The field measurement site is located in Heshan, Guangdong Province, southern China, near the Heshan National Field Research Station (Heshan-NFRS) of Forest Ecosystems (22.683°N, 112.900°E). The E. urophylla forests selected for this study covered an area of more than 25 hm², and were planted and completely natural growth. All individuals in the forests were similar in overall growth, with an age of about 2 years and an average height of about 12–15 m. In-situ measurements were carried out in September 2022 (Sep. 2022), March 2023 (Mar. 2023), and October 2023 (Oct. 2023), and more detailed information on sampling sites can be found in the previous study (Zeng et al., 2024).

The CO₂ response curves for isoprene emissions from E. urophylla were measured using a portable photosynthesis system (LI-6800, Li-Cor Biosciences, Lincoln, NE, United States). Three trees were randomly selected for each experiment, and one healthy, sunlit and mature leaf per tree was randomly selected for the control experiments, all leaves being 2–2.5 m above the ground. The CO₂ concentration gradients were set at 50, 100, 200, 300, 400, 500, 600, 800, 1,000 ppm, with other environmental factors such as temperature, light (Photosynthetic active radiation, PAR), and relative humidity carefully maintained at 30°C, 1,000 μmol m⁻² s⁻¹, and 55%, respectively. We used a standard leaf chamber with an area of 6 cm² and the circulating air inlet flow rate was set at 500 μmol s⁻¹ (∼735 mL min⁻¹). The recycle gas passed through an active charcoal VOCs-scrubber and Na₂S₂O₃ (Sodium thiosulfate) column to remove BVOCs and O₃ before entering the instrument. The gas from the leaf chamber was divided into two paths, one of which (flow rate is 200 μmol m⁻²s⁻¹) passed through an infrared gas analyzer (IRGA), and the other (flow rate is 300 μmol m⁻²s⁻¹) flowed out of the sampling port, and thus could be sampled from the outlet of the chamber. Sampling was initiated 5 min after leaf photosynthesis had fully stabilized, and BVOCs samples were collected using an adsorbent cartridge bedded with Tenax TA and Carbograph 5TD (C2-AAXX-5149, Marks International Ltd., United Kingdom) combined with a portable dual-channel sampler (ZC-QL, Zhejiang Hengda Instrumentation Ltd., China). The sampling flow rate is 200 mL min^-1 with duration of 5 min, resulting in a sampling volume of 1 L. The collected adsorbent cartridge samples were meticulously stored and analyzed within 1 week, ensuring the thoroughness of our data collection process.

Adsorbent cartridge samples were analyzed using an automated thermal desorption-gas chromatography/mass spectrometry system (TD-GC/MS, Markes TD-100/Agilent 5975/Agilent 7890). The TD-100 thermally desorbed the adsorbent tubes at 280°C for 10 min, and then the desorbed analytes were transferred by high-purity helium through a transfer line at 180°C to a cryogenic cold trap at −10 °C (U-T11PGC-2S, Markes International Ltd., United Kingdom). The cold trap was then rapidly heated to 320°C, and the high-temperature desorbed analytes were carried by helium at a flow rate of 1 mL min⁻¹ into an HP-5MS capillary column (30 m × 0.25 mm × 0.25 μm, Agilent Technologies, Inc., CA, United States) for separation and then analyzed by mass selective detector (MSD, model 5975, Agilent Technologies, Inc., CA, United States). The initial gas chromatography (GC, model 7890, Agilent Technologies, Inc., CA, United States) oven temperature was 35°C (3 min hold), then increased to 100°C at 5°C min⁻¹ and hold for 1 min, then to 120°C at 10°C min⁻¹ and hold for 12 min, and then to 260°C at 20°C min⁻¹ with a final hold for 2 min. The MSD was operated in both scan mode with electron impacting ionization at 70 eV. Target compounds were identified by their retention times and mass spectra and quantified with the standard calibration curves. More details on the identification and quantification methods were described in the previous articles of our group (Zeng et al., 2022a; 2022b; 2024).

