Coordinated volatile isoprenoid production and leaf turnover strategy protect central Amazon Forest trees against stress

The Amazon Forest covers more than half of the tropical forest area in the world, and stores up to 200 Pg of carbon^1,2. Yet, climate stress and associated disturbances are increasingly shifting parts of the forest toward a net carbon source³. Therefore, understanding the future carbon balance of the Amazon Forest depends not only on quantifying disturbance rates but also on elucidating the ecophysiological mechanisms that regulate how trees respond to stress and disturbance. Among these mechanisms, leaf turnover strategy and volatile isoprenoid (VI; i.e., isoprene, monoterpenes and sesquiterpenes) production are particularly important, as they link plant stress responses, carbon balance, and climate regulation through forest-atmosphere feedbacks.

In the central Amazon Forest, precipitation seasonality drives leaf turnover (i.e., the rate at which leaf biomass is replaced through leaf production and senescence), with substantial crown renewal occurring during the driest months of the year⁴. This seasonal crown renewal is largely driven by brevideciduous trees, which shed large fractions or all of their leaves synchronously every year, and remain leafless for up to one month before flushing a new leaf cohort^5,6. Evergreen trees, although predominant in this forest^4,7,8, exhibit smaller and more irregular leaf flushing events^5,6.

Studies have proposed drought stress and herbivore avoidance as evolutionary drivers of this massive dry-season leaf turnover⁴. Brevideciduous trees are less resistant to embolism, and losing leaf area or replacing old leaves for new ones with better stomatal control is a proposed drought resistance strategy^6,7,9,10. Also, young leaves are softer and more vulnerable to herbivory, thus exchanging leaves during the dry season may reduce exposure to higher wet-season herbivore pressure^4,11,12. Importantly, these strategies are not strictly conserved at the species level and can shift in response to stress events and disturbances^13,14,15. This plasticity suggests that climate change may reorganize canopy carbon dynamics and associated physiological responses at regional scales, including VI production.

Isoprene (C₅H₈), and some light-dependent monoterpenes (C₁₀H₁₆)^16,17, are produced from recently assimilated photosynthetic carbon and immediately emitted^18,19. They are associated with increased photosynthetic thermotolerance and have been proposed to act through multiple mechanisms. These include scavenging reactive oxygen species (ROS)²⁰, acting as sinks for excess reducing power²¹, and enhancing photochemical efficiency^22,23. Recent studies also emphasize signaling roles linking growth and defense responses^24,25,26,27.

Sesquiterpenes (C₁₅H₂₄) and most monoterpenes accumulate in specialized storage structures (e.g., resin ducts or glandular trichomes) and are emitted gradually under normal conditions or rapidly upon structural damage^28,29,30,31. They are associated with biotic stress protection, acting as herbivore deterrents and plant signaling molecules^32,33. Because VI emissions draw on both recently assimilated and stored carbon pools, they represent a stress-mitigation mechanism while also incurring a carbon cost, linking physiological stress protection to ecosystem carbon balance.

VIs are also the most abundantly emitted class of plant biogenic volatile organic compounds (BVOCs)³⁴ and significantly impact atmospheric processes. After emission, they react with hydroxyl radicals (OH), contributing to secondary organic aerosol production, affecting cloud formation and radiative balance^35,36,37,38. In the presence of nitrogen oxides (NO_x), they also promote tropospheric O₃ formation³⁹. Although isoprene dominates global fluxes³⁴, monoterpenes and sesquiterpenes have higher particle formation potential and carbon costs^40,41,42. Changes in VI emission composition may therefore alter plant carbon budgets and forest-atmosphere feedbacks.

Warmer and drier climates are expected to favor the selection of brevideciduous trees⁷ and strong isoprene or light-dependent monoterpene emitters capable of sustaining higher photosynthetic performance under heat stress^43,44,45,46. At the same time, more frequent and intense stress events will likely promote stronger fluxes from heavier and more reactive, temperature-sensitive, monoterpenes and sesquiterpenes^43,47,48. Such composition shifts have already been observed in the Amazon during El Niño years^42,48,49. These observations indicate that climate stress can reorganize both the magnitude and chemical composition of VI emissions.

The Amazon Forest represents the largest and most chemically diverse source of VIs to the atmosphere^34,39,50. However, current BVOC emission models⁵¹ rely on simplified emission factors (i.e., emission rates measured under standard conditions of incident photosynthetic photon flux density, PPFD = 1000 µmol m⁻² s⁻¹; and leaf temperature = 30 °C) tied to generalized plant functional types (PFTs) (e.g., CLM4 model⁵²) and response parameters derived from limited flux tower data. These approaches mask variability in tropical species physiology and environmental responses⁵³. Moreover, they do not account for how leaf turnover strategy modulates emission responses to light and heat stress, nor how such differences propagate to canopy-scale flux estimates under future climate scenarios.

Hence, in this study, we asked: (i) how do VI emission profiles and their responses to light and temperature change across different leaf turnover strategies; (ii) how do photochemical activity traits vary between isoprene emitters and non-emitters from different leaf turnover strategies; and (iii) how may the interaction of leaf turnover strategy and light and temperature changes affect canopy-level isoprene flux estimates? By addressing these questions, our study provides mechanistic insight into how leaf turnover strategy modulates the composition and temperature sensitivity of VI emissions and their scaling to canopy fluxes, providing constrained parameters that improve representation of Amazon forest-atmosphere feedbacks in current models.

VI emission responses to changes in light and temperature in different leaf turnover strategies

All compounds showed increases in emission rates with leaf temperature (Fig. 1 and Supplementary Table 1). Isoprene and monoterpene temperature responses were similar across leaf turnover strategies (Fig. 1A, B). Sesquiterpene emissions, on the other hand, showed steeper temperature-driven increases in brevideciduous trees (p (temp:pheno) = 0.04, Fig. 1C). At 45 °C, brevideciduous trees also showed higher sesquiterpene emission rates (p (T₄₅) = 0.02). Carbon partitioning shifted from lighter to heavier compound emissions with increasing leaf temperature (p (temp) = 0.06; Fig. 1D).

