ABC All-Hands 2025 Recap

ABC Global Center convenes all-hands at Pennsylvania field station, charts next phase of ‘AI for Nature’

As Henry David Thoreau wrote, “In wildness is the preservation of the world.” For the ABC Global Center, that wildness is also a living laboratory where field notes, birdsong and camera traps become data, and data becomes decisions.

The all-hands meeting at the Pymatuning Lab of Ecology began with a dawn bird hike before shifting to sessions on multimodal sensing—camera traps, bioacoustics, drones, GPS and satellite imagery—and the infrastructure to make those streams FAIR, reproducible and AI-ready. Breakouts examined sensor placement and sampling design, masking methods, and strategies for merging datasets across scales, from neighborhood plots to national observatories, with a focus on explainable AI that highlights the traits scientists actually use.

“Our goal is simple and hard: turn messy, cross-modal field data into trustworthy tools that help people make decisions on the ground,” said Justin Kitze, whose lab at the University of Pittsburgh hosted the meeting. “From birdsong to drone imagery to environmental records, we’re building workflows that others can rerun, test and improve—because there’s no AI without data, and no impact without reproducibility.”

Participants reviewed progress on digital-twin pilots (including work at The Wilds), sketched shared benchmarks for image and audio datasets, and aligned outreach plans—from classroom materials to community field days—to widen participation. Plain-language communication remained a priority, with teams workshopping quick definitions for “camera trap,” “bioacoustics” and “digital twin” to reach funders and policymakers.

The gathering coincided with national momentum around research rigor and the launch of Year Two of the NSF HDR Machine Learning Challenge, inviting computer scientists and ecologists to tackle out-of-distribution modeling on vetted, containerized workflows.

Next steps include expanding field deployments, publishing merged datasets with clear documentation, and inviting collaborators to co-develop benchmarks and decision-support tools. The through-line: AI for Nature—explainable, open and actionable science that connects data to conservation outcomes.