Industries / Mining and materials / Mining and quarrying
Tailings and water
Tailings are the ground-up waste left after processing; this work stores them safely and keeps mine water clean.
5
Milestones achieved, from 1966 to 2020.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1966
The Aberfan tip slide
On 21 October 1966 a tip of coal waste slid onto the village of Aberfan in Wales and killed 144 people, 116 of them children. The tribunal of inquiry concluded that "Blame for the disaster rests upon the National Coal Board", and on its recommendation full information on every Coal Board tip was gathered for the first time.
LessonA waste heap needs the survey and records of any other structure.
Credit
Aberfan Tribunal
1992
A cyanide mine abandoned
Summitville Consolidated Mining ran a heap-leach gold mine, which uses cyanide to draw gold out of heaped ore, at Summitville, Colorado, from July 1986 to October 1991, and abandoned the site in December 1992. The US Environmental Protection Agency took over on 16 December 1992, and the state and EPA have worked since then to slow acid mine drainage from the site.
LessonA mine's water problem can outlast the company, and then the public pays.
Credit
Colorado Department of Public Health and Environment and US EPA (clean-up)
2014
The Mount Polley breach
On 4 August 2014 the tailings dam at Mount Polley mine in British Columbia failed. An independent panel found the main cause in the design: a weak layer of glacial lake sediment under the dam, which site investigations had not found. It called for a move to best available technology, such as filtered, compacted tailings, and for independent tailings review boards.
LessonThe panel found that inspections would not have prevented the failure, so safety has to start with what lies under the dam.
Credit
Mount Polley Independent Expert Engineering Investigation and Review Panel
2020
A public tailings dam database
After a tailings dam at Brumadinho, Brazil, collapsed on 25 January 2019 and killed at least 259 people, investors asked mining companies for details of their dams. More than 100 companies answered, and in January 2020 the Global Tailings Portal put details of more than 1,800 dams in a free, searchable database.
Credit
GRID-Arendal, with the Investor Mining and Tailings Safety Initiative led by the Church of England Pensions Board and the Swedish National Pension Funds' Council on Ethics
2020
A global tailings standard
The International Council on Mining and Metals (ICMM), the UN Environment Programme and the Principles for Responsible Investment launched the Global Industry Standard on Tailings Management in August 2020. ICMM members had to disclose the status of each of their tailings facilities against it by August 2025.
Credit
ICMM, UNEP and PRI
Watch for
Two-thirds of dams conform
ICMM says 67% of the 836 tailings facilities run by its members fully conform to the global tailings standard, and 33% conform in part. ICMM says full alignment is taking longer than it expected.
Achieved whenevery facility fully conforms and an independent party confirms it.
Claimed by
ICMM, November 2025
Open
01
Radar checks of past failures
A test of whether free satellite radar data showed the ground moving before each tailings dam failure since 2015, using one method for every case, published with a list of which failures gave warning and how early.
The briefCredit · unclaimed02
One table of dam disclosures
One open, comparable table of the dam-by-dam disclosures companies published under the global tailings standard in August 2025, with location, height, consequence class and conformance status, each row linked to its source page.
The briefCredit · unclaimed
Open 01 · the brief
Radar checks of past failures
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A test of whether free satellite radar data showed the ground moving before each tailings dam failure since 2015, using one method for every case, published with a list of which failures gave warning and how early.
Why it is still open
The studies found for this brief each look at a single dam. One, by researchers at the University of Nottingham and the University of Sri Jayewardenepura, looked back at Brumadinho alone and found signs of risk at least five months before the collapse.
Built with what exists
Copernicus Sentinel-1 radar images, which are free; InSAR processing tools such as ASF HyP3, ESA SNAP or MintPy (InSAR measures ground movement by comparing radar images taken at different times); Python; public lists of tailings dam failures.
First step
List tailings dam failures since 2015 with dates and coordinates, and check which ones Sentinel-1 covered in the year before.
Done looks like
Every failure since 2015 that Sentinel-1 covered, analysed with the same code, in a table giving whether movement crossed a set threshold before the failure, how many days before, and how many dams that did not fail crossed the same threshold.
Open 02 · the brief
One table of dam disclosures
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
One open, comparable table of the dam-by-dam disclosures companies published under the global tailings standard in August 2025, with location, height, consequence class and conformance status, each row linked to its source page.
Why it is still open
ICMM members disclosed the status of their facilities company by company in August 2025. The one public cross-company database, the Global Tailings Portal, is a beta version launched in January 2020 from companies' answers to investors.
Built with what exists
Company disclosure pages; a large language model such as Claude or Gemini to pull fields from PDFs with page citations (the note "Seven more flagged" covers Claude reviewing records); a spreadsheet or SQLite database; a hand check.
First step
Gather the disclosure links for every ICMM member company into one list.
Done looks like
All 836 facilities that ICMM counts in one public table, with a 10% random sample checked by hand and the error rate published.