The isoprene emission rate (E, nmol m⁻² s⁻¹) can be calculated as (Equation 1): E = F C o u t   ‐ C i n ⋅ 10 ‐ 6 S ⋅ 10 ‐ 4 (1)

Where F (μmol s⁻¹) is the flow rate of circulating air, here F = 500 μmol s^-1; C _out (nmol mol⁻¹) and C _in (nmol mol⁻¹) are the isoprene volume mixing ratio at the outlet and inlet air, respectively; S (cm²) is the area of the leaf chamber, which was 6 cm²; 10⁻⁶ is the conversion factor from μmol to mol; 10⁻⁴ is the conversion factor from cm² to m².

In order to fit the ISOP-CO₂ response curves using the model algorithm, the emission rates measured at different concentration conditions need to be normalized to 400 ppm measurements (Equation 2), respectively, to obtain normalized isoprene-CO₂ response curves: γ = E C O 2 E 400 (2)

Where E _CO2 is the actual emission rate at different CO₂ concentrations and E ₄₀₀ is the actual emission rate at 400 ppm CO₂ concentration; γ is the activity factor. The normalized response curves were fitted using the CO₂ response algorithm (Equation 3) in MEGAN v2.1: γ c , i = I s , ⁡ max ‐ I s , ⁡ max × 0.7 C a h C q h + 0.7 C a h (3)

Where I _s,max is the maximum standardized isoprene emission rate predicted by the algorithm and is unitless; γ _c,i is the activity factor of isoprene emission; C _a is the CO₂ concentration (ppm); h and C _q are unitless empirical parameters, and the initial values of the fit are the default values of the MEGAN model, specifically, I _s,max = 1.072; h = 1.7; C ^q = 1,218 (Wilkinson et al., 2009).

The ISOP-CO₂ response curves for E. urophylla were obtained through in situ control experiments on subtropical Eucalyptus across different seasons. Figure 1 shows the ISOP-CO₂ response curves for three different months, along with the parameter fits derived from the data. The variations in these parameters explain the differences in the peak points and subsequent downward trends of the curves. Although the E. urophylla individuals measured each month were randomly selected, the isoprene emission trends in response to CO₂ concentration were consistent among different individuals within the same month. This suggests that the ISOP-CO₂ response pattern is consistent among individuals of the same species during the same month. As illustrated in Figure 1, isoprene emissions peaked as CO₂ concentrations increased from 50 ppm to approximately 200–300 ppm and then decreased, indicating that isoprene emissions from tropical/subtropical E. urophylla are suppressed by high levels of CO₂ concentration. The similar suppress phenomenon were also observed for temperate plants (Rosenstiel et al., 2003; Heald et al., 2009; Potosnak et al., 2014; Sharkey and Monson, 2014). While this suppression is common among tree species, there are significant differences in the degree of suppression reported in different studies, indicating species-specific differences in the ISOP-CO₂ response curves (Niinemets et al., 2021).

