SkyTruth Annual Report

If you can see it, you can CHANGE it.

Artisanal gold mining in Madre de Dios, Peru. Google Earth ©2020 CNES / Airbus.

Annual Report

Artisanal gold mining in Madre de Dios, Peru. Google Earth ©2020 CNES / Airbus.

Message From the President

In the huge auditorium of the MGM Grand Hotel in Las Vegas, surrounded by some of the smartest computer scientists and software engineers in the country, I waited for my name to be called. It was December 2, 2019 and SkyTruth was one of four worthy nonprofits chosen to participate in the re:Invent 2019 Nonprofit Hackathon for Good.

What did that mean? It meant SkyTruth was competing for brainpower in a day-long event, hosted by Amazon Web Services (AWS), to address some of the world’s most pressing problems with computer technology. As I listened to our competitors pitch their projects — each hoping to inspire the best and brightest to work on their goals (for free) that day — I reviewed SkyTruth’s major talking points in my head. First and foremost, I told myself, remember what we are trying to accomplish:

Stop oil pollution at sea.

Why? Every year, more than 230 million gallons of oil are intentionally dumped into the ocean. That’s a BP-sized spill that nobody hears about, every year. The culprits? Thousands of ships at sea, mostly big cargo ships and tankers, that are discharging untreated bilge waste. Bilge is the nasty, oily slop that accumulates in a vessel during normal operations. The quickest, cheapest way to get rid of it is to flush it directly into the sea, a process called bilge dumping.

This has been illegal under international maritime law since the 1970s.

And yet it continues to happen with alarming frequency, mostly afflicting countries that can’t afford a robust navy or coast guard to patrol their waters. This stuff is toxic, harmful to marine life, and all too often causes “mystery spills” that wash up on shorelines around the world, along with dead and dying sea birds, turtles and other marine life. Tourists flee, fisheries close, and thousands of people suffer losses to their livelihoods and wellbeing.

For years, SkyTruth has observed this phenomenon via satellite imagery. In 2019 we began a more strategic and systematic search for polluters, focusing on heavily trafficked shipping lanes that are visible from publicly available ship-tracking data. In total, between January and December 2019, our staff and interns found 163 slicks averaging 56 kilometers (almost 35 miles) in length. We were not able to cover all of the Earth’s oceans — we were limited by satellite capabilities and staff time. But the frequency of this activity tells us that it is routine in many places in the world’s oceans, and that the problem is much larger than the incidents we uncovered.

SkyTruth is working to expose just how big a problem bilge dumping is. 

How? By automating the detection of bilge dumping on satellite imagery, allowing us to monitor the entire ocean every day, and identifying the polluters responsible. In 2019 we began using Artificial Intelligence techniques such as machine learning to train computers to recognize the tell-tale slicks that trail behind a polluting vessel. We are building programs to correlate Automatic Identification System (AIS) vessel-tracking radio broadcasts with those slicks to help us identify the polluters in many cases. As we move forward, we will implement a notification system to inform, in near-real time, enforcement agencies, port authorities, supply chain managers, shipping companies, flag state compliance officers, activists, journalists, and anyone else who is in a position to take action.

This was the vision racing through my mind that December morning in Las Vegas.  When my name was called, I stepped onto the stage, smiled at the assembled crowd of techies and made my pitch. It must have worked. After a brief presentation and a little Q and A, SkyTruth attracted seven different teams with a total of 35 computer scientists and engineers to work on different components of our project.

And perhaps most importantly, we gained a valuable new friend: Amazon Web Services awarded SkyTruth an Imagine Grant to start building the toolkit that will help the world end bilge dumping. In addition to direct support, the grant includes generous in-kind donations of cloud computing assets, technical training for our team, hands-on help from AWS engineers, as well as marketing and promotional assistance.

It’s an exciting time at SkyTruth, reminiscent of the early days of our Global Fishing Watch collaboration, which fundamentally changed how the world sees and tracks commercial fishing.

We will be applying many of the same techniques developed to detect bilge dumping — a major part of our Conservation Vision program — to terrestrial conservation challenges in our new partnership with Wildlife Conservation Society (WCS).  SkyTruth will be helping WCS monitor and manage protected areas around the world. We are identifying specific projects to demonstrate how automated analysis of satellite images can assist efforts to detect and stop illegal logging, land clearing, oil and gas drilling, mining, poaching, and other activities that threaten biodiversity and destabilize communities.

