The SkyTruth app that maps potential drillout scenarios across the landscape of Allegheny County, PA has officially received its first update! In an effort to make the experience more user-friendly, explanatory text and tips have been added. Our app has also been updated to remove from the drillout scenario areas such as major highways and the Pittsburgh International Airport, where drilling would obviously not take place.
A screenshot of the app when first initialized.
At the request of some users, we’ve also tabulated the results for the potential drillout scenarios by municipality. See the results in this table showing the number of occupied structures within two miles of a hypothetical drilling site, based on a given setback distance (in feet) and drilling site spacing (in acres), for every township and borough.
We were also asked to calculate the number of occupied structures located within 500 feet, and within two miles, of existing Marcellus Shale drilling and fracking sites. According to our analysis, 78 occupied structures fall within 500 feet of an active drilling site in Allegheny County and 67,673 occupied structures sit within two miles of an active drilling site. Recent scientific research has found human health impacts for people living within 2 miles of a drilling site.
Be sure to check out these insightful new updates for yourself. Give the app a try and let us know what you think by contacting Brendan at info@skytruth.org with any feedback you might have!
https://skytruth.org/wp-content/uploads/2019/04/Aco_AppUpdate_image.jpg8571920Brendan Jarrell/wp-content/uploads/2016/08/test_logo.pngBrendan Jarrell2019-04-02 10:05:272019-04-02 10:10:11Allegheny County Drilling App Receives Its First Update
Possible Bilge Dumping by Indonesian Cement Carrier in the Strait of Malacca
By Lucy Meyer
On February 15, 2019, a vessel that appeared to be releasing oily waste was captured by satellite almost 10 kilometers offshore Peureulak, a small town in Aceh Province, on the northern tip of the Indonesian island of Sumatra. Radar imagery from the European Space Agency’s Sentinel-1 satellite shows an 18-kilometer slick trailing a northbound ship, visible as a bright spot at the end of the dark slick.
Figure 1. Sentinel-1 radar satellite image showing suspected bilge-dumping (dark, linear slick) off Sumatra on February 15, 2019.
The ship is traveling through the Strait of Malacca, a narrow strip of water between Sumatra and the Malay Peninsula. The Strait is one of the world’s busiest shipping lanes as it is both the shortest and most convenient path between the Indian and Pacific Oceans. Due to the Strait’s high density of marine traffic of all types, oil spills — accidental and intentional — are likely to occur. Figure 1 illustrates suspected bilge dumping, a typically intentional discharge of oily waste from ships to reduce ballast water or free up space in the cargo holds. Typically, bilge-dumps form distinctive linear slicks visible on satellite imagery.
While radar satellite images are very useful tools for detecting slicks, they are typically not detailed enough to allow identification of the responsible vessel. However, many vessels broadcast their identity and other information using the radio-frequency Automatic Identification System (AIS). AIS use is required for all large cargo vessels and tankers. By studying the AIS broadcasts in this area using exactEarth’s ShipView service, which collects the signals using satellites and ground-based receivers, SkyTruth analyst Bjorn Bergman determined the Indonesian cement carrier PERKASA (Figure 2) was at this location when the Sentinel-1 radar image was acquired. Formerly known as KOEI MARU NO 7, the vessel was built in 1981 by Ube Industries, Ltd., a Japanese chemical manufacturing company. Today, the ship is operated by PT Indobaruna Bulk Transport (IBT), an Indonesian shipping company based in Jakarta.
Figure 3 shows the PERKASA’s AIS-derived track overlain on the Sentinel-1 image, revealing a very close match between the vessel’s path and the suspected bilge slick. The AIS signal immediately to the south of the vessel location on the image indicates it was traveling 11 knots (~20.4 km/h) at 11:17 UTC; the signal immediately following at 12:10 UTC indicate the vessel was traveling 10.8 knots (~20.0 km/h). Using the location data encoded with these AIS signals, we calculated the likely position of PERKASA at the instant the image was acquired (11:43 UTC). The ship’s predicted location closely matches the vessel’s position in the Sentinel-1 image, and no other vessels broadcasting AIS were likely candidates for a match. This leads us to infer that PERKASA is the vessel seen apparently discharging oily bilge waste in the satellite image.
