Canopy gap monitoring in Central African tropical forests using pocket drones: Reliability depends on scale

Canopy gaps play a fundamental role in shaping light regimes, regeneration dynamics, and carbon stocks in tropical forests. Yet, the high cost and logistical constraints of monitoring 3D forest structure at large scales severely limit our understanding of gap structure and dynamics in Central Africa. We evaluated the performance of pocket-drone (∼ 250 g, ∼ 500 USD and Passive optical RGB sensor) derived digital surface models (RGB-DSM) against airborne LiDAR-derived canopy height models (LiDAR-CHM), used as a reference, over the Lileko site (∼240 ha) within the Yangambi forest landscape located in the Democratic Republic of the Congo (DRC). Results revealed a consistent scale-dependent pattern in gap estimation performance relative to airborne LiDAR reference data. Agreement between methods increased with canopy height, yet local-scale spatial correspondence remained limited across all vertical strata, indicating persistent uncertainty in the precise delineation of individual gaps. However, at plot scale, estimates of gap area per 1-ha plot converged between sensor instruments, suggesting that local over- and under-estimations compensate each other when aggregated. This compensatory effect was more pronounced in structurally heterogeneous old-growth mixed forests than in monodominant Gilbertiodendron dewevrei forests, highlighting the role of forest structural complexity in shaping cross-sensor agreement. These results confirm that Pocket drones provide a low-cost, spatially consistent proxy for LiDAR-derived canopy structure despite a systematic bias. Although this bias limit fine-scale individual gap delineation, they reliably estimate plot-scale total gap area, supporting applications in gap dynamics, carbon accounting, REDD + , and canopy monitoring in LiDAR-scarce Central African forests.