ABSTRACT

Greenhouse gas emissions from hydroelectric dams are often portrayed as nonexistent by the hydropower industry, and have been largely ignored in global calculations of emissions from land-use change. Brazil’s Tucuruí Dam provides an example with important lessons for policy debates on Amazonian development and on how to assess the global warming impact of different energy options. Tucuruí is better from the point of view of power density, and hence greenhouse gas emissions per unit of electricity, than both the average for existing dams in Amazonia and the planned dams that, if all built, would flood 3% of Brazil’s Amazon forest. Tucuruí’s emission of greenhouse gases in 1990 is equivalent to 7.0-10.1 × 10⁶ tons of CO₂-equivalent carbon, an amount substantially greater than the fossil fuel emission of Brazil’s biggest city, São Paulo. Emissions need to be properly weighed in decisions on dam construction. Although many proposed dams in Amazonia are expected to have positive balances as compared to fossil fuels, substantial emissions indicated by the present study reduce the benefits often attributed to the planned dams.

1. Introduction

Hydropower is often promoted by government authorities as a "clean" source of energy, in contrast to fossil fuels (e.g., de Souza, 1996). While fossil fuel contributions to global warming are well known, hydroelectric dams are not free of impact. Hydroelectric dams in tropical forest areas emit greenhouse gases such as carbon dioxide (CO₂) and methane (CH₄). The ratio of impact to benefit varies tremendously among dams depending on their power output. Tucuruí, the dam examined in this paper, has a more favorable balance than either the average existing dam or the average planned dam in Brazilian Amazonia.

Tucuruí serves as a testing site for the country’s plans for hydroelectric development in Amazonia. The possibility of claiming carbon credit for planned hydroelectric dams often arises in discussions in Brazil on the Kyoto Protocol’s Clean Development Mechanism. Greenhouse gas emissions clearly cannot be ignored in such discussions. Even more important are the social and environmental impacts caused by many dams, which have caused these projects to be questioned as a forms of “clean development” (Fearnside, 1999a). The Tucuruí Dam provides an excellent example of these problems (Fearnside, 1999b, 2001).

This paper estimates greenhouse gas emissions for Tucuruí for 1990, the base year for national inventories of greenhouse gas emissions under the United Nations Framework Convention on Climate Change (UN-FCCC). The estimate includes emissions from various sources of emission that have been ignored in previous estimates for Amazonian reservoirs, such as methane release from water passing through the spillway and turbines.

2. The Tucuruí Reservoir

Brazil’s Tucuruí Dam was completed in 1984 on the Tocantins River, a north-flowing tributary to the Amazon River located in Pará (Figure 1). The reservoir’s area is officially 2430 km² at the normal operating level of 72 m above mean sea level (msl). LANDSAT satellite measurements estimated the area at 2247 km² in 1989 (INPE, see Fearnside, 1995, p. 13) and 2800 km² in June and July 1996 (de Lima et al., 2000). The current dam (Tucuruí-I) has 3960 MW of installed capacity. A second phase (Tucuruí-II) is expected to double the installed capacity to 8085 MW by 2002 (Brazil, Programa Avança Brasil, 1999).

[Figure 1 here]

The Tucuruí Dam is one of 10 focal studies of the World Commission on Dams, particularly for understanding greenhouse gas emissions from hydroelectric development (WCD, 1999). Tucuruí has the largest reservoir in Brazilian Amazonia after Balbina (which is often dismissed by electrical authorities as atypical because its low power density). Balbina has especially high emissions due to its large reservoir relative to the generating capacity that could be installed at a location with flat topography and low streamflow (Fearnside, 1995, 1996a; Rosa et al., 1996a). Tucuruí-I (the present configuration of Tucuruí) has 1.63 watts (W) of installed capacity per m² of reservoir surface, whereas Brazil’s national power authority (ELETROBRÁS) has calculated the average power density for the entire hydroelectric potential of the Amazon Region to be only 1 W/m² (Rosa et al., 1996b, p. 6). This refers to the full list of dams planned in Brazil's 2010 Plan, regardless of the intended date of construction (Brazil, ELETROBRÁS, 1987; see Fearnside, 1995). The equivalent figure for the 5537 km² of water surface in the four existing large dams (whose total installed capacity is 4490 MW) is 0.81 W/m², or only half the power density of Tucuruí-I.

3. Reservoir Emissions

3.1. Emissions from Above-Water Decay

When tropical forests are flooded by reservoirs, the trees are left standing in the water with the exception of small areas cleared near the dams. A substantial part of the biomass is left projecting above the water surface and decays aerobically. This source of emission has been ignored in most discussions of the global warming impact of hydroelectric development. Parameters for above-water emissions calculations for the Tucuruí reservoir are given in Table I.

[Table I here]

3.2. Emissions from the Reservoir Surface

Several recent studies in reservoirs indicate that methane emissions show a large peak in the first years after filling, followed by a decline. At age one year, the Petit-Saut reservoir in French Guiana released 1300 mg CH₄/m²/day from the water surface (530 from bubbling and 770 from diffusion) (Galy-Lacaux et al., 1997). The Curuá-Una reservoir, at age 21 years, released 66 mg CH₄/m²/day (16 from diffusion and 50 from bubbling) (Duchemin et al., 2000). Gatun Lake in Panama at age 84 years released 412 mg CH₄/m²/day (12 from diffusion and 400 from bubbling) (Keller and Stallard, 1994). Measurements of methane emissions specific to Tucuruí indicate wide spatial and temporal variations (Table II).

[Table II here]

Considerable controversy has surrounded the methodology for estimating emissions from the reservoir surface. An inverted funnel sampling device used by the University of São Paulo at São Carlos measures bubbling emissions only (e.g., Rosa et al., 1996b, c, 1997a). Diffusion chambers used by INPE (de Lima and Novo, 1999; de Lima et al., 2000) and by studies by the University of Quebec at Montreal at other Amazonian reservoirs (Duchemin et al., 2000) captured both bubbles and diffusion emissions. However, these chambers are less efficient than funnels for estimating the bubbling component because chambers make collections using a series of brief (typically 15-minute) emplacements that can easily miss the sporadic bursts of bubbling activity (Keller and Stallard, 1994). Recent work at Tucuruí (age 14-15 years) by the University of São Paulo at São Carlos with both funnels and diffusion chambers indicates that bubbling only contributes 6-16% of total surface emissions of 14.6-205.3 mg CH₄/m²/day (Matvienko et al., 2000). At the Curuá-Una reservoir (age 21 years), Duchemin et al. (nd) measured both bubbling and diffusion and found bubbling accounts for 32-81% of total surface emissions of 37.5-80.2 mg CH₄/m²/day. At Petit-Saut (age 1 year), Galy-Lacaux et al. (1997) found a bubbling contribution of 59% to a total surface emission of 1300 mg CH₄/m²/day. In coves at Gatun Lake (age 84 years), bubbling contributed 97% of the 400 mg CH₄/m²/day average surface emission (Keller and Stallard, 1994). Given the lack of any consistent proportionality between bubbling and diffusion, only estimates that include both bubbling and diffusion were used in the present study (Table II).

The area covered by macrophytes (mostly floating weeds such as water hyacinth, Eichhornia crassipes) is an essential determinant of methane flux. At the high-water period (14 August 1988), Novo and Tundisi (1994, p. 149) found that 21% of the Tucuruí reservoir was covered with macrophytes based on LANDSAT imagery. Based on the data and assumptions of Novo and Tundisi (1994), for the high water-level period, and on the assumptions of these authors for the remainder of the year, the average area occupied by macrophytes over the annual cycle was 286.4 km², or 13.1% of the 2188 km² average area of the reservoir that can be computed for the same year. At Tucuruí, macrophytes exploded in the first year after filling and then died back to lower levels as the nutrient content of the water declined. In 1986 (two years after closing) the area of macrophytes was estimated at 620 km² (Brazil, ELETRONORTE, 1988a, p. 94), or about 26% of the reservoir area when full. A study by de Lima et al. (2000) shows June-August (high water level) macrophyte cover decreasing from 39% in 1986 to 11% in 1994, which appears to be a stable level. The cover would have been 21% in 1988, corresponding to the scenario in Table III. Initial explosive growth followed by decline has also been the pattern for macrophytes at other tropical reservoirs such as Brokopondo in Surinam (Leentvaar, 1966), Guri in Venezuela (Vilarrubia and Cova, 1993) and Balbina (Fearnside, 1989; Walker et al., 1999), Curuá-Una (Junk et al., 1981) and Samuel (Bohdan Matvienko, public statement, 24 February 2000) in Brazil.

[Table III here]

In seven studies in várzea (floodplain) lakes, areas with macrophytes had 3.25 times more CH₄ emissions than open water (see Fearnside, 1995, p. 15). At Tucuruí in September 1992, an area with macrophytes had 1056 times more CH₄ emission by bubbling than open water in the river channel, 0.8 times as much as open water with standing trees, and 5.8 times as much as open water in a cove without standing trees (Rosa et al., 1996c, p. 150). The greater areas of macrophytes in a reservoir’s early years contribute to a greater pulse of methane emissions during these years.

