For the past For three decades, geologist Carlos Sousa has worked for the Brazilian nonprofit Imazon, researching how he and his teams can use applied science to protect the Amazon rainforest. For most of that time, satellite imagery made up the bulk of his work.
In the early 2000s, Souza and his colleagues came to the realization that 90 percent deforestation occurs within 5 km of the newly created roads. While satellites have been able to track road expansion for a long time, the old way of doing things required people to label those results by hand, accumulating what eventually became training data. Those years of work paid off last fall with the release of an AI system that Imazon says detects 13 times more roads than the previous method, with 70 to 90 percent accuracy.
The proponents of satellite imagery and machine learning have ambitious plans for solving big problems at scale. Technology can play a role in anti-poverty campaigns, protect the environment, help billions of people. get street addressesand increase crop yields in the face of increasing climate change. A UNESCO report published this spring highlights the highlights 100 AI models with the potential to change the world for the better. But despite recent advances in deep learning and satellite image quality, and a record number of satellites expected to enter orbit over the next few years, ambitious attempts to use AI to solve big problems still stumble upon traditional barriers such as government bureaucracy or lack of political will or resources.
For example, stopping deforestation requires more than finding the problem from space. Brazilian federal government program helped reduce deforestation from 2004 to 2012 by 80 percent over previous years, but then federal support waned. In line with a campaign promise, President Jair Bolsonaro relaxed enforcement and called for the rainforest to be opened up to industry and livestock settlers. As a result of deforestation in the Amazon reached the highest levels seen for over ten years.
Other AI conservation groups have faced similar challenges. Global Fishing Watch uses machine learning models to determine vessels that turn off GPS systems to avoid detection; they can predict the type of ship, the types of fishing gear, and the direction of its movement. Ideally, this information helps authorities around the world detect illegal fishing and inform decisions to board vessels for inspection at sea, but controlling large swathes of the ocean is difficult. Global Fishing Watch specialists have recorded hundreds of boats involved in illegal squid fishing in 2020, data that head of research David Krodsma attributes to growing cooperation between China and South Korea, but this has not resulted in any specific prosecution. Port enforcement is “the key to making deterrence scalable and affordable,” he said.
Back on land, consulting firm Capgemini is working with non-profit environmental group The Nature Conservancy to track trails in the Mojave Desert and protect endangered animal habitats from human activity. As part of a pilot program last year, the initiative mapped trails created by SUVs on hundreds of square miles of satellite imagery in Clark County, Nevada, to create an AI model that can automatically identify newly created roads. Based on this work, The Nature Conservancy intends to expand the project to control the entire desert, which spans over 47,000 square miles across four US states.
However, as with Amazon, identifying problem areas only gives you definite results if there aren’t enough resources to act on those findings. The Nature Conservancy is using its artificial intelligence model to inform land managers of potential threats to wildlife or biodiversity. The Mojave Desert Conservancy is overseen by the US Bureau of Land Management, which has only about 270 rangers and special agents on duty.
In northern Europe, Iceye has begun tracking ice buildup in the waters off Finland using microsatellites and machine learning. But in the past two years, the company has begun predicting flood damage using microwave imaging, which allows you to see through clouds at any time of the day. According to Iceye vice president of analytics Shay Strong, the biggest challenge right now is not spacecraft development, data processing, or improving machine learning models that have become commonplace. It deals with institutions stuck on age-old ways of doing things.
“We can more or less understand where things are going to happen, we can get images, we can do analysis. But right now, our biggest problem is still dealing with insurance companies or governments,” she says.
“This is the next step in local coordination and implementation that is needed to take action,” says Hamed Alemohammad, chief data scientist at the nonprofit Radiant Earth Foundation, which uses satellite imagery to achieve sustainable development goals such as eradicating poverty and hunger. . “This is where I think the industry should put more focus and effort. It’s not just about a fancy blog post and a deep learning model.”
Often it is not only about attracting politicians. AT analysis 2020, a group of academic, government and industry researchers highlighted the fact that the African continent contains most of the world’s uncultivated arable land and is expected to account for a significant portion of global population growth in the coming decades. Satellite imagery and machine learning can reduce dependence on food imports and turn Africa into the breadbasket of the world. But lasting change, they say, will require a build-up of professional talent with technical expertise and government support so that Africans can build technology to meet the needs of the continent rather than importing solutions from elsewhere. “The path from satellite imagery to public policy decisions is not easy,” they wrote.
Labali Toure is co-author of this paper and Head of the Geospatial Faculty at the University of Agriculture in Senegal. In this capacity, and as the founder of Geomatica, a company providing automated satellite imagery for farmers in West Africa, he has seen how satellite imagery and machine learning help decision makers understand how salt flow can affect irrigation and affect crop yields. He also saw that it helped with questions about how long a family was on the farm and helped with land management issues.
Sometimes free satellite imagery from services such as NASA’s LandSat or the European Space Agency’s Sentinel programs is sufficient, but some projects require high-resolution photographs from commercial vendors and cost can be an issue.
“If decision makers know [the value] it can be easy, but if they don’t know, it’s not always easy,” Toure said.
Back in Brazil, in the absence of federal support, Imazon is now building ties with more politicians at the state level. “There is currently no evidence that the federal government will lead efforts to conserve or deforest the Amazon,” says Souza. In October 2022, Imazon signed cooperation agreements with prosecutors to gather evidence of environmental crimes in four Brazilian states on the border of the Amazon rainforest to share information that can help prioritize law enforcement resources.
When you go after people who cut down protected areas, the damage has already been done. Now Imazon wants to use AI to stop deforestation before it happens, intertwining that road detection model with a model designed to predict which Amazon-bordering communities are most at risk of deforestation over the next year.
Deforestation continued in historical courses in early 2022, but Sousa hopes that through work with non-profit partners, Imazon can expand its deforestation AI to the other seven countries in South America that touch the Amazon rainforest.
And this fall, Brazil will hold presidential elections. The current leader in the polls, former President Luiz Inácio Lula da Silva, is expected to strengthen law enforcement weakened by Bolsonaro and restore the Amazon Fund for Foreign Investment in Reforestation. Lula’s environmental plan isn’t expected for months, but environment ministers from his previous term predict it will make reforestation the cornerstone of its platform.
Credit: www.wired.com /