Using open-access geospatial data to automate tailings dam management

Major tailings dam failures in recent years, along with environmental, social and governance (ESG) compliance requirements and the rise in ESG investing have renewed interest in tailings-facility management for mining. Changing weather patterns pose a further risk of tailings dam failures and increase the complexity of monitoring these facilities. This makes new monitoring models, such as the one discussed here, all the more important.

Thousands of mine tailings dams (or tailings storage facilities) dot the globe, storing gigantic volumes of toxic mine waste. Tailings management expert Chad LePoudre believes mining companies can better be described as waste-management companies, since ore constitutes a miniscule portion of mined material (typically 5 – 16 g/ton for gold), with the vast bulk (e.g. 999 995 – 999 984 g/t) ending up as processed waste.[1]

Overview of tailings facilities globally, as displayed in GRID-Arendal’s Global Tailings Portal database in August 2021. The background thematic map shows earthquake distribution peak ground acceleration.

Scaling up and complementing traditional monitoring approaches

Traditionally, tailings dams are surveyed by mine surveyors with GPS and total stations at quarterly or annual intervals. Aerial photogrammetry and lidar surveys can also provide high-accuracy measurements for monitoring tailings dam safety, compliance and volumes. Other monitoring methods include internet-connected sensors, distributed fibre-optic sensing, lithology and geochemical measurements. Remote and unmanned vehicles such as drones and hydrographic survey craft have also become commonly used monitoring solutions.

The geographic span and remote nature of many tailings dams have made satellite-based remote sensing an effective way to monitor and measure active and historic tailings facilities for anything from conformance to design, to surface water accumulation, volume changes and reconciliations, stability and subsidence. This technology also provides ESG investors (high-net-worth investors demanding greater accountability and environmental responsibility from the firms they invest in) and others with an independent means of verification.

Satellites as an investor’s tool

Investors and NGOs are driving the adoption and use of satellite-based data for financial and risk management, as well as environmental-impact accountability. The international Investor Mining and Tailings Safety Initiative is an investor-led effort with support from the UN Environment Programme, with more than $13-trillion in funds under their management.[2] As part of the initiative, the Norwegian environmental organisation GRID-Arendal created the Global Tailings Portal, a database of tailings dams and their related information.[3]

There are similar initiatives that also apply satellite data for monitoring issues such as value chains for ethical sourcing, deforestation, and polluting emissions. The Spatial Finance Initiative, for example, promotes the integration of geospatial analysis into financial decision-making for information markets, financial products, and risk management. Organisations such as WattTime in turn use satellite-based data for emissions-monitoring at the facility/factory level.

Pinpointing failure

According to the environmental consultancy GRID-Arendal, tailings dams fail for many reasons, the most common causes being overtopping, slope instability, seismic instability, structural inadequacies, and seepage and internal erosion.[4]

Causes of tailing dams failures 1915-2016. (Credit: GRID-Arendal)

Many of these conditions can be monitored and measured using earth-observation satellite data.

Commercial companies and services such as Airbus OneAtlas’s Stack Insight and PinkMatter’s FarEarth Change Monitor Service already offer mining companies solutions to monitor and measure change and subsidence, and conduct volume calculations.[5], [6] These solutions typically rely on proprietary medium to high-resolution optical and radar (SAR) satellite imagery and digital elevation data. SAR data and InSAR processing are gaining popularity in monitoring applications, as they don’t suffer from the limitations that cloud cover poses to optical observations.

A new satellite-based remote sensing study (discussed below) goes a step further by predicting the waste flow path and affected areas in case a tailings dam fails. Making use of publicly available satellite imagery and digital elevation data, this study paves the way for governments, the public and investors to hold mining companies accountable with independent, verifiable information.

Creating a predictive flow-path model

Following the tragic tailings dam failure at the Brumadinho Córrego do Feijão iron-ore mine in Brazil, Dr Iqra Atif at the Wits Mining Institute developed a geographic information system, or GIS-based tailings spill-path model as a first step for mines and others to map and quantify the potential damage and affected area in case of a tailings dam failure.

Numerical models are successful for risk and design assessments of tailings dams, notably the Smoothed Particle Hydrodynamics (SPH) method for modelled tailings slurry runout. But these models tend to be computationally intensive and expensive to run.