A large area of the reservoir bottom is seasonally exposed. Considering the 58 m above msl minimum operating level for Tucuruí-I (Brazil, ELETRONORTE, 1989, p. 64), this area occupies 858 km² (Fearnside, 1995, p. 13), while considering the drawdown to 68 m above msl in August 1988 (before all turbines were operational), the area occupied 397 km² (Novo and Tundisi, 1994). When flooded, the drawdown area offers ideal conditions for generation of methane, as well as for methylation of mercury in the soil. In the Samuel reservoir, for example, areas like this released 15.3 g C/m²/year as CH₄ through bubbling when seasonally flooded, as compared to 7.2 g C/m²/year among standing dead trees in permanently flooded areas and only 0.00027 g C/m²/year in the main channel (Rosa et al., 1996c, p. 150).

Based on information on habitat areas and emission rates (Tables II and III), one can calculate approximate emissions through bubbling and diffusion of CH₄ from Tucuruí (Table IV). This assumes that the area covered by macrophytes throughout the annual cycle follows the assumptions of Novo and Tundisi (1994, p. 150), which are that the maximum macrophyte area (505.4 km²) applies to four months (assumed to be May-August), while for two months (assumed to be March and April) the area is 50% of the maximum and is replaced by open water (25%) and exposed drawdown (25%), and for six months (assumed to be September-February) the macrophyte area is 30% of the maximum and is replaced by open water (15%) and exposed drawdown (15%). The year is divided into two seasons on the basis of streamflow: a low-flow period (January-May) and a high-flow period (June-December). The year could also be divided on the basis of water level (low = September-February, high = March-August) or rainfall (dry = July-November, wet = December-June).

[Table IV here]

3.3. Emissions from the Turbines

In 1991 Tucuruí produced 18.03 TWh of electricity (Brazil, ELETRONORTE, 1992, p. 3), or 2058 MW. The dam's annual production was expected to be 2476 MW by 1991 (Brazil, ELETRONORTE, 1989, p. 58), or 20.3% more than was actually produced; had production been higher, CH₄ emissions from water passing through the turbines would also have been proportionately higher. Each turbine has a nominal capacity of 350 MW and a power factor of 95% (i.e., 332.5 MW of effective production), and uses 575 m³/s of water (Brazil, ELETRONORTE, 1989, p. 17). Each turbine therefore uses 18.1 × 10⁹ m³/year of water, and 18.3 MW is generated per 10⁹ m³ of water. A total of 112.2 × 10⁹ m³ of water passed through the turbines in 1991. The methane concentration at 30-m depth was 6 mg CH₄/liter in March 1989 (unpublished data of J.G. Tundisi cited by Rosa et al., 1997a, p. 43). Work at the Petit-Saut Dam by Galy-Lacaux et al. (1999, p. 508) shows that CH₄ concentrations fluctuate on a seasonal basis in a pattern that corresponds to the balance between water inflow and outflow in the reservoir. The amplitude of the oscillation is such that the maximum concentration is at least 50% higher than the minimum in each annual cycle. The one available profile for CH₄ concentrations in the water at Tucuruí is from March 1989, which is during the high water-flow period when the time series at Petit-Saut (Galy-Lacaux et al., 1999) indicates that CH₄ concentrations are at a minimum. If the relative magnitude of the seasonal oscillation in CH₄ concentration at Petit-Saut applies to Tucuruí, the concentration at 30-m depth should vary (at least) between 6 and 9 mg CH₄/liter, with a mean value of 7.5 mg CH₄/liter (Figure 2).

[Figure 2 here]

This can be regarded as a conservative estimate of the concentration in the water passing through the turbines, since CH₄ concentration increases with depth and the intake is at a depth of 35.4 m when the reservoir is at the operating level of 72 m above msl (Brazil, ELETRONORTE, 1989, p. 157). Based on an annual average methane concentration of 7.5 mg CH₄/liter at the level of the turbines, one can calculate that the amount of CH₄ exported from the reservoir through the turbines in 1991 would have been 0.842 × 10⁶ t. However, the seasonal oscillation acts to reduce the amount of CH₄ exported below this value because power generation is greatest during the high-flow period, when the concentration of CH₄ in the water is least. An adjustment for this effect is calculated in Table V, reducing the 1991 export of CH₄ by 6.7% to 0.785 × 10⁶ t.

[Table V here]

The fate of the CH₄ in water passing through the turbines can be estimated based on data from the Petit-Saut Dam (Galy-Lacaux et al., 1997). Totaling three measurements at Petit-Saut, an average of 87.1% of the methane was degassed immediately when the water emerged from the turbines; of the remaining methane, 18.4% was degassed from the river water and 81.6% was oxidized to CO₂ in the first 40 km downstream of the dam. Based on these data, the 1991 release from water passing through the turbines at Tucuruí totaled 0.702 × 10⁶ t CH₄ (0.684 × 10⁶ t at the turbines and 0.019 × 10⁶ t in the river).

A significant difference between Tucuruí and Petit-Saut is an aerating device built about 100 m downstream of the Petit-Saut Dam. When dam operation began in June 1994, almost all of the fish died in the river below the dam, motivating suspension of generation while the device, a 4-m-high dual-nappe weir, was built to create an artificial waterfall and provide more oxygenated water to the river below. The weir was completed in February 1995 (Gosse, 1999). An unintended byproduct of this is the release of additional methane, some of which would otherwise have been oxidized to CO₂ by bacterial activity in the river or in the ocean (40 km downstream). In the case of Tucuruí, however, one can assume that most of the CH₄ in the water is also released when water passes through the turbines because of the sudden drop in pressure. For example, at Balbina, water samples from the reservoir bottom (29 m maximum depth) foam with CH₄ and CO₂ bubbles when brought to the surface (Bohdan Matvienko, public statement, 24 February 2000).

The Petit-Saut data do not allow separation of the amount degassed immediately as the water emerges from the turbines from that degassed at the artificial waterfall. Galy-Lacaux et al. (1997, p. 479) calculate the combined release at these two points from the concentrations of CH₄ in the water column just above the dam and in the water below the artificial waterfall. Of the three such paired measurements reported by Galy-Lacaux et al. (1997, p. 497), the average CH₄ concentration drops from 8.11 mg/liter to 0.77 mg/liter, or 90.5%. The average amount degassed at the turbines and waterfall totaled 98.2 t CH₄/day (89.9% of the CH₄ exported through the turbines, or 97.7% of the 100.5 t CH₄/day total emissions from turbined water). All release in the river occurs in the first 20-30 km below the weir; the average of the three measurements of this release was 2.3 t CH₄/day, which represents 2.1% of the CH₄ exported through the turbines, or 2.3% of the total emissions from turbined water. If the artificial waterfall did not exist, the amount degassed in the river would probably be higher than the 11.0 t/day measured in the Petit-Saut case (21.0% of the CH₄ entering the river below the weir) because of the higher concentration of CH₄ that would enter the river at this point. This makes it possible to calculate lower and upper bounds for the emission from water passing through the turbines at Tucuruí. Considering the percentages released as 21.0%-89.9% based on the Petit-Saut results, the release from the turbines at Tucuruí in 1990 was 0.165-0.702 × 10⁶ t CH₄. The total methane released from water passing through the turbines at Tucuruí was 2-8 times the total release from bubbling and diffusion in the reservoir itself.

3.4. Emissions from the Spillway

An additional major source of CH₄ emissions at Tucuruí is water released through the spillway. This water is not taken from the surface, but rather comes from a level 52 m above msl (Brazil, ELETRONORTE, 1989, p. 146). Water released from the spillway comes from a depth of 20 m when the reservoir is at the normal operating level of 72 m above msl. Water shoots out from under a series of 23 steel doors when they are raised; normally this is a thin sheet from the bottom of the spillway bay, although these gigantic doors (each 21-m high and weighing 220 t) can be pivoted upward to allow major floods to pass. With the exception of such flood events, the offtake is therefore the 52 m above msl elevation of the spillway crest. In each spillway, the water descends a chute to 30 m above msl, where it is thrown into the air by a ski jump-like device and plunges an additional 28 m to a concrete-lined dissipation basin below. The great white plume of spray formed when all 23 of the 20-m wide spillways are open is undoubtedly the most spectacular and frequently photographed sight at the dam. The completeness and instantaneous nature of the aeration make it a safe assumption that all CH₄ dissolved in the water is released immediately to the atmosphere.

