An integrated Earth observation-based framework for preliminary ecological screening of marine oil spills

Satellite remote sensing has significantly advanced the detection and monitoring of marine oil spills; however, translating observations of surface contamination into ecologically meaningful assessments remains a fundamental challenge. Here, we present an integrated framework that combines Sentinel-1 synthetic aperture radar (SAR) imagery, a machine-learning-based oil-spill detection model, and indicators of habitat vulnerability, ecosystem functioning (chlorophyll-a), and recovery potential to provide a preliminary assessment of ecological exposure and vulnerability associated with marine oil contamination. The framework was applied to the February 2026 Abkatún oil spill in the western Gulf of Mexico, with NOAA WebGNOME trajectory simulations used to characterize oil transport pathways and provide process-based context for interpreting contamination persistence and exposure patterns. The SAR-based Random Forest classifier showed high agreement with manually interpreted SAR reference samples across the two study areas, with overall accuracies of 95.27-99.55% and oil-class F1-scores of 86.65-99.33%. Integration of satellite observations and trajectory simulations indicated that areas of higher estimated ecological pressure generally coincided with greater contamination persistence and more vulnerable ecosystems, rather than being determined by surface oil extent alone, although quantitative exposure estimates remained sensitive to the selected distance-decay formulation. The framework identified the highest estimated ecological pressure where persistent contamination coincided with vulnerable habitats, particularly coral reefs and mangrove forests, suggesting that spills with comparable surface extents may be associated with different levels of ecological pressure depending on exposure duration, transport processes, and ecosystem characteristics. Chlorophyll-a concentrations were lower during the 2026 study period relative to the pre-spill period and the 2024-2025 seasonal baseline, although these differences do not establish a causal relationship between the observed productivity anomaly and the oil spill. By integrating Earth observations with ecological indicators and transport-based interpretation, the proposed framework extends satellite-based spill assessment beyond contamination mapping toward ecosystem-oriented characterization of exposure and sensitivity following marine oil spills.