Ozone-vegetation feedback through dry deposition and isoprene emissions in a global chemistry-carbon-climate model

Ozone-vegetation feedback is essential to tropospheric ozone (O3) concentrations. The O3 stomatal uptake damages leaf photosynthesis and stomatal conductance and, in turn, influences O3 dry deposition. Further, O3 directly influences isoprene emissions, an important precursor of O3. The effects of O3 on vegetation further alter local meteorological fields and indirectly influence O3 concentrations. In this study, we apply a fully coupled chemistry-carbon15 climate global model (ModelE2-YIBs) to evaluate changes in O3 concentrations caused by O3–vegetation interactions. Different parameterizations and sensitivities of the effect of O3 damage on photosynthesis, stomatal conductance, and isoprene emissions (IPE) are implemented in the model. The results show that O3-induced inhibition of stomatal conductance increases surface O3 on average by +2.1 (+1.4) ppbv in eastern China, +1.6 (-0.5) ppbv in the eastern U.S., and +1.3 (+1.0) ppbv in western Europe at high (low) damage sensitivity. Such positive feedback is dominated by reduced O3 dry deposition, 20 in addition to the increased temperature and decreased relative humidity from weakened transpiration. Including the effect of O3 damage on IPE slightly reduces surface O3 concentrations by influencing precursors. However, the reduced IPE weakens surface shortwave radiative forcing of secondary organic aerosols leading to increased temperature and O3 concentrations in the eastern U.S. This study highlights the importance of interactions between O3 and vegetation with regard to O3 concentrations and the resultant air quality. 25


Introduction
Tropospheric ozone (O 3 ) is generated by photochemical reactions involving nitrogen oxides (NO x ) and volatile organic compounds (VOCs) under strong solar radiation (Sillman, 1999;Atkinson, 2000;Jacob and Winner, 2009). It is one of the most important air pollutants and has been of widespread concern (Wang et al., 2017;Li et al., 2019). High O 3 concentrations at the surface can not only injure human respiratory health (Gauderman et al., 2004;Lelieveld et al., 2015) but also lead to considerable damage to plants and crops, which further changes the land carbon budget (Fuhrer et al., 1997;Yue and Unger, 2014;Lombardozzi et al., 2015). In turn, vegetation can modulate O 3 concentrations via influencing dry deposition processes, precursor emissions (such as those of isoprene, monoterpene, and sesquiterpene), and meteorological fields. Studying O 3 -vegetation interactions is of great importance to better understand the variations in O 3 concentrations as well as the ecosystem carbon cycle, particularly for regions with high O 3 levels and vegetative cover.
Ground-level O 3 reduces vegetation photosynthesis by stomatal uptake (Fuhrer et al., 1997;Ainsworth et al., al. (2007) showed that the elevated O 3 since the preindustrial period depressed photosynthesis and stomatal conductance of trees by 9 %-13 % and 11 %-15 %, respectively. A recent global meta-analysis on poplar showed that current O 3 concentrations reduced the CO 2 assimilation rate and stomatal conductance by 33 % and 25 %, respectively, compared to that of charcoal-filtered air (Feng et al., 2019a). In model studies, an offline process-based vegetation model (the Yale Interactive Terrestrial Biosphere model, or YIBs) estimated that the present-day effect of O 3 damage reduced gross primary productivity (GPP) by 4 %-8 % on average over the eastern US during the summer (Yue and Unger, 2014) and annual net primary productivity (NPP) by approximately 14 % in China (Yue et al., 2017). Lombardozzi et al. (2015) also showed that the present-day O 3 exposure reduces GPP globally by 8 %-12 % using the Community Land Model (CLM).