Figure 4. Zoomed-out view of Sentinel-1 image showing a series of patchy slicks along the coast of Aceh Province, Indonesia. Dark, linear slick at upper left is the suspected bilge slick from PERKASA shown in Figures 1 and 3.
To the south, a chain of less-distinctive slicks along the coast are roughly aligned with PERKASA’s track (Figure 4). These slicks are broad and striated as opposed to the slender 18-kilometer long slick, which could be a result of wind and current blowing apart what had originally been a series of discharges from the vessel. The AIS transmissions from PERKASA are infrequent in this region (Figure 5), making us somewhat less confident that this vessel was also the source of these patchy slicks.
Figure 5. PERKASA’s AIS-derived track overlain on Figure 4.
The operator of PERKASA, IBT, claims “we put high priority in safety by adhering to policies, practices, and procedures in our Safety Management System to ensure the safety of crews, staffs, cargoes, vessels, as well as environment.” In addition nearly all of IBT’s fleet is registered with classification societies. According to The International Association of Classification Societies (IACS), the purpose of a classification society is “to provide classification and statutory services and assistance to the maritime industry and regulatory bodies as regards maritime safety and pollution prevention.” IACS is a non-governmental organization composed of twelve classification societies. PERKASA is registered with Biro Klasifikasi Indonesia (BKI) and Nippon Kaiji Kyokai (ClassNK), which is a member of IACS.
One of the certification services provided by ClassNK is the Verification for Clean Shipping Index (CSI). The objective of CSI is to verify the environmental performance of a vessel’s operations in five areas, including water and wastes. Ballast water, sewage/black water, garbage, sludge oils, and bilge water are covered under this category.
Bilge dumping — intentional or otherwise — would seem to violate the principles touted by the vessel operator, and call into question the effectiveness of the classification societies.
[This is a guest post about the ongoing Taylor Energy oil spill from Dr. Ian MacDonald, oceanographer at Florida State University. Ian helped SkyTruth make independent estimates of the size of the Deepwater Horizon oil spill in 2010 that dwarfed the estimates told to the public by BP.]
As recently as two days ago — March 13, 2019 — pollution experts at the National Oceanic and Atmospheric Administration were reporting a 14 square-mile oil slick that originated out in the Gulf of Mexico about 12 miles from the Birdfoot Delta’s farthest bit of land. By now there are hundreds of satellite and aerial images telling the same, sorry story. The source is the wreck of MC20A, an oil platform owned by Taylor Energy Company that was destroyed by winds, waves, and mudslides spawned by Hurricane Ivan in 2004. Last fall, the Coast Guard and other agencies federalized the response to an oil spill that has been going on for fourteen years and counting, disinviting the company from the latest effort to stem the flow by attaching a massive containment dome to what remains of the platform. Although the company has long insisted that the spill is trivial–no more than 10 gallons per day–a growing chorus of scientists have disagreed, by orders of magnitude. My personal estimate is 96 barrels (4032 gallons) per day, and I tend toward the low end of the scientific opinions.
Why the Feds changed their mind, and how come it took so long, are questions I address in a report on the longest offshore oil spill in U.S. history. I tell the story from my perspective as an oceanographer who studies natural and unnatural oil inputs to the ocean, and based on what is now over seven years of funded research on MC20A.
Storms like Ivan seem to be growing more common. The sediments lost from the drastic reduction of Louisiana wetlands have been deposited on the slope in huge mud lobes–some of which will inevitably slide toward the sprawling network of aging platforms and pipelines that surrounds the Delta. The lessons we learn from MC20A, and the response by a unified command under the direction of the US Coast Guard, may be put to the test again, possibly much more severely than with MC20A.
https://skytruth.org/wp-content/uploads/2019/03/Screen-Shot-2019-03-15-at-5.46.26-PM.png381648John Amos/wp-content/uploads/2016/08/test_logo.pngJohn Amos2019-03-15 22:58:012019-03-19 12:14:05What can we learn from the longest oil spill in US history?
The trends aren’t encouraging: Industrialization, urban development, deforestation, overfishing, mining and pollution are accelerating the rate of global warming and damaging ecosystems around the world. The pace of environmental destruction has never been as great as it is today. Despite this grim assessment, I believe there’s reason to be hopeful for a brighter future.
I’m optimistic because of a new and powerful conservation opportunity: the explosion of satellite and computing technology that now allows us to see what’s happening on the ground and on the water, everywhere, in near real-time.
