Residues from the processing of ores and minerals, called tailings, are usually collected in large tailings dams.
The Brumadinho dam failure, which killed at least 259 people and irreversibly destroyed nearby ecosystems, highlights the high risks and potential environmental damage of such tailings dams. The disaster led to several initiatives to more closely monitor tailings storage facilities (TSFs), create greater transparency, and better assess risks, such as the Global Tailings Portal. Investors are also required to report regularly on the safety of tailings dams.
Monitoring of tailings dams can be automated through the use of remote sensing methods, such as using multispectral satellite data to calculate the soil moisture index (SMI).
The stability of the dams can be determined based on various factors, such as detection of displacements, defects, change in water and mineral content, etc. All of these factors can be monitored by the automated analysis of multispectral satellite data.
A machine learning (ML) model is able to recognize and classify objects in image and radar data by training with labeled data. As a result, tailings dams can be reliably identified and their change detected over a period of time. However, this requires a satellite data source that regularly provides high-quality imagery and radar data. Open source satellite data, such as ESA's Sentinel data, is usually low resolution, while commercial data can be very costly.
Object detection and classification in image data can be achieved with Computer Vision techniques, for which convolutional neural networks (CNN), such as the U-Net architecture, are typically used. Such a network is able to segment an image, meaning that each pixel of an image is classified, allowing the classification of different objects. As a result, tailings dams and their changes can be detected.
Radar data can be used to determine the moisture content of tailings, as well as minerals in surrounding waters and soils that could indicate a leak in the dam. The soil moisture index (SMI), for example, can be used to monitor changes in the water content of tailings dams. The normalized difference vegetation index (NDVI) can map changes in surrounding vegetation. Significant changes in these indices can be indicators that the dam is leaking or about to breach.
Receive news about Machine Learning and news around dida.
Successfully signed up.
Valid email address required.
Email already signed up.
Something went wrong. Please try again.