Isoprene emissions (IPE) from vegetation can be affected by surface O 3 . Isoprene is the most dominant species among biogenic VOCs (BVOCs) and accounts for approximately one-half of global BVOC emissions (Guenther et al., 2012). The effect of O 3 on IPE is complex. Calfapietra et al. (2009) reviewed observational experiments in Italy and proposed a hypothesis that there might be a detoxification effect resulting from O 3 -IPE interactions. Vegetation under a low accumulated O 3 dose can be simulated to increase the levels of IPE to reduce oxidative damage, but months of O 3 exposure are harmful to metabolism and reduce IPE. Several studies have showed that O 3 fumigation over a short time (days to weeks) but at high concentrations (100-300 ppbv) led to increased IPE (Velikova et al., 2005;Fares et al., 2010), while some other experiments conducted over an entire growing season (at least 3 months) under controlled O 3 concentrations (approximately 80 ppbv) showed that O 3 reduced IPE (Calfapietra et al., 2008;Yuan et al., 2016Yuan et al., , 2017. A recent global meta-analytic review showed that IPE negatively responded to elevated O 3 (91 ppbv on average) by −8 % (Feng et al., 2019b). Overall, consecutive exposure to high O 3 levels has a negative impact on IPE, although there are large uncertainties resulting from vegetation type (Tiiva et al., 2007;Ryan et al., 2009), temperature (Hartikainen et al., 2009), and CO 2 concentration (Calfapietra et al., 2008).
O 3 dry deposition is one of the important sinks of tropospheric O 3 and mainly occurs over vegetation (Wesely, 1989). The stomatal uptake of vegetation plays an important role in this removal process. (Wesely and Hicks, 2000). Val Martin et al. (2014) showed that the O 3 dry deposition velocity in the Community Earth System Model (CESM) significantly increased and was more reasonable when the original scheme (Wesely, 1989), which assumed that stomatal resistance was only related to temperature and water vapor, was replaced with a scheme coupled to vegetation (Collatz et al., 1991;Sellers et al., 1996). In addition, BVOC emissions can change the local NO x /VOC ratio and, in turn, influence O 3 concentrations. For example, Fu and Liao (2012) showed that the interannual variations in BVOCs alone can lead to 2 %-5 % differences in simulated O 3 over China during the summer using the Model of Emissions of Gases and Aerosols from Nature (MEGAN; Guenther et al., 2006) module embedded within the global three-dimensional chemical transport model (GEOS-Chem). Calfapietra et al. (2013) reviewed the role of BVOCs emitted by urban trees in O 3 concentrations in cities and showed that BVOCs generally promoted O 3 formation because of the VOC-limited condition. Furthermore, the modifications of meteorological fields caused by vegetation (Liu et al., 2006;Wu et al., 2011)  In this study, we apply a semi-mechanistic O 3 damage scheme (Sitch et al., 2007)  The dynamic and physical processes are calculated every 30 min. Gas-phase chemistry in the troposphere includes basic NO x -HO x -O x -CO-CH 4 chemistry as well as peroxyacyl nitrates and the following hydrocarbons: terpenes, isoprene, alkyl nitrates, aldehydes, alkenes, and paraffins. Chlorine-containing and bromine-containing compounds, chlorofluorocarbons (CFCs), and N 2 O source gases are all included in the stratospheric gas-phase chemistry. Dry deposition of gases is calculated by using a resistance-in-series scheme, which was updated to include coupling to stomatal resistance (Val Martin et al., 2014). In addition, the model interactively simulates aerosols such as sulfate, nitrate, elemental and organic carbon, sea salt, and dust considering the climate through direct (Koch et al., 2006) and indirect effects (Menon et al., 2008(Menon et al., , 2010 and gas-phase chemistry by affecting photolysis rates (Bian et al., 2003). Meteorological and hydrological variables in this model have been fully validated via observations and a reanalysis dataset (Schmidt et al., 2014). The anthropogenic emission inventory for the present-day (2010) from the IPCC RCP8.5 scenario (van Vuuren et al., 2011) is utilized in this study.
