The risk of a tipping point for Amazon forests — a perturbation threshold beyond which abrupt, irreversible (or difficult to reverse) changes in forest function and large-scale tree die-off occur — motivates much recent Amazon forest science and policy work to understand and reduce the risk. However, virtually all the science to date focuses on tipping points as a basin-wide phenomenon. To understand how large-scale tipping points may be triggered, we urgently need to study mechanisms of tipping point onsets at local-scales in pivotal forests. Here, we used 12+ years of observations (spread over two decades from 2001-2020) of water and energy fluxes from eddy covariance measurements, and associated ecological and meteorological observations in the eastern Amazon basin, to investigate the potential for hydrological extremes to induce a “local scale” tipping point. We focused on forest transpiration capacity (the capacity of vegetation to convert available energy into latent heat, quantified as the ratio of transpiration to incoming radiation, T/R) because transpiration in eastern Amazon forests is the basis for precipitation recycling on which forests to the west depend.Our observations encompassed two strong El Niño droughts, in 2002-2003 and in 2015-2016. Both events were characterized by similarly reduced rainfall; however, the 2015 El Nino was further amplified by an ongoing warming trend, which made for a hotter drought with higher atmospheric vapor demand, exacerbating the drought effects on the forest. This amplification of drought by warming was apparently sufficient to cause widely divergent responses to the two droughts. The forest responded positively to the 2002 drought, with increases in canopy conductance (Gs) and in evapotranspiration, consistent with a stable forest transpiration capacity (T/R) that saw proportional increases in T in response to higher R. By contrast, the transpiration capacity (sustained through two decades of previous dry seasons and the 2002 El Nino drought) collapsed during the 2015 drought. Notably, the forest’s ability to transpire did not return with the rain as the drought ended, but remained low for several years. Thus, we concluded that the difference between the 2002 and 2015 El Nino’s was sufficient to push the forest past a “tipping point” threshold in forest transpiration function, into an alternate state of reduced function in which it remained trapped until the forest could regenerate the canopy with new leaves. This discovery of the phenomenon of local-scale short-term tipping point dynamics in forest canopy function opens the door to investigating and understanding how basin-scale tipping points may emerge from local phenomena, and how careful local observations may provide early warnings of impending larger-scale forest vulnerability.

Forest structure plays an important role in determining habitat suitability for plants and animals, but these relationships are poorly characterized for different biological communities in tropical forests. We used ground-based lidar to quantify structural metrics and determine their contribution in predicting species diversity and compositional changes between plots for nine biological groups in an Amazonian forest. For each group, we calculated Fisher’s alpha index and summarized community composition using Principal Coordinates Analysis. As biological organisms may also react directly to hydro-edaphic conditions, we carried out variance partitioning analysis using linear regressions to disentangle the relative contribution of structural metrics and hydro-edaphic variables. Forest structure was related to species diversity and composition of some groups, specifically for plants, anurans, and birds. Mean canopy height, leaf area height volume, and skewness explained more than one-third of species diversity of palms and trees, with higher values relating to higher species diversity. Hydro-edaphic variables were the most important predictors of the main compositional axis for plant groups, but some structural metrics explained more than 30

