17Drought threatens tropical rainforests over seasonal to decadal timescales [1][2][3][4] , but the drivers 18 of tree mortality following drought remain poorly understood 5,6 . It has been suggested that 19 reduced availability of non-structural carbohydrates (NSC) critically increases mortality risk 20 through insufficient carbon supply to metabolism ('carbon starvation') 7,8 . However little is 21 known about how NSC stores are affected by drought, especially over the long term, and 22 whether they are more important than hydraulic processes in determining drought-induced 23 mortality. Using data from the world's longest-running experimental drought study in tropical 24 rainforest (in the Brazilian Amazon), we test whether carbon starvation or deterioration of the 25 water-conducting pathways from soil to leaf trigger tree mortality. Biomass loss from 26 mortality in the experimentally-droughted forest increased substantially after >10 years of 27 reduced soil moisture availability. The mortality signal was dominated by the death of large 28 trees, which were at a much greater risk of hydraulic deterioration than smaller trees. 29However, we find no evidence that the droughted trees suffered carbon starvation, as their 30 NSC concentrations were similar to those of un-droughted trees, and growth rates did not 31 decline in either living or dying individuals. Our results indicate that hydraulics, rather than 32 carbon starvation, triggers tree death from drought in tropical rainforest. 34Drought-response observations from both field-scale experiments and natural droughts have 35 demonstrated increased mortality over the short-term (1-3 years), with notably higher 36 vulnerability for some taxa, and for larger trees 6,9,10 . After several years of drought, 37 recovering growth rates in smaller trees, dbh (diameter at breast height) <40 cm, and reduced 38 mortality have been recorded at different locations 6,11,12 . However, the long-term (>10 yr) 39 sensitivity of tropical forests to predicted prolonged and repeated water deficit [1][2][3] we synthesise these data to test whether long-term soil moisture deficit alters NSC storage 64 and use in tropical rainforest trees, and if this, or hydraulic processes, are most strongly 65 associated with increased mortality rates. 66By 2014, following 13 years of the TFE treatment, cumulative biomass loss through mortality 67 was 41.0±2.7% relative to pre-treatment values (Fig. 1a), and the rate of loss had increased 68 substantially since the previous reported value of 17.2±0.8%, after 7 years of TFE 6 . 69Accelerating biomass loss and failure to recover substantially, or to reach a new 70 equilibrium 13 , has led to a committed flux to the atmosphere from decomposing necromass of 71 101.9±19.1 Mg C ha -1 (Fig. 1a). This biomass loss has been driven by elevated mortality in 72 the largest trees (Fig. 1b), as previously observed over shorter timescales 6 , and has created a 73 canopy that has had a persistently lower average leaf area index during 2010-2014 74 (12.0±1...

Abstract. Forest ecosystem models based on heuristic water stress functions poorly predict tropical forest response to drought partly because they do not capture the diversity of hydraulic traits (including variation in tree size) observed in tropical forests. We developed a continuous porous media approach to modeling plant hydraulics in which all parameters of the constitutive equations are biologically interpretable and measurable plant hydraulic traits (e.g., turgor loss point πtlp, bulk elastic modulus ε, hydraulic capacitance Cft, xylem hydraulic conductivity ks,max, water potential at 50 % loss of conductivity for both xylem (P50,x) and stomata (P50,gs), and the leaf : sapwood area ratio Al : As). We embedded this plant hydraulics model within a trait forest simulator (TFS) that models light environments of individual trees and their upper boundary conditions (transpiration), as well as providing a means for parameterizing variation in hydraulic traits among individuals. We synthesized literature and existing databases to parameterize all hydraulic traits as a function of stem and leaf traits, including wood density (WD), leaf mass per area (LMA), and photosynthetic capacity (Amax), and evaluated the coupled model (called TFS v.1-Hydro) predictions, against observed diurnal and seasonal variability in stem and leaf water potential as well as stand-scaled sap flux. Our hydraulic trait synthesis revealed coordination among leaf and xylem hydraulic traits and statistically significant relationships of most hydraulic traits with more easily measured plant traits. Using the most informative empirical trait–trait relationships derived from this synthesis, TFS v.1-Hydro successfully captured individual variation in leaf and stem water potential due to increasing tree size and light environment, with model representation of hydraulic architecture and plant traits exerting primary and secondary controls, respectively, on the fidelity of model predictions. The plant hydraulics model made substantial improvements to simulations of total ecosystem transpiration. Remaining uncertainties and limitations of the trait paradigm for plant hydraulics modeling are highlighted.

