Bilge Dumping off the Coast of Brazil

The cause of the massive oil spill plaguing Brazil’s beaches is still unknown, but monitoring reveals a potential new bilge dumping incident

We still haven’t found the cause of the massive oil spill that’s been plaguing Brazil’s beaches since early September.  

But SkyTruth’s continued surveillance of the coast of northeastern Brazil, in response to one of the country’s worst oil-related environmental disasters ever, has uncovered what appears to be another previously unreported bilge dumping incident off the coast of Joao Pessoa in the state of Paraiba. Located about 20 km offshore, a 25 km-long slick appears to originate from the Grajau, a Brazil-flagged liquefied petroleum gas (LPG) tanker. Slicks such as this are a hallmark of the intentional dumping of untreated, oily bilge wastes from vessels underway at sea, although there may be other explanations for this slick (for example, the ship was experiencing a serious mechanical problem). The slick (a long, dark streak) and vessel (a bright spot at the south end of the slick) are shown on this Sentinel-1 radar satellite image taken on the 19th of July. We identified the vessel using their public AIS tracking broadcasts, extracted from the ShipView vessel-tracking platform. The image was captured at 07:53 UTC; a careful look at the AIS broadcasts from Grajau just before and after the image was taken show that the vessel we can see on the radar image is very likely Grajau.

Recent discoveries of bilge dumping in the Atlantic Ocean along Brazil’s coast reveal that this is a persistent problem that — as in many places — lacks effective enforcement. None of the slicks we’ve seen appear big enough to be the source of the oil plaguing Brazil’s beaches. This potential bilge slick from Grajau is no exception: it’s a modest-sized slick compared with the dozens of bilge slicks we’ve seen from other places around the world that are occasionally more than 100 km long. And this slick, just 20 km offshore, probably would have dissipated or washed ashore several weeks before the thick globs of heavy oil began to appear on the beaches in early September.

Nevertheless, bilge dumping is a chronic source of oil pollution in the ocean that has been hidden for too long. Now that we can see it, and can identify the likely polluters, it’s time for governments to take action to bring this illegal practice to an end.

AIS ship-tracking broadcasts (red dots) from the Brazil-flagged LPG tanker Grajau, overlain on a Sentinel-1 radar satellite image showing an apparent bilge-dumping slick (dark streak) and the vessel that appears to be responsible (bright spot, indicated within the red circle). Based on the AIS data, we think this vessel is likely the Grajau. See inset map at upper right for detail. Image was collected at 07:53 on July 19.

The location of the boat, relative to Brazil’s coastline.

More oil pollution in southeast Asia: suspected bilge dumping off Indonesia and The Philippines

[This analysis of oil pollution in the waters of southeast Asia was written as part of a collaborative effort between SkyTruth team members Lucy Meyer and Brendan Jarrell.]

Our routine monitoring of the world’s oceans has led to some extraordinary findings. For example, in previous updates, we’ve identified oil slicks in traffic-heavy locations like the Strait of Malacca. But as you’ll see in this post, bilge dumps occur elsewhere in southeast Asia. 

Those who follow our posts are probably familiar with how we identify vessels at sea. To new readers, let us explain what bilge dumping is and how we identify potentially responsible vessels. Bilge dumping is the disposal of waste water from a ship’s lower hull. Bilge water is supposed to be treated before it’s discharged, but sometimes vessel operators will bypass the pollution control equipment and flush oily, untreated bilge into the ocean – in direct violation of marine pollution law. We use images from satellites to monitor for illegal bilge dumping. In satellite imagery, oily bilge dumps usually form distinctive linear slicks. By matching the time of the imagery to broadcasts from a vessel tracking service called automatic identification system (AIS), we can determine the identity of vessels that appear to be causing the slicks. We used this process to identify the vessel associated with a long bilge slick in Figure 1 below.


Figure 1: A vessel shown passing through the Sunda Strait, identified as the Sungai Gerong, apparently trailing a long oily bilge slick.