The YIBs model is a dynamic vegetation model that includes nine plant functional types (PFTs; Table S1 in the Supplement) and can simulate biophysical processes of photosynthesis, transpiration, and respiration with variations in meteorological fields. Since the higher leaf photosynthesis requires larger stomatal conductance to allow more CO 2 enter the leaves, leaf photosynthesis and stomatal conductance are closely related and calculated using the Farquhar and Ball-Berry models (Farquhar et al., 1980;Ball et al., 1987) as follows: where the total leaf photosynthesis (A tot ) is the minimum value of the ribulose-1,5-bisphosphate carboxylase-limited (RuBisCO-limited) rate of carboxylation (J c ), light-limited rate (J e ), and export-limited rate (J s ). Stomatal conductance for H 2 O (g s ) is calculated by the A tot , dark respiration rate (R d ), relative humidity (RH), and CO 2 concentration at the leaf surface (c s ). The values of m and b are different for different PFTs (Table S1). A canopy radiation scheme is ap-plied in YIBs to separate diffuse and direct light for sunlit and shaded leaves (Spitters et al., 1986). The LAI and tree growth are dynamically simulated with the allocation of carbon assimilation. The emissions of isoprene are calculated online as a function of J e photosynthesis (Eq. 1), canopy temperature, intercellular CO 2 , and CO 2 compensation point (Arneth et al., 2007;Unger, 2013) and have been fully validated by Unger et al. (2013). Carbon fluxes, phenology, LAI, GPP, and net ecosystem exchange (NEE), as well as other parameters of vegetation in ModelE2-YIBs, have been previously extensively evaluated and agree well with the observations (Yue and Unger, 2015). In addition, ModelE2-YIBs shows good performance in simulating O 3 -vegetation interactions such as O 3 -GPP and O 3 -g s relationships (Yue et al., 2016;Yue et al., 2018).
The O 3 dry deposition velocity (V d ) in ModelE2-YIBs is calculated following the multiple-resistance approach originally described by Wesely (1989): where R a , R b , and R c are the aerodynamic resistance, quasilaminar sublayer resistance above the canopy, and surface resistance, respectively. R c is computed as follows: where R s , R lu , R cl , and R g represent the stomatal resistance, leaf cuticle resistance, lower-canopy resistance, and the ground resistance, respectively. In this study, the original parameterization for R s , which is empirically expressed by solar radiation, surface air temperature, and the molecular diffusivities for water vapor, has been substituted by the reciprocal of g s from Eq.
(2) following Val Martin et al. (2014). In this case, O 3 dry deposition can be interactively influenced by the stomatal O 3 uptake process for vegetation. Isoprene and α-pinene are considered to be the precursors for biogenic secondary organic aerosols (SOAs) in ModelE2-YIBs, which are computed online based on the two-product scheme developed by Chung and Seinfeld (2002). Isoprene can be oxidized by O 3 as follows: Changes for semivolatile product P i (i = 1, 2) at each time step (dt) are calculated by where H is the enthalpy of vaporization and is set as 42.0 kJ mol −1 for isoprene (Chung and Seinfeld, 2002;Henze and Seinfeld, 2006) and 72.9 kJ mol −1 for α-pinene. K sc is the saturation concentrations at the temperature T sc (295 K) and set as 1.62 m 3 µg −1 (0.064 m 3 µg −1 )and 0.0086 m 3 µg −1 (0.0026 m 3 µg −1 ) for the two products formed by oxidation of isoprene (α-pinene), respectively (Presto et al., 2005;Henze and Seinfeld, 2006 (2007) is applied in this study that simulates the effect of O 3 damage to the photosynthesis rate via the following formula: where A totd (g sd ) and A tot (g s ) are the O 3 -affected and original total leaf photosynthesis (stomatal conductance), respectively. F is the ratio between affected and original photosynthesis. It depends on the instantaneous leaf uptake of O 3 as follows: where parameter a represents the O 3 -damaging sensitivity dependent on vegetation types with a range from low to high values. F O 3 ,crit is a critical threshold for damage (Table S1). F O 3 is the O 3 uptake rate by the stomata, which is calculated by where [O 3 ] is the surface O 3 concentrations and R a is the aerodynamic resistance in Eq. (3). k O 3 is 1.67, which is the ratio of leaf resistance for O 3 to leaf resistance for water vapor. This scheme has been utilized in many previous studies, which have reported that O 3 reduces GPP by 4 %-8 % on an annual mean basis in the eastern US and by 10 %-20 % during the summer in China (Yue and Unger, 2014;Yue et al., 2017).

The effect of O 3 damage to IPE
To date, there are no mature parameterizations that calculate the contributions of O 3 damage to IPE. Here, we propose two schemes based on observations to quantify the changes in surface O 3 concentrations resulting from O 3 damage to IPE.
