SkyTruth’s Cerulean: Public Database of Near-Real-Time Oil Detections

We introduce Cerulean, a fully automated machine learning pipeline designed to detect anthropogenic oil pollution across the world’s oceans in near-real-time from Sentinel-1 data. Cerulean is anticipated to provide a vital resource to researchers, environmental organizations, governments, journalists, and other stakeholders, by providing a free and public global monitoring and reporting system for oil pollution.

Cerulean was recently launched and is now accessible to the public via a state-of-the-art map-based interface (cerulean.skytruth.org) as well as an Open Geospatial Consortium (OGC) compliant API (api.cerulean.skytruth.org). These two interfaces allow unfettered access to our continuously growing database of potential oil slicks, able to be filtered against many different criteria including event timestamp, geographic area of interest, and slick area. Additionally, we cross-correlate each high-confidence slick with our vessel-tracking AIS database and our oceanic stationary infrastructure database to highlight the top three probable sources of each potential slick.

Cerulean’s code is open-source and has been published on GitHub to encourage public scrutiny for review and feedback. Though the model itself is continually being improved upon, and individual slicks should be subjected to expert review, we hope that this tool can immediately become a regular source of high-quality slick candidates for users seeking to test or corroborate their own models and systems, or to review their own sovereign waters. We are also looking for potential partners interested in collaboration to demonstrate novel value pipelines that leverage this new datasource.