Transpiration (E) and leaf-air vapor pressure deficit (VPD) changed in response to light and temperature (Supplementary Figs. 1 and 2). Including E and VPD as covariates significantly improved model performance for monoterpene emissions during temperature response curves (ΔAIC = −19.0; ΔR² = +0.20; p < 0.001 for nested model comparison; Supplementary Table 2). In contrast, sesquiterpene emissions showed no consistent improvement in model fit when E or VPD were included (ΔAIC ≥ 0; p > 0.1; Supplementary Table 2).

Monoterpene emissions in Brosimum parinarioides, Cariniana decandra, and Croton matourensis (Fig. 2A, D, J) peaked near 40 °C. Monoterpene emissions in B. parinarioides were dominated by trans-β-ocimene, with smaller proportions of cis-β-ocimene (Fig. 2A). α-Pinene dominated emissions in C. decandra and C. matourensis. C. decandra also emitted lower proportions of β-pinene, camphene, myrcene, sabinene, and tricyclene (Fig. 2D). C. matourensis emitted small proportions of sabinene, camphene, and myrcene starting from 37.5 °C (Fig. 2J).

While assimilation rate (A_n) and stomatal conductance (g_s) declined sharply above 37.5–40 °C (Fig. 3A), monoterpene emission composition exhibited pronounced temperature-dependent restructuring (Fig. 3B). Monoterpenes were grouped according to their biosynthetic carbocation intermediates and pathway temperature sensitivities. Emissions of group 1 monoterpenes (cis-/trans-β-ocimene, myrcene) increased with leaf temperature and dominated the emission profile up to 37.5 °C. At the highest temperatures (40–42.5 °C), the contribution of group 1 declined, while group 5 compounds (camphene, tricyclene) increased sharply, becoming dominant at 45 °C. Groups 2–4 showed comparatively minor changes. Across the full temperature range, brevideciduous trees exhibited higher proportional contributions of Group 1 compounds relative to evergreen trees (p = 0.046; Supplementary Table 3).

Percentages of photosynthetic carbon (%C) loss to VI emissions increased with leaf temperature up to 40 °C, after which patterns became more variable (Table 1 and Supplementary Fig. 3). This trend was independent of isoprene emission capacity or leaf turnover strategy. For many species, %C loss to isoprene emissions became negative at a given leaf temperature, indicating continued emissions after photosynthesis ceased and production likely derived from alternative carbon pools (carbon storage, CS; Table 1). No significant differences were detected in photosynthetic thermal limits (i.e., highest observed leaf temperature where assimilation rates were still positive) between isoprene emitters and non-emitters from different leaf turnover strategies (Supplementary Fig. 4).

Photochemical activity traits of isoprene emitters and non-emitters from different leaf turnover strategies

To facilitate interpretation, photochemical activity traits presented in this section were grouped according to the type of analysis they support. Figure 4 shows parameters derived as single representative values from light or temperature response curves, allowing direct comparisons between leaf turnover strategies and isoprene emission groups (emitters vs. non-emitters), including their interaction. Figure 5 shows variables that describe continuous responses to changes in leaf temperature, for which dynamic patterns are more informative than single summary values.

Photochemical activity traits did not differ between isoprene emitters and non-emitters from different leaf turnover strategies, but results suggested higher light saturation point (LSP) in brevideciduous isoprene emitters (p (isop:pheno) = 0.08; Fig. 4A). Across all groups, electron transport rate (J), photochemical quenching (qP), quantum efficiency of photosystem II (ϕ_PSII), A_n, and g_s, declined progressively with leaf temperature (Fig. 5). Non-linear temperature response models revealed group-level differences in baseline photosynthetic parameters and thermal response traits (Supplementary Table 4).

The highest k₂₅ (modeled baseline rate at 25 °C) values for A_n and g_s were observed in brevideciduous isoprene emitters and evergreen non-emitters, respectively. For J, ϕ_PSII, and qP, the highest k₂₅ values occurred without a consistent pattern. Across all groups and variables, bootstrapped confidence intervals (CIs) for k₂₅ were strictly positive, supporting the reliability of these baseline estimates. E_α (activation energy) showed high uncertainty, as CIs included zero in nearly all cases. For J, qP, ϕ_PSII, and g_s, ΔS (entropy term) tended to be higher in non-emitters, particularly among evergreen trees. However, many ΔS CIs approached the model’s upper constraint (2000 J mol⁻¹ K⁻¹), indicating uncertainty in upper-bound estimates. Hd (deactivation energy) was consistently higher in evergreen non-emitters across most variables, although brevideciduous non-emitters showed the highest Hd for A_n. CIs for Hd estimates were broad but did not reach model bounds.

Light and temperature response parameters of isoprene emissions from different leaf turnover strategies and canopy-level flux projections

Isoprene light and temperature response parameters α (initial light-response slope), C_L1 (light-saturation coefficient), C_T1, and C_T2 (activation and deactivation energy parameters, respectively) derived from this study are presented in Supplementary Table 5. Brevideciduous isoprene emitters had higher C_T2 values relative to evergreen emitters (p = 0.001; Fig. 6D). Isoprene LSP values were higher in brevideciduous trees, although the difference was weakly supported statistically (p = 0.09; Fig. 6E). Pool sizes of the immediate isoprene synthesis substrate dimethylallyl pyrophosphate (DMADP) (Supplementary Fig. 5A) and isoprene synthase activity rates (Supplementary Fig. 5B) did not differ between leaf turnover strategies.

Isoprene fluxes estimated with parameters derived from this study were generally above 2000 µg m⁻² h⁻¹ at 09:00, peaked at 13:00 and decreased again at 17:00 (Fig. 7). When fluxes were estimated with default parameters⁵⁴, values were consistently higher (Fig. 8). Kruskal–Wallis pairwise comparisons confirmed the statistical significance of this pattern at 09:00 and 13:00 for evergreen trees (Supplementary Fig. 6B), with a similar trend for brevideciduous trees (Supplementary Fig. 6A).

Model comparisons (Supplementary Table 6 and see “Methods” for details) further showed that, relative to Model 1 (observed emission factors^55,56 + derived parameters), Model 2 (observed emission factors^55,56 + default parameters⁵⁴) and Model 3 (default emission factors³⁴ + default parameters⁵⁴) produced flux estimates that were 98.4–256.3% and 61.1–218.3% higher, respectively, across leaf turnover strategies.

