Revisiting mechanisms underlying tree mortality induced by drought in the Amazon: from observation to modeling

Abstract:

In the past decade, two extreme droughts experienced by the Amazon rainforest led to a perturbation of carbon cycle dynamics and forest structure, partly through an increase in tree mortality. While there is a relatively strong consensus in CMIP5 projections for an increase in both frequency and intensity of droughts across the Amazon, the potential for forest die-off constitutes a large uncertainty in projections of climate impacts on terrestrial ecosystems and carbon cycle feedbacks.

Two long-term through fall exclusion experiments (TFE) provided novel observations of Amazonian ecosystem responses under drought. These experiments also provided a great opportunity to evaluate and improve models' behavior under drought by comparing simulations and observations.

While current DGVM use a wide array of algorithms to represent mortality, most are associated with large uncertainty for representing drought-induced mortality, and require updating to include current information of physiological processes. During very strong droughts, the leaves desiccate and stems may undergo catastrophic embolism. However, even before that point, stomata close, to minimize excessive water loss and risk of hydraulic failure, which reduces carbon assimilation. To maintain respiration and other functions, plants may eventually deplete stored non-structural carbon compounds (NSC), which may have negative impacts on plant and eventually increase the probability of mortality.

Here, we describe a new parameterization of the mortality process induced by drought using the ORCHIDEE-CAN dynamic vegetation model and test it using the two TFE results. We first updated and evaluated both the representation of hydraulic architecture and the NSC pool dynamics using in situ data. We implemented a direct climate effect on mortality through catastrophic stem embolism, based on hydraulic vulnerability curves. In addition, we explored the role of NSC on hydraulic failure and mortality by coupling in the model NSC content and vulnerability curves, following the idea that stored NSC serves a critical osmotic function.

Our results suggest that models have the capacity to represent individual mortality from a mechanistic perspective, providing a framework for informing future experiments and data collection for model development.