How satellites, drones and artificial intelligence could help reforestation

How satellites, drones and artificial intelligence could help reforestation

This morning I spoke with Diego Saez Gil, the co-founder and CEO of Pachama – a Y Combinator graduate who brings machine learning to the delicate business of reforestation. In short, Pachama ensures that when a company says it offsets emissions by planting trees, the planting is actually done.

“We use remote sensing, including satellite imagery and lidar data, to monitor and verify the claims of forest carbon projects,” says Saez Gil, adding that the platform aims to bring more transparency to the carbon offset market – which is increasingly trading trees.

Just last week, the World Economic Forum urged leaders to support its 1 trillion trees – a platform that hopes to unite the world’s various reforestation campaigns, including the similar-sounding Trillion Trees initiative, which aims to plant one trillion trees.

Critics have warned that the PR push for forests distracts from the larger and more difficult challenge of actually reducing carbon emissions, and I am inclined to agree. Tree-growth is also an arduous process, so that the promised compensations could only occur in the future, and even then could be wiped out by wildfire, drought or other disaster. But there is another problem.

“Today there is more demand than certified supply for planting trees as carbon offsetting,” says Saez Gil. But that doesn’t mean that there isn’t enough potential land to plant a trillion trees. According to ETH Zurich researchers, there are 0.9 billion hectares of land worldwide suitable for reforestation, on which 1.2 trillion trees could be planted and 205 billion tonnes of carbon could be stored.

According to Saez Gil, the problem lies in the certification process, which assesses how much carbon a proposed reforestation project would capture. The current process, which relies heavily on manual controls, takes years. Pachama is trying to speed up the system.

“We are training algorithms for depth learning using lidar data – collected during overflights by drones or an aircraft -, field plots – data collected on the ground by forestry services – and also satellite images. The algorithms begin to learn that a certain combination of colours, shapes and tree species contains a certain amount of carbon, and eventually we can start making predictions about carbon absorption rates,” says Saez Gil.

Once Pachama has certified a plot of land, it can be sold on the carbon market, and then the growth of the plot can be tracked through the start-up’s user dashboard. So if a tree falls, someone knows about it.

 

Microsoft, which promised last week to reduce carbon dioxide emissions by 2030, has teamed up with Pachama to achieve this goal, and if Pachama’s system is scalable, it could help hundreds of others plant trees conscientiously. But Saez Gil readily admits that reforestation is “not a silver bullet”.

“We must reduce emissions and get away from fossil fuels. We can’t just keep going just because we keep planting trees,” says Saez Gil.

***

Another carbon sequestration start-up that caught my eye this week is Texas-based Hypergiant Industries, which is experimenting with algae blooms as a means of removing CO2 from the environment. According to CEO Ben Lamm, the company’s Eos bioreactor – an AI-monitored, refrigerator-sized unit that houses an algae bloom – is 400 times more efficient at absorbing carbon than trees.

The product, which is currently in the beta phase, is connected to the exhaust pipe of the heating, ventilation and air conditioning (HVAC) system of an industrial plant, where the algae can feed on the CO2.

“We are dealing with a ticking clock, and unfortunately the trees are not growing fast enough to quickly reduce the growing atmospheric carbon content… In the context of carbon sequestration and combating climate change, there are more efficient and faster alternatives that should be given priority,” says Lamm.

Leave a Reply

Your email address will not be published.