Enhanced Subsea Environment Analysis for Oil and Gas Decommissioning Through Clustering

The decommissioning of offshore infrastructure in environmentally sensitive regions, such as the Campos Basin in Brazil, requires detailed seabed characterization to avoid damaging deep-sea coral habitats. Traditional methods, relying on manual interpretation of sonar and ROV data, are labor-intensive, costly, and limited in scalability. This study addresses this gap by introducing a data-driven framework that applies the DBSCAN clustering algorithm to georeferenced seabed targets obtained from side scan sonar and multibeam surveys. The methodology converts polygonal reflective features into structured point clouds and employs K-Nearest Neighbors to estimate clustering parameters, followed by sensitivity analyses to explore spatial and computational behavior. Two case studies were conducted: a local-scale application around platform P-33 (∼3,000 targets) and a field-scale analysis across the Marlim and Voador fields (∼29,000 targets). Results showed that DBSCAN effectively identified ecologically coherent clusters, with optimal configurations achieving close alignment with expert-defined zones. The method proved flexible, allowing adjustments to clustering granularity for different operational needs. This unsupervised, scalable approach advances seabed zoning methodologies by reducing subjectivity and improving reproducibility. It offers actionable insights for decommissioning planning, including defining exclusion zones, prioritizing ROV inspections, and supporting risk-based decision frameworks. The framework lays the foundation for future enhancements, including integration of biological descriptors, adaptive parameter tuning, and temporal monitoring. This contribution represents a significant step toward intelligent, sustainable seabed classification in support of offshore environmental stewardship and operational efficiency.