Industries / Construction and property / Heavy civil / Dredging
Contaminated sediment
Some harbor and river bottoms hold mercury, PCBs (polychlorinated biphenyls, long-lasting industrial oils) and other poisons from old factory discharges, and this work digs them out, covers them or treats them in place, then deals with what was removed.
6
Milestones achieved, from 1977 to 2015.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1977
Mercury dredged from Minamata Bay
From 1977 to 1990 the Kumamoto Prefectural Government dredged about 1.5 million m³ of bottom sediment holding more than 25 ppm of mercury from Minamata Bay in Japan, where mercury in factory wastewater had caused Minamata disease, and sealed it into 58 ha of reclaimed land. By Japan's Ministry of the Environment's figures, the chemical company Chisso paid about ¥30 billion and the national and prefectural governments about ¥9 billion each. Nets put up in 1974 to keep polluted fish inside the bay came down in October 1997, after mercury in seven fish species had stayed below the control values (0.4 ppm total mercury, 0.3 ppm methylmercury) for three years running.
LessonFish were declared safe seven years after the dredging ended, not when it ended.
Credit
Kumamoto Prefectural Government; Chisso; Government of Japan
1993
Hamburg washes its harbor mud
In March 1993 the port of Hamburg started METHA (Mechanical Treatment of Harbour Sediment), a plant for dredged port mud. Pollutants cling mostly to the finest grains, so the plant uses hydrocyclones, spinning separators, to split silt finer than 63 micrometers from the sand, then dewaters sand on sieves, fine sand on a vacuum belt and silt in filter presses. The cleaner sand can be reused. The Central Dredging Association (CEDA) gives its capacity as up to 0.9 million m³ of sediment a year. The title of a 1993 paper by its engineers calls it "the first large-scale plant for treatment of harbour sediments."
Credit
Port of Hamburg (today the Hamburg Port Authority); engineer H.-D. Detzner and colleagues
2005
No default cleanup method
The US Environmental Protection Agency (EPA) issued its Contaminated Sediment Remediation Guidance for Hazardous Waste Sites in December 2005. It compares three approaches: monitored natural recovery (letting clean sediment bury the old while it is checked), capping (covering it with clean material) and dredging. It states that "there is no presumptive remedy for any contaminated sediment site." It warns that dredging leaves residual contamination behind and can release pollutants by stirring up sediment, and asks project managers to predict residuals from pilot studies.
Credit
US EPA, with NOAA, the US Army Corps of Engineers (USACE) and the US Fish and Wildlife Service
2006
Charcoal mixed into riverbeds
Activated carbon is a very porous charcoal that binds PCBs so that worms and fish take up less. In 2006 Alcoa, under EPA oversight, mixed or spread activated carbon into PCB-contaminated sediment in plots of the lower Grasse River at Massena, New York. By 2009, PCBs in worms exposed in the field were 85% to 97% lower where carbon had been mixed in. Cores taken in 2016 by USACE's research center (ERDC) and the University of Maryland, Baltimore County, found the carbon still in place under about 2 cm of new sediment a year, cutting PCBs in the water between sediment grains by 96% to 98%. A 2015 review counted pilot or full-scale carbon projects at more than 25 sites in the United States, Norway and the Netherlands.
LessonA fine carbon slurry placed under water tends to wash away, and the Grasse River plots needed steel enclosures, which led to pellets that sink whole and break apart in the sediment.
Credit
Alcoa and US EPA (pilot); Upal Ghosh, Todd Bridges and colleagues (10-year study)
2009
Hudson's first season overshoots
General Electric (GE) dredged PCB-contaminated sediment from the Upper Hudson River near Fort Edward, New York, from May to October 2009, the first of two phases. The contamination was deeper than the design showed. About 35% of the design cores had not reached the bottom of the contaminated layer, so crews removed nearly twice the planned volume in the areas they worked and finished 10 of 18 planned areas, 284,000 cubic yards in all, by EPA's count. EPA measured about 440 kg of PCBs carried downstream past Thompson Island, against a limit of 117 kg for the phase. A panel of seven independent scientists reviewed the season, and EPA rewrote its standards for releases, leftover contamination and output before the second phase.
LessonShort cores led to repeated passes, and repeated passes stirred up more PCBs.
