A team of scientists is studying the sound of the forest in Ecuador to learn how artificial intelligence (AI) could follow animal life in recovering environments.
When scientists want to measure new forest growth, they can study large areas of land with tools like satellites and lidar.
But understanding how fast and in what amount wildlife is returning to an area is more difficult. Sometimes it requires an expert to listen through sound recordings and pick out animal calls.
Jorg Muller is a field expert on birds at the University of Wurzburg Biocenter in Germany. He wondered if there was a different way.
Muller told the French news agency AFP: "I saw the gap that we need, particularly in the tropics, better methods to quantify the huge diversity... to improve conservation actions.”
So, he turned to bioacoustics, which uses sound to learn more about animal life and the environments in which they live.
The tool has been used by scientists for some time. But more recently, researchers are using it with computer learning to study large amounts of data more quickly.
Muller and his team recorded wildlife sounds at sites in Ecuador’s Choco area. The environments they recorded included areas that were once used for agriculture and raising livestock to old-growth forests.
They first had experts listen to the recordings and index the sounds of different animals. Then, they examine the sound quality to measure the environment.
Finally, they ran two weeks of recordings through an AI computer program trained to understand 75 different bird calls.
More recordings needed
The program was able to pick out the calls on which it was trained. However, scientists wondered if the program could correctly identify the number of different kinds of plants and animals in each environment.
To see if the program could do that, the team used two different controls. One was from the experts who listened to the audio recordings, and the second was based on examples from each environment, which can be used to understand biodiversity.
Since the amount of available sounds used to train is limited, the AI program could only identify one-fourth of the bird calls that experts could. But it was still able to correctly measure biodiversity levels in each environment, the study said.
The research was published recently in Nature Communications. The study said the scientists’ results show that the AI program is a powerful tool to measure the recovery of animal communities in tropical forests.
The research noted that biodiversity found from recordings can be quantified in a cost-effective and complete way. And it said that it can measure environments, “… from active agriculture to recovering and old-growth forests.”
There are still areas for improvement, including the lack of animal sounds on which to train AI models. And the method can only capture animals that use sound to communicate.
I’m Gregory Stachel.