Our findings indicate that leaf turnover strategy structures the temperature sensitivity of sesquiterpene emissions and the composition of monoterpene emissions, and coordinates differences in photochemical temperature responses and carbon allocation under heat stress. Incorporating empirical parameterization that accounts for leaf turnover strategy into canopy simulations yielded more realistic flux estimates than default approaches^34,54, which substantially overestimated isoprene fluxes. Accounting for this phenological organization is therefore essential for mechanistically representing forest-atmosphere coupling in a warming climate.

VI emission responses to changes in light and temperature in different leaf turnover strategies

Isoprene emissions increased with temperature across all species, confirming the thermal sensitivity of this pathway under Amazon Forest conditions^23,53. Even species with no detectable isoprene emissions at 30 °C began emitting as the temperature increased. These observations suggest that the fraction of functionally isoprene-emitting species in this forest is likely larger than previously assumed (~20%⁵⁷), consistent with recent studies^53,58. Isoprene emission factors observed here also exceeded values reported previously for the same site⁵⁶ (Supplementary Fig. 7), reflecting seasonal differences in leaf developmental stage and photosynthetic activity during early wet season conditions^50,59,60,61.

Monoterpene and sesquiterpene emissions showed strong temperature sensitivity, as indicated by the estimated β coefficients⁶², which frequently exceeded the ±20% temperature sensitivity range adopted in MEGAN v2.1³⁴ (Supplementary Fig. 8). This suggests that current parameterizations underestimate thermal amplification of heavier compounds in this system^29,47.

Notable exceptions were α- and β-pinene and trans-β-ocimene, whose temperature responses deviated from the exponential behavior expected for storage pool release. Trans-β-ocimene emissions in B. parinarioides and α-pinene emissions in C. decandra and C. matourensis peaked at 40 °C rather than increasing exponentially, suggesting light-dependent enzymatic production^62,63,64. Monoterpene light responses in B. parinarioides and C. matourensis further support this interpretation, as emissions increased concurrently with assimilation rates under increasing light (Supplementary Fig. 9). Moreover, ¹³C-labeling studies in the central Amazon Forest have demonstrated light-dependent emissions of β-ocimene⁴⁸, supporting the physiological control inferred here. Given the high reactivity of trans-β-ocimene with O₃, OH, and NO₃ radicals⁶⁵, such enzymatically regulated emissions may disproportionately influence atmospheric chemistry under warm conditions.

Highly reactive group 1 emissions increased with leaf temperature and became dominant at ~37.5–40 °C, in agreement with the “monoterpene thermometer” framework⁴⁸. At higher temperatures (42.5–45 °C), however, group 1 contributions declined, and emissions became increasingly dominated by group 5 compounds. This shift suggests a transition from physiologically controlled, light-dependent emissions toward volatilization-driven release from storage pools. This transition is likely promoted by elevated temperature and leaf-air VPD, an important driver of canopy BVOC emissions⁶⁶. Because group 5 compounds have longer atmospheric lifetimes than group 1^16,67, this shift implies physiological changes within leaves as well as altered atmospheric processes, particularly during extreme heat events.

Increasing temperature was accompanied by substantial carbon allocation to VI emissions, particularly near the upper thermal limits of photosynthesis. Carbon losses to isoprene reached up to 17.3% of assimilated carbon prior to photosynthetic collapse—well above the 1–2% typically reported under non-stress conditions^68,69. In many cases, emissions persisted even after photosynthesis ceased, indicating the use of alternative carbon sources^70,71. Monoterpene and sesquiterpene emissions accounted for even larger carbon losses in some species (e.g., 100% and 40% in Pouteria guianensis, respectively), reflecting strong mobilization of stored carbon pools. Together, these findings show that rising leaf temperature reorganizes not only emission composition but also the magnitude of carbon investment in VI production.

Given that Amazon Forest canopies already experience leaf temperatures exceeding 40 °C^48,72, these temperature-driven shifts in emission magnitude, composition and carbon allocation are ecologically relevant under current climate conditions. Many species studied here rank among the most important contributors to regional biomass⁷³, suggesting that observed emission responses may scale to ecosystem-level fluxes. Monoterpenes and sesquiterpenes impose higher carbon costs than isoprene and contribute disproportionately more to secondary organic aerosol formation and O₃ chemistry^40,41,74. Warmer and drier climates are projected to favor brevideciduous trees⁷—which here showed higher temperature sensitivity of sesquiterpene emissions and larger proportional emissions of highly reactive group 1 compounds. These patterns suggest that climate-driven shifts in leaf turnover strategy could amplify forest-atmosphere feedback through coordinated changes in emission magnitude and chemical reactivity.

Physiological changes in isoprene emitters and non-emitters across different leaf turnover strategies

Brevideciduous isoprene emitters showed higher baseline photosynthetic performance and tended to reach higher light saturation points. This pattern suggests an enhanced capacity to operate under high irradiance, consistent with the proposed protective role of isoprene in mitigating excess energy and oxidative stress⁷⁵.

Concurrently, brevideciduous trees displayed stronger temperature-driven increases in sesquiterpene emissions and higher emissions of highly reactive monoterpenes. These carbon-rich compounds play important roles in herbivore deterrence and rapid plant signaling^32,33. Such functions are particularly relevant during the synchronous, massive dry-season leaf renewal typical of brevideciduous trees^4,15. This suggests that these trees rely on a VI-mediated stress-response strategy. Such interpretation can be further supported by evidence of terpene synthase gene upregulation in isoprene-fumigated Arabidopsis⁷⁶, positive correlations between isoprene and sesquiterpene emissions⁷⁷, and increased diversity of stored sesquiterpenes with higher isoprene emission factors⁵⁶.