To accomplish these amazing feats, we hired a machine learning engineer, Jona Raphael. Jona is advancing our ability to automatically scan thousands of satellite images a day and find the needle in the haystack of data; that is, detect features of interest or concern that might otherwise remain hidden from view.

We are improving our capabilities every day, applying these techniques across a broad spectrum of our work.  Read on to learn more about our activities in 2019 and how we will be building on them for years to come, addressing escalating  issues such as environmental justice and coping with new challenges presented by the global pandemic.

SkyTruth’s machine learning team, working furiously at the re:Invent 2019 Nonprofit Hackathon for Good.

Program Updates

Securing healthy waters and oceans

Bilge dumping: Our newest body of work in oceans focuses on stopping bilge dumping at sea (as described in the President’s Message above). In 2019 our interns and staff visually identified 163 incidents of bilge dumping around the world —  just by viewing satellite images of likely hotspots manually. By the end of the year, we had begun developing the machine learning techniques that will detect bilge dumps around the world with an algorithm that scans thousands of images every day and flags illegal activity automatically. This will increase our ability to detect polluters, alerting law enforcement agencies, concerned citizens, policymakers and journalists to this chronic pollution problem. We believe that as shipping companies are finally held accountable, these incidents — which appear to be routine — will come to a halt. You can learn more about this often overlooked environmental practice in our blog series exploring how and why vessels pollute, and what can be done to stop it.

We continued to work closely with our partners at Global Fishing Watch (which we launched as a separate entity in 2017 with our partners Google and Oceana) to monitor the so-called “dark fleet” — vessels not broadcasting their location, identity and activity via the Automatic Identification System (AIS). Dark fleet vessels could be engaged in illegal or unauthorized activity. As part of this work, we mapped the world’s offshore infrastructure, and were the first in the world to do so. Oil platforms and other structures like wind turbines appear as bright white spots on satellite radar imagery – much the same as large fishing or container vessels. So, we decided to convert this constant frustration into a novel opportunity: we made a new data set of global offshore infrastructure. Our paper describing our new technique, Automating offshore infrastructure extractions using synthetic aperture radar & Google Earth Engine,” was published in the peer reviewed journal Remote Sensing of Environment, and our partners at Global Fishing Watch are using our dataset routinely to help with their work.

Harmful Algal Blooms: In 2019 we also began a new project to map harmful algal blooms (HABs) in freshwater and along coasts. HABs affect every U.S. coastal and Great Lakes state. Normally, algae are harmless — simple plants that form the base of aquatic food webs. But under the right conditions, algae can grow out of control causing toxic blooms that can kill wildlife and sicken people.

We’ve been working with researchers at Kent State University who have developed a technique for detecting cyanobacteria (blue-green algae) and other harmful algae in the western basin of Lake Erie, a known hotspot for harmful algal blooms. Currently, their method uses infrequently collected, moderate resolution satellite imagery to identify harmful blooms and the factors that trigger them. This technology limits the utility of their tool. Additionally, their method is written using proprietary software, making it difficult for others to use. We worked to implement Kent State’s method on the Google Earth Engine platform, making it much easier to generate updates to the analysis, and offering the possibility of automating the update on a regular basis. With greater image availability provided by Earth Engine in the cloud, it should be feasible to create a near real-time algal bloom alerts system. Such a system could be used by water quality managers and public health officials to monitor water quality and reduce threats to wildlife and public health by taking action in the early stages of an evolving algal bloom. 

Other actions:  Our oceans work in 2019 included a few other notable activities. 

In the wake of one of the worst offshore oil spills in Brazil’s history, we posted imagery and analysis of unusual vessel activity off the country’s northeast coast in an effort to help identify the source of this oil spill. 

We used a combination of satellite imagery and vessel tracking data to show systematic GPS manipulation occurring at Chinese oil facilities and other government installations. Regrettably, we weren’t able to determine a cause for the GPS manipulation. But we were able to use vessel tracking data to demonstrate that GPS interference and suspicious vessel activity can be pinpointed very precisely in both time and space.

Algal bloom in Lake Erie. Image courtesy NOAA.

“Working with Brendan on the SkyTruth team was a pleasure for me and my students. SkyTruth learned our complex analytical approach to mapping Harmful Algal Blooms and helped us implement it in the cloud using Google Earth Engine and Python programming. And they provided our team at KSU with intensive hands-on training, so now we can expand our work with new sources of satellite data and apply our analysis anywhere in the world.”

