Carbon–Water Coupling in Forest Ecosystems Under Climate Change: Advances in Water Use Efficiency and Sustainability Perspectives

Abstract

Climate change is reshaping how forests balance carbon uptake and water loss. This review aims to clarify how climate change alters forest carbon–water coupling. Using water-use efficiency (WUE) as a unifying lens, we synthesize mechanisms from leaves to ecosystems and evaluate evidence from studies screened in 2000–2025 spanning eddy covariance, tree-ring isotopes, remote sensing and models. Globally, tree-ring data indicate ~40% intrinsic WUE increases since 1901, yet ecosystem-scale gains are usually <20% after accounting for mesophyll conductance. Under drought, heat and high vapor-pressure deficit, photosynthesis declines more than evapotranspiration, producing partial carbon–water decoupling and lower WUE_e. Responses vary with hydraulic traits, forest type/age and site water balance, with notable tropical data gaps. We identify when WUE gains translate into true resilience: stomatal regulation and canopy structure jointly maintain GPP, prevent hydraulic failure and ensure post-event recovery. Management options include thinning, species/provenance choice, mixed stands and adaptive rotations to balance carbon storage with water yield. Key uncertainties stem from sparse long-term observations, tropical satellite biases and models that overestimate WUE or underplay extremes. We recommend integrating multi-source, multi-scale data with interpretable hybrid models, expanding tropical networks and strengthening MRV frameworks to support risk-aware, climate-smart forestry.

1. Introduction

Human activities have unequivocally driven global warming, with atmospheric carbon dioxide (CO₂) concentrations in 2023 reaching 151% of pre-industrial levels—the highest in over two million years [1]. This escalation is pushing multiple planetary boundaries—including climate and freshwater—beyond their safe operating space, increasing the risk of large-scale, irreversible environmental shifts [2]. Forest ecosystems, covering ~30% of global land area, are central to this planetary crisis [3]. They function as the largest terrestrial carbon sink and a key modulator of hydrological and energy cycles [4,5,6], storing roughly twice the carbon currently in the atmosphere and removing about one-third of annual anthropogenic CO₂ emissions [7,8]. Yet the capacity of forests to provide these critical services is increasingly threatened by the pace and magnitude of ongoing climate change [9,10].

The traditional view of forests as reliable, perpetual carbon sinks is therefore becoming precarious. Although their mitigation value is undisputed and conservation/restoration targets are ambitious [11], climate change is inducing positive feedbacks that can erode sink strength. Rising temperatures, altered precipitation, and more frequent extremes intensify wildfires, insect outbreaks, and drought-induced mortality [12,13,14]. These disturbances reduce carbon storage and, at times, flip forests from net sinks to net sources via combustion and accelerated decomposition [15]. Recent evidence indicates that parts of the southeastern Amazon may already have transitioned to a net carbon source under combined climatic and disturbance pressures [16]. This possibility shifts the scientific imperative from simply quantifying the sink to assessing its stability and resilience, and cautions against treating forests as permanent offsets without aggressive mitigation and robust protection of ecosystem integrity.

At the core of these dynamics lies the tight coupling of the carbon and water cycles through photosynthesis and transpiration [17]. Stomata—microscopic pores on leaf surfaces—regulate gas exchange: opening admits atmospheric CO₂ for fixation but simultaneously increases water-vapor loss [18,19]. This physiological trade-off underpins carbon–water coupling from leaves to ecosystems [20], shaping land–atmosphere feedback by partitioning surface energy into latent and sensible heat, with consequences for surface temperature, humidity, boundary-layer development, and regional precipitation [21]. Via moisture recycling, these interactions propagate downwind and can trigger cascading vegetation–rainfall feedback (Figure 1) [22].

Figure 1. Cascading effects in the vegetation–rainfall system. (a) Vegetation–atmosphere system in equilibrium. (b) Initial forest loss triggered by decreasing oceanic moisture inflow. This reduces local evapotranspiration and the resulting downwind moisture transport. (c) As a result, the rainfall regime is altered in another location, leading to further forest loss and reduced moisture transport. Copyright Nature Communications (2017) [22], CC BY 4.0.

Water-use efficiency (WUE) provides an integrative lens for analyzing this coupling across scales [23]. Defined consistently but with scale-specific metrics—e.g., intrinsic WUE at the leaf scale, canopy-level WUE, and ecosystem-scale WUE (WUE_e)—WUE captures how physiological regulation and structural attributes jointly respond to climate drivers and management [23,24]. Spatiotemporal changes in WUE reveal trade-offs between carbon sequestration and water availability, inform vulnerability to heat and drought, and help anticipate implications for ecosystem services in water-limited regions [25].

Scope and objectives: This review synthesizes advances in carbon–water coupling in forest ecosystems under climate change, using WUE as the unifying metric. We first clarify the mechanistic basis—from stomatal control, mesophyll and hydraulic constraints, and canopy structure to ecosystem fluxes—and how CO₂ enrichment, warming, drought, heatwaves, and high vapor-pressure deficit (VPD) alter coupling. We then evaluate methods for assessing forest WUE, including eddy-covariance and sap-flow observations, tree-ring isotopes, remote sensing (e.g., SIF-based photosynthesis proxies), and process-based as well as hybrid modeling, highlighting strengths, limitations, and key sources of uncertainty. Next, we synthesize global assessments and regional case studies to characterize WUE patterns and their climatic/biotic drivers, including legacy effects after extremes. Building on this evidence, we examine management and policy-relevant trade-offs between carbon storage and water yield, and outline how WUE can guide risk-aware, climate-smart forestry.

Contribution relative to prior reviews: While recent syntheses emphasize either physiological mechanisms or large-scale modeling of WUE and carbon–water coupling [23,24], our review advances the field in three ways: (i) it explicitly links cross-scale mechanisms to climate extremes and compound events; (ii) it delineates conditions under which WUE gains fail to improve ecosystem-scale WUE (e.g., heat-driven “emergency cooling,” structural increases in evaporation fractions, erosion of hydraulic safety margins); and (iii) it translates the synthesis into multi-objective management and governance guidance. Table 1 compares our scope with prior syntheses in terms of scale integration, treatment of extremes, and policy implications.

Table 1. Comparison between this review and previous syntheses on forest WUE and carbon–water coupling.