The first scheme assumes that O 3 leads to the same percentage of damage to photosynthesis and IPE because IPE are observed to linearly vary with photosynthesis (Yuan et al., 2016). The affected IPE (IPE d ) can be calculated as follows: where F is calculated by using Eq. (10) and IPE is the original level of IPE. Hereafter, this scheme is termed the "F scheme". Another scheme is based on open-top chamber (OTC) observations. Although many experiments have studied the effects of O 3 on IPE, most have applied a limited range of O 3 levels (e.g., 7.3-56.6 ppbv in Hartikainen et al., 2009, or > 100 ppbv in Fares et al., 2010. In reality, surface O 3 concentrations can vary from several parts per billion by volume (e.g., in the polar region during the winter) to over 100 ppbv (e.g., in megacities of China during the summer).
To date, only one study (Yuan et al., 2017) has explored the responses of IPE to different levels of O 3 damage for two poplar clones; a linear regression between the percentage damage of IPE (PDI) and the cumulative stomatal uptake of O 3 > 1 nmol O 3 m −2 s −1 (POD 1 ) was derived as follows: The POD 1 is calculated by the following formula: where F O 3 is the O 3 uptake rate by stomata (nmol O 3 m −2 s −1 ), which is the same as that in Eq. (11). dt indicates the time integration step, and n indicates the total number of time steps during the growing season. In this study, the POD 1 accumulated over the growth season is defined as April to October north of 23.5 • N (e.g., Tucker et al., 2001;White et al., 2002;Yin et al., 2014;Wang et al., 2019), November to March south of 23.5 • S (e.g., Broich et al., 2015;Moore et al., 2016), and 200 d between 23.5 • N and 23.5 • S because the leaf phenology in tropical evergreen forests is not determined by seasonality (Xiao et al., 2006). Limited by the data availability, we apply the PDI function (Eq. 13) for poplar to all vegetation types as follows: Hereafter, this scheme is termed a "linear scheme." Different from the F scheme, the linear scheme calculates IPE damage using accumulated O 3 instead of instantaneous O 3 concentrations.

Descriptions for sensitivity experiments
Seven experiments (Table 1) where S i and O i are the simulated and observed O 3 concentrations, respectively, and n is the total number of observational sites.
3 Results Figure 1 shows a comparison of the simulated summer O 3 concentrations to the observations. The model reasonably reproduces spatial patterns, with a correlation coefficient of 0.41. The NMBs between simulations and observations in US and Europe are 11.7 % and 13.2 %, respectively, which are comparable with the simulation performed by CESM (Lamarque et al., 2012;Sadiq et al., 2017) (Yue et al., 2017).

CTRL simulation and model evaluation
To further compare the performance of ModelE2-YIBs with other chemistry-climate models, we select six simulated cases performed by different model members in the Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP; Lamarque et al., 2013) and implement the evaluation with the same observational data (Fig. S2 in the Supplement). The correlation coefficient (0.41) and NMB (29.3 %) for ModelE2-YIBs are located in the ranges of 0.36 % to 0.60 % and −16.0 % to 45.1 % by the model ensembles, suggesting that ModelE2-YIBs has comparable performance to other state-of-the-art models. However, most of the current chemistry-climate models lack the interactive vegetation growth module, let alone studying O 3 -vegetation interactions. The vegetation variables (e.g. GPP and LAI) in ModelE2-YIBs have been fully evaluated in previous studies (Yue and Unger, 2015), making ModelE2-YIBs a suitable tool for this work. Figure 2 shows the global June-July-August (JJA) surface O 3 concentrations, O 3 dry deposition velocity, GPP, and IPE. Simulated O 3 is high in the eastern US, western Europe, India, and eastern China (Fig. 2a). The spatial pattern of O 3 dry deposition velocity (Fig. 2b) resembles that of the GPP (Fig. 2c) because the O 3 stomatal uptake dominantly contributed to the dry deposition. Both are high in the eastern US, western Europe, Amazon, eastern China, and Indonesia and show a reasonable magnitude consistent with previous modeling studies (Val Martin et al., 2014;Yue and Unger, 2015;Sadiq et al., 2017). The spatial pattern of IPE (Fig. 2d)  also resembles that of the GPP (Fig. 2c) except that the IPE in Europe are lower than those in other regions. Such discrepancies are likely attributed to the lower fraction of deciduous broadleaf forest, which provides a high yield of IPE (Potter et al., 2001).  (Fig. 3a). However, the F scheme with low sensitivity predicts low damage of ∼ 10 % in these regions (Fig. 3b). At a global scale, IPE decrease by 1.2 %-3.2 % because of the O 3 effect. The damage using the linear scheme is generally within the low-to-high range of predictions by using the F schemes. For the linear scheme, IPE in eastern China show the greatest damage of ∼ 15 %. Figure 4 shows seasonal variations in the effect of O 3 damage to IPE in eastern China, the eastern US, and western Europe. The magnitude of IPE changes is generally within the range of 10 %-29 %, as summarized by the observational meta-analysis (Feng et al., 2019b). The F scheme is dependent on instantaneous O 3 uptake, which peaks during the summer when both surface O 3 and stomatal conductance are high. In contrast, the linear scheme depends on the accumulated O 3 flux, which increases from zero to high levels during the growth season. As shown, the percentage of O 3 damage to IPE is low during April and May but increases to a similar magnitude to that in the F scheme with high sensitivity during August; it reaches a maximum in October. The differences in the F (instantaneous) and linear (accumulated) schemes cause distinct seasonal variations in the IPE damage, which might cause different feedback to the O 3 concentrations. However, the IPE peaks during summer (Fig. S3 in the Supplement), suggesting that absolute changes in IPE are most significant during this season (Fig. S4 in the Supplement). Meanwhile, since the surface O 3 concentrations and the vegetation growth both peak during boreal summer in the Northern Hemisphere, the O 3 -vegetation interactions are supposed to be the strongest in this season. As a result, we focus our analyses on the summer to explore the O 3 -vegetation interactions and feedback.

O 3 -vegetation feedbacks on surface O 3 concentrations
The effect of O 3 damage to stomatal conductance inhibits dry deposition (Fig. S5 in the Supplement), leading to significant increases in summer surface O 3 , particularly in eastern China, Japan, the eastern US, and western Europe ( Fig. 5a  and b). The positive feedback can be greater than 5 ppbv in eastern China with high sensitivity (Fig. 5a). Smaller changes are predicted for low sensitivity, which shows limited perturbations in the US and Japan (Fig. 5b). Including the effect of O 3 damage to both stomatal conductance and IPE maintains the spatial pattern of O 3 changes but occurs at a lower magnitude ( Fig. 5c-f) because these two effects offset each other.
With high damage to stomatal conductance, surface O 3 remains increasing in eastern China, Japan, the eastern US, and western Europe even with reduced IPE (Fig. 5c and e). However, with low damage to stomatal conductance, surface O 3 shows limited changes in Europe, China, and Japan when IPE are simultaneously reduced ( Fig. 5d and f). Surprisingly, sur- face O 3 increases over the eastern US in these cases ( Fig. 5d and f) compared to the limited changes when IPE remain unperturbed (Fig. 5b). Figure 6 summarizes the changes in surface O 3 over sensitive regions. Without IPE feedback, the effect of O 3 damage to stomatal conductance leads to changes in regionally averaged surface O 3 by +2.1 ppbv (+1.2 ppbv) in eastern China, +1.8 ppbv (−0.3 ppbv) in the eastern US, and +1.3 ppbv (+1.0 ppbv) in western Europe for high (low) damage sensitivity (Table 2). Changes in eastern China are the greatest compared to those of the other two regions, mainly because of the high O 3 level (Fig. 1a) and sensitive tree species (the high a and low F O 3 ,crit for deciduous broadleaf forest; Table S1). Surface O 3 is predicted to decrease in the eastern US with the low damage sensitivity, though such a change is not significant over most grids (Fig. 5b). The inclusion of the effect of O 3 damage for both stomatal conductance and IPE slightly weakens the O 3 feedback, leading to changes in O 3 concentrations of +1.5 ppbv (+0.02 ppbv) with the F scheme and +2.0 ppbv (−0.3 ppbv) with the linear scheme in eastern China for high (low) sensitivity. The regional maximum O 3 changes can reach 7.4 ppbv (4.6 ppbv) in eastern China. Further, the effect of O 3 damage to IPE weakens the positive feedback in western Europe by approximately 1-2 ppbv. The average O 3 changes in the eastern US due to high (low) O 3 damage are +1.4 ppbv (+1.6 ppbv) with the F scheme and +1.8 ppbv (+1.1 ppbv) with the linear scheme when IPE feedback is included.