Up until now we’ve been inspiring people to take action by using satellites to show them what’s already happened to the environment, typically months or even years ago. But technology has evolved dramatically since I started SkyTruth, and today we can show people what’s happening right now, making it possible to take action that can minimize or even stop environmental damage before it occurs. For example, one company, Planet, now has enough satellites in orbit to collect high-resolution imagery of all of the land area on Earth every day. Other companies and governments are building and launching fleets of satellites that promise to multiply and diversify the stream of daily imagery, including radar satellites that operate night and day and can see through clouds, smoke and haze.
Just a few of the Earth-observation satellites in orbit. Image courtesy NASA.
The environmental monitoring potential of all this new hardware is thrilling to our team here at SkyTruth, but it also presents a major challenge: it simply isn’t practical to hire an army of skilled analysts to look at all of these images, just to identify the manageable few that contain useful information.
Artificial intelligence is the key to unlocking the conservation power of this ever-increasing torrent of imagery.
Taking advantage of the same machine-learning technology Facebook uses to detect and tag your face in a friend’s vacation photo, we are training computers to analyze satellite images and detect features of interest in the environment: a road being built in a protected area, logging encroaching on a popular recreation area, a mining operation growing beyond its permit boundary, and other landscape and habitat alterations that indicate an imminent threat to biodiversity, ecosystem integrity, and human health. By applying this intelligence to daily satellite imagery, we can make it possible to detect changes happening in the environment in near real-time. Then we can immediately alert anyone who wants to know about it, so they can take action if warranted: to investigate, to document, to intervene.
We call this program Conservation Vision.
And by leveraging our unique ability to connect technology and data providers, world-class researchers and high-impact conservation partners, we’re starting to catalyze action and policy success on the ground.
We’re motivated to build this approach to make environmental information available to people who are ready and able to take action. We’ve demonstrated our ability to do this through our partnership with Google and Oceana with the launch and rapid growth of Global Fishing Watch, and we’re already getting positive results automating the detection of fracking sites around the world. We have the technology. We have the expertise. We have the track record of innovation for conservation. And we’ve already begun the work.
Stay tuned for more updates and insights on how you can be part of this cutting-edge tool for conservation.
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.
https://skytruth.org/wp-content/uploads/2019/02/image4-1.png8111856Ry Covington/wp-content/uploads/2016/08/test_logo.pngRy Covington2019-02-13 13:45:372019-04-04 10:25:51Using machine learning to map the footprint of fracking in central Appalachia
SkyTruth and/or its third party tools use cookies to monitor traffic in the aggregate on our websites and applications and to improve user experience. If you continue browsing, we'll assume you agree to the use of cookies. You can change your cookie settings at any time in your browser settings. Find out more in our privacy policy.
Allegheny County Drilling App Receives Its First Update
/in FrackFinder, Hydraulic Fracturing, Natural Gas /by Brendan JarrellThe SkyTruth app that maps potential drillout scenarios across the landscape of Allegheny County, PA has officially received its first update! In an effort to make the experience more user-friendly, explanatory text and tips have been added. Our app has also been updated to remove from the drillout scenario areas such as major highways and the Pittsburgh International Airport, where drilling would obviously not take place.
A screenshot of the app when first initialized.
At the request of some users, we’ve also tabulated the results for the potential drillout scenarios by municipality. See the results in this table showing the number of occupied structures within two miles of a hypothetical drilling site, based on a given setback distance (in feet) and drilling site spacing (in acres), for every township and borough.
We were also asked to calculate the number of occupied structures located within 500 feet, and within two miles, of existing Marcellus Shale drilling and fracking sites. According to our analysis, 78 occupied structures fall within 500 feet of an active drilling site in Allegheny County and 67,673 occupied structures sit within two miles of an active drilling site. Recent scientific research has found human health impacts for people living within 2 miles of a drilling site.
Be sure to check out these insightful new updates for yourself. Give the app a try and let us know what you think by contacting Brendan at info@skytruth.org with any feedback you might have!
PERKASA Caught Bilge-Dumping?