Emissions from the spillway would be very large if the reservoir were always kept at its full water level. Considering the long-term average streamflow of 11,107 m³/s (Brazil, ELETRONORTE, 1989, p. 17), or 350.5 × 10⁹ m³/year, and the seasonally adjusted CH₄ concentration of 3.75 mg CH₄/liter at 20-m depth, the spillway would have emitted 0.893 × 10⁶ t CH₄ in 1990, equivalent to 5.1 × 10⁶ t of CO₂-equivalent C. However, several factors reduce the annual emission from water released through the spillway. One is the seasonal cycle in CH₄ concentration, the greatest release from the spillway being during the high-flow period when the CH₄ concentration is lowest. Another factor is the effect of drawdowns: as the water level falls, the depth of the spillway intake relative to the water surface decreases, presumably with a corresponding reduction in CH₄ concentration at the spillway level. These two effects reduce the estimated export of CH₄ through the spillway to 0.535 × 10⁶ t (Table VI), a decrease of 40.1%. An effect not corrected for is the thickness of the sheet of water allowed to pass through the spillway: although normally only a narrow slit is opened, during larger floods the floodgates can be raised higher, allowing water nearer the surface (with lower CH₄ content) to escape.

[Table VI here]

3.5. Loss of Sources and Sinks in Living Forest

When standing tropical forest is flooded and killed, the forest’s natural sources and sinks of greenhouse gases are lost. These include the loss of an annual uptake of carbon by the standing forest. Studies using eddy-correlation techniques indicate that intact Amazonian forests have a net uptake carbon at present (e.g., Grace et al., 1995; Mahli et al., 1998). Although this effect cannot be permanent (over the long-term the forest C stocks cannot continue to grow), the uptake effect nevertheless constitutes an addition to the impact of killing large areas of forest by flooding. Other losses include loss of a small methane sink in tropical forest soil and loss of a very small methane source from forest termites. On the other hand, a source of nitrous oxide (N₂O) emissions is eliminated by the flooding. Amazonian forest soils at Paragominas (where the seasonal distribution of precipitation is similar to that at Tucuruí) emit an estimated 8.68 kg of N₂O/ha/year (Verchot et al., 1999, p. 37), equivalent to 0.73 t CO₂-equivalent C/ha/year considering the 100-year global warming potential of 310 adopted by the Kyoto Protocol for N₂O. The 1926 km² of forest flooded at Tucuruí (Fearnside, 1995, p. 11) therefore emitted 0.117 × 10⁶ t of CO₂-equivalent C annually as N₂O prior to flooding. The area flooded by Tucuruí, as with most hydroelectric dams, was not a wetland prior to flooding, but rather was an area of rapids on the river that had topography sloping steeply enough to maintain well-drained soils. The pre-reservoir emission was therefore not the much larger source of either CH₄ or N₂O that has sometimes been implied. The net effect of losses of sources and sinks in living forest is calculated in Table VII.

[Table VII here]

3.6. Global Warming Impact of 1990 Emissions

In summary, the major sources of methane emissions at Tucuruí in 1990 were as follows, in 10⁶ t CH₄: 0.0937 from bubbling and diffusion, 0.1649-0.7025 from the turbines, and 0.5353 from the spillway (Table VIII). Small additional contributions were made by termites in above-water decay and from loss of the sink in forest soil, and a small reduction in the flux resulted from loss of forest termites. The CH₄ emission totaled 0.79-1.33 × 10⁶ t of gas; considering a global warming potential of 21 (Schimel et al., 1996, p. 121), this is equivalent to 4.5-7.6 × 10⁶ t of CO₂-equivalent C. Emissions of CO₂ in 1990 were estimated at 9.68 × 10⁶ t of CO₂ gas, or 2.64 × 10⁶ t of C. Adjustment for loss of the N₂O source in forest soil decreases the emission by 1-2%. The contribution of methane represented 64-75% of the total greenhouse gas impact of 7.0-10.1 × 10⁶ t of CO₂-equivalent C in 1990 (Table VIII). As explained in Table VIII (note a), 1990 emissions are calculated from parameters from the various years for which information is available.

[Table VIII here]

4. Discussion

4.1. Uncertainty

The reliability of the present estimate is most sensitive to the value of two parameters: the concentrations of CH₄ in the water passing through the spillway and through the turbines. Here a set of values is used that was measured at Tucuruí in March 1989 by José G. Tundisi (cited by Rosa et al., 1997a, p. 43). These values are adjusted for seasonal oscillations based on the series of measurements at Petit-Saut (Galy-Lacaux et al., 1999). The existence of seasonal oscillations indicates the importance of a series of measurements to capture this source of variation. Most research effort aimed at quantifying greenhouse gas emissions from hydroelectric dams, including Tucuruí, has been devoted to measurements of fluxes at the reservoir surface. However, the calculations in the present paper show clearly that the greatest gains in reducing uncertainty in the overall estimate would come from better information on the CH₄ concentrations in water entering the turbines and the spillway, and the fate of CH₄ in the river below the dam.

Methane fluxes from the reservoir’s surface, particularly through bubbling, are also subject to cycles. On a seasonal basis, emissions per unit area are higher at any given location in the reservoir when the water level is low. The frequent drawdowns in reservoir management can be expected to result in greater CH₄ release through bubbling. The large releases that occur when water levels fall are likely not to be detected by the brief measurement “campaigns” that are the basis of currently available data.

Bubbling emissions are greater in shallower water because it has less vertical distance over which CH₄ bubbles released from the sediments can be oxidized before reaching the surface. Also, hydrostatic pressure on the sediments is less, leading to greater release of bubbles from this supersaturated environment. In addition, methanogenesis rates are sensitive to temperature, and the cooler sediments at greater depths would produce less CH₄ than sediments in shallow areas. At Gatun Lake, for example, over a depth gradient from 0.5 to 10 m, bubbling rate decreased by a factor of 10, of which a factor of 2.3-3.9 could be explained by temperature and pressure differences (Keller and Stallard, 1994, p. 8315). The substantial additional effect of depth may be due to greater inputs of terrestrial C in the shallow near-shore areas (Keller and Stallard, 1994). In addition to the effect of depth on emission variations over time, this factor also shows the great spatial variation that exists over the reservoir surface and the care needed to obtain representative samples and interpret these through appropriate weighting by the area of each habitat and depth category. The three-zone division used in the current paper is the maximum level of detail that current data permit, but as more measurements become available, a finer breakdown of depth and habitat classes could improve the reliability of the estimates.

On a diurnal basis, emissions are higher during the day than at night due to greater wind and wave action (Duchemin et al., 2000; Keller and Stallard, 1994). Greater bubbling fluxes in the afternoon at Tucuruí may also be due to a diurnal cycle in atmospheric pressure that is equivalent to an 18-cm water level fluctuation in terms of hydrostatic pressure exerted at the bottom (de Lima and Novo, 1999). Since many reported measurements do not specify that a 24-hour monitoring cycle was included, this is a source of additional uncertainty.

The Galy-Lacaux et al. (1999) study at Petit-Saut indicates that CH₄ concentrations decline over time, falling from 14 to 10 mg/liter in the first four years of impoundment (the measurement period at Petit-Saut), and expected to decline to 0.3 mg/liter at age 20 years based on present CH₄ levels at a comparable dam in the Ivory Coast. The projected concentration at Petit-Saut six years after impoundment (the reservoir age for the current estimate for Tucuruí) is 4 mg/liter. However, Galy-Lacaux et al. (1999) used an average CH₄ concentration over the full vertical profile of the water column at a sampling station near the dam as the estimate of the concentration in the water passing through the turbines. Petit-Saut differs from Tucuruí in some significant ways affecting the choice of a CH₄ value. The reservoir at Tucuruí is approximately twice as deep at the dam as Petit-Saut, with the midpoint of the intakes for the turbines located at a depth of 35.4 m (below the 34-m total depth of the reservoir at Petit-Saut). In addition, Petit-Saut has a special structure built to minimize the discharge of anoxic water (which is also the most methane rich). This is an underwater dike built parallel to the dam 60 m upstream as a measure to immobilize the lower half of the water column and draw only relatively well-oxygenated surface water into the turbine intakes (Sissakian and Desmoulins, 1991). Tucuruí has no such structure, making a CH₄ concentration value measured as close as possible to the level of the turbine intakes a more appropriate choice than the mean for the water column.

The present estimate of emissions from Tucuruí is conservative for several reasons. The estimate ignores "unusual" events, such as storms, that result in much higher than average emissions from the reservoir surface. These events have been found to represent a substantial portion of the annual emissions in reservoirs in northern Canada (Duchemin et al., 1995). Storms can cause large inputs of organic matter from the watershed, such as leaves, twigs and other debris; they can also create seiches that bring the oxycline to the surface, allowing release from methane-saturated deep water (Donald D. Adams, public statement, 24 February 2000).

The use of data from different years to produce an approximate estimate for 1990 adds to the uncertainty. Some of the effects result in overestimation of the 1990 emission, such as using macrophyte areas from 1988 and CH₄ content of the water from 1989, turbine and spillway depths from 1988, and turbine water flow from 1991. Other factors underestimate 1990 emission, such as bubbling and diffusion per unit of area from 1996-1997 and spillway flow from 1991.

The present estimate does not include emissions from deforestation by the displaced population. The substantial emissions from dam construction that would be needed for a full-chain energy analysis (“life-cycle analysis”) are also not included. Future impacts would also include emissions from upstream dams planned to regulate the flow of the Tocantins River.

