Ecology is entering an era where the limiting factor is no longer collecting data, but converting it into reliable ecological measurements. A major knowledge gap is that many core variables we care about—tree-level mortality and turnover, species composition change, animal abundance and distribution, and outcomes relevant to carbon accounting—are still estimated with methods that are labor-intensive, site-specific, and difficult to scale. This creates bottlenecks for tracking rapid disturbance impacts, evaluating restoration and management actions, and verifying claims such as carbon credits with transparent, auditable evidence. Our lab is working at the cutting edge by developing computer vision and deep learning pipelines that translate high-resolution airborne or drone imagery into standardized ecological layers across broad areas. For forests, we delineate individual tree crowns and use them as the basic unit for monitoring survival, biomass change, and species turnover over time, enabling both fundamental ecological inference and practical reporting for carbon verification. In parallel, we build models to detect and identify animals from airborne imagery, which helps close the gap between plot-based wildlife surveys and landscape-scale estimates needed for conservation and rewilding. Together, these approaches push AI toward repeatable, uncertainty-aware ecological monitoring that can connect fine-scale biology to regional decision-making.

• Individual tree species mapping across the National Ecological and Observatory Network (Li et al., in preparation)

We have generate species maps at individual tree level.