The Colorado pikeminnow was once a prized catch. The toothless, torpedo-shaped fish had dark, tender, flavorful meat and came in absurdly large sizes, especially for a minnow. In the early 1900s, fishers reported catching some up to six feet long, using mice, frogs and cottontail rabbit heads to lure in the voracious eaters. They were so plentiful that people could catch them using pitchforks.
Today, these sightings are recounted as if from a history book. In the 1960s, the fish were federally listed as endangered — in part because of the construction of several dams.
Hydropower dams blocked the fish’s migrations for spawning, altered river flow and churned cooler water downstream. The Colorado pikeminnows, which were not accustomed to the cooler waters, were soon outcompeted for food by nonnative fish. Now, most Colorado pikeminnows reach only two to three feet long.
“Those fish are now endangered and have been replaced with cold-water-adapted trout,” said Gordon Holtgrieve, a fish ecologist at the University of Washington. “The river doesn’t look anything like it used to, and Native Americans of the region, who traditionally used these fish, have lost part of their culture.” He said the trout are part of highly prized recreational fishing in the area.
The ubiquitous dams around the world are built to guard against extreme flooding, meet steadily increasing water demands and provide hydroelectric power. They also alter river ecosystems — such as by changing temperatures downstream — and can substantially change nearby fish populations.
In China, the Xinanjiang and Danjiangkou hydroelectric dams caused the peak summer temperature to decrease 7.2 to 10.8 degrees (4 to 6 degrees Celsius) in the downstream reaches of nearby rivers. Fish spawning was delayed by three to eight weeks, causing the local extinction of many of the warm-water fish. The Keepit Dam in Australia also reduced temperatures in the Namoi River, disrupting thermal spawning cues for many native fish.
Worldwide, at least 3,700 medium and large hydropower dams are planned in the coming decades or under construction, heavily concentrated in South America, Africa and South and East Asia. Hundreds of millions of people in large river basins in these areas rely heavily on the river for their livelihoods, Holtgrieve said. For example, Cambodians receive about 80 percent of their animal protein from primarily wild-caught freshwater fish from the Mekong River.
Now, in a recent study, researchers have created a first-of-its-kind machine learning model that can predict temperature changes as a result of dams planned around the globe and could help planners and engineers mitigate the environmental impact. Analyzing future dams worldwide, the team found some dams changed downstream temperatures by as much as 10.8 degrees Fahrenheit (6 degrees Celsius).
Based on the research, the team created a public tool that allows people to plug in the dimensions of a future dam and learn how it will affect downstream temperatures.
“The Congo, Amazon and Mekong basins are going to have a large number of dams, and that is inevitable,” said Shahryar Ahmad, the lead author of the study. “We don’t want to repeat the same mistakes, or at least some of the disadvantages, that we are seeing from the dams that have been built over the past century.”
Cooler in the summer, warmer in the winter
Like layers of a cake, large bodies of water typically have different temperatures at different depths, known as thermal stratification. Colder, denser layers gravitate toward the bottom, while a relatively warmer layer heated by the sun sits near the surface.
Hydropower dams generally operate by drawing water from the deeper layers of a reservoir into a turbine for energy. This brings colder waters downstream and causes a cooling effect in the summer; the effect reverses in the winter. Some also draw water from the surface or have shallower reservoirs, which could create warmer downstream temperatures.
“This kind of has been known, but in developing countries where you have so little data, it’s hard to track,” said Faisal Hossain, a professor at the University of Washington who along with Holtgrieve was a study co-author. “It was the extent of the cooling or the warming that we could detect that was surprising, sometimes 6 degrees, which can be pretty substantial.”
The team first analyzed the thermal impact from more than 100 existing dams of various sizes, depths and characteristics in the United States, South Asia, Europe and South America. In the United States, the team examined historical records from temperature monitoring stations upstream and downstream near the dams. In places without ground observations, they used thermal satellite observations to investigate temperature differences.
They found that most U.S. dams produced a cooling effect downstream in the summertime and a warming effect in the winter.
The extent of the cooling or warming is related to the size and depth of a dam’s reservoir. Dams connected to larger and deeper storage pools, which had stronger thermal stratification, had the largest temperature differences from upstream to downstream. Smaller storage reservoirs had weaker stratified water layers, which could be easily mixed by winds, and led to smaller downstream cooling effects or even warming.
“Depending on … the storage capacity of the reservoir and the ability to stratify the water, that’s where the thermal cooling and thermal warming arises from,” said Ahmad, who conducted this research as a doctoral candidate at the University of Washington and is now a researcher at NASA.
The location of the dams also mattered. Dams in arid and warm regions with hotter summers, and thus warmer surface layers, tended to have more downstream cooling. Dams in humid and snowy climates, and thus cooler air temperatures, did not show a strong cooling impact and sometimes produced downstream warming.
Using this information, the team trained its machine learning model to predict the impact of future dams in the summer and winter.
The future of dams — and fish
Of the 216 future dams studied, about 73 percent would potentially cool downstream rivers during the summer by up to 11.9 degrees (6.6 degrees Celsius) compared with upstream. About 25 percent were predicted to warm downstream rivers by up to 8.5 degrees (4.7 Celsius). In the winter, most dams warmed the downstream rivers by up to 3.6 degrees (2 degrees Celsius).
Some of the areas most profoundly affected by dams appeared in the biodiverse Amazon, Paraná, Niger and Mekong basins. Basins in the Amazon will experience moderate cooling and warming at various dams. Dams in the Niger basin are likely to cause warming downstream. The Paraná basin will experience moderate cooling in the summers.
The team also looked at how river temperatures near dams would become altered with climate change if humans continued emitting greenhouse gases. Under increased warming, warmer downstream river temperatures would get warmer, and cooler temperatures would also get warmer in both summer and winter.
“I was most surprised just at the scale, where a lot of these predicted dams are going in, and just the scale of a problem,” said Lindsey Bruckerhoff, a fish ecologist with the U.S. Geological Survey who was not involved in the research. “In a lot of the basins in South America and Asia where they want to put some of these dams, there are still important fisheries there that people actually do have livelihoods on.”
The team designed the machine learning model as a publicly available tool so people can model the effect of their future dams on river temperature.
“Somebody can just plug in the size of a dam, and then they can know, ‘If my dam is this big and this wide and has this massive capacity, it will cause thermal change of this much range,’” Ahmad said. “This is where our study maybe comes in really handy.”
Hossain said the monitoring tool could help inform dam operations so downstream temperatures remain in a tolerable range for the wildlife. He said perhaps some dams can draw less water from the deeper, colder layers and spill more from warmer waters near the surface. Previous work by the researchers also showed that hydropower dams can be more efficient by taking weather forecasts into account.
Ahmad said replacing dam power using smaller turbines, or several smaller ones, can also reduce the environmental footprint in some cases.
Perhaps, Hossain said, a solution is not building the dam at all, if the environmental effects are too disruptive in that region.
“Maybe we can look at what the temperature changes might be, bring the community together and explore other alternatives in a cost-effective way,” Hossain said, “and then thereby minimize the negative impacts on Mother Nature.”
An earlier version of this article misidentified Bruckerhoff as Bruckeroff. The article has been corrected.