4.2. Comparison with Previous Estimates

Greenhouse gas emissions from the Tucuruí reservoir for a single year (1990) have been estimated (Fearnside, 1995), and the analysis was subsequently extended from a single year to compute the amount and timing of emissions over a 100-year time period, which could then be compared to the emissions that would be produced by generating the same amount of energy from fossil fuels (Fearnside, 1997a). Factors considered included the initial stock and distribution of carbon, decay rates and pathways (leading to CO₂ and CH₄), and losses of power in transmission lines. Factors not considered included forest degradation on islands and reservoir shores, nitrous oxide sources in drawdown zones and transmission lines, additional methane emission pathways for release from standing trees, water passing through the turbines, etc. Construction-phase emissions were also not included, nor were emissions from deforestation by people displaced by (and attracted to) the project.

Earlier calculations assumed that only 10% of the water surface was covered by macrophytes (Fearnside, 1997a). The average area used in the current calculation was 13.1% (Table III). However, the emission from macrophyte beds is much lower under the current calculation (72 mg CH₄/m²/day at high water and 68 mg CH₄/m²/day at low water) than the 174.7 mg CH₄/m²/day used in previous calculations (Fearnside, 1995, 1997a). This probably indicates that the present calculation is conservative, since the previous ones, although based on data from várzea lakes rather than from Tucuruí, were based on many more observations (e.g., Bartlett et al., 1990; Devol et al., 1990; Wassmann and Thein, 1989).

Most of the global warming impact in the earlier calculations (Fearnside, 1995) came from CO₂ released by above-water decomposition of wood: in 1990 CO₂ contributed 83% and CH₄ 17%, considering the global warming potential of 21 for CH₄ for the impact of a ton of this gas relative to a ton of CO₂ currently used by the Intergovernmental Panel on Climate Change (IPCC) (Schimel et al., 1996, p. 121). In the above analysis, methane emissions were assumed to be relatively constant over the time horizon, rather than having an initial peak followed by a decline to a lower plateau.

Inclusion of CH₄ emissions from water released through the turbines and the spillway, which were not included in previous estimates, substantially increases the reliability of the present calculation. These sources increase total emission of CH₄ as compared to earlier emissions estimates (Fearnside, 1995, 1997a), which included CH₄ from decomposition of submerged forest, for which assumptions had been used that now appear to be conservative. The CH₄ production estimates based on assumptions about decomposition rates and pathways have been replaced with more reliable estimates based on measurements of CH₄ concentration in the water and release through the turbines and spillway. This significantly changes previous estimates for 1990 (Fearnside, 1995), in which CO₂ contributed 83% and CH₄ 17%. The revised estimate indicates lower methane emissions from the reservoir surface (mainly due to lower values for emission from macrophytes per m²).

The results of the present study are compared with those of previous studies in Table IX. Studies vary widely, not only in their final results but also in the completeness of their coverage of emissions sources. Estimates producing very low final results ignore CH₄ emissions from water passing through the turbines and spillway and CO₂ emissions from decay of above-water biomass. Mainly due to the inclusion of emissions from the turbines and spillway, calculations in the present study more than doubles this author’s previous estimate for emissions in 1990 (Fearnside, 1995) from 3.1 × 10⁶ t CO₂-equivalent C (considering the current global warming potential of 21 for CH₄) to 8.6 × 10⁶ t CO₂-equivalent C, this being the midpoint of the 7.0-10.1 × 10⁶ t CO₂-equivalent C range that results from uncertainty concerning the percentage of CH₄ released in turbined water. Despite high and poorly quantified uncertainty, the finding of substantial emission is believed to be quite robust. The results of this study indicate emissions one to two orders of magnitude greater than the reservoir-surface emission studies that currently form the basis of Brazilian policy on global warming and hydroelectric dams (Table IX).

[Table IX here]

4.3. Time Path of Emissions

A key question for the future will be whether the concentration of CH₄ in the water will decline to a plateau at a very low level, such as the 0.32 mg/liter found by Galy-Lacaux et al. (1999) at a 20-year-old African reservoir. One factor determining this will be the relative importance of different sources of the carbon that decays to methane. The rapid decay of soft plant parts from the original forest is probably complete in all of these reservoirs by age six years, but continued inputs come from the watershed in the form of dissolved organic carbon and organic debris brought by inflowing water. It is also generated within the reservoir by primary production, especially by macrophytes, using nutrients supplied to the reservoir from the inflow. In a reservoir like Tucuruí, with large-scale deforestation and consequent soil erosion in the watershed upstream of the dam, these inputs of nutrients and of organic carbon can be expected to continue over the long term at high levels. The coverage of macrophytes declined in the reservoir over the 1986-1994 period, but appears to have stabilized at the level of coverage observed in 1994, when these plants covered 11% of the water surface during the high-water period (de Lima et al., 2000).

Emissions from Tucuruí today would differ from those in 1990. An important factor increasing emissions is that the 1991 power generation data used in the estimate for 1990 was for a period before all of the turbines had been installed in the Tucuruí-I phase of the dam. On the other hand, above-water biomass decay would have slowed over the years as this carbon stock disappears, and the coverage of macrophytes declined from 21% to the plateau level of 11% of the high-water period surface area.

Greater fluctuation in the water level (with more turbines installed) also leads to greater emissions. When the water level in the reservoir falls, vegetation quickly regrows on the exposed land. This soft green biomass rapidly decomposes when the water level subsequently rises and floods the drawdown area, releasing methane under the anoxic conditions that prevail on the bottom. Because these areas are relatively shallow, a substantial portion of the bubbles that form can reach the surface before the methane can be oxidized in the water column. Green vegetation in flooded drawdown zones was found to be a significant source of methane bubbling at Balbina (Bohdan Matvienko, public statement, 24 February 2000).

4.4. Time Preference and Energy Choices

Brazil as a whole emitted 53 × 10⁶ t of carbon annually from fossil fuels in 1990 (La Rovere, 1996). The 7.0-10.1 × 10⁶ t emission of CO₂-equivalent C from Tucuruí in 1990 therefore represents 13-19% of the fossil fuel emission from the entire 170 million Brazilian population. The Tucuruí emission is 1.3-1.9 times that of the fossil fuel burned by the 17 million population of Brazil’s largest city, São Paulo (10% of Brazil’s population).

The above-water wood that produced 25-36% of the emission from Tucuruí in 1990 will eventually disappear. The methane emission that makes up the remainder of the dam’s global-warming impact will decline to a lower plateau, but a poorly quantified part of this will continue as a permanent source. A São Paulo-sized emission source may therefore be permanent. These impacts consider the 100-year global warming potentials without discounting (currently used by the Kyoto Protocol); were discounting or other time-preference weighting mechanisms to be applied, the relative impact of hydroelectric dams could be higher than those calculated here by a factor of two or more (Fearnside, 1997a).

Hydroelectric power generation produces large pulses of CO₂ and CH₄ emissions in the first years after filling the reservoir, while thermal generation produces a constant flux of gases in proportion to the power generated. The analysis of the timing of emissions (Fearnside, 1997a) indicates that the average CO₂ molecule in the atmospheric load contributed by Tucuruí enters the atmosphere 15 years earlier than the average molecule in the comparable load from fossil fuel generation. This means that, considering a 100-year time horizon, a ton of CO₂ emitted by Tucuruí has more global warming impact than a ton emitted by fossil fuel, whether or not discounting is applied to greenhouse gases. If discounting is applied, then the relative impact of the hydroelectric option is increased.

Decisions on the time scale over which dams and their global warming impacts are assessed, and on the weighting for time (as by discounting) over the course of the time horizon, will have dramatic influence on the choices that are made on energy development options. They will also influence the assessment of the worldwide contribution to global warming made by reservoirs. Decisions on time horizons and discount rates should be made to best represent the interests of society. If long time horizons are applied without discounting (or other forms of time-preference adjustment) within the time horizon, the result would be to give little value to delaying global warming. Although no agreement on these issues has yet been reached, this author has advocated using a time horizon of 100 years in conjunction with discounting at an annual rate of about 1%, or its equivalent under an alternative time-preference weighting system (Fearnside, 2000a, b; Fearnside et al., 2000).

When global warming is delayed, the impacts (including human life and other non-monetary impacts) that would have occurred over the course of the delay represent benefits to society. Translating this societal value into the decision-making tools represented by time horizon and discounting will result in giving greater weight to short-term impacts such as the peak of emissions from dam construction and the first few years of impoundment and the short-lived gases such as methane produced by reservoirs. Choice of a 100-year time horizon would be consistent with many dam life-cycle analyses and with the global warming potentials currently adopted in an addendum to the Kyoto Protocol (Decision 2/CP.3) for the Protocol’s first commitment period (2008-2012). A decision for the first commitment period is expected to be made at the Sixth Conference of the Parties in November 2000. Regardless of what decision is made, the increasingly unavoidable human impacts and the very long-lasting nature of global warming mean that international negotiations will continue for many years beyond the first commitment period. This author believes that this process will tend towards increasing weight being placed on time, and consequently to an increase in the impact attributed to emissions from hydroelectric dams relative to those from many other energy alternatives.