Amazon forests are characterized by rich structural diversity. However, the influence of factors such as topography, soil attributes, and external disturbances on structural variability is not always well characterized, and traditional structural metrics may be inadequate to capture this type of complexity. While LiDAR offers expanded structural insights, traditional parameters used in LiDAR analysis, such as mean or maximum canopy height, are not always well directly linked to environmental variables like topography. Emerging approaches merge LiDAR with machine learning to uncover deeper structural complexities. However, work to date may fail to fully utilize the potential of fine-scale LiDAR information. Here we introduce a novel approach, leveraging 2D point cloud images derived from a profiling canopy LiDAR (PCL). The technique targets intricate details within LiDAR point clouds by using deep learning algorithms. With a dataset from the Central Amazon comprising 18 multitemporal transects of 450 m in length, our objective was to detect structural "fingerprints" of varied topographical types along a hillslope, comprising: Riparian, White-sand, and Plateau, and to detect any gradient of structural shifts based on terrain variations here represented by the height above the nearest drainage (HAND). The dataset was trained and tested using a leave-one-group-out approach (LOGO) in which, for each iteration, a complete 450 m multitemporal transect was excluded from training and tested after each iteration. The fast.ai platform and a ResNet-34 architecture, coupled with transfer learning, were used to perform a classification to distinguish between three topographical types. Furthermore, a hybrid model combining a Convolutional Autoencoder, and Partial Least Square (PLS) regression was designed to detect forest structural gradient correlations with HAND variation. Cross-validation achieved a promising high weighted F1 score of 0.83 to classify forests based on the topographical types. Additionally, a combined Convolutional Autoencoder and PLS regression revealed a strong correlation (R-2 = 0.76) between actual and predicted HAND. Innovatively combining deep learning with groundbased PCL LiDAR, our study revealed unique Amazon Forest structures connected to topographic variation. Our findings underscore the transformative potential of such integrative approaches for investigating forest dynamics and promise a powerful new tool for understanding climate-related forest structure change.

Vegetation growth is affected by past growth rates and climate variability. However, the impacts of vegetation growth carryover (VGC; biotic) and lagged climatic effects (LCE; abiotic) on tree stem radial growth may be decoupled from photosynthetic capacity, as higher photosynthesis does not always translate into greater growth. To assess the interaction of tree-species level VGC and LCE with ecosystem-scale photosynthetic processes, we utilized tree-ring width (TRW) data for three tree species: Castanopsis eyrei (CE), Castanea henryi (CH, Chinese chinquapin), and Liquidambar formosana (LF, Chinese sweet gum), along with satellite-based data on canopy greenness (EVI, enhanced vegetation index), leaf area index (LAI), and gross primary productivity (GPP). We used vector autoregressive models, impulse response functions, and forecast error variance decomposition to analyze the duration, intensity, and drivers of VGC and of LCE response to precipitation, temperature, and sunshine duration. The results showed that at the tree-species level, VGC in TRW was strongest in the first year, with an average 77% reduction in response intensity by the fourth year. VGC and LCE exhibited species-specific patterns; compared to CE and CH (diffuse-porous species), LF (ring-porous species) exhibited stronger VGC but weaker LCE. For photosynthetic capacity at the ecosystem scale (EVI, LAI, and GPP), VGC and LCE occurred within 96 days. Our study demonstrates that VGC effects play a dominant role in vegetation function and productivity, and that vegetation responses to previous growth states are decoupled from climatic variability. Additionally, we discovered the possibility for tree-ring growth to be decoupled from canopy condition. Investigating VGC and LCE of multiple indicators of vegetation growth at multiple scales has the potential to improve the accuracy of terrestrial global change models.

Tropical forest photosynthesis can decline at high temperatures due to (1) biochemical responses to increasing temperature and (2) stomatal responses to increasing vapor pressure deficit (VPD), which is associated with increasing temperature. It is challenging to disentangle the influence of these two mechanisms on photosynthesis in observations, because temperature and VPD are tightly correlated in tropical forests. Nonetheless, quantifying the relative strength of these two mechanisms is essential for understanding how tropical gross primary production (GPP) will respond to climate change, because increasing atmospheric CO2 concentration may partially offset VPD-driven stomatal responses, but is not expected to mitigate the effects of temperature-driven biochemical responses. We used two terrestrial biosphere models to quantify how physiological process assumptions (photosynthetic temperature acclimation and plant hydraulic stress) and functional traits (e.g., maximum xylem conductivity) influence the relative strength of modeled temperature versus VPD effects on light-saturated GPP at an Amazonian forest site, a seasonally dry tropical forest site, and an experimental tropical forest mesocosm. By simulating idealized climate change scenarios, we quantified the divergence in GPP predictions under model configurations with stronger VPD effects compared with stronger direct temperature effects. Assumptions consistent with stronger direct temperature effects resulted in larger GPP declines under warming, while assumptions consistent with stronger VPD effects resulted in more resilient GPP under warming. Our findings underscore the importance of quantifying the role of direct temperature and indirect VPD effects for projecting the resilience of tropical forests in the future, and demonstrate that the relative strength of temperature versus VPD effects in models is highly sensitive to plant functional parameters and structural assumptions about photosynthetic temperature acclimation and plant hydraulics.

