The PeatPic project: Predicting plot-scale green leaf phenology across peatlands

This dataset comprises standardized plot-scale photographs and associated metadata collected from 42 peatland research sites during the 2021 and 2022 growing seasons. Using the PeatPic protocol, participants photographed dominant vegetation communities weekly to bi-weekly from pre-growth through senescence, capturing seasonal dynamics across bog, fen, and swamp ecosystems. Metadata includes vegetation type, microform, site characteristics, and disturbance history. RGB imagery was analyzed using the Phenopix R package to derive a greenness index, and seasonal trajectories were modeled using a Gaussian function to estimate the timing and magnitude of peak greenness. These data support cross-site comparisons of peatland green leaf phenology (GLP) and facilitate broader investigations into vegetation responses to climate and disturbance.

Methods
Using the standardized PeatPic protocol, participants collected nadir-view photographs (0.036–1 m² plots) of dominant vegetation communities at 27 peatland research sites. Photographs were taken weekly to bi-weekly throughout the growing season and into autumn, under consistent lighting (auto white balance, overcast or clear skies, no shadows). Plots (4–12 per site) were either permanent or temporarily marked. Each site used a single camera or phone model. Metadata collected for each plot included peatland type, dominant vegetation group, microform type, experimental treatments, tree cover, permafrost presence, and disturbance history.

RGB images were analysed using the Phenopix R package (v2.4.4; Filippa et al. 2016). Greenness Index was calculated as GDN / (RDN + GDN + BDN), omitting images with snow or standing water. Regions of interest were delineated using DrawROI and greenness indices extracted per pixel with Extract VIs. A Gaussian model was fitted to greenness values by day-of-year (DOY) for each plot-year (n = 319), with parameters describing peak greenness (a), DOY at peak (b), and seasonal progression (c). Models with p > 0.05 were excluded (n = 96). To align hemispheric seasons, southern hemisphere DOYs were offset by ±182.5 days as appropriate.