Canopy deciduousness and leaf phenological strategies in Central Africa explored with phenocams

Canopy deciduousness and leaf phenological strategies remain largely unexplored in central Africa though their highly seasonal functioning. Field observations conducted at a few sites over the past decades are time-consuming, costly, and often miss short episodes of leaf loss, flush, or color change. Here, we present findings from the first phenocam-equipped site in central Africa, providing continuous, high-frequency monitoring of canopy dynamics in Lopé National Park, Gabon. Using 4,221 images collected by two phenocams between 2019 and 2023, we characterized leaf phenology at the individual crown level for three easily identified canopy tree species-Ceiba pentandra, Lophira alata and Aucoumea klaineana-that exhibit contrasting phenological strategies. We tested how rainfall and irradiance influence the leaf loss and flush. Analysis from image annotations revealed diverse strategies, ranging from full deciduousness in C. pentandra, triggered by low rainfall, to evergreen persistence in A. klaineana, with L. alata displaying a wet season deciduous strategy. We further explored the potential of signal analysis approaches, specifically Hidden Markov Models (HMMs), to automatically identify phenological phases and transitions from phenocam image time series. HMMs are particularly suited to underlying hidden processes - mirroring the way ecological transitions unfold progressively based on both current conditions and prior states. Our approach, combining high-frequency imaging with probabilistic modeling, paves the way for automated tools for the monitoring of tropical forest phenology in a rapidly changing climate.