STRESZCZENIE

Aircraft condensation trails (contrails) are one of the largest non-CO2 contributors to aviation’s climate impact. Because only a small fraction of flights produces persistent contrails, identifying which specific aircraft is responsible for each observed trail is a necessary first step towards targeted mitigation, and is the subject of active operational research at EUROCONTROL. This thesis develops a vision-based attribution pipeline for ground-based images captured at the Maastricht Upper Area Control Centre. A YOLOv11 instance-segmentation model, trained on 6,253 annotated images across four morphological classes, detects contrails with a mask mAP@50 of 0.720 on the validation set. Detected masks are then linked to source aircraft through a four-stage pipeline. The wind-advection step is the main methodological contribution: it extends prior nearest-aircraft attribution methods to wind-advected trails. The pipeline is evaluated on two case studies validated against the visual record. The work is presented as a research feasibility study and identifies the validation and tuning steps required for deployment at population scale.