Artificial intelligence models capture the biodiversity of the Chocó

In the Chocó forests, in the Canandé and Tesoro Escondido reserves, a research team led by Dr. Jörg Müller from the University of Würzburg (Germany) applied bioacoustics, the study of animal sounds, to measure the biodiversity of the forest soundscapes. The results highlight the high regeneration capacity of the forest, which needs just one human generation to recover most of its biodiversity, without the need for active restoration.

The study started in October 2021, when Fundación Jocotoco opened its Chocó Lab in the Canandé reserve, located in the Esmeraldas province. Here, the Reassembly* research unit, with the participation of more than 15 German and Ecuadorian universities, works in a four-year project to study the dynamics of natural forest recovery and the resilience of ecosystem processes across time. The lower Chocó tropical forest of Ecuador is one of the most diverse and threatened ecosystems on the planet. Deforestation, climate change and the advance of the agricultural frontier are the main drivers of biodiversity loss in this region.

The bioacoustics project used recordings of vocalizing species that were analyzed by a group of experts, and their identifications were combined with two different types of automated measurements, one of which employs deep learning. Metabarcoding (DNA sequencing of environmental samples) was also used to contrast the results of the artificial intelligence models, as well as to include non-vocalizing species, such as the majority of nocturnal insects.

The study area comprised 43 plots in every stage of the forest recovery spectrum, from pastures and cacao plantations to mature forests, including areas with regrowth of 1 to 32 years. The results indicate that both the community composition (i.e., the biodiversity) of the vertebrate and insect species increases as the forest regenerates. Species composition is a better indicator of forest recovery than the number of individuals of each species.

For Dr. Martin Schaefer, CEO of the Fundación Jocotoco and co-author of the paper, "the use of artificial intelligence in biodiversity monitoring is a scalable, efficient and cost-effective method. Quantifying the biodiversity that we protect in our reserves will allow us to evaluate our conservation efforts, in addition to serving as a basis for developing green financing mechanisms, such as biodiversity credits, using reliable data." Jocotoco's Conservation department is developing the methodology for landscape level monitoring using soundscapes, which will be implemented in 2024.

The paper is available here at no cost.

*Reassembly is funded by the German Research Foundation (DFG - Deutsche Forschungsgemeinschaft)