School of Civil & Environmental Engineering (ETDs)

Permanent URI for this community

https://hdl.handle.net/10539/37962

Browse

Browsing School of Civil & Environmental Engineering (ETDs) by Author "O’Donovan, Christopher Galen"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Use of Multispectral Satellite Imagery to Monitor the Decant Pond of Tailings Dams
    (University of the Witwatersrand, Johannesburg, 2023-08) O’Donovan, Christopher Galen; Adam, Elhadi; Torres-Cruz, Alberto
    Tailings dam failures, such as the Jagersfontein failure in the Free State province and the Fundão and Feijão failures in Brazil, have brought into question the ability of the mining industry to operate safely, threatening its social license. To improve the safety of tailings dams, leading indicators of dam safety should be monitored. The location and historical behaviour of the tailings decant pond provides insight into several such leading indicators and can be used as a proxy to flag potential construction issues. This work investigates the use of public multispectral data collected by the Sentinel-2 satellite mission to monitor the supernatant tailings dam decant pond. This is achieved by leveraging the cloud-based Google Earth Engine platform and open-source GIS tools. Sentinel-2 acquires visible and near infrared spectrum data with a spatial resolution of 10 m and a revisit time of 5 days. Pond data is obtained by visual assessment and automated thresholding of Sentinel-2 imagery. Thresholds of near-infrared (NIR) reflectance and the normalised difference water index (NDWI) obtained by a least square error analysis are investigated. Implementation of the method at three South African tailings dams, constituting four decant ponds, illustrates the capabilities and limitations of Sentinel-2 imagery. High spatial resolution (<5 m) multispectral satellite imagery and natural colour aerial orthophotos (<0.25 m) serve as reference data. Visually assessed Sentinel-2 pond data presented a bias towards slight over estimation of the pond area compared to reference data. Other leading indicators did not show systematic bias across all sites. In general, the deviation between Sentinel-2 and the reference measurements was high, indicating that Sentinel-2 imagery should be used with caution for measurements critical to dam safety. Site-specific thresholds of NIR and NDWI indicated that automated thresholding of the NDWI is superior to NIR reflectance alone. It is shown that Sentinel-2 timeseries imagery can be used in tailings dam monitoring to supplement existing construction surveillance frameworks and provide historical pond data in the absence of such information.