Leveraging Remote Sensing and Theory to Predict Tree Size Abundance Distributions Across Space

ABSTRACT Background Remote sensing (RS) technologies provide unprecedented opportunities to assess forest structure at broad spatial scales. Light detection and ranging (LiDAR) is a powerful tool that offers detailed vertical information, but its consistent high‐resolution coverage can be limited across vast areas and quantifying understory vegetation remains challenging due to occlusion. Conversely, while high‐resolution RGB imagery is more accessible and valuable for large‐scale analyses, it comes with higher uncertainty and only captures canopy‐level information. Integrating RGB data with ecological theory and existing LiDAR‐derived products (e.g., canopy height models, CHM) allows us to enhance predictions and broaden the applicability of forest size structure mapping from the understory to the canopy, particularly through tree size–abundance relationships. Theory Tree size and abundance generally follow known distributions, where smaller trees are exponentially more common than larger trees. This pattern emerges from fundamental ecological constraints, including volume packing, metabolic scaling and light competition, which collectively govern forest self‐organisation across spatial scales. Proposed Approach We present a workflow that integrates size‐abundance scaling theory with RGB data and CHMs to improve predictions of forest structure. By estimating tree size distributions from RGB imagery, we extend the applicability of ecological scaling models to broader spatial scales and offer an approach that complements traditional methods. Assessment We apply this approach to forest monitoring networks, including NEON and ForestGEO, to assess its accuracy and generalisability across a range of forests (e.g., subtropical pine, boreal, low and dry, tall and dense) in North America. Our results show that RGB‐based estimates can successfully recover size‐abundance distributions from the understory to the canopy. Future Directions Further refinement of our approach could enhance predictions by incorporating species classifications from hyperspectral data and using more spatially or taxonomically specific allometric equations. These additions would enable more precise scaling of tree diameter from remote sensing data.