Although the extent of isoprene emission suppression by high CO₂ concentrations varied significantly between months, the overall pattern of the ISOP-CO₂ response curve remained relatively consistent within each month. The seasonal variations of ISOP-CO₂ responses showed that relatively stronger suppression in cooler seasons and weaker suppression in warmer seasons. Similar seasonal variations were also found from temperate trees (Monson et al., 2016), and the growth temperature were supposed to contribute to the observed seasonal changes in ISOP-CO₂ response (Potosnak et al., 2014; Lantz et al., 2019b; Sahu et al., 2023). Laboratory experiments on poplar have similarly found that elevated temperatures may counteract this inhibitory effect (Potosnak et al., 2014; McClain and Sharkey, 2019; Sahu et al., 2023). Similarly, from our field experiments conducted in Sep. 2022, Mar. 2023, and Oct. 2023, activity factor of isoprene emission (devoted as γ in Figure 1) decreased by 10%, 55%, and 39% on average in Sep. 2022, Mar. 2023, and Oct. 2023, respectively, at 800 ppm CO₂ when compared to the emissions at 400 ppm CO₂, with daytime mean temperatures of 32.1°C, 22.9°C, and 28.7°C, respectively. These findings align with other studies suggesting that growth temperature is a potential factor driving the seasonal changes in the ISOP-CO₂ response and demonstrating a synergistic effect of temperature and CO₂ on isoprene emissions. However, the effects of temperature and CO₂ on isoprene emissions are currently treated as independent modules in the MEGAN model (Guenther et al., 2012). As global temperatures and CO₂ concentrations increase together, incorporating the synergistic effects of temperature and CO₂ into the model will be crucial for accurately simulating and understanding the future impacts of climate change on isoprene emissions.

To further investigate the coupling between temperature and ISOP-CO₂ response, we linearly fitted the ISOP-CO₂ response curves within this CO₂ concentration range (e.g., 400–1,000 ppm), using the slope (k value) to characterize the degree of suppression, and explored the seasonal variations of the k value in relation to different temperature. The study on subtropical Eucalyptus demonstrates that the seasonality of the ISOP-CO₂ response curve is correlated with growth temperature. The linear fitting results revealed that the k values of the ISOP-CO₂ response curves for Sep. 2022, Mar. 2023, and Oct. 2023 were −2.4 × 10⁻⁴, −11.0 × 10⁻⁴, and −8.0 × 10⁻⁴, respectively (Supplementary Figure S1). This correlation may arise from changes in the sensitivity of isoprene emissions to CO₂ as a result of plant adaptation to growth temperature or modifications in isoprene emissions (Niinemets and Sun, 2015; Lantz et al., 2019a; 2019b; Sahu et al., 2023).

Previous studies have shown that isoprene emission factors are influenced by the average temperature over the past 1–10 days, indicating an adaptive mechanism of isoprene emission to growth temperature (Sharkey et al., 1999; Rapparini et al., 2004; Guenther et al., 2012). We selected the average daily mean temperature (TEMP_mean) and the average daily maximum temperature (TEMP_max) for the past 1-, 5-, and 10-day, respectively, as indicators of growth temperature. The temperature data were obtained from the Heshan National Field Research Station of Forest Ecosystems (http://hsf.cern.ac.cn/). As shown in Figure 2, a significant correlation (p < 0.05) between the k values and temperature with correlation coefficient (r²) values ranging from 0.85 to 0.96 were obtained. The overall correlation between the k value and the TEMP_max was slightly higher than with the TEMP_mean. Additionally, the correlation between k values and the temperature over the past 10-day was generally higher than in the past 5- and 1-day. Moreover, the r² values between the k value and the TEMP_mean and TEMP_max over the past 10-days exceeded 0.90 (0.93 and 0.96, respectively), suggesting that the past 10 days temperature can be used as indicators for predicting the seasonal pattern of the ISOP-CO₂ response.

Besides the environment factors, the variation in plant physiological parameters, such as photosynthesis (Pn), stomatal conductance (gs), and electron transfer rate (ETR), that could influence the ISOP-CO₂ response curve were also investigated. As shown in Figure 3, these parameters were expressed as relative changes. The Pn-CO₂ curves for Sep. 2022, Mar. 2023, and Oct. 2023 displayed consistent trends. Photosynthetic activity increased rapidly with CO₂ concentration from a negative value to approximately 1.2 when CO₂ reached 600 ppm, after which it plateaued, indicating a saturation point under high CO₂ conditions (Figure 3A). The gs showed considerable variation at low CO₂ concentrations (<300 ppm), but the gs-CO₂ curves exhibited high consistency as CO₂ concentration increased (Figure 3B). However, the ETR-CO₂ curves followed a similar trend to the Pn-CO₂ curves but with noticeable seasonal changes. Specifically, the ETR-CO₂ curves were consistent across months at lower CO₂ concentrations (<400 ppm), but diverged significantly at higher CO₂ concentrations, stabilizing in the high-temperature month (Sep. 2022) and decreasing in cooler months (Mar. 2023 and Oct. 2023). The ETR-CO₂ curves for Mar. 2023 and Oct. 2023 decreased by 12.4% and 5.0%, respectively, at 800 ppm compared to 400 ppm CO₂ (Figure 3C).

