To the Editor —

Dams change downstream inundation dynamics, interrupt fish migration and compromise aquatic biodiversity1. These and some other ecological impacts of the hydropower boom now underway at low latitudes are receiving increasing attention1,2. However, greenhouse gas (GHG) emissions, although sometimes substantial, are frequently not considered as part of the environmental licensing process. Carbon dioxide (CO2) and methane (CH4) production in hydroelectric reservoirs stem from the decomposition of flooded organic matter — especially in the first years after impoundment3 — although emissions can remain substantial for many years if organic matter inputs are sufficient. In thermally stratified reservoirs, CH4 produced in sediments, anoxic water and flooded soils tends to accumulate in anoxic bottom waters, eventually outgassing to the atmosphere4,5. Given the warm temperatures, large supplies of organic matter and prevalence of anoxia in Amazonian reservoirs, high rates of GHG emissions are both expected and observed3,4. In some cases, emissions can surpass those from fossil fuel-based power plants4.

After the imbroglio over the approval of the Belo Monte dam5, now under construction, Brazil is planning an equally controversial hydroelectric complex7. To be located in the pristine and densely forested Tapajós River basin in the eastern Amazon6, the five new dams of the planned Tapajós Hydroelectric Complex (THC) will flood an area three times larger than Belo Monte (∼2,000 km2) to generate about the same amount of energy (∼11,000 MW).

On the basis of the average emissions of Amazonian reservoirs3 (updated with data from Balbina reservoir4,8), if built, the THC would probably release approximately 6.7 Tg of CO2-equivalent (CO2e) emissions per year from the surface through diffusive flux. This is more than twice that estimated for the highly criticized hydroelectric complex of the same size planned in the Mekong River in tropical Asia9. Considering two different scenarios (that is, the highest and lowest CO2 and CH4 emission rates reported for Amazonian reservoirs4), the emissions could range from 2.4 to 36.4 Tg CO2e. Hence, even the most optimistic scenario suggests large emissions.

Although the contributions of CO2 and CH4 to total emissions can vary between reservoirs, each gas would probably be responsible for 50% of CO2e evasion. Usually, CH4 is more problematic than CO2, as the conditions created in reservoirs (for example, low oxygen concentrations, formation of littoral marshes) frequently favour CH4 production3,5,8.

Our estimates of GHG emissions, although significant, are probably underestimates. First, we did not consider GHG emissions from the outflowing river or degassing on passage through the turbines. These pathways could nearly double estimated GHG emissions from the THC8,10. Second, evasion of GHGs from newly created marshes in the drawdown area of the reservoirs is not taken into account, and may represent up to 20% of surficial emissions5.

Future studies considering the variability within and between reservoirs will certainly improve estimates. Nonetheless, although hydropower generally results in lower GHG emissions than fossil fuel-based electric power11, the existing assessment clearly indicates that hydroelectric reservoirs (especially Amazonian ones) are far from being GHG-neutral. Several further dams are planned in the Amazon1,6, and if GHG emissions from Amazonian reservoirs continue to be overlooked in the environmental licensing process, Brazil's reputation of promoting GHG-neutral economic growth will be seriously compromised. Using models to project GHG emissions from future reservoirs is more than an academic exercise, such projections can guide measures to reduce emissions — possibly through the formation of reservoirs that are morphometrically less favourable to GHG emission — and help to avoid constructing dams in biomes where emissions tend to be high.