5. Conclusions

Hydroelectric dams in tropical forest areas produce substantial emissions of greenhouse gases. Although uncertainty regarding the amount of emission is still high, the magnitude of emissions involved is sufficient to both affect global levels of greenhouse gases and to demonstrate the need for careful comparisons of hydroelectric and other energy options as a part of the decision-making process. Tucuruí, with a global warming impact in 1990 greater than that of the fossil fuel burned by the city of São Paulo, provides a reminder of the potential scale of emissions from the dozens of reservoirs that are planned for construction in Amazonia over the coming decades.

6. Acknowledgments

The National Council of Scientific and Technological Development (CNPq AIs 350230/97-98 & 523980/96-5) and the National Institute for Research in the Amazon (INPA PPI 5-3150 and 1-3160) provided financial support. I thank Evlyn M.L.M. Novo and Ivan Tavares de Lima for use of unpublished information. An earlier version of this discussion was presented at the World Commission on Dams Workshop on Greenhouse Gas Emissions from Reservoirs, Montreal, Canada, 24-25 Feb. 2000. I thank John J. Magnuson for comments.

7. References

Bartlett, K.B., Crill, P.M., Bonassi, J.A., Richey, J.E. and Harriss, R.C.: 1990, ‘Methane flux from the Amazon River floodplain: Emissions during rising water’, J. Geophys. Res. (Atmos.) 95(D10), 16,773-16,778.

Brazil, ELETROBRÁS: 1987, Plano 2010: Relatório Geral. Plano Nacional de Energia Elétrica 1987/2010 (Dezembro de 1987), Centrais Elétricas do Brasil (ELETROBRÁS), Brasilia, Brazil, 269 pp.

Brazil, ELETRONORTE: 1987, Estudos Ambientais do Reservatório de Balbina, Relatório "Diagnóstico" BAL-50-1001-RE, Centrais Elétricas do Norte do Brazil (ELETRONORTE), Brasilia, Brazil, 308 pp.

Brazil, ELETRONORTE: 1988, UHE Tucuruí: Plano de Utilização do Reservatório, Caracterização e Diagnóstico do Reservatório e de sua Area de Influência, TUC-10-26346-RE, Volume I – Texto, Centrais Elétricas do Norte do Brasil (ELETRONORTE), Brasilia, Brazil, 228 pp.

Brazil, ELETRONORTE: 1989, Usina Hidrelétrica Tucuruí: Memória Técnica, Diretoria Técnica (DT), Departamento de Projetos (TPR), Projeto Memória, Centrais Elétricas do Norte do Brasil (ELETRONORTE), Brasilia, Brazil, 681 pp.

Brazil, ELETRONORTE: 1992, Ambiente, Desenvolvimento, Tucuruí, Centrais Elétricas do Norte do Brasil (ELETRONORTE), Brasilia, Brazil, 32 pp.

Brazil, Programa Avança Brasil: 1999, http://www.abrasil.gov.br.

de Lima, I.B.T. and Novo, E.M.L.M.: 1999, ‘Carbon flows in the Tucuruí reservoir’, in L.P. Rosa and M.A. dos Santos (eds.), Dams and Climate Change, Coordenação dos Programas de Pós-Graduação de Engenharia (COPPE), Universidade Federal de Rio de Janeiro (UFRJ), Rio de Janeiro, Brazil, pp. 78-84.

de Lima, I.B.T., Novo, E.M.L.M., Ballester, M.V.R. and Ometto, J.P.: 2000, ‘The role of macrophyte community in the CH₄ production and emission in the tropical reservoir of Tucuruí, Pará state, Brazil’, Verh. Internat. Verein. Limnol. (in press).

de Souza, J.A.M.: 1996, ‘Brazil and the UN Framework Convention on Climate Change’, in International Atomic Energy Agency (IAEA), Comparison of Energy Sources in Terms of their Full-Chain Emission Factors: Proceedings of an IAEA Advisory Group Meeting/Workshop held in Beijing, China, 4-7 October 1994, IAEA-TECDOC-892, IAEA, Vienna, Austria, pp. 19-21.

Devol, A.H., Richey, J.H., Forsberg, B.R. and Martinelli, L.A.: 1990, ‘Seasonal dynamics in methane emissions from the Amazon River floodplain to the troposphere’, J. Geophys. Res. (Atmos.) 95(D10), 16,417-16,426.

Duchemin, E., Lucotte, M., Canuel, R. and Chamberland, A.: 1995, ‘Production of the greenhouse gases CH₄ and CO₂ by hydroelectric reservoirs of the boreal region’, Global Biogeochem. Cycles 9(4), 529-540.

Duchemin, E., Lucotte, M., Canuel, R., Queiroz, A.G., Almeida, D.C., Pereira, H.C., Dezincourt, J. and Ribeiro, L.E.: 2000, ‘Comparison of greenhouse gas emissions from an old tropical reservoir with those of other reservoirs worldwide’, Verh. Internat. Verein. Limnol. 27: 1-5.

Fearnside, P.M.: 1989, ‘Brazil's Balbina Dam: Environment versus the legacy of the pharaohs in Amazonia’, Environ. Manage. 13(4), 401-423.

Fearnside, P.M.: 1995, ‘Hydroelectric dams in the Brazilian Amazon as sources of 'greenhouse' gases’, Environ. Conserv. 22(1), 7-19.

Fearnside, P.M.: 1996a, ‘Hydroelectric dams in Brazilian Amazonia: Response to Rosa, Schaeffer & dos Santos’, Environ. Conserv. 23(2), 105-108.

Fearnside, P.M.: 1996b, ‘Amazonia and global warming: Annual balance of greenhouse gas emissions from land-use change in Brazil's Amazon region’, in J. Levine (ed.), Biomass Burning and Global Change, Volume 2: Biomass Burning in South America, Southeast Asia and Temperate and Boreal Ecosystems and the Oil Fires of Kuwait, MIT Press, Cambridge, MA, USA, pp. 606-617.

Fearnside, P.M.: 1997a, ‘Greenhouse-gas emissions from Amazonian hydroelectric reservoirs: The example of Brazil's Tucuruí Dam as compared to fossil fuel alternatives’, Environ. Conserv. 24(1), 64-75.

Fearnside, P.M.: 1997b, ‘Greenhouse gases from deforestation in Brazilian Amazonia: Net committed emissions’, Clim. Change 35(3), 321-360.

Fearnside, P.M.: 1999a, ‘The potential of Brazil's forest sector for mitigating global warming under the Kyoto Protocol's "Clean Development Mechanism"’, in J.D. Kinsman, C.V. Mathai, M. Baer, E. Holt and M. Trexler (eds.), Global Climate Change: Science, Policy, and Mitigation/Adaptation Strategies: Proceedings of the Second International Specialty Conference, Washington, DC, 13-15 October 1998, Air & Waste Management Association (AWMA), Sewickley, PA, USA, pp. 634-646.

Fearnside, P.M.: 1999b, ‘Social impacts of Brazil's Tucuruí Dam’, Environ. Manage. 24(4), 485-495.

Fearnside, P.M.: 2000a, ‘Why a 100-year time horizon should be used for global warming mitigation calculations’, (manuscript).

Fearnside, P.M.: 2000b, ‘Time preference in global warming calculations: A proposal for a unified index’, (manuscript).

Fearnside, P.M.: 2001, ‘Environmental impacts of Brazil's Tucuruí Dam: Unlearned lessons for hydroelectric development in Amazonia’, Environ. Manage. 27(3): 377-396.

Fearnside, P.M., Lashof, D.A. and Moura-Costa, P.: 2000, ‘Accounting for time in mitigating global warming through land-use change and forestry’, Mitig. Adapt. Strateg. Global Change 5(3), 239-270.

Fearnside, P.M., Leal Filho, N. and Fernandes, F.M.: 1993, ‘Rainforest burning and the global carbon budget: Biomass, combustion efficiency and charcoal formation in the Brazilian Amazon’, J. Geophys. Res. (Atmos.) 98(D9), 16,733-16,743.

Galy-Lacaux, C., Delmas, R., Jambert, C., Dumestre, J.-F., Labroue, L., Richard, S. and Gosse, P.: 1997, ‘Gaseous emissions and oxygen consumption in hydroelectric dams: A case study in French Guyana’, Global Biogeochem. Cycles 11(4), 471-483.

Galy-Lacaux, C., Delmas, R., Kouadio, J., Richard, S. and Gosse, P.: 1999, ‘Long-term greenhouse gas emissions from hydroelectric reservoirs in tropical forest regions’, Global Biogeochem. Cycles 13(2), 503-517.

Gosse, P.: 1999, ‘A system for reoxygenating the water at Petit-Saut’ http://www.edf.fr/der/html/der/environnement/ptiso.en.htm

Grace, J., Lloyd, J., McIntyre, J., Miranda, A.C., Meir, P., Miranda, H.S., Nobre, C., Moncrieff, J., Massheder, J., Malhi, Y., Wright, I. and Gash, J.: 1995, ‘Carbon dioxide uptake by an undisturbed tropical rain forest in southwest Amazonia, 1992 to 1993’, Science 270, 778-780.

