More than 300 teams from 70 countries competed in the XPRIZE Rainforest, a global competition that promotes new technologies for the exploration, evaluation and conservation of biodiversity in tropical rainforests. The results were announced during the G20 summit in Rio de Janeiro. The three winners are 'Limelight Rainforest' (first place), 'Map of Life Rapid Assessments' (second place), and our 'Brazilian Team' (third place). They will now share a prize pool of 10 million dollars among themselves.
Dr Thiago Gouvêa, from Brazil and senior scientist at DFKI Niedersachsen, where he heads the computer-aided sustainability and technology team in the research department Interactive Machine Learning (IML), says: We celebrate our top-three achievement in the XPRIZE Rainforest competition, but the real challenge lies ahead: halting and reversing the global decline in biodiversity. As an AI research group, we can make a contribution. The competition amplified the impact of our work by fostering collaboration with domain experts on the frontlines of biodiversity conservation. Moving forward, our focus is on refining our methods, strengthening these collaborations, and ensuring that cutting-edge technologies remain accessible to those tackling this critical issue.”
The Brazilian Team was one of twelve teams to qualify for the semi-finals in the first test phase in Singapore in 2023. In July 2024, it competed in the final together with five other teams in the Brazilian Amazon. The team also included researchers from DFKI Niedersachsen and doctoral students from the Applied Artificial Intelligence (AAI) chair at the University of Oldenburg. Dr Thiago Gouvêa, Hannes Kath, Bengt Lüers, Rida Saghir and Ilira Troshani jointly developed an AI system that filters out the acoustic signals that ecologists need to determine biodiversity from a large amount of audio data recorded in the Brazilian rainforest.
Working in the Brazilian Team of the XPRIZE Rainforest was a special experience, summarises scientist Hannes Kath: “Using machine learning methods, we worked to interactively quantify biodiversity using audio data and to combine knowledge of nature and technology. It was inspiring and very educational to work with scientists from a wide range of disciplines and to test our models under real-world conditions. Our success highlights the importance of our work and the effectiveness of our methods.”
In the ultimate stage of the challenge, which took place in July 7th to 30th this year, in the Tumbira Community of the Rio Negro Sustainable Development Reserve, the finalist teams had to survey the biodiversity in an area of 100 hectares of Amazon Forest, without ever entering the study area. Each team had 24 hours to collect data from the plot, and a further 48 hours to report key information concerning the biodiversity found there.
In total, the Brazilian Team documented 418 taxa (different kinds of living organisms), of which 266 were identified to the level of species. Their results included three species possibly new to science. Their report also documented complex interactions between the species and identified those that provide valuable ecosystem services to the forest bioeconomy. The fieldwork in this phase was conducted by 18 members of the Brazilian Team over the course of 24 hours and, in the subsequent 48 hours, the other members of the team collaborated to process and analyse the data and prepare the final report which was delivered to the coordinators of the challenge.
“The team made great efforts to identify the plants and animals they found to the most detailed taxonomic level possible, and analysed these identifications to provide insights on the local biodiversity”, explained the team coordinator Vinicius Souza. “The discoveries of the teams in the final will benefit not only Brazil, but the countries of Latin America, Africa and Asia that are home to tropical forests.”
DNA, Biodiversity and Insights and includes more than 100 collaborators with multidisciplinary technological knowledge and expertise (such as biologists, engineers, informaticians and lawyers). Most are Brazilians, but the team includes scientists from 10 other countries, so collectively the team represents dozens of Brazilian and international institutions.
The Bioacoustics group, to which the Oldenburg researchers belong, working with the Robotics group, developed solutions for the collection of audio recordings in the forest canopy, in clearings and even in aquatic environments. Protocols and software using AI were developed for the automated detection, classification and identification of vocalizations (calls) of different groups of animals such as birds, amphibians, bats, primates, and even the sounds of insects. In addition, more than 16 thousand sound recordings of animal were captured in Amazonia by BT and use for training these AI algorithms.
To organize this biodiversity information, the Insights group created a new platform Contextual Ecosystem Services on Amazon of Brazil (CESSABR). The CESSABR platform is a prototype database, which includes 4996 plant species and 6290 animal species from the Brazilian Amazon.
The Brazilian Team also emphasized their commitment to Indigenous and traditional communities, with a respectful approach aligned with Brazilian legislation. Initiatives were proposed to train young people from these communities in the use of technologies which offer potential for the sustainable economic development of the regions where they live, promoting their empowerment and increasing the productivity of forest resources.