Abstract. Forest ecosystem models based on heuristic water stress functions poorly predict tropical forest response to drought because they do not capture the diversity of hydraulic traits (including variation in tree size) observed in tropical forests. We developed a Richards’ equation-based model of plant hydraulics in which all parameters of its constitutive equations are biologically-interpretable and measureable plant hydraulic traits (e.g., turgor loss point πtlp, bulk elastic modulus ε, hydraulic capacitance Cft, xylem hydraulic conductivity ks,max, water potential at 50 % loss of conductivity for both xylem (P50,x) and stomata (P50,gs), and the leaf:sapwood area ratio Al:As). We embedded this plant hydraulics model within a forest simulator (TFS) that modeled individual tree light environments and their upper boundary condition (transpiration) as well as provided a means for parameterizing individual variation in hydraulic traits. We synthesized literature and existing databases to parameterize all hydraulic traits as a function of stem and leaf traits wood density (WD), leaf mass per area (LMA) and photosynthetic capacity (Amax) and evaluated the coupled model’s (TFS-Hydro) predictions against diurnal and seasonal variability in stem and leaf water potential as well as stand-scaled sap flux. Our hydraulic trait synthesis revealed coordination among leaf and xylem hydraulic traits and statistically significant relationships of most hydraulic traits with more easily measured plant traits. Using the most informative empirical trait-trait relationships derived from this synthesis, the TFS-Hydro model parameterization is capable of representing patterns of coordination and trade-offs in hydraulic traits. TFS-Hydro successfully captured individual variation in leaf and stem water potential due to increasing tree size and light environment, with model representation of hydraulic architecture and plant traits exerting primary and secondary controls, respectively, on the fidelity of model predictions. The plant hydraulics model made substantial improvements to simulations of total ecosystem transpiration under control conditions, but the absence of a vertically stratified soil hydrology model precluded improvements to the simulation of drought response. Remaining uncertainties and limitations of the trait paradigm for plant hydraulics modeling are highlighted.

Determining climate change feedbacks from tropical rainforests requires an understanding of how carbon gain through photosynthesis and loss through respiration will be altered. One of the key changes that tropical rainforests may experience under future climate change scenarios is reduced soil moisture availability. In this study we examine if and how both leaf photosynthesis and leaf dark respiration acclimate following more than 12 years of experimental soil moisture deficit, via a through‐fall exclusion experiment (TFE) in an eastern Amazonian rainforest. We find that experimentally drought‐stressed trees and taxa maintain the same maximum leaf photosynthetic capacity as trees in corresponding control forest, independent of their susceptibility to drought‐induced mortality. We hypothesize that photosynthetic capacity is maintained across all treatments and taxa to take advantage of short‐lived periods of high moisture availability, when stomatal conductance (g s) and photosynthesis can increase rapidly, potentially compensating for reduced assimilate supply at other times. Average leaf dark respiration (R d) was elevated in the TFE‐treated forest trees relative to the control by 28.2 ± 2.8% (mean ± one standard error). This mean R d value was dominated by a 48.5 ± 3.6% increase in the R d of drought‐sensitive taxa, and likely reflects the need for additional metabolic support required for stress‐related repair, and hydraulic or osmotic maintenance processes. Following soil moisture deficit that is maintained for several years, our data suggest that changes in respiration drive greater shifts in the canopy carbon balance, than changes in photosynthetic capacity.

Summary The tropics are predicted to become warmer and drier, and understanding the sensitivity of tree species to drought is important for characterizing the risk to forests of climate change. This study makes use of a long‐term drought experiment in the Amazon rainforest to evaluate the role of leaf‐level water relations, leaf anatomy and their plasticity in response to drought in six tree genera.The variables (osmotic potential at full turgor, turgor loss point, capacitance, elastic modulus, relative water content and saturated water content) were compared between seasons and between plots (control and through‐fall exclusion) enabling a comparison between short‐ and long‐term plasticity in traits. Leaf anatomical traits were correlated with water relation parameters to determine whether water relations differed among tissues.The key findings were: osmotic adjustment occurred in response to the long‐term drought treatment; species resistant to drought stress showed less osmotic adjustment than drought‐sensitive species; and water relation traits were correlated with tissue properties, especially the thickness of the abaxial epidermis and the spongy mesophyll.These findings demonstrate that cell‐level water relation traits can acclimate to long‐term water stress, and highlight the limitations of extrapolating the results of short‐term studies to temporal scales associated with climate change.