Credit
GE (dredging); US EPA (oversight); the independent peer review panel
2015
Hudson's fish lag the dredges
GE removed about 2.7 million cubic yards of PCB-contaminated sediment from 40 miles of the Upper Hudson between 2009 and 2015. EPA's 2002 plan expected fish to reach its first health target, 0.4 mg of PCBs per kg of fish, within five years of dredging, by 2020. In 2021 the river-wide average was still above it, though the share of fish samples below the target had risen from 21% to 37%. In its third five-year review, finalized in January 2025, EPA again put off deciding whether the cleanup protects health. It says it needs at least eight years of fish data after dredging and will decide in an addendum by the end of 2027.
LessonFish can take longer to recover than a cleanup plan predicts; EPA expects more than 50 years of natural recovery before people can eat river fish once a week.
Credit
GE; US EPA
Watch for
Clearing a dam pool first
In Cuyahoga Falls, Ohio, EPA is dredging PCBs, PAHs (tar-like chemicals from burning fuel), oil and metals such as cadmium and lead from the pool behind the Gorge Dam, before the dam is removed in a separate project. The sediment is pumped two miles downstream to a placement area built to stabilize and store it. EPA expects to remove more than 850,000 cubic yards; dredging resumed in early March 2026 after a winter pause.
Achieved whendredging reaches its planned volume, the dam is removed, and monitoring by Ohio EPA or another outside party finds no rise in PCB or PAH levels downstream.
Claimed by
US EPA, March 2026
Open
01
Fish recovery scorecard
A public table and chart for at least ten waters where contaminated sediment was dredged, such as the Upper Hudson, the Grasse River and Minamata Bay. For each, it gives the fish targets and dates the cleanup plan promised beside the fish levels later measured, with a link to the page each number came from.
The briefCredit · unclaimed02
Debris versus the bucket
A tank test, using only clean kaolin clay and sand, of how much sediment a closed "environmental" clamshell bucket lets out when wood scraps, rope or stones stop it from shutting, and of whether a camera or closure sensor on the bucket catches each failed closure.
The briefCredit · unclaimed
Open 01 · the brief
Fish recovery scorecard
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A public table and chart for at least ten waters where contaminated sediment was dredged, such as the Upper Hudson, the Grasse River and Minamata Bay. For each, it gives the fish targets and dates the cleanup plan promised beside the fish levels later measured, with a link to the page each number came from.
Why it is still open
The Upper Hudson plan expected its first fish target by 2020; the 2021 average was still above it, and EPA has put its verdict off to 2027. Each site's numbers sit in separate review reports. The National Research Council compared sites in its 2007 report Sediment Dredging at Superfund Megasites, and no maintained update was found.
Built with what exists
EPA five-year review reports and their fish appendices; state fish data such as New York's; Japan's Ministry of the Environment report for Minamata; a web agent to collect the reports and an AI model to pull numbers out of PDF tables, with every number checked by a person (see the Note "Deleting 80% to start over"); Python and a plotting library.
First step
Take the Upper Hudson's yearly fish averages from Appendix 3 of EPA's third five-year review and plot them against the 0.4 and 0.2 mg/kg targets.
Done looks like
A public CSV and chart covering at least 10 sites, each row giving the target, the promised date, the latest measured value and its year, and the source page; a second person checks 50 numbers drawn at random and at least 48 match their source.
Open 02 · the brief
Debris versus the bucket
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A tank test, using only clean kaolin clay and sand, of how much sediment a closed "environmental" clamshell bucket lets out when wood scraps, rope or stones stop it from shutting, and of whether a camera or closure sensor on the bucket catches each failed closure.
Why it is still open
After the 2009 Hudson season, EPA named bucket closure as one of six main causes of PCB release and reported buckets failing to close on debris, such as slab wood left by old sawmills. Patents and suppliers describe pressure, magnetic and electronic closure sensors, but a search found no published test that counts the sediment released per grab with and without debris, or how often the sensors miss.
Built with what exists
An acrylic tank about 1 m long; a 1:10 scale clamshell bucket made on a 3D printer and worked by servo motors; kaolin clay and fine sand; two optical turbidity sensors such as the DFRobot SEN0189, read by an Arduino or Raspberry Pi and calibrated against weighed, filtered water samples; a Raspberry Pi camera and an image model such as Ultralytics YOLO trained to flag a bucket that is not fully shut.
First step
Build the tank and sediment bed, calibrate the turbidity sensors against weighed kaolin suspensions, and film 20 clean grabs.
Done looks like
200 filmed grabs, 100 clean and 100 with a debris item, each with the mass of sediment released, published with the video; the camera model's hits and misses on the grabs where the bucket stayed open.