Joe Ortiz
Kent State University

Image provided by NASA/GSFC/METI/ERSDAC/JAROS and U.S./Japan ASTER Science Team, via NASA Earth Observatory, ©2020.

Restoring clean air and stable climate

Global natural gas flaring detected by satellite. 

Natural Gas Flaring: We updated our Annual Flaring Volume map first launched in 2017 to provide site specific estimates of the annual volume of gas flared during oil and gas production worldwide — to include data from 2017–2018. Flaring is the act of burning off excess natural gas from oil wells when it can’t economically be stored and sent elsewhere. Flaring is also used to burn gases that would otherwise present a safety problem. But flaring from oil wells is a significant source of carbon dioxide, emits soot into neighboring communities, and wastes natural gas. SkyTruth’s map allows users to search virtually any geographic area they’re interested in, then easily compare and download flaring volume totals from 2012 through 2018 to observe trends. Residents, researchers, journalists and others concerned about flaring in their city or study area can easily determine the sources of the problem and how much gas has been flared using the latest data available. 

Oil and gas drilling sites: In 2019 we applied machine learning techniques to automate the detection of oil and gas drilling sites. We built a model that could detect oil and gas drilling sites in Pennsylvania with an accuracy of 84%  and low rates of false positive or  false negative results.

We’re continuing to advance our work by reconfiguring our model for image segmentation, a machine learning process that “clips out” the exact boundaries of drilling sites. We’ve also expanded this work geographically, focusing on public lands in the U.S. West. We’ve developed a new training data set of 2,000 oil and gas drilling sites throughout the Colorado River Basin. Our new model is able to extract – or segment – the complete boundaries of each drilling site in an image. This allows us to make direct measurements of the landscape footprint of oil and gas development: a key metric for quantifying the impact of drilling on the intactness and quality of wildlife habitat, and the proximity of drilling-related construction activity to homes, schools and businesses. Our initial segmentation model for oil and gas drilling sites in the Colorado River basin returned 86% accuracy. We then used this model to identify the perimeter and affected area of 5,000 new wellpads. Our next step is to take the models that we’ve built and tested and deploy them in real-world applications that conservation advocates, the media, researchers, and policymakers can use to demonstrate the growth of a fossil fuel industry that continues to threaten global climate.

We are also developing new training data to detect roads associated with drilling activities in Pennsylvania. Although roads are relatively well mapped in the state, recently built private roads that service oil and gas drilling sites remain a glaring omission from the currently available data sets. Mapping these private roads is essential to get a full picture of the footprint of drilling and  fracking in the state, and to understand how much important wildlife habitat has been fragmented by these activities.

Moreover, our work to detect drilling roads in Pennsylvania is serving as the foundation for our work with WCS to monitor protected areas around the world for illegal activities like poaching, illegal logging, or mining. Once our processing pipeline and model are returning consistently accurate results in Pennsylvania, it will be much easier to apply those tools to parts of the world where data on roads are less reliable — or nonexistent.

“Good timing. Super helpful for the New Mexico methane rule work that’s underway.”

Thomas Singer, Senior Policy Advisor
Western Environmental Law Center, commenting on SkyTruth’s global natural gas flaring

Puerto Carreño in Columbia, ©2020 Google Earth View, imagery TerraMetrics.

Safeguarding lands and communities

Updated mountaintop mining data:  We updated our mountaintop removal mining data set to include information from 2016, 2017 and 2018. This update makes our data set, and the accompanying map, the most comprehensive available. The map gives users the ability to quickly generate simple reports on mining activity by county or their area of interest. We continue to make this data available to research partners with active new projects underway, promising to add to the body of research that has already influenced public mining policy and practices. Most recently, scientists at West Virginia University published a study in the peer reviewed International Journal of Environmental Research and Public Health that relied on this dataset to document an association between mining and dementia-related deaths. As under-resourced communities in Appalachia fight against this destructive activity, SkyTruth’s data can support research and analysis that strengthens their arguments and turbo-charges their advocacy. 

Drilling setback app: One of our most exciting achievements this year was developing our first Google Earth Engine web application. Our app focuses on mapping drilling setbacks in Allegheny County, Pennsylvania and allows users to explore the number of homes at risk from oil and gas drilling emissions under different setback scenarios. To publicize the release of the app, we hosted a webinar demonstrating how to use it. The webinar was attended by conservation advocates, members of the media, and researchers examining the issue of fracking in Pennsylvania. 

