Ghost roads threaten Asian rainforest

When roads come, there is destruction.

The majority of the unmapped and unseen land was not mapped or seen until now

AI could speed up detection, a growing hope

By Marianne Bray

HONG KONG (Thomson Reuters Foundation), June 13, 2008 - Researchers say that newly discovered ghost road pose a threat unprecedented to the world’s dwindling forests.

They also found that AI could map secret roads and help humans avoid some of the worst dangers.

Where roads go, people will follow, and soon, there will be mass destruction. As loggers and poachers invade an already fragile environment, millions of trees are cut down, and people, birds, and wildlife flee.

The Thomson Reuters Foundation was told by Daniel Carillo of the Rainforest Action Network in San Francisco that "road development is often the initial fatal step" to forest destruction.

Researchers spent 7,000 hours combing through millions of high-resolution satellite images to discover a maze-like network of ghost roads that winded their way across Southeast Asia and Melanesia.

They hope that AI will help them track deforestation more accurately in the future, and thus be able to stop illegal activities with greater speed.

William Laurance told the Thomson Reuters Foundation that "there's no doubt AI could be a crucial arrow in forest conservationists' quiver to help map roads, and slow down forest destruction."

If they don't have a clear understanding of where these activities are taking place, developing nations cannot stop illegal roads, mining, logging and poaching.

Laurance's Team - whose findings appeared in Nature magazine in the month of April - stated that the secret network was one of the gravest direct threats to tropical forests around the world, which are already under threat from logging and mining, as well as agriculture, climate change, and agricultural practices.

Laurance said, "We call these roads 'ghost roads', because they are hidden and the destruction of the forest they cause is often out of control. As ecologists, this scares us to death."

Researchers in Nature magazine said that while planned roads help trade and flow vital resources, their unregulated cousins can "unleash Pandora's Box of environmental ills" and social challenges.

Researchers and volunteers from the 210-strong study checked images of 1,42 million plots to map forests in Indonesia, Malaysia and New Guinea.

They found that in this small part of the globe, there are nearly one million kilometers of roads that have not been mapped on OpenStreetMap and 1,16 million more km than the Global Roads Inventory Project.

They found enough ghost roads to circle the globe 23-29 times.

AI HELPS HUMANS

Researchers said that deforestation usually increases shortly after roads are constructed. This pattern is common on three of the largest continent islands in the world: Borneo Sumatra, and New Guinea.

Carillo said that when roads are cut into forests, they disrupt important ecosystems and wildlife's habitat.

They suddenly give access to poachers or loggers who want to exploit the forest that was previously inaccessible. This can be disastrous.

The money that can be made is what makes the road so appealing.

Around 40% of roads in Borneo and Sumatra, and New Guinea, run through plantations for palm oil, wood pulp, and rubber trees. The remaining roads run through intact forests and agricultural lands.

A Nature article from 2019 shows that the destruction of tropical forests has a large impact on climate change, as they account for 68% of carbon in the world.

NASA's study found that these trees absorb more carbon dioxide than other forest types, but they also release more of it when they are removed.

The team stated that better monitoring is essential to keep up with the rapid pace of road construction and that AI can be used to map the changing landscape.

Researchers from the team tested how well AI worked by training three machine-learning algorithms to identify uncharted roads in satellite images of rural areas and semi-forested regions in equatorial Asia-Pacific.

The AI models they used were correct up to 81% the time.

Manual mapping is laborious, even though human accuracy can reach 90%.

Jayden Engert is the lead author of the Ghost Road Study. He says that human research can take years to cover a small area.

Engert said, "Our data will be usable by the time we are able to map accurately using the imagery that we have." "Roads continue to be built, so it is impossible to update data manually.

Laurance said that ecologists would benefit from a program which would automatically take snapshots of vast areas.

Laurance said, "We needed AI desperately because we spent 7,000 hours mapping a tiny piece of the globe."

We estimated that it would take 73 years to map all the roads on Earth, or 640,000 hours.

Carlos Souza is a researcher with the non-profit conservation organization Imazon. He has already been working on AI detection in Amazon.

Laurance says that this kind of monitoring in conjunction with credible law enforcement is the key to stopping new road construction. She cites the success of human AI efforts in Brazil.

He said, "They found they only needed to fine or jail just a few offenders and then everyone else falls into line." (Editing done by Lyndsay Griffiths. Thomson Reuters Foundation, the charitable arm Thomson Reuters. Visit https://www.context.news/))