Abstract :
[en] While phenology is an essential biodiversity variable to monitor ecosystem response to
climate change, land surface phenology (satellite observations) is difficult to implement in
cloudy tropical regions and field phenology (crown observations) is generally limited to a few species monitored at low frequency, (bi)monthly. Time-lapse phenological cameras, or PhenoCams, offer new opportunities that were barely explored in Africa, allowing the
estimation of the frequency, length and intensity of canopy deciduousness, an important
mechanism for canopy trees to cope with seasonal drought. Here, we present the early results of a PhenoCam network installed in the forest-savanna mosaic of Lopé NP, Gabon, and the analytical framework developed for deciduousness detection. Regions of Interest (ROI), including individual tree crowns and forest/savanna extents, were manually digitized for a reference image used to align all daily images. For each ROI, phenological cycles were estimated using variations in the Green Chromatic Index (GCC) and Red Chromatic Index (RCC) among the daily images. The three years of images taken by three PhenoCams show that forest canopy seasonal functioning is bimodal, mirroring rainfall seasonality, with two peaks in the GCC corresponding to rainy seasons. Conversely, savanna shows a unimodal pattern, with a minimum GCC during the long dry season. Leaf loss and leaf growth in the tree crowns are brief events, typically lasting less than 10 days for most deciduous species. We detected more leaf flush or greening events than leaf loss events indicating leaf renewal without leaf loss. This work is encouraging for long-term monitoring of tree phenology, although the challenge is to maintain PhenoCams over time. In terms of analytical perspectives, we aim to enhance phenocam data analysis for better ecological dynamics understanding.
