Neural Tree Reconstruction for the Open Forest Observatory

A university-led consortium is building the Open Forest Observatory, a publicly accessible database of geospatial forest data and open-source tools that use uncrewed aerial vehicles to produce three-dimensional tree maps for climate and wildfire applications. The Open Forest Observatory (OFO) aims to make low-cost forest mapping available to ecologists, land managers, and the public [1]. The collaboration is assembling a database of geospatial forest data alongside open-source methods for mapping forests with uncrewed aerial vehicles [2]. The resulting data can help prioritize reforestation, inform wildfire hazard reduction, and monitor carbon sequestration [2]. In its current form, the OFO database generates 3D tree maps using classical structure-from-motion techniques [1]. That approach is prone to artifacts, lacks detail, and struggles to reconstruct the forest floor because overhead imagery has limited visibility there [2]. Reconstruction errors can propagate into downstream scientific tasks, including wildfire simulations [2]. Advances in 3D reconstruction such as Neural Radiance Fields, or NeRF, produce higher-quality results that are more robust to sparse views and support data-driven priors [1]. The OFO team is exploring ways to incorporate NeRFs into the dataset and outlining future work to support additional state-of-the-art 3D vision models [2]. The paper describing the effort was submitted on 16 June 2026 by Marissa Ramirez de Chanlatte and colleagues [1]. The submission file size was 7,496 KB [1]. The research underscores the importance of high-fidelity 3D reconstructions for forestry applications, where even small errors in tree structure can alter the output of models used to predict fire behavior or estimate carbon stocks [2].