This Sentinel-1 radar satellite image from July 2nd shows a slick about 177 kilometers long around the southwest tip of Banten Province, Island of Java, Indonesia (Figure 1). In the yellow box, you can see a vessel at the head of the slick. By investigating AIS broadcasts from exactEarth’s ShipView service, we identified an Indonesian oil products tanker named the Sungai Gerong as the likely vessel. The satellite scene, captured at 22:33 UTC (Coordinated Universal Time), shows a slick that closely aligns to the AIS broadcasts from the Sungai Gerong.

You’ll probably notice that the tail-end of the slick is a bit contorted and offset from the track of the Sungai Gerong. The slick’s appearance was likely influenced by ocean currents and local weather conditions between the time of the ship’s passing and when the image was taken. Global wind maps show that there were 10-15 knot winds blowing northwest up to six hours before the image was acquired. This data suggests that wind likely impacted the slick’s appearance. As a result, we believe that the Sungai Gerong is the likely source of this slick.

Using AIS, we tracked the Sungai Gerong as it traveled north through the Sunda Strait — the body of water between the Indonesian islands of Java and Sumatra — to the port of Jakarta. Similar to the Strait of Malacca, the Sunda is an important waterway that connects the Indian Ocean to the Java Sea. Though not as dense with marine traffic as the Malacca Strait, the Sunda is still subjected to pollution from vessels. 

We also recently identified two suspected bilge dumps in the Philippines (Figure 2). Occurring on July 6th in the South China Sea, a 238 kilometer long slick behind the vessel in this Sentinel-1 radar image looks like a bilge dump. The Philippine island of Palawan, a popular tourist destination for its beautiful natural landscape, appears on the right side of the map frame. Another smaller slick without a known source is visible to the left of the larger slick.


Figure 2: The Ulaya makes its way through the South China Sea. Palawan Island, a part of the Philippines, can be seen to the right.


Using AIS broadcasts from ShipView, we identified the Ulaya, a Thai oil tanker, as a possible source of the slick. The last AIS broadcast from the Ulaya (seen directly above the ship) was transmitted fifteen minutes before the image was captured. These AIS broadcasts give us reason to believe that the Ulaya could be responsible for this slick. Moreover, ShipView shows that the vessel was headed towards the Port of Belawan in the Strait of Malacca with a shipment of  Dangerous Goods. According to the International Maritime Organization, a United Nations agency that regulates global shipping, chemicals falling under this classification are “hazardous to marine environments.” Thus, a slick from this ship could be of greater concern than usual.

These examples show that bilge dumping continues to be a problem in the waters of southeast Asia. But with satellite imagery, anyone, anywhere can see what’s happening on the water and help to raise the alarm. We hope that our persistent and careful surveillance will inspire others to pressure policy makers, government regulators, and the shipping industry to take strong, coordinated action to stop bilge dumping.

“Well Kick” Causes Spill in Java Sea

Following up on recent reports of oil in the water off the north coast of Karawang Regency, West Java, Indonesia, SkyTruth has picked up a slick in Sentinel-1 radar imagery. In the image from July 18th, an unidentified platform (circled in red) located roughly 12 km north of the Karawang shore is shown emitting a 34.7 km-long slick into the Java Sea. A story written by the local Jakarta Post on July 18th describes state-owned energy firm Pertamina’s decision to evacuate personnel and halt operations at an offshore production rig in their Offshore Northwest Java (ONWJ) block. The evacuation was ordered after a dangerous “well kick”, or unplanned release of gas caused by low pressure in a wellbore, initiated a large slick on the 16th of July. A separate report released by the Jakarta Post five days later indicated that the Indonesian Transportation Ministry teamed up with Pertamina in response to the oil-related event, along with several other smaller entities in the area. The response vessels were able to set up a boom around the perimeter of the offshore platform. Unfortunately, this didn’t stop oil from reaching villages and beaches on West Java’s coast. Given the fact that several vessels surround the unidentified object in the Sentinel-1 image, we believe that this could be the affected drilling platform. Pertamina’s upstream director Dharmawan Samsu estimated that it will take approximately eight weeks for the oil and gas leakage to be plugged.

