Correct understanding and modelling of terrestrial carbon cycles is a critical first step for accurate prediction of large-scale climate variability and change, which impacts the lives and living quality of all organisms of our planet, including us human beings. However, considerable uncertainties remain, particularly associated with key model parameters and processes that determine the strengths of land plants to interact with current and changing climate. With such uncertainties, there is no doubt that current Terrestrial Biosphere Models (TBMs) fail to adequately capture the magnitude, spatial variation and seasonality of terrestrial carbon uptake. Here, with tropical evergreen forest ecosystems as an example study system, I show how the integration of multi-scale remote sensing with field observations can help 1) reveal the cryptic phenology pattern (“landscape evergreen, individual dynamics”) in Amazonian evergreen forests, 2) improve process understanding of the phenology-physiology relationships in the tropics, and 3) advance modelling skills to better represent tropical photosynthesis seasonality in TBM-type land models. Collectively, I hope to show that such an integrated approach, which connects multi-scale observations with ecological theories and models, can be an important way towards a more predictive future of terrestrial ecosystems with climate change.