Field Museum Team In Xprize Rainforest Finals

The catch was that the competition rules prohibited any team member from setting foot in the forest. As the countdown began, a fleet of drones lifted off from the clearing to begin data collection. One of six teams competing in the XPRIZE Rainforest finals, the Map of Life team included Field Museum staff (ichthyologists Lesley de Souza and Sophie Picq, botanist Nigel Pitman, and newly hired ornithologist Cameron Rutt) and Associates (ornithologist Ramiro Melinski and botanists Luis Torres and Marcos Ríos). The MOLRA team used drone-based audio, photo, and eDNA sampling to make over 5,000 identifications across 225 unique species and 109 plant morphospecies. The fleet of lightweight drone, collected 4,000-plus photos, 26 hours of audio recordings, and 24 separate eDNA samples from the study site in a single day. Getting from raw samples to species identities was achieved through the team’s advanced new modeling technology, state-of-the-art artificial intelligence algorithms, innovative eDNA processing techniques, and collaboration with biodiversity experts not just in Brazil, but all over the world.  The goal of the XPRIZE Rainforest competition was to spur the development of new technologies to survey biodiversity remotely, and the finals saw a parade of drones and robots and AI-powered canopy platforms. As Nigel Pitman noted, for the Field Museum, the most important lessons coming out of XPRIZE are much more down-to-earth. Historically, the data collected in the Museum’s rapid inventories has taken years to process and share. During XPRIZE Rainforest we saw data pipelines operating more than 100 times faster than ours, capable of transferring high-quality, expert-validated biodiversity data from a remote field site to the Global Biodiversity Information Facility in a matter of hours. Indeed, after the finals all of the photographs and audio recordings collected by all of the XPRIZE teams were posted here. We learned that experts can do a great deal of species identifications remotely; not a single one of our plant and animal experts were on-site during the competition, and some were as far away as Denmark. We also learned that it is possible to deliver a comprehensive report 100 times faster than we do; at the end of the 72-hour period the Map of Life team delivered a gorgeous 230-page report on our findings, with more than 100 tables and figures, as well as a photographic field guide and a digital dashboard displaying our findings. The competition also showed how environmental DNA will boost our rapid inventory field work, by proving that the time to convert an eDNA sample into a list of species can be measured in hours. The Map of Life team pioneered some other drone-based sampling techniques that have the potential to improve rapid inventory results. For example, the team programmed drones to fly predetermined routes above the canopy, stopping at waypoints to take an array of photographs, with no piloting required even for takeoff or landing. The captured images were first run through an AI model that flagged photos containing fruits or flowers, then through another model that offered preliminary identifications of those plants. Automated sampling like this has the potential to give rapid inventory botanists a low-cost “eye in the sky” to spot canopy plants we would not otherwise detect in our ground-based surveys. Funding for the MOLRA team was provided by the Keller Science Action Center, the Negaunee Integrative Research Center, the Grainger Bioinformatics Center, and the Walder Foundation. Winners of the competition are expected to be announced in November!
September 27. 2024