere or by resulting in a for CO2. estimate this sink of CO2 by from of carbon et al. to of carbon of carbon as by the anthropogenic CO2 emissions to the atmosphere. that fluxes of greenhouse gases to the atmosphere from all reservoirs in the because they are as a relatively to energy produced by and because of for water and as well as the and of a global The global terrestrial area flooded by reservoirs in the because of potential hydroelectric sites have been developed is that greenhouse gas fluxes from hydroelectric reservoirs of the global warming potential of all other current anthropogenic emissions and the surface area of hydroelectric reservoirs of all reservoir types (ICOLD 1998). more determine the of reservoir fluxes as compared to other greenhouse gas more CO2 and CH4 flux measurements are from reservoirs in all global with an on tropical reservoirs. an is on areas of and reservoirs, and of terrestrial and areas flooded in each reservoir, so that the net change in greenhouse gas flux to flooding can be is for greenhouse gas studies on reservoirs to continue, given the of our estimate for the flux resulting from reservoir and the of in the sink for CO2 in the then we be to determine the impact of reservoir on the global carbon and the flux data in the form of in are for the of R. and This was by the and of Fluxes of CO2 and CH4 from the surface of temperate and tropical reservoirs of different and fluxes of CO2 and CH4 from the surfaces of different fluxes of CH4, and from the surface of reservoirs and snags the of a from these trees, and soil carbon such as that found in and after of decompose little over time at the bottom of reservoirs. of L. St. A static floating chamber in the of the Experimental Lakes Area Reservoir Project Fluxes of CO2 and CH4 from the surface of the reservoir are calculated by the rate of buildup of these gases over time the of L. St. The Experimental Lakes Area Reservoir Project reservoir years after experimentally flooding a wetland by the water The wetland before flooding consisted of a pond by a that most of the peat is of L. St. between the of temperate reservoirs and the average flux of CO2 of of are from 1