First step towards the development of an assessment tool for tailings dams

Published: 30 October 2019

Geotechnical experts and data scientists from Brazil and the Netherlands met for a week in early October at a hackathon. The aim was to design a system for monitoring the safety of tailings dams. The meeting was in the Deltares iD-Lab and it resulted in the first step towards an assessment tool for tailings dams.

Seven incidents worldwide this year

Tailings dams are earth structures (sometimes up to 100 metres high) that are used to retain waste from mines (tailings) for decades. Just a few weeks ago, a dam in the state of Mato Grosso in Brazil failed and 580,000 cubic metres of sludge flowed a distance of up to two kilometres. Disasters like this have been more frequent in recent years in Brazil. So far this year, seven major incidents have been reported worldwide.

Combination of different datasets

The Dutch and Brazilian team collected data from three tailings dams in Brazil. The result was a collection of large and very different datasets combining local and satellite data. Examples of local data are in situ monitoring (pore pressure meters, inclinometers), the dam structure, and the geotechnical and rheological properties of the material. The satellite data included optical Landsat data and Sentinel-1 InSAR data.

Deltares Aquamonitor

Optical satellite images serve as a basis for detecting changes in the humidity of the earth’s surface. This is an important parameter for detecting piping and the internal erosion of dams, two failure mechanisms that can compromise the safety of tailings dams. The Deltares Aqua Monitor can be used to detect changes in land and water coverage.


During the hackathon, a ‘mock-up’ of an Assessment Tool for Tailings Mines was made on the basis of the national mining agency’s database and other sources of information. Using optical satellite images, it was possible to look back and detect the geotechnical failure of a slope three months ago. The idea is to develop the mock-up further into a warning system by combining the model with other observations and computer forecasts.

Monitoring dams and early warning

The Brazilian and Dutch experts developed a statistical algorithm to detect changes in the Brumadinho tailings dam. This 80-metre-high dam failed earlier this year with catastrophic consequences. The animation presented here detected changes in the pattern of dam deformation three days before the actual failure of the dam on 25 January. An early warning system for dam failures will come within reach if this proof of concept is developed further and combined with other observations and numerical forecasts.

Detecting changes in the dam’s deformation pattern using Sentinel-1 InSAR data that made it possible to issue an early warning of the dam failure three days in advance.

Impact of dam failures

If a tailings dam collapses, a flood of sludge is released that flows downstream at a given speed. The animation presented here is a computer simulation of a mudslide from the Delft3D-Slurry model. Combining the results of the model with information about land use (including factors such as the population, the environment and economic activity) makes it possible to determine the potential impact.

Computer simulation of a mudslide after a dam failure

Ton Peters, Deltares dam expert & the leader of the hackathon: “The results of the hackathon (datasets, simulations and the mock-up of the assessment tool) will generate a lot of opportunities to monitor the safety of tailings dams in Brazil better in the short term. ”

Participants at the hackathon

The following organisations were present at the hackathon:

  • Geological & mining experts from Brazilian organisations: the National Mining Agency (ANM), the Minas Gerais environmental agency (SEMAD) and the University of Ouro Preto (UFOP).
  • Remote sensing specialists, data scientists and geotechnical experts from the Netherlands: Deltares, University of Twente, Delft University of Technology and SkyGeo.