Near Real-Time Coming to Cerulean
New feature surfaces recent oil slicks to Cerulean users with exceptional timeliness
What’s new: Faster pollution detection
We are updating Cerulean, our tool for tracking global marine oil pollution, to enhance support for users tracking very recent pollution events. To date, our most reliable detections were constrained to satellite imagery older than three days, after which time we could match potential pollution to potential vessel sources. The new release evaluates oil slick detections in near real-time as public satellite imagery becomes available to support faster identification of potential incidents within a few hours of image capture. The ability to filter recent detections for higher-confidence events will allow swifter action on oil slicks by groups such as enforcement agencies and pollution watchdogs.
Why this is an improvement: New leverage in existing data
Cerulean uses machine learning models to scan and process all Sentinel-1 radar satellite images collected over the ocean. It is trained specifically to detect oil slicks, which appear as dark signatures on the images.
To date, the model has used ship location data from vessel automatic identification systems (AIS) and offshore oil platform locations to find the possible sources that most closely match the location, timing, and shape of each slick. This results in a delay for Cerulean users because that data is available from our AIS provider about 72 hours after transmission.
Our new data processing step targets vessel slick detections within the 72-hour window–when we have identified potential pollution events, but don’t have access to AIS data yet. We measure how closely the geometry of a potential slick matches the shape of previously confirmed vessel-related slicks, bridging the critical gap when vessel AIS data is unavailable. The geometry is analyzed as soon as satellite imagery becomes available, resulting in 15% of all detections being available in under two hours, 50% within four and a half hours, and 85% within twelve hours. We are calling the resulting score from the detections “slick confidence.”
An oil slick that was identified by Cerulean 1 hour and 27 minutes after a vessel was at this location. It received a slick confidence score of 0.89.
Approximately 72 hours later, a vessel’s track can be seen in the area of the slick, and that vessel’s details are listed under the “Potential Sources” header in the panel.
The algorithm also helps weed out false detections caused by natural seeps, ice, algal blooms, sargassum, and other artifacts in the imagery caused by weather events.
Where to find slick confidence today
Currently, our near real-time analysis is available in the Cerulean platform by focusing the date range filter to the last three days or less and making sure the “No potential sources” checkbox is selected under the Source Filters heading.
It is also available through the API. If you are a developer interested in using this new data, be sure to check out our public guide on how to use our API, including a near real-time update example! We are actively seeking new opportunities to integrate this new methodology with our slick-source-match, so stay tuned for more exciting developments on that front.
We’re always improving our tools to serve our mission of promoting conservation for people and the planet. Visit Cerulean to learn more about our work in ocean oil detection and check out our API guide for more examples and full documentation.
Stay connected to hear about our latest releases by signing up for our updates and newsletter. Questions or feedback? Connect with us at support@skytruth.org.
