ABSTRACT Hurricanes strongly regulate vegetation structure and dynamics in coastal tropical forests by causing tree mortality and damaging the crowns of surviving trees. Despite the importance of this damage for long‐term forest demography and carbon dynamics, the patterns and drivers of variation in hurricane‐induced damage across heterogeneous landscapes remain poorly quantified. Here, we integrate field inventory data with repeated airborne LiDAR measurements to quantify tree crown damage following Hurricane María (Category 4) in the Luquillo Experimental Forest, Puerto Rico. We applied two approaches: manually delineated crowns and an automatic 3D crown delineation algorithm (AMS3D) to quantify individual crown damage from Hurricane Maria over the 16‐ha Luquillo Forest Dynamics Plot (LFDP). Both methods consistently revealed that taller canopies experienced less damage compared with shorter canopy trees, contradicting previous pixel‐based remote sensing assessments reporting greater damage in taller trees and forests. Analysis of crown‐level responses to hurricanes showed that tall canopy trees (≥ 25 m) suffered approximately 30% less crown area damage than shorter canopy trees (~15 m), while other tree‐level factors had negligible effects. At the stand level, canopy height was also negatively associated with damage. Quadrats dominated by Prestoea acuminata var montana and Dacryodes excelsa experienced less damage. The effect of stand canopy rugosity and topography was limited. Across the landscape, canopies at low elevations experienced higher damage, possibly due to shifts in species composition along the elevation gradient. As climate change is expected to increase hurricane frequency and intensity in the North Atlantic, our findings highlight the need for incorporating organismal attributes (tree size and species identity) into process‐based ecosystem models to accurately predict hurricane‐induced damage. Furthermore, moving from pixel‐based to crown‐based remote sensing approaches is essential to correctly capture forest responses and reduce prediction biases in tropical forest vulnerability and resilience under intensifying storm regimes.
Han et al. (Thu,) studied this question.