Dr Atif’smodel, on the other hand, runs in Python and draws on open-source data, which make it affordable and efficient. She selected Sentinel-2A data (with thirteen spectral bands and a medium [10 m] spatial resolution) since it is publicly available, and used it in combination with the Global Digital Elevation Model (GDEM) version 2.0, an open-access topographical dataset. To prepare for the imagery analysis, the pre-processing of atmospheric, radiometric, and geometric corrections was automated. To “smooth” the digital elevation model (DEM) – a process to remove sinks/depressions that could distort the data – she employed a linear interpolation method for its simplicity and low computational requirements. The processed DEM was then used in the model to determine hydrological parameters such as the flow direction and flow accumulation of spilled tailings.

The model incorporates international standards and practices on the design, management and monitoring of tailings storage facilities, as well as South African guidelines and regulations. Dr Atif also notes that satellite data fulfils the ICMM’s (International Council on Mining and Metallurgy) requirement for data to be independently verifiable.

The model was tested on satellite imagery captured prior to the Brumadinho tailings dam disaster, and validated against post-failure imagery of the Samarco tailings dam. The predicted flow path matched the actual flow path of the tailings dam failures with 95% accuracy.

The modelled tailing spill path (red) overlaid on real tailing spill extent of the Brumadinho Córrego do Feijão tailings dam failure.

In the peer-reviewed paper that followed, Dr Atif and her co-authors state that the model is capable of making acceptable predictions of the potential tailings flow path in the case of tailings-storage-facility failures in mountainous regions, and is suitable for mapping a first pass of the affected area, which in turn can be used in the development of mitigation strategies and to guide search and rescue missions. The model is best suited for predicting longitudinal flow, with further research into the prediction of cross-sectional characteristics being recommended by the study’s authors.

Like all commercial solutions, the accuracy of the tailings spill-path model is highly dependent on the spatial resolution of the input DEM, and can be improved with higher-resolution satellite imagery. Complementary data and inputs such as lidar, SAR and in-situ monitoring could also supplement it and help overcome the limitations of cloud-cover.

Application in the South African

South Africa is considered to have the highest number of environmentally dangerous tailings storage facilities in the world according to a Reuters investigation in 2019.[7] Most of the country’s tailings dams built using the upstream construction method, considered the most environmentally risky. For this reason, the investor community also doesn’t favour it.

Active and inactive high-risk upstream tailings globally in 2019. (Credit: Reuters)

Monitoring and prediction methods and models like Dr Atif’s and others’, which draw on free, publicly available data, could prove especially useful when the Southern African Institute of Mining and Metallurgy (SAIMM) and South Africa’s Minerals Council set out to develop best-practice guidelines on the design, management and monitoring of tailings facilities.

Independent risk assessment and public accountability

The success of Dr Atif’s tailings spill-path model lies in providing an independent means to measure and assess risk, with the potential to scale and automate the near-real time detection and risk modelling of other dams. In doing so, it can promote transparency between mining companies, investors, and the general public and regulators. The model remains the intellectual property of Dr Atif and the Wits Mining Institute, but there is talk of open-sourcing it once the project is completed.

Using open data holds cost advantages not only for the mining companies and third parties, but also for regulators and investors. There’s also the promise of even greater-resolution data on the horizon as new sensors are developed and launched by national space agencies such as the European Union’s ESA and the US’s NASA. African countries too are charting paths in the space sector with new investment such as the South African Space Infrastructure Hub, which received R4.6-billion in funding in 2020, the uptake of Digital Earth Africa, and other projects that align with the Africa Union’s Agenda 2063.


[1] ESRS CIM, Tailings Failure Case Studies, Statistics and Failure Modes Webinar, (Nov. 02, 2018). Accessed: Mar. 29, 2021. [Online Video]. Available:

[2] ‘ICMM: Tailings waste’. (accessed Aug. 24, 2021).

[3] ‘World’s first public database of mine tailings dams aims to prevent deadly disasters | GRID-Arendal’. (accessed Aug. 24, 2021).

[4] ‘Causes of tailing dams failures 1915-2016 | GRID-Arendal’. (accessed Aug. 24, 2021).

[5] ‘Stack Insight – monitor mines & calculate volume extraction’. (accessed Jul. 27, 2021).

[6] ‘Farearth | Pinkmatter Solutions’, (accessed Jul. 27, 2021).

[7] ‘The looming risk of tailings dams’, Reuters. (accessed Aug. 24, 2021).

Related Articles