How AI is improving the monitoring of nature and land

Regulations such as the EU’s Nature Restoration Regulation are introducing new requirements not only to restore ecosystems, but also to monitor the development of nature more effectively. At present, this is done through field surveys, where researchers go out into the countryside to observe biodiversity and how land is grazed. 

Covering such large areas in this way and gaining a comprehensive overview of the situation is both a difficult and time-consuming task. 

Satellite imagery and AI complement fieldwork and make it more accurate

Satellite imagery makes it possible to cover large areas, and AI helps to analyse the information and identify where a closer look is needed.

“With this type of AI application, we can scale up and make parts of the work more time-efficient. Instead of always sending people out, we can monitor locations from space over time,” says Aleksis Pirinen, AI researcher at RISE.

The amount of information that can be extracted depends on the resolution of the images. RISE, in collaboration with Arla, has used drone images from farms to investigate what can be identified in satellite images and when higher-resolution images are required.

Through the EU’s Copernicus Earth observation programme, vast amounts of satellite data are freely available. However, having access to the images is not the same as being able to interpret them. 

“The AI model can be trained to become an expert in its specific task. It can do the job faster and more accurately than humans,” says Aleksis Pirinen.

AI models that can be used anywhere on Earth

But that does not mean that humans will become redundant. Fieldwork remains important for investigating things that satellites cannot see. It is also needed to train and verify AI models.

“In order for us to get an AI model to recognise nature restoration from satellite images, we need to provide the model with a reference set that shows it is getting it right. And that reference set can only be obtained at ground level, by having a person there to check how things are going,” says Aleksis Pirinen.

AI can also make fieldwork more accurate. Instead of relying on random sampling, efforts can be targeted at areas where the analysis indicates that something appears to be out of the ordinary.

At the same time, AI models may need to be adapted to local conditions. Weather, cloud cover and other factors affect the appearance of satellite images.

"If a model has been trained on images from another part of the world, it may need to be adapted and retrained to perform well on Swedish images. At the same time, more general AI methods are being developed that can be used successfully almost anywhere on Earth."

Among other things, RISE has been working with SMHI on AI models to recognise the types of clouds found in Scandinavia. In addition, RISE runs Digital Earth Sweden, a platform where public authorities and other stakeholders can access satellite data, which is being developed with Swedish conditions in mind.

The technology can provide a basis for urban planning

The combination of AI and satellite data can be used for much more than just nature restoration. RISE is working with various stakeholders on green infrastructure in cities and methods for monitoring changes to coastlines.

For local authorities, this technology can provide a basis for urban planning.

"Is this car park used as much as we think? To answer that, we can monitor it over time and provide the local authority with data to help them decide whether the site should remain a car park or be used for something else."

"That sort of interdisciplinary work is our strength at RISE. We have our own experts in a wide range of fields, which in this case can range from biodiversity and pollinators to AI and geology – all under one roof. Furthermore, we often act as a convener, bringing together different organisations, companies and public authorities to collaborate. That is what is needed to successfully complete such complex projects and work towards long-term change,” says Aleksis Pirinen.