Long-term measurements of ecosystem-atmosphere exchanges of carbon, water, and energy, via eddy flux towers, give insight into three key questions about Amazonian forest function. First, what is the carbon balance of Amazon forests? Some towers give accurate site-specific carbon balances, as validated by independent methods, but decisive resolution of the large-scale question will also require integration of remote sensing techniques (to detect and encompass the distribution of naturally induced disturbance states across the landscape of old growth forests) with eddy flux process studies (to characterize the association between carbon balance and forest disturbance states). Second, what is the seasonality of ecosystem metabolism in Amazonian forests? Models have historically simulated dry season declines in photosynthetic metabolism, a consequence of modeled water limitation. Tower sites in equatorial Amazonian forests, however, show that photosynthetic metabolism increases during dry seasons ("green up"), perhaps because deep roots buffer trees from dry season water stress, while phenological rhythms trigger leaf flush, associated with increased solar irradiance. Third, how does ecosystem metabolism vary across biome types and land use patterns? As dry season length increases from equatorial forest, to drier southern forests, to savanna, fluxes show seasonal patterns consistent with increasing water stress, including a switch from dry season green up to "brown down." Land use change in forest ecosystems removes deep roots, artificially inducing the same trend toward brown down. In the final part, this review suggests that eddy tower network and satellitebased insights into seasonal responses provide a model for detecting responses to extreme interannual climate variations that can test whether forests are vulnerable to model-simulated Amazonian forest collapse under climate change.