Repeated RGB-Colorized Handheld SLAM for Height-Resolved Seasonal Observed Occupancy in Contrasting Deciduous Forest Sectors

Seasonal forest phenology is commonly summarized as canopy greenness or phenophase timing, although leaf development also redistributes observed plant material through three-dimensional space. We evaluated whether repeated RGB-colorized handheld simultaneous localization and mapping (SLAM) can provide height-resolved trajectories of seasonal observed occupancy in adjacent Ailanthus altissima-dominated and native-dominated sectors of a young deciduous forest. We acquired 149 scans on 25 dates from March 2025 to March 2026 at six permanent locations. Point clouds were restricted to date-invariant common support, normalized to a March terrain model, and voxelized at 0.20 m. New occupancy was referenced to the union of two strict March leaf-off scans. A weakly supervised foliage likeness proxy combined geometry-first pseudo-labels with relative color, intensity, and local three-dimensional features; its outputs were interpreted as relative scores rather than leaf fraction, LAI, or biomass. The strongest and most persistent invaded positive signal was localized to 2–4 m. Continuous-time models supported the integrated 1–5 m contrast from late April through October, whereas formal support for the 5–12 m crown domain was limited to the late season invaded positive phase; the earlier native positive crown feature remained descriptive. Height-integrated SLAM showed broad seasonal concordance with intercepted PAR (rrm = 0.881), GCP-linked Sentinel-2 EVI2 (rrm = 0.670) and the five-date litterfall comparison (rrm = 0.914). However, correlations of the invaded minus native trajectories were positive but imprecise. The independent observations therefore supported the broad seasonal cycle rather than the detailed sector-specific or height-specific pattern. The workflow provides a conservative means of localizing relative seasonal observed occupancy in three dimensions, but the resulting contrasts remain site-specific and hypothesis-generating.