Junk, W.J., Robertson, B.A., Darwich, A.J. and Vieira, I.: 1981, ‘Investigações limnológicas e ictiológicas em Curuá-Una, a primeira represa hidrelétrica na Amazônia Central’, Acta Amazonica 11(4), 689-716.

Keller, M., Kaplan, W.A. and Wofsy, S.C.: 1986, ‘Emissions of N₂O, CH₄ and CO₂ from tropical forest soils’, J. Geophys. Res. (Atmos.) 91, 11,791-11,802.

Keller, M. and Stallard, R.F.: 1994, ‘Methane emission by bubbling from Gatun Lake, Panama’, J. Geophys. Res. (Atmos.) 99(D4), 8307-8319.

La Rovere, E.L.: 1996, ‘The prevention of global climate changes and sustainable energy development in Brazil’, in L.P. Rosa and M.A. dos Santos (eds.), Greenhouse Gas Emissions under a Developing Countries Point of View, Coordenação dos Programs de Pós-Graduação de Engenharia (COPPE), Universidade Federal do Rio de Janeiro (UFRJ), Rio de Janeiro, Brazil, pp. 215-225.

Leentvaar, P.: 1966, ‘The Brokopondo Lake in Surinam’, Verh. Internat. Verein. Limnol. 16, 680-684.

Malhi, Y., Nobre, A.D., Grace, J., Kruijt, B., Pereira, M.G.P., Culf, A. and Scott, S.: 1998, ‘Carbon dioxide transfer over a central Amazonian rain forest’, J. Geophys. Res. (Atmos.) 103(D24), 31,593-31,612.

Martius, C., Fearnside, P.M., Bandeira, A.G. and Wassmann, R.: 1996, ‘Deforestation and methane release from termites in Amazonia’, Chemosphere 33(3), 517-536.

Matvienko, B., Rosa, L.P., Sikar, E., dos Santos, M.A., Menezes, F. and Lourenço, R.: 2000, ‘Carbon dioxide and methane emission from some Brazilian reservoirs’, Paper presented at the World Commission on Dams Workshop on Greenhouse Gas Emissions from Reservoirs, Montreal, Canada, 24-25 February 2000, 8 pp.

Matvienko, B. and Tundisi, J.G.: 1996, ‘Biogenic gas release by reservoirs in the Amazon’, Report to Centrais Elétricas do Brasil (ELETROBRÁS), Rio de Janeiro, Brazil, 11 pp.

Matvienko, B. and Tundisi, J.G.: 1997, ‘Biogenic gases and decay of organic matter’, in L.P. Rosa and M.A. dos Santos (eds.), Hydropower Plants and Greenhouse Gas Emissions, Coordenação dos Programas de Pós-Graduação em Engenharia (COPPE), Universidade Federal do Rio de Janeiro (UFRJ), Rio de Janeiro, Brazil, pp. 34-40.

Novo, E.L.M. and Tundisi, J.G.: 1994, ‘Contribution of remote sensing techniques to the assessment of methane emission from large tropical reservoirs’, Remote Sensing Reviews 10, 143-153.

Polunin, N.V.C.: 1984, ‘The decomposition of emergent macrophytes in fresh water’, Advan. Ecological Res. 14, 115-168.

Revilla Cardenas, J.D., Kahn, F.L. and Guillamet, J.L.: 1982, ‘Estimativa da fitomassa do reservatório da UHE de Tucuruí’ in Projeto Tucuruí, Relatório Semestral, Período janeiro/junho 1982, Vol. 2: Limnologia, Macrófitas, Fitomassa, Degradação da Fitomassa, Doenças Endêmicas, Solos, Centrais Elétricas do Norte do Brasil (ELETRONORTE) and Instituto Nacional de Pesquisas da Amazônia (INPA), Manaus, AM, Brazil, pp. 1-11.

Rosa, L.P., dos Santos, M.A., Tundisi, J.G. and Sikar, B.M.: 1997a, ‘Measurements of greenhouse gas emissions in Samuel, Tucuruí and Balbina Dams’, in L.P. Rosa and M.A. dos Santos (eds.), Hydropower Plants and Greenhouse Gas Emissions, Coordenação dos Programas de Pós-Graduação em Engenharia (COPPE), Universidade Federal do Rio de Janeiro (UFRJ), Rio de Janeiro, Brazil, pp. 41-55.

Rosa, L.P. and Schaeffer, R.: 1995, ‘Global warming potentials: The case of emissions from dams’, Energy Policy 23, 149-158.

Rosa, L.P., Schaeffer, R. and dos Santos, M.A.: 1996a, ‘Are hydroelectric dams in the Brazilian Amazon significant sources of 'greenhouse' gases?, Environ. Conserv. 23(2), 2-6.

Rosa, L.P., Schaeffer, R. and dos Santos, M.A.: 1996b. A Model of Greenhouse Gas Emissions from Hydroelectric Plants and an Application to Dams in the Amazon Region of Brazil, Coordenação dos Programs de Pós-Graduação de Engenharia (COPPE), Universidade Federal do Rio de Janeiro (UFRJ), Rio de Janeiro, Brazil, 17 pp.

Rosa, L.P., Schaeffer, R. and dos Santos, M.A.: 1996c, Emissões de metano e dióxido de carbono de hidrelétricas na Amazônia comparadas às termelétricas equivalentes. Cadernos de Energia 9, 109-157.

Rosa, L.P., Sikar, B.M., Sikar, E.M. and dos Santos, M.A.: 1997b, ‘A model for CH₄ and CO₂ emission mean life in reservoir based on data from an Amazonian hydroplant’ in L.P. Rosa and M.A. dos Santos (eds.), Hydropower Plants and Greenhouse Gas Emissions, Coordenação dos Programas de Pós-Graduação em Engenharia (COPPE), Universidade Federal do Rio de Janeiro (UFRJ), Rio de Janeiro, Brazil, pp. 102-111.

Schimel, D. and 75 others: 1996, ‘Radiative forcing of climate change’, in J.T. Houghton, L.G. Meira Filho, B.A. Callander, N. Harris, A. Kattenberg and K. Maskell (eds.), Climate Change 1995: The Science of Climate Change, Cambridge University Press, Cambridge, pp. 65-131.

Sissakian, C. and Desmoulins, G.: 1991, ‘Impacts des retenues en site tropical: Actions entreprises a l'occasion de la réalisation du barrage de Petit-Saut en Guyane Française’, in Seventeenth Congress on Large Dams, Vienna, 1991, International Commission on Large Dams (ICOLD), Paris, France, Q. 64, R. 1, pp. 1-18.

Tian, H., Mellilo, J.M., Kicklighter, D.W., McGuire, A.D., Helfrich III, J.V.K., Moore III, B. and Vörösmarty, C.: 1998, ‘Effect of interannual climate variability on carbon storage in Amazonian ecosystems’, Nature 396, 664-667.

Verchot, L.V., Davidson, E.A., Cattânio, J.H., Akerman, I.L., Erickson, H.E. and Keller, M.: 1999, ‘Land use change and biogeochemical controls of nitrogen oxide emissions from soils in eastern Amazonia’, Global Biogeochem. Cycles 13(1), 31-46.

Vilarrubia, T.V. and Cova, M.: 1993, ‘Estudio sobre la distribución y ecologia de macrófitos aquáticos en el embalse de Guri’, Interciencia 18(2), 77-82.

Walker, I., Miyai, R. and de Melo, M.D.A.: 1999, ‘Observations on aquatic macrophyte dynamics in the reservoir of the Balbina hydroelectric power plant, Amazonas state, Brazil’, Acta Amazonica 29(2), 243-265.

Wassmann, R. and Thein, U.G.: 1989, ‘Spatial and seasonal variation of methane emission from an Amazon floodplain lake’, Paper presented at the Workshop on 'Cycling of Reduced Gases in the Hydrosphere,' SIL Congress, Munich, Germany, 17 August 1989, (manuscript).

WCD (World Commission on Dams): 1999, ‘World Commission on Dams to study Brazil’s Tucurui Dam and Amazon/Tocantins River Basin’, http://www.dams.org.

FIGURE LEGENDS

Fig. 1 – Locations mentioned in the text.

Fig. 2 – Methane concentration profile in Tucuruí. Observed March 1989; data are from measurements by J.G. Tundisi, cited by Rosa et al. (1997a, p. 43). Adjusted annual mean values are calculated as described in the text, based on proportional seasonal variations at Petit-Saut (Galy-Lacaux et al., 1997, 1999).

TABLE I

Parameters for Tucuruí reservoir emission from above-water biomass

Parameter

Value

Units

Source

Above-ground fraction

0.759

Fearnside (1997b, p. 37)(*)

Average depth of surface water zone

meter

Assumption, based on commercial timber spoilage

Leaf decay rate in seasonally inundated zone

-0.5

fraction yr^-1

Assumption; note seasonal drying accelerates rate (Polunin 1984, p. 129).

