A Digital Twin enabled satellite workflow for automated oil spill detection and forecasting

The concept of the Digital Twin of the Ocean (DTO) has transitioned from a research vision to an operational paradigm in the ILIAD project. Several of the mature Digital Twin components are available as reusable, findable (through the Iliad Registry: https://iliad-registry.inesctec.pt) and interoperable application packages, enabling automated environmental monitoring and decision support. This contribution presents the Cretan Sea oil spill DTO, focusing on near-real-time oil spill detection and forecast.

The presented system implements an end-to-end workflow based on Sentinel-1 SAR imagery, orchestrated through Common Workflow Language (CWL). Incoming satellite data are automatically ingested, processed, and analysed using containerized application packages, enabling scalable and reproducible execution across cloud and HPC infrastructures. Oil spill detection is performed using a deep learning approach based on a combination of FCOS and U-Net convolutional neural networks, trained to discriminate oil slicks from look-alike phenomena in SAR imagery. The results are systematically compared against a statistical detection methodology implemented via the SNAPpy library, providing robustness and methodological benchmarking.

Detected oil spill events trigger downstream Digital Twin services, including high-resolution marine forecasting and oil spill transport modelling. The forecasting framework integrates dynamically downscaled atmospheric forcing from WRF, hydrodynamic fields from NEMO, and sea state information from WAVEWATCH III, providing coastal-scale predictions at kilometer resolution. Oil spill transport and fate are simulated using the already established and validated MEDSLIK-II software [1], with results visualized through operational web platforms to support rapid situational awareness. Additionally, a 4D immersive visualization tool is introduced to present the oil spill evolution and fate in an intuitive spatio-temporal environment, enhancing operational readiness and enabling first responders and non-expert stakeholders to rapidly interpret complex model outputs without reliance on conventional map-based products.

By packaging satellite analytics, numerical modelling, and orchestration logic into reusable application packages, the system demonstrates how post-project DTO assets can be operationalized beyond the ILIAD lifecycle. The Cretan Sea DTO illustrates a transferable Digital Twin workflow for automated oil spill detection and response, supporting environmental monitoring authorities with timely, data-driven decision support.

References
[1] M. De Dominicis, N. Pinardi, G. Zodiatis, and R. Archetti, “MEDSLIK-II, a Lagrangian marine surface oil spill model for short-term forecasting – Part 2: Numerical simulations and validations,” Geosci. Model Dev., vol. 6, pp. 1871–1888, 2013. doi: 10.5194/gmd-6-1871-2013