In Amazonian tropical forests, seasonal photosynthetic activity is influenced by meteorology and leaf phenological cycles. Nevertheless, our understanding of the relationship between these drivers is limited. Do all tropical forests exhibit similar ecosystem responses to rainfall and irradiance? How do soil and tree-crown functional characteristics (composition of leaf lifespans and emergence/abscission times) influence phenological cycles? Here, we compared two rainforests located in central (Tapajós-K67) and southern (Jaru-RJA) Amazonia that share similar seasonality in climatic drivers such as annual precipitation and dry-season length, however different temperature, irradiance, and carbon exchange seasonal cycles. We used phenocameras to characterize canopy phenology, hypothesizing that variations in leaf-flush and senescence mediated the different functional relations between climate and carbon cycling. At both sites, maximum litter-fall occurred at the onset of the wet period, while canopy greenness peaked at the end of the dry-season -- suggesting similar upper-canopy leaf allocation and turnover strategies. At Tapajós, photosynthetic activity (GEP) and leaf area (LAI) increased as the dry-season progressed, whereas at Jaru, they declined. These seasonal patterns were associated with differences in (1) soil depth, with greater water storage capacity in the Tapajós; (2) leaf area density profiles (Tapajós exhibited a higher gap fraction, resulting in a larger number of low-height individuals, while Jaru's leaf area was concentrated closer to the crowns); and (3) the timing of crown green-up, with Tapajós experiencing leaf-out within a few days, whereas a less synchronized leaf flush occurred at Jaru (possibly due to contrasting tree species functional diversity, with K67 exhibiting more homogenous canopy traits than RJA). At forest level, young canopy leaves correlated with high radiation and albedo. However, photosynthesis-leaf age relationships exhibited hysteresis, complicating direct regressions. Our findings contribute to a better understanding of how ecosystems respond to meteorology and the importance of phenology on tropical forests' biogeochemical cycles.

Rising temperatures are influencing forests on many scales, with potentially strong variation vertically across forest strata. Using published research and new analyses, we evaluate how microclimate and leaf temperatures, traits, and gas exchange vary vertically in forests, shaping tree, and ecosystem ecology. In closed-canopy forests, upper canopy leaves are exposed to the highest solar radiation and evaporative demand, which can elevate leaf temperature (Tleaf ), particularly when transpirational cooling is curtailed by limited stomatal conductance. However, foliar traits also vary across height or light gradients, partially mitigating and protecting against the elevation of upper canopy Tleaf . Leaf metabolism generally increases with height across the vertical gradient, yet differences in thermal sensitivity across the gradient appear modest. Scaling from leaves to trees, canopy trees have higher absolute metabolic capacity and growth, yet are more vulnerable to drought and damaging Tleaf than their smaller counterparts, particularly under climate change. By contrast, understory trees experience fewer extreme high Tleaf 's but have fewer cooling mechanisms and thus may be strongly impacted by warming under some conditions, particularly when exposed to a harsher microenvironment through canopy disturbance. As the climate changes, integrating the patterns and mechanisms reviewed here into models will be critical to forecasting forest-climate feedback.

Amazon forests are being degraded by myriad anthropogenic disturbances, altering ecosystem and climate function. We analyzed the effects of a range of land‐use and climate‐change disturbances on fine‐scale canopy structure using a large database of profiling canopy lidar collected from disturbed and mature Amazon forest plots. At most of the disturbed sites, surveys were conducted 10–30 years after disturbance, with many exhibiting signs of recovery. Structural impacts differed in magnitude more than in character among disturbance types, producing a gradient of impacts. Structural changes were highly coordinated in a manner consistent across disturbance types, indicating commonalities in regeneration pathways. At the most severely affected site – burned igapó (seasonally flooded forest) – no signs of canopy regeneration were observed, indicating a sustained alteration of microclimates and consequently greater vulnerability to transitioning to a more open‐canopy, savanna‐like state. Notably, disturbances rarely shifted forests beyond the natural background of structural variation within mature plots, highlighting the similarities between anthropogenic and natural disturbance regimes, and indicating a degree of resilience among Amazon forests. Studying diverse disturbance types within an integrated analytical framework builds capacity to predict the risk of degradation‐driven forest transitions.