Evergreen non-emitters, on the other hand, had higher baseline stomatal conductance and greater thermal stability. This likely reflects a different strategy, focused on sustained gas exchange and photosynthetic stability rather than VI-mediated protection. Sustained stomatal opening may increase exposure to hydraulic risk under high evaporative demand. We therefore exploratorily examined whether hydraulic vulnerability varied between leaf turnover strategy and isoprene emission groups. While evergreen trees exhibited lower hydraulic vulnerability (more negative P₅₀), consistent with greater hydraulic safety¹⁰, this was not associated with isoprene emissions (Supplementary Fig. 10). However, given the limited overlap between hydraulic and VI measurements in our dataset, this result should be interpreted cautiously. This lack of association may be attributable to the small sample size for combined measurements of hydraulic vulnerability, leaf turnover strategy, leaf gas exchange and isoprene emission. Therefore, we propose that future studies combining such measurements across contrasting environments will be critical to determine whether isoprene emitters and non-emitters differ in their operating proximity to hydraulic failure thresholds.

Nevertheless, our results do not support a uniform stress-tolerance benefit from isoprene emission^{23,45,46,78,79}. Instead, the patterns observed here indicate that central Amazon Forest trees employ contrasting, yet coherent, stress-response strategies. Brevideciduous trees combine enhanced volatile responsiveness and chemical investment to support massive dry-season leaf turnover. Evergreen trees, in contrast, maintain higher baseline stomatal conductance, thermally stable photochemistry, and greater hydraulic safety, indicating a strategy centered on sustained physiological stability rather than VI-mediated protection.

Evaluation of MEGAN parameterizations using canopy-scale isoprene fluxes

We evaluated how isoprene light and temperature response parameters derived from this study compared to the default parameterization used in MEGAN⁵⁴. Model 1 (observed emission factors + derived parameters) predicted 13:00 fluxes of ~4.5 mg m⁻² h⁻¹ from October to December, assuming a forest composition of ~70% evergreen and ~30% brevideciduous trees. In contrast, Model 3 (default emission factors + default parameters) predicted ~12.7 mg m⁻² h⁻¹ under the same conditions, indicating a substantial overestimation of isoprene fluxes. Our results confirm that inappropriate emission factors contribute strongly to this bias, in agreement with previous findings⁵⁰. We further show that differences in isoprene light and temperature response parameters between leaf turnover strategies amplify this effect. This leads to particularly strong overestimation in evergreen trees, with important implications for regional isoprene flux estimates, given their predominance in this forest.

Model 1 fluxes were consistent in magnitude, albeit slightly lower than, independent eddy covariance measurements conducted at the same site during the extreme 2015 El Niño event⁵⁰. These measurements reported midday canopy-scale isoprene fluxes of ~6–8 mg m⁻² h⁻¹ under elevated temperature and drought conditions, which likely enhanced thermal and radiative emission drivers relative to those in our study. This agreement in order of magnitude supports the interpretation that Model 1 provides realistic flux estimates under non-extreme conditions. Accordingly, our results indicate that combining observed emission factors with leaf turnover strategy-informed light and temperature response parameters yields more realistic canopy-scale isoprene flux estimates than default plant functional type-based MEGAN parameterizations for the central Amazon Forest.

From a global perspective, this response contrasts with observations from higher-latitude ecosystems. In boreal systems, eddy covariance measurements over a fen showed maximum fluxes of ~46.8 mg m⁻² h⁻¹ under warm conditions and indicated that MEGAN underestimated observed fluxes by more than a factor of nine⁸⁰. Similarly, tundra measurements showed midday fluxes of ~0.5–1.5 mg m⁻² h⁻¹, while MEGAN simulations predicted ~0.07–0.5 mg m⁻² h⁻¹⁸¹. Together, these results demonstrate systematic, biome-specific biases in MEGAN parameterization, with underestimation in boreal and tundra ecosystems and overestimation in the central Amazon Forest. These comparisons highlight the importance of leaf-level measurements for constraining emission factors and response parameters, thereby improving canopy-scale flux estimates and reducing uncertainty in global isoprene budgets.

Conducting field experiments in remote and isolated areas of the Amazon Forest is inherently challenging, particularly when sampling from very tall trees (>30 m). These logistical constraints often limit the number of available observations and contribute to high data variability. Such limitations become especially pronounced when numerous variables are investigated simultaneously, impacting statistical significance. Nevertheless, despite these constraints, our results revealed significant relationships and clear trends consistent with physiological expectations. These findings establish an important foundation for future studies exploring ecophysiological differences between isoprene emitters and non-emitters across leaf turnover strategies and emphasize the importance of linking leaf-level observations to broader biogeochemical processes.

In sum, our findings provide robust empirical evidence that light and temperature responses of volatile isoprenoid emissions vary not only with species physiology but also with leaf turnover strategy—a dimension largely overlooked for tropical tree species in current isoprene emission models⁵¹. We show that default parameterization based on coarse plant functional types misrepresents emission dynamics, particularly for the predominant evergreen canopy trees in the central Amazon Forest. We further show that brevideciduous trees may employ a coordinated chemical defense strategy involving isoprene, highly reactive monoterpenes, and sesquiterpenes.

As climate change continues to alter thermal and hydrological regimes across the Amazon^82,83,84, these shifts in emission chemistry and canopy phenology could drive changes in atmospheric reactivity and feedback processes³⁹. Given that the Amazon Forest harbors the largest area of tropical forest globally² and is the largest contributor to global volatile isoprenoid fluxes^34,50, understanding these emission dynamics is critical. Our study, therefore, lays a foundation for incorporating ecophysiological nuance into global Biogenic Volatile Organic Compound models and highlights the urgent need for gridded, trait-based parameter maps rooted in leaf-level measurements. Although very challenging, this approach will be critical for improving the accuracy of predictions related to atmospheric composition, biosphere resilience, and climate-vegetation feedbacks in tropical forests.

