Cambridge researchers are leveraging the capabilities of drones, lasers, and AI to try to transform the speed and accuracy of assessing biodiversity and carbon in forests.

Drones deployed near ground level equipped with light-detection lasers – usually used on satellites – can create 3D “point cloud” models of forest structures, said Emily Lines, university associate professor in physical geography at the University of Cambridge.

The Cambridge team aims to publish the research free-to-access online to make the method more accessible, while improving assessments of the value of different ecosystems, Lines told Carbon Pulse.

The resulting data is much more detailed than information from satellites, and can save the large amount of time it takes to manually collect the information on the ground, she said.

The 3D structures can measure the exact amount of biomass in a tree to within more than 95% accuracy, she said. However, previously it had been laborious to understand the models.

When translated using AI, the drone-gathered data can reveal metrics usable by ecologists about tree trunks, branches, twigs, and leaves.

A project that a few years ago would have taken months, is now taking a couple of days with the help of AI.

“Tackling these bottlenecks in processing, that’s where the AI approaches have been really promising for us – showing a lot of potential and power.”

Although this kind of technology has been more expensive historically, it is becoming cheaper – with relevant drones now available for around £300 ($385), she said.

A form of the method is already being used at national level. Some Eastern European governments collect 3D models via backpack-mounted sensors for forest surveys, she said.

On the private side, some companies use the technology to measure the biomass of forests to inform carbon credit programmes.

If the researchers can improve the method, then using the technology for biodiversity credits seems like an “obvious extension”.

The project is one of several at University of Cambridge focusing on AI for climate and nature.

LIMITATIONS

Lines acknowledges the technology has limitations, such as its inability to measure animal species, and for some metrics there is no substitute yet for human-gathered data.

“We’re basically measuring proxies here. We’re not detecting individual butterflies, but we think that the structure of the forest in that level of detail should provide much better proxies,” said Lines.

“There’s plenty of things that we can’t measure with these. We’re basically just measuring the structure of things, where the material is within a forest. These technologies give us better proxies that better relate to the habitat experience.”

Indeed, Enda Keane, CEO of Irish data company Treemetrics, said claims made by data companies about the ability to gather accurate AI-powered data remotely via satellites is “complete and utter hype, it’s very frustrating”.

“All the science will tell you that you have to have ground data to predict, and you can’t cut those corners,” he told Carbon Pulse.

Data from AI based on satellite images can be wrong by up to 30%, he claimed.

AI can make a process more efficient, less labour intensive, but is often no substitute for on-the-ground data collection, he said.

“It’s more about the person on the ground, how much effort they put in.” The key is to measure the diameter of many trees throughout the forest, he said.

In June, Treemetrics announced a partnership with the climate risk arm of French insurer Axa Group on natural disaster forestry insurance.

By Thomas Cox – t.cox@carbon-pulse.com

*** Click here to sign up to our twice-weekly biodiversity newsletter ***