Amazonia contains the most extensive tropical forests on Earth, but the role of the region as a carbon appears to be declining. Increasing deforestation, fire and climate change-associated increases in drought, threaten to push forests past a tipping point. However, forests are complex, exhibiting drought responses indicative of both resilience (photosynthetic “greening”) and vulnerability (browning and tree mortality) that are difficult to explain by climate variation alone. Still needed is a framework for understanding and predicting how different regions will respond to different kinds of future drought. Here, we combine remotely-sensed photosynthetic vegetation indices (enhanced vegetation index, EVI, corrected for sun-sensor geometry; and solar-induced chlorophyll fluorescence, SIF) with ground-based tree demography to test recent ecological hypotheses about forest drought resilience and vulnerability for different forest ecotypes across the basin, defined by their water-table depth, soil fertility and vegetation characteristics. In high-fertility southern Amazonia, drought response was importantly structured by water-table depth, with resilient greening in shallow-water-table-forests (where greater water availability heightened responsiveness to excess sunlight) contrasting with vulnerability (“browning” and excess tree mortality) over deeper water tables. Notably, shallow-water-table-forest resilience weakened as droughts lengthened. By contrast, low-fertility northern Amazonia, with slower-growing but drought-hardy trees (or tall trees, with deep-rooted water access), supported more drought-resilient forests independent of water-table depth. This work reveals a new biogeography of forest drought response that provides a framework for conservation decisions and improved prediction of heterogeneous forest responses to future climate changes, but warns that longer/more frequent droughts undermine these multiple ecohydrological strategies of Amazon forest resilience.

Predictions of the magnitude and timing of leaf phenology in Amazonian forests remain highly controversial. Here, we use terrestrial LiDAR surveys every two weeks spanning wet and dry seasons in Central Amazonia to show that plant phenology varies strongly across vertical strata in old-growth forests, but is sensitive to disturbances arising from forest fragmentation. In combination with continuous microclimate measurements, we find that when maximum daily temperatures reached 35 °C in the latter part of the dry season, the upper canopy of large trees in undisturbed forests lost plant material. In contrast, the understory greened up with increased light availability driven by the upper canopy loss, alongside increases in solar radiation, even during periods of drier soil and atmospheric conditions. However, persistently high temperatures in forest edges exacerbated the upper canopy losses of large trees throughout the dry season, whereas the understory in these light-rich environments was less dependent on the altered upper canopy structure. Our findings reveal a strong influence of edge effects on phenological controls in wet forests of Central Amazonia.

Tropical forest function is of global significance to climate change responses, and critically determined by water availability patterns. Groundwater is tightly related to soil water through the water table depth (WT), but historically neglected in ecological studies. Shallow WT forests (WT < 5 m) are underrepresented in forest research networks and absent in eddy flux measurements, although they represent c. 50% of the Amazon and are expected to respond differently to global-change-related droughts. We review WT patterns and consequences for plants, emerging results, and advance a conceptual model integrating environment and trait distributions to predict climate change effects. Shallow WT forests have a distinct species composition, with more resource-acquisitive and hydrologically vulnerable trees, shorter canopies and lower biomass than deep WT forests. During 'normal' climatic years, shallow WT forests have higher mortality and lower productivity than deep WT forests, but during moderate droughts mortality is buffered and productivity increases. However, during severe drought, shallow WT forests may be more sensitive due to shallow roots and drought-intolerant traits. Our evidence supports the hypothesis of neglected shallow WT forests being resilient to moderate drought, challenging the prevailing view of widespread negative effects of climate change on Amazonian forests that ignores WT gradients, but predicts they could collapse under very strong droughts.

