Use Case: Oil Spill Detection

WITOIL: Predicting Oil Spills for Cleaner Seas

Outsmart oil spills: WITOIL - AI-powered forecasting for faster response

Overview

WITOIL (Where Is the Oil) is a multi-model Decision Support System (DSS) on-demand service that forecasts transport and weathering of actual or hypothetical oil spills in the global, regional European Seas, and in the selected coastal areas. WITOIL uses the MEDSLIK-II oil spill model forced by operational meteo-oceanographic services. The service has been further refined by using labelled image datasets from Sentinel 1, Sentinel 2 and Landsat 8 satellites to improve existing deep-learning algorithms and the AI-supported detection of oil spills. WITOIL is now an operational service in the iMagine-AI platform and interfaces with existing operational marine pollution oil spill monitoring and modelling services running at CMCC. Relevant guidelines and documentation are also available to promote the service uptake.

Users will be able to access past oil spill events, observe their results in the platform, and perform simulations on their own, provided they can upload georeferenced oil spill area files. The system will be accessible through a simple user interface, currently aiming for it to be Streamlit, so a less experienced user can perform simulations on their own without the need for extensive computing skills.

Main Features

  • The inference cluster to execute the requests asked by end users usually does not require intensive computation and, in rare cases, will need the Bayesian optimization workflow, given the difficult nature of providing oil spill imagery into the system.
  • No authentication required
  • Users’ submissions should be available in the system for visualisation. The platform will provide access to the output (forecast) of the Medslik model related to simulations performed, considering as input the set of oil spills provided by OrbitalEOS and published on Zenodo.

Explore the AI Models Underlying the Services