Industries / Energy, water and waste / Power generation / Hydroelectric
Fish and river health
This work limits the harm a hydro plant does to the river: getting fish past dams and turbines, releasing water in ways closer to a natural river, and taking dams out when the river is worth more than the power.
7
Milestones achieved, from 1938 to 2024.
2
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1938
Ladders for a big dam
When Bonneville Dam on the Columbia River was completed in 1938, the US Army Corps of Engineers opened three fish ladders, channels of stepped pools that fish swim up, with fish traps and elevators as backup. They were built to carry millions of adult salmon and steelhead a year over a 65-foot (20 m) dam. The Northwest Power and Conservation Council's history says this had never been done on such a scale, "and it worked."
Credit
US Army Corps of Engineers; Harlan Holmes of the US Bureau of Fisheries helped design the fish passage
1947
Turbines do kill fish
A 1938 Corps report said passing through a turbine was of little danger to fish. In 1947 Ivan Donaldson, one of the Corps' first fish biologists, became the first to study how many fish die on that route, by the Northwest Power and Conservation Council's account, and in 1952 he put the death rate of young fish at Bonneville at about 15%. His 1942 proposals, screens to steer young fish away from turbine intakes into bypass channels, are still in use. Later work found that sudden drops in pressure, more than the blades, cause most of the deaths.
LessonHarm to fish has to be measured; the design-stage belief that turbines were safe was wrong.
Credit
Ivan Donaldson, US Army Corps of Engineers
1996
A flood on purpose
Glen Canyon Dam, finished in 1963, trapped the sand that floods once carried to the sandbars of the Grand Canyon downstream, and the bars shrank. In 1996 the US Department of the Interior released the first controlled flood from the dam, through its turbines and bypass outlets. Sandbars grew within days, but much of the new sand washed away within 6 to 12 months of normal dam operation. The study team learned that sand from side streams moves on quickly, so later floods, in 2004 and 2008, were timed to follow side-stream floods and built large bars. Since 2012 the department has scheduled floods by measuring how much new sand has arrived.
LessonA planned flood helps only when it is timed to the sand that side streams have just delivered.
Credit
US Department of the Interior; US Geological Survey Grand Canyon Monitoring and Research Center (Paul Grams, John Schmidt, David Topping, Theodore Melis and colleagues)
2011
Switzerland sets a deadline
Switzerland's revised Waters Protection Act of 2011 requires owners of existing hydro plants to remove serious harm to rivers by 2030. The cantons counted about 1,000 hydro plant obstacles to fish migration, 100 plants that must damp hydropeaking (sharp daily rises and falls in river flow as a plant starts and stops), and 140 hydro plants and 360 other works that starve rivers of gravel. Owners get the full eligible cost back from a fund paid by a surcharge on the national grid; CHF 471.5 million had been promised by the end of 2023. By the end of 2022, 111 of about 1,000 reported projects, about 10%, were under way or finished, by the Federal Office for the Environment's (BAFU) count.
LessonEven fully funded, refitting old plants for fish is slow: after eleven years about a tenth of the work had started.
Credit
the cantons and the Federal Office for the Environment (BAFU), under the 2011 Waters Protection Act
2019
A runner shaped for salmon
In 2019 a new fixed-blade runner (the spinning part of a turbine), designed for safer fish passage, went into Unit 2 of Ice Harbor Dam on the Snake River in Washington. Researchers at Pacific Northwest National Laboratory (PNNL) measured it with Sensor Fish, fish-sized instruments that record pressure and impacts on their way through. At the highest flow setting tested, the share of readings with pressure low enough to harm salmon (below 86.2 kPa) fell from 19.2% with the old turbine to 0%, and the new turbine was about 4% more efficient. Ice Harbor is the first of 90 large turbines in the Federal Columbia River Power System planned for replacement with fish passage as the main goal.
Credit
US Army Corps of Engineers; PNNL (J. J. Martinez, T. Fu and colleagues, published May 2025)
2022
One fish in five dies
Researchers at the Leibniz Institute of Freshwater Ecology and Inland Fisheries (IGB) in Berlin pooled field tests of more than 275,000 fish of 75 species at 122 hydro sites in 15 countries. On average 22.3% of fish passing through a turbine were killed or badly injured. Larger fish fared worse; slow-turning very-low-head turbines and water wheels did less harm, and some conventional plants did far better than others. An earlier review at Oak Ridge National Laboratory found that Francis turbines, which make 56% of US hydropower, had the highest fish deaths of any turbine type but had been studied less than Kaplan turbines.
Credit
Johannes Radinger, Christian Wolter and colleagues, IGB; Brenda Pracheil and colleagues, Oak Ridge National Laboratory
2024
Four Klamath dams removed
On 2 October 2024 the Klamath River Renewal Corporation announced that four hydroelectric dams on the Klamath River in California and Oregon, Copco No. 1, Copco No. 2, J.C. Boyle and Iron Gate, had been fully removed. They had blocked fish for more than a century and supplied no irrigation, drinking water or flood control; federal regulators approved their removal in November 2022. On 16 October, biologists of the Oregon Department of Fish and Wildlife (ODFW) found a fall-run Chinook salmon in a stream above the former J.C. Boyle Dam, which ODFW called the first anadromous fish (one that lives in the sea and spawns in rivers) back in Oregon's part of the basin since 1912.