Predictions of the magnitude and timing of leaf phenology in Amazonian forests remain highly controversial, which limits our understanding of future ecosystem function with a changing environment. Here, we use biweekly terrestrial LiDAR surveys spanning wet and dry seasons in Central Amazonia to show that plant phenology of old-growth forests varies strongly across strata but that this seasonality is sensitive to disturbances arising from forest fragmentation. In combination with continuous microclimate measurements, we found that when maximum daily temperatures reached 35 °C in the latter part of the dry season, the upper canopy of large trees in undisturbed forests shed their leaves and branches. By contrast, the understory greens-up with increased light availability driven by the upper canopy loss alongside more sunlight radiation, even during periods of drier soil and atmospheric conditions. However, persistently high temperatures on forest edges exacerbated the upper canopy losses of large trees throughout the dry season, and the understory seasonality in these light-rich environments was disrupted as a result of the altered canopy structure. These findings demonstrate the plant-climate interactions controlling the seasonality of wet Amazonian forests and show that forest fragmentation will aggravate forest loss under a hotter and drier future scenario.

Abstract Tropical ecosystems are undergoing unprecedented rates of degradation from deforestation, fire, and drought disturbances. The collective effects of these disturbances threaten to shift large portions of tropical ecosystems such as Amazon forests into savanna‐like structure via tree loss, functional changes, and the emergence of fire (savannization). Changes from forest states to a more open savanna‐like structure can affect local microclimates, surface energy fluxes, and biosphere–atmosphere interactions. A predominant type of ecosystem state change is the loss of tree cover and structural complexity in disturbed forest. Although important advances have been made contrasting energy fluxes between historically distinct old‐growth forest and savanna systems, the emergence of secondary forests and savanna‐like ecosystems necessitates a reframing to consider gradients of tree structure that span forest to savanna‐like states at multiple scales. In this Innovative Viewpoint, we draw from the literature on forest–grassland continua to develop a framework to assess the consequences of tropical forest degradation on surface energy fluxes and canopy structure. We illustrate this framework for forest sites with contrasting canopy structure that ranges from simple, open, and savanna‐like to complex and closed, representative of tropical wet forest, within two climatically distinct regions in the Amazon. Using a recently developed rapid field assessment approach, we quantify differences in cover, leaf area vertical profiles, surface roughness, albedo, and energy balance partitioning between adjacent sites and compare canopy structure with adjacent old‐growth forest; more structurally simple forests displayed lower net radiation. To address forest–atmosphere feedback, we also consider the effects of canopy structure change on susceptibility to additional future disturbance. We illustrate a converse transition—recovery in structure following disturbance—measuring forest canopy structure 10 yr after the imposition of a 5‐yr drought in the ground‐breaking Seca Floresta experiment. Our approach strategically enables rapid characterization of surface properties relevant to vegetation models following degradation, and advances links between surface properties and canopy structure variables, increasingly available from remote sensing. Concluding, we hypothesize that understanding surface energy balance and microclimate change across degraded tropical forest states not only reveals critical atmospheric forcing, but also critical local‐scale feedbacks from forest sensitivity to additional climate‐linked disturbance.

Tropical forests may be vulnerable to climate change1–3 if photosynthetic carbon uptake currently operates near a high temperature limit4–6. Predicting tropical forest function requires understanding the relative contributions of two mechanisms of high-temperature photosynthetic declines: stomatal limitation (H1), an indirect response due to temperature-associated changes in atmospheric vapour pressure deficit (VPD)7, and biochemical restrictions (H2), a direct temperature response8,9. Their relative control predicts different outcomes—H1 is expected to diminish with stomatal responses to future co-occurring elevated atmospheric [CO2], whereas H2 portends declining photosynthesis with increasing temperatures. Distinguishing the two mechanisms at high temperatures is therefore critical, but difficult because VPD is highly correlated with temperature in natural settings. We used a forest mesocosm to quantify the sensitivity of tropical gross ecosystem productivity (GEP) to future temperature regimes while constraining VPD by controlling humidity. We then analytically decoupled temperature and VPD effects under current climate with flux-tower-derived GEP trends in situ from four tropical forest sites. Both approaches showed consistent, negative sensitivity of GEP to VPD but little direct response to temperature. Importantly, in the mesocosm at low VPD, GEP persisted up to 38 °C, a temperature exceeding projections for tropical forests in 2100 (ref. 10). If elevated [CO2] mitigates VPD-induced stomatal limitation through enhanced water-use efficiency as hypothesized9,11, tropical forest photosynthesis may have a margin of resilience to future warming. Photosynthesis in tropical forests shows an apparent sensitivity to temperature. This Letter teases apart the effects of temperature and correlated atmospheric water demand on ecosystem productivity.