Study/Review	Scope	Scale Emphasis	Treatment of Climate Extremes	Policy/Management Implications	Distinct Contribution
Zhang et al. (2023) [23]	WUE responses to climate change and management	Mainly physiological and ecosystem	Limited	Discussed management but not water–carbon trade-offs	Valuable synthesis of WUE drivers
Montibeller et al. (2022) [24]	Carbon assimilation vs. water use in European forests	Regional focus (Europe)	Minimal	Policy implications indirect	Highlights European case studies
This Review (2025)	Carbon–water coupling under climate change, with WUE as core metric	Multi-scale (leaf → ecosystem → global)	Explicit focus on droughts, heatwaves, and compound extremes	Strong emphasis on trade-offs between carbon sequestration and water yield; climate-smart forestry strategies	Provides integrative, cross-scale and policy-relevant perspective

To orient readers, Figure 2 contrasts present-day and climate-change regimes. Under present-day conditions, adequate rainfall and moderate temperatures support soil infiltration, root water uptake, and stomatal opening, coupling photosynthesis (GPP) with evapotranspiration (ET) and yielding relatively high WUE_e (WUE_e = GPP/ET). With warming and drying, atmospheric aridity (higher VPD) increases atmospheric demand while soil moisture declines; photosynthesis and growth typically fall faster than ET as evaporation fractions rise, producing partial carbon–water decoupling and lower WUE_e. Arrows indicate directions of change; insets are schematic.

Figure 2. Conceptual shifts in forest carbon–water coupling and water-use efficiency under climate change.

This article is a critical narrative review; no new, unpublished data are presented. Methods (evidence synthesis)—Databases: Web of Science, Scopus, Google Scholar, CNKI. Timeframe: 2000–2025. Queries combined WUE/carbon–water coupling + (forest* OR eddy covariance OR isotope* OR remote sensing OR model*) + (drought OR heatwave OR VPD). Inclusion: studies quantifying WUE (iWUE/WUE_e/uWUE) or carbon–water coupling under climate drivers; extremes explicitly analyzed or implicated; management/policy relevance. Exclusion: non-forest, no WUE metric, commentary only. Screening: dual-review, two-stage (title/abstract → full text), DOI/title deduplication. Extraction: metrics, scale, biome, drivers, methods, uncertainty; double-entry cross-check.

2. Mechanisms of Carbon–Water Coupling in Forests

2.1. Stomatal Regulation: The Nexus of Photosynthesis and Transpiration

Stomata are the primary regulators of gas exchange between plants and the atmosphere, placing them at the nexus of the carbon and water cycles [25]. The aperture of these pores, controlled by a pair of guard cells, responds dynamically to a suite of environmental cues to optimize carbon uptake while minimizing water loss. A key driver of stomatal behavior in the current era is the rising concentration of atmospheric CO₂. Elevated CO₂ generally induces partial stomatal closure, a response observed across a wide range of forest species [18]. A comprehensive meta-analysis of Free Air CO₂ Enrichment (FACE) experiments found that elevated CO₂ reduces stomatal conductance (g_s) by an average of 22% across all plant species [26]. Figure 3 visualizes how rising CO₂ shifts guard-cell behavior and g_s [27]. Panel B contextualizes leaf-level responses within the long-term rise in atmospheric CO₂. Panels C–E show that step increases in ambient CO₂ induce a rapid decline in stomatal conductance (g_s; conductance to water vapor, mmol m⁻² s⁻¹) in wild-type leaves, whereas carbonic-anhydrase mutants display attenuated responses, consistent with CA-mediated CO₂ sensing. Together with Panel A (light–dark shifts in intercellular CO₂), the figure substantiates our claim that elevated CO₂ typically reduces g_s (~22% on average) and increases intrinsic water-use efficiency [27].

Figure 3. CO₂-driven stomatal signaling and closure pathways. (A) Light-driven changes in intercellular CO₂ in Vicia faba at ambient 350 ppm, measured with a potentiometric CO₂ microprobe inserted through open stomata. (B) Atmospheric CO₂ record. (C–E) Effects of CO₂ on gas exchange in carbonic anhydrase mutants: raw stomatal conductance. Copyright Cell (2016) [27].

This response is not universal; its magnitude varies significantly with species, plant functional type, and age. For instance, the reduction in g_s is often stronger in deciduous trees compared to coniferous trees and in younger trees compared to older ones [28]. In addition to this short-term physiological response, long-term exposure to elevated CO₂ can trigger developmental changes, such as a reduction in stomatal density on new leaves, further altering the leaf’s capacity for gas exchange [18,29]. The underlying mechanisms for these responses are complex, involving biochemical signaling pathways where enzymes like carbonic anhydrases play a role in sensing CO₂ changes, and hormonal signals, particularly abscisic acid (ABA), mediate guard cell movements [27].

2.2. Leaf-to-Ecosystem Scaling of Gas Exchange Processes

Translating physiological responses from the leaf level to the entire ecosystem is a significant challenge, as the relationships are complex and nonlinear [30]. A reduction in stomatal conductance at the leaf level does not necessarily scale proportionally in ecosystem-scale evapotranspiration. One major confounding factor is the ecosystem’s structural response to environmental change, particularly changes in the total amount of leaf area per unit ground area (Leaf Area Index, LAI). The CO₂ fertilization effect can stimulate plant growth, leading to an increase in LAI [31], which can offset or even overwhelm the water-saving effect of individual stomata closing.

The physical coupling between the forest canopy and the atmosphere plays a critical role. The efficiency with which turbulence can exchange heat and water vapor between the canopy and the overlying air, known as aerodynamic coupling, influences how changes in stomatal conductance translate to canopy transpiration. For example, aerodynamically well-coupled systems like tall forests may show a more direct relationship between stomatal closure and reduced transpiration, whereas poorly coupled systems like agricultural crops may experience a buildup of humidity within the canopy that dampens the effect of stomatal closure on overall water loss [32]. These scaling complexities explain why WUE measured at the leaf level (iWUE) can exhibit different trends and drivers compared to WUE measured at the ecosystem scale (WUE_e) [30].

2.3. Impacts of Key Climate Drivers (Elevated CO₂, Temperature, VPD) on Physiological Coupling

Forest ecosystems are responding to a suite of concurrent and interacting climate drivers. Elevated CO₂ is the primary driver of increased intrinsic water use efficiency (iWUE) through its effect on stomatal closure [18]. However, this effect does not occur in isolation. Rising temperatures can increase photosynthetic rates up to a thermal optimum but also exponentially increase respiration and the atmospheric demand for water, thus driving higher transpiration rates [20]. The reduction in transpirational cooling that accompanies CO₂-induced stomatal closure can lead to higher leaf temperatures, which may push leaves beyond their thermal optimum and potentially exacerbate heat stress [33].

Vapor Pressure Deficit (VPD), the difference between the amount of moisture in the air and how much moisture the air can hold when saturated, is another powerful driver of stomatal closure. As air warms, its capacity to hold water increases, often leading to a higher VPD, which creates a steeper gradient for water to diffuse out of the leaf [34]. This can induce stomatal closure independently of CO₂ levels. The interplay between these factors is complex. For example, while the general assumption is that stomata close in response to elevated CO₂, some field observations have shown that under hot and dry conditions, stomatal conductance in certain woody species may paradoxically increase with rising CO₂, challenging over-simplified notions of plant response [35]. The net effect on a forest’s carbon-water balance is therefore a result of the dynamic interplay between the water-saving effects of CO₂ and the water-spending effects of warming and increased atmospheric aridity.

Quantitatively, across tower–tree-ring paired sites, uWUE remains effectively stable with VPD up to the site-specific 75th percentile (log–log slope within −0.05 to +0.05) but turns negative at the 90th–95th percentiles. By contrast, many PBMs and ML upscalers retain near-zero slopes across the full VPD distribution, underestimating drought-time sensitivity. For clarity, we term uWUE stable when the slope lies between −0.05 and +0.05 and unstable when it is <−0.05, noting that this unstable, high-VPD regime often coincides with ecosystem-scale WUE_e declines.

2.4. Decoupling Phenomena Under Climatic Extremes

Forest responses to climate change occur in a multi-stressor context in which any potential benefits of elevated CO₂ are offset by constraints imposed by warming and drying. Under moderate forcing, stomatal regulation can preserve a functional balance between carbon uptake and water loss; during heatwaves and severe droughts this coupling frequently breaks down [20]. In intense heat some species express an emergency thermoregulatory mode in which photosynthesis is strongly downregulated to protect the photosynthetic apparatus, while transpiration is maintained or even enhanced to provide evaporative cooling [34]. Under drought a comparable decoupling is evident, with secondary growth, indexed by tree-ring width, typically more sensitive to water stress than ecosystem photosynthesis (GPP), implying a short-term reallocation away from long-lived woody biomass toward metabolic maintenance [36,37].

These nonlinear stress-emergent behaviors are poorly represented in most land-surface and Earth-system models, which commonly impose persistently tight carbon–water coupling [21,34,38], thereby introducing substantial uncertainty in projections under intensifying extremes. Table 2 juxtaposes empirical evidence from FACE, FLUXNET and isotopic records with current model representations and shows a systematic underestimation of nonlinear responses and decoupling during extreme events.

Table 2. Empirical evidence versus model representation of carbon–water decoupling phenomena under climate extremes.

Climate Extreme	Empirical Evidence (FACE, FLUXNET, Isotopes)	Model Representation (PBMs, ESMs)	Key Gap and References
Drought	Stomatal closure increases iWUE but tree growth declines; evidence of GPP–growth decoupling	Many PBMs assume tight coupling of GPP and growth; allocation dynamics simplified	Models fail to represent drought legacies and allocation trade-offs [37,38].
Heatwaves	Photosynthesis shuts down while transpiration continues for evaporative cooling [34]	Models typically assume proportional reduction in transpiration with reduced photosynthesis	Emergency-mode thermoregulation missing [15,34].
High VPD	Eddy covariance: stomata close, GPP decreases, but transpiration decline weaker than expected	PBMs use fixed stomatal conductance functions; lack flexibility under high VPD	WUE gains overestimated, water loss underestimated [30,39].
Compound Stress (drought + heat)	Observed synergistic effects: mortality risk sharply increases, decoupling amplified	Models often simulate stress additively, rarely include nonlinear interactions	Missing thresholds and tipping points [21].

3. Advances in Understanding Forest Water Use Efficiency

Figure 4 provides a scale-aware overview of forest water-use efficiency (WUE): it introduces the three core metrics—intrinsic WUE (iWUE), canopy-scale WUE (WUEc) and ecosystem-scale WUE (WUE_e)—with their respective numerators and denominators, lists the main biotic and abiotic drivers that shape these metrics (functional type, stand age, heat, drought, vapour-pressure deficit) and visualizes the “efficiency paradox”, in which CO₂-driven gains in iWUE are offset by climate-stress constraints on growth and WUE_e, thereby framing the global/regional patterns discussed in Section 3.2, Section 3.3 and Section 3.4.

Figure 4. Defining and measuring forest water-use efficiency across scales: metrics, drivers and the efficiency paradox.

3.1. Defining and Measuring WUE Across Scales

The term WUE is applied across multiple scales, and its definition varies accordingly. This diversity of metrics is essential for understanding different aspects of the carbon-water relationship but requires careful distinction to avoid misinterpretation when comparing studies. For instance, a physiological adaptation at the leaf level might not translate directly to an equivalent change in water use at the watershed scale. The primary definitions and their scales of application are summarized in Table 3. Responses to climate drivers can differ markedly across these scales; for example, an increase in precipitation might decrease leaf-level WUE but increase ecosystem-level WUE if the resulting boost in productivity outweighs the total increase in evapotranspiration [40].

Table 3. Definitions and measurement scales of WUE.

Metric	Abbreviation	Formula	Numerator Represents	Denominator Represents	Typical Scale
Intrinsic Water Use Efficiency	iWUE	A/gs	Net CO2 Assimilation	Stomatal Conductance	Leaf [23]
Instantaneous Water Use Efficiency	WUEi	A/T	Net CO2 Assimilation	Transpiration	Leaf [23]
Canopy Water Use Efficiency	WUEc	GPP/Tc	Gross Primary Productivity	Canopy Transpiration	Canopy/Stand [40]
Ecosystem Water Use Efficiency	WUEe	GPP/ET	Gross Primary Productivity	Evapotranspiration	Ecosystem [41]
Net Ecosystem Water Use Efficiency	WUE_nee	NEE/ET	Net Ecosystem Exchange	Evapotranspiration	Ecosystem [23]
Underlying Water Use Efficiency	uWUE	(GPP × VPD)/gs	GPP weighted by VPD	Canopy Conductance	Ecosystem [23]

3.2. Global and Regional Trends in Forest WUE: Evidence from Multi-Proxy Records

A consistent global trend emerging from multi-proxy records, particularly tree-ring isotope analysis, is a substantial increase in intrinsic water use efficiency (iWUE) over the past century [41]. A global meta-analysis of tree-ring chronologies demonstrated a ~40% increase in tree iWUE since 1901, a trend that coincides with the rise in atmospheric CO₂ and is considered to be largely driven by it [42]. However, the precise magnitude of this trend is a subject of active research. Recent studies that incorporate the effects of mesophyll conductance—the rate of CO₂ diffusion from intercellular airspaces to the sites of carboxylation within chloroplasts—suggest that conventional models have likely overestimated the historical gains in WUE [42,43]. When adjusted for mesophyll conductance, the estimated increase in global forest WUE during the 20th century was found to be considerably smaller than previously thought, which has significant implications for projections of the CO₂ fertilization effect [43].

Regional patterns reveal considerable heterogeneity in WUE trends and their drivers:

• Boreal Forests: In the vast boreal forests of Canada and Siberia, studies consistently show a strong increase in iWUE [44]. This suggests that other factors, such as nutrient limitations or increasing temperature-induced drought stress, are constraining productivity, preventing trees from capitalizing on the potential benefits of higher CO₂ [44]. In Siberia, the net effect of climate change on productivity is complex, with warming potentially increasing growth in some cold-limited northern regions while increasing fire risk and drought stress in southern regions [45].
• Temperate Forests: In Europe, tree-ring studies also show a sustained increase in iWUE, with little evidence of the response plateauing at higher CO₂ concentrations as had been previously hypothesized [46]. In the diverse forests of Northeast China, WUE trends are highly variable, depending on the specific forest type (e.g., coniferous vs. broad-leaved) and the local climatic gradient. For example, in the Daxing’anling mountains, a key driver of WUE patterns is the latitudinal gradient, which influences temperature and moisture availability, leading to distinct local adaptations to drought [47].
• Tropical Forests: As a global carbon–water hotspot, tropical forests remain data-poor relative to their importance. These highly productive ecosystems are characterized by large carbon and water fluxes but remain relatively under-studied. Available data suggest that drought events can lead to enhanced WUE, but this is often an adaptive response to water stress rather than a sign of increased productivity and may be accompanied by reduced growth [48].

3.3. Biotic and Abiotic Drivers of WUE Variability: Forest Type, Age, and Environmental Stress

The variability in forest WUE is controlled by a combination of biotic characteristics and abiotic environmental factors. Forest functional type is a primary determinant. For instance, a study across the Northern Hemisphere found that the recent increase in WUE in deciduous broadleaf forests was primarily driven by reductions in stomatal conductance (a water-saving strategy), likely in response to rising VPD. In contrast, the WUE increase in evergreen needleleaf forests was mainly driven by enhanced GPP (a carbon-gain strategy), suggesting different physiological responses to the same large-scale environmental changes [49].

Forest age and successional stage also play a crucial role. Ecosystem-level WUE dynamics are influenced by factors like interspecific competition, self-thinning, and changes in canopy structure over time. Consequently, young, rapidly growing forests often exhibit different WUE patterns and sensitivities to climate compared to mature, old growth stands [23].

The widespread evidence of rising iWUE has led to a paradox of efficiency that challenges simplistic interpretations of the CO₂ fertilization effect. Numerous studies, particularly in water- or nutrient-limited biomes, document a clear decoupling between rising leaf-level efficiency and whole-tree growth [50,51]. A tree that is becoming more “water-efficient” is not necessarily becoming more productive. This apparent contradiction is resolved by understanding that the iWUE increase is often not a result of stimulated photosynthesis but rather a defensive response of stomatal closure forced by other climatic stressors, such as rising VPD or declining soil moisture. In these cases, the tree is prioritizing survival (avoiding hydraulic failure) overgrowth (carbon accumulation) [52]. This insight is critical, as it suggests that the positive effect of CO₂ on plant growth is being significantly constrained or even negated by the negative impacts of concurrent warming and drying in many of the world’s forests. Consequently, policies and models that assume a direct, positive link between rising iWUE and a strengthening forest carbon sink may be dangerously overestimating the capacity of forests to mitigate future climate change [50].

3.4. Tropical Forests and Carbon–Water Coupling

Tropical forests, particularly those in the Amazon, Congo Basin, and Southeast Asia, play a crucial role in global carbon sequestration and water cycling. However, the current understanding of carbon–water coupling in these regions remains limited due to significant data scarcity and the complexity of their responses to climate change. Tropical forests are especially vulnerable to extreme climatic events, such as droughts and heatwaves, which are becoming more frequent and intense due to climate change [53]. These events not only reduce the forests’ capacity to sequester carbon but can also trigger a shift from carbon sinks to carbon sources, as observed in some areas of the Amazon during severe droughts [16]. Multi-year aircraft and tower constraints indicate parts of the southeastern Amazon have already become net carbon sources during recent severe droughts, underscoring sink-to-source risks under compounding heat and water stress. Evidence further suggests tropical iWUE responses can mask growth declines and heightened mortality risk when hydraulic failure thresholds are approached [54,55,56].

One of the most pressing challenges in tropical forests is drought-induced tree mortality. Recent research shows that while trees can temporarily increase water-use efficiency (WUE) during moderate droughts, extreme droughts often result in hydraulic failure, mortality, and long-term declines in productivity [57]. These processes undermine forest resilience and contribute to carbon cycle instability. Despite the importance of these dynamics, the lack of long-term monitoring networks across tropical regions makes it difficult to quantify ecosystem-level water and carbon flux responses with high confidence.

To address these gaps, future research should focus on expanding isotopic and dendrochronological networks in the tropics, especially in the Amazon, Congo, and Southeast Asia. Isotope-based records of intrinsic WUE, when integrated with flux tower and satellite data, could provide a more comprehensive understanding of how carbon–water coupling responds to extreme events [58]. In parallel, incorporating multi-method approaches that combine isotopic analyses, high-resolution remote sensing, and ecosystem modeling will improve predictions of tropical forest dynamics and their role in global carbon and water cycles [59].

4. Methodological Frontiers in Quantifying Carbon-Water Dynamics

The study of forest carbon-water coupling relies on a diverse suite of methodologies, each with unique strengths and limitations. A critical analysis of these tools is essential for interpreting the scientific literature and understanding the sources of uncertainty in our knowledge. Recent decades have seen significant advances in in situ observations, remote sensing platforms, and modeling paradigms, with the greatest progress now emerging from their integration. A comparative summary of these approaches is provided in Table 4.

4.1. In Situ Observations: Advances in Eddy Covariance, Stable Isotope Analysis, and Dendrochronology

Eddy Covariance (EC): The global FLUXNET network of micrometeorological towers provides the most direct, continuous measurements of net ecosystem exchange of CO₂ and evapotranspiration at the ecosystem scale (typically covering ~1 km²) [60]. These high-frequency data are invaluable for calculating various WUE metrics and understanding their variability from diurnal to interannual timescales [61]. However, the technique is expensive, spatially sparse, and subject to challenges such as data gaps and uncertainties in partitioning net fluxes into component parts like GPP and respiration [23].

Stable Isotope Analysis: The analysis of stable isotopes of carbon (δ¹³C) and oxygen (δ¹⁸O) in plant tissues provides a powerful tool for reconstructing past physiological behavior [62]. The δ¹³C signature serves as a time-integrated proxy for iWUE, as photosynthetic discrimination against the heavier ¹³C isotope is related to the ratio of intercellular to ambient CO₂ concentrations [62,63,64]. When combined with δ¹⁸O, which is influenced by transpiration and source water, dual-isotope approaches can help disentangle the relative contributions of changing photosynthetic capacity versus stomatal conductance to observed iWUE trends [42]. Major uncertainties remain around accurately accounting for mesophyll conductance and post-photosynthetic fractionation processes that can alter the isotopic signal [62].

Dendrochronology: The study of tree rings provides exceptionally long-term, annually resolved archives of forest dynamics [65]. Ring widths serve as a proxy for tree growth, while the stable isotope composition of the wood cellulose provides a record of physiological functioning (i.e., iWUE) [50]. This combination allows for retrospective analyses of forest responses to environmental changes over centuries, placing modern trends in a crucial historical context and enabling robust tests of the long-term CO₂ fertilization effect [50].

4.2. Remote Sensing Applications: From MODIS and Landsat to Next-Generation Sensors

Remote sensing is indispensable for scaling up site-level understanding to regional and global scales. Satellites like NASA’s Moderate Resolution Imaging Spectroradiometer (MODIS) provide global products of GPP and ET, which can be divided to estimate ecosystem WUE. While invaluable for observing large-scale patterns, these products are derived from algorithms and have inherent uncertainties stemming from their coarse spatial resolution (~500 m to 1 km), cloud cover, and the use of generalized parameters for different vegetation types [66].

Sensors with higher spatial resolution, such as Landsat (30 m) and the Sentinel-2 constellation (10–20 m), allow for monitoring at the forest stand level and in more fragmented landscapes [67]. The inclusion of red-edge spectral bands on Sentinel-2 has particularly enhanced the ability to assess vegetation stress and health [68]. Next-generation technologies are further revolutionizing the field. Light Detection and Ranging (LiDAR), both airborne and spaceborne (e.g., GEDI), provides detailed 3D information on forest structure, while hyperspectral sensors offer rich information on plant biochemistry. These advanced tools, often deployed on Unmanned Aerial Vehicles (UAVs) for ultra-high-resolution mapping, are providing unprecedented detail for forest monitoring [69].

4.3. Modeling Approaches: Process-Based Models vs. Data-Driven Machine Learning

Many ecosystem models tend to overestimate gains in water-use efficiency (WUE) because of oversimplified structural assumptions. Most neglect the strong physiological constraints imposed by heat and high vapor-pressure deficit (VPD) on stomatal conductance and omit essential processes such as canopy structural feedback, plant hydraulic failure, and drought legacies. Achieving more realistic predictions requires the explicit integration of VPD-sensitive stomatal regulation, hydraulic feedback, and dynamic carbon allocation mechanisms with memory.

Process-based models (PBMs)—such as Biome-BGC and the Community Land Model (CLM)—simulate carbon and water fluxes from first physical and biogeochemical principles [70,71]. Their primary strength lies in their mechanistic transparency, enabling projections under novel climate scenarios [72]. The high dimensionality of parameters and limited representation of structural and belowground processes often result in significant uncertainty and limit their predictive robustness [73,74].

In contrast, machine-learning (ML) and artificial-intelligence (AI) approaches infer relationships directly from large observational datasets [75]. These models have become increasingly valuable for upscaling flux-tower observations, mapping forest attributes from remote sensing, and uncovering nonlinear interactions that may be overlooked by PBMs [76]. Nonetheless, their lack of explicit physical processes constrains interpretability and often degrades performance under extrapolated climatic conditions. The emerging frontier is hybrid or physics-informed modeling, which embeds mechanistic constraints within data-driven frameworks, combining the interpretability of PBMs with the flexibility of ML to enhance both physical consistency and predictive power [77].

As this is a narrative synthesis, no new model training or evaluation is performed here. Instead, we summarize key methodological insights: future modeling efforts should explicitly account for cross-scale structural feedback, incorporate plant-hydraulic and stomatal regulation schemes, and systematically quantify both structural and data-driven uncertainties to enhance reproducibility, transparency, and comparability across modeling frameworks.

4.4. The Power of Integration: Multi-Scale Data Fusion for Comprehensive Monitoring

No single methodology can capture the full complexity of forest carbon-water dynamics. The future of forest monitoring lies in integrated approaches that strategically fuse data from multiple sources across different scales [78]. This creates a synergistic system where the strengths of one method compensate for the weaknesses of another. For example, high-resolution UAV-LiDAR data can be used to calibrate and validate coarser-resolution satellite products; these satellite data, in turn, can be assimilated into process-based models to constrain parameters and improve regional predictions [79].

This tension between the high precision of local, in situ measurements and the broad scale of remote sensing and modeling creates significant challenges but also highlights a critical path forward [51]. Discrepancies between a FLUXNET tower measurement and a corresponding satellite pixel are common and arise from a combination of factors: the tower’s limited footprint may not represent the heterogeneous pixel, the satellite algorithm may be biased, and the models used to fill data gaps in either dataset introduce their own errors [80,81,82]. This reveals that our scientific conclusions about large-scale forest WUE trends are not merely observations of nature but are also shaped by the chosen methodological framework. The grand challenge, therefore, is not just to collect more data with any single method, but to develop robust model-data fusion frameworks that can formally reconcile these disparate data streams, propagate uncertainties through the analysis chain, and produce a more holistic and honest assessment of our understanding of forest ecosystems [78,83].

While each approach provides unique insights, their reliability varies substantially across spatial scales and climatic regimes. Figure 5 summarizes the comparative strengths and uncertainties of the major methodologies discussed, highlighting the need for integrated, multi-scale approaches. Horizontal bars indicate the main scales where each method is most reliable, from leaf to global levels.

Figure 5. Spatial applicability of major methodologies for assessing forest carbon–water coupling.

Table 4. Comparison of methodologies for assessing forest carbon–water fluxes and WUE.

Methodology	Principle	Key Variables Measured/Derived	Spatiotemporal Scale	Advantages	Limitations and Uncertainties
Eddy Covariance	Micrometeorological measurement of turbulent fluxes	NEE, ET, GPP	Ecosystem (~1 km2), continuous (sub-hourly to decadal)	Direct ecosystem-scale flux measurement; high temporal resolution; non-destructive.	High cost; limited spatial; data gaps; energy balance closure issues; assumptions in flux partitioning [60].
Stable Isotopes	Isotopic fractionation during gas exchange and water transport	iWUE, gs	Leaf to tree, integrated over tissue formation (annual to centennial)	Provides long-term historical records; Integrates physiological responses over time; can separate drivers of iWUE change.	Indirect measurement; destructive sampling; uncertainties from mesophyll conductance and post-photosynthetic fractionation [62].
Remote Sensing	Spectral reflectance/emittance related to vegetation properties	GPP, ET, LAI, NDVI	Stand to Global (>10 m to >1 km), daily to bi-weekly	Broad spatial coverage; Cost-effective for large areas; Repeated observations for monitoring.	Indirect estimation via models; cloud contamination; algorithm uncertainties; scale mismatch with ground validation [84].
Process-Based Models	Mathematical simulation of biophysical and ecological processes	All C & H2O fluxes and stocks	Plot to Global, any temporal resolution	Predictive capability for future scenarios; mechanistic understanding; integration of multiple drivers.	High complexity and parameter uncertainty; can misrepresent key processes; requires extensive calibration/validation [85].
Machine Learning/AI	Data-driven statistical pattern recognition	Any predictable variable	Any scale (dependent on training data)	Excellent at capturing complex nonlinear patterns; can integrate diverse data types.	“Black box” nature lacks interpretability; poor extrapolation beyond training data; requires large datasets [75].

Table 4. Comparison of methodologies for assessing forest carbon–water fluxes and WUE.

Methodology	Principle	Key Variables Measured/Derived	Spatiotemporal Scale	Advantages	Limitations and Uncertainties
Eddy Covariance	Micrometeorological measurement of turbulent fluxes	NEE, ET, GPP	Ecosystem (~1 km2), continuous (sub-hourly to decadal)	Direct ecosystem-scale flux measurement; high temporal resolution; non-destructive.	High cost; limited spatial; data gaps; energy balance closure issues; assumptions in flux partitioning [60].
Stable Isotopes	Isotopic fractionation during gas exchange and water transport	iWUE, gs	Leaf to tree, integrated over tissue formation (annual to centennial)	Provides long-term historical records; Integrates physiological responses over time; can separate drivers of iWUE change.	Indirect measurement; destructive sampling; uncertainties from mesophyll conductance and post-photosynthetic fractionation [62].
Remote Sensing	Spectral reflectance/emittance related to vegetation properties	GPP, ET, LAI, NDVI	Stand to Global (>10 m to >1 km), daily to bi-weekly	Broad spatial coverage; Cost-effective for large areas; Repeated observations for monitoring.	Indirect estimation via models; cloud contamination; algorithm uncertainties; scale mismatch with ground validation [84].
Process-Based Models	Mathematical simulation of biophysical and ecological processes	All C & H2O fluxes and stocks	Plot to Global, any temporal resolution	Predictive capability for future scenarios; mechanistic understanding; integration of multiple drivers.	High complexity and parameter uncertainty; can misrepresent key processes; requires extensive calibration/validation [85].
Machine Learning/AI	Data-driven statistical pattern recognition	Any predictable variable	Any scale (dependent on training data)	Excellent at capturing complex nonlinear patterns; can integrate diverse data types.	“Black box” nature lacks interpretability; poor extrapolation beyond training data; requires large datasets [75].

Table 4 summarizes the principles and limits of each approach, yet key discrepancies persist. Flux towers provide high-resolution fluxes but often overestimate WUE in drought years due to partitioning and energy-closure issues. Isotopes reveal long-term iWUE trends, but miss short-term variation. Remote sensing enables global mapping yet dampens extremes through algorithmic and cloud biases. Process-based models integrate drivers but have high parameter uncertainty. Machine-learning models capture nonlinearities yet remain black boxes with poor interpretability and extrapolation.

4.5. Critical Comparison of Methodological Reliability

Although diverse methodologies are available to study forest carbon–water dynamics, their reliability varies substantially across climatic regimes and spatial scales. For example, eddy covariance provides highly accurate flux measurements at the ecosystem scale but is limited in arid and heterogeneous landscapes where tower footprints may not represent regional variability [86]. Stable isotopes are powerful for reconstructing long-term physiological signals, yet uncertainties in mesophyll conductance and post-photosynthetic fractionation remain particularly problematic in water-stressed ecosystems. Remote sensing offers unmatched spatial coverage, but its reliability decreases under persistent cloud cover in the tropics and when transferring global algorithms to drylands. Process-based models (PBMs) capture mechanistic detail but often underperform under extreme droughts or heatwaves because hydraulic failure and decoupling phenomena are poorly represented [87]. In contrast, machine learning and AI approaches excel in pattern recognition under data-rich conditions but suffer from poor extrapolation to novel climate extremes [88]. These comparisons underscore the need for integrated approaches that combine the strengths of each method to achieve robust multi-scale assessments.

4.6. Evidence Integration: Cross-Method Consistency Under Extreme Events

Building on Section 4.4 and Section 4.5, which motivate multi-scale data fusion and compare methodological reliability, we conclude the methods–synthesis part with two brief case studies. These cases triangulate tower-based WUE_e (GPP/ET), tree-ring iWUE, and satellite indicators (SIF/structure), and benchmark observations against state-of-the-art model ensembles [89,90].

• Case 1—Central/Northern Europe, summer 2018 (“hot drought”). Independent lines of evidence indicate widespread photosynthetic depression and a weakened land CO₂ sink. OCO-2 SIF showed pronounced negative anomalies during the heat–drought episode across vegetation types, consistent with reduced light-use efficiency [91]. At the carbon-budget scale, atmospheric inversions estimated temperate Europe’s annual NEE in 2018 to be less negative by about 0.09 ± 0.06 Pg C yr⁻¹ relative to the previous decade [92]. Multi-sensor analyses further documented productivity losses with legacy effects extending into autumn [93]. Concurrent dendro- and isotope-based studies show increases in iWUE during extreme drought while growth remains constrained, consistent with stomatal conservatism under high VPD and low soil moisture [94,95]. Together, these lines of evidence imply that GPP declined more than ET in many forests, yielding lower WUE_e at the ecosystem scale while iWUE increased at the tree/leaf scale. We juxtapose these observations with data-driven flux products and DGVM ensembles; the ensembles capture regional anomaly signs but can underestimate GPP deficits in hotspots (ensemble mean ± 1σ) [89].
• Case 2—California, 2012–2015 (“multi-year hot drought”). Remote sensing and field campaigns revealed progressive loss of canopy water content and widespread mortality across California forests, providing structural evidence of physiological strain [96,97]. In parallel, recent evaluations show that satellite SIF closely tracks seasonal and interannual GPP variability, supporting its use as a drought indicator in water-limited regions of the western United States [98,99]. At the ecosystem scale, satellite-based assessments report episodes where WUE_e increases under supply-limited conditions because ET reductions outpace GPP declines—an effect also found in global analyses of WUE response to drought [100]. As in Europe 2018, we benchmark observations against machine-learning flux products and model ensembles to display anomaly trajectories with ensemble mean ± 1σ envelopes [89].

Across both events, SIF and GPP co-vary as a common physiological signal, while iWUE increases at the tree scale reflect stomatal regulation under elevated atmospheric dryness. Whether WUE_e decreases (when GPP losses dominate) or increases (when ET reductions dominate) hinges on the relative magnitudes of GPP versus ET anomalies. These cases underscore that cross-scale WUE metrics are complementary rather than interchangeable and should be interpreted jointly when assessing forest responses to hot droughts and their management/policy implications.

5. Sustainability Perspectives: Implications for Ecosystem Services and Management

To guide this section, Figure 6 presents a roadmap of sustainability perspectives at the carbon–water nexus. It illustrates the trade-offs and synergies among carbon sequestration, water yield, and forest productivity (Section 5.1); highlights forest resilience under drought and adaptive water-use strategies (Section 5.2); outlines climate-smart forestry options for enhancing carbon–water co-benefits (Section 5.3); summarizes the policy implications for conservation and sustainable resource management (Section 5.4); and introduces governance, equity, and policy frameworks (Section 5.5).

Figure 6. Carbon–water trade-offs, resilience, climate-smart forestry and policy implications.

5.1. Trade-Offs and Synergies: Carbon Sequestration, Water Yield, and Forest Productivity

When do WUE gains translate into real resilience? (1) Hydraulics: safety margins remain ≥ xylem vulnerability thresholds; (2) Structure: LAI increases do not over-escalate ET under high VPD; (3) Legacy: post-event growth and NEP recover within 1–2 years. Fail cases include heatwave “emergency cooling” where ET persists but GPP collapses, and drought-legacy allocation shifts that decouple ring growth from GPP, both lowering WUE_e despite higher iWUE [101,102].

Balancing carbon sequestration with water yield is a central challenge in sustainable forest management. The tension is most acute in water-limited regions and should guide site selection for large-scale afforestation and reforestation; if poorly sited, such projects can exacerbate local water scarcity [23]. Forest management practices, such as thinning, can be employed to mediate this trade-off. By reducing stand density, thinning can decrease total ecosystem water use and increase water yield, though this may come at the expense of reducing the total carbon stored on the landscape at a given time [23].

This conflict highlights that two major global sustainability goals—climate change mitigation through carbon sequestration and ensuring water security—can be at odds. A forest management strategy that is “climate-smart” cannot be singularly focused on maximizing carbon. The optimal solution is not universal but is highly context specific. In a water-abundant region, management might prioritize maximizing carbon stocks. Conversely, in a water-scarce basin, the most sustainable strategy may be to manage for lower-density, more drought-resilient forests that prioritize water conservation and yield, even if this means lower carbon sequestration rates. Global policies that incentivize carbon sequestration without accounting for these hydrological trade-offs risk creating perverse outcomes, such as promoting water-intensive plantations in drought-prone areas, undermining long-term ecological and social sustainability.

Regional case studies underscore these trade-offs. In semi-arid northern China, large-scale afforestation under the Grain-for-Green program has enhanced aboveground carbon storage but substantially reduced river discharge and groundwater recharge, demonstrating the risks of planting water-demanding species in arid zones [103,104]. In the Amazon Basin, evidence shows that poorly planned large-scale plantations can disrupt regional rainfall recycling by intensifying evapotranspiration, thereby undermining long-term hydrological resilience [16,104,105]. By contrast, European mixed-species forestry trials illustrate adaptive pathways: tree diversity enhances drought resilience, optimizes water use across species, and sustains ecosystem productivity better than monocultures [106,107]. These cases suggest that maladaptive afforestation may exacerbate water scarcity, while adaptive practices—such as species diversification, silvicultural thinning, and site-specific planning—offer pathways to reconcile carbon sequestration with water security.

Soil conservation as a co-driver. Increased canopy cover raises interception and ET but also improves soil aggregate stability and infiltration, reducing overland flow and sediment export. Where rainfall intensity is rising, practices that maintain litter/understory and mixed roots can co-optimize carbon storage, water yield timing, and erosion control, mitigating trade-offs [108,109].

Comparative assessments from Southwest China and Kentucky, USA show that projected land-use change can rival or exceed climate impacts on water-related ecosystem services; mechanisms include species choice altering ET and rooting depth, and management shaping infiltration/runoff partitions—consistent with our trade-off framework [110,111,112,113,114].

5.2. Forest Resilience in the Face of Drought: The Role of Adaptive Water Use Strategies

Ecosystem resilience refers to the capacity of a forest to withstand a disturbance like drought and recover its structure and function afterward [115]. A key physiological mechanism underpinning this resilience is the ability of trees to adapt their water use. Increasing WUE is a primary adaptive strategy under water stress, as plants regulate their stomata to conserve limited moisture. The nature of this response, however, is complex. While moderate drought often leads to an increase in WUE, severe or prolonged drought can overwhelm a plant’s regulatory capacity, causing both stomatal and non-stomatal damage to the photosynthetic apparatus and leading to a sharp decline in both WUE and productivity [116].

Resilience varies significantly among forest types. For example, studies suggest that forests in drier climates may be more resilient to water losses (i.e., they recover hydrologic function quickly), whereas forests in more humid climates may be more resilient to productivity losses (i.e., they recover carbon uptake rates quickly) [117]. Furthermore, biodiversity within a forest can enhance overall ecosystem resilience. A diversity of species employing a range of water-use strategies—from “water-saving” isohydric species that close stomata early to “risk-taking” anisohydric species that maintain photosynthesis for longer—can buffer the entire ecosystem against the impacts of drought [118].

5.3. Climate-Smart Forestry: Adaptive Management for Enhancing Carbon-Water Co-Benefits

In response to the growing threats of climate change, the concept of “climate-smart forestry” has emerged. This approach seeks to integrate climate change adaptation and mitigation with sustainable forest management objectives, aiming to create resilient ecosystems that continue to provide essential goods and services [119]. Adaptive management is central to this paradigm, requiring a flexible suite of strategies to reduce forest vulnerability to impacts like water scarcity [120]. Key adaptive practices that directly address the carbon–water nexus include the following:

• Silvicultural Thinning: Proactively reducing stand density to lessen competition for water among remaining trees, thereby increasing their individual drought resilience and potentially increasing water yield from the watershed [121].
• Species and Genotype Selection: In reforestation and afforestation efforts, prioritizing the selection of tree species and provenances that are better adapted to projected future climate conditions, particularly those exhibiting greater drought tolerance and higher intrinsic WUE [122].
• Promoting Mixed-Species Forests: Moving away from monocultures toward planting and managing for mixed-species stands. This increases structural and functional diversity, which can enhance overall ecosystem resilience to disturbances and optimize the use of resources like water and nutrients [123].
• Modifying Rotation Lengths and Harvesting Regimes: Adjusting the timing and intensity of harvesting can influence long-term water use. For instance, shorter rotation cycles may reduce the period during which a stand is at maximum water consumption, while selective harvesting can maintain continuous canopy cover, which helps protect soil and water resources [124].