Although damage to stomatal conductance and IPE exert opposite effects, surface O 3 in general increases after including both processes (Fig. 6), suggesting that dry deposition inhibition plays the dominant role. For the same high O 3 damage sensitivity to stomatal conductance, changes in surface O 3 remain similar over eastern China and the eastern US between the F and linear schemes in terms of the responses of the IPE (Table 2). However, responses in western Europe are weaker for the linear scheme (Fig. 5e) compared to that of the F scheme (Fig. 5c), though the former predicts lower reductions in IPE (Fig. 3). Nevertheless, inclusion of IPE reductions helps increase surface O 3 over the eastern US ( Fig. 5d and f vs. Fig. 5b), which is unexpected, since the reduction in IPE is supposed to decrease O 3 concentrations.  These changes are speculated to be indirectly related to O 3vegetation feedback to meteorology and would be further examined in the next section. Figures 7 and 8 show the changes in surface air temperature and relative humidity (RH) between different sensitivity experiments and the CTRL simulation, respectively. When considering the effect of O 3 damage on stomatal conductance alone, eastern China becomes warmer ( Fig. 7a and b) and drier ( Fig. 8a and b), favoring O 3 chemical production and increasing surface O 3 concentrations (Jacob and Winner, 2009). The damaged stomatal conductance weakens leaflevel transpiration and thus reduces the latent heat flux at the surface (Fig. S6 in the Supplement), leading to a higher tem-  Fig. 8c and e vs. 8a and Fig. 8d and f vs. 8b) but significantly increases surface air temperature in the eastern US (as shown in Fig. 7c and e vs. 7a and Fig. 7d and f vs. 7b). The temperature in western Europe also slightly increases when IPE reductions are included, particularly when utilizing the F scheme with high sensitivity (Fig. 7c). Isoprene is among the most important precursors for the formation of SOAs (Claeys et al., 2004), which are able to reduce surface air temperature by light extinction (Charlson et al., 1992). As a result, the O 3 -induced reduction of IPE decreases SOA loading and weakens the "cooling effect" of aerosols, leading to a higher temperature at the surface. The positive changes in shortwave radiative forcing following SOA reduction are the strongest in the eastern US when considering the effect of O 3 damage to IPE, particularly for the F schemes with high sensitivity (Fig. 9). Such warming explains why the reduced IPE help increase the surface O 3 in the eastern US (Fig. 6). However, aerosols in regions with high anthropogenic emissions (such as eastern China) are more dominated by inorganic components (Sun et al., 2006;Yang et al., 2011); thus, the changes in SOAs are less important. As a result, the feed-back of O 3 -induced IPE reductions on temperature is not significant in eastern China compared to that of other regions.

Effects of O 3 -vegetation interactions on meteorology and vegetation
In addition to the direct damage (Fig. 3), IPE are indirectly affected by perturbations in the LAI and meteorology. Figure S5 shows that the LAI decreases in three polluted regions (eastern China, the eastern US, and western Europe) because of the O 3 -mediated inhibition of photosynthesis, although the magnitude is typically within 5 %. Moderate changes in the LAI by O 3 have also been reported in previous studies (Yue and Unger, 2015;Sadiq et al., 2017), suggesting that LAI feedback is too low to effectively influence IPE and the consequent surface O 3 . Furthermore, the warming effects resulting from the O 3 -induced inhibition on stomatal conductance (Fig. 7) and the changes in the LAI (Fig. S7 in the Supplement) cause limited changes in IPE (Fig. S8 in the Supplement), suggesting that O 3 -vegetation feedback does not significantly change IPE. In comparison, Sadiq et al. (2017) reported a strong positive feedback (3-5 times greater than our results) on IPE caused by increased temperature from reduced transpiration when the effect of O 3 damage to stomatal conductance is considered. However, Sadiq et al. (2017)   in a global model. Compared to their results, we find an ultimate positive feedback with a similar magnitude of surface O 3 concentrations but different spatial pattern. The strongest feedback is in eastern China rather than western Europe, which is more reasonable, as the O 3 level in China is much higher than that in Europe (Lu et al., 2018). In addition, the effect of O 3 -vegetation feedback on temperature is lower in our study. The fixed decoupled scheme in Sadiq et al. (2017) may have overestimated the effect of O 3 damage to stomatal conductance, leading to stronger feedback on O 3 concentrations and temperature. Furthermore, the mechanisms of O 3 effects on IPE are different. Sadiq et al. (2017) showed increased IPE because of the warming feedback. However, such warming is not significant in our study (Fig. S8 in the Supplement). Instead, we include the direct effect of O 3 damage to IPE based on observations. Although the simulations show limited impacts of reduced IPE on surface O 3 , the simultaneously reduced SOAs contribute to increased surface O 3 by weakening shortwave radiative forcing and increasing temperature in the eastern US.
