Automating Mining Site Classification with Machine Learning: Near-surface raw materials in Bavaria

Economic development requires substantial raw mineral resources. The construction industry, in particular, is dependent on the supply of domestic raw materials such as sand, gravel and clay. Monitoring the evolution of extraction sites is crucial for assessing current resource usage and future availability. The Landesamt für Umwelt - Bayern currently employs manual analysis of remote sensing data (e.g., aerial imagery, digital elevation models) and GIS software to classify open-pit mining regions as active or inactive – a resource-intensive process. This work presents a machine learning approach to partially automate this classification. We developed and trained a Random Forest model using existing manually-classified data. Initial validation demonstrates the significant potential of this methodology, achieving an accuracy exceeding 90% in preliminary tests. This contribution will demonstrate the first promising results of our automated classification system and discuss the challenges and future directions for model refinement and enhanced classification accuracy. This approach offers a pathway to reduce the resources required for monitoring mineral extraction activities.