LessonSalmon reached the reopened river within two weeks of the last work in the channel.
Credit
Klamath River Renewal Corporation; Kiewit (contractor); PacifiCorp (former owner); the Klamath Tribes and other tribes, who led the push for removal
Watch for
AI reads hydro licenses
Oak Ridge National Laboratory (ORNL) published a database of the environmental measures, such as minimum river flows and fish ladders, that the Federal Energy Regulatory Commission (FERC) requires in hydropower licenses. It covers 497 licenses issued from 1998 to 2025, which make up 478 of the 1,006 FERC projects active in September 2026, and lists 19,113 mentions in 128 categories, each with the license text behind it. A BERT language model found the measures and a subject-matter expert reviewed and corrected its results; ORNL calls the method reproducible and plans to add new licenses.
Achieved whensomeone outside ORNL checks a random sample of entries against the licenses and publishes how many were missed or misfiled.
Claimed by
Thomas Ruggles, Hong-Jun Yoon, Arjun Bhattacharyya and Debjani Singh, ORNL, July 2025, updated September 2026
Salmon breed above the lake
In autumn 2025 Chinook salmon passed Keno Dam (24 September) and the Link River fish ladder (6 October) and reached the Williamson and Sprague rivers above Upper Klamath Lake, ODFW reported. On 18 March 2026 the Klamath Tribes' fisheries department caught young Chinook in traps on the Sprague and Williamson rivers, which it calls the first naturally born salmon in Upper Klamath Lake's tributaries in more than a century.
Achieved whenadult salmon that hatched in the upper basin come back from the ocean to spawn there, counted by ODFW or the Klamath Tribes.
Claimed by
the Klamath Tribes, March 2026, after ODFW's October 2025 report
Open
01
Check releases against license limits
For hydro projects whose license sets a minimum flow or a ramping rate (how fast the flow below the plant may rise or fall), and that have a public stream gauge at the point where the limit applies, a table of how often the river met each limit, hour by hour, over 2015 to 2025. Each limit links to the license text, and each long breach is matched to any deviation report the operator filed.
The briefCredit · unclaimed02
A national hydropeaking map
A public map and table of how sharply river flow rises and falls each day below every US hydro plant that has a USGS gauge with 15-minute records downstream, using published hydropeaking measures, for 2015 to 2025, with each plant matched to its operating mode.
The briefCredit · unclaimed
Open 01 · the brief
Check releases against license limits
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
For hydro projects whose license sets a minimum flow or a ramping rate (how fast the flow below the plant may rise or fall), and that have a public stream gauge at the point where the limit applies, a table of how often the river met each limit, hour by hour, over 2015 to 2025. Each limit links to the license text, and each long breach is matched to any deviation report the operator filed.
Why it is still open
ORNL's 2022 dataset lists the flow requirements in 50 randomly chosen FERC licenses but records what licenses require, not whether projects meet them. The checks that exist cover one region or one river and rely on operators' own reports: a 2012 review by British Columbia's natural resources ministry of 16 small hydro plants counted 92 and 101 ramping breaches in 2010 at the two Clowhom plants alone, and found periods when operators' gauges recorded nothing. A UC Davis student paper compared flows below Iron Gate Dam with recommended river flows for one river. Web and GitHub searches found no check across many projects against independent gauges.
Built with what exists
ORNL's environmental flow dataset and its AI-built mitigation database; the US Geological Survey (USGS) Water Data API for 15-minute river flow; license orders and deviation reports from FERC's eLibrary; a model that reads long legal documents and turns each limit into a testable rule, such as Mistral Large 4, whose weights are free to download (see the Note "A trillion open parameters"); Python with pandas.
First step
Download ORNL's environmental flow dataset, list the projects that have a USGS gauge at their compliance point, and check one project's 2024 record by hand.
Done looks like
At least 30 projects checked, each with yearly hours below its minimum flow and counts of ramping breaches; every breach longer than one hour compared with the operator's filed reports; code and results published, and five projects rechecked by hand by a second person with the same results.
Open 02 · the brief
A national hydropeaking map
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public map and table of how sharply river flow rises and falls each day below every US hydro plant that has a USGS gauge with 15-minute records downstream, using published hydropeaking measures, for 2015 to 2025, with each plant matched to its operating mode.
Why it is still open
A September 2025 ORNL and PNNL study used 15-minute data from 69 USGS gauges, only in the Southeast and Pacific Northwest. A 2021 study of 500 gauges in the US and Canada measured weekly cycles from daily data, which miss swings within a day. Web and GitHub searches found free tools that compute hydropeaking measures, such as the HEDA R package and scripts for Carolli and colleagues' HP1 and HP2 indicators, but no national table of sub-daily swings.
Built with what exists
The USGS Water Data API; ORNL's Existing Hydropower Assets plant list to match gauges to plants; the HEDA R package or Python code for HP1 and HP2; an AI coding agent to write the download and matching scripts.
First step
Pick ten USGS gauges below hydro plants in Georgia and North Carolina, run HEDA on their 2024 15-minute records, and compare the results with the 2025 study's figures for the Southeast.
Done looks like
At least 300 plant and gauge pairs in at least 30 states, each with yearly hydropeaking values; results for the Southeast and Pacific Northwest set beside the 2025 study's, with any difference explained; data and code published under an open license.