Study site and experimental design

Data were collected at an upland forest (locally called terra firme) permanent plot located in the Amazon Tall Tower Observatory (ATTO) site in the central Amazon Forest region. The ATTO site is located about 150 km northeast of Manaus (02° 08.9′ S, 59° 00.2′ W), at the Uatumã Sustainable Development Reserve. The climate is humid tropical, with a mean annual temperature of 26.7 °C and precipitation of 2376 mm, characterized by a pronounced wet season between December and May, and a drier season between July and October, with transitions in between⁵. Mean temperatures are around 25.6 °C during the wet season and 27.1 °C during the dry season⁵⁰. Vegetation in the terra firme plot is dense (leaf area index of 5.3 m² m⁻²), mature, and non-flooded, with a mean canopy height of 35 m⁵⁰. The soil is a highly weathered and well-drained ferralsol⁸⁵. For more details on the experimental site, see ref. ⁸⁶

We sampled trees and performed measurements from November 29 to December 6, 2022. This period corresponds to the beginning of the wet season, when canopies are dominated by mature leaves^50,60, and variation in leaf age is expected to be low. We measured light and temperature response of VI emissions and leaf gas exchange and chlorophyll fluorescence characteristics, including: assimilation rate (A_n), stomatal conductance (g_s), transpiration (E), leaf-air vapor pressure deficit (VPD), temperature optimum for photosynthesis (T_opt), light saturation point (LSP), photosynthetic capacity (A_sat), electron transport rate (J), quantum efficiency of photosystem II (ϕ_PSII), and photochemical quenching (qP). Measurements were conducted on 12 selected trees representing 12 angiosperm species (one tree per species, Table 2), including six isoprene emitters, and six non-emitters. Each group contained three evergreen and three brevideciduous trees. Detailed leaf turnover strategy classifications and isoprene emission factor measurements are described in previous work⁵⁶.

All trees occupied the upper canopy layer of the plot. Due to logistical challenges of measuring intact leaves from trees exceeding 30 m in height, we performed all measurements on leaf samples from cut branches immediately placed in unfiltered, room-temperature tap water that was sourced from a well in the field and therefore consisted of clean groundwater. A tree climber collected a branch at least 2 cm in diameter from a sun-exposed area of the upper canopy. After collecting, the branch was immediately cut underwater to prevent embolism, stored in a water bottle for transport, and re-cut at the field camp before VI emission and gas exchange measurements. The interval between cutting and leaf enclosure was short and did not exceed 15 min.

This method provides a practical solution for conducting VI emission and gas exchange measurements without compromising leaf viability^{42,58,87,88,89,90,91}, and numerous studies have shown that it does not significantly compromise measurements. For instance, terpene and lipoxygenase-derived compound emissions from distant foliage were found to remain stable for several hours after branch cutting^92,93. Likewise, with the exception of a small amount of acetaldehyde, no other BVOC emissions were observed from broadleaf plant species when mechanical wounding occurred at a location remote from the site of measurement, such as in the case of branch cutting⁹⁴. Furthermore, no significant differences were observed between measurements performed in cut and intact branches, with assimilation rates of 15–24 and 12–18 μmol m⁻² s⁻¹, and isoprene emissions of 50–60 and 40–50 nmol m⁻² s⁻¹, respectively⁹⁵. These findings collectively validate the use of cut branches as a reliable proxy for in situ canopy measurements.

Measurements of light and temperature response curves of photochemical activity traits and volatile isoprenoid emissions

For each tree, we selected one visibly mature and healthy leaf of the branch to measure the responses of photochemical activity traits (Methods Section “Photochemical activity traits”) and VI emissions (Methods Section “Identification and quantification of volatile isoprenoid emissions”) to changes in light and temperature. Light response curves were conducted under both ambient (21%) and reduced (2%) O₂ conditions, the latter used to suppress photorespiration. Reduced O₂ was achieved by injecting a controlled mix of ultrahigh-purity N₂ at a rate of 450 ml min⁻¹ and ambient air at 50 ml min⁻¹ into the leaf chamber. We performed gas exchange and chlorophyll fluorescence measurements with a LI-6800 portable gas exchange system (LiCor Inc., USA). A LI-6400XT was used only for light response curves of C. decandra due to logistical constraints in equipment availability.

At the beginning of each measurement day, a chamber blank sample was obtained from the empty leaf chamber to characterize background concentrations, which were subtracted from all subsequently quantified VI emissions. To minimize possible carryover of semi-volatile sesquiterpenes between trees, we measured all response curves in a fixed sequence for each tree: ambient O₂ light response, reduced O₂ light response, and temperature response. Before starting measurements on a new tree, the leaf chamber was flushed with ambient air for at least 20 min. This procedure, together with the two light curves performed prior to each temperature curve, ensured any sesquiterpenes that possibly adhered to the leaf chamber during the previous tree’s temperature measurements were effectively removed, preventing cross-contamination. As sesquiterpene emissions are not light-dependent, no additional flushing was required between light and temperature response curves.

During temperature response curves, chamber relative humidity gradually decreased as leaf-air VPD increased (Supplementary Fig. 11). Sesquiterpene emissions increased smoothly with temperature, without short-lived emission bursts indicative of storage pool mobilization from enclosure-induced stress (Supplementary Fig. 12A). All measurements were conducted using the 2 cm² leaf chamber and at a flow rate of 500 μmol s⁻¹, corresponding to multiple chamber air exchanges per minute, ensuring rapid flushing and minimizing accumulation or memory effects. Empty-chamber blank measurements showed consistently low sesquiterpene concentrations and no evidence of accumulation or memory effects in the chamber across temperature steps (Supplementary Fig. 12B).

Before each response curve sequence, we separately enclosed the leaf (for compound leaves, we considered a leaflet as the equivalent of a simple leaf lamina) in the leaf chamber with the following environmental conditions: photosynthetic photon flux density (PPFD) of 1000 μmol m⁻² s⁻¹, leaf temperature of 30 °C, CO₂ and H₂O concentrations of 420 μmol mol⁻¹ and 21 mmol mol⁻¹, respectively, and relative humidity of ~60%. The CO₂ concentration was set to approximate ambient atmospheric levels at the time of measurement, while H₂O concentration followed the standard configuration used in comparable gas exchange protocols. Measurements began after acclimating the leaf to these conditions for at least 20 min, or until assimilation rate (A_n), stomatal conductance (g_s), and internal CO₂ concentration (C_i) reached a stable, positive plateau. If stability was not reached, the leaf was replaced, or a new branch was sampled.

