Marine oil spills detection and monitoring in the Great Barrier Reef
Project outline
The European Union’s Copernicus Programme has developed a family of six satellite missions called “Sentinels” that provides global, near-real-time satellite observations of a wide range of environmental parameters. One of the key benefits of the Copernicus Programme is that the continuous satellite imagery acquisition, assured until at least 2030, is made available free-of-charge to all users.
Each Sentinel mission carries a range of technologies, such as SAR for the Sentinel-1 mission (2014-) and multispectral imaging and thermal instruments for land, ocean and atmospheric monitoring for the Sentinel-2 and Sentinel-3 missions (2016-). The near-coincident, high spatial and temporal resolutions provided by the Sentinel-1, -2 and -3 satellites allows for synergistic use of SAR, optical and thermal information for environmental monitoring. Multi-satellite information of this nature was not available to the research community until 2014, when the Sentinel missions started. This synergistic use is further explored in this research but only Sentinel-1 SAR for oil spills detection is presented in this article.
The need of remote sensing-based oil spills monitoring systems in Australia was dramatically illustrated in August 2009 by the Montara oil spill incident in the Timor Sea. An estimated 30,000 tonnes of heavy fuel oil (HFO) leaked from the Montara wellhead platform, 250 km off the Western Australian coast, during 75 days between the 21st August and the 03rd November 2009. The monitoring of this incident did not beneficiate from remote sensing support at the time.
Historically, oil spills are best detected using satellite imagery from synthetic aperture radar sensors (SAR). With HFO illegal discharges often occurring at night and ship accidents frequently resulting from stormy weather, SAR sensors are well suited for detecting such incidents because they are active satellite instruments that operate day and/or night and in any weather, with wide swaths (> 100 Km) that can cover large areas of the ocean and high spatial resolution (10-20 m) that is sufficient to detect small size oil slicks.
One of the main limitations of using SAR imagery however, is the difficulty to distinguish false-positives, such as wind or bathymetric features, or biogenic films from true-positive HFO pollution events. All appear as low backscattering area, called a “dark spot”, which contrast with the surrounding waters that often appear brighter.
This can be illustrated by these two Sentinel-1 SAR scenes: the first (left) was acquired on 20 January 2018 over the China Sea following the Sanchi oil spill incident of January 6, 2018. The oil slicks are circled in red in the image. The second (right) was acquired over the Central Great Barrier Reef on the 3rd September 2019 and shows biogenic slicks north of Townsville.
Sentinel-1 SAR scenes showcasing oil spills and look-alikes.
The Great Barrier Reef
The Great Barrier Reef marine park is recognized as a high-risk area for oil spills by the Australian Maritime Safety Authority (AMSA). Over the past 10 years (i.e., 2009-2019), three known major oil spill incidents have occurred across the Great Barrier Reef: two were ship accidents, first in 2009 with the Merchant Vessel “Pacific Adventurer” off Cape Moreton, then in 2010 with the Merchant Vessel “Shen Neng 1” off Rockhampton. Even more insidious are illegal discharges from ships and drilling platforms because not officially reported unless detected, and the Great Barrier Reef is no exempt to such incidents: in July 2015, a series of illegal discharges contaminated the shores south of Townsville (Cape Upstart). Illegal oil discharges are not currently monitored routinely in the GBR and may be more frequent than anticipated.
The use of SAR for monitoring the Great Barrier Reef is innovative as mostly optical satellite sensors have been used so far, with the latest and most successful example being the eReefs water quality dashboard. The Sentinel-1 SAR satellite is orbiting the earth every hundred minutes at 700 km in altitude and moving at a velocity of 7 km/sec, providing imagery over the Great Barrier Reef every 12 days. For this research, Sentinel-1 SAR imagery was automatically processed and visualised daily to search for any oil spill signature. As of June 2020, a total of 3,500 Sentinel-1 SAR scenes, imaging parts of the 2,600 km stretch of the Great Barrier Reef marine park, have been acquired at around 05.30am local time (AEST) by the satellite since 2014.
Currently, our system provides a semi-automated, ongoing background level monitoring in the Great Barrier Reef, but we are actively pursuing this research to improve the detection system and make it fully operational by exploring machine learning and computer vision approaches for a fully automated detection of oil spills. Other applications are also explored, such as the monitoring of river floods as illustrated by this Sentinel-1 SAR image of 30th March 2017 following Cyclone Debbie. The river flood discharges freshwater into the GBR lagoon following the landfall of severe Tropical Cyclone Debbie, likely creating a different water mass characterised by lower signal backscatter (darker colour).
Burdekin river runoff during Cyclone Debbie, March 2017, captured by Sentinel-1 SAR.
Project team