Upgrading SkyTruth Alerts: One of our goals for 2018–2019 was to rebuild and re-launch our SkyTruth Alerts system. SkyTruth Alerts show people where environmental incidents have recently occurred, particularly for the oil and gas industry. Any user can monitor areas they care about, and get notified about environmental incidents in those areas. This tool helps level the playing field for citizen  activists in vulnerable communities who are fighting ongoing pollution in their backyards.

We created a new landing page for Alerts that makes it easier for people to get started using the platform, and we’ve added documentation — such as how-to guides — about how our alerts are generated. We’ve also added a new “Timeline” feature that lets users see where alerts have occurred over time so that they can start to identify hotspots. The new landing page and how-to guides have been helpful for users navigating the data sets that we’ve added. Over the course of the year we added oil and gas drilling permit alerts for Colorado, Wyoming, New Mexico, Montana, and Utah. We’ve also added pollution reports from Florida and New Mexico, and additional data sets on methane gas flaring, pipeline routes, and other layers for added context for users.

SkyTruth Alerts now gives users the ability to browse recent and historical satellite imagery.

“I want to thank you both for helping me learn to use the various programs to download and analyze the Sentinel-3 aerosol optical depth data. I worked through dozens of images to find the one below. It demonstrates how aerosols accumulate during an inversion. A picture is worth a thousand words and I presented this to the WVDEP Advocate last week. It seemed to ‘knock his socks off’ …”

Mike Glenn, retired plant physiologist
USDA

Aksu in China, 2020 Google Earth View, imagery Maxar Technologies.

Protecting a legacy of biodiversity

Habitat Change Detection: We have been working with Defenders of Wildlife for two years to automate the detection of land cover changes. We now have automated change detection running as a new “beta” feature in our SkyTruth Alerts platform. Working with Defenders and Upstream Tech, a public benefit corporation that harnesses software and technology for conservation and water, we explored options to introduce the automated change detection algorithm to the U.S. Fish and Wildlife Service and encourage them to integrate this work with the existing tools used by the Service. If adopted, this will mean Service biologists responsible for managing endangered plants and wildlife will have better information about habitat changes.

Conservation Vision:  Our advances in machine learning to detect bilge dumping and oil and gas wells all contribute to our Conservation Vision program — a long-term initiative to apply Artificial Intelligence techniques such as machine learning to a broad range of conservation problems.  We anticipate countless opportunities to apply the skills we are developing now, to conservation problems in the future. As we identify needs on the ground with our new partners at WCS, we will be applying these cutting edge techniques and creating new solutions to seemingly intractable challenges, such as poaching, habitat fragmentation, and more. Stay tuned!

Using satellites to track down polluters at sea.

“Please keep up the fine work on this mystery. You are providing a service.”

Christopher Reddy, Senior Scientist
Woods Hole Oceanographic Institution, responding to our work on a major oil spill in Brazil

Irawaddy Delta in Myanmar. ©2020 USGS Earthshots.

2019 Board of Directors

It’s no small role that SkyTruth’s board of directors plays in ensuring we continue to carry out our mission. We are fortunate to have assembled a group of highly qualified individuals from across the country and many sectors.

John Amos
Walter Ailes
Mary Anne Hitt
Catherine Irwin
Darshan Karwat
Sue Kemnitzer
Monica Medina
Barbara Morgan
Stephen J. Paradis
Alan Septoff
David Shearer
Henry “Hank” Willard
Paul Woods

2019 Funders

SkyTruth gratefully acknowledges the generous support of our funders, including individuals from around the world. Your support makes it possible to illuminate human impact on the planet and inspire conservation solutions through technology. Thank you!

Anonymous
Catena Foundation
Clif Bar Family Foundation

Cornell Douglas Foundation
Google
Herbert W. Hoover Foundation/Kent State University
Patagonia
The Keith Campbell Foundation
The Nora Roberts Foundation
The Tilia Fund
Wallace Genetic Foundation
Walmart Foundation

Financials

SkyTruth in the News

SkyTruth uses the view from space to inspire people to protect the environment. We are a 501(c)(3) nonprofit organization headquartered in Shepherdstown, West Virginia, with offices across the United States and in Indonesia. We believe more transparency leads to better management and better outcomes. By sharing our findings – stunning imagery and robust science-based data – with the public for free, we move policy makers, governments and corporations towards more responsible behavior and more accountability for the environment.