Ambient and reduced O₂ light response curves were both performed under decreasing light intensity steps: 2000, 1500, 1000, 750, 500, 250, 100, 50, and 0 μmol m⁻² s⁻¹, while leaf temperature, CO₂, and relative humidity were fixed at the standard conditions defined at the start of the measurements. At each light step, we measured gas exchange characteristics every 30 s for 2.5 min. Temperature curves were performed under increasing leaf temperature steps: 30, 35, 37.5, 40, 42.5, and 45 °C, while PPFD, CO₂, and relative humidity were fixed at the standard conditions defined at the start of the measurements. At each temperature step, we measured gas exchange characteristics every 30 s for 5 min.

We used a leaf chamber fluorometer to simultaneously quantify chlorophyll fluorescence variables during all response curves. At the final logged data point of each successive light and temperature step, an actinic light pulse of 10,000 μmol m⁻² s⁻¹ (10% blue light and 90% red light), modulated at 20 kHz, was applied for 1 s, and steady-state (F_s), light-adapted maximal (F_m’) and light-adapted minimal (F_o’) fluorescence yields were recorded. After measuring all response curves, the leaves were scanned with a table scanner to obtain leaf area, dried in an oven at 60 °C for 72 h, and weighed to obtain leaf dry mass. We analyzed the images of scanned leaves with ImageJ software⁹⁶ to obtain leaf area, and calculated specific leaf area (SLA) as the ratio of leaf area to leaf dry mass.

Photochemical activity traits

ϕ_PSII (Eq. 1) and qP (Eq. 2) were calculated as⁹⁷:

Ambient and reduced O₂ light response curves were fitted to a hyperbolic function⁹⁸ (Supplementary Figs. 13 and 14). Ambient O₂ curves were used to define LSP and estimate A_sat of each tree. LSP was defined as the first PPFD (above the inflection point) where the rate of increase in mean A_n (i.e., the derivative of the logistic function) fell below 5% of maximum A_n. A_sat was estimated as the predicted A_n at the LSP.

For each tree, we calibrated J (µmol e⁻ m² s⁻¹) as:

where the lumped parameter k is the slope of the linear regression between A_n and PPFD × ϕ_PSII/4 at 2% O₂ (Supplementary Fig. 15) derived from the linear part of the light response curve (just before the inflection point)^99,100. Due to limitations in the number of observations in low-light intensities (0–100 µmol m⁻² s⁻¹), the linear regression was performed with values obtained from fitted hyperbolic curves.

A quadratic function was fitted to the relationship between A_n and leaf temperature for each tree (Supplementary Fig. 16)¹⁰¹, and T_opt (°C) was estimated based on the fitted model parameters. Since observed data for Peltogyne catingae and Scleronema micranthum did not follow a quadratic function, their T_opt was determined as the leaf temperature where the highest A_n was observed. The estimated T_opt for C. decandra and Manilkara bidentata fell before the first observed data point.

Identification and quantification of volatile isoprenoid emissions

Air exiting the leaf chamber of the gas analyzer was redirected to a proton transfer reaction-quadrupole mass spectrometer (PTR-QMS, IONICON Analytik, Innsbruck, Austria) to obtain real-time VI emission measurements under controlled leaf chamber conditions. The PTR-QMS operated in standard conditions with a drift tube voltage of 600 V, drift tube pressure of 2.2 mbar, and E/N 120 Td. A hydrocarbon filter (Restek Pure Chromatography, Restek Corporation, USA) was installed at the air inlet of the gas analyzer to remove VIs from incoming ambient air. All tubing in contact with the sampling air were made of PTFE, a material inert to VIs.

The flow rate of air going inside the PTR-QMS was 200 ml min⁻¹. Measurements were performed for 2.5 min and 5 min at each light and temperature step, respectively. During each PTR-QMS measurement cycle, the following mass-to-charge ratios (m/z) were monitored: 21 (H₃¹⁸O⁺), 32 (O₂⁺), and 37 (H₂O-H₃O⁺) with a dwell time of 500 ms each; 41 (isoprene fragment), 69 (isoprene), 81 (monoterpene fragment), 137 (monoterpenes), 149 (sesquiterpene fragment) and 205 (sesquiterpenes) with a dwell time of 1 s each. Humidity-dependent calibrations (using water-bubbled N₂ to dilute standard gas, simulating ambient relative humidity) were performed with a cylinder of certified standard gases provided by Apel-Riemer Environmental, Inc. (Supplementary Table 7) at the beginning and end of the measurement campaign. Mixing ratios of VIs were calculated from the calibration curves (R² ≈ 0.99). PTR-QMS detection limits were calculated as three times the standard deviation of isoprene, total monoterpenes, and total sesquiterpenes (ppb) detected in the water-bubbled N₂ background of the calibration curves and were equal to 0.93, 2.14, and 2.83 ppb, respectively. Mixing ratios of isoprene, total monoterpenes, and total sesquiterpenes (ppb) were used to calculate emission rates per unit leaf area as:

where F is the leaf-level VI emission rate per area (nmol m⁻² s⁻¹); R_ppb is the VI concentration of the outgoing air (ppb); Q is the flow rate of air into the leaf chamber (500 μmol s⁻¹); and S is the leaf area within the chamber (0.0002 m²).

Since individual monoterpenes and sesquiterpenes cannot be quantified with the PTR-QMS, air exiting the leaf chamber during the temperature curves was also routed to fill adsorbent cartridges (stainless steel tubes filled with Tenax TA and Carbograph 5 TD adsorbents) at a rate of 200 ml min⁻¹ for 5 min. This resulted in the collection of VIs from 1 L chamber air for individual monoterpene and sesquiterpene identification and quantification in the lab. VIs accumulated in the adsorbent cartridges were determined by gas chromatography—time-of-flight–mass spectrometry (GC–ToF–MS), at the Atmospheric Chemistry Department of the Max Planck Institute for Chemistry (Mainz, Germany).

