Contrasting carbon cycle responses to dry (2015 El Nino) and wet (2008 La Nina) extreme events at an Amazon tropical forest

Land surface models diverge in their predictions of the Amazon forest's response to climate change-induced droughts, with some showing a catastrophic collapse of forests, while others simulating resilience. Therefore, observations of tropical ecosystem responses to real-world droughts and other extreme events are needed. We report long-term seasonal dynamics of photosynthesis, respiration, net carbon exchange, phenology, and tree demography and characterize the effect of dry and wet events on ecosystem form and function at the Tapajos National Forest, Brazil, using over two decades of eddy covariance observations that include the 2015-2016 El Nino drought and La Nina 2008-2009 wet periods. We found strong forest responses to both ENSO events: La Nina saw forest net carbon loss from reduced photosynthesis (due to lower incoming radiation from increased cloudiness) even as ecosystem respiration (Reco) was maintained at mean seasonal levels. El Nino induced the opposite short-term effect, net carbon gains, despite significant reductions in photosynthesis (from a droughtinduced halving of canopy conductance to CO2 and significant losses of leaf area), because drought suppression of Reco losses was even greater. However, long-term responses to the two climate perturbations were very different: transient during La Nina -the forest returned to its "normal" state as soon as the climate did, and longlasting during El Nino -leaf area loss and associated declines in photosynthetic capacity (Pc) and canopy conductance were exacerbated and extended by feedbacks from higher temperatures and atmospheric evaporative demand and persisted for similar to 3+ years after normal rainfall resumed. These findings indicate that these forests are more vulnerable to drought than to excess rain, because drought drives significant changes in forest structure (e.g., leaf-abscission and mortality) and ecosystem function (e.g. reduced stomatal conductance). As future Amazonian climate change increases frequencies of hydrological extremes, these mechanisms will determine the long-term fate of tropical forests.