/in Oceans, Uncategorized /by Lucy MeyerPossible Bilge Dumping by Indonesian Cement Carrier in the Strait of Malacca
By Lucy Meyer
On February 15, 2019, a vessel that appeared to be releasing oily waste was captured by satellite almost 10 kilometers offshore Peureulak, a small town in Aceh Province, on the northern tip of the Indonesian island of Sumatra. Radar imagery from the European Space Agency’s Sentinel-1 satellite shows an 18-kilometer slick trailing a northbound ship, visible as a bright spot at the end of the dark slick.
The ship is traveling through the Strait of Malacca, a narrow strip of water between Sumatra and the Malay Peninsula. The Strait is one of the world’s busiest shipping lanes as it is both the shortest and most convenient path between the Indian and Pacific Oceans. Due to the Strait’s high density of marine traffic of all types, oil spills — accidental and intentional — are likely to occur. Figure 1 illustrates suspected bilge dumping, a typically intentional discharge of oily waste from ships to reduce ballast water or free up space in the cargo holds. Typically, bilge-dumps form distinctive linear slicks visible on satellite imagery.
While radar satellite images are very useful tools for detecting slicks, they are typically not detailed enough to allow identification of the responsible vessel. However, many vessels broadcast their identity and other information using the radio-frequency Automatic Identification System (AIS). AIS use is required for all large cargo vessels and tankers. By studying the AIS broadcasts in this area using exactEarth’s ShipView service, which collects the signals using satellites and ground-based receivers, SkyTruth analyst Bjorn Bergman determined the Indonesian cement carrier PERKASA (Figure 2) was at this location when the Sentinel-1 radar image was acquired. Formerly known as KOEI MARU NO 7, the vessel was built in 1981 by Ube Industries, Ltd., a Japanese chemical manufacturing company. Today, the ship is operated by PT Indobaruna Bulk Transport (IBT), an Indonesian shipping company based in Jakarta.
Figure 3 shows the PERKASA’s AIS-derived track overlain on the Sentinel-1 image, revealing a very close match between the vessel’s path and the suspected bilge slick. The AIS signal immediately to the south of the vessel location on the image indicates it was traveling 11 knots (~20.4 km/h) at 11:17 UTC; the signal immediately following at 12:10 UTC indicate the vessel was traveling 10.8 knots (~20.0 km/h). Using the location data encoded with these AIS signals, we calculated the likely position of PERKASA at the instant the image was acquired (11:43 UTC). The ship’s predicted location closely matches the vessel’s position in the Sentinel-1 image, and no other vessels broadcasting AIS were likely candidates for a match. This leads us to infer that PERKASA is the vessel seen apparently discharging oily bilge waste in the satellite image.
To the south, a chain of less-distinctive slicks along the coast are roughly aligned with PERKASA’s track (Figure 4). These slicks are broad and striated as opposed to the slender 18-kilometer long slick, which could be a result of wind and current blowing apart what had originally been a series of discharges from the vessel. The AIS transmissions from PERKASA are infrequent in this region (Figure 5), making us somewhat less confident that this vessel was also the source of these patchy slicks.
The operator of PERKASA, IBT, claims “we put high priority in safety by adhering to policies, practices, and procedures in our Safety Management System to ensure the safety of crews, staffs, cargoes, vessels, as well as environment.” In addition nearly all of IBT’s fleet is registered with classification societies. According to The International Association of Classification Societies (IACS), the purpose of a classification society is “to provide classification and statutory services and assistance to the maritime industry and regulatory bodies as regards maritime safety and pollution prevention.” IACS is a non-governmental organization composed of twelve classification societies. PERKASA is registered with Biro Klasifikasi Indonesia (BKI) and Nippon Kaiji Kyokai (ClassNK), which is a member of IACS.
One of the certification services provided by ClassNK is the Verification for Clean Shipping Index (CSI). The objective of CSI is to verify the environmental performance of a vessel’s operations in five areas, including water and wastes. Ballast water, sewage/black water, garbage, sludge oils, and bilge water are covered under this category.
Bilge dumping — intentional or otherwise — would seem to violate the principles touted by the vessel operator, and call into question the effectiveness of the classification societies.
What can we learn from the longest oil spill in US history?