Sample desorption of the VIs accumulated in adsorbent cartridges was achieved with a two-stage automated thermal desorber (TD100-xr, MARKES International, UK), with He 5.0 as the carrier gas. Adsorbent cartridges were purged with carrier gas for 5 min at a flow of 50 ml min⁻¹, followed by sample desorption at 250 °C at a flow of 50 ml min⁻¹ for 5 min onto a focusing cold trap (materials emissions, MARKES International, UK) for pre-concentration at 30 °C. The cold trap was purged with carrier gas for 1 min with a flow of 50 ml min⁻¹, then rapidly heated to 250 °C. The sample was removed from the cold trap at a flow of 2 ml min⁻¹ and injected into the GC column. The sampled compounds were separated using a 60 m DB-1 column (0.25 mm internal diameter, film thickness 1 μm, Agilent Technologies, UK). The temperature program used was as follows: 50–150 °C at 4 °C min⁻¹, and 150–250 °C at 8 °C min⁻¹, the temperature was then held for 5 min. The column flow was set to 2 ml min⁻¹. Detection was achieved using a time-of-flight—mass spectrometer (Bench TOF-Select, MARKES International, UK).

All compounds were identified using the NIST library and headspace tests from liquid standards, when available. All monoterpenes were quantified using compound-specific calibration factors, except for cis-β-ocimene, which was quantified using the trans-β-ocimene calibration factor, and eucalyptol, which was quantified using the α-pinene calibration factor. Sesquiterpenes were initially quantified using the calibration factor of α-pinene and subsequently back-calibrated using relative response factors derived from the ratio of the calibration slopes of α-pinene and the individual sesquiterpene liquid standards. When a liquid standard was unavailable, an average response factor derived from other sesquiterpenes was applied. Full calibration details and representative GC–ToF–MS chromatograms are presented in the Supplementary Table 8 and Supplementary Figs. 17 and 18, respectively.

VI emission light and temperature response models

We modeled light responses of isoprene and light-dependent monoterpenes as⁵⁴:

where the emission factor ε₀ is the observed leaf-level emission rate at standard conditions (PPFD of 1000 µmol m⁻² s⁻¹ and leaf temperature of 30 °C), L is incident PPFD (µmol m⁻² s⁻¹), and C_p5 = 1.0¹⁰². We estimated empirical coefficients α and C_L1 based on nonlinear least-square fits to the observed data (Supplementary Fig. 9). Isoprene LSP was defined as the first PPFD where the rate of increase in mean isoprene emission rates fell below 5% of maximum isoprene emission rates.

We modeled the temperature response of normalized isoprene emissions as⁵⁴:

where

R is the gas constant (0.008314 kJ K⁻¹ mol⁻¹), T_opt (K) is leaf temperature at the highest observed isoprene emission rate (E_opt), and T_L (K) is leaf temperature. We estimated empirical coefficients C_T1 and C_T2 based on nonlinear least-square fits to the observed data (Supplementary Fig. 19).

We estimated β parameters of total and individual monoterpene and sesquiterpene temperature responses as⁶²:

where E_T1 and E_T2 correspond to monoterpene and sesquiterpene emission rates at leaf temperatures T₁ = 30 °C and T₂ = 45 °C.

We estimated isoprene fluxes from 175 canopy-dominant trees in the upland forest plot at 09:00, 13:00, and 17:00 for the months of October–December/2022. Average air temperature (°C) and PPFD (µmol m⁻² s⁻¹) values were obtained at 36 m above the ground, which corresponds to the plot’s average canopy height (Supplementary Table 9). We compared fluxes estimated with light and temperature response parameters (α, C_L1, C_T1, C_T2) derived from this study against fluxes estimated with default parameters⁵⁴. Isoprene fluxes were estimated by multiplying γ_{P, i} (Eq. 5) and γ_{T, i} (Eq. 6).

Following that, we compared canopy-level isoprene fluxes estimated for the same time period with: observed leaf-level emission factors^55,56 and parameters derived from this study (Model 1); observed leaf-level emission factors^55,56 and default parameters⁵⁴ (Model 2); and default emission factors³⁴ and default parameters⁵⁴ (Model 3). Fluxes were estimated using the Model of Emissions of Gases and Aerosols from Nature (MEGAN) v2.1 (Eq. 9)³⁴:

where isoprene fluxes (γ_i) are modeled as a function of a canopy environmental coefficient (C_CE = 0.57), canopy leaf area index (LAI = 5.3 m² m⁻²⁵⁰), and light (γ_{P, i}), temperature (γ_{T, i}), leaf age (γ_{A, i}), soil moisture (γ_{SM, i}), and environmental CO₂ concentration (γ_{C, i}) algorithms. While newer updates of this model are available (MEGAN3⁵¹), core parameterizations used here did not change from the previous version (MEGAN v2.1³⁴). For each month, we assumed γ_{SM, i} and γ_{C, i} were constant and equal to 1. Similarly, we assumed that, during this period, canopies were mostly composed of mature leaves and γ_{A, i} was equal to 1. Seasonal phenology was not explicitly modeled because simulations were designed to isolate trait-based, leaf turnover strategy differences under comparable environmental conditions.

Measurements of post-illumination isoprene emissions

We measured post-illumination isoprene emissions according to Rasulov et al.¹⁰³. Following light and temperature response curves, leaves were again acclimated to standard environmental conditions described in the Methods Section “Measurements of light and temperature response curves of photochemical activity traits and volatile isoprenoid emissions” until isoprene emission rates reached a stable, positive plateau. Once steady-state isoprene emissions were achieved, the light was switched off, and the isoprene emission decay was monitored until reaching zero. The first peak in the dark-decay kinetics corresponds to the rapid consumption of DMADP formed prior to darkening¹⁰³. We estimated DMADP pool sizes for each tree by integrating the first dark decay of isoprene emissions using the Peak Analyzer module from OriginPro 2019b (OriginLab) (Supplementary Fig. 20). We integrated DMADP pool sizes at different moments of the first dark decay and obtained paired values of isoprene emission rate and DMADP pool size. Isoprene synthase activity rate was then estimated as the slope of the linear regression between isoprene emission rates and DMADP pool sizes¹⁰⁴.

Air exiting the gas analyzer leaf chamber was directly redirected to the PTR-QMS. The leaf chamber reaches steady state rapidly after changes in gas concentration (~4 s)¹⁰⁵. Nevertheless, we applied a chamber finite time response correction¹⁰³. Chamber response time was estimated in a separate calibration¹⁰⁶ by injecting a small stream of isoprene standard of known concentration into the empty leaf chamber and monitoring its decay with a PTR–ToF–MS after rapid removal of the source. The resulting chamber-clearing trace characterized the intrinsic response time of the measurement system. Post-illumination isoprene emission signals were corrected by subtracting this chamber-response function from the observed dark-decay traces prior to estimating DMADP pool sizes and isoprene synthase activity rate.