The unidentified platform (circled in red) can be seen leaking oil into the Java Sea. Several small vessels are in the platform’s proximity.

Using machine learning to map the footprint of fracking in central Appalachia

Fossil fuel production has left a lasting imprint on the landscapes and communities of central and northern Appalachia.  Mountaintop mining operations, pipeline right-of-ways, oil and gas well pads, and hydraulic fracturing wastewater retention ponds dot the landscapes of West Virginia and Pennsylvania.  And although advocacy groups have made progress pressuring regulated industries and state agencies for greater transparency, many communities in central and northern Appalachia are unaware of, or unclear about, the extent of human health risks that they face from exposure to these facilities.  

A key challenge is the discrepancy that often exists between what is on paper and what is on the landscape.  It takes time, money, and staff (three rarities for state agencies always under pressure to do more with less) to map energy infrastructure, and to keep those records updated and accessible for the public.  But with advancements in deep learning, and with the increasing amount of satellite imagery available from governments and commercial providers, it might be possible to track the expansion of energy infrastructure—as well as the public health risks that accompany it—in near real-time.

Figure 1.  Oil and gas well pad locations, 2005 – 2015.

Mapping the footprint of oil and gas drilling, especially unconventional drilling or “fracking,” is a critical piece of SkyTruth’s work.  Since 2013, we’ve conducted collaborative image analysis projects called “FrackFinder” to fill the gaps in publicly available information about the location of fracking operations in the Marcellus and Utica Shale.  In the past, we relied on several hundred volunteers to identify and map oil and gas well pads throughout Ohio, Pennsylvania, and West Virginia.  But we’ve been working on a new approach: automating the detection of oil and gas well pads with machine learning.  Rather than train several hundred volunteers to identify well pads in satellite imagery, we developed a machine learning model that could be deployed across thousands of computers simultaneously.  Machine learning is at the heart of today’s companies. It’s the technology that enables Netflix to recommend new shows that you might like, or that allows digital assistants like Google, Siri, or Alexa to understand requests like, “Hey Google, text Mom I’ll be there in 20 minutes.”

Examples are at the core of machine learning.  Rather than try to “hard code” all of the characteristics that define a modern well pad (they are generally square, generally gravel, and generally littered with industrial equipment), we teach computers what they look like by using examples.  Lots of examples. Like, thousands or even millions of them, if we can find them. It’s just like with humans: the more examples of something that you see, the easier it is to recognize that thing later. So, where did we get a few thousand images of well pads in Pennsylvania?  

We started with SkyTruth’s Pennsylvania oil and gas well pad dataset. The dataset contains well pad locations identified in National Agriculture Imagery Program (NAIP) aerial imagery from 2005, 2008, 2010, 2013, and 2015 (Figure 1).  We uploaded this dataset to Google Earth Engine, and used it to create a collection of 10,000 aerial images in two classes: “well pad” and “non-well pad.” We created the training images by buffering each well pad by 100 meters, clipping the NAIP imagery to the bounding box, and exporting each image.

The images above show three training examples from our “well pad” class. The images below show three training examples taken from our “non-well pad” class.

We divided the dataset into three subsets: a training set with 4,000 images of each class, a validation set with 500 images of each class, and a test set with 500 images of each class.  We combined this work in Google Earth Engine with Google’s powerful TensorFlow deep learning library.  We used our 8,000 training images (4,000 from each class, remember) and TensorFlow’s high-level Keras API to train our machine learning model.  So what, exactly, does that mean? Well, basically, it means that we showed the model thousands and thousands of examples of what well pads are (i.e., images from our “well pad” class) and what well pads aren’t (i.e., images from our “non-well pad” class).  We trained the model for twenty epochs, meaning that we showed the model the entire training set (8,000 images, remember) twenty times.  So, basically, the model saw 160,000 examples, and over time, it “learned” what well pads look like.

