If you’ve got as little as 30 seconds and a decent internet connection, you can help combat the deforestation of the Amazon.
“You don’t have to be a climate scientist, you don’t have to be a data scientist, you just have to be a citizen that is concerned about the issue of deforestation,” said Elliot Inman, a researcher at systems analysis company SAS, which created an app that uses humans’ sharp eyes to teach computers what deforestation looks like.
Some 15% of the Amazon, the world’s largest rainforest and a crucial carbon repository, has been cut or burned down. Around two-thirds of the Amazon lie within Brazil’s borders, where almost 157 square miles of forest were cleared in April alone. In addition to storing billions of tons of carbon, the Amazon is home to tens of millions of people and some 10% of the Earth’s biodiversity.
Scientists are warning of an impending tipping point beyond which the Amazon will be irreparably transformed.
To stem the destruction, it is necessary to track where and when it’s happening. To that end, SAS formed a nonprofit partnership with the International Institute for Applied Systems Analysis, or IIASA, a research institution based in Austria, to build an artificial intelligence model that can recognize signs of deforestation. That data can be used to alert governments and conservation organizations where intervention is needed and to inform policies that protect vital ecosystems. It may even one day predict where deforestation is likely to happen next.
“Forests are by definition remote and far away and vast,” explained Fred Stolle, deputy director of forest programs at World Research Institute, who was not involved in the SAS project. To monitor deforestation, conservation organizations need an eye in the sky.
WRI’s Global Forest Watch online tracking system receives images of the world’s forests taken every few days by NASA satellites. A simple computer algorithm scans the images, flagging instances where before there were trees and now there are not. But slight disturbances, such as clouds, can trip up the computer, so experts are increasingly interested in using artificial intelligence.
To train its computer vision model to reliably recognize tree loss, SAS and IIASA turned to crowdsourcing. Visitors to the program website flip through pictures of the Amazon taken by European Space Agency satellites hundreds of miles above the Earth, looking for clues.
The interface is a lot like those reCAPTCHA boxes that ask you to prove you’re “not a robot” by clicking on the parts of a picture that contain a truck or a stoplight. Here, the pictures show 31 square miles of forest divided into nine sections. Users simply select the squares in which they’ve spotted some indication of human impact: the tell-tale quilt of farm plots, a highway, a suspiciously straight edge of tree line.
IIASA has run many citizen science projects to monitor deforestation around the globe. The goal here is to use human-generated data to automate the process.
“The task that we’re asking them to do is very simple: Look at these nine blocks and make a judgment about each one. Does that satellite image look like a situation where human beings have transformed the landscape in some way?” Inman explained.
“We’ve been gathering data from human beings making these judgments and building an artificial intelligence model that will attempt to make judgments in the same way,” he added.
It’s not always easy; that’s the point. For example, a brown patch in the trees could be the result of burning to clear land for agriculture (earning a check mark for human impact), or it could be the result of a natural forest fire (no check mark). Keen users might be able to spot subtle signs of intervention the computer would miss, like the thin yellow line of a dirt road running through the clearing.
The app serves up confounding images to more and more users until a clear answer emerges. This is important, said Stolle. “If the data is inconsistent you can imagine that the algorithm gets inconsistent.” In his experience, citizen scientists do much better after a couple of days of training on what to look for. SAS’s website offers a handful of examples comparing natural forest features and manmade changes.
If users can’t reach a consensus as to whether or not human impact is visible in a given image, SAS sends it to experts for analysis.
Since the app was launched on Earth Day in April, users have analyzed almost 41,000 images, covering an area of rainforest nearly the size of the state of Montana. Deforestation caused by human activity is evident in almost 2 in 5 photos.
The SAS/IIASA team will add new images of the Amazon in Brazil, and expand west to Peru and south into Bolivia. The researchers hope to use historical images of these new geographies to create a predictive model that could identify areas most at risk of future deforestation. If they can show that their AI model is successful, it could be useful for NGOs, governments and forest monitoring bodies, enabling them to carefully track forest changes and respond by sending park rangers and conservation teams to threatened areas. In the meantime, it’s a great educational tool for the citizen scientists who use the app.
SAS doesn’t track individual users, but judging by the data they’ve generated, there are a lot. Inman was surprised how willing people have been to spend their time clicking on abstract-looking pictures of the Amazon. The app — while sleek and easy to use — is not what you’d call exciting or especially motivating. There are no prizes, no leaderboard, no follow-up emails encouraging you to come back tomorrow. You don’t even get to know if you were “right” — just a pop-up saying, “Thanks for your help so far!”
What’s more, SAS and IIASA released the app as COVID-19 cases were surging worldwide. Inman worried about being insensitive to or distracting from the growing crisis.
“Despite all of that, we have still had people from 80 different countries come onto the app and make literally hundreds of judgments that enabled us to resolve 40,000 images, which is hundreds of thousands of square kilometers of space,” he said. “What is the number one thing that I learned from this? We can depend on citizen scientists to help fight climate change.”
HuffPost’s “Work In Progress” series focuses on the impact of business on society and the environment and is funded by Porticus. It is part of the “This New World” series. All content is editorially independent, with no influence or input from Porticus. If you have an idea or tip for the editorial series, send an email to firstname.lastname@example.org.