/in Uncategorized /by John AmosAs recently as two days ago — March 13, 2019 — pollution experts at the National Oceanic and Atmospheric Administration were reporting a 14 square-mile oil slick that originated out in the Gulf of Mexico about 12 miles from the Birdfoot Delta’s farthest bit of land. By now there are hundreds of satellite and aerial images telling the same, sorry story. The source is the wreck of MC20A, an oil platform owned by Taylor Energy Company that was destroyed by winds, waves, and mudslides spawned by Hurricane Ivan in 2004. Last fall, the Coast Guard and other agencies federalized the response to an oil spill that has been going on for fourteen years and counting, disinviting the company from the latest effort to stem the flow by attaching a massive containment dome to what remains of the platform. Although the company has long insisted that the spill is trivial–no more than 10 gallons per day–a growing chorus of scientists have disagreed, by orders of magnitude. My personal estimate is 96 barrels (4032 gallons) per day, and I tend toward the low end of the scientific opinions.
Why the Feds changed their mind, and how come it took so long, are questions I address in a report on the longest offshore oil spill in U.S. history. I tell the story from my perspective as an oceanographer who studies natural and unnatural oil inputs to the ocean, and based on what is now over seven years of funded research on MC20A.
Storms like Ivan seem to be growing more common. The sediments lost from the drastic reduction of Louisiana wetlands have been deposited on the slope in huge mud lobes–some of which will inevitably slide toward the sprawling network of aging platforms and pipelines that surrounds the Delta. The lessons we learn from MC20A, and the response by a unified command under the direction of the US Coast Guard, may be put to the test again, possibly much more severely than with MC20A.
Will we be ready?
Read my report to learn more.
CONSERVATION VISION
/in Conservation Vision /by John AmosUsing Artificial Intelligence to Save the Planet
A letter from our founder, John Amos
The trends aren’t encouraging: Industrialization, urban development, deforestation, overfishing, mining and pollution are accelerating the rate of global warming and damaging ecosystems around the world. The pace of environmental destruction has never been as great as it is today. Despite this grim assessment, I believe there’s reason to be hopeful for a brighter future.
I’m optimistic because of a new and powerful conservation opportunity: the explosion of satellite and computing technology that now allows us to see what’s happening on the ground and on the water, everywhere, in near real-time.
Up until now we’ve been inspiring people to take action by using satellites to show them what’s already happened to the environment, typically months or even years ago. But technology has evolved dramatically since I started SkyTruth, and today we can show people what’s happening right now, making it possible to take action that can minimize or even stop environmental damage before it occurs. For example, one company, Planet, now has enough satellites in orbit to collect high-resolution imagery of all of the land area on Earth every day. Other companies and governments are building and launching fleets of satellites that promise to multiply and diversify the stream of daily imagery, including radar satellites that operate night and day and can see through clouds, smoke and haze.
The environmental monitoring potential of all this new hardware is thrilling to our team here at SkyTruth, but it also presents a major challenge: it simply isn’t practical to hire an army of skilled analysts to look at all of these images, just to identify the manageable few that contain useful information.
Artificial intelligence is the key to unlocking the conservation power of this ever-increasing torrent of imagery.
Taking advantage of the same machine-learning technology Facebook uses to detect and tag your face in a friend’s vacation photo, we are training computers to analyze satellite images and detect features of interest in the environment: a road being built in a protected area, logging encroaching on a popular recreation area, a mining operation growing beyond its permit boundary, and other landscape and habitat alterations that indicate an imminent threat to biodiversity, ecosystem integrity, and human health. By applying this intelligence to daily satellite imagery, we can make it possible to detect changes happening in the environment in near real-time. Then we can immediately alert anyone who wants to know about it, so they can take action if warranted: to investigate, to document, to intervene.
We call this program Conservation Vision.
And by leveraging our unique ability to connect technology and data providers, world-class researchers and high-impact conservation partners, we’re starting to catalyze action and policy success on the ground.
We’re motivated to build this approach to make environmental information available to people who are ready and able to take action. We’ve demonstrated our ability to do this through our partnership with Google and Oceana with the launch and rapid growth of Global Fishing Watch, and we’re already getting positive results automating the detection of fracking sites around the world. We have the technology. We have the expertise. We have the track record of innovation for conservation. And we’ve already begun the work.
Stay tuned for more updates and insights on how you can be part of this cutting-edge tool for conservation.
Using machine learning to map the footprint of fracking in central Appalachia
/in FrackFinder, Hydraulic Fracturing, Watchdog /by Ry CovingtonFossil 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.
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.
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.