Our best model run returned an accuracy of 84%, precision and recall measures of 87% and 81%, respectively, and a false positive rate and false negative rate of 0.116 and 0.193, respectively.  We’ve been pleased with our initial model runs, but there is plenty of room for improvement. We started with the VGG16 model architecture that comes prepackaged with Keras (Simonyan and Zisserman 2014, Chollet 2018).  The VGG16 model architecture is no longer state-of-the-art, but it is easy to understand, and it was a great place to begin.  

After training, we ran the model on a few NAIP images to compare its performance against well pads collected by SkyTruth volunteers for our 2015 Pennsylvania FrackFinder project.  Figures 4 and 6 depict the model’s performance on two NAIP images near Williamsport, PA. White bounding boxes indicate landscape features that the model predicted to be well pads.  Figures 5 and 7 depict those same images with well pads (shown in red) delineated by SkyTruth volunteers.

Figure 4.  Well pads detected by our machine learning algorithm in NAIP imagery from 2015.
Figure 5.  Well pads detected by SkyTruth volunteers in NAIP imagery from 2015.
Figure 6.  Well pads detected by our machine learning algorithm in NAIP imagery from 2015.
Figure 7.  Well pads detected by SkyTruth volunteers in NAIP imagery from 2015.

One of the first things that stood out to us was that our model is overly sensitive to strong linear features.  In nearly every training example, there is a clearly-defined access road that connects to the well pad. As a result, the model regularly classified large patches of cleared land or isolated developments (e.g., warehouses) at the end of a linear feature as a well pad.  Another major weakness is that our model is also overly sensitive to active well pads.  Active well pads tend to be large, gravel squares with clearly defined edges. Although these well pads may be the biggest concern, there are many “reclaimed” and abandoned well pads that lack such clearly defined edges.  Regrettably, our model is overfit to highly-visible active wells pads, and it performs poorly on lower-visibility drilling sites that have lost their square shape or that have been revegetated by grasses.

Nevertheless, we think this is a good start.  Despite a number of false detections, our model was able to detect all of the well pads previously identified by volunteers in images 5 and 7 above.  In several instances, false detections consisted of energy infrastructure that, although not active well pads, remain of high interest to environmental and public health advocates as well as state regulators: abandoned well pads, wastewater impoundments, and recent land clearings.  NAIP imagery is only collected every two or three years, depending on funding. So, tracking the expansion of oil and gas drilling activities in near real-time will require access to a high resolution, near real-time imagery stream (like Planet, for instance).  For now, we’re experimenting with more current model architectures and with reconfiguring the model for semantic segmentation — extracting polygons that delineate the boundaries of well pads which can be analyzed in mapping software by researchers and our partners working on the ground.

Keep checking back for updates.  We’ll be posting the training data that we created, along with our initial models, as soon as we can.

Monitoring the tailings dam failure of the Córrego do Feijão mine

On Friday, January 25th, the tailings dam to the Córrego do Feijão mine burst near Brumadinho, State of Minas Gerais, Brazil (the moment of failure was captured on video). Operated by Brazilian mining company Vale S.A., this incident recalls the collapse of Vale’s Samarco Mine in 2015 which unleashed 62 million cubic meters of toxic sludge downstream. As of Monday, the death toll reached 120, however, the full extent of damage is unknown. To monitor the impact, here is a Sentinel-2 scene of Córrego do Feijão from eighteen days before and seven days after the dam’s failure. As of February 2nd, approximately 2.85 km2 of sludge surrounds the region.

Sentinel 2 scene showing the extent of flooding as a result of the tailings dam failure. As a result of the failure, 3 billion gallons of mining waste were spilled.

This slider, below, shows the area near the town of Brumadinho before and after the dam failure with the inundation highlighted in yellow, it can be accessed here.