Freshwater ecosystems are essential to our planet, teeming with an abundance of life that sustains both nature and humanity. These shimmering lakes, meandering rivers, and tranquil ponds are not just serene havens; they support biodiversity, providing a sanctuary for countless species. Yet they are fragile and under threat from pollution, overuse, and climate change. Researchers, determined to preserve them, are finding new ways to study and protect these ecosystems, and could even get some help from artificial intelligence.

“There is such an alarming loss of freshwater biodiversity because most human populations around the world exist next to, and depend upon, freshwater ecosystems. This creates conflict between our interests for survival, which often involves over-exploitation, and the long-term survival of the ecosystem,” says Jack Greenhalgh, an ecologist and science communicator at the Pyrenean Institute of Ecology. “As human populations rapidly expand, there is an ever-increasing pressure placed on freshwater ecosystems.”

To preserve and protect these ecosystems, ecologists need to understand the level of biodiversity that exists in that area. Simply put, abundant biodiversity signifies a healthy ecosystem. However, conventional methods used to monitor biodiversity in freshwater ecosystems are usually invasive, often resulting in researchers inadvertently disturbing or harming the habitat and organisms. Therefore, finding non-invasive methods to monitor and understand them is vital for their conservation.

Greenhalgh recently investigated a novel and unique way of measuring freshwater biodiversity that mitigates the impacts from conventional methods, by simply listening. Greenhalgh published a study that investigated the daily acoustic cycles of British ponds, which used pond soundscapes.

“Observing nature by simply listening offers a totally new perspective, which is revealing lots of new and interesting information that can help conservation efforts,” Greenhalgh says. “A healthy pond sounds like a rainforest, with lots of different sounds made by the frantic rattling of insects and persistent whining and popping of plants.”

Information gathered by monitoring soundscapes, or all the sounds in an environment are beneficial in a number of ways. They can be used to assess the biodiversity of an area, detect environmental changes such as deforestation or habitat degradation, and promote the overwhelming diversity of life that occurs within an ecosystem.

Greenhalgh used an underwater microphone, called a hydrophone, to listen to the cacophony of sounds created by plants and insects in British ponds. Aquatic insects make sounds by rubbing hard body parts together in the hope of attracting mates, and during respiration, submerged plants release oxygen bubbles that create rhythmic whining and ticking sounds.

In Greenhalgh’s study, he submerged microphones into the water and recorded the sounds from five ponds over three months, collecting 840 hours worth of underwater sound recordings. He discovered that British ponds possess daily acoustic cycles, with singing insects at night and popping photosynthesising plants during the day.

“We knew that different species of aquatic insects must be able to produce different sounds because old identification books focus on the sound-producing anatomy that they have to distinguish between species, so if they’ve got different anatomical instruments, then they must be able to make different noises. But there’s very little work done to catalogue those sounds,” Greenhalgh says.

Cataloguing insect sounds is no easy feat, and Greenhalgh has spent hours trying to decipher different sounds created by different insects. So far, only a very small number of freshwater sounds have been catalogued reliably to species level, but Greenhalgh is currently working with people from across the globe to establish an extensive audio library.

“One of the great, but also problematic aspects, of soundscape monitoring is that it generates vast amounts of data,” he says. “Imagine you’ve got fifty different microphones out there all recording for five hours day and night for months, maybe even years. You’d quickly end up with an inconceivable amount of data that you would never be able to listen to.”

However, there might be an answer when it comes to deciphering all this complex information: artificial intelligence. Built with the capability to analyse large amounts of data, AI could be just what conservationists need to understand the complexities of our ecosystems. Much progress has been made recently surrounding the development of machine learning models to distinguish between images. And AI is finding increased uses in sound recognition.

For example, AI is playing a transformative role in assisting individuals with voice impairments by facilitating innovative communication solutions. Through voice recognition technology, AI systems can interpret and convert text input, allowing people with speech challenges to express themselves verbally and communicate effectively. In this way, AI promotes inclusivity and fosters a more accessible and supportive environment.

The use of AI in the analysis of sound recordings from terrestrial and marine environments is rapidly gaining traction, too. For example, Abu Dhabi-based Palmear is harnessing the power of AI and acoustics for early pest detection, which could help preserve palm tree populations globally. Their solution works by inserting a needle into a tree for 50 seconds; AI analyses tree sounds, pinpointing pest infections and enabling precise interventions.

The continuous learning capability of the AI, fed by incoming data, enhances its ability to recognize and distinguish different pest sounds, ultimately improving its effectiveness over time. Additionally, the system’s aptitude for interpreting sounds shows potential in assessing soil biodiversity and promoting regenerative agriculture, contributing to the pursuit of healthier and more productive ecosystems.

“Artificial intelligence is a very powerful and non-invasive tool in measuring biodiversity using bio-acoustics, it serves the ability to detect specific species in an acoustically complex soundscape which provides invaluable insight into the taxonomy of the sample,” says Zeid Sinokrot, CEO and founder at Palmear.

However AI is yet to be fully applied in the analysis of freshwater soundscapes. By having methods that can automatically recognise different species, conservationists could rapidly gain an understanding of the biodiversity in an area, providing new insights into how freshwater ecosystems are changing in response to climate change and other human-induced pressures.

“Different species produce unique sounds,” Greenhalgh says. “Using machine learning, you could teach a computer model to automatically recognize each species’ unique acoustic signature.”

Understanding freshwater soundscapes could also help shift public attitudes towards freshwater conservation. Greenhalgh pointed out the power that recordings of whale song had on conservation efforts to stop whaling and how public interest in these aquatic giants grew as a result.

“I understand a water beetle is not as charismatic as a whale,” Greenhalgh says, “but by highlighting the surprising beauty and complexity of freshwater soundscapes, it might be possible to shift attitudes towards the vital conservation of our freshwater habitats.”