Tropical forests are important for storing carbon, supporting biodiversity, protecting soil, and regulating water and climate. These benefits depend not just on how much forest there is, but on the structure of the forest—how tall the trees are, how dense the canopy is, and how complex the layers of vegetation are from top to bottom.

As forests regrow after being cleared or burned, it’s important to assess how well they are recovering. Most large-scale studies focus only on tree height, which misses important aspects of forest structure. In this study, we used laser data from NASA’s GEDI mission to measure forests in 3D. We developed a machine learning method that compares each forest area to a reference set of well-preserved forests and assigns a score that reflects how structurally “mature” it is.

We tested this method in areas with different land uses and found that conserved forests scored highest, while agricultural and degraded areas scored lowest. We also found that forests burned multiple times over the last decades had poor structural recovery. This method provides a new way to track forest recovery more accurately, helping guide restoration, carbon programs like REDD+, and forest policy decisions.