Plant physiological states can be influenced by factors such as PAR, temperature, CO₂ concentration, or biotic stressors (e.g., viruses, pests) (Mu et al., 2022). However, since we chose healthy and sunlight leaves during our experiments, the effects of factors like PAR, viruses, and pests can be disregarded. The consistent Pn-CO₂ curves across the three experiments suggest that photosynthetic activity remained stable despite seasonal variations (Figure 3), likely due to isoprene synthesis utilizing only 1%–2% of the carbon fixed by photosynthesis (Niinemets et al., 2021). The gs also showed consistent trends at high CO₂ concentrations across months, while variations at low CO₂ concentrations were likely due to unstable stress responses affecting stomatal behavior (Monson et al., 2013; Oikawa and Lerdau, 2013; De Souza et al., 2018). These findings align with other studies showing that gs does not significantly impact isoprene emissions (Sahu et al., 2023), implying that gs is not a major factor in the seasonal variation of ISOP-CO₂ curves.

A prevailing hypothesis suggested that the suppression of the ISOP-CO₂ response curve is due to limitations in the availability of key substances such as adenosine triphosphate (ATP) and nicotinamide adenine dinucleotide phosphate (NADPH) (Niinemets et al., 1999; Rosenstiel et al., 2003; Rasulov et al., 2009; 2016; 2018; Morfopoulos et al., 2014). Based on this, the triose phosphate utilization (TPU) limitation hypothesis was proposed, which posits that under high CO₂ concentrations, the rate of increase in photosynthesis outpaces the rate at which glyceraldehyde 3-phosphate (GAP) is utilized by the methylerythritol 4-phosphate (MEP) pathway. This leads to inorganic phosphate (Pi) accumulation in chloroplasts, which is then incorporated into phosphorus-containing sugars, thereby inhibiting ATP synthesis (McClain and Sharkey, 2019). Although Lantz et al. (2019b) argued that TPU limitation is not essential for suppressing isoprene emissions under high CO₂ concentrations, its presence can still influence the degree of suppression. Furthermore, TPU limitation may be alleviated at higher temperatures, as increased respiration rates and upregulated enzyme activity in the MEP pathway facilitate the release of accumulated Pi, thereby removing the TPU limitation (Monson et al., 2016). Evidence for this was observed in our experimental results, where the ETR-CO₂ curves did not exhibit suppression at high CO₂ concentrations during the warm month (Sep. 2022). In contrast, such suppression was evident in cooler months (March and October 2023) (Figure 3C). The more significant decrease in Mar. 2023 compared to Oct. 2023 is likely attributable to the lower average daytime mean temperatures in March (22.9°C vs. 28.7°C, respectively).

The reduction in the ETR-CO₂ curve at high concentrations is considered characteristic of TPU limitation (McClain and Sharkey, 2019). It can be used to infer whether leaves are experiencing this limitation, which rarely occurs at temperatures above 30°C. This suggests that temperature is likely a key factor controlling the seasonality of the ISOP-CO₂ response, as other major influences did not significantly alter the plant’s physiological state. Our study further supports this, showing that this acclimatization mechanism is reflected in the ISOP-CO₂ response curve.

