The seasonal variability in carbon fluxes between tropical rainforests and atmosphere is still poorly understood despite their crucial contribution to global biogeochemical cycles. In particular, dynamic global vegetation models (DGVMs) typically simulate a decrease in productivity during the dry season, while observations from eddy‐covariance data in light‐limited rainforests point to a dry‐season increase in gross primary productivity. Some previous studies highlighted the importance of the variation of leaf properties (e.g. leaf photosynthetic capacity) with leaf age and leaf demography to explain fluxes seasonality, but data is critically lacking to develop evidence-based knowledge on the underlying mechanisms and allow their integration into predictive models. In this project, we propose to (i) substantially increase the data availability on the variation of leaf properties with leaf age for Amazonian trees by benefitting from a characterization of the diversity of phenological patterns within French Guyanan forests ; (ii) extend it to physiological traits, such as the leaf water potential at turgor loss point, which has been related to leaf shedding, but, to our best knowledge, never measured on various leaf ages; (iii) upscale such data at the community scale and use available eddy-covariance data on fluxes and remotely-sensed data on leaf quantity to tease apart the relative contribution of climate, leaf quantity and leaf properties in explaining the seasonality of ecosystem fluxes in French Guiana; and, (iv) assimilate such data into a trait-based and individual-based model of tropical forest dynamics to improve our predictive ability of biogeochemical cycles.