Eavesdropping on fish to measure river health

Often hidden from sight, and definitely not heard by humans, fish grunt, squeal and gurgle as they hunt, make love and pass the time of day with neighbours.

For the first time, we are able to use these sounds in order to detect the health of fresh waterways, the rivers, lakes and streams.

In research published in Frontiers in Ecology and the Environment, researchers used underwater microphones to record the sounds of fish, other animals and the water itself. Analysing this soundtrack allowed scientists to count the number and variety of fish and animals present, and measure river health.

Dr Simon Linke from the Australian Rivers Institute at Griffith University says the advantage of the new approach is that it provides a continuous measurement that does not hurt fish.

“Traditional methods of monitoring freshwater systems rely on taking samples, and these are expensive, unreliable and intrusive. They are only accurate at the time they are taken, in contrast to an audio monitoring system which gives a continuous readout,” he said.

New technology has been developed that uses artificial intelligence (AI) to analyse these continuous river recordings.

Dr Toby Gifford – now at Monash University – is an expert in real-time processing of audio signals. He argues that we can now use computers to automatically process biological sounds, removing the need to manually annotate and count fish calls.

“Acoustic monitoring is becoming mainstream in many terrestrial ecosystems, and even in oceans,” he said. “It is exciting to start applying these methods in rivers, particularly now that advances in deep learning are making automated analysis feasible.”

The recording system used by Dr Linke and his colleagues also works with insects, as Dr Camille Desjonqueres from the Paris Natural History Museum explains:

“We record the sounds of water insects, another important indicator of water quality. When these soundtracks are fed into a computer, our acoustic monitoring system provides an instant readout.”

The next logical step would be to build a database of fish calls, so the sounds can be matched with a fish species.

“There are 120,000 bird calls on record and it would be great to have a similar underwater sound database, like a museum reference collection,” Dr Linke said.

The research was carried out Dr Linke and colleagues from Australia, Switzerland and France. The researchers hope their paper will be a call to action for the global research community to gather forces and make this operational in the next 10 years.