AI and biodiversity and nature

Nature action is climbing up the boardroom agenda. Yet significant data gaps remain, and sustainability professionals are increasingly turning to AI to fill them.

Global research has highlighted that nature loss could cost eight industries up to $430bn per year, with food production, retail and forestry among the hardest hit. The pressure to act is growing.

Companies are under increasing scrutiny to measure and report on how their operations affect biodiversity. Investors are asking ever more questions, and 190+ nations have promised mandatory biodiversity reporting for corporates by 2030.

AI can process large amounts of information that would otherwise be impossible to collect manually, helping sustainability teams make data-driven decisions. But alongside these opportunities come questions about the environmental cost of the infrastructure supporting AI, from energy-intensive data centres to water use and emissions.

edie spoke with leading institutions and sustainability professionals about how they are using AI to support conservation, agriculture and habitat management, while carefully considering its environmental impact. These include the Royal Veterinary College, Vodafone and the UK National Parks, and M&S’s dairy farms.

The Royal Veterinary College (RVC) has partnered with biodiversity monitoring start-up Chirrup.AI to survey wildlife across its 200-hectare estate.

Alongside its veterinary teaching, the college manages farmland, ancient woodland and ponds, and has long sought ways to track biodiversity change. With limited staff capacity, carrying out surveys across such varied land has been difficult.

“It’s simply impossible for one sustainability manager to cover that much ground without technology like this,” says Rachel Ward, environmental sustainability manager at RVC. “The sensors are out for three weeks, recording constantly. We don’t have the staff to do that kind of monitoring, nor the time to analyse all that data.”

Chirrup.AI has developed weatherproof acoustic recorders. The devices capture birdsong and convert it into spectrograms, which are analysed by an AI model trained to identify up to 100 bird species. The results are compiled into reports showing which species are present, with benchmarking against similar locations.

All data collected by the devices is reviewed by qualified ornithologists, ensuring a human-in-the-loop element remains in the monitoring process.

Chirrup.AI’s founder Conrad Young describes birds as “strong bioindicators”, as woodland bird species indicate habitat diversity.

However, both RVC and Chirrup recognise that the technology has limitations. The AI can misidentify species or miss unusual ones, and a sample of results is checked by ecologists each year. Young says the focus is on building baselines that can be compared over time, rather than seeking perfect data. Ward notes that for RVC, the alternative would be not collecting the information at all.

Environmental considerations are also part of the partnership. Chirrup reuses devices several times a year, has moved assembly to the UK to improve recyclability, and runs its processing on renewable-powered servers.

At RVC, questions remain about how to account for the emissions of AI within sustainability reporting, but Ward says the overall benefits outweigh the footprint in this case. She also notes that clearer regulation could help organisations account for the emissions of AI more consistently, particularly within Scope 3 reporting, where data gaps remain.

The partnership has a teaching dimension as well. Biodiversity data will be used in student projects, ranging from analysing which bird species are present on the estate to exploring how this information links to land management decisions.

Ward says this could extend to practical outcomes, such as identifying natural pest control and reducing chemical use in fields. “With AI, we’ll get long-term value in terms of student projects, biodiversity insights and better land management,” she explains.

On a larger scale, AI is being applied across all 15 UK National Parks, where Vodafone and conservation teams are combining machine learning with aerial imagery to map habitats, monitor biodiversity and guide restoration work.

David Alexander, research lead at the Landscape Observatory and Peak District National Park, explains why.

“National Parks are huge and varied,” he says. “There are rush pastures, small woodlands, ponds—all scattered across the landscape. Traditional surveys take years. AI lets us look across entire areas, spot patterns and identify habitats much faster. It doesn’t replace ecologists, but it gives us another layer of insight to decide where to focus our efforts.”

The project combines aerial imagery with machine learning to detect subtle features in the landscape.

Alexander explains: “A convolutional neural network can see patterns the way a human does, in terms of associations of colour, texture, shape and edges.

“Traditional methods just look at single pixels or each point in isolation. With AI, we can process large areas quickly and accurately, even across parks as different as Dartmoor and Northumberland.”

Vodafone provides the network infrastructure, sensors and anonymised mobility data to support the research. Sophia Rose, sustainable business executive at Vodafone UK, notes, “This isn’t about surveillance. The data is aggregated to protect privacy. What it does is help park managers understand habitats, visitor movement and ecological changes in real time.”

Alexander emphasises that the insights are meant to inform practical action on the ground.

He says: “For example, if we map rare habitats like chalk grasslands, we can see where they’re fragmented and decide where to focus restoration efforts. It helps us allocate dwindling resources efficiently and intervene where it will have the greatest impact.”

The project is research-led, currently run through Cranfield University, with Vodafone supporting with funding and technical expertise.

Alexander stresses sustainability: “We want to use AI without creating huge carbon costs. Our approach processes data efficiently, and the insights are designed to be applied across all parks.”

AI’s environmental applications aren’t limited to wild spaces. On M&S farms, it’s helping farmers monitor the health of livestock.

At an M&S farm in Chippenham, farmers are using AI to monitor the health and wellbeing of their cows. Tools such as CattleEye and Vet Vision AI track lameness and other health concerns in real time, helping staff identify problems early and reduce the need for intensive treatment.

M&S farmers tell edie that AI alerts allow them to catch issues before they become serious.

“Before, we might not notice lameness until it was more advanced,” one explains. “Now we can act sooner, which helps the cows recover faster and reduces the workload.”

According to CattleEye’s website, using AI to select cows for hoof trimming and treatment can lead to a 10% decrease in lameness, with an estimated $200 saved per cow per year and a 0.57-ton reduction in carbon footprint per cow annually.

But as these examples illustrate AI’s promise, there are growing concerns over the very infrastructure required to support it.

The rapid expansion of data centres, which require vast amounts of electricity and water, has sparked debates over whether the benefits of AI in areas come at too high a cost for nature. Data centre infrastructure can encroach on green spaces and strain local water resources.

The UK Government recently launched an eight-week consultation to streamline permits for energy infrastructure, including data centres. Officials say the proposals could reduce approval times from months to days, citing lessons from the US. But in the States, rapid data centre growth has contributed to water stress and pollution, raising concerns about potential impacts if similar approaches are applied in the UK.

Then-Housing Secretary Angela Rayner recently approved a hyperscale data centre on green belt land near the M25 in Buckinghamshire, overruling a local council. Campaigners, including Foxglove and Global Action Plan, argue no environmental impact assessment was carried out for the 90-megawatt (MW) facility. While smaller than other planned projects, even modest developments could increase local energy demand and electricity costs.

Dominic Dyer, chair of Nature 2030, says: “We need to make sure that what’s happened in the US doesn’t happen here in the UK. That means long-term planning around energy supply, environmental safeguards and how technology is deployed.”

He adds: “I think there is a technological innovation revolution going on. It’s every bit as life-changing as the Industrial Revolution, which changed all our lives.

“So you’ve got to basically come up with a way by which AI and data banks can be developed, but not by reliance on fossil fuels. That’s an opportunity, but also a challenge for business, for tech companies, for those people that are building those data banks.”

“Rather than just let all the tech leaders sit down and top up their billions, we’ve got to think about the entrepreneurs of tomorrow, and make sure this technology delivers real-world benefits for society and nature,” he concludes.