Statistical analyses

Due to a lack of within-species replicates, statistical comparisons were conducted at the group level between isoprene emitters and non-emitters and between leaf turnover strategies (brevideciduous vs. evergreen). To evaluate how carbon partitioning among isoprene, total monoterpenes and total sesquiterpenes varied as a function of leaf temperature across leaf turnover strategies, we calculated an isoprenoid emission metric⁴². Mass-based emission rates (µg C g⁻¹ h⁻¹) of each emitted compound group were multiplied by the number of carbon atoms in their molecule. Carbon-weighted values were then summed and divided by the total sum of the mass-based emission rates. Values close to 5, 10, or 15 indicate that isoprene, monoterpenes, or sesquiterpenes, respectively, are predominant in the emission profile.

To evaluate how VI emissions and carbon partitioning responded to increases in leaf temperature across leaf turnover strategies, we performed general linear regression models of the following response variables (y): isoprene, monoterpene, and sesquiterpene emission rates, and isoprenoid emission metric; varying as a function of leaf temperature (y ~ temp; p (temp)), leaf turnover strategy (y ~ pheno.type; p (pheno.type)), the interaction between leaf temperature and leaf turnover strategy (y ~ temp × pheno.type; p (temp:pheno)), and leaf turnover strategy at each leaf temperature step (T_i) (y at T_i ~ pheno.type; p (T_i)). To assess whether E and leaf VPD explained additional variance in emissions beyond light or temperature effects, we compared baseline linear models including only the primary driver (light or leaf temperature) with extended models additionally including E and VPD as covariates.

To evaluate temperature-driven shifts in monoterpene composition across leaf turnover strategies, we classified individual monoterpenes into groups 1–5 (group 1: cis-/trans-β-ocimene, myrcene; group 2: α-/β-pinene; group 3: sabinene, α-/γ-terpinene, α-phellandrene; group 4: limonene, terpinolene, eucalyptol; group 5: camphene, tricyclene), according to their biosynthetic carbocation intermediates and pathway temperature sensitivities⁴⁸. We summed compound emission rates within each group and normalized these sums by total monoterpene emissions and calculated group-level proportional contributions (% of total monoterpenes). We then performed general linear regression models with group-level proportions as the response variable (y) varying as a function of leaf temperature (y ~ temp; p (temp)), leaf turnover strategy (y ~ pheno.type; p (pheno.type)), and the interaction between leaf temperature and leaf turnover strategy (y ~ temp × pheno.type; p (temp:pheno)).

To estimate percentages of photosynthetic carbon (%C) loss to VI emissions in each species, we divided observed mass-based emission rates (µg C g⁻¹ h⁻¹) of each compound group—adjusted for the biosynthetic carbon cost (i.e., multiplied by 1.2 to account for 6/5 for isoprene, 12/10 for monoterpenes, and 18/15 for sesquiterpenes) —by corresponding observed mass-based assimilation rates (µg C g⁻¹ h⁻¹), multiplied by 100. This correction accounts for the fact that VIs are synthesized from C₅ precursors via the 2-C-methyl-D-erythritol 4-phosphate (MEP) pathway in chloroplasts or the mevalonic acid (MVA) pathway in the cytosol, requiring greater photosynthetic carbon input than is retained in the emitted molecule¹⁰⁷.

To evaluate whether photochemical activity traits differed between isoprene emitters and non-emitters across leaf turnover strategies, we performed general linear regression models of the following response variables (y): LSP, A_sat, J, A_n at 45 °C, qP, ϕ_PSII, g_s at A_sat; varying as a function of isoprene emission (emitter, non-emitter; y ~ isoprene; p (isoprene)), leaf turnover strategy (y ~ pheno.type; p (pheno.type)), and the interaction between isoprene emission and leaf turnover strategy (y ~ isoprene × pheno.type; p (isop:pheno)).

To evaluate how photochemical activity traits changed with leaf temperature in isoprene emitters and non-emitters across leaf turnover strategies, we modeled the temperature responses of J, qP, ϕ_PSII, and A_n using a non-linear regression framework. We fitted a peaked Arrhenius temperature response model to each variable:

where T is leaf temperature (K), k₂₅ is the modeled baseline rate at 25 °C, E_α is the activation energy (J mol⁻¹), ΔS is the entropy term (J mol⁻¹ K⁻¹), Hd is the deactivation energy (J mol⁻¹), T_ref = 298.15 K (25 °C), and R is the universal gas constant. While our temperature ramp started at 30 °C, we used 25 °C as a reference since k₂₅ is a standardized model-derived parameter widely used to compare physiological baseline performance across studies and conditions¹⁰⁸. The model was fitted separately for each combination of isoprene emission (emitter vs. non-emitter) and leaf turnover strategy (brevideciduous vs. evergreen). To account for within-group variability and obtain confidence intervals (CIs) for parameter estimates, we performed bootstrap resampling (n = 300) within each group and recorded the fitted parameter (k₂₅, E_α, ΔS, and Hd) distributions. From bootstrapped fits, we calculated and compared means and 95% CIs for each parameter and group.

To evaluate whether isoprene light (α and C_L1) and temperature (C_T1 and C_T2) response parameters derived from this study, observed isoprene T_opt, LSP, DMADP pool size, and isoprene synthase activity rate varied between leaf turnover strategies, we performed general linear regression models of each response variable (y) varying as a function of leaf turnover strategy (y ~ pheno.type). Lastly, we performed Kruskal–Wallis pairwise comparisons between isoprene fluxes estimated with parameters derived from this study and default parameters⁵⁴ (see Methods Section “VI emission light and temperature responses models”). All statistical analyses were performed in Python 3 (Python Software Foundation, 2023) with Jupyter Notebook as the primary environment. The Python libraries pandas, NumPy, and SciPy, were used for data handling, fitting, and model construction; math for mathematical equations; statsmodels for general linear regression models; and matplotlib and seaborn for visualizing results.