Title: Challenges in Big-Earth Observation Data Analytics for Global-Scale Mining Land Use Surveying Slides 📎

The mining industry is fundamental to our society and economy, yet we have limited knowledge of its intricate global environmental and social impacts and risks. A fundamental prerequisite for assessing these concerns is the availability of updated land use maps that cover mining infrastructures. Unfortunately, such maps are neither consistently disclosed by mining corporations nor covered in global-scale land use data products. Visual interpretation of satellite images has attempted to bridge this data gap. However, this approach is inefficient for real-time mapping due to its labour-intensive nature, which makes it unfeasible to track the rapid proliferation of mining sites globally. This presentation will highlight recent advancements and persistent obstacles in creating comprehensive mining land use maps. As a land use category, mining encompasses diverse surface elements like open pits, tailings dams, processing plants, and rock waste dumps. Such a varied composition induces pronounced intra-class variability, intensified by diverse spatial patterns, geological contexts, and landscape nuances. Additionally, the spectral signatures of mining elements are frequently similar to non-mining entities such as exposed soil, rock surfaces, and other industrial structures, complicating differentiation. Furthermore, a significant challenge lies in the limited and biased training data available for large-scale mining land-use mapping applications. Therefore, to efficiently map global mining expansion, the semantic classifications of mining land use must be refined. This should be complemented by bolstered training datasets and new algorithms to address imbalanced classification scenarios with a minority class that has diverse spectral responses. Addressing these issues is urgent as the lack of comprehensive geospatial data on mining prevents the understanding of the effects of mining and can ultimately have significant environmental and societal implications through misinformed policy and decision-making.