Developments in artificial intelligence (AI) are speeding up the time it takes to process large amounts of scientific data and could potentially help scientists make better decisions about how to preserve biodiversity, say industry experts.

Researchers at the Zoological Society of London (ZSL) use AI to reduce the time it takes to process the large amounts of data gathered from using camera traps, tracking devices, and audio systems to analyse animal populations and ecosystem interaction.

Thanks to continual improvements in cognitive AI, they are then able to make judgements about that data, such as detecting species, and make predictions about the environmental drivers of animal behaviour, said Robin Freeman, head of the indicators and assessments unit at ZSL.

“We can use AI to understand how factors like climate change, habitat loss, and habitat degradation are impacting populations, and perhaps make decisions about how to mitigate some of those impacts,” Freeman said during a recent panel discussion on the use of AI in nature-related fields hosted by Google.

Global wildlife populations have declined by as much as 69% on average since 1970, according to the WWF’s Living Planet Report 2022, largely driven by anthropogenic causes such as habitat degradation, habitat loss, and climate change.

Recognition of biodiversity decline and the detrimental impact it has upon the global economy, particularly in nature-reliant sectors like agriculture and manufacturing, is growing though action to curb nature decline isn’t happening fast enough, said Freeman.

Under the Kunming-Montreal Global Biodiversity Framework (GBF) agreed at COP15 in Dec. 2022, signatory countries agreed to protect 30% of land and marine areas for nature by 2030, known as “30X30”, though critics say that government action isn’t happening fast enough and that more policy consistency is needed.

Measuring biodiversity richness and how it changes depending on different anthropogenic and environmental drivers is key to ascertaining which areas to prioritise for protection and to understanding overall ecosystem health.

AI allows scientists to automate this data collection, know where to focus, and collect more data overall, said Freeman.

SPEEDIER ANALYSIS

The use of camera traps to collect data about animal populations produces vast amounts of data that is incredibly time consuming and laborious for people to process alone.

One survey in London’s Regent’s Park with 150 cameras will generate about 1.6 million images, which could take over a year to process using people alone, whereas with the use of AI, the time taken to process that data can be reduced tenfold, by detecting animals’ presence in an image, said Freeman.

The development of new AI techniques will enable the identification of what species there are, and therefore reduce the overhead again, he added.

“We’ll still need to have humans in the loop to confirm it, but the impact of automation and AI on the day-to-day is phenomenal,” said Freeman.

APPLYING AI TO AUDIO

AI is also applied to audio data by ZSL researchers, by reducing the time it takes to process the data collected from open-source acoustic monitors like AudioMoth that are particularly useful for detecting species like bats and birds.

Using acoustic monitoring devices to collect data about ecosystem health based on animal sounds is a powerful and cost-effective way of measuring biodiversity and should be one of the primary data collection methods in new biodiversity crediting systems, say experts.

AI models can be used to automatically detect where and when particular species occur across an entire dataset using their specific acoustic signatures.

Interpreting data from tracking devices, such as those attached to seabirds on their migration between the UK and Argentina, is also made easier through the application of AI, according to ZSL researchers.

“We have recorded thousands of birds on their migrations between the UK and Argentina, and this has completely changed our understanding of how they engage with their environment,” said Freeman.

We can use deep learning to look at the different traces we get, for whether they’re moving, whether they’re touching the water, the light levels, and diving under the water, to make predictions about when the dives occur, and what the environmental drivers of those dives are.”

Deep learning is a subset of machine learning that uses multi-layered neural networks to simulate the complex decision-making power of the human brain. Some form of deep learning powers most of the AI around today.

MAKING BETTER DECISIONS

ZSL analysis aided by AI has shown that locations warming the fastest see bird populations decline more rapidly than elsewhere, which highlights the detrimental impact of climate change on biodiversity, in addition to anthropogenic land use change as a negative driver, said Freeman.

AI can be applied to these global population models and potentially used in future to make better decisions about how to reduce biodiversity decline, he said.

The technology can also be applied to locate ocean biodiversity hotspots, such as the AI model developed by the Wildlife Conservation Society (WCS) that allows scientists to map the locations of underwater biodiversity hotspots across 11 countries in the Western Indian Ocean.

The model, announced in late April, found that of the 119 biodiversity hotspots identified in the Western Indian Ocean, most are not currently protected or otherwise conserved, highlighting the importance of using technological tools like AI for countries drawing up their 30X30 biodiversity plans.

The researchers paired high-resolution oceanographic data with detailed in-water surveys by field scientists, creating an AI model to identify underwater biodiversity hotspots, with the region broken down into 6.25-kilometre reef cells, with cells identified for the highest numbers of fish and coral species.

“We found that among the highest biodiversity locations in these 11 countries, many were not protected at all. Most marine protected areas (MPAs) don’t have sufficient data to back up their designation,” said Tim McClanahan, director of marine science at WCS.

“Many MPAs were designated based on ‘expert opinion’ and observational anecdotes rather than data and models. What’s often lacking is real data telling us: Where are the highest-biodiversity areas in each country? Which places will be the most climate-resilient? Which areas do people like fishers rely on the most for food and income? Those are the types of data that we need in order to make the best decisions. This new model advances the ability to make the right decisions,” he said.

Top ranked locations for underwater biodiversity in Western Indian Ocean

Source: WCS

Meanwhile in the UK, a study using AI alongside acoustic monitoring will assess pollinator levels on solar farms, to gauge how these projects can boost pollinator biodiversity compared to transitional farmland.

The study, led by renewable energy company Low Carbon in partnership with Lancaster University and the UKRI Engineering and Physical Sciences Research Council, will be carried out at Westmill Solar Park in Oxfordshire, which was developed by Low Carbon in 2011 and operates as a community-owned solar project.

The research aims to provide key evidence for emerging policies such as the UK’s Biodiversity Net Gain (BNG) policy, which became mandatory in February, and for how renewable energy firms measure the impact of development on nature.

By Bryony Collins – bryony@carbon-pulse.com

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