A healthy coral reef is not only characterized by its vibrant colors but also by the sounds it emits. If you have had the opportunity to dive near one, you may have heard the clicks emitted by the different creatures that live underwater, such as shrimp or fish. As this background hum is such a characteristic of the soundscape of coral reefs, we could pay attention to it to monitor their health status, as they are in danger.
This is the idea that a team of researchers has had that has used machine learning (automatic learning) to train an algorithm that recognizes the subtle acoustic differences between a healthy reef and a degraded one, a contrast so tenuous that it may be impossible for our ears pick it up.
The researchers say that the tool offers interesting advantages when compared to other procedures used to determine the health status of reefs. The tool saves a lot of time and work , and you just have to think about what the operation would mean without it: periodic visits by divers who visually assess how the coral cover is, then listen to the recordings that are made… You also have to Keep in mind that many animals that live in this underwater ecosystem hide or are only seen at night, further complicating visual studies.
“Our findings show that a computer can pick up patterns that are undetectable to the human ear, ” says marine biologist Ben Williams of the University of Exeter in the UK. “It can tell us more quickly and more accurately how the reef is doing.”
What Williams and his colleagues did to capture the sounds of corals was to record at seven different locations in the Spermonde archipelago, off the southwestern coast of Sulawesi, Indonesia, which is also home to the Mars Coral Reef Restoration project for the restoration of coral reefs. The recordings included four types of reef habitat: healthy, degraded, mature restored, and newly restored . Each of them had a different amount of coral cover and therefore a different noise.
“Previously we relied on listening to and manually annotating these recordings to make reliable comparisons,” Williams explained in a Twitter thread. “However, this is a very slow process and the size of marine soundscape databases is skyrocketing with the advent of low-cost recorders.”
To automate the process, the team trained a machine learning algorithm to be able to differentiate the different types of coral recordings. Subsequent tests showed that the AI tool could identify reef health from audio recordings with 92% accuracy .
“This is a really exciting development,” says co-author and marine biologist Timothy Lamont of Lancaster University in the UK. “In many cases it is easier and cheaper to deploy an underwater hydrophone on a reef and leave it there than to have expert divers repeatedly visit the reef to study it, especially in remote locations.”
According to the researchers, the results of the algorithm depend on a combination of factors in the underwater soundscape, such as the abundance and diversity of fish vocalizations, invertebrate sounds, and even faint noises thought to be produced by algae. along with contributions from abiotic sources, such as subtle differences in the sound of waves and wind in different types of coral habitat.
As experts have pointed out, our ears are not able to easily identify those faint sounds that can occur on a reef, but a machine can do it reliably. However, they acknowledge that the method has room for refinement to offer more sound sampling and “a more nuanced approach to classifying ecostate.”
The bad news is that time is just what coral reefs don’t have. We must act quickly if we want to save them.
Referencia: Williams B. et al. 2022. Enhancing automated analysis of marine soundscapes using ecoacoustic indices and machine learning. Ecological Indicators. DOI: https://doi.org/10.1016/j.ecolind.2022.108986
