This article originally appeared in the online magazine iBestuur on 18 juni 2024.
"Carto" used to be the name of the "Spheer App".
The Netherlands is exceptionally far ahead in developing artificial intelligence that uses aerial imagery to monitor nature reserves. The northern provinces did the groundwork; the technology is now ready for nationwide use.
Purple moor-grass (pijpestrootje) and wavy hair-grass (bochtige smele) thrive on nitrogen-rich soil. So where these plants grow abundantly, there is a lot of nitrogen in the ground. Instead of pulling on boots and mapping these plants in the field, AI models are trained to do this using satellite images, aerial photographs, and weather data. This “remote sensing” provides up-to-date information faster about the presence of certain vegetation types and the condition of nature than is possible with fieldwork.
Groundbreaking application
The provinces of Fryslân, Groningen, Drenthe, and Noord-Brabant are working with the Dutch company Spheer.ai on nature monitoring using technology that is also used by generative AI such as ChatGPT and Midjourney. On Carto, Spheer.ai’s own platform, data is prepared so that AI models can be trained on it. Spheer.ai can do this faster than anyone else and needs less data. Their model can also work with older images, unlocking historical information, which in turn is input for calculating forecasts.
Innovator Gerard Kema of the province of Friesland is hugely enthusiastic about the collaboration and the results so far. “What’s happening here is groundbreaking; I haven’t seen anything like this anywhere in Europe.” The collaboration arose from the question of whether AI is suitable to support the mandatory twelve-yearly Natura 2000 monitoring. The answer is an emphatic “yes.” However, it does affect how people work, because nature monitoring still involves a lot of manual work.
“What’s happening here is groundbreaking; I haven’t seen anything like this anywhere in Europe.”
Henk Pieter Sterk is a policy researcher for Nature Monitoring at the province of Friesland and trained in the traditional way. “That means going into the field, observing, counting, and putting dots on a map. Then we apply protocols to that. I still work that way, but I combine it with new techniques. I call it blended monitoring.” According to Sterk, provinces are ready to deploy AI in this area in a larger collaborative framework. He says: “We hear from administrators that they want forecasts and faster results. Researchers tell us they’re itching for tools to make that happen. That has never been the case before. And now we can give both groups what they want.”
“Let’s join forces and collaborate between provinces.”
Removing bias
Another important advantage of the AI version of monitoring is that human bias has been removed from the data. Sterk: “A recent study showed that three different field researchers produced three different map results, because they looked in three different ways. Now we can offer administrators consistent data on which to base policy.” Kema adds that this new way of working also produces up-to-date data much faster, and that adjusting policy and/or retraining AI models can therefore also happen much faster. “This way we get a much better view of our nature.”
Business model
From IPO, Babette Bakker is the coordinator of the Dutch Societal Innovation Hub. “The Netherlands has many different regional initiatives,” she says. “Let’s join forces and collaborate between provinces. We are already holding talks about bringing together the different initiatives and budgets. If there is one ecosystem in which all relevant parties work together, we can also achieve more.”
Besides the provinces, these include RVO, IPO, Staatsbosbeheer, BIJ12, and the Dutch AI Coalition. Bakker sees many more possibilities for this technology, also for “BV Nederland” (the Dutch economy). “We work together with market parties in software, drones, robots, and other hardware. That’s a revenue model. This technology can be deployed for much more monitoring than just the mandatory Natura 2000 areas.”
“With this technology we get more facts about the state and development of nature.”
“We are much better able to monitor, manage, and preserve nature. We have proven that the technology works. Now the task is to focus on strategic collaboration and development. The time is ripe.”
At the table
To move the conversation forward in the right way, different kinds of people need to be at the table: policy and implementation, but also people who understand finance, methodologies and standards, and the latest developments in innovation. That takes time. Kema says: “Nature change is a long-term issue. We’re not talking about a little project that has to be finished within a year. Innovative solutions simply need time.”
Bakker adds: “With this technology we get more facts about the state and development of nature. That requires not only technological innovation, but also new ways of collaborating and of policy- and decision-making based on this data.”
Nationwide deployment
Kema believes the technology is ready for nationwide deployment for producing habitat and vegetation maps. “We are much better able to monitor, manage, and preserve nature. We have proven that the technology works. Now the task is to focus on strategic collaboration and development. The time is ripe for it.”
