A Review of Oil Spill Detection Techniques on Water
Sana Kazilbash posted on July 15, 2020 |
Technologies include SLED-W, visual and infrared cameras, laser spectroscopy and radar.

Satellite image of the Deepwater Horizon spill in the Gulf of Mexico. The white rectangle illustrates the oil spill, which can be difficult to distinguish against the various clouds in the photo. (Image courtesy of National Center for Biotechnology Information.)
Satellite image of the Deepwater Horizon spill in the Gulf of Mexico. The white rectangle illustrates the oil spill, which can be difficult to distinguish against the various clouds in the photo. (Image courtesy of NASA.)

We have seen some highly publicized oil spills lately—from the massive Russian oil disaster at Norilsk, to the Trans Mountain pipeline leak that occurred in June. Many of these petroleum accidents typically cost billions of dollars to clean up, and contribute significantly to the contamination of lakes, rivers and seas. With over 80,000 miles of oil pipelines running through the United States, this begs the question: how are spills even being monitored and detected?

Smart Leak Detection on Water (SLED-W)

Let’s start by exploring recent innovations around the Smart Leak Detection on Water (SLED-W) system, which was unveiled just last month by the Southwest Research Institute (SwRI).

SLED-W is a computer-based machine learning solution that uses algorithms to combine visual and thermal data for the recognition of unique oil and water properties. This can be achieved cost-effectively by affixing inexpensive thermal and visible cameras to aircraft, stationary devices or watercraft.

“Labeling oil is a significant challenge,” says Ryan McBee, the SLED-W project leader at SwRI. “For SLED-W, we had to account for different behaviors so it would know what to consider and what to ignore to avoid false positives.”

Oil reacts with water in different ways; it can spread over water as a separate layer or it can mix together with water. This makes it challenging for sensors to detect oil leaks under a variety of weather conditions. Visible cameras have their limitations, and cannot always distinguish transparent oils that blend with water. Thermal sensors often show false positives near animals. By using a multidisciplinary approach involving machine learning algorithms, SLED-W analyzes different perspectives and evaluates the most relevant information from combined thermal and visual data—mitigating the weaknesses of each sensor on its own.

Training SLED-W’s machine learning algorithms to detect oil under various conditions. (Image courtesy of Phys.org.)
Training SLED-W’s machine learning algorithms to detect oil under various conditions. (Image courtesy of Phys.org.)

SLED-W is still an up-and-coming technology though, and it will be some time before it is fully established. So far, the computer vision solution has showed positive initial results when utilized to detect two different types of oil. Upon the gathering of further data from field testing, the algorithm will be updated to handle more varied real-world conditions.

Visual Imaging

According to a scientific paper authored by M. Fingas and C.E. Brown in the Sensors journal, visual imaging is broadly used in oil spill monitoring due to its relatively low cost. A thin sheen of oil on water is known to reflect light over a wide spectrum (often appearing silver), while bulk oil is easily observed to be brown or black.

However, oil doesn’t have specific spectral differences from water, and certain processes are often required for discriminating between the two fluids. One technique employs a push-broom scanner with a CCD detector in conjunction with an optical system. The signal from the CCD scanner is processed and enhanced in order to produce the desired results.

Another method involves the use of polarized lenses on video cameras to improve contrast (as oil on water has a polarizing effect on light). Contrast can also be maximized by setting detectors at an angle of 53 degrees—the Brewster angle—in order to obtain optimal oil-on-water data from reflected light.

Visible oil detection techniques come with their limitations. Some barriers to accurate visual imagery include darkness, cloud cover, adverse weather conditions and sun glitter—which can often be misinterpreted as oil sheens.

Sun glint is frequently responsible for confounding an entire scene. (Image courtesy of National Center for Biotechnology Information.)
Sun glint is frequently responsible for confounding an entire scene. (Image courtesy of Environment Canada.)

Historically, there have been issues in the application of satellite imaging. In the late ’80s and early ’90s, there were only a handful of satellites that would pass over an area as infrequently as once every 26 days. What were the chances that a satellite would glide over an oil spill on a cloudless day? The massive Exxon Valdez spill in 1989, for example, was detected over a month after it occurred, on one clear day that corresponded with a satellite overpass. Even after the exact slick coordinates were determined, it took two months for experts to produce a decent image of the oil spill due to the unavailability of algorithms for visualizing the oil.

