Answers are grounded in this paper and cite their sources.
Globally, the long-term stability of tailings storage facilities is a critical environmental and safety concern, particularly in mining countries such as Chile, which hosts over 836 deposits distributed across highly variable geotechnical and climatic conditions. Traditional assessment frameworks rely heavily on field inspections and site-specific data, limiting their scalability and frequency of evaluation by the national geology and mining service. This study proposes a simplified mechanism for evaluating physical stability based on open-access data processed using artificial intelligence and geographic information systems to obtain an overall condition assessment. This mechanism quantifies two key indicators—the potential impact on the environment and the potential for physical instability—which are integrated into a hierarchical classification to prioritize regulatory supervision. The proposed methodology combines satellite imagery, geographic information layers, and deep learning models for the automatic detection, segmentation, and classification of tailings storage facilities. The systematic injection of rotational symmetries significantly increased the robustness of the segmentation model, evidenced by improvements across multiple evaluation metrics for different facility sizes. In addition, a classification model successfully categorized the distinct topographic configurations of these deposits. The methodology was validated in two major Chilean regions, achieving strong agreement with expert technical evaluations and providing a structured prioritization scheme that enables authorities to improve continuous monitoring and regulatory supervision. Finally, a large language model was implemented to assist in interpreting the simplified evaluation process for nontechnical users, structuring the analytical information into an accessible technical report format.
No takes yet. Share an insight, caveat, or question.
Cocca-Guardia et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: