Literature Review: Forest Decline Dynamics: Drought, Fire, and Ecosystem Collapse

Forests are experiencing accelerated mortality due to interconnected pressures from drought, fire, and biotic agents, with key mechanisms and thresholds still being elucidated. This review synthesizes 71 studies (2007–2025) across global ecosystems, integrating findings from physiology, ecology, and modeling. It reveals that drought-induced hydraulic failure and carbon starvation concurrently weaken trees, also impairing resin defenses and escalating bark beetle outbreaks. Critically, atmospheric vapour pressure deficit drives water stress independently of soil moisture, while intensified fire-climate interactions, compounded by beetle-altered fuels, can push landscapes past resilience thresholds into non-forest states. Climate warming has expanded beetle ranges and voltinism, transforming outbreaks into continental-scale carbon-climate feedbacks. Post-disturbance recovery is increasingly failing due to seed source loss, climatic unsuitability, short-interval reburning, and degraded mycorrhizal networks, locking sites into alternative stable states. Consequently, tree regeneration succeeds in only about 21% of documented cases globally following compound disturbances. The analysis concludes that forest decline involves mechanistically linked threats requiring integrated management, as forest loss represents a potentially irreversible outcome of climate change.

1. Introduction

Forests cover nearly one-third of Earth’s terrestrial surface and underpin a disproportionate share of global biodiversity, carbon storage, and hydrological regulation. Yet mounting observational and experimental evidence accumulated over the past two decades points toward a troubling trajectory: forests worldwide are dying faster than previously recorded, and the mechanisms, feedbacks, and tipping dynamics driving this acceleration are only partially understood [1, 2]. Drought, heat, fire, and biotic disturbance agents — forces that have always shaped forest ecosystems — are intensifying in concert and at scales that challenge longstanding assumptions about forest resilience. What makes the current moment in this literature particularly consequential is not simply that these stressors are becoming more severe in isolation, but that their interactions are generating outcomes that earlier frameworks of disturbance ecology were not designed to anticipate. Bark beetle outbreaks that were once understood as localized responses to windthrow or stand-level stress are now documented as continent-spanning eruptions driven by compound heat-drought events [3, 4], converting vast tracts of forest from carbon sinks to carbon sources [5] and restructuring fuel loads in ways that amplify subsequent fire risk. Equally significant, the regeneration processes that historically allowed forests to recover from such disturbances are themselves being undermined — by the elimination of seed sources across expanding high-severity burn perimeters, by climatic conditions during post-disturbance establishment windows that are increasingly hostile to seedling survival, and by disruption of the belowground biological legacies on which juvenile trees depend. A systematic synthesis of the science as it has developed between 2007 and 2025 is therefore both timely and necessary.

The period under review has witnessed a series of landmark shifts. Physiological research has moved from descriptive accounts of tree death toward mechanistic hypotheses centered on hydraulic failure and carbon starvation [6, 7, 8], two pathways that have generated sustained debate and increasingly sophisticated empirical testing [9, 10]. Simultaneously, atmospheric scientists and ecophysiologists have converged on vapour pressure deficit as a critical and underappreciated driver of tree water stress [11, 12] — one whose influence may rival or exceed that of soil moisture deficit under certain conditions [13]. At the same time, remote sensing platforms and global forest inventory networks have enabled the documentation of die-off events across biomes at a resolution and geographic breadth previously impossible [14], revealing that forest mortality is neither idiosyncratic nor confined to any single climatic zone. In parallel, entomologists and forest pathologists have demonstrated that climate warming is restructuring the population dynamics, voltinism, and geographic range of bark beetles at continental scales [15], creating compound disturbance regimes in which drought-weakened trees face biotic attack pressures that exceed any historical precedent [3, 16]. More recently, a growing body of field-based and modeling research has documented that post-disturbance forest regeneration — long treated as an ecological given — is failing at landscape scales across western and boreal North America, with seed source limitation, climate moisture stress, and short-interval reburning operating as compounding bottlenecks that foreclose recovery trajectories previously considered reliable [17, 18]. These parallel developments have not yet been comprehensively integrated, and the relationships between process-level mechanisms, atmospheric drivers, biotic disturbance feedbacks, regeneration failure, and landscape-scale outcomes remain a central unresolved tension in the field.

Against this backdrop, the present review addresses four interrelated research questions. First, what physiological mechanisms lead to tree mortality during drought and heat stress? Second, how do fire regimes interact with forest structure and climate to produce ecosystem-scale transitions? Third, when the hydrological cycle intensifies, how do changes in precipitation patterns and atmospheric moisture demand interact to stress forest ecosystems? And fourth, under what conditions do forests recover following disturbance versus transition to alternative stable states — and what demographic and belowground mechanisms determine which trajectory prevails?

To address these questions systematically, this review synthesizes 71 peer-reviewed studies published between 2007 and 2025, spanning experimental physiology, atmospheric science, landscape ecology, entomology, forest pathology, ecosystem modeling, and post-disturbance regeneration ecology. The scope is deliberately broad in geographic terms — encompassing temperate, boreal, subtropical, and semi-arid forest systems — while remaining focused on the mechanistic and dynamical processes that connect climate forcing to forest outcomes. Studies examining agricultural systems, non-woody vegetation, or climate projections without grounding in observed or experimental forest dynamics fall outside the scope of this synthesis.

The review is organized around five thematic clusters. The first addresses the physiological mechanisms of drought-induced tree mortality, examining the hydraulic failure and carbon starvation frameworks [6, 19], the growing body of work that seeks to resolve their relative contributions [8, 9], and the downstream consequences of physiological stress for trees’ chemical defense capacity against biotic agents. The second cluster turns to vapour pressure deficit and atmospheric moisture demand, exploring how rising evaporative stress interacts with stomatal regulation and canopy-level water use to impose constraints on forest function independent of — and in conjunction with — soil moisture limitation [11, 12]. The third cluster shifts in scale to document global patterns and regional case studies of forest die-off, drawing on observational networks and remote sensing analyses to characterize the climatic fingerprints and spatial distribution of mortality events [14, 1], including the expanding role of bark beetle outbreaks driven by warming-induced voltinism shifts and range expansion [15, 3]. The fourth cluster examines land-atmosphere feedbacks, compound extremes, and fire-climate interactions, integrating research on how vegetation and climate are coupled in ways that can amplify drought, alter ignition regimes, and generate conditions for cascading disturbance — including the complex fuel dynamics created by bark beetle mortality. The fifth and final cluster addresses forest ecosystem transitions, resilience thresholds, and the concept of alternative stable states, evaluating the theoretical and empirical evidence for threshold behavior and the conditions that determine whether forests recover or undergo irreversible compositional change — with particular attention to the demographic and belowground mechanisms that mediate non-recovery [17, 18].

Taken together, these themes reflect a field that has matured considerably in its mechanistic grounding while grappling with the inherently complex, cross-scale nature of forest decline [20]. The urgency of this synthesis stems from the recognition that forest systems may be approaching — and in some regions may already be crossing — thresholds beyond which conventional expectations of recovery no longer apply [2, 1].

2. Methodology

The literature informing this review was identified and assembled through a structured search of the OpenAlex database, combining targeted keyword queries with quality-based filtering to produce a focused corpus spanning the period 2007 to 2025.

Search Strategy

Seven thematic queries were constructed to capture the principal dimensions of forest decline dynamics. These addressed, respectively, the physiological mechanisms of drought-induced mortality including hydraulic failure and carbon starvation [8, 7, 10]; fire regime interactions with forest structure under climate change and the emergence of alternative stable states [21, 22]; hydrological cycle intensification and the role of vapour pressure deficit in driving forest stress [12, 13]; post-disturbance recovery trajectories and resilience [23, 24]; more recent work on ecosystem collapse under compounding drought and heat stress [14, 1]; the role of biotic disturbance agents — bark beetles, fungal pathogens, and drought-pest interaction feedbacks — in amplifying forest mortality under changing climatic conditions [3, 25]; and post-disturbance regeneration failure, recruitment bottlenecks, and the demographic and belowground mechanisms mediating forest non-recovery following compound disturbances [17, 26, 27]. Each query was submitted to OpenAlex, which returned an initial pool of candidate papers. Relevance scoring was then applied with a threshold of 0.6, and papers meeting this threshold were taken forward for full analysis.

Citation network expansion was also initiated as a supplementary discovery mechanism. Although the expansion stage formally examined forward and backward citations, it contributed additional relevant papers to the candidate pool, achieving meaningful coverage before the process was terminated upon reaching predefined collection targets.

Selection and Filtering

Quality filters were applied consistently across the candidate pool to balance scholarly rigour with currency. Papers older than the two-year recency window were required to meet a minimum citation threshold of five, ensuring that older contributions had accrued sufficient scholarly engagement to warrant inclusion.

To ensure the corpus remained responsive to rapidly evolving conditions in this field — where unexpected events of elevated tree mortality continue to emerge globally [1] and hotter-drought fingerprints are being documented across forest biomes in near real time [14] — a recency quota of 35% was enforced. This quota ensured that a meaningful proportion of the final selection derived from work published within the most recent two years. This approach is particularly important given that critical forest system transitions and tipping-point dynamics remain an active frontier of ongoing research [28, 20].

Where retrieval of a selected paper failed, an alternative from the candidate pool was substituted, and all papers in the final corpus were available for full-text analysis.

Processing and Corpus Characteristics

All 71 papers in the corpus underwent full-text analysis; no papers were processed from abstracts or metadata alone, and there were no unresolved retrieval failures. The date range of 2007 to 2025 reflects a period of intensifying scientific attention to forest vulnerability. This period encompasses foundational mechanistic work on tree mortality [7, 8, 13], early empirical characterisations of post-fire regeneration failure [17, 29], and the most recent empirical and modelling contributions on ecosystem transitions and compound disturbance dynamics [30, 1, 26].

Thematic Organisation

The final corpus was organised into five thematic clusters corresponding broadly to the query structure: physiological drought mortality mechanisms [8, 9], fire–climate–forest interactions, atmospheric and hydrological drivers of stress, post-disturbance resilience and successional dynamics [23], and emerging evidence of ecosystem collapse [31] — with biotic disturbance agent dynamics and post-disturbance regeneration failure integrated across all five themes rather than treated as independent clusters, reflecting the fundamentally interactive role that bark beetles, pathogens, regeneration bottlenecks, and their climatic drivers play across the full spectrum of forest decline processes [32, 3, 33, 17, 29]. This clustering provided the organisational scaffold for the synthesis, allowing patterns of convergence and disagreement across the literature to be examined systematically rather than study by study. Together, the 71 papers provide sufficient disciplinary breadth — spanning ecophysiology, disturbance ecology, entomology, forest pathology, hydrology, regeneration ecology, soil biology, and global change science — to support a coherent and evidence-grounded account of forest decline dynamics under contemporary and projected climate conditions.

3. Physiological Mechanisms of Drought-Induced Tree Mortality

Understanding why trees die during drought has become one of the most consequential questions in forest ecology, particularly as drought events intensify in frequency and severity under accelerating climate change. Two principal mechanistic pathways have dominated the research landscape: hydraulic failure, in which xylem cavitation and progressive loss of hydraulic conductivity deprive trees of water transport capacity, and carbon starvation, in which prolonged stomatal closure depletes nonstructural carbohydrate (NSC) reserves below the thresholds required for metabolic function. These mechanisms were formally articulated as a conceptual framework by [6], whose foundational modeling work projected that temperature increases of 1.1–6.4°C could raise transpiration demand by 7–48%, driving both isohydric and anisohydric species toward their respective mortality thresholds—isohydric piñon toward carbon starvation within seven months and anisohydric juniper toward hydraulic failure after fifteen months under a simulated severe drought scenario. This dual-mechanism model structured two decades of subsequent inquiry, provoking both productive refinement and persistent debate.

Hydraulic Failure as a Near-Universal Mortality Hallmark

Early empirical work largely supported hydraulic failure as a central, if not exclusive, driver of mortality. [9] synthesized evidence from over 130 sources to demonstrate that catastrophic failure of the hydraulic system is a principal proximate mechanism during severe drought. Species characteristically operate within narrow hydraulic safety margins that leave little buffer between minimum operational water potentials and cavitation thresholds. The structural basis for this hydraulic vulnerability was established globally by [34], who compiled embolism resistance data for 480 woody species across 81 sites. Their analysis found that 70% of forest species operate with hydraulic safety margins of less than one megapascal—margins that are largely independent of mean annual precipitation. This indicates that wet-climate forests are as hydraulically precarious as their arid-zone counterparts. According to [34], this finely tuned risk acceptance reveals an evolutionary calculus that tolerates meaningful hydraulic vulnerability in exchange for competitive efficiency under non-lethal conditions.

The most rigorous quantitative support for hydraulic failure as a near-universal feature of drought mortality came from the multi-species synthesis of [8], which aggregated data across 19 experimental and observational studies encompassing 26 tree species. Their meta-analysis found that all species showed at least 60% loss of hydraulic conductivity (PLC) at the time of death, with a mean of 84.3%—establishing xylem hydraulic dysfunction as what they termed a universal hallmark of drought-induced mortality. Critically, this finding reframed earlier equal-weighting models: carbon starvation, defined by significantly lower NSC concentrations relative to healthy controls, was present in only 48% of mortality cases, indicating it operates as a conditional rather than obligate pathway. This asymmetry challenged the dual-mechanism framing of [6] and repositioned hydraulic failure as the primary mechanism. Complementary experimental work by [7], using piñon pine subjected to severe drought versus shading treatments, demonstrated that fast-dying drought trees exhibited rapid xylem hydraulic conductivity loss to near-zero with relatively preserved carbohydrate reserves. In contrast, carbon starvation signatures were confined to shade-stressed individuals, further confirming the condition-dependence of each pathway.

Corroborating field evidence reinforced this hydraulic hierarchy. [35] documented the widespread die-off of trembling aspen (Populus tremuloides) across western North America and found through direct in situ measurements that hydraulic failure of roots and branches was the primary driver, with carbohydrate depletion playing no significant role. This finding placed the aspen case firmly in the hydraulic-failure category and directly contested interpretations that attributed contemporaneous die-offs to carbon starvation. More recently, [36] compared declining and healthy larch (Larix principis-rupprechtii) trees and found that declining individuals exhibited 43.6% higher native PLC and 21.0% lower sapwood-specific hydraulic conductivity than their healthy counterparts. These hydraulic impairments were accompanied by modest but significant reductions in NSC concentrations in leaves, branches, and twigs—reinforcing the pattern that hydraulic dysfunction is the dominant and most consistent signal of mortality trajectory.

Carbon Starvation as a Conditional Mortality Pathway

While hydraulic failure dominates mortality patterns in synthesized datasets—[8] found that all 26 species examined across 19 studies exhibited 60% or greater loss of xylem hydraulic conductivity at death, with a cross-species mean of 84.3%—carbon starvation remains an important explanatory pathway under specific conditions. This pathway becomes particularly relevant when elevated temperatures amplify respiratory carbon demand [6]. A landmark controlled experiment by [37] using mature piñon pines at the Biosphere 2 facility demonstrated this temperature effect. Trees exposed to elevated temperatures during drought died approximately 28% faster (18 weeks versus 25 weeks) without showing significantly different xylem water potentials. This pattern pointed to respiratory carbon drain rather than accelerated cavitation as the underlying mechanism. The authors argued that such temperature sensitivity portends amplified regional die-off rates under warming scenarios, particularly through the carbon starvation pathway.

Further synthesis by [38] confirmed that temperature strongly influences mortality timing for both isohydric and anisohydric species. However, this work simultaneously challenged the predictive value of these functional classifications, showing them to be insufficient for capturing the full range of species-specific responses [6]. The complexity deepened with experimental results from [7], who subjected piñon pine to both drought and shade treatments. The researchers found that these two stress types induced entirely different mortality pathways within the same species: drought-stressed trees exhibited hydraulic failure, while shade-stressed trees showed carbon starvation. Crucially, they identified that phloem turgor pressure collapse preceded mortality in drought-stressed trees and correlated strongly with carbohydrate availability in bark and phloem tissue. This finding suggests that phloem dynamics serve as a mechanistically distinct signal that bridges, rather than separates, the two dominant hypotheses [9].

Notably, [8] found that carbon starvation signals were most prevalent among gymnosperms, appearing in 83% of cases, compared to roughly 56% in boreal and temperate angiosperms. This taxonomic pattern underscores that the conditional nature of the carbon starvation pathway is itself structured by plant lineage.

The Carbon-Water Nexus and Phloem as Proximate Trigger

Recent synthetic work has increasingly emphasized the interdependence of hydraulic and carbon processes rather than treating them as separable pathways. [19] articulated a concept of “carbon-water interdependency,” arguing that hydraulic failure typically kills trees faster than carbon starvation, but that the two processes are mutually reinforcing. Specifically, impaired hydraulic function reduces photosynthetic carbon assimilation, while depleted non-structural carbohydrate (NSC) reserves constrain the osmotic adjustments and repair mechanisms needed to restore hydraulic function.

Within this framework, phloem transport has emerged as a proximate mechanistic link between the two failure modes. As xylem water potential drops, phloem viscosity rises and sugar transport slows, effectively starving sink tissues of carbon even before bulk NSC pools are exhausted [7, 9]. [39] similarly positioned vegetation mortality as the outcome of interacting hydraulic and metabolic cascades, noting that regional-scale die-off events are increasing across multiple biomes. These events are consistently associated with temperature-drought interactions rather than either factor alone—a pattern corroborated by global field observations spanning hundreds of forest die-off events across six continents [14, 2].

[13] extended this integrative perspective into the contemporary context of rising atmospheric CO₂. They argued that elevated CO₂ reduces mortality risk through enhanced stomatal closure and improved water-use efficiency, but that these benefits are frequently offset by compensatory increases in leaf area that restore whole-canopy transpiration to near-ambient levels [12]. This synthesis reinforced the dual-depletion model—simultaneous water and carbon loss—while introducing anthropogenic atmospheric chemistry as a moderating variable with uncertain net effects at the stand scale.

Mutually inclusive interactions between hydraulic and carbon failure pathways, rather than competition between them, are now increasingly recognized as the norm across species and drought regimes [10, 8].

Root Water Uptake Depth as a Reframing Factor

Field observations during the extreme 2018 European drought revealed a significant conceptual challenge to the hydraulic-versus-carbon framing. [40] monitored leaf water status and native embolism formation across nine co-occurring temperate tree species and found that root water uptake depth—not the position of hydraulic thresholds in vulnerability curves or the size of NSC reserves—was the critical determinant of whether a species exceeded its hydraulic damage threshold. Three species—Fagus sylvatica, Carpinus betulus, and Picea abies—exceeded critical water potential thresholds and sustained canopy embolism, while six species with deeper root water access maintained safe water potentials throughout the event. This pattern implies that soil water access, governed by rooting depth and soil texture, may precede and override physiological traits in determining drought survival outcomes, at least under single acute drought events.

Corroborating evidence comes from intraspecific comparisons in Mediterranean oak forests, where stable isotope analyses revealed that non-declining individuals of Quercus pubescens and Quercus cerris consistently accessed deeper soil water at peak summer drought than their declining neighbors. This finding demonstrates that variation in rooting depth can determine survival even among conspecifics sharing the same site [41].

The coordination among rooting depth, xylem hydraulic capacity, and stomatal regulation further suggests that belowground water acquisition is functionally integrated with—rather than independent of—the hydraulic traits foregrounded in mortality frameworks [42]. If robust across systems, this reframing shifts predictive emphasis away from leaf- and stem-level hydraulic parameters toward belowground water acquisition architecture. This domain remains severely underrepresented in controlled mortality experiments and global synthesis datasets [20].

Drought-Impaired Defense Chemistry and the Threshold for Biotic Attack

The physiological consequences of drought extend beyond the hydraulic and carbon pathways that directly cause tree death to encompass the impairment of chemical defense systems. This impairment connects individual-tree physiology to landscape-scale mortality through the mediation of biotic disturbance agents, particularly bark beetles. Trees in drought-stressed states face not only the risk of direct physiological failure but also a dramatically heightened vulnerability to mass attack by insects whose success depends on overwhelming host resistance [3]. The mass-attack strategy exploited by eruptive bark beetle species such as Dendroctonus ponderosae is itself threshold-dependent: beetle populations must achieve sufficient density to saturate resin-based defenses before individual trees can be successfully colonized, and drought-induced defense impairment lowers that colonization threshold substantially [3, 43].

The biochemical foundation of this vulnerability was clarified by [44], who compared phloem and foliar defense chemistry in two mountain pine beetle (Dendroctonus ponderosae) host species — lodgepole pine and jack pine — under controlled water deficit and fungal inoculation treatments. Water deficit inhibited both phenolic production and total monoterpene accumulation in lodgepole pine, directly undermining the resin-based defenses that constitute the primary barrier against beetle colonization. The response was species-contingent, however: jack pine showed minimal alteration in defense compound production under equivalent water stress, yet critically failed to mount any induced monoterpene response to inoculation with beetle-associated fungi [44]. This failure of induced defense in a species that lacks a co-evolutionary history with D. ponderosae helps explain why the beetle’s climate-driven range expansion into boreal jack pine forests has proceeded with such lethality [15, 3]. The novel host simply cannot mount the chemical counterattack that even stressed lodgepole individuals partially sustain. Warming winters have relaxed the thermal barriers that historically constrained D. ponderosae at high latitudes and elevations, enabling this geographic expansion and exposing climatically naïve host communities to an agent against which they have no effective induced-defense repertoire [15].

These tree-level defense failures translate to landscape-scale mortality through cross-scale interactions that compound climatic and structural vulnerabilities. [45] used drone-based remote sensing across 32 sites in California’s Sierra Nevada during and after the catastrophic 2012–2016 drought to track ponderosa pine mortality at fine spatial resolution. Climatic water deficit exerted a strong positive main effect on host mortality rates, but this drought effect interacted non-additively with local forest structure: the proportion of ponderosa pine in a local neighbourhood strongly amplified mortality probability, and larger trees suffered disproportionately higher mortality rates in moisture-stressed settings [45]. Larger trees may be preferential targets because they represent higher-quality resources for beetle reproduction, and their loss can trigger mass-attack dynamics that overwhelm even physiologically intact neighbours. Complementary field evidence from piñon–juniper woodlands further confirms that drought-predisposed trees experience bark beetle colonization and mortality at rates far exceeding those attributable to either stressor in isolation [43], while experimental work on Norway spruce similarly demonstrates that water-limiting conditions measurably reduce resin flow and pitch tube formation — the first physical lines of resistance [46]. These findings collectively underscore that the drought–beetle interaction is not merely additive but synergistic, producing mortality rates that exceed the sum of either stressor acting alone.

This synergistic framing was formalized by [32], who proposed a two-axis counterfactual framework for mortality attribution — asking separately how many trees would have died from drought alone and how many from insects alone, thereby illuminating interaction effects that single-stressor assessments systematically obscure. Their synthesis demonstrated that using drought metrics as straightforward proxies for insect-mediated mortality fundamentally misrepresents causation and underestimates compound vulnerability. The critical implication for the physiological mechanisms discussed throughout this section is that hydraulic failure and carbon starvation do not merely kill trees directly [6, 10]; they also erode the defense capacity that prevents biotic agents from delivering the proximate cause of death. In many die-off events, the tree crosses a defense threshold before it crosses a hydraulic or carbon lethal threshold — a sequencing that has profound implications for understanding mortality timing and for developing predictive models that integrate physiological stress with biotic hazard [32, 43].

Outstanding Gaps and Unresolved Debates

Despite substantial progress in understanding tree mortality, critical knowledge gaps persist. First, the role of phloem turgor collapse as a distinct proximate trigger, initially identified by [7], has not been systematically integrated into mortality models or multi-species syntheses [20, 47]. This integration gap stems partly from the technical challenges of measuring phloem pressure in situ [9].

Second, experimental evidence on how sequential droughts interact with hydraulic and carbon recovery trajectories remains scarce [48, 49]. This scarcity limits understanding of legacy effects that may progressively lower thresholds for subsequent mortality events.

Third, tropical species are dramatically underrepresented in controlled mortality experiments [50, 8], despite comprising the majority of global forest carbon. This underrepresentation creates a significant bias in current understanding.

Fourth, belowground hydraulic processes—including root cavitation, mycorrhizal water transport, and the hydraulic coupling between roots and soil—remain poorly linked to the aboveground indicators that dominate current monitoring frameworks [42, 51].

Fifth, the integration of defense chemistry impairment into process-based mortality models remains nascent. Most frameworks that predict tree death from hydraulic failure or carbon starvation do not simultaneously track the erosion of resin-based resistance, which determines whether biotic agents deliver the proximate mortality event [32, 44].

Addressing these gaps will require improved model–data integration that can capture nonlinear interactions across mechanisms [19, 31]. Ultimately, resolving these knowledge gaps will necessitate coordinated experimental designs that simultaneously track hydraulic, carbon, phloem, defense chemistry, and rooting dynamics across species, sizes, and drought regimes [20].

4. Vapour Pressure Deficit, Atmospheric Moisture Demand, and Stomatal Regulation

Vapour pressure deficit (VPD) — the difference between the water vapour pressure that air could hold at saturation and the amount it actually holds — has emerged as one of the most consequential yet underappreciated drivers of plant water stress, stomatal regulation, and forest mortality. While soil moisture has long occupied the centre of drought research, a growing body of evidence reveals that the atmospheric demand for water constitutes an independent and amplifying stressor that operates even when soil water supply remains adequate. This section traces the development of that understanding, from early syntheses establishing VPD’s role in mortality to the most recent experimental findings that challenge foundational assumptions embedded in standard gas exchange methodology.

VPD as an Amplifier of Drought Stress: Establishing the Framework

Early theoretical and empirical work established that transpiration is driven not by soil moisture alone but by the vapour pressure gradient between leaf interiors and the surrounding atmosphere. Foundational mechanistic research [6] formalized how projected temperature increases of 1.1–6.4°C could raise transpiration demand by 7–48%, pushing both isohydric and anisohydric species toward their respective mortality thresholds through carbon starvation and hydraulic failure. This conceptual framing was pivotal because it positioned VPD not as a secondary correlate of heat but as a primary forcing variable with distinct mechanistic pathways to tree death — pathways now understood to involve cytorrhysis and irreversible cell wall damage as the terminal consequence of sustained water depletion [13].

Building directly on this foundation, Breshears and colleagues [11] provided what remains one of the most explicit arguments for VPD’s independent amplifying role. By synthesizing ecophysiological and climate data, they demonstrated that increasing atmospheric moisture demand intensifies tree water stress and mortality risk beyond what soil moisture deficits alone would predict. Their critical insight was that high VPD can sustain or accelerate hydraulic deterioration even when precipitation is sufficient to maintain moderate soil water content — a finding with profound implications for projecting die-off events under warming climates where VPD trends are steepening independently of precipitation changes [12].

By the early 2010s, concurrent synthetic work reinforced the argument that VPD deserves co-equal status with precipitation as a landscape disturbance driver [19]. McDowell and colleagues’ broader treatment of drought-induced mortality through experimental, observational, and modelling approaches confirmed that VPD accelerates mortality through stomatal closure and the cascading carbon-water interdependency that links hydraulic failure and carbon starvation — processes that interact rather than operate in isolation [13]. Critically, rising VPD also triggers a land-atmosphere feedback loop: increased evaporative demand dries soils, prompting further stomatal closure, redirecting energy toward sensible heat, and elevating VPD still further [12]. This self-amplifying dynamic underscores why atmospheric moisture demand must be treated as a first-order driver of forest decline.

Physiological Mechanisms: Stomatal Responses and Carbon–Water Trade-offs

The mechanistic pathway through which VPD operates is well characterised at the leaf level. Stomata respond to elevated VPD within minutes, closing to limit water loss but simultaneously suppressing CO₂ uptake and photosynthetic carbon gain [12]. Progressive internal leaf unsaturation under increasing VPD further impairs mesophyll conductance and electron transport, compounding the direct stomatal limitation on assimilation [52]. This dual constraint — a source limitation through reduced carbon assimilation and a sink limitation through declining water potential — means that VPD exerts compound effects on tree function that simple hydraulic models tend to underestimate [13]. Novick and colleagues’ 2024 synthesis further highlighted that larger trees face disproportionate mortality risk under rising VPD because height-related water potential gradients make it increasingly difficult to sustain hydraulic continuity from root to canopy [12], a point that connects VPD stress to known size-dependent mortality patterns [9].

The CO₂–VPD interaction introduces a critical and unresolved tension in projections of future forest health. Rising atmospheric CO₂ tends to reduce stomatal aperture, lowering transpiration and theoretically reducing mortality risk through improved water use efficiency [13]. However, this apparent benefit is frequently offset at the ecosystem scale because reduced stomatal conductance often triggers compensatory increases in leaf area index, which collectively maintain or even increase canopy-level water demand [13, 53]. More fundamentally, recent evidence emphasises that rising VPD may negate or overwhelm the stomatal closure benefits of elevated CO₂ [12], particularly under compound warming conditions where extreme heat has been shown to paradoxically force stomata open, accelerating hydraulic depletion and mortality risk [54]. This dynamic is most acute under the compound warming and drying trajectories most consistent with current climate projections. Few experimental studies have simultaneously manipulated CO₂, VPD, and temperature to identify the net mortality threshold [19], leaving this interaction as a pressing open question in the field.

Residual Conductance After Stomatal Closure: Species Variation and Seasonal Stability

Even when stomata are fully closed, water continues to escape across the leaf surface through cuticular pathways and incompletely sealed stomatal pores — a flux quantified as minimum or residual conductance (g_min). This trait has attracted increasing attention as a determinant of survival during prolonged drought or extreme heat events when stomata remain shut for extended periods [9, 13]. After stomatal closure, xylem water potential continues to fall toward critical cavitation thresholds, and the rate at which it does so is governed substantially by g_min — making this trait a direct link between leaf-surface permeability and whole-plant hydraulic failure [9]. A rigorous 2024 study quantifying g_min across nine temperate European tree species during severe drought [55] documented more than a sixfold range in g_min across species — from Pinus at approximately 0.8 mmol m⁻² s⁻¹ to Acer and Sorbus at approximately 4.8 mmol m⁻² s⁻¹ — establishing the substantial interspecific variation in post-stomatal water loss that earlier syntheses had inferred but rarely quantified with this precision [8]. Critically, seasonal variation in g_min was largely absent across most species once leaves had matured, suggesting that this trait is relatively fixed within a growing season and cannot be dynamically upregulated in response to acute drought. This finding has direct implications for understanding why certain species or populations face catastrophic desiccation during extreme events: when stomata close and atmospheric demand remains high [12], g_min becomes the rate-limiting constraint on survival, and species with high cuticular permeability are structurally disadvantaged irrespective of their hydraulic architecture [56].

Uncertainty persists regarding the relative contributions of cuticular versus stomatal pathways to g_min and how each component responds to chronic high VPD acclimation over longer timescales [12]. Data on how g_min varies with leaf age, canopy position, or developmental acclimation to elevated VPD remain limited [55], representing a clear gap that will require controlled factorial experiments across diverse species and canopy positions to resolve.

Challenging a Foundational Assumption: Intercellular Airspace Humidity

Perhaps the most conceptually disruptive recent contribution to this theme is the finding by [52] that the intercellular airspaces of leaves are not fully saturated under elevated VPD — directly contradicting an assumption that has been embedded in standard gas exchange models for decades. Using a combination of gas exchange measurements and carbon and oxygen isotope discrimination across four species, Diao and colleagues found that intercellular relative humidity declined by approximately 0.10 per 1 kPa increase in VPD, reaching a mean of 0.73 at the highest VPD tested. When this unsaturation was explicitly accounted for in model calculations, the inferred decline in stomatal conductance with VPD was less pronounced than standard models suggested, while mesophyll conductance showed significant decline with VPD only when unsaturation was incorporated.

The implications of this finding extend well beyond the immediate study. Rising VPD is already recognised as a pervasive and intensifying driver of forest water stress [12], and stomatal and mesophyll conductance are central parameters in the mechanistic models used to diagnose and predict woody plant mortality under drought [13, 19]. If published conductance values across the literature have been calculated under the assumption of full intercellular saturation, a systematic bias may pervade our understanding of how leaves regulate gas exchange under high VPD. This raises urgent questions about the accuracy of model parameterisations used in process-based mortality models [57, 8] and the reliability of trait databases compiled from standard gas exchange protocols. The 2024 vintage of this finding means that the broader community has only begun to grapple with its ramifications, and systematic reassessment of published conductance data under this revised framework has not yet occurred.

Integration into Mortality Modelling: Gaps and Trajectories

Despite considerable mechanistic progress, VPD remains insufficiently integrated into process-based forest mortality models, which continue to treat soil moisture as the dominant drought variable [12, 11, 13]. This structural omission likely causes systematic underestimation of mortality risk in warming climates where VPD trends are accelerating independently of soil moisture trajectories [14, 12], a decoupling that global tree die-off records increasingly confirm [14]. Notably, VPD’s role extends beyond direct physiological stress on trees to influence the population dynamics of biotic disturbance agents: recent empirical work has demonstrated that VPD during the critical June–July swarming period serves as a threshold trigger for bark beetle outbreaks, with exceedance of 0.84 kPa VPD increasingly likely under projected warming [58]. Drought-stressed trees are simultaneously rendered more susceptible to beetle colonisation through reductions in resin-based defences [32, 46], while warming-driven phenological shifts further synchronise beetle emergence with vulnerable host conditions [3]. This dual pathway — VPD simultaneously stressing host trees and facilitating beetle outbreak initiation — represents a compound hazard that current models rarely represent [4]. Resolving these gaps will require coordinated multi-factor experiments [59], improved representation of atmospheric demand in mortality model architecture, and the integration of newly characterised traits such as g_min [55] and intercellular humidity unsaturation [52] into next-generation frameworks.

Synthesis: Atmospheric Demand as a Structural Driver of Forest Vulnerability

Taken together, the findings reviewed in this section establish VPD not as a peripheral modifier of soil moisture stress but as a structurally independent driver of forest mortality operating through at least three distinct mechanisms: the direct amplification of hydraulic and carbon stress even under moderate soil water availability [11, 12], the imposition of a fixed within-season floor on water loss once stomata close — a floor that varies sixfold across species and cannot be rapidly adjusted to acute atmospheric demand [55] — and the systematic distortion of how stomatal and mesophyll conductance respond to high VPD, now that intercellular airspace unsaturation has been demonstrated to invalidate a foundational modelling assumption [52]. These three findings collectively imply that forests are more vulnerable to atmospheric drying than either physiological models or mortality projections have generally acknowledged [2].

That vulnerability acquires particular urgency when considered alongside the global-scale evidence examined in the following section. Global field observations spanning 1,303 plots across five decades and all forested biomes have identified a consistent “hotter-drought fingerprint” — a compound signal of elevated maximum temperatures, VPD, and cumulative water deficit combined with reduced precipitation and soil moisture — as the dominant climatic precondition for widespread tree die-off [14]. Under a +4°C warming scenario, the frequency of climate conditions meeting this mortality threshold is projected to increase by approximately 140%, rising from around 1.62 years per decade under current conditions to 3.88 years per decade [14]. If VPD trends are steepening independently of precipitation across forested biomes [12], then the atmospheric demand dynamics characterised here are likely to operate as an accelerant within precisely the climatic conditions most predisposed to triggering the landscape-level mortality cascades and post-die-off compositional shifts [26, 1] that Section 5 documents.

5. Global Patterns, Climatic Fingerprints, and Regional Case Studies of Forest Die-Off

Forest die-off events have emerged as one of the most consequential ecological signals of anthropogenic climate change, yet characterising their global distribution, climatic triggers, and biome-specific dynamics has required decades of cumulative synthesis. Early observational work was scattered across regions and disciplines, but a growing body of research now documents the spatial and temporal fingerprints of tree mortality with increasing rigour — from global databases of field observations to fine-grained isotopic case studies in individual stands. The most recent contributions, particularly from 2024 and 2025, have both confirmed and complicated this picture, revealing that even forests historically adapted to their local climates now face unprecedented mortality risk.

The Hotter-Drought Fingerprint: Global Evidence and Its Implications

The foundational synthesis for understanding global tree mortality patterns comes from Hammond and colleagues, whose analysis of 1,303 plots drawn from 154 peer-reviewed studies spanning five decades and all forested biomes identified a globally consistent ‘hotter-drought fingerprint’ associated with die-off events [14]. Across diverse elevations, climate zones, and forest types, tree mortality was characteristically preceded by atmospheric conditions combining anomalous heat with water deficit — not drought alone. The physiological basis for this dual sensitivity is well established: heat exacerbates hydraulic failure by elevating vapour pressure deficit and accelerating xylem embolism, while simultaneously depleting non-structural carbohydrate reserves through increased respiratory demand [9, 13]. Crucially, the frequency with which such mortality-triggering conditions occur increases nonlinearly with warming, rising from approximately 1.62 events per decade under current conditions to 3.88 per decade under +4°C warming, a projected 140% increase [14]. This nonlinearity is significant: it implies that incremental temperature rises will produce disproportionate increases in die-off risk, undermining any assumption that forest systems will track warming gradually.

That this hotter-drought fingerprint operates not only through direct physiological damage but also through the activation of biotic disturbance cascades has become increasingly clear. Drought stress reduces trees’ capacity to produce defensive secondary metabolites, lowering resistance thresholds for bark beetle attack and defoliator outbreaks — interactions that produce nonadditive, compounding mortality well beyond what either stressor alone would generate [32, 58]. [2] argued forcefully that global vulnerability to tree mortality was being systematically underestimated — not merely by models but by the broader scientific and policy community — because recent hot droughts represent the most severe moisture conditions in 800–1,200 years for multiple regions based on tree-ring evidence, and because tree mortality consistently occurs faster than recovery growth can compensate [49]. This asymmetry establishes inherent frequency thresholds beyond which ecosystems cannot recover between disturbance events, a concern compounded by the observation that conventional assessments treat biotic agents as secondary factors rather than as co-drivers whose synergistic interaction with hotter drought amplifies mortality beyond what any single-stressor framework can capture [2, 32].

Post-Mortality Replacement Patterns: Self-Replacement as the Exception

The consequences of these die-off events for forest persistence depend critically on what follows mortality — a dimension that global fingerprint analyses have only recently begun to address. A systematic analysis of 131 globally distributed sites with documented drought-related tree mortality revealed that self-replacement of dominant species occurred at only 21% of sites, while replacement by shrubs or other woody species — often representing more xeric functional types — occurred at 69% of sites [26]. This finding fundamentally reframes forest die-off from a temporary setback to a gateway for persistent structural reorganisation. The rarity of self-replacement implies that assumptions embedded in forest recovery models — that disturbed forests will recapitulate pre-disturbance composition — are empirically unjustified across a wide range of systems. Shrub-dominated replacement states typically store substantially less aboveground carbon, and where they represent alternative stable states maintained by drought-fire feedbacks, may resist reversion to forest cover under continued climatic stress [26, 29, 22]. Global field observations spanning 675 georeferenced sites further reinforce this concern, showing that the frequency of mortality-triggering climate conditions is projected to increase by up to 140% under a +4°C warming scenario, compressing the recovery windows available for self-replacement [14]. This global replacement pattern provides the essential demographic context for the regional case studies that follow: the question is not merely whether trees die, but whether forests return.

A persistent limitation of these global analyses is the underrepresentation of African and South and Southeast Asian forests in mortality databases, despite these regions containing substantial forest area and facing accelerating climate pressures [14, 50]. This sampling bias means global fingerprint characterisations are weighted toward temperate and boreal systems, and mortality thresholds derived from these datasets may not translate reliably to understudied biomes [20]. Compounding this, global analyses have generally treated biotic agents — bark beetles, root pathogens, mistletoes — as secondary or coincidental factors rather than as co-drivers interacting dynamically with climatic stress [32, 1], an integration that remains largely absent from fingerprint-scale syntheses. Indeed, drought-weakened trees face sharply elevated susceptibility to bark beetle attack and fungal pathogen proliferation, interactions that can amplify and spatially extend mortality far beyond what climate stress alone would predict [32, 3].

Temperate European Forests: The 2018–2022 Stress Sequence

While global databases establish statistical patterns, regional case studies reveal the mechanisms and management feedbacks that aggregate mortality data cannot. The sequence of heat and drought events affecting European forests between 2018 and 2022 provides one of the most intensively documented recent examples. A 2025 interdisciplinary analysis integrating peer-reviewed literature, government reports, and remote sensing data found that Central Europe was the most severely affected region, with widespread crown thinning, bark beetle outbreaks of exceptional scale, and economic losses — in the Czech Republic, bark beetle infestations produced timber volumes exceeding the country’s normal annual roundwood production [60]. These cascading impacts exemplify the compound mortality pathway: drought weakens tree defences by reducing resin production and carbon reserves available for defensive chemistry [46], enabling biotic agents to cross damage thresholds they would not reach under hydraulic stress alone [43]. The mechanistic basis of this interaction is now well established — water stress impairs xylem transport and depletes non-structural carbohydrates, progressively compromising the constitutive and induced defence systems on which conifers rely to resist phloem-feeding beetles [13].

Process-based modelling has confirmed that these European outbreaks represent a fundamental shift in outbreak causation. [58] employed the iLand landscape model with a newly developed drought-triggered outbreak module to simulate Ips typographus dynamics in Central European spruce forests and found that incorporating drought as the primary inciting factor — rather than relying solely on windthrow-based initiation as prior models had done — increased explained cumulative mortality variance from 6.6% to 47.5%, achieving an R² of 0.86. This is not a marginal improvement but a reframing: drought has displaced windthrow as the dominant contemporary outbreak trigger for European spruce bark beetle, and its importance will grow as vapour pressure deficit during the critical June–July period increasingly exceeds the identified outbreak threshold of 0.84 kPa, with up to 46% of years projected to surpass this value under RCP8.5 by 2071–2100 [58]. Complementary empirical evidence from a network of 158 pheromone-baited traps distributed across seven million hectares found that hotter drought conditions increase I. typographus catch by approximately 2,000 beetles per trap per degree Celsius of temperature increase under drought, that critical swarming thresholds are reached 4–7 days earlier per degree of warming, and that beetle populations exhibit rare landscape-scale synchronization extending across hundreds of kilometres during drought years [61]. This spatial synchronization is particularly alarming because it means that the natural buffering provided by asynchronous local outbreak dynamics — a key stabilizing feature of bark beetle population ecology at regional scales [3] — breaks down precisely when climatic conditions are most severe.

One finding from the broader European analysis runs counter to prevailing expectations from global fire-weather research: no statistically significant increase in wildfire incidence was observed across the study zones during 2018–2022 relative to 2010–2014 [60]. This is a notable tension. Global analyses of fire-weather trends project rising fire hazard and burned area as drought intensifies and temperatures climb [62], yet the European record suggests that management interventions — including altered fuel management, early warning systems, and firefighting capacity — may partially offset drought-driven fire risk at regional scales. Whether this suppression of the fire signal is a durable management achievement or a temporary lag in fire regime transition remains an open and consequential question [62].

Scots Pine: Mortality Shifting into Optimal Range Areas

The Scots pine (Pinus sylvestris) literature provides a particularly instructive case of how die-off patterns can confound prior ecological assumptions. A 2024 Tamm Review synthesising the drivers, mechanisms, and trends of drought-induced Scots pine mortality documents that high mortality rates are now being reported from the species’ climatic optimum areas — zones where growth conditions were historically more favourable and where locally adapted populations were expected to be most resilient [63]. This reversal challenges the assumption that mortality risk concentrates at range margins. Drought triggers self-thinning and amplifies the activity of biotic agents including bark beetles and mistletoes [32, 33], with post-mortality stand densities falling far below pre-drought levels — a structural legacy that persists well beyond the mortality event itself [63]. The drought–bark beetle interaction is particularly consequential: water stress compromises resin-based defences, lowering the threshold at which beetle populations can overwhelm host trees and transition from endemic to epidemic dynamics [58, 32].

The long-term physiological precursors of Scots pine mortality are captured with particular precision in dendrochronological and isotopic analyses. Timofeeva and colleagues, working in a dry Swiss environment, demonstrated that trees that ultimately died showed reduced radial growth beginning in the mid-1980s — decades before mortality — concurrent with rising temperatures and vapour pressure deficit, and that this decline was accompanied by increasingly conservative water-use strategies as reflected in stable carbon isotope ratios (δ¹³C) in tree rings [64]. Complementary dendrochronological work confirms that pre-mortality growth suppressions detectable in ring-width series can serve as early warning signals distinguishing trees destined to die from survivors years in advance [65]. The implication is that mortality events represent the endpoint of gradual physiological deterioration rather than sudden system failure, and that ring-width and isotopic records can serve as early warning archives if monitored at sufficient temporal resolution.

Complementary evidence from the dry Alpine valleys of Switzerland demonstrates that these gradual mortality trajectories can culminate in species-level replacement. Analysis of Swiss National Forest Inventory data across 201 grid points between 1983 and 2003, combined with annual mortality records, reveals that Scots pine mortality was highest at low elevations below 1,000 m and correlated strongly with summer drought indices, with mortality rates three to four times higher than Swiss national averages [66]. Mortality was further amplified on drier sites with high stand competition and concentrated disproportionately among small-diameter trees — a pattern suggesting that the replacement process operates through differential vulnerability within the pine population rather than wholesale stand collapse [66]. The beneficiary of this transition is pubescent oak (Quercus pubescens), a more drought-tolerant species progressively filling the space vacated by declining pine. This elevation-gradient pattern foreshadows a broader upslope displacement of drought-adapted communities [67], making the Alpine case an early-detection sentinel for dynamics likely to intensify across temperate mountain systems and reinforcing the global pattern in which only a minority of drought-killed forests undergo self-replacement [26, 66].

Climate-Driven Bark Beetle Outbreaks, Voltinism Shifts, and Novel Host Encounters

The geographic expansion and intensification of bark beetle outbreaks constitute one of the most consequential global patterns of climate-driven forest mortality, yet they have historically been treated as regional entomological phenomena rather than as an integral component of the global die-off picture. Recent work has overturned this separation, establishing that climate warming is restructuring beetle population dynamics, voltinism, and host range at continental scales through mechanisms that are both direct — operating on beetle physiology — and indirect — operating through drought-mediated suppression of host defences.

The conceptual architecture for understanding outbreak dynamics as threshold-based cascade phenomena was substantially advanced by [3], whose cross-scale framework reconceptualised eruptions as processes governed by opposing positive and negative feedbacks operating simultaneously at the levels of individual tree entry, stand-level colonisation, and landscape-scale spread. A central insight of that work is that the conditions initiating an outbreak need not persist to sustain it: once positive feedbacks — such as mass-attack pheromone signalling that overwhelms tree resin defences — become self-reinforcing, the eruption develops internal momentum. The colonisation sequence typically progresses from initial host location via kairomones, to mass aggregation driven by aggregation pheromones, to eventual cessation of attack signalled by anti-aggregation compounds, a communication architecture that climate-driven population growth can push beyond the threshold at which tree defences are overwhelmed [68]. Critically, forest management history acts as an anthropogenic amplifier of these dynamics: commercially managed monocultures and high-density even-aged stands increase the proportion of physiologically susceptible trees and reduce the structural heterogeneity that would otherwise constrain landscape-scale spread [3, 4].

The direct physiological pathway through which warming reshapes beetle biology centres on temperature-dependent life-history traits. [69] provided foundational synthesis demonstrating that species differ substantially in their cold-induced mortality thresholds and developmental timing requirements, meaning that warming does not translate uniformly into increased beetle performance across all species and regions. For Dendroctonus ponderosae and Dendroctonus rufipennis in western North America, warming was projected to reduce overwinter mortality and enable more synchronised, univoltine adult emergence — the latter being critical because population-level synchrony is required for the mass-attack coordination that overcomes tree defences [69]. Northward and elevational range expansions were projected as formerly climatically inhospitable zones crossed thermal suitability thresholds, predictions that subsequent monitoring data have broadly confirmed [15]. Dendroctonus ponderosae alone has killed more than 11 million hectares of pine across North America over a thirteen-year period, with its range continuing to expand northward under warming temperatures [68]. In Europe, analogous dynamics for Ips typographus have been documented, with rising temperatures enabling multiple voltine generations per season — from a historical one or two up to three in some central European localities — where thermal accumulation now exceeds the degree-day thresholds required for additional cohort development [4, 15]. Additional generations compress the temporal window between disturbance events, reduce the opportunity for host-tree recovery between attack cycles, and alter phenological synchrony with natural enemies.

The consequences of range expansion are compounded where beetles encounter novel hosts that lack co-evolutionary defences. The mountain pine beetle’s incursion into boreal jack pine forests east of its historical lodgepole pine range exemplifies this dynamic: jack pine fails entirely to upregulate monoterpene defence in response to beetle-associated fungal inoculation, whereas even drought-stressed lodgepole pine retains a partial induced response [44]. This immunological asymmetry in novel hosts suggests that as beetle ranges shift poleward and to higher elevations under continued warming, the forests they encounter may be structurally and biochemically unprepared for mass attack [15] — a vulnerability that the limited genetic diversity in key hydraulic and defense traits among conifers [56] is unlikely to resolve on management-relevant timescales.

A further layer of complexity arises from the role of mutualistic fungal symbionts in mediating bark beetle virulence. Many aggressive bark beetle species maintain obligate fungal symbionts housed in specialised mycangia, which provide nutritional supplementation for larvae and actively contribute to host-tree mortality by disrupting resin flow [68]. [70] demonstrated that altered atmospheric and moisture conditions differentially affect beetle reproduction and the performance of associated fungal associates, raising the possibility that climate change could either strengthen or disrupt the mutualistic interactions that underpin beetle colonisation success. This introduces a biotic interaction pathway into outbreak risk that operates independently of — and potentially in tension with — the thermal developmental pathways emphasised in earlier literature, and whose trajectory under future climates remains deeply uncertain.

The continental scope of these changes was synthesised by [16], whose comprehensive North American review established that climate change was already driving expansions of insect and disease disturbances that met or exceeded predictions from national climate assessments. The physiological sensitivity of forest insects to temperature — through short generation times, high reproductive potential, and direct metabolic effects on voltinism — makes them among the most climate-responsive components of forest ecosystems [16, 15]. Together with the evidence from European systems, this body of work establishes that bark beetle outbreaks are no longer peripheral to the global forest die-off narrative but are among its most dynamic and consequential expressions.

Pacific Northwest: Changing Fire, Changing Forests

The Pacific Northwest of the United States provides a regionally focused synthesis of how changing wildfire regimes interact with drought and insect outbreaks to reshape forest trajectories. [71] integrated historical records, observed trends, and projected futures to demonstrate that wildfires in the region are increasingly associated with warm, dry atmospheric conditions that simultaneously reduce fuel moisture, extend fire seasons, and increase the probability of high-severity burning over large contiguous areas. Critically, their synthesis highlights that fire does not operate in isolation: compound events combining fire with drought and bark beetle outbreaks are projected to become primary drivers of ecosystem state change, with reburns occurring more frequently on landscapes where insect-killed stands have substantially elevated fuel loads [71, 72, 3]. Wildfire-driven conversion away from conifer-dominated forest cover is already documented across western North America, with stand-replacing fires and subsequent regeneration failures interacting to push some sites toward persistent non-forest states [29, 22]. Post-fire conifer regeneration in these systems faces a narrowing establishment window as warming temperatures suppress seedling survival at lower elevations and on drier aspects [17, 73] — a dynamic whose demographic consequences are explored in greater depth in Section 7. Short-interval reburns compound this problem further, as successive fires remove seed sources and exhaust soil seedbanks before recovery can occur [18, 23]. These projected trajectories reinforce the broader pattern that fire regime change is most consequential not in its direct effects but in its interaction with other stressors to create compound disturbance sequences from which recovery is uncertain.

Aspen Decline and Mediterranean Oak Dieback: Mechanistic Debates

In western North America, sudden aspen decline (SAD) following the 2002 drought in Colorado has generated a productive mechanistic debate. Anderegg and colleagues linked the die-off to unprecedented growing-season temperatures and evaporative deficit, with isotopic analysis of xylem water demonstrating that aspens depend primarily on shallow soil moisture and have limited plasticity in their water-use strategies [74]. When shallow stores were depleted to record lows, mortality cascaded at landscape scales. Critically, subsequent work demonstrated that drought induces cumulative, multiyear hydraulic damage — a process of “cavitation fatigue” in which xylem becomes progressively more vulnerable to water transport failure, leaving even surviving trees more susceptible to subsequent droughts [48]. An alternative framing, associated with Worrall and colleagues, emphasises evaporative demand rather than hydraulic failure as the proximate driver — a distinction with practical implications for projecting mortality under future climates. If hydraulic failure mechanisms dominate [56], then traits governing cavitation resistance become primary targets for understanding population-level vulnerability; if evaporative demand and shallow soil moisture depletion prevail [74], then rooting depth and soil water-holding capacity become more decisive. These drivers are not mutually exclusive [10], but the relative weight assigned to each shapes both modelling approaches and management responses.

Mediterranean oak dieback introduces a further spatial dimension to the driver debate. A study of Quercus cerris and Quercus pubescens in southern Italy used stable water isotope analysis of soil and xylem water sampled at peak summer drought to demonstrate that non-declining trees consistently accessed deeper soil water pools than their declining counterparts [41]. Growth divergences between declining and non-declining trees were detectable as far back as 2002 for some species [41], reinforcing the pattern of protracted pre-mortality decline seen in Scots pine [65], and pointing to intraspecific variation in rooting behaviour — rather than species-level climate tolerance — as a critical determinant of individual survival. This finding has implications for assisted migration and selective management strategies: within a single stand, hydraulic access to deep water may predict survival more reliably than species identity.

Tropical Forests and Planted versus Natural Forest Resilience

Tropical forest mortality introduces a distinct set of challenges, particularly regarding the separation of background mortality from drought-driven pulses. A comprehensive 2018 review documented multi-decadal increases in individual tree mortality rates across the Amazon Basin, with parallel trends in Southeast Asia and remotely sensed canopy loss signals extending to the Congo [50]. A pantropical synthesis further underscores how drought and fire interactions accelerate these carbon cycling disruptions across all major tropical biomes [75]. Particularly concerning is the disproportionate mortality of large trees during drought events — individuals that contribute disproportionately to forest carbon stocks and structural complexity [50, 9]. Large trees are especially vulnerable due to their greater hydraulic path lengths and the outsized proportion of stand-level biomass they represent, meaning their loss triggers non-linear reductions in carbon sink capacity [5]. However, insufficient long-term monitoring networks in tropical regions mean that distinguishing baseline demographic change from acute climate-driven mortality episodes remains deeply problematic [20]. The African tropical forest system is especially data-sparse, creating a significant blind spot in global assessments — a gap that broad-scale remote sensing has only partially addressed [50].

In China, the contrast between planted and natural forest responses provides a policy-relevant complement to these biogeographical analyses. Using satellite-derived productivity data across 2001–2020, Ma and colleagues found that planted forests exhibit significantly lower drought resistance and resilience than natural forests, particularly in subtropical broad-leaved evergreen and warm temperate deciduous zones [76]. Between 2011 and 2020, planted forest drought resistance increased while resilience declined relative to the prior decade, indicating a shift toward conservative stress-avoidance strategies at the cost of recovery capacity — a trajectory with implications for China’s large-scale afforestation programmes [76]. Declining forest resilience under anthropogenic climate change more broadly mirrors this pattern, with satellite-era analyses identifying widespread signals of reduced post-disturbance recovery capacity across multiple forest biomes [77].

Adaptive Capacity and the Limits of Local Adaptation

Running through this body of work is a recurring tension concerning adaptive capacity. Brodribb and colleagues argue that forests have limited ability to adapt rapidly to novel drought regimes because migration rates are slow and genetic diversity in key hydraulic traits — especially in conifers — is low [56]. This vulnerability is compounded by a striking finding from global hydraulic trait databases: approximately 70% of forest species worldwide operate with narrow hydraulic safety margins of less than one megapascal, functioning perilously close to the threshold of damaging embolism under natural drought conditions — and these margins are largely independent of mean annual precipitation, suggesting that even species in historically moist climates are structurally exposed [34]. The Scots pine review provides striking empirical corroboration: mortality is now concentrated precisely where local adaptation should have been strongest, suggesting that even well-adapted populations cannot buffer against the pace of contemporary climate change [63]. The bark beetle range expansion literature reinforces this concern from the biotic dimension: novel host species encountered by range-shifting beetles lack the co-evolutionary defence capacity that even imperfect adaptation provides [44, 69, 78], meaning that adaptive limitations operate simultaneously on trees’ abiotic stress tolerance and their biotic resistance. Together, these findings push against optimistic scenarios in which local genetic adaptation buffers mortality risk, and reinforce the importance of understanding forest die-off not as an exceptional disturbance but as a structurally predictable consequence of warming trajectories already underway.

Cross-Biome Synthesis and Implications for Land–Atmosphere Dynamics

The regional and taxonomic diversity of evidence assembled across this section converges on a set of patterns that are consistent enough to constitute a global signal. The hotter-drought fingerprint documented by [14] is not merely a statistical abstraction: it finds expression in Central European spruce stands overwhelmed by beetle synchronisation [61, 58], in Swiss pine populations replaced by oak along elevation gradients [66], in drought-killed aspen clones across Colorado [74], and in tropical canopies losing their largest individuals to intensifying moisture stress [50]. The 21% self-replacement rate documented globally [26] means that the dominant outcome of die-off — across biomes, management histories, and forest types — is structural reorganisation toward less complex, lower-carbon vegetation states, a transition that planted forests appear particularly ill-equipped to resist [76].

What this cross-biome convergence implies for land–atmosphere interactions is consequential: dying and structurally simplified forests alter surface energy balances, reduce transpirational cooling, and modify fuel loads in ways that feed back directly onto the atmospheric conditions — elevated vapour pressure deficit, intensified fire weather, amplified drought — that drove their mortality in the first place [79, 12]. Plant functional trait composition in replacement communities governs the magnitude of these biophysical feedbacks, with low-stature shrub assemblages substantially reducing latent heat flux and amplifying sensible heating relative to the closed-canopy forests they displace [79]. Insect-killed standing deadwood, meanwhile, undergoes well-documented transitions in fuel structure and flammability over post-outbreak decades, shaping fire behaviour and atmospheric smoke loading at landscape scales [72, 80].

The shrub-dominated replacement states and insect-killed standing deadwood that increasingly characterise post-mortality landscapes represent not endpoints but altered initial conditions for the fire–climate and drought–heat feedbacks examined in the following section. Understanding the land–atmosphere consequences of these vegetation transitions — and the degree to which they amplify or suppress the compound extremes already accelerating forest decline [13] — is therefore a necessary complement to the mechanistic mortality picture assembled here.

6. Land–Atmosphere Feedbacks, Compound Extremes, and Fire–Climate Interactions

The preceding sections established, at the level of individual trees and stands, how hydraulic failure, carbon starvation, and VPD-driven atmospheric demand interact to produce mortality across biomes — from Scots pine at its southern range margin to tropical giants in the Amazon and drought-stressed conifers in semi-arid North America. Yet these tree-level and species-level mechanisms do not operate in isolation from the land surface on which they occur. Forests exchange energy, water, and carbon with the atmosphere in ways that can amplify or suppress the very climatic drivers — soil moisture deficits, VPD, surface temperature — that determine whether individual trees survive or die. Understanding land–atmosphere coupling is therefore the logical next analytical layer: it situates the physiological and biogeographical patterns reviewed in Sections 3–5 within a regional climate system whose feedbacks can either moderate or accelerate the mortality trajectories already described.

Land–atmosphere feedbacks, compound extreme events, and fire–climate interactions form an interconnected system in which vegetation, soil, and atmosphere exchange energy and water in ways that can amplify or suppress climatic extremes, reshape fire regimes, and ultimately reconfigure ecosystem state. Understanding these feedbacks has advanced substantially over the past decade, though recent work has both deepened mechanistic insight and introduced new paradoxes that complicate earlier, more optimistic interpretations of vegetation–climate coupling.

Soil Moisture–Atmosphere Feedbacks and the Role of Plant Functional Traits

Early observational work demonstrated that land–atmosphere feedbacks are not spatially uniform, but vary systematically with climate regime and vegetation type. Analysing data from 40 FLUXNET2015 eddy covariance towers, [79] showed that positive drought-intensification feedbacks — in which low soil moisture amplifies temperature, vapour pressure deficit (VPD), and sensible heat flux — were substantially more prevalent at warmer, drier sites, while negative feedbacks dominated at cooler sites. Critically, climate and plant functional traits together explained 54–67% of cross-site variation in soil moisture feedbacks [79], establishing that vegetation is not a passive responder but an active modulator of regional drought dynamics. This finding elevated plant functional trait diversity from a community-ecology concern to a land-surface physics problem: species composition shapes whether drought self-intensifies or self-limits, with consequences that cascade from the leaf to the regional atmosphere.

VPD occupies a pivotal position within these feedback pathways. Early synthesis work recognised that atmospheric moisture demand amplifies tree water stress and mortality risk well beyond what soil moisture deficits alone would predict [11]. More recently, [12] extended this understanding by documenting that elevated VPD triggers rapid stomatal closure within minutes, simultaneously constraining photosynthesis through both carbon source limitation and hydraulic sink limitation, and that larger trees face disproportionate mortality risk because height-related water potential gradients intensify vulnerability to hydraulic failure. The underlying mechanisms span carbon starvation, hydraulic failure, and their interaction [13], with the relative importance of each pathway varying with species traits and drought duration [9]. Together, these studies position VPD as the critical land–atmosphere coupling variable that bridges leaf-scale physiology and regional climate: rising VPD is simultaneously a consequence of warm, dry land surfaces and a driver of further stomatal restriction, reduced transpirational cooling, and continued surface warming — a positive feedback loop whose intensification potential is amplified by compound soil drought–heat extremes projected to increase under future warming [53, 81]. These dynamics are most acute at the warmer sites identified by [79], where the transition from negative to positive land–atmosphere coupling is already underway.

Root architecture adds a further layer of within-stand heterogeneity to these feedbacks. Tracking nine co-occurring temperate European species through the exceptional 2018 drought, [40] found that root water uptake depth — rather than hydraulic thresholds or carbohydrate reserves — was the primary determinant of species-level drought vulnerability. Species drawing water from deeper horizons maintained water potentials above critical embolism thresholds even as shallow-rooted neighbours experienced canopy dieback [40]. This hydrological partitioning within a single stand implies that community-level land–atmosphere feedbacks depend not only on which plant functional types are present but on how they partition the soil water column — a dimension rarely incorporated into Earth system models that treat canopy conductance as a bulk property [82], and one whose omission may systematically bias projections of regional drought intensification in structurally diverse forests.

The Paradox of Vegetation Greening and Compound Drought–Heat Extremes

The recognition that increasing atmospheric CO₂ and warming drive widespread vegetation greening — often framed as a beneficial carbon sink enhancement — has been substantially complicated by more recent modelling work. Using 17 CMIP6 Earth System Models, [53] demonstrated that vegetation-driven changes are projected to increase global compound soil drought–heat likelihood by a factor of 2.2 by the end of the 21st century, equivalent to approximately 14% of the total increase attributable to climate change. The dominant mechanism is albedo reduction: current-season greening decreases surface reflectance, warming the land surface and increasing sensible heat flux in ways that reinforce soil drying and heat accumulation [53]. This biophysical pathway is compounded by a secondary feedback in which reduced soil moisture suppresses plant transpiration, further shifting the surface energy balance toward sensible rather than latent heat — a process whose strength is significantly mediated by plant functional traits governing hydraulic water transport capacity and maximum leaf gas exchange [79]. This finding inverts a widespread assumption and introduces a counterintuitive tension: the same vegetation expansion celebrated for sequestering carbon simultaneously intensifies the compound extremes that threaten vegetation survival. The result is a feedback architecture in which greening and drought-heat co-occurrence are not independent trends but mutually reinforcing processes — a dynamic that compound extreme projections rarely incorporate alongside dynamic vegetation feedbacks [53], and one that current land surface models may systematically underestimate by failing to represent plant hydraulic diversity [79].

Fire Weather, Burned Area, and the Human–Climate Paradox

Fire represents the most dramatic expression of ecosystem–atmosphere coupling failure, converting decades of carbon accumulation into atmospheric forcing within days. A comprehensive synthesis of satellite observations and meteorological reanalyses documented that global fire weather season length increased by 27% and extreme fire weather days by 54% from 1979 to 2019 [62] — a trajectory consistent with the VPD intensification and drought self-amplification mechanisms described above [12, 79]. Rising VPD directly elevates fire weather danger by desiccating live and dead fuel moisture, while simultaneously suppressing photosynthesis and accelerating the accumulation of combustible biomass [12]. Yet the same synthesis documented a 27% decline in global burned area from 2001 to 2019, with African savannahs accounting for 50–59% of the reduction, driven predominantly by agricultural expansion and landscape fragmentation rather than climate amelioration [62]. This apparent disconnect — fire weather intensifying while fire occurrence declines — represents one of the more provocative findings in contemporary fire ecology [83], because it reveals the extent to which human landscape modification suppresses the climatic signal in burned area statistics at the global scale.

Regional analyses have further complicated this picture. Evidence from Europe suggests that even during periods of exceptional drought, fire activity need not increase detectably [62, 60], implying strong regional heterogeneity in fire–climate relationships that global analyses may obscure. In contrast, climate-driven fire intensification is strongly expressed in regions such as California and British Columbia, where anthropogenic warming has doubled burned forest area and driven abrupt, step-change increases in fire activity since the mid-2000s [84, 85]. The divergence between fire weather potential and realised fire activity underscores a fundamental methodological tension: fire weather indices capture climatic predisposition, but actual ignition, spread, and suppression are mediated by land cover, human behaviour, and fire management capacity in ways that currently defy simple integration with physical climate projections [86].

Bark Beetle Mortality, Fuel Dynamics, and Compound Fire Risk

The interaction between bark beetle outbreaks and wildfire introduces a compound disturbance pathway whose consequences for fire behaviour are temporally structured, stage-dependent, and deeply entangled with the same drought dynamics that drive both processes. Early intuitions equated beetle-killed forests with uniformly higher fire risk; research over the past decade has substantially complicated this picture [3].

Field-based chronosequence research in Colorado lodgepole pine forests systematically compared fuel structure and expected fire behaviour across four mountain pine beetle (MPB) impact stages: Green (pre-outbreak or live), Red (recently killed with retained red needles), Grey (needles shed, snags standing), and Old-MPB (advanced decay, substantial fallen woody debris) [72]. Surface fireline intensity was highest not in the Red stage — when public concern about beetle-killed forests typically peaks — but in the Grey and Old-MPB stages, driven by the progressive accumulation of large-diameter surface fuels as snags collapse and fine needle material is lost from the canopy [72]. The more acute fire hazard in recently killed Red-stage stands lies not in surface fire intensity but in the lowered threshold for crown fire initiation: dead foliage retains its physical structure but loses the moisture content that buffers live needles against ignition, substantially reducing the critical surface fireline intensity required to transition a surface fire into an active crown fire. Wind speed requirements for sustaining active crown fire are correspondingly reduced during this stage. However, under extreme burning conditions — high temperature, low humidity, strong winds — weather emerges as the dominant driver of crown fire activity regardless of beetle-impact stage [72], suggesting that the marginal contribution of beetle outbreak stage to crown fire risk is most meaningful under moderate rather than extreme fire weather.

In Grey and Old-MPB stages, needle loss opens the canopy and reduces canopy bulk density, which paradoxically lowers the probability of sustained crown fire even as surface fire intensity rises [72, 23]. The canopy, stripped of fine fuels, becomes less capable of propagating fire laterally through the aerial fuel bed, shifting the dominant fire type toward intense but surface-confined burning. This stage-dependent dynamic underscores the inadequacy of treating beetle-affected forests as a homogeneous high-risk category and has direct implications for the timing and prioritisation of fuel management interventions.

Drought functions as the common driver and threat multiplier linking beetle outbreaks and fire in this system: it weakens tree resin defences — reducing the oleoresin exudation pressure that constitutes a primary physical barrier to beetle gallery establishment — facilitating successful beetle attack and mass mortality [43, 46]; it desiccates fuels, elevating flammability; and it drives the atmospheric conditions associated with extreme fire weather [87]. Cross-scale analyses confirm that climatic water deficit is a primary determinant of both the spatial extent and severity of beetle-induced tree mortality across western forests [45], reinforcing the tight coupling between moisture stress and outbreak dynamics. The result is a tightly coupled feedback system in which the same climatic conditions predispose forests to beetle outbreaks and then, in subsequent years, prime the resulting fuel loads for ignition [3, 58]. Synthesis of drought impacts across U.S. forest systems confirms that western forests are already undergoing large stand-level shifts driven by the interaction of drought, insects, and fire, producing die-offs and compositional changes at scales that challenge existing ecological baselines and generate novel fuel configurations that historical fire behaviour models were not calibrated to characterise [87, 32].

Post-Fire Succession and the Reshaping of Pyrophyte Communities

Fire is not merely an endpoint in the drought–vegetation–climate feedback chain but also a reset mechanism that restructures community composition and, through it, future land–atmosphere coupling. A twelve-year longitudinal study across a burn severity gradient in upland oak–pine forest found that higher burn severities produced greater species diversity and richness, with yellow pines and non-oak-pine pyrophytes maintaining positive relationships with burn severity through year 12 post-fire, while oak recruitment no longer tracked this relationship over the same period [88]. These trajectories imply that high-severity fire progressively shifts stand composition toward fire-adapted species, potentially altering the functional trait distribution — rooting depth, stomatal sensitivity, leaf area — that determines future land–atmosphere feedbacks. Indeed, community-weighted maximum photosynthetic rate and stem hydraulic vulnerability diversity have been shown to mediate the strength of soil moisture–atmosphere feedbacks across temperate and boreal forests, with plant functional traits and climate together explaining 54–67% of cross-site variation in drought-intensification signals [79]. Where beetle-killed stands serve as the fuel base for such high-severity fires, the compound disturbance sequence — drought, beetle outbreak, fire — can produce compositional outcomes qualitatively different from those following either agent alone [26], a cascading dynamic with implications for the alternative stable state transitions examined in the following section.

Gaps and Outstanding Questions

Despite substantial progress, critical integrative gaps remain. Compound extreme projections that incorporate dynamic vegetation feedbacks rarely also include fire as a vegetation state-change mechanism [53], yet fire-driven forest loss plausibly feeds back to regional VPD and precipitation patterns through reductions in transpiration and surface roughness [79, 75]. The hydraulic physiology of fire-damaged trees and how altered hydraulic function modulates subsequent drought vulnerability remains almost entirely unexamined at a mechanistic level, despite the clear relevance of hydraulic architecture to drought survival [40, 12]. The temporal sequencing of beetle-fire interactions — including how the stage-dependent fuel dynamics documented by [72] interact with the accelerating beetle phenology documented by [61] under hotter drought conditions — requires landscape-scale modelling that couples beetle population dynamics with fire behaviour models [3, 58], an integration that remains at the frontier of the field. Post-fire landscapes increasingly host invasive species whose distinct functional traits and rooting strategies may substantially alter local water cycling and energy partitioning [83, 79], yet their role as modulators of land–atmosphere coupling has received minimal quantitative attention. Addressing these gaps requires frameworks that couple tree hydraulic physiology, bark beetle ecology, fire behaviour, and land-surface modelling within a common accounting of the water and energy balance.

7. Forest Ecosystem Transitions, Resilience Thresholds, and Alternative Stable States

Forest ecosystems do not degrade linearly. Under sufficient pressure, they can transition abruptly to alternative stable states — a recognition that has reshaped how ecologists conceptualise disturbance, recovery, and the limits of ecosystem resilience. This section traces the evolution of thinking about forest resilience thresholds, from the foundational theoretical frameworks that predict non-linear transitions, through the landmark synthesis of Amazon tipping points, to empirical evidence of resilience erosion and state change across biomes.

Theoretical Foundations: Ecological Resilience, Bistability, and Critical Transitions

The conceptual basis for understanding abrupt, difficult-to-reverse ecosystem transitions draws on several decades of theoretical development in resilience science and nonlinear dynamics. Central to this framework is the distinction between engineering resilience — the speed at which a system returns to a single equilibrium following perturbation — and ecological resilience — the magnitude of disturbance a system can absorb before reorganising around a fundamentally different set of structures, functions, and feedbacks [89, 90]. Engineering resilience assumes that forests will always return to their prior state given sufficient time, treating perturbations as temporary deviations from a single stable configuration. Ecological resilience, by contrast, acknowledges that multiple stable states — forest, savanna, grassland — may coexist as possible configurations of the same landscape under identical broad climatic conditions, separated by self-reinforcing feedback boundaries [89, 91]. Within this framework, resilience is not a measure of recovery speed but of the size of the basin of attraction — the range of conditions over which a particular ecosystem state can persist — and its erosion under chronic stress is what makes forests vulnerable to abrupt transformation rather than gradual decline.

The mathematical architecture underlying this multi-state view is the fold bifurcation, in which a gradually changing environmental driver progressively shrinks the basin of attraction for the forested state until it disappears entirely at a critical threshold, forcing an abrupt transition to an alternative configuration [92]. Crucially, these transitions exhibit hysteresis: once a system has crossed a tipping point and reorganised around new self-reinforcing feedbacks, simply restoring the environmental driver to pre-transition levels is insufficient to recover the original state [93, 24]. Recovery requires conditions substantially more favourable than those that existed before the transition, a property that renders many ecosystem shifts effectively irreversible on management-relevant timescales. This hysteretic structure is what distinguishes true regime shifts from linear degradation: the system’s history determines its current state, and two landscapes experiencing identical present-day conditions can occupy entirely different stable states depending on whether a tipping point has been crossed [93, 94].

A further critical insight concerns critical slowing down — the progressive decrease in a system’s rate of recovery from small perturbations as it approaches a tipping point [92]. These signatures — detectable as increasing autocorrelation and variance in ecosystem state variables — form the theoretical basis for early-warning signals now being actively sought in satellite-derived vegetation metrics and forest monitoring datasets [77, 95, 96]. Empirical applications of this framework have documented measurable losses of resilience in both tropical and temperate forests under intensifying climate stress [77, 95]. The novel concept of late warning signals — indicators that a tipping point has already been passed but the system remains temporarily in a transient state — introduces both new detection opportunities and new management complexities [92].

Two further theoretical advances are of particular relevance. First, rate-induced critical transitions can occur when the pace of environmental change exceeds a system’s capacity to track its shifting stable state, triggering a critical transition even when the same magnitude of slow change would not have caused one [97]. This mechanism is acutely relevant for long-lived tree species, whose demographic and physiological adjustment rates are inherently slow relative to the velocity of current warming, VPD increase, and fire regime alteration — a mismatch that exposes even forests not yet at a fold bifurcation boundary to transition risk [97, 1]. Second, the interactive effects of press and pulse disturbances on regime shifts depend critically on disturbance duration, not merely intensity: ecosystems may recover from intense but brief perturbations while succumbing to moderate but sustained ones once a critical duration threshold is exceeded [98]. For forests subjected to repeated drought-fire sequences, this duration-dependence framework provides a mechanistic explanation for why sequential moderate disturbances can be more transformative than single extreme events [30].

[24] provided an early effort to bridge resilience theory and forest ecology, demonstrating that forest tipping points manifest differently across spatial and temporal scales and that spatial heterogeneity in environmental drivers likely prevents synchronised continental-scale transitions while still permitting abrupt local and regional shifts — a pattern corroborated by analyses of the Amazon system, where spatially heterogeneous deforestation and drought exposure produce asynchronous rather than uniform resilience loss [28, 99]. The operational quantification of resilience in forest systems remains a formidable challenge, requiring data and modelling integration that most forest monitoring programmes have not yet achieved [100, 101, 96]. Indeed, undesirable ecosystem states — such as degraded shrublands replacing former forest — can themselves exhibit high resilience by resisting restoration efforts [101].

Critical Transitions and the Architecture of Amazon Tipping Points

The Amazon forest system provides the most extensively studied empirical test case for applying critical transitions theory to a specific biome. Reviews of drought physiology identified hydraulic failure and carbon starvation as the two dominant pathways to mortality [56], while long-term forest monitoring documented multi-decadal increases in individual tree mortality rates across the Amazon Basin, with parallel signals emerging in Southeast Asia and remotely sensed canopy loss in the Congo [50]. This foundational work established that tropical forest decline was already underway, but it did not directly address whether such mortality could cross into self-reinforcing ecosystem transformation.

Empirical evidence that such a transformation may already be in progress comes from satellite-based resilience analyses. Boulton et al. [95] analyzed vegetation optical depth data across 1991–2016 and found that more than 76% of the Amazon basin exhibited rising lag-1 autocorrelation — a critical slowing down signal indicating reduced recovery capacity — with decline accelerating sharply after 2003 in drier regions and areas proximate to human land use. Two positive feedback mechanisms were implicated: fire amplification and the suppression of moisture recycling by deforestation, both of which can propel abrupt forest loss beyond critical thresholds once initiated.

The most integrative synthesis to date, [28], fused paleoclimatic records, satellite observations, and field monitoring to identify specific quantitative thresholds beyond which the Amazon forest system faces critical transition. Five water-stress drivers were identified, each with an approximate threshold: global warming at 2–6°C, annual rainfall at approximately 1,000 mm, rainfall seasonality at roughly −450 mm water deficit, dry season length at eight months, and accumulated deforestation at 20% of biome area. Critically, the study estimated that 38% of the Amazon biome has already been degraded when repeated extreme drought events are incorporated into accounting. The recognition that multiple stressors interact synergistically — such that deforestation and warming together push forests closer to transition than either alone — is consistent with both the theoretical expectation that multiple press drivers can narrow the basin of attraction simultaneously [98] and with empirical modelling showing that combined drought and deforestation can trigger up to 6.6 times more local regime shifts from forest to savanna than either factor alone [99]. The wide ranges embedded in these thresholds — most notably the 2–6°C warming window — reflect genuine scientific uncertainty about where tipping points lie, encoding heterogeneity in forest composition, soil water-holding capacity, and local feedback dynamics that current models cannot fully resolve.

Resilience Deficits, Early-Warning Signals, and Fire as a Transition Amplifier

The empirical detection of resilience loss before actual transition represents a critical methodological frontier. As a forest system approaches its tipping point, theory predicts declining recovery rates producing rising autocorrelation and variance in remotely sensed vegetation indices [92]. [28] drew on such signals to argue that resilience has deteriorated significantly since the early 2000s, with forests in drier, more fragmented regions showing the clearest evidence of approaching threshold behaviour — though this satellite-based inference identifies trajectories toward transition rather than demonstrated completions of one. Complementary work by [95] similarly documented pronounced Amazon rainforest resilience loss since the early 2000s, particularly in regions exposed to moisture stress and human pressure, reinforcing the convergence of independent lines of evidence for system-wide destabilisation.

Fire interacts with these trajectories in complex ways. Global fire weather season length increased by 27% and extreme fire weather days rose by 54% between 1979 and 2019, even as global burned area paradoxically declined by 27% over the satellite era due to agricultural expansion fragmenting fire-prone savannas [62]. Within the Amazon context, fire functions as both a symptom of resilience loss and an accelerant that prevents recovery. Drought conditions fundamentally alter forest flammability by drying fuels, thinning canopies, and raising understory temperatures, while fire kills tropical trees primarily through heat damage to the vascular cambium — with mortality rates rising sharply alongside fire intensity and aboveground biomass losses exceeding 50% in severely burned regions [75]. Carbon emissions then operate across two timescales: immediate combustion and slower decomposition of fire-killed trees over subsequent decades, both of which suppress net primary productivity and obstruct canopy recovery [75]. In the language of bistability theory, fire operates as a positive feedback mechanism that deepens the basin of attraction for the non-forest state while further eroding the basin for the forested state, thereby lowering the deforestation threshold identified by [28]. Critically, combined drought and deforestation can trigger up to 6.6 times more local regime shifts from forest to savanna than either stressor alone, underscoring the strongly nonlinear synergy between these forces [99].

Land-atmosphere feedbacks further complicate the picture. Positive drought-intensification feedbacks, whereby low soil moisture amplifies temperature, vapour pressure deficit, and sensible heat flux, are more prevalent at warmer sites, while negative feedbacks dominate cooler environments [79]. This spatial heterogeneity in feedback structure means that as warming progresses, previously buffered forests may cross into a positive-feedback regime, accelerating moisture stress beyond what precipitation trends alone would predict. The transition from negative to positive land-atmosphere feedback thus constitutes its own form of threshold behaviour operating within the broader tipping point architecture — a nested dynamic consistent with the panarchy concept of cross-scale interactions [90].

Compound Drought–Beetle–Fire Cascades and the Compression of Recovery Windows

The theoretical expectation that sequential compound disturbances can be more transformative than single extreme events [98] finds increasingly direct empirical support in the dynamics of drought–beetle–fire cascades now documented across temperate and boreal forests. The chain typically initiates with drought-mediated impairment of host defences — including reductions in resin duct production and oleoresin exudation pressure that underpin primary resistance — enabling bark beetle outbreaks of exceptional scale and synchrony [46, 43, 32]. The resulting mass mortality restructures fuel loads in ways that elevate subsequent fire risk — a compound disturbance sequence in which each link amplifies the next rather than operating independently [87, 72]. Warming-driven voltinism shifts, whereby species such as Dendroctonus ponderosae and Ips typographus transition from univoltine to semivoltine or bivoltine life cycles under elevated temperatures [15, 4], compress the interval between successive beetle generations, while drought-driven synchronisation of beetle swarming concentrates attack pressure across spatial scales of hundreds of kilometres [61, 3]. This temporal compression directly narrows the recovery windows available between disturbance events — the very parameter that the press-pulse framework identifies as critical for determining whether ecosystems recover or undergo regime shift [98].

The carbon-cycle consequences of these cascading disturbances provide a further positive feedback to the climate forcing that drives them. The mountain pine beetle outbreak in British Columbia demonstrated that insect-driven mortality can transform large forested landscapes from net carbon sinks into net carbon sources, releasing cumulative emissions rivalling those of national transportation sectors [102, 15]. More broadly, drought-induced tree mortality has been shown to produce measurable, landscape-scale decreases in carbon sink capacity across forest biomes [5], underscoring that the British Columbia case is not anomalous but representative of a wider pattern. Spatially explicit modelling in the Lake Tahoe Basin confirmed that warmer, drier trajectories produce the highest cumulative ecosystem service losses, with carbon sequestration accounting for approximately 90.8% of aggregated annual economic losses; enhanced management through thinning and prescribed fire reduced average annual biomass losses by approximately 184,000 to 320,000 tonnes relative to business-as-usual across all climate projections [80].

The systematic underestimation of compound dynamics in global vegetation models has been identified as a structural deficiency with profound implications for projections of forest futures. [2] demonstrated that current-generation models lack the mechanistic fidelity to represent hydraulic failure, carbon starvation, and insect-mediated mortality as distinct and interacting processes [9, 8, 13], instead relying on coarse stress indices that consequently underestimate both the probability and pace of large-scale forest die-off. Process-based landscape models that explicitly incorporate drought as a beetle outbreak trigger have achieved markedly better correspondence with observed mortality — explaining up to 47.5% of cumulative disturbance impact compared with only 6.6% for wind-only models [58] — yet these approaches remain computationally intensive and region-specific. The gap between the demonstrated importance of compound drought–beetle–fire cascades and their representation in the models used to project forest futures constitutes one of the most consequential disconnects in contemporary forest science.

Demographic Mechanisms of Non-Recovery: Regeneration Failure, Seed Source Loss, and Belowground Legacies

A critical complement to tipping point theory concerns the demographic and belowground mechanisms that determine whether a forest actually recovers or remains locked in an alternative state once canopy mortality has occurred. The transition from mortality event to permanent state change is mediated by a series of recruitment bottlenecks — seed availability, seedling establishment, and belowground biological support — each of which is now demonstrably under threat.

Seed source limitation as a spatial bottleneck

A synthesis of 49 publications spanning hundreds of wildfires across the western United States established that distance to living seed sources is among the most robust predictors of post-fire conifer recruitment, with regeneration declining measurably at distances of 40 to 400 metres from surviving trees depending on the species [17]. As fire severity increases and burn perimeters expand [71, 62], the interior of high-severity patches increasingly lies beyond the effective dispersal shadow of surviving trees — a problem that is especially acute for obligate-seeding species such as ponderosa pine, where large high-severity patches can outstrip seed dispersal distances entirely and leave interior regeneration zones beyond the reach of any viable seed source [29]. The result is extensive recruitment voids that passive recovery cannot fill. Warmer and drier post-fire conditions independently suppress seedling establishment, particularly at lower elevations [17], and critically, high fire severity amplifies these climatic effects by reducing seed availability while simultaneously degrading the microsite conditions needed for germination and early survival [73]. Modelling across 10,230 field plots and 334 wildfires found that in 40–42% of the western United States study area, conifer regeneration was projected as likely under low-severity but not high-severity fire through 2031–2050, and that the proportion of the landscape considered unlikely to support any conifer regeneration regardless of fire severity grew from 5% in 1981–2000 to 26–31% by mid-century [73]. A parallel synthesis focused on boreal systems found that black spruce (Picea mariana) faces the broadest regeneration vulnerability under climate change, with both germination success and seedling survival declining under projected temperature and drought regimes [27]. Together, these findings establish that seed source distance sets the ceiling on potential regeneration while climate moisture stress determines how much of that potential is realised — a two-stage bottleneck that tightens as fires grow larger and post-fire conditions grow hotter and drier.

Short-interval reburning and the elimination of regenerating cohorts

If single high-severity fires create recruitment bottlenecks through seed source removal and climate stress [17, 27], reburning within ecologically short intervals can render those bottlenecks effectively permanent. A paired-site study of 50 locations in the northwestern Canadian boreal zone demonstrated that short-interval reburn sites (fire return intervals of 4 to 17 years) exhibited conifer seedling densities only 51.9% of those at long-interval sites, and 72% of short-interval sites had fewer than half their post-fire stems composed of conifers, compared with just 12% of long-interval sites [18]. Because reburning consumes not only residual seed trees but also the juvenile cohorts that established after the first fire, successive disturbances effectively reset the successional clock before any meaningful seed bank can accumulate [29, 103, 17]. This depletion of both aerial and soil seed sources is compounded by the shortened intervals between fires, which are insufficient for regenerating conifers to reach reproductive maturity and replenish local seed pools [27, 73]. Drought conditions further compounded this effect, suggesting that the interaction between shortened fire return intervals and climatic aridity creates a compound disturbance regime [83, 23] that pushes boreal stands across a compositional threshold into persistent broadleaf-dominated states [18, 29, 22].

Landscape-scale disequilibrium and the fire-suppression legacy

Stand-level regeneration failures accumulate into regional compositional change when projected across entire landscapes. Using the process-based LANDIS-II model across a 2.94-million-hectare study area in the Klamath region, [22] showed that even under historical baseline climate conditions, the legacy of twentieth-century fire suppression has already placed these forests in structural disequilibrium, with approximately 580,000 hectares (31% of current conifer extent) projected to transition toward shrubland dominance. This disequilibrium reflects the broader pattern across western North America, where altered disturbance regimes — shaped by decades of fire exclusion — have decoupled stand structure from the climate and fire conditions under which these forests developed [23]. Under climate change scenarios, summer drought suppresses forest growth and reduces post-fire conifer establishment probabilities by 26–39%, compared with only 7–20% reductions for shrubland and hardwood species — a competitive asymmetry that progressively favours non-forest vegetation after fire [22]. Such wildfire-driven forest conversion to shrubland and non-forest states is increasingly documented across western North American landscapes as a discrete and potentially irreversible outcome [29]. Earlier empirical work in the Klamath-Siskiyou region had documented substantial spatial variability in post-fire conifer recruitment [104], foreshadowing what simulation and review work subsequently confirmed: low-elevation, dry-forest sites and recently reburned areas represent the leading edge of regeneration failure [17], while higher-elevation refugia may temporarily buffer recruitment under contemporary conditions — a dynamic corroborated by modelling across 334 western US wildfires showing that the northern Rockies and higher-elevation forests retain the greatest near-term regeneration probability [73]. Critically, high fire severity and warm, dry post-fire climate interact synergistically to compound recruitment failure, and the proportion of the western US unlikely to support conifer regeneration regardless of fire severity grew from 5% in 1981–2000 to 26–31% by mid-century [73].

Resprouting limits, fire-induced deforestation, and belowground legacies

Beyond seed-dependent recruitment, vegetative resprouting also operates within strict physiological boundaries that are increasingly being exceeded. Increasing fire frequency progressively damages the cambium of surviving trees, with repeated thermal insults ultimately causing structural collapse that disproportionately affects large, old trees [103]. These limits to resprouting have been formalised within a hierarchical conceptual model of fire-induced deforestation, which identifies three conditions that must co-occur for permanent forest loss: individual tree mortality, population-level regeneration failure, and the absence of effective resprouting understory species [105]. A critical mechanism within this hierarchy is post-fire xylem cavitation, which can kill fire-surviving trees weeks to months after the burn event [105, 9], bridging the fire and drought literatures as trees that survive initial fire injury succumb to hydraulic dysfunction under high-VPD conditions [106]. High-severity fires that eliminate surviving seed sources across large patches further compound this trajectory, as passive regeneration from the forest interior becomes geometrically less probable with increasing patch size [29, 17, 73] — a constraint that is tightening as the areal extent of high-severity burn patches expands under warming-driven fire regimes [23].

Perhaps the most consequential mechanism locking disturbed forests into alternative states operates belowground. Fire eliminates or severely degrades the ectomycorrhizal (ECM) hyphal networks and spore banks that facilitate colonisation of establishing seedlings [105, 107]. Into the mycorrhizal vacuum created by severe fire, post-fire herbaceous pioneers establish arbuscular mycorrhizal (AM) networks that are functionally incompatible with the ECM dependency of tree seedlings, creating a priority effect in which the identity of the first colonising mycorrhizal guild determines the vegetation trajectory for decades [105, 108, 107]. Experimental evidence on soil microbial diversity-function relationships deepens this concern: microbial diversity loss significantly impaired nitrogen mineralisation and nutrient cycling, and the functional consequences of that loss were amplified under drought conditions [109] — precisely the conditions that characterise post-fire environments in warming climates. Soil bacterial and fungal communities moreover differ markedly in their sensitivity to such post-disturbance stress, with fungal communities exhibiting slower recovery trajectories than bacterial ones [110], further prolonging the window during which ECM networks remain absent or dysfunctional. Restoration interventions focused narrowly on reseeding or replanting may therefore be insufficient if the belowground community has transitioned to an AM-dominated or microbially impoverished state, since the identity of plant functional groups used in restoration plantings may matter for their capacity to rebuild ECM networks and sustain microbially driven nutrient cycling [109, 105].

Taken together, these findings identify the specific feedbacks that lock in non-forest states following compound disturbance: seed source elimination across expanded high-severity patches, climatic unsuitability during seedling establishment windows, short-interval reburning that prevents cohort maturation [18], and mycorrhizal guild displacement that forecloses ECM-dependent tree recruitment [17, 18, 22, 105]. Their co-occurrence under compound disturbance regimes creates interlocking constraints that make passive forest regeneration increasingly improbable across large portions of fire-prone landscapes [29, 103, 23].

Planted Forests, Biological Constraints, and the Limits of Management Optimism

[56] argued compellingly that trees face slow migration rates relative to the pace of climate change [26], and low genetic diversity in key hydraulic traits severely limits phenotypic adjustment — a constraint that the rate-induced transition framework formalises by demonstrating that collapse can occur purely because the pace of change exceeds the response rate [97]. Such hydraulic constraints are compounded by the fact that drought stress operates through mutually reinforcing mechanisms including carbon starvation and hydraulic failure [10], narrowing the physiological envelope within which planted monocultures can respond. Comparing satellite-derived drought resistance and resilience metrics across China from 2001 to 2020, [76] found that planted forests exhibited significantly lower drought resilience than natural forests, particularly in subtropical broad-leaved evergreen and warm temperate deciduous systems. Paradoxically, planted forest drought resistance increased while resilience decreased between 2011–2020 compared to 2001–2010, suggesting a shift from recovery-oriented to resistance-oriented coping strategies that may be unsustainable under continued drought intensification — a pattern consistent with broader evidence that positive effects of tree species diversity on productivity can switch to negative following severe drought [111]. Evidence from European systems reinforces this concern: dense commercial monocultures, particularly Norway spruce plantations, are structurally predisposed to severe bark beetle outbreaks because intraspecific competition compromises individual tree resin defence capacity precisely when climatic stress is highest [4, 46]. Water-limiting conditions have been shown experimentally to directly predispose Norway spruce to Ips typographus attack by reducing resin flow and monoterpene concentrations [46], while drought acting as an inciting factor can trigger large-scale forest dieback through compound bark beetle–drought interactions [58]. Silvicultural transitions toward mixed-species stands and reduced stand densities have been identified as structural interventions that reduce both bottom-up and top-down drivers of outbreak intensity [4], and modelling work confirms that such interventions consistently reduce biomass losses across a range of plausible climate futures [80]. However, the distinction between specified resilience to particular drought events and the broader general resilience that enables systems to absorb unforeseen disturbances [91] — and the risk that management optimises the former at the expense of the latter — warrants careful consideration in evaluating these findings.

Non-Recovery Trajectories: From Mortality to Transformation

The transition from reversible mortality to irreversible state change requires integrating tree-level physiology with landscape-scale dynamics. The 2002 Colorado drought illustrates the vulnerability of even temperate forest systems: it produced unprecedented combinations of growing-season temperature and evaporative deficit, depleting shallow soil moisture and triggering widespread aspen mortality [74]. Aspen’s limited plasticity in water use and dependence on shallow soil moisture reserves meant that an extreme but temporally bounded drought event produced landscape-scale die-off with slow or absent recovery trajectories in the most affected stands — a pattern also documented in aspen populations along the southern edge of the Canadian boreal forest following severe drought [112]. Broader assessments confirm that such non-recovery outcomes are becoming more frequent as climate-driven mortality events outpace forest regeneration capacity [1].

[50] documented disproportionate mortality among the largest tropical trees during drought, a size-class effect with significant implications for carbon storage and forest structure. Because large trees account for a disproportionate share of forest biomass and contribute to microclimatic buffering of the understory [32], their selective loss may lower the threshold at which a remaining forest community becomes susceptible to savannification — a transition increasingly documented as forests lose resilience under repeated drought stress [95, 28]. Declining forest resilience has now been detected across biomes, with remote-sensing analyses revealing that recovery rates following disturbance have slowed significantly since the early 2000s [77]. Post-mortality vegetation replacement surveys further confirm that drought-killed forests do not reliably recruit back to their prior composition, with woodland and shrubland states substituting for closed-canopy forest in the most severely impacted sites [26]. The size-dependent mortality pattern also intersects with bark beetle ecology: larger trees suffer disproportionately higher mortality during beetle outbreaks in drought-stressed systems [45], and drought-primed physiological stress is now recognized as a principal inciting mechanism for beetle population irruptions [43, 58]. The cross-scale dynamics linking drought, host-tree physiology, and beetle population eruptions further amplify non-recovery risk at the landscape level [3], suggesting that the selective removal of large individuals through compound drought-beetle disturbance may erode structural resilience more rapidly than either stressor acting alone [32].

Outstanding Gaps

Despite this substantial progress, several integrative gaps remain consequential. Tropical forest resilience thresholds outside the Amazon — in the Congo Basin, the Mekong region, and insular Southeast Asia — are poorly constrained, and it is not clear whether the rainfall and deforestation thresholds identified by [28] transfer to structurally different biomes; satellite-derived resilience metrics suggest that declining forest stability is a multi-biome phenomenon [77, 96], yet the critical transition thresholds themselves remain Amazon-centric. The role of compound drought-beetle-fire cascades in driving state transitions has been increasingly documented in temperate and boreal systems [3, 32, 24] but remains poorly integrated into the tipping point frameworks developed for tropical forests; whether the self-reinforcing feedback dynamics observed in beetle-fire-drought sequences [87, 58] share the mathematical structure of the fold bifurcations that characterise savannification is an open theoretical question with significant implications for early-warning signal detection [92, 113]. The belowground dimension of non-recovery — particularly the role of mycorrhizal guild displacement as a self-reinforcing feedback maintaining alternative states [105] — warrants formal integration into tipping point models that currently represent only aboveground vegetation-climate feedbacks. The demographic mechanisms now linking tree-level mortality to landscape-scale transitions — seed source elimination, establishment failure, short-interval reburning, and mycorrhizal disruption [17, 18, 105] — await integration into a unified framework that bridges critical transitions theory with the empirical ecology of regeneration failure.

8. Discussion

The cumulative picture that emerges from this review is one of converging vulnerabilities. No single mechanism explains forest mortality under climate change, and no single process determines whether a forest recovers or transitions to an alternative state. What has become clear over the past two to three years is that the field has moved decisively beyond single-driver frameworks — away from treating hydraulic failure and carbon starvation as competing hypotheses, and away from treating drought, heat, fire, biotic agents, and atmospheric moisture demand as separable stressors. The more consequential advance is recognizing that these mechanisms and drivers interact across scales in ways that compress the time available for forests to recover between disturbances.

A field reshaped by multimechanism and multidriver thinking

Perhaps the most substantive shift in recent understanding concerns vapour pressure deficit. VPD has been repositioned from a secondary modifier of drought stress to a primary driver of mortality risk in its own right, operating partly independently of soil moisture availability [13, 12, 11]. This reframing matters for models and for management: forests in regions where soils retain adequate moisture may nonetheless face escalating atmospheric demand that drives stomatal closure, reduces carbon assimilation, and depletes nonstructural carbohydrate reserves over successive seasons [19, 9]. That this effect remains poorly represented in process-based mortality models — which continue to treat soil moisture as the dominant drought variable — represents one of the most consequential mismatches between empirical understanding and predictive tools [13]. Correcting it is not merely a technical refinement; it changes which forests are identified as vulnerable and over what timescales.

The interaction between hydraulic failure and carbon starvation has similarly been reframed. Rather than competing explanations, these processes are now better understood as co-occurring and mutually reinforcing pathways whose relative contributions depend on drought duration, intensity, species’ hydraulic safety margins, and prior stress history [8, 9, 7]. Evidence from multi-species syntheses confirms that both mechanisms frequently operate in parallel, with the balance shifting depending on the pace and severity of water deficit [8, 49]. The largely unresolved role of phloem transport failure sits at the intersection of both pathways — capable of disrupting carbon delivery to roots and meristems before xylem cavitation reaches lethal thresholds [7, 47] — and represents a mechanistic gap with direct implications for understanding mortality timing and for species-level differences in drought sensitivity.

Biotic agents as compound disturbance amplifiers, not secondary consequences

A central finding of this review is that biotic disturbance agents — particularly bark beetles — must be recognised not as secondary consequences of drought stress but as active participants in compound disturbance cascades whose synergistic interactions with abiotic stressors produce mortality outcomes that neither driver can explain alone. The evidence that drought impairs resin-based chemical defenses below the thresholds required to repel mass beetle attack [44, 32] establishes a direct mechanistic link between the physiological stress pathways described in Section 3 and the landscape-scale die-off events documented in Section 5. In many mortality events, the tree crosses a defense threshold before it crosses a hydraulic or carbon lethal threshold — the proximate cause of death is biotic, but the ultimate cause is the abiotic stress that eroded resistance.

The implications of this compound framing extend to how vulnerability is assessed and projected. Global vegetation models that lack explicit representation of insect-mediated mortality systematically underestimate forest die-off rates [2], and the magnitude of this underestimation has been quantified: incorporating drought as a beetle outbreak trigger in process-based landscape models increased explained mortality variance from 6.6% to 47.5% in Central European spruce forests [58]. This is not a marginal correction but a fundamental reframing of projected forest futures. The concurrent acceleration of beetle voltinism — warmer winters reducing cold-induced larval mortality and enabling additional generations per year in species such as Ips typographus and the mountain pine beetle [15, 4] — range expansion into forests lacking co-evolutionary defenses [69, 44, 15, 68], and landscape-scale synchronisation of swarming during drought years [61] create a compound rate-of-change problem that current single-stressor frameworks are structurally unable to capture. Singh et al. [4] further demonstrate that the combined influence of shifting temperature regimes and drought frequency is restructuring outbreak dynamics in ways that decouple beetle population dynamics from historical climatological baselines, rendering retrospective risk models unreliable for future projections. Cross-scale analyses further demonstrate that bark beetle outbreaks are governed by bottom-up host stress, landscape-level fuel connectivity, and top-down climate forcing simultaneously — interactions that render outbreak dynamics fundamentally non-linear [3].

The carbon-climate feedback dimension of bark beetle outbreaks adds a further layer of systemic concern. The demonstration that the British Columbia mountain pine beetle outbreak converted forests from carbon sinks to carbon sources at a scale comparable to national transportation emissions [102] established that beetle-driven mortality is not merely an ecological disturbance but a positive feedback mechanism within the global carbon cycle [80]. This feedback operates in the same direction as the fire-climate feedbacks already well documented in the literature, yet the two are rarely modelled within a common framework despite operating on the same landscapes and being driven by the same drought dynamics. The interaction is not merely additive: beetle-killed stands alter fuel structure across successive post-outbreak stages — elevating fine surface fuels in the red stage, reducing canopy moisture, and accumulating large-diameter surface fuels in grey and old-MPB stages — each presenting a distinct and elevated fire hazard profile through different mechanisms [72, 80], a compounding sequence that shared beetle-fire models have yet to fully operationalise.

Regeneration failure as the mechanism that makes compound disturbances irreversible

Perhaps the most significant conceptual advance integrated in this review is the recognition that compound drought-beetle-fire cascades produce irreversible outcomes not merely because they kill trees at unprecedented rates, but because they systematically eliminate the demographic and biological infrastructure on which forest recovery depends. The evidence reviewed in Section 7 establishes that post-disturbance regeneration is failing across broad swathes of western and boreal North America due to the simultaneous operation of multiple recruitment bottlenecks: seed source elimination across expanding high-severity burn perimeters, where regeneration declines sharply beyond 40–400 metres from surviving trees depending on species [17]; climatic unsuitability during critical seedling establishment windows, with post-fire drought independently suppressing recruitment even where seeds arrive [17, 27]; and short-interval reburning that destroys regenerating cohorts before they reach reproductive maturity, reducing conifer seedling densities by nearly half and shifting stand composition toward persistent broadleaf dominance [18]. Empirical documentation of wildfire-driven forest conversion across western North American landscapes confirms that these bottlenecks are already producing measurable, landscape-scale transitions away from conifer dominance [29], with post-fire vegetation recovery increasingly constrained by climatic conditions that exceed the physiological tolerances of regenerating seedlings [103].

These demographic bottlenecks operate in concert with belowground processes that have received insufficient attention in earlier disturbance frameworks. The disruption of ectomycorrhizal networks by severe fire creates a belowground priority effect in which post-fire herbaceous colonisers establish arbuscular mycorrhizal communities functionally incompatible with tree seedling establishment, effectively locking disturbed sites into non-forest states through a self-reinforcing biotic feedback [105]. Specialist plant–microbial interactions of this kind are increasingly recognised as key drivers of plant-community trajectories following disturbance [108], reinforcing the view that the mycorrhizal guild shift is not merely an incidental consequence of fire but an active mechanism of state maintenance. That the functional consequences of soil microbial diversity loss are amplified under drought conditions [109] — precisely the conditions characterising post-fire environments under warming — creates a cascading belowground vulnerability that compounds the aboveground constraints on recruitment. Landscape-scale modelling confirms the cumulative consequence: even under historical climate baselines, fire-suppression legacies have placed some forest systems in structural disequilibrium, with climate change widening the competitive asymmetry between drought-stressed conifer seedlings and the shrub and hardwood species that increasingly dominate post-fire environments [22, 23].

This regeneration failure evidence fundamentally strengthens the empirical basis for the irreversibility that tipping point theory predicts. The fold bifurcation framework explains why transitions should exhibit hysteresis [24, 21], but the specific feedbacks maintaining non-forest states — seed source absence, mycorrhizal guild displacement, climatic establishment failure, competitive exclusion by shrubs — are now empirically identified rather than merely hypothesised. The global finding that self-replacement of dominant tree species occurs at only 21% of drought-mortality sites, with shrub replacement at 69% [26], provides the broadest quantitative confirmation that forest die-off under contemporary climate conditions leads predominantly to compositional reorganisation rather than recovery.

Compounding disturbances and the erosion of recovery windows

The interaction between fire regimes and drought now appears central to understanding ecosystem-level transitions in ways that were underappreciated even recently. Fire does not simply follow drought as a lagged consequence; it interacts with vegetation structure altered by prior drought mortality — including the stage-dependent fuel loads created by bark beetle outbreaks [72] — to reshape fuel loads, microclimate, and post-fire regeneration conditions. When fire return intervals shorten under warming and drying [18, 29] — and when post-fire recovery occurs under a higher-VPD atmosphere — the probability that recovering vegetation re-establishes the same structural type diminishes substantially [17, 73]. This is the mechanism through which many modeled and observed ecosystem transitions are now understood to operate: not as abrupt threshold crossings but as sequential disturbances that progressively erode resilience before a transition becomes visible at landscape scales [22]. The press-pulse interaction framework formalises this insight: chronic warming constitutes the press that narrows the basin of attraction, while successive fire, drought, and beetle events act as pulses whose cumulative effect depends on the duration and frequency of the press, not merely their individual intensity [98]. The drought-beetle-fire cascade is among the most potent expressions of this compounding dynamic, because each element amplifies the next: drought weakens defenses, beetle mortality restructures fuels, and the resulting fires prevent the regeneration that would otherwise restore the system to its pre-disturbance state [87]. Short-interval reburning is particularly damaging in this regard, as post-fire seedling establishment is acutely sensitive to atmospheric moisture demand, and rising VPD under continued warming is projected to push an increasing fraction of post-fire environments beyond the regeneration niche of incumbent species [73, 13].

The evidence from regional case studies reinforces this picture. Die-off events that once appeared anomalous are increasingly interpretable as early expressions of climatically driven transitions, particularly where background mortality rates had already been elevated by chronic VPD stress [1, 12] or repeated moderate droughts. The European 2018–2022 sequence provides a particularly instructive example: drought-driven bark beetle outbreaks of unprecedented scale cascaded across Central Europe, yet no corresponding increase in wildfire was detected [60] — raising the question of whether the fire component of the compound cascade is suppressed, delayed, or will eventually manifest as beetle-killed fuel loads enter their most flammable successional stages. The consistent underrepresentation of African and South and Southeast Asian forests in global mortality databases [14, 20] is therefore not a minor geographic gap — it likely means that the global extent of ongoing transitions is being underestimated, and that the climatic fingerprints identified from better-studied regions may not transfer reliably to systems with distinct species compositions and disturbance histories [50].

Biodiversity, genetic erosion, and the hidden costs of compositional change

The consequences of compound disturbance extend beyond the binary question of whether forests persist to encompass the composition, diversity, and adaptive capacity of the forests that remain or re-establish. Evidence from a long-term temperate forest biodiversity experiment revealed that the positive relationship between species diversity and productivity — documented over more than a decade — reversed following the catastrophic 2018 European drought, with diverse mixtures suffering greater productivity losses than monocultures because drought-sensitive species disproportionately represented in mixed plots dragged plot-level performance below that of drought-tolerant monocultures [111]. This finding fundamentally qualifies the insurance hypothesis: diversity provides resilience insurance only insofar as it incorporates functionally distinct, stress-tolerant species [114]; a diverse assemblage of drought-sensitive taxa may amplify rather than dampen the impacts of extreme events. Functional trait screening — particularly traits associated with hydraulic safety margins and stomatal regulation — is increasingly recognised as essential for distinguishing assemblages that confer genuine insurance value from those that are merely species-rich [79]. The policy-relevant implication is that the operative question shifts from “how much diversity?” to “diversity of what?” — demanding closer integration of functional trait-based approaches into biodiversity-ecosystem functioning frameworks.

At landscape scales, analysis of over two decades of Forest Inventory and Analysis data across 22 eastern U.S. states reveals that disturbance consistently facilitates the establishment of more thermophilic regeneration — a process of compositional thermophilisation that could be interpreted as adaptive tracking of warming climates [115]. Yet this apparent resilience is purchased at a cost: adult tree species richness declined measurably in disturbed forests, and sapling recruitment did not reliably compensate for overstory mortality, raising concerns about future biomass potential and taxonomic diversity [115]. The net effect is a forest estate that may be better matched to future climates in terms of species thermal niches, yet simultaneously depleted in structural complexity, diversity, and carbon storage capacity. This trajectory mirrors patterns documented more broadly across the Anthropocene, wherein rapid compositional turnover driven by simultaneous stressors consistently favours generalist and disturbance-adapted taxa at the expense of specialist species with narrower functional roles [116].

These compositional dynamics intersect with a less visible but equally consequential erosion of genetic diversity. Forest fragmentation isolates populations into patches too small to maintain the allelic richness required for adaptive response, driving inbreeding and generating an extinction debt in which genetic impoverishment manifests not in the parent generation — which may persist for centuries — but in the younger cohorts that follow [117, 118]. This temporal displacement means that forests currently appearing demographically stable may already have crossed genetic thresholds from which recovery without intervention is unlikely. Climate change further compounds this erosion by disrupting the mast seeding dynamics on which foundation species depend for effective reproduction, reducing the genetic diversity entering regeneration pathways precisely when adaptive capacity is most needed [117]. Read alongside the regeneration failure evidence, the implication is sobering: the forests that do manage to re-establish after compound disturbance may be taxonomically simplified, genetically impoverished, and functionally reorganised in ways that reduce both their carbon storage capacity and their ability to withstand the next disturbance cycle.

Implications for theory and practice

For ecological theory, these findings challenge resilience frameworks that implicitly assume a single stable reference state toward which forests return following disturbance — an assumption aligned with engineering resilience rather than the ecological resilience concept that recognises multiple possible stable configurations of the same landscape [89, 24, 113]. Empirical evidence for alternative stable states in mountain and boreal forests — where repeated disturbance under shifting climate pushes systems across thresholds into non-forested configurations — reinforces the view that basin-of-attraction geometry is itself dynamic [21]. Where background climate has shifted sufficiently, there may be no stable forest state to return to, and resilience metrics derived from historical baselines will overestimate recovery probability [100]. Management interventions aimed at accelerating post-disturbance recovery — including assisted regeneration and species compositional adjustment — need to account for the possibility that the target state itself is moving, or that the basin of attraction for a forested state has contracted to the point where recovery requires conditions more favourable than any achievable through management alone. Silvicultural interventions that reduce stand density and increase structural and compositional diversity show promise for moderating compound disturbance severity — including bark beetle outbreak intensity — and their benefits appear robust across a range of climate projections [80, 4], but the decadal timescales required for such structural transformations may not be available where compound cascades are already underway.

For practice, the most immediate implication is that drought-monitoring systems and early-warning frameworks calibrated primarily to soil moisture are likely to underdetect mortality risk in regions where VPD trends are driving stress. Rising atmospheric moisture demand has been identified as a critical amplifying mechanism in tree mortality that operates independently of, and interactively with, soil water limitation [11, 13], and its increasing dominance in the global tree die-off record [14] means that soil-moisture-only monitoring frameworks carry a systematic detection gap. Incorporating atmospheric demand alongside soil moisture in operational forest health monitoring is achievable with existing data streams and warrants prioritization [12]. Equally, early-warning systems for bark beetle outbreaks should be integrated with drought monitoring infrastructure rather than operating as separate detection networks; the empirical demonstration that VPD during critical swarming periods serves as a threshold trigger for outbreak initiation [58, 61] provides a direct operational pathway for such integration. The regeneration failure evidence adds a further practical imperative: passive post-disturbance recovery strategies are no longer sufficient across broad areas of fire-prone western and boreal North America [29, 17, 27], and must be supplemented by assisted regeneration using climate-informed seed sourcing — selecting reproductive material matched to projected rather than historical climatic conditions — and by soil restoration approaches that address belowground biological legacies, including mycorrhizal inoculation where ECM networks have been eliminated [105, 17, 117]. Reduced fire severity has been identified as a near-term buffer that can preserve regeneration potential in some conifer systems [73], pointing to fire management as a lever for extending the window of intervention opportunity. The scale of area requiring such intervention under projected fire and drought regimes, however, far exceeds current restoration capacity, underscoring the need for triage frameworks that prioritise intervention in areas most likely to recover and most valuable ecologically.

Limitations and future directions

This review is constrained by the same geographic and taxonomic imbalances it identifies: the synthesis reflects a literature that overrepresents temperate conifers and Mediterranean-type systems and underrepresents tropical broadleaf forests in controlled experimental settings [50, 19]. Findings about hydraulic safety margins, carbon starvation timelines [9, 7], defense chemistry thresholds, and recovery capacity [8] may not generalize across this gap. The bark beetle literature is similarly weighted toward several well-studied beetle-host systems in North America and Europe; the role of biotic disturbance agents in tropical, subtropical, and Southern Hemisphere forest mortality remains comparatively understudied [32]. The regeneration failure literature is concentrated in western and boreal North America [17, 29], and whether the seed source distance thresholds, climate establishment windows [27], and mycorrhizal disruption dynamics documented in these systems transfer to tropical, Southern Hemisphere, or Asian forest contexts remains largely untested.

The most productive frontier for future research lies at the intersection of belowground hydraulics, atmospheric demand, biotic disturbance dynamics, regeneration ecology, and sequential disturbance. Root cavitation, mycorrhizal mediation of water transport [119], and the legacy effects of prior droughts on hydraulic and carbohydrate reserves [49] remain poorly constrained experimentally, particularly across the repeated-drought scenarios that will characterize coming decades [13]. The mechanistic quantification of water-deficit thresholds for irreversible defense impairment across ecologically important tree species — the point at which resin production can no longer sustain resistance to mass beetle attack [43, 46] — represents a critical gap that bridges physiology and disturbance ecology. The role of fungal pathogens and root diseases, which interact with both drought stress and beetle ecology but remain comparatively understudied as compound mortality drivers [70, 33], warrants substantially greater integration into monitoring and modelling frameworks. The long-term trajectories of stands currently experiencing regeneration failure — whether they represent permanent state transitions [115] or delayed recovery — cannot yet be resolved from available data and demand multi-decadal monitoring designs. Equally pressing is the need to integrate the well-developed mathematical theory of critical transitions — including fold bifurcations, hysteresis, and rate-induced tipping [92, 113, 24] — with the species-level physiological, entomological, demographic, and belowground data that determine where thresholds actually lie in specific forest systems [94]. Formal validation of predictive vulnerability frameworks against observed post-disturbance outcomes remains a high-priority need. Closing these gaps will require coordinated multi-site experiments that span climatic gradients and explicitly manipulate disturbance sequence, not just single-event intensity — a research design that the trajectory of current findings makes urgent.

9. Conclusions

This systematic review of 71 papers, organized across five integrative themes, synthesizes current understanding of forest vulnerability under accelerating climate change and provides a foundation for projecting ecosystem trajectories in an era of compounding disturbances.

Physiological Mechanisms of Tree Mortality

The evidence is unambiguous: tree death under drought and heat stress proceeds through two primary, often concurrent pathways — hydraulic failure and carbon starvation [13, 7, 9]. Hydraulic failure occurs when accumulated embolisms irreversibly block xylem water transport, while carbon starvation reflects the depletion of non-structural carbohydrate reserves below the threshold needed to sustain respiration, osmotic regulation, and cellular maintenance [13, 10]. Prolonged vapor pressure deficit and soil moisture deficit push xylem conduits beyond their embolism thresholds while simultaneously exhausting these carbohydrate reserves needed for defense and respiration [8, 13]. Critically, these physiological stress pathways also erode the resin-based chemical defenses that constitute trees’ primary resistance to bark beetle mass attack [43, 46], meaning that drought compromises survival both directly and by lowering the threshold for successful biotic colonization. Biotic agents, particularly bark beetles and fungal pathogens, exploit these physiologically weakened states, accelerating mortality at landscape scales through synergistic interactions that produce die-off rates exceeding those attributable to either abiotic or biotic stressors alone [32, 3]. Management strategies must therefore address water dynamics, carbon dynamics, and biotic disturbance risk as interconnected rather than independent concerns.

Bark Beetle Outbreaks as Climate-Driven Compound Disturbance

Climate warming has fundamentally restructured bark beetle population dynamics at continental scales — shifting voltinism, reducing overwinter mortality, expanding geographic and elevational range, and enabling landscape-scale synchronization of swarming during drought years [3, 15]. These changes are not incremental adjustments to historical disturbance regimes but qualitative transformations that create novel host-pest interactions in forests lacking co-evolutionary defense capacity; bark beetles already affect larger forest areas than fire in western North America, with over 47 million hectares of coniferous forest impacted in a single decade [3]. The conversion of beetle-killed forests from carbon sinks to carbon sources constitutes a positive climate feedback that amplifies the warming driving further outbreaks [80, 32]. Single-stressor models that omit drought-beetle coupling may underestimate large-scale forest transformation by an order of magnitude — a finding confirmed by landscape simulations showing that a wind-only outbreak model reproduced just 6.6% of observed spruce mortality compared to 44–46% observed, while a drought-coupled model achieved an R² of 0.86 and projected loss of ~92% of initial spruce growing stock by 2100 [58].

Fire Regimes, Forest Structure, and Ecosystem Transitions

Fire no longer operates as a periodic reset within stable forest systems. Instead, increasing fire frequency, severity, and spatial extent — driven by fuel accumulation, drought, and extreme heat — are reshaping the structural and compositional baseline of forests [29, 86]. The fuel dynamics created by bark beetle mortality are temporally structured and stage-dependent: crown fire risk is elevated in recently killed (Red-stage) stands through reduced foliar moisture — with effective canopy moisture approximately one-third lower than in unattacked stands — while surface fire intensity peaks in later decay (Old-MPB) stages where 1,000-hour surface fuel loads can drive fireline intensities up to 28 times higher than standard operational models predict [72]. Active crown fire transition occurs at wind speeds 25 km/h lower in beetle-affected stands than in green stands under equivalent drought conditions [72]. The interaction between climate, beetles, and fire is not additive but multiplicative [3], meaning that even modest additional warming can push fire-affected landscapes across critical thresholds toward non-forest states [22, 29] through compound disturbance cascades in which each element amplifies the next.

Hydrological Intensification and Atmospheric Moisture Demand

An intensifying hydrological cycle does not uniformly benefit forests. While some regions receive increased total precipitation, forests face greater stress from the combination of longer dry intervals, more intense precipitation events that reduce soil infiltration, and rising atmospheric evaporative demand [1, 59]. Vapor pressure deficit (VPD) emerges as a particularly consequential driver: elevated VPD triggers rapid stomatal closure, constrains photosynthesis, and can induce hydraulic failure even when soil moisture appears adequate, as atmospheric desiccating strength operates independently of root-zone water availability [12, 13]. Critically, VPD also serves simultaneously as a threshold trigger for bark beetle outbreak initiation by weakening tree resin defenses under sustained atmospheric stress [58, 46]. The land–atmosphere coupling further amplifies these dynamics: rising VPD accelerates evapotranspiration, drying soils and elevating sensible heat, which drives VPD still higher [12]. Forest water balance projections must therefore explicitly incorporate atmospheric demand, not precipitation alone.

Recovery Versus Transition to Alternative Stable States

Recovery is conditional, not guaranteed. The review identifies forest age, seed source proximity, soil integrity, mycorrhizal network continuity, browsing pressure, compound disturbance history, and the frequency of subsequent disturbances as the primary determinants of whether a disturbed forest regenerates or transitions to shrubland, grassland, or savanna [24]. Post-fire conifer recruitment is now failing at landscape scales across western and boreal North America due to the simultaneous operation of multiple bottlenecks: seed source elimination beyond effective dispersal distances, climatic unsuitability during seedling establishment windows, and short-interval reburning that destroys regenerating cohorts before reproductive maturity [17, 18]. Wildfire-driven forest conversion to non-forest vegetation states has been documented across millions of hectares in western North America, with warming and drought projected to accelerate this trajectory [29, 73]. Belowground, the disruption of ectomycorrhizal networks by severe fire creates priority effects in which non-forest mycorrhizal guilds foreclose tree re-establishment for decades [105]. Globally, self-replacement of dominant tree species occurs at only 21% of drought-mortality sites [26], confirming that forest die-off under contemporary conditions leads predominantly to compositional transformation rather than recovery. Critically, repeated disturbances within short intervals — a pattern made more likely by climate change and by the temporal compression inherent in accelerating beetle voltinism and fire regime shifts — dramatically reduce recovery probability [23]. The drought-beetle-fire cascade represents a particularly potent pathway toward irreversible transformation [3, 58], because each disturbance element modifies the conditions governing the next. Once alternative stable states are established, they are self-reinforcing through hysteretic feedback dynamics that resist restoration [94, 67], making early intervention the only cost-effective management response.

Implications and Call to Action

This review establishes that the primary threats to global forest stability are mechanistically linked and climatically amplified [32, 75]. Isolated management responses — whether focused on fire suppression, beetle containment, or drought adaptation alone — are insufficient because they fail to address the compound and cascading nature of the disturbance regime forests now face [3]. Conservation and forest management communities must adopt integrated frameworks that simultaneously address physiological stress thresholds, biotic disturbance agents, fire risk, hydrological change, and post-disturbance regeneration capacity. Silvicultural strategies that promote structural and compositional diversity offer demonstrable benefits for moderating compound disturbance severity [23, 115], but require implementation timescales that may not be available where cascading disturbances are already underway. Passive recovery strategies must be supplemented by assisted regeneration using climate-informed seed sourcing [117, 17] and soil restoration approaches that address belowground biological legacies, including mycorrhizal inoculation in severely degraded post-fire environments [107, 120]. Researchers must prioritize long-term observational networks capable of detecting early transition signals grounded in critical slowing down theory [92, 24] — an approach supported by satellite-based evidence of declining forest resilience across multiple biomes [77] — must integrate bark beetle and pathogen monitoring with drought and fire early-warning systems, and must expand regeneration monitoring programmes to detect non-forest transition dynamics [29, 17] before they become irreversible. Policymakers must treat forest loss not as an ecological footnote to climate change, but as one of its most consequential and potentially irreversible outcomes.

---

References

[1] Hartmann, H., Bastos, A., Das, A., et al. (2022). Climate Change Risks to Global Forest Health: Emergence of Unexpected Events of Elevated Tree Mortality Worldwide. Annual Review of Plant Biology. [2] Allen, C., Breshears, D., McDowell, N. (2015). On underestimation of global vulnerability to tree mortality and forest die-off from hotter drought in the Anthropocene. Ecosphere. [3] Raffa, K., Aukema, B., Bentz, B., et al. (2008). Cross-scale Drivers of Natural Disturbances Prone to Anthropogenic Amplification: The Dynamics of Bark Beetle Eruptions. BioScience. [4] Singh, V., Naseer, A., Mogilicherla, K., et al. (2024). Understanding bark beetle outbreaks: exploring the impact of changing temperature regimes, droughts, forest structure, and prospects for future forest pest management. Reviews in Environmental Science and Bio/Technology. [5] Liu, Q., Peng, C., Schneider, R., et al. (2023). Drought-induced increase in tree mortality and corresponding decrease in the carbon sink capacity of Canada’s boreal forests from 1970 to 2020. Global Change Biology. [6] McDowell, N., Pockman, W., Allen, C., et al. (2008). Mechanisms of plant survival and mortality during drought: why do some plants survive while others succumb to drought?. New Phytologist. [7] Sevanto, S., McDowell, N., Dickman, L., et al. (2014). How do trees die? A test of the hydraulic failure and carbon starvation hypotheses. Plant Cell & Environment. [8] Adams, H., Zeppel, M., Anderegg, W., et al. (2017). A multi-species synthesis of physiological mechanisms in drought-induced tree mortality. Nature Ecology & Evolution. [9] Choat, B., Brodribb, T., Brodersen, C., et al. (2018). Triggers of tree mortality under drought. Nature. [10] Hajek, P., Link, R., Nock, C., et al. (2022). Mutually inclusive mechanisms of drought-induced tree mortality. Global Change Biology. [11] Breshears, D., Adams, H., Eamus, D., et al. (2013). The critical amplifying role of increasing atmospheric moisture demand on tree mortality and associated regional die-off. Frontiers in Plant Science. [12] Novick, K., Ficklin, D., Grossiord, C., et al. (2024). The impacts of rising vapour pressure deficit in natural and managed ecosystems. Plant Cell & Environment. [13] McDowell, N., Sapes, G., Pivovaroff, A., et al. (2022). Mechanisms of woody-plant mortality under rising drought, CO2 and vapour pressure deficit. Nature Reviews Earth & Environment. [14] Hammond, W., Williams, P., Abatzoglou, J., et al. (2022). Global field observations of tree die-off reveal hotter-drought fingerprint for Earth’s forests. Nature Communications. [15] Pureswaran, D., Roques, A., Battisti, A. (2018). Forest Insects and Climate Change. Current Forestry Reports. [16] Aaron S. Weed, Matthew P. Ayres, Jeffrey A. Hicke (2013). Consequences of climate change for biotic disturbances in North American forests. Ecological Monographs. [17] Stevens‐Rumann, C., Morgan, P. (2019). Tree regeneration following wildfires in the western US: a review. Fire Ecology. [18] Whitman, E., Parisien, M., Thompson, D., et al. (2019). Short-interval wildfire and drought overwhelm boreal forest resilience. Scientific Reports. [19] McDowell, N., Ryan, M., Zeppel, M., et al. (2013). Improving our knowledge of drought-induced forest mortality through experiments, observations, and modeling. New Phytologist. [20] Hartmann, H., Moura, C., Anderegg, W., et al. (2018). Research frontiers for improving our understanding of drought-induced tree and forest mortality. New Phytologist. [21] Stritih, A., Seidl, R., Senf, C. (2023). Alternative states in the structure of mountain forests across the Alps and the role of disturbance and recovery. Landscape Ecology. [22] Serra‐Diaz, J., Maxwell, C., Lucash, M., et al. (2018). Disequilibrium of fire-prone forests sets the stage for a rapid decline in conifer dominance during the 21st century. Scientific Reports. [23] Hessburg, P., Miller, C., Parks, S., et al. (2019). Climate, Environment, and Disturbance History Govern Resilience of Western North American Forests. Frontiers in Ecology and Evolution. [24] Reyer, C., Brouwers, N., Rammig, A., et al. (2015). Forest resilience and tipping points at different spatio-temporal scales: approaches and challenges. Journal of Ecology. [25] Aukema, B., Zhu, J., Möller, J., et al. (2009). Predisposition to bark beetle attack by root herbivores and associated pathogens: Roles in forest decline, gap formation, and persistence of endemic bark beetle populations. Forest Ecology and Management. [26] Batllori, E., Lloret, F., Aakala, T., et al. (2020). Forest and woodland replacement patterns following drought-related mortality. Proceedings of the National Academy of Sciences. [27] Boucher, D., Gauthier, S., Thiffault, N., et al. (2019). How climate change might affect tree regeneration following fire at northern latitudes: a review. New Forests. [28] Flores, B., Montoya, E., Sakschewski, B., et al. (2024). Critical transitions in the Amazon forest system. Nature. [29] Coop, J., Parks, S., Stevens‐Rumann, C., et al. (2020). Wildfire-Driven Forest Conversion in Western North American Landscapes. BioScience. [30] Connor, S., Araújo, J., Boski, T., et al. (2021). Drought, fire and grazing precursors to large-scale pine forest decline. Diversity and Distributions. [31] Au, J., Bloom, A., Parazoo, N., et al. (2023). Forest productivity recovery or collapse? Model-data integration insights on drought-induced tipping points. Global Change Biology. [32] Anderegg, W., Hicke, J., Fisher, R., et al. (2015). Tree mortality from drought, insects, and their interactions in a changing climate. New Phytologist. [33] Desprez‐Loustau, M., Marçais, B., Nageleisen, L., et al. (2006). Interactive effects of drought and pathogens in forest trees. Annals of Forest Science. [34] Choat, B., Jansen, S., Brodribb, T., et al. (2012). Global convergence in the vulnerability of forests to drought. Nature. [35] William R. L. Anderegg, Joseph A. Berry, Duncan D. Smith, et al. (2011). The roles of hydraulic and carbon stress in a widespread climate-induced forest die-off. Proceedings of the National Academy of Sciences. [36] Song, H., Zhong, T., Zhu, J., et al. (2025). Higher risk of hydraulic dysfunction and carbohydrate depletion of declining Larix principis‑rupprechtii trees. Ecological Processes. [37] Adams, H., Guardiola‐Claramonte, M., Barron‐Gafford, G., et al. (2009). Temperature sensitivity of drought-induced tree mortality portends increased regional die-off under global-change-type drought. Proceedings of the National Academy of Sciences. [38] Zeppel, M., Adams, H., Anderegg, W. (2011). Mechanistic causes of tree drought mortality: recent results, unresolved questions and future research needs. New Phytologist. [39] Nathan G. McDowell (2011). Mechanisms Linking Drought, Hydraulics, Carbon Metabolism, and Vegetation Mortality. PLANT PHYSIOLOGY. [40] Ansgar Kahmen, David Basler, Günter Hoch, et al. (2022). Root water uptake depth determines the hydraulic vulnerability of temperate European tree species during the extreme 2018 drought. Plant Biology. [41] Ripullone, F., Camarero, J., Colangelo, M., et al. (2020). Variation in the access to deep soil water pools explains tree-to-tree differences in drought-triggered dieback of Mediterranean oaks. Tree Physiology. [42] Ding, J., Buotte, P., Bales, R., et al. (2023). Coordination of rooting, xylem, and stomatal strategies explains the response of conifer forest stands to multi-year drought in the southern Sierra Nevada of California. Biogeosciences. [43] Gaylord, M., Kolb, T., Pockman, W., et al. (2013). Drought predisposes pinon—juniper woodlands to insect attacks and mortality. New Phytologist. [44] Nadir Erbilgin, Jonathan A. Cale, Inka Lusebrink, et al. (2016). Water-deficit and fungal infection can differentially affect the production of different classes of defense compounds in two host pines of mountain pine beetle. Tree Physiology. [45] Koontz, M., Latimer, A., Mortenson, L., et al. (2021). Cross-scale interaction of host tree size and climatic water deficit governs bark beetle-induced tree mortality. Nature Communications. [46] Netherer, S., Matthews, B., Katzensteiner, K., et al. (2015). Do water-limiting conditions predispose Norway spruce to bark beetle attack?. New Phytologist. [47] Anderegg, W., Berry, J., Field, C. (2012). Linking definitions, mechanisms, and modeling of drought-induced tree death. Trends in Plant Science. [48] Anderegg, W., Plavcová, L., Anderegg, L., et al. (2013). Drought’s legacy: multiyear hydraulic deterioration underlies widespread aspen forest die-off and portends increased future risk. Global Change Biology. [49] Trugman, A., Detto, M., Bartlett, M., et al. (2018). Tree carbon allocation explains forest drought-kill and recovery patterns. Ecology Letters. [50] McDowell, N., Allen, C., Anderson‐Teixeira, K., et al. (2018). Drivers and mechanisms of tree mortality in moist tropical forests. New Phytologist. [51] Abdalla, M., Bitterlich, M., Jansa, J., et al. (2023). The role of arbuscular mycorrhizal symbiosis in improving plant water status under drought. Journal of Experimental Botany. [52] Diao, H., Cernusak, L., Saurer, M., et al. (2024). Dry inside: progressive unsaturation within leaves with increasing vapour pressure deficit affects estimation of key leaf gas exchange parameters. New Phytologist. [53] Li, J., Zhang, Y., Bevacqua, E., et al. (2024). Future increase in compound soil drought-heat extremes exacerbated by vegetation greening. Nature Communications. [54] Marchin, R., Backes, D., Ossola, A., et al. (2022). Extreme heat increases stomatal conductance and drought-induced mortality risk in vulnerable plant species. Global Change Biology. [56] Brodribb, T., Powers, J., Cochard, H., et al. (2020). Hanging by a thread? Forests and drought. Science. [57] McDowell, N., Williams, P., Xu, C., et al. (2019). Multi-scale predictions of massive conifer mortality due to chronic temperature rise. Nature Climate Change. [58] Das, A., Baldo, M., Dobor, L., et al. (2025). The increasing role of drought as an inciting factor of bark beetle outbreaks can cause large‑scale transformation of Central European forests. Landscape Ecology. [59] Knapp, A., Condon, K., Folks, C., et al. (2024). Field experiments have enhanced our understanding of drought impacts on terrestrial ecosystems—But where do we go from here?. Functional Ecology. [60] Knutzen, F., Averbeck, P., Barrasso, C., et al. (2025). Impacts on and damage to European forests from the 2018–2022 heat and drought events. Natural hazards and earth system sciences. [61] Mária Potterf, Tobias Frühbrodt, Dominik Thom, et al. (2025). Hotter drought increases population levels and accelerates phenology of the European spruce bark beetle Ips typographus. Forest Ecology and Management. [62] Jones, M., Abatzoglou, J., Veraverbeke, S., et al. (2022). Global and Regional Trends and Drivers of Fire Under Climate Change. Reviews of Geophysics. [63] Bose, A., Geßler, A., Büntgen, U., et al. (2024). Tamm review: Drought-induced Scots pine mortality – trends, contributing factors, and mechanisms. Forest Ecology and Management. [64] Galina Timofeeva, Kerstin Treydte, Harald Bugmann, et al. (2017). Long-term effects of drought on tree-ring growth and carbon isotope variability in Scots pine in a dry environment. Tree Physiology. [65] Camarero, J., Gazol, A., Sangüesa‐Barreda, G., et al. (2015). To die or not to die: early warnings of tree dieback in response to a severe drought. Journal of Ecology. [66] Rigling, A., Bigler, C., Eilmann, B., et al. (2012). Driving factors of a vegetation shift from Scots pine to pubescent oak in dry Alpine forests. Global Change Biology. [67] Stritih, A., Seidl, R., Senf, C. (2023). Alternative states in the structure of mountain forests across the Alps and the role of disturbance and recovery. Landscape Ecology. [68] Gómez, D., Riggins, J., Cognato, A. (2021). Chapter 10 Bark Beetles. [69] Barbara Bentz, Jacques Régnière, Christopher J. Fettig, et al. (2010). Climate Change and Bark Beetles of the Western United States and Canada: Direct and Indirect Effects. BioScience. [70] Zaman, R., Shah, A., Shah, A., et al. (2024). Unraveling the multifaceted effects of climatic factors on mountain pine beetle and its interaction with fungal symbionts. Global Change Biology. [71] Jessica E. Halofsky, David L. Peterson, Brian J. Harvey (2020). Changing wildfire, changing forests: the effects of climate change on fire regimes and vegetation in the Pacific Northwest, USA. Fire Ecology. [72] Schoennagel, T., Veblen, T., Negrón, J., et al. (2012). Effects of Mountain Pine Beetle on Fuels and Expected Fire Behavior in Lodgepole Pine Forests, Colorado, USA. PLoS ONE. [73] Davis, K., Robles, M., Kemp, K., et al. (2023). Reduced fire severity offers near-term buffer to climate-driven declines in conifer resilience across the western United States. Proceedings of the National Academy of Sciences. [74] Leander D. L. Anderegg, William R. L. Anderegg, John T. Abatzoglou, et al. (2013). Drought characteristics’ role in widespread aspen forest mortality across Colorado, USA. Global Change Biology. [75] Brando, P., Paolucci, L., Ummenhofer, C., et al. (2019). Droughts, Wildfires, and Forest Carbon Cycling: A Pantropical Synthesis. Annual Review of Earth and Planetary Sciences. [76] Longlong Ma, Jun Ma, Pu Yan, et al. (2025). Planted Forests in China Have Higher Drought Risk Than Natural Forests. Global Change Biology. [77] Forzieri, G., Dakos, V., McDowell, N., et al. (2022). Emerging signals of declining forest resilience under climate change. Nature. [78] Desprez-Loustau, M., Aguayo, J., Dutech, C., et al. (2015). An evolutionary ecology perspective to address forest pathology challenges of today and tomorrow. Annals of Forest Science. [79] Anderegg, W., Trugman, A., Bowling, D., et al. (2019). Plant functional traits and climate influence drought intensification and land-atmosphere feedbacks. Proceedings of the National Academy of Sciences. [80] Chen, H., Sloggy, M., Flake, S., et al. (2025). Impacts of climate-driven insect population change on sawtimber provisioning, carbon sequestration, and water retention: a case study of bark beetle outbreaks in the USA. Frontiers in Forests and Global Change. [81] Sippel, S., Reichstein, M., Ma, X., et al. (2018). Drought, Heat, and the Carbon Cycle: a Review. Current Climate Change Reports. [82] Ruiz, J., Cochard, H., Delzon, S., et al. (2024). Plant hydraulics at the heart of plant, crops and ecosystem functions in the face of climate change. New Phytologist. [83] McLauchlan, K., Higuera, P., Miesel, J., et al. (2020). Fire as a fundamental ecological process: Research advances and frontiers. Journal of Ecology. [84] Williams, P., Abatzoglou, J., Gershunov, A., et al. (2019). Observed Impacts of Anthropogenic Climate Change on Wildfire in California. Earth s Future. [85] Parisien, M., Barber, Q., Bourbonnais, M., et al. (2023). Abrupt, climate-induced increase in wildfires in British Columbia since the mid-2000s. Communications Earth & Environment. [86] Wasserman, T., Mueller, S. (2022). Climate influences on future fire severity: a synthesis of climate-fire interactions and impacts on fire regimes, high-severity fire, and forests in the western United States. Fire Ecology. [87] James S. Clark, Louis R. Iverson, Christopher W. Woodall, et al. (2016). The impacts of increasing drought on forest dynamics, structure, and biodiversity in the United States. Global Change Biology. [88] Culbert, S., Arthur, M., Cotton, C., et al. (2025). Higher severity fire increases the long-term competitiveness of pyrophytes in an upland oak-pine forest, Kentucky, USA. Fire Ecology. [89] Folke, C. (2016). Resilience (Republished). Ecology and Society. [90] Walker, B., Gunderson, L., Kinzig, A., et al. (2006). A Handful of Heuristics and Some Propositions for Understanding Resilience in Social-Ecological Systems. Ecology and Society. [91] Folke, C., Carpenter, S., Walker, B., et al. (2010). Resilience Thinking: Integrating Resilience, Adaptability and Transformability. Ecology and Society. [92] Sardanyés, J., Ivančić, F., Vidiella, B. (2024). Identifying regime shifts, transients and late warning signals for proactive ecosystem management. Biological Conservation. [93] Walker, B., Carpenter, S., Anderies, J., et al. (2002). Resilience Management in Social-ecological Systems: a Working Hypothesis for a Participatory Approach. Conservation Ecology. [94] Bestelmeyer, B., Ellison, A., Fraser, W., et al. (2011). Analysis of abrupt transitions in ecological systems. Ecosphere. [95] Boulton, C., Lenton, T., Boers, N. (2022). Pronounced loss of Amazon rainforest resilience since the early 2000s. Nature Climate Change. [96] Runge, K., Tucker, M., Crowther, T., et al. (2025). Monitoring Terrestrial Ecosystem Resilience Using Earth Observation Data: Identifying Consensus and Limitations Across Metrics. Global Change Biology. [97] Koen Siteur, Maarten B. Eppinga, Arjen Doelman, et al. (2016). Ecosystems off track: rate‐induced critical transitions in ecological models. Oikos. [98] Zak Ratajczak, Paolo D’Odorico, Scott L. Collins, et al. (2017). The interactive effects of press/pulse intensity and duration on regime shifts at multiple scales. Ecological Monographs. [99] Staal, A., Dekker, S., Hirota, M., et al. (2015). Synergistic effects of drought and deforestation on the resilience of the south-eastern Amazon rainforest. Ecological Complexity. [100] Baho, D., Allen, C., Garmestani, A., et al. (2017). A quantitative framework for assessing ecological resilience. Ecology and Society. [101] Angeler, D., Allen, C. (2016). Quantifying resilience. Journal of Applied Ecology. [102] Werner A. Kurz, Caren C. Dymond, G. Stinson, et al. (2008). Mountain pine beetle and forest carbon feedback to climate change. Nature. [103] Nolan, R., Collins, L., Leigh, A., et al. (2021). Limits to post-fire vegetation recovery under climate change. Plant Cell & Environment. [104] Jeff Shatford, David E. Hibbs, Klaus J. Puettmann (2007). Conifer Regeneration after Forest Fire in the Klamath-Siskiyous: How Much, How Soon?. Journal of Forestry. [105] Karavani, A., Boer, M., Baudena, M., et al. (2018). Fire-induced deforestation in drought-prone Mediterranean forests: drivers and unknowns from leaves to communities. Ecological Monographs. [106] Hammond, W., Yu, K., Wilson, L., et al. (2019). Dead or dying? Quantifying the point of no return from hydraulic failure in drought-induced tree mortality. New Phytologist. [107] Asmelash, F., Bekele, T., Birhane, E. (2016). The Potential Role of Arbuscular Mycorrhizal Fungi in the Restoration of Degraded Lands. Frontiers in Microbiology. [108] Semchenko, M., Barry, K., de Vries, F., et al. (2022). Deciphering the role of specialist and generalist plant-microbial interactions as drivers of plant-soil feedback. New Phytologist. [109] Jayaramaiah, R., Martins, C., Egidi, E., et al. (2024). Soil function-microbial diversity relationship is impacted by plant functional groups under climate change. Soil Biology and Biochemistry. [110] Jaeger, A., Hartmann, M., Six, J., et al. (2023). Contrasting sensitivity of soil bacterial and fungal community composition to one year of water limitation in Scots pine mesocosms. FEMS Microbiology Ecology. [111] Shovon, T., Auge, H., Haase, J., et al. (2024). Positive effects of tree species diversity on productivity switch to negative after severe drought mortality in a temperate forest experiment. Global Change Biology. [112] Michaelian, M., Hogg, E., Hall, R., et al. (2011). Massive mortality of aspen following severe drought along the southern edge of the Canadian boreal forest. Global Change Biology. [113] Dakos, V., Kéfi, S. (2022). Ecological resilience: what to measure and how. Environmental Research Letters. [114] Skelton, R., West, A., Dawson, T. (2015). Predicting plant vulnerability to drought in biodiverse regions using functional traits. Proceedings of the National Academy of Sciences. [115] Germain, S., Hansen, W., Canham, C. (2025). Reassembly of Disturbed Forests Portends Climate Resilience but Diversity Loss. Ecosystems. [116] Jarne, P. (2025). The Anthropocene and the biodiversity crisis: an eco-evolutionary perspective. Comptes Rendus Biologies. [117] Konrad, H., George, J., Szukala, A. (2023). The State of Forest Genetic Diversity: Anthropogenic Impacts and Conservation Initiatives. [118] Pardini, R., Bueno, A., Gardner, T., et al. (2010). Beyond the Fragmentation Threshold Hypothesis: Regime Shifts in Biodiversity Across Fragmented Landscapes. PLoS ONE. [119] Gorzelak, M., Asay, A., Pickles, B., et al. (2015). Inter-plant communication through mycorrhizal networks mediates complex adaptive behaviour in plant communities. AoB Plants. [120] Chen, M., Arato, M., Borghi, L., et al. (2018). Beneficial Services of Arbuscular Mycorrhizal Fungi – From Ecology to Application. Frontiers in Plant Science.