Thankfully, satellite imaging today has come a long way since then, and there are now numerous satellites providing coverage of the Earth, such as QuickBird, WorldView-1 and -2. Multispectral satellites such as MODIS and MERIS are capturing data for a variety of wavelengths, and correcting images using a classification system for enhancing and characterizing oil on water. Advanced image processing systems are now being used to automatically rectify images for abnormalities and correspondence to graphical information system (GIS) requirements. This has improved the delivery time of images, making them more useful to oil spill responders.

Advances are also being made in the removal of sun glitter through signal processing techniques. The advent of light-enhanced video cameras has led to adequate functionality in darkness. However, even these modifications do not serve to completely eliminate the issues surrounding visual oil detection, and further research is underway.

Thermal Imaging

Infrared (IR) sensing can also be used to detect oil spills in a cost-effective manner.

The process involves the absorption of light in the visible region, a portion of which is then re-radiated to the infrared spectrum (usually in the 8–14μm wavelengths). Since oil demonstrates greater infrared emissivity than water, oil heated by the sun emits radiation that shows up on infrared images based on oil thickness.

An infrared image of the same oil spill as the picture above. Note how IR addresses the sun glitter issue and creates better contrast based on oil thickness. (Image courtesy of National Center for Biotechnology Information.)
An infrared image of the same oil spill as the picture above. Note how IR addresses the sun glitter issue and creates better contrast based on oil thickness. (Image courtesy of Environment Canada.)

Thermal imaging presents its own limitations. For one, IR images taken at night do not have the definition of those captured in the daytime. Furthermore, IR sensors are prone to interference due to the fact that natural objects like seaweed, sediment and organic matter often appear like oil.

Satellites such as AVHRR and MODIS SST (sea surface temperature) have been used in the past to collect IR data for the mapping of oil spills.

Laser Spectroscopy

Laser fluorosensors are highly effective for detecting oil spills not only on water, but also on soil, ice and other surfaces.

The technology provides a depth of valuable chemical analysis information to oil spill response workers. Laser spectroscopy involves oil absorbing UV light and releasing the extra energy at wavelengths that are unique to different oil classes. As a result, substances like oil and seaweed can be differentiated due to their fluorescence at distinctive wavelengths. For example, chlorophyll fluoresces at 685nm, while crude oil fluoresces within a 400–650nm range after activation with a 308nm excimer laser.

Fluorescence spectra of various types of oil after a 308nm excimer laser excitation. (Image courtesy of National Center for Biotechnology Information.)
Fluorescence spectra of various types of oil after a 308nm excimer laser excitation. (Image courtesy of Environment Canada.)

UV lasers in the 300–355nm region, such as the XeCl excimer laser (308nm), the nitrogen laser (337nm), the XeF excimer laser (351nm), and the frequency-tripled Nd:YAG laser (355nm), have become available commercially in the last few years and are now considered essential for oil spill detection. Further advancements of the technology are ongoing.

Radar

Last but definitely not least, we have radar imagery. Radar is uniquely equipped for the surveillance of large areas at any time of the day (or night), and even during rainy weather or fog. All that’s essentially needed is a sea state that is not too high or low, and wind speeds between 1.5–10m/s.

Radar is widely used for oil spill mapping—not only of large spills but also of oil platform leaks. Both ship-borne and satellite radar are dominant methods of monitoring offshore oil spills. Oil on water attenuates capillary waves, yielding a radar image where oil shows up as dark spots.

Satellite image of oil platform discharges in the Caspian Sea. (Image courtesy of National Center for Biotechnology Information.)
Satellite image of oil platform discharges in the Caspian Sea. (Image courtesy of Macdonald-Detwiler.)

One major limitation of radar is that it is susceptible to interference from many slick look-alikes, such as wave shadows, topographical features, seaweed beds and even sea-life sperm. Data processing advancements are aiming to eliminate these look-alikes and automate the oil detection process. Steps for successful radar data processing involve an initial assessment of an image for quality, and the removal of noise. Next, wind fields are removed using wind maps of the same scale as the radar image. Fixed geographic features and known seaweed beds are blanked out using GIS data. Finally, algorithms are used to differentiate oil dark spots from their look-alikes. For example, texture analysis shows oil having a consistent texture versus the sea’s rougher texture, while shape analysis illustrates how oil contours to different shapes from its look-alikes. In the past decade, intelligent machine learning algorithms have become readily available to oil response workers, leading to vast improvements in radar image processing.

And there you have it: a brief overview of the technologies that are currently being used for the detection of oil spills on water. While each technique has its practical uses and drawbacks, these methods are still economical and serve to protect our waterways from becoming polluted oil zones.

Recommended For You