5.4. Policy Implications for Conservation and Sustainable Resource Management

A nuanced understanding of forest carbon-water coupling has profound implications for policy. International and national policies aimed at climate change mitigation through forestry, such as REDD+ (Reducing Emissions from Deforestation and Forest Degradation), must be designed with careful consideration of their potential impacts on water resources [125,126]. Simply maximizing carbon stocks without evaluating hydrological trade-offs can lead to unintended negative consequences for water security and other ecosystem services.

Long-term success requires sustained political will, capable institutions, and policies that decouple commodity production from deforestation [127]. Equally critical is investment in fire prevention, control of illegal logging, and empowerment of local and Indigenous communities, who are often the most effective stewards of forest landscapes [128]. Ultimately, policy frameworks should manage forests as complex, adaptive socio-ecological systems, valuing their full suite of services—carbon storage, water regulation, biodiversity, and social resilience—rather than carbon stocks alone [128,129].

5.5. Governance, Equity, and Policy Frameworks

International frameworks (e.g., REDD+ MRV requirements) and regional strategies (e.g., European Green Deal Forest initiatives) can mainstream WUE-aware planning by conditioning incentives on water-security safeguards (basin-scale caps, drought triggers) and equity criteria (tenure, benefit sharing). In developing countries, coupling WUE tracking with water-security indicators can guide investment toward fuel-load reduction and biomass recovery with co-benefits for rural energy. At watershed vs. urban scales, differentiated measures are needed: upstream thinning/diversification and fuel-breaks to stabilize baseflow and reduce fire risk; peri-urban greenbelts managed for lower ET and erosion control to protect water supply and air quality; and community-led monitoring to ensure environmental justice and adaptive governance [130,131].

6. Synthesis, Grand Challenges, and Future Research Directions

6.1. Synthesizing the Current State of Knowledge on Forest Carbon-Water Coupling

This review has synthesized the multifaceted relationship between carbon and water cycles in forest ecosystems under the growing pressure of global climate change. Rising CO₂ alters the fundamental trade-off between carbon uptake and water loss, increasing intrinsic WUE globally. However, this physiological signal of enhanced efficiency is often a deceptive indicator of forest health. It is frequently decoupled from ecosystem-scale productivity and growth, which are increasingly constrained by concurrent warming, atmospheric drying, and altered precipitation regimes [42,50].

This paradox of efficiency underscores that the benefits of CO₂ fertilization are often outweighed by climate-related stressors. Recent innovations in in situ observation, remote sensing, and modeling have greatly enhanced our capacity to monitor these complex dynamics. Yet, each method possesses inherent limitations, and the greatest progress is being made through their integration. The implications of these changing carbon-water dynamics are profound, creating critical trade-offs between managing forests for carbon sequestration versus water yield and challenging the long-term resilience of these vital ecosystems [23,132,133,134,135].

6.2. Identifying Critical Knowledge Gaps and Methodological Uncertainties

Despite significant progress, critical knowledge gaps and uncertainties persist, hindering our ability to accurately predict the future of forest ecosystems. Key challenges include:

• Belowground Processes: The role of root systems, including competition for soil water, hydraulic redistribution, and carbon allocation belowground, represents a major frontier in ecohydrology that is inadequately represented in most ecosystem models.
• Interacting Disturbances: Forests are increasingly subject to multiple, interacting disturbances (e.g., drought weakening trees, making them more susceptible to insect attack and fire). Quantifying the synergistic effects of these compound disturbances on carbon-water coupling and post-disturbance recovery trajectories is a major uncertainty.
• Model Deficiencies: Process-based models often lack robust representations of key physiological mechanisms, such as hydraulic failure, plant-level optimization of water use, and the decoupling of photosynthesis and transpiration under extreme stress.
• Remote Sensing Uncertainties: While powerful for spatial monitoring, remote sensing products for GPP and ET—the core components of ecosystem WUE—contain significant uncertainties related to algorithm assumptions, sensor limitations, and validation challenges. These errors propagate directly into our large-scale assessments of carbon-water dynamics.

6.3. Proposing a Research Agenda for the Next Decade

To address the critical knowledge gaps identified, we propose a research agenda centered on four grand challenges. The path forward requires a paradigm of “systems integration,” where multi-platform observation networks, multi-paradigm models, and multi-objective management frameworks are co-developed. The four main challenges are as follows:

• Develop a new generation of integrated monitoring systems that fuse remote sensing with in situ networks.
• Improve the predictive capacity of ecosystem models by incorporating more realistic plant hydraulics and hybridizing them with machine learning.
• Quantify the resilience and tipping points of forest ecosystems under combined stressors [136].
• Translate science into actionable, adaptive management and policy through co-developed decision-support tools [137].

Addressing these challenges is critical for advancing our fundamental understanding and providing the scientific foundation needed to steward forests sustainably through the 21st century.

7. Conclusions and Outlook

Forests play a crucial role in regulating carbon and water cycles, and WUE provides an integrative framework to examine their interactions under climate change. Evidence indicates that WUE has increased in many regions as rising CO₂ enhances photosynthesis relative to transpiration, although these gains are highly context-dependent and can be offset by drought and heat stress. Species-specific hydraulic traits and stomatal regulation strongly influence ecosystem-level outcomes, underscoring the importance of forest composition in determining resilience.

Advances in flux measurements, isotopic analyses, remote sensing, and modeling have established a coherent picture of rising WUE, yet significant uncertainties remain, particularly regarding extreme events, drought legacies, and belowground processes. Improved integration of multi-source datasets, enhanced process representation, and inter-model comparison are needed to constrain these uncertainties and assess the robustness of WUE responses under variable climate regimes.

From a management perspective, maximizing carbon sequestration without considering water constraints risks maladaptive outcomes. Incorporating WUE into forest management and policy offers a quantitative means to balance carbon storage with water yield and broader sustainability objectives.

Future research should move beyond short-term experiments to establish long-term observation networks and realistic climate-change scenarios that explicitly capture compound extremes and progressive warming. Comparative evaluation and optimization of WUE models—spanning process-based, remote-sensing, and data-driven frameworks—will be critical for reconciling scale mismatches and improving predictive accuracy. Translating WUE knowledge into practical decision tools will be essential for sustaining forests as resilient carbon sinks and reliable water providers in a changing climate.

Limitations include literature bias toward temperate/boreal regions and the absence of a meta-analysis. Future work should prioritize tropical networks, long-term compound-extreme records, and systematic WUE model intercomparisons under realistic scenarios. Addressing these gaps requires prioritizing tropical monitoring networks, developing compound-extreme archives, and conducting coordinated intercomparisons of WUE models under realistic future scenarios.