Factors Affecting the Practicability of UAV Monitoring During Early Tropical Forest Restoration: The Roles of Size, Season and Species

Management of forest restoration relies on early tree-performance data from labour-intensive ground surveys. Unmanned Aerial Vehicles (UAVs) offer a labour-saving alternative. However, the factors affecting when planted saplings first become detectable in UAV imagery are poorly understood. We investigated how sapling size, species and season affect UAV monitoring of a tropical forest-restoration trial on a limestone quarry. Using an off-the-shelf RGB quadcopter flying 10 m above ground, we compared sapling detectability and size in UAV imagery with ground measurements, over the first year after planting. One year after planting, 85% of live saplings were detected in UAV orthomosaics. However, generalised linear mixed models of detectability in 3D models, accounting for season and species, predicted high detectability for most species only at 2 m tall. For detectable saplings, mean 3D model heights matched ground data (mean difference −2.0 cm, 95% CI −5.5 to 1.5 cm, after one year), but individual measurements were imprecise. Sapling detectability was lowest in the dry-season surveys, consistent with leaf loss in deciduous species. However, leaf condition was not recorded, and seasonal effects could not be separated from survey-specific factors. Orthomosaics may support sapling counts by the second rainy season, but height measurement from 3D models may require saplings to reach 2 m tall or taller, which few saplings had reached during the study. Further testing in more complex environments is recommended.