Our results are subject to uncertainties in modeled O 3 and damaging schemes. ModelE2-YIBs overestimates summer O 3 , particularly in China (Fig. 1), which may exacerbate the damage to stomatal conductance and the consequent feedback. The O 3 damage parameterization by Sitch et al. (2007) is a semiphysical scheme that couples photosynthesis and stomatal conductance. However, some observational studies have showed that the sluggish stomatal responses under chronic O 3 exposure lead to stomata losing function and decoupling from photosynthesis (Paoletti and Grulke, 2005;Gregg et al., 2006). The decoupled parameterization proposed by Lombardozzi et al. (2012) has been applied to estimate the effect of O 3 damage to photosynthesis and stomatal conductance (Lombardozzi et al., 2015;Sadiq et al., 2017;Zhou et al., 2018). Nevertheless, we apply the parameterization by Sitch et al. (2007) because the damage is reasonably associated with the ambient O 3 level, and the scheme has been extensively evaluated against available observations (Yue et al., 2017;Yue and Unger, 2018). Fixed damage for low (even zero) O 3 included in some PFTs in the decoupled scheme may result in overestimation of O 3vegetation feedback in the global model.
To our knowledge, this is the first time that the effect of O 3 damage to IPE is included in a fully coupled global chemistry-carbon-climate model. Both the F and linear schemes can simulate reasonable reductions in IPE compared to global meta-analysis, although with large uncertainties. The reduced IPE, as precursors, have insignificant effects on surface O 3 concentrations in eastern China (Figs. 5 and 6), likely because of high anthropogenic emissions that undermine the feedback of IPE changes to surface O 3 . However, the reduced IPE weaken SOA radiative forcing and increase surface temperature in the eastern US, where biogenic SOAs provide important contributions to total aerosols (Fine et al., 2008;Goldstein et al., 2009). These results suggest that IPE feedback to the surface O 3 is quite uncertain and dependent on ambient precursors (anthropogenic vs. biogenic) and oxidizing capacity (NO x -saturated vs. NO x -limited).
Variations in meteorological parameters may also influence O 3 -vegetation feedback. Plant stomata tend to close under drought stress to prevent water loss. As a result, dry climate may weaken O 3 -vegetation feedback through regulation of stomatal conductance (Lin et al., 2019). The effects of drought cannot be evaluated using ModelE2-YIBs, which simulates climatology with small interannual variability. In the future, a chemical transport model (CTM) coupled with a dynamic vegetation model (such as GC-YIBs developed by Lei et al., 2020) will be used to examine drought impacts by using observation-based meteorological forcings.
Despite these uncertainties, our analyses highlight the importance of O 3 -vegetation interactions in surface O 3 concentrations. The feedback should be considered in regional and global air quality models for more realistic simulations. Furthermore, the effect of positive feedback on surface O 3 may potentially aggravate O 3 pollution in the future with increased ambient O 3 under a warming climate (Lei et al., 2012;Doherty et al., 2013).
Author contributions. XY conceived the study. CG carried out the simulations and performed the analysis. YL and YM provided useful comments on the paper. CG, XY, and HL prepared the paper, with contributions from all coauthors.
Financial support. This research has been supported by the National Key Research and Development Program of China (grant no. 2019YFA0606802), the National Natural Science Foundation of China (grant nos. 41975155 and 91744311), and the Startup Foundation for Introducing Talent of NUIST.
Review statement. This paper was edited by Frank Dentener and reviewed by two anonymous referees.