Above-water decay rate (0-4 yr)

-0.1680

fraction yr^-1

Assumed same as felled forest (Fearnside 1996b, p. 611)(*)

Above-water decay rate (5-7 yr)

-0.1841

fraction yr^-1

Assumed same as felled forest (Fearnside 1996b, p. 611)

Above-water decay rate (8-10 yr)

-0.0848

fraction yr^-1

Assumed same as felled forest (Fearnside 1996b, p. 611)

Above-water decay rate (>10 yr)

-0.0987

fraction yr^-1

Assumed same as felled forest (Fearnside 1996b, p. 611)

Carbon content of wood

0.50

Fearnside et al. (1993)

Rate of wood fall from above-water zone

0.1155

fraction yr^-1

Assumption: average lifetime = 6 yr

Average total biomass of forest at Tucurui

519

t ha^-1

Revilla Cardenas et al. (1982)

Average water depth at minimum level

9.7

meter

Uses 58.0 m above msl as minimum normal operating level (Brazil, ELETRONORTE, 1989, p. 64).

Initial biomass present: leaves

8.8

t ha^-1

Calculated from total biomass and Fearnside (1995, p. 12).

Initial biomass present: wood above water

291.0

t ha^-1

Calculated from total biomass and Fearnside (1995, p. 12).

Initial biomass present: below ground

125.1

t ha^-1

Calculated from total biomass and Fearnside (1995, p. 12).

Methane release by termites

0.687

kg CH₄ ha^-1

yr^-1

Martius et al. (1996, p. 527).

Table 2: AVAILABLE DATA ON EMISSIONS FROM THE TUCURUÍ RESERVOIR SURFACE IN DIFFERENT HABITATS


Habitat		Date	Season	Season	Emission	Emission			Source
			(water	(water	type	(mg CH4/
			level)	flow)		m2/day)

						mean	sd	n
HABITAT DIVISIONS USED IN CALCULATION:

Open water		May 1996	High	High	Bubbling+diffusion	12			(a)
		Aug. 1996	High	Low	Bubbling+diffusion	33.5			(a)
		Dec. 1996	Low	Low	Bubbling+diffusion	65			(a)
		Aug. 1997(b)	High	Low	Bubbling+diffusion	86.5			(c)
	Mean: high water flow					12.0		1
	Mean: Low water flow					61.7	26.6	3

Macrophyte beds		May 1996	High	High	Bubbling+diffusion	73			(a)
		Aug. 1996	High	Low	Bubbling+diffusion	63			(a)
		Dec. 1996	Low	Low	Bubbling+diffusion	72			(a)
	Mean: high water flow					73		1
	Mean: Low water flow					67.5	6.4	2


Standing dead trees		May 1996	High	High	Bubbling+diffusion	56.4			(a)
		Aug. 1996	High	Low	Bubbling+diffusion	59			(a)
		Dec. 1996	Low	Low	Bubbling+diffusion	960			(a)
		Aug. 1997	High	Low	Bubbling+diffusion	74.8			(d)
	Mean: high water flow					56.4		1
	Mean: Low water flow					364.6	515.7	1

OTHER MEASUREMENTS:

Open water		Mar. 1989	High	High	Bubbling	0			(e)
		Sept. 1993(f)	Low	Low	Bubbling	0.018			(g)
		Mar.1993(h)	High	High	Bubbling	14.2			(g)
		Sept. 1993(h)	Low	Low	Bubbling	3.3			(g)

Macrophyte beds		Sept. 1993	Low	Low	Bubbling	19.0			(g)

Standing dead trees		Mar. 1993	High	High	Bubbling	3.3			(g)
		Sept. 1993	Low	Low	Bubbling	24.8			(g)


Weighted averages		Feb.-Mar. 1993	Low	High	Bubbling+diffusion	5.6			(i)
for whole reservoir		Sept. 1993	Low	Low	Bubbling+diffusion	15.8			(i)
		1998 (j)	?		Bubbling	13.1			(k)
		1998 (j)	?		Diffusion	192.2			(k)
		1999 (j)	?		Bubbling	2.4			(k)
		1999 (j)	?		Diffusion	12.2			(k)
-------------
(a) E.M.L.M. Novo, personal communication 1999.
(b) Open water: "tributaries" < 10 m depth.
(c) de Lima et al. nd.
(d) de Lima and Novo 1999.
(e) Rosa et al. 1996b,c 1997a.
(f) Open water: channel.
(g) Rosa et al. 1997a: 48.
(h) Open water: protected cove.
(i) Matvienko and Tundisi 1996: 10.

(j) Month unspecified.
(k) Matvienko et al. 2000.

Table III Calculated areas of habitats in the Tucuruí reservoir in 1988
Month	Mean streamflow (10³m³s^‑1) (a)	Residence time (days) (b)	Volume (10⁹ m³)	Macrophyte area (km²) (c)	Water without macrophytes (km²)	Exposed drawdown area (km²)		Permanently flooded with emergent trees (km²)	Water without trees or macrophytes (km²)	Macrophyte area
										Macrophyte area
										as % of total area	as % of water area



Jan.	15.3	37	48.9	151.6	1,879.8	398.0		396.8	1,483.0	6.2	7.5
Feb.	20.8	27	48.5	151.6	1,879.8	398.0		396.8	1,483.0	6.2	7.5
Mar.	24.3	23	48.3	252.7	1,924.0	252.7		396.8	1,527.2	10.4	11.6
Apr.	23.8	24	49.4	252.7	1,924.0	252.7		396.8	1,527.2	10.4	11.6
May	15.3	37	48.9	505.4	1,924.0	0.0		396.8	1,527.2	20.8	20.8
Jun.	7.7	74	49.2	505.4	1,924.0	0.0		396.8	1,527.2	20.8	20.8
Jul.	4.5	126	49.0	505.4	1,924.0	0.0		396.8	1,527.2	20.8	20.8
Aug.	3.2	177	48.9	505.4	1,924.0	0.0		396.8	1,527.2	20.8	20.8
Sep.	2.4	236	48.9	151.6	1,879.8	398.0		396.8	1,483.0	6.2	7.5
Oct.	2.7	210	49.0	151.6	1,879.8	398.0		396.8	1,483.0	6.2	7.5
Nov.	4.6	123	49.0	151.6	1,879.8	398.0	(d)	396.8	1,483.0	6.2	7.5
Dec.	8.8	64	48.7	151.6	1,879.8	398.0		396.8	1,483.0	6.2	7.5

Mean	11.1	96.5	48.9	286.4	1,901.9	241.1		396.8	1,505.1	11.8	13.1

(a) Brazil, ELETRONORTE (1989, p. 51).
(b) Brazil, ELETRONORTE (1988, p. 87).
(c) In 1989 the macrophyte maximum was in July, when water level was 72 m above msl, and minimum was in November when water level was 68 m above msl (Novo and Tundisi, 1994, p. 150). Intervening months are interpolated based on the assumptions of Novo and Tundisi (1994).
(d) From Novo and Tundisi (1994, p. 149); Fearnside (1995, p. 13) used 858 km² based on water volumes.

TABLE IV Methane sources in the Tucuruí reservoir surface
	High- water area (km²)	Average area (km²)		Share of average area (%)		Average area (km²)				Emission (mg CH₄m^-2day^-1)				Emission (t CH₄)

						High flow (Jan-May)		Low flow (Jun-Dec)		High-flow period (a)		Low-flow period (a)		High-flow period 151 days		Low-flow period 214 days		Total 365 days

Open water
No trees or macrophytes	1,545.5	1,505.1		68.8		1,509.5		1,502.0		12.0		61.7		14,055		19,819		33,873
Standing tree area	407.4		396.8		18.1		396.8		396.8		56.4		364.6		21,844		30,958		52,802
Open water total	1,952.9		1,901.9		86.9		1,906.3		1,898.7								50,777		86,675

Macrophyte beds	294.1		286.4		13.1		262.8		303.2		73.0		67.5		2,679		4,380		7,059

Whole reservoir	2,247.0		2,188.3		100.0		2,169.1		2,202.0						38,578		55,157		93,734

Average emission			114.3								117.8		117.1
(a) Table II.

