Fig. 3: Predicted relative Plant area index (PAI, %) time-series.

a Forest phenology acquired using a Terrestrial Laser Scanner (TLS) within the Biological Dynamics of Forest Fragments Project (BDFFP) was vertically stratified, with the understory (<15 m aboveground) and upper canopy (≥15 m aboveground) presenting different trajectories of growth during the dry season. However, the vegetation structure and phenology were both significantly altered by edge effects up to 40 m from the forest fragment margins. PAI predictions from linear mixed modelling used measurement date (time), a categorical variable that indicates whether plots were near an edge (edge effects) and an interaction term time × edge effects as fixed variables. Edge effects nested within transect identity were included as random variables (Eqs. 1 and 2). Predicted relative PAI of the understory, upper canopy and total PAI (that combined both vertical strata) in b forest edges and c undisturbed interior forests was calculated as the PAI at any time divided by initial PAI (PAI_i) collected in April 2019. Each point (and lines corresponding to linear interpolations between points) represents the mean relative PAI obtained by fitting 200 randomised permutations of subsets split into 80/20 for calibration and validation, respectively. The shaded areas represent 95% confidence intervals based on uncertainty in those parameter estimates. While transects exhibited similar phenological trends, PAI differed significantly between transects—thereby we here show the predicted values. See Supplementary Fig. 4 for measured PAI and Supplementary Fig. 5 for absolute predicted PAI values and model uncertainty. The shaded area represents the dry season between mid-June and mid-October.