Site icon Conservation news

AI model shows how Amazon dams can be made less environmentally damaging

  • Researchers have developed a model using artificial intelligence to analyze the environmental impacts of 351 hydropower dam projects currently under evaluation in the Amazon Basin.
  • The model aims to provide information that would help planners and policymakers optimize the capacity and location of new dams to minimize their negative impacts.
  • It also shows, however, that no proposed dam could ever have zero impact across all the environmental criteria, and that social impacts on local communities remain far too complex to model with AI.
  • While other researchers have welcomed the new way of modeling the risks, they recommend an end to high-capacity hydropower projects in the Amazon and a greater focus on solar and wind power instead.

Hydropower accounts for more than 60% of the electricity generated in Brazil. And the network of mighty rivers in the Amazon Basin is responsible for most of this potential. Today, there are more than 300 proposals for new dams under consideration in the region.

But while hydropower is considered a source of clean energy — it emits no greenhouse gases when generating power — the construction of the dams, and the environmental and social impacts entailed in the process, can be devastating to the environment and to traditional communities.

A new study published in Science magazine shows that artificial intelligence can be used to help identify the best and worst locations for building new hydropower plants. What sets it apart from other studies on the issue is that it considers the entire gamut of impacts — not only from new undertakings, but also of existing ones — whereas convention analyses have tended to focus only on individual projects.

The study, led by Alexander Flecker, a biology professor at Cornell University in the U.S., presents a method to evaluate the pros and cons related to a hydropower plant’s capacity, its impacts, and a set of environmental criteria including water flow, sediment transport, waterway connectivity and fish biodiversity.

According to Flecker, the main difference between this method and ones that evaluate dams individually is that the latter fail to consider the broader scenario of consequences that accumulate over time with multiple undertakings. These can manifest themselves on much larger physical scales than projections focused on individual dams.

“Sometimes projects can seem like they have less impact on a small scale but are rarely seen in solutions with greater physical scale,” Flecker says. “Our approach is able to carry out a retroactive analysis and show the environmental costs or the lost benefits of dam-by-dam planning. In some cases, the lack of coordinated planning revealed a great loss of benefits.”

The map shows the loss of connectivity resulting from the existing hydroelectric dams in the Amazon Basin (in yellow) and the foreseen impact of the dams proposed for construction (in purple). If all were built, only half the basin would still have connected rivers (in blue). Image by Flecker et al. (2022).

At least 158 hydroelectric dams are either operating or under construction today in the five nations composing 90% of the Amazon Basin (Brazil, Bolivia, Peru, Ecuador and Colombia). Another 351 proposed dams are under evaluation for the region.

The study’s authors note that the existing distribution and the hydropower potential are not equal throughout the Amazon. Most of the new projects being eyed are for the Tapajós River in the Brazilian Amazon or the Marañón River in the Peruvian Amazon.

Gordon Holtgrieve, a professor of ecology at the University of Washington in the U.S., says the study offers a new paradigm for public policy around dam building. “The information shows which dam configurations would have the most or the least impact from an environmental perspective,” says Holtgrieve, who was not involved in the study. “This makes it possible to filter out very damaging situations, which should be avoided.”

He says a key finding from the study is that none of the dam configurations was considered good with regard to all of the environmental factors. “This means that difficult decisions must be made, and this study offers new data to inform these choices. It also became evident that achieving the greatest benefits at the lowest cost will call for cooperation between nations,” Holtgrieve says.

An Amazonian question

While the study’s results show there are ways to plan new dams so as to minimize their impacts, the history of hydropower projects in the Amazon is marked by environmental and social problems that should call into question the sustainability of continuing to generate electricity at scale this way, others say.

Philip Martin Fearnside, a researcher at Brazil’s National Institute of Amazonian Research (INPA) and a member of the Network for Nature Conservation Specialists (RECN), says building dams to have minimal impacts could work in theory. But their negative consequences would still far outweigh any benefits that may come from their construction, he says.

“Instead of suggesting that dams be built in places where they create fewer impacts, I’m of the opinion that we should reject any hydropower plant in the Amazon with installed capacity over 10 megawatts,” Fearnside says.

He notes that installed capacity of 10 MW is the most commonly used global limit for classifying small hydropower plants, a standard used in Brazil until 2004. Since then, however, “small” plants have been redefined to include those with a capacity of up to 50 MW, creating a loophole for the construction of dams without federal licensing; small plants only need to be licensed at the state level, which avoids the need for a more rigorous approval process

Distribution of the 158 hydropower dams currently operating in the Amazon Basin today (in red) and the 351 under consideration (in yellow) with their installed capacity. Image by Flecker et al. (2022).

As things stand, many of the projects being proposed meet Brazil’s definition of a small hydropower plant, and could in theory be approved and built. But rather than add capacity to meet demand for electricity, planners and policymakers should be looking at ways to reduce demand, Holtgrieve says.

“It’s simply impossible for 8 billion people to live on the planet with a high standard of living without degrading the environment,” he says. “The real questions are: ‘How can we minimize environmental impact and still have healthy, stable human societies?’ and “What are we willing to give up in order to preserve the very important ecosystem services that people enjoy?’”

In Brazil, the poster child for the hydro-as-clean-energy paradox is Belo Monte, built in the Xingu River in the state of Pará. Construction ran from 2011 to 2019.

“Belo Monte cost more than double its estimated budget. This is merely a financial problem. The social and environmental impacts don’t figure into this calculation,” Fearnside says. “Plus, the climate changes we are expecting worsen the scenario for hydropower plants in the Amazon, including the ones operating today.”

Fearnside says Brazil enormous potential for generating power from other renewable sources, including solar and wind. “There is huge potential for installing tall [wind] towers along the coast of the continental shelf. And every rooftop in Brazil could be used to generate solar power.”

Study co-author Carla Gomes, director of Cornell’s Institute for Computational Sustainability, says a certain amount of dam building could provide high gains in terms of energy generation and economic development, at an acceptable cost of environmental impacts.

But she acknowledges that the complex issue of building hydropower dams in the Amazon includes factors that can’t be measured by artificial intelligence. There are social, political, economic and cultural aspects that technology can’t yet account for.

“The proposal [in the study] is part of a process that is larger and that we feel should be adopted for strategic planning in large hydrographic basins,” Gomes says. “The artificial intelligence approach functions as an initial filter in the first stages of the process. Our model plays an essential role in directing the focus toward the best locations to build, but it doesn’t serve as a substitute for field work with traditional communities.”


Flecker, A. S., Shi, Q., Almeida, R. M., Angarita, H., Gomes-Selman, J. M., García-Villacorta, R., … Gomes, C. P. (2022). Reducing adverse impacts of Amazon hydropower expansion. Science375(6582), 753-760. doi:10.1126/science.abj4017

Banner image of the Tucuruí hydroelectric plant in the state of Pará, Brazil, by PGEDAM (NUMA/UFPA) (CC BY-NC).

This article was reported by Mongabay’s Brazil team and first published here on our Brazil site on March 10, 2022.

Exit mobile version