Table V Calculation of methane export by the Tucuruí turbines in 1991 (a)
Month	Turbine flow (b) (10⁹ m³)	Water level (c) (m above msl)	Correction for seasonal oscillation (%) (d)	Turbine intake depth (m below surface)	Corrected CH₄ concentration in water released by turbines (e) (mg CH₄ liter^‑1)	CH₄ exported by turbines (10⁶ t)



Jan.	10.9	67.5	-17	30.9	6.2	0.0676
Feb.	12.4	67.5	-33	30.9	5.0	0.0622
Mar.	12.4	69.3	-50	32.7	3.8	0.0464
Apr.	12.4	69.3	-33	32.7	5.0	0.0622
May	12.4	72.0	-17	35.4	6.2	0.0770
Jun.	9.4	72.0	0	35.4	7.5	0.0702
Jul.	9.4	72.0	17	35.4	8.8	0.0821
Aug.	6.3	72.0	33	35.4	10.0	0.0632
Sep.	4.8	67.5	50	30.9	11.3	0.0542
Oct.	4.8	67.5	33	30.9	10.0	0.0481
Nov.	9.4	67.5	17	30.9	8.8	0.0821
Dec.	9.4	67.5	0	30.9	7.5	0.0702

Mean	9.5	69.3	0	32.7	7.5	0.0655
Total	113.8					0.7854
(a) Based on 1991 power generation and 1988 storage changes and evaporation (from water and macrophyte areas).
(b) Allocated among months by adding or subtracting from average monthly mean in units of one turbine, so as to maintain positive spillway flow, within the constraints of turbine capacity and total annual electricity generation.
(c) Based on areas (Table 3), interpolated from 10-m intervals (Brazil, ELETRONORTE, 1989, Fig. MT-TUC-05).
(d) Percentage deviation from annual average based on approximate amplitude of oscillations at Petit-Saut from Galy-Lacaux et al. (1999).
(e) Corrected for spillway depth with adjustment for annual mean CH₄ concentration at the turbine depth (Fig. 2) and for seasonal oscillations in CH₄ concentration. For example, in January the annual mean CH₄ concentration for >30 m depth is 7.5 mg CH₄ liter^-1 and the corrected concentration (-17%) is 6.2 mg CH₄ liter^-1.

TABLE VI Calculation of methane export by the spillway in 1991(a)
Month	Inflow (10⁹ m³) (b)	Evapora-tion (10⁹ m³) (c)	Storage change (10⁹ m³) (d)	Spillway flow (10⁹ m³) (e)	Spillway depth (m below surface) (f)	Annual mean CH₄concen-tration at spillway depth (mg CH₄ liter^‑1) (g)	Corrected CH₄ concentration in water released by spillway (mg CH₄ liter^‑1) (h)	CH₄ exported by spillway (10⁶ t)



Jan.	41.0	0.26	0.2	29.7	15.5	2.6	2.2	0.0648
Feb.	55.8	0.26	-0.4	43.5	15.5	2.6	1.8	0.0767
Mar.	65.1	0.28	-0.2	52.7	17.3	3.1	1.5	0.0813
Apr.	63.8	0.28	1.0	50.1	17.3	3.1	2.1	0.1035
May	41.0	0.31	-0.4	28.7	20.0	3.7	3.1	0.0894
Jun.	20.6	0.31	0.2	10.8	20.0	3.7	3.7	0.0403
Jul.	12.1	0.31	-0.2	2.5	20.0	3.7	4.4	0.0111
Aug.	8.4	0.31	-1.2	2.9	20.0	3.7	5.0	0.0145
Sep.	6.3	0.26	-0.1	1.3	15.5	2.6	3.9	0.0050
Oct.	7.1	0.26	0.6	1.5	15.5	2.6	3.5	0.0052
Nov.	12.3	0.26	0.5	2.2	15.5	2.6	3.1	0.0067
Dec.	23.6	0.26	-0.1	14.1	15.5	2.6	2.6	0.0370
Mean	29.7	0.28	0.0	20.0	17.3	3.1	3.1	0.0446
Total	357.0	3.39	0.0	239.8				0.5353
(a) Based on 1991 power generation and 1988 storage changes and evaporation (from water and macrophyte areas).
(b) Based on long-term streamflow (Table 3).
(c) Evaporation without macrophytes is 1548 mm year^‑1 (Brazil, ELETRONORTE, 1989, p. 47); evapotranspiration of macrophyte areas is assumed to be twice this rate.
(d) Based on storage volumes (Table 3).
(e) Calculated by difference from inflow and evaporation + turbines (from Table 5) + storage change.
(f) Based on water levels from Table 5.
(g) Figure 2, using values for the spillway depth with adjustment for seasonal oscillations in CH₄ concentration.
(h) Adjusted with correction for seasonal oscillations in CH₄ concentrations from Table 5.

TABLE VII Net emissions from loss of living forest sources and sinks (a)
	Per-hectare flux					Tucuruí emission (10⁶ t CO₂-equivalent C yr^‑1)			Source of per-hectare value
	Gas (t gas ha^‑1 yr^‑1)		Equivalent carbon (t CO₂-equivalent C ha^‑1 yr^‑1)
Item

Loss of CO₂ carbon uptake from standing forest			1.2		0.3				0.06		Tian et al. (1998) (b)

Loss of N₂O emission from forest soil		-0.0087		-0.734				-0.14		Verchot et al. (1999, p. 37).

Loss of CH₄ uptake from forest soil		0.0005		0.00015				0.000028		Keller et al. (1986).

Loss of CH₄ emission from forest termites		-0.014		-0.104				-0.020		Fearnside (1996b).

Total			-0.52				-0.10

(a) Considering area of lost forest as 1926 km² (Fearnside, 1995, p. 11). IPCC SAR 100-year global warming potentials are used: CO₂ = 1, CH₄ = 21, N₂O = 310 (Schimel et al., 1996). Negative values represent reduced emission to the atmosphere when forest is lost.
(b) Based on modeled 1980-1994 average.

TABLE VIII Greenhouse gas emissions from Tucuruí in 1990 (a)
Gas	Emission source		Flux (10⁶ t gas)					CO₂-equivalent C					Relative contribution
						(10⁶ t C) (b)					(%)
			High	Low		High		Low		High		Low
			scenario	scenario		scenario		scenario		scenario		scenario

CH₄	Bubbling + diffusion		0.0937		0.0937		0.537		0.537		5%		8%
	Above-water decay (c)		0.0005		0.0005		0.003		0.003		0.03%		0.04%
	Loss of forest soil sink		0.0001		0.0001		0.001		0.001		0.01%		0.01%
	Loss of forest termites		-0.0027		-0.0027		-0.015		-0.015		-0.15%		-0.22%
	Turbines		0.7025	0.1649		4.023		0.945		40%		13%
	Spillway		0.5353	0.5353		3.066		3.066		30%		44%

		Total CH₄	1.3294	0.7919		7.61		4.54		75%		64%

CO₂	Above-water decay		9.34		9.3400		2.55		2.55		25%		36%
	Below-water decay		0.11		0.1100		0.03		0.03		0.30%		0.43%
	Loss of uptake from forest		0.23		0.2300		0.06		0.06		1%		1%
		Total CO₂	9.68	9.68		2.64		2.64		26%		38%

N₂O	Loss of forest soil source		-0.00167		-0.00167		-0.14		-0.14		-1%		-2%

Total						10.11		7.03		100%		100%
(a) Components are from various years: habitat areas and water levels from 1988, per-area bubbling and diffusion from 1996-1997, turbine and spillway water flow from 1991, CH₄ content of water from 1989, decay emissions from 1990.
(b) Global warming potential of CH₄ = 21; N₂O = 310 (Schimel et al., 1996).
(c) Fearnside (1995) based on above-ground decay in forest felled for agriculture and ranching (Martius et al., 1996).

TABLE IX Comparison with other estimates of greenhouse gas emissions from Tucurui
Author	Year of emission	Factors included (a)	CH₄ flux per unit area (mg CH₄ m^-2 day^‑1)	Net annual emission (10⁶ t gas)				CO₂-equivalent C (10⁶ t C yr^‑1)		Method for CH₄ estimate
				CO₂	CH₄	N₂O

This study: Low scenario	1990	1,2,3,4,5,6,7		9.7	0.79	-0.00167		7.0		(b)
This study: High scenario	1990	1,2,3,4,5,6,7		9.7	1.33	-0.00167		10.1		(b)

Fearnside, 1995	1990	(1,2,3), 4		9.5	0.09	--	3.1		(c)

Rosa and Schaffer, 1995	1990	(1,2,3)		--	0.52	--		3.0		(c,d)

Novo and Tundisi, 1994	1988	1,2	96	--	0.085	--		0.49		(b)

Rosa et al., 1996c, 1997b	1993	1	15	--	0.013	--		0.07		(b,e)

Matvienko et al., 2000	1998-99	1,2	112	--(f)	0.099	--		0.57		(b,e)

Matvienko and Tundisi, 1997	Sep.1993	1,2	15	--(f)	0.013	--		0.08		(b,e)
(a) Factors: 1 = bubbling from surface, 2 = diffusion from surface, 3 = turbines, 4 = above-water decay, 5 = forest soil CH₄, 6 = forest soil N₂O, 7 = forest termites; parentheses ( ) = implicitly included.
(b) Based on flux data.
(c) Based on assumptions regarding rates of decomposition and fraction emitted as CH₄.
(d) CH₄ emission calculated for 1990 from assumptions of Rosa and Schaffer (1995, p. 155) as mean of two scenarios, and converted to CO₂-equivalent C using the IPCC 100-year GWP of 21 (Schimel et al., 1996).
(e) Reservoir emissions calculated from reported per-m² value using an area of 2430 km².
(f) CO₂ bubbling measured, but cannot be considered a net emission because much is derived from carbon input from the watershed and from primary production in the reservoir.