Experimental research shows that isoprene emission by plants can improve photosynthetic performance at high temperatures. But whether species that emit isoprene have higher thermal limits than non-emitting species remains largely untested. Tropical plants are adapted to narrow temperature ranges and global warming could result in significant ecosystem restructuring due to small variations in species' thermal tolerances. We compared photosynthetic temperature responses of 26 co-occurring tropical tree and liana species to test whether isoprene-emitting species are more tolerant to high temperatures. We classified species as isoprene emitters versus non-emitters based on published datasets. Maximum temperatures for net photosynthesis were ~1.8°C higher for isoprene-emitting species than for non-emitters, and thermal response curves were 24% wider; differences in optimum temperatures (Topt ) or photosynthetic rates at Topt were not significant. Modelling the carbon cost of isoprene emission, we show that even strong emission rates cause little reduction in the net carbon assimilation advantage over non-emitters at supraoptimal temperatures. Isoprene emissions may alleviate biochemical limitations, which together with stomatal conductance, co-limit photosynthesis above Topt . Our findings provide evidence that isoprene emission may be an adaptation to warmer thermal niches, and that emitting species may fare better under global warming than co-occurring non-emitting species.

Assessing the persistent impacts of fragmentation on aboveground structure of tropical forests is essential to understanding the consequences of land use change for carbon storage and other ecosystem functions. We investigated the influence of edge distance and fragment size on canopy structure, aboveground woody biomass (AGB), and AGB turnover in the Biological Dynamics of Forest Fragments Project (BDFFP) in central Amazon, Brazil, after 22+ yr of fragment isolation, by combining canopy variables collected with portable canopy profiling lidar and airborne laser scanning surveys with long-term forest inventories. Forest height decreased by 30% at edges of large fragments (>10 ha) and interiors of small fragments (<3 ha). In larger fragments, canopy height was reduced up to 40 m from edges. Leaf area density profiles differed near edges: the density of understory vegetation was higher and midstory vegetation lower, consistent with canopy reorganization via increased regeneration of pioneers following post-fragmentation mortality of large trees. However, canopy openness and leaf area index remained similar to control plots throughout fragments, while canopy spatial heterogeneity was generally lower at edges. AGB stocks and fluxes were positively related to canopy height and negatively related to spatial heterogeneity. Other forest structure variables typically used to assess the ecological impacts of fragmentation (basal area, density of individuals, and density of pioneer trees) were also related to lidar-derived canopy surface variables. Canopy reorganization through the replacement of edge-sensitive species by disturbance-tolerant ones may have mitigated the biomass loss effects due to fragmentation observed in the earlier years of BDFFP. Lidar technology offered novel insights and observational scales for analysis of the ecological impacts of fragmentation on forest structure and function, specifically aboveground biomass storage.

Plant phenology-the timing of cyclic or recurrent biological events in plants-offers insight into the ecology, evolution, and seasonality of plant-mediated ecosystem processes. Traditionally studied phenologies are readily apparent, such as flowering events, germination timing, and season-initiating budbreak. However, a broad range of phenologies that are fundamental to the ecology and evolution of plants, and to global biogeochemical cycles and climate change predictions, have been neglected because they are "cryptic"-that is, hidden from view (e.g., root production) or difficult to distinguish and interpret based on common measurements at typical scales of examination (e.g., leaf turnover in evergreen forests). We illustrate how capturing cryptic phenology can advance scientific understanding with two case studies: wood phenology in a deciduous forest of the northeastern USA and leaf phenology in tropical evergreen forests of Amazonia. Drawing on these case studies and other literature, we argue that conceptualizing and characterizing cryptic plant phenology is needed for understanding and accurate prediction at many scales from organisms to ecosystems. We recommend avenues of empirical and modeling research to accelerate discovery of cryptic phenological patterns, to understand their causes and consequences, and to represent these processes in terrestrial biosphere models.