Over time, the correlation between temperature indicators and the k value of the ISOP-CO₂ response curve was higher over the past 10 days compared to the past 5 and 1 days, suggesting that this acclimation mechanism has a delayed effect. This delay may be related to physiological adjustments made by plants in response to temperature changes, such as alterations in membrane permeability, the expression of specific functional genes, and the synthesis of secondary metabolites like isoprenoid compounds (Jagadish et al., 2021; Wang et al., 2024). The slightly higher overall correlation between the k value and TEMP_max compared to TEMP_mean suggests that heat stress may also affect the ISOP-CO₂ response, likely due to the plant’s stress response to sudden temperature changes, which could involve regulating isoprene emission, stomatal conductance, and respiration rate (Li and Sharkey, 2013; Li et al., 2020). Given the strong correlation between the k value with the TEMP_mean and the TEMP_max over the past 10 days (with correlation coefficients larger than 0.90), it is possible to model the relationship between temperature and the ISOP-CO₂ response by further incorporating the effects of temperature on the k value (Figure 2).

The comparison of the ISOP-CO₂ response curves for each month with the model default response curves and found that the model (MEGAN default k = −3.7 × 10⁻⁴) predicts stronger suppression of isoprene emissions at high CO₂ concentrations than what was observed in Sep. 2022, but weaker suppression compared to Mar. 2023 and Oct. 2023 (Supplementary Figure S2). This leads to an underestimation of isoprene emissions during the high-temperature months and an overestimation during the low-temperature months. For example, at a CO₂ concentration of 800 ppm, the model underestimates emissions by 5% in Sep. 2022 but overestimates the emisions by 38% in Mar. 2023 and 21% in Oct. 2023, respectively. Therefore, to lower the uncertainties of isoprene emission simulation, it is important to account for seasonal variations in the ISOP-CO₂ response, such as by incorporating the effect of the TEMP_max over the past 10 days, as suggested above.

Even though the strong suppression of isoprene emissions only occurred under high CO₂ concentrations, not the current state. Under climate change, wildfires’ extent, frequency, intensity, and duration have increased recently, leading to significant spikes in atmospheric CO₂ concentrations and regional temperatures in localized areas. For example, during the Australian wildfires of 2019–2020, CO₂ concentrations in smoke-covered regions increased by 5- to 80-fold (Van Der Velde et al., 2021; Hong et al., 2023). Under such conditions, isoprene emissions from vegetation surrounding the wildfire perimeter may be severely suppressed by high CO₂ concentrations. However, the accompanying increase in temperature could mitigate the extent of this suppression. Our results provide insights into assessing the impacts of isoprene emissions from vegetation in wildfire-affected regions, which is essential for accurately estimating the consequences of wildfire disasters.

Moreover, according to the IPCC report, global average CO₂ concentrations are expected to reach 900 ppm or higher by the end of the 21st century, and global average temperatures could rise by 1°C–4.5°C (IPCC, 2018). Under future conditions, the suppression of isoprene emissions by CO₂ will become more pronounced. However, this study shows that this effect must account for seasonal variability. Our results also suggest that rising global temperatures may reduce the suppression effect of CO₂ on isoprene emissions in the future. However, current models do not include the coupling mechanism between temperature and the ISOP-CO₂ response, which could introduce significant uncertainty into predictions of future isoprene emissions. The current model may exhibit a slight prediction bias (e.g., <10%) during future high-temperature seasons or extreme heatwaves, potentially overestimating emissions during cooler seasons. To more accurately predict future isoprene emissions, it is necessary to incorporate the coupling mechanism between temperature and the ISOP-CO₂ response into the model.

Additionally, many studies have shown that elevated atmospheric CO₂ concentrations fertilize terrestrial vegetation, leading to increased net primary productivity and leaf area index (Piao et al., 2012; Forkel et al., 2016; Chen et al., 2019). This increase may partially offset the suppression of isoprene emissions caused by higher CO₂ levels. Therefore, assessing the impact of elevated CO₂ on future isoprene emissions requires a comprehensive approach that considers multiple factors and coupled mechanisms.