Satellite-enabled early warning system for geotechnical structures

Thesis

There is a critical need to protect communities, natural habitats and the ecosystems from toxic mine waste pollution that can result from the collapse of tailings storage facilities (TSFs). With increasing role of mining in the transition away from fossil fuels, the need to do mining safely and sustainably is of critical importance. Satellite Synthetic Aperture Radar Interferometry (InSAR) shows much promise towards this monitoring ambition on the global scale. In this work, the opportunities and challenges for developing a satellite based early warning system for TSFs are explored. This necessarily brings together three separate fields - geotechnical engineering, satellite remote sensing and deep learning.

Firstly, ground deformation measurements from InSAR and ground-based prism monitoring are compared to Finite Element (FE) simulation results for a recent tailings dam collapse. The study highlights the efficacy and complementarity of geotechnical and remote sensing techniques for the monitoring of TSFs. Moreover, extracting meaningful information and interpreting the deformation patterns from InSAR data can be a challenging task. One approach to address this challenge is through the use of data science techniques. The representation of InSAR metadata as Embedded Entities within a Deep Learning framework (EE-DL) is proposed for modelling the spatiotemporal deformation response. This study demonstrated that EE-DL can detect and predict the fine spatial movement patterns that eventually resulted in the TSF collapse. Overall, EE-DL is proposed as a promising approach for building early-warning systems for critical infrastructures that use InSAR to predict ground deformations. This research is designed to push the limits of what is possible from the globally available open source data from the Sentinel-1 satellite. Moreover, the value and improvement in the early warning framework offered by commercial, high resolution data from the satellite Radarsat-2 are also investigated.

Authors: Bayaraa, M

• DOI: 10.5287/ora-7jqxemrzy
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: early warning sytems; Geotechnical engineering; InSAR; Remote sensing for infrastructure; Radarsat-2 InSAR; gaussian process regression; TSF; deep learning; Sentinel-1 InSAR; AI; making mining safer and sustainable
• Subjects: Remote sensing; Geotechnical engineering