Evaluating GEDI for quantifying forest structure across a gradient of degradation in Amazonian rainforests

Abstract Forest structure is key to understanding the resilience of tropical forests (their ability to recover from disturbance) and predicting how these ecosystems will respond to future environmental and climatic fluctuations. Current resilience studies in the Amazon rely on passive and active remote sensing forest cover metrics that offer limited insight into nuanced forest canopy structural changes associated with degradation. The Global Ecosystem Dynamics Investigation (GEDI) spaceborne lidar provides detailed information on canopy structure, a key factor in forest recovery, resilience, and ability to provide ecosystem services. We evaluate GEDI spaceborne lidar’s capability to characterise forest structure along a gradient of degradation (e.g. primary unburned (PU), secondary recovering, fire frequency), and investigate the potential of quantifying forest structure to advance understanding of forest responses to disturbance across the Amazon. We assess the correspondence of GEDI structural metrics such as relative height (RH) and canopy cover to airborne lidar metrics across the Brazilian Amazon using Lin’s concordance correlation coefficient (CCC). We evaluate GEDI forest structure variation along a gradient of forest degradation. We explore the potential of principal component (PC) analysis applied to GEDI data to derive a continuous descriptor of forest structural state that characterises the forest degradation continuum, and use a multinomial logistic regression model (MNLR) to further evaluate this descriptor. The strongest positive correspondence for all sampled GEDI and airborne lidar footprints occurs at RH96 (CCC; 0.57) of the canopy height profile, with strongest agreement in primary forest burned at least three times (CCC; 0.91). Whilst canopy cover showed significant recovery 15–25 years after disturbance, forest canopy height and aboveground biomass density had not fully recovered to pre-disturbance levels within 38 years. The PCA identified the importance of RH75, RH96, foliage height diversity index and canopy directional gap probability, with PC1 and PC2 explaining 80% and 15% of variance, respectively. The MNLR suggests that the ratio of these PCs effectively characterises the forest degradation gradient. We demonstrate the ability of GEDI 2A/B structural metrics to differentiate forest structure along a gradient of PU to secondary severely degraded forest in the Amazon rainforest, despite some overlap in structural characterisations between similar degradation classes. Our new method for deriving a forest structural state metric supports future research on forest monitoring, conservation, and the study of Amazon-wide ecosystem resilience.