Competing effects of nitrogen deposition and ozone exposure on Northern hemispheric terrestrial carbon uptake and storage, 1850-2099

Tropospheric ozone (O3) and nitrogen deposition affect vegetation growth and thereby the ability of the land biosphere to take up and store carbon. However, the magnitude of these effects on the contemporary and future terrestrial carbon balance is insufficiently understood. Here, we apply an extended version of the O-CN terrestrial biosphere model that simulates the atmosphere to canopy transport of O3, its surface and stomatal uptake, the O3-induced leaf injury, as well as the coupled ter5 restrial carbon and nitrogen cycles. We use this model to simulate past and future impacts of air pollution against a background of concurrent changes in climate and carbon dioxide concentrations (CO2) for two contrasting representative concentration pathways (RCP) scenarios (RCP2.6 and RCP8.5). The simulations show that O3-related damage considerably reduced Northern hemispheric gross primary production (GPP) and long-term carbon storage between 1850 and the 2010s. The simulated O3 effect on GPP in the Northern hemisphere peaked 10 towards the end of the 20 century with reductions of 4 %, causing a reduction in the Northern hemispheric carbon sink of 0.4 PgCyr. During the 21 century, O3-induced reductions in GPP and carbon storage are projected to decline through a combination of direct air pollution control methods that reduce near surface O3 and the indirect effects of rising atmospheric CO2, which reduces stomatal uptake of O3 concurrent with increases of leaf-level water-use efficiency. However, in hotspot regions such as East Asia, the model simulations suggest a sustained decrease of GPP by more than 8 % throughout the 21 15 century. O3 exposure reduces projected carbon storage at the end of the 21 century by up to 15 % in parts of Europe, the US and East Asia. Our simulations suggest that the stimulating effect of nitrogen deposition on regional GPP and carbon storage is lower in magnitude compared to the detrimental effect of O3 during most of the simulation period for both RCPs. In the second half of the 21 century, the detrimental effect of O3 on GPP is outweighed by nitrogen deposition, but the effect of nitrogen deposition on land carbon storage remains lower than the effect of O3. Accounting for the stimulating effects of nitrogen 20 deposition but omitting the detrimental effect of O3 may lead to an over estimation of projected carbon uptake and storage.

Productivity and carbon storage in many Northern hemispheric terrestrial ecosystems are affected by the limited availability of nitrogen (N; Vitousek and Howarth, 1991;LeBauer and Treseder, 2008;Zaehle, 2013). As a side-effect of air pollution, increased deposition of reactive nitrogen from e.g. anthropogenic fossil fuel burning and increased soil emissions associated 25 with fertiliser use (Galloway et al., 2004) have the potential to fertilise these N-limited ecosystems and thereby enhance productivity and carbon storage (Norby, 1998;Zaehle et al., 2011;Thomas et al., 2010). However, oxidised forms of reactive nitrogen in the atmosphere (collectively referred to as NO y ) are also a precursor of tropospheric ozone (O 3 is a toxic air pollutant that enters plants primarily though the leaves' stomata where it can induce cellular damage (Fiscus et al., 2005;Tausz et al., 2007;McAinsh et al., 2002). Commonly observed effects are visible injury (Langebartels et al., 30 1991;Wohlgemuth et al., 2002), reductions in photosynthetic capacity (Tjoelker et al., 1995;Wittig et al., 2007) and growth or yield (Grantz et al., 2006;Hayes et al., 2007;Feng and Kobayashi, 2009;Wittig et al., 2009;Leisner and Ainsworth, 2012). Ozone induced plant damage can reduce the terrestrial carbon uptake and storage and through this cause an increase in atmospheric CO 2 concentrations and an intensification of climate change (Sitch et al., 2007;Ainsworth et al., 2012). Ozone mixing ratios in Europe have approximately doubled during the 20 th century (Cooper et al., 2014). The anthropogenic increase 35 in NO y emissions primarily from combustion sources has been identified as the major cause for the increasing near-surface O 3 concentrations between 1970-1995 in the mid-latitudes of the Northern Hemisphere (Fusco and Logan, 2003). Ozone levels are projected to decline until the end of the 21 s t century due to assumed stringent air pollution policies, but future climate conditions with increasing temperatures as well as reduced cloudiness and precipitation will tend to increase O 3 formation with increasing daily O 3 peaks and average concentrations in summer (Meleux et al., 2007;van Vuuren et al., 2011). The 40 application of the RCP scenarios (Moss et al., 2010;van Vuuren et al., 2011) in 14 global chemistry transport models results in the projection of declining annual global mean surface O 3 concentrations in most regions of the globe except South Asia where increases are simulated (Wild et al., 2012). Projections of nitrogen deposition in the 21 st century suggest little change across all scenarios of the Representative concentration pathways (RCP), despite notable regional differences (Lamarque et al., 2013). A small decline in deposition rates is proposed only under the scenario RCP2.6. 45 Simulations with nitrogen-enabled terrestrial biosphere models suggest that N deposition may be responsible for 10 to 50 % of the global residual land carbon uptake (Zaehle et al., 2011;Quéré et al., 2018). Several models including the O 3 effect on carbon cycle suggest that simulated present-day and future O 3 exposure can reduce regional and global scale productivity (Felzer et al., 2005;Sitch et al., 2007;Franz et al., 2017;Lombardozzi et al., 2015;Oliver et al., 2018). For instance, modelling studies by Sitch et al. (2007) and Oliver et al. (2018) suggest a reduction in O 3 induced damage of gross primary production Here, we assess the combined effect of O 3 and nitrogen deposition on the Northern hemispheric terrestrial biosphere against the background of simulated changes due to increasing atmospheric CO 2 and climate change. Elevated levels of atmospheric CO 2 stimulate leaf photosynthesis and reduce stomatal conductance (Medlyn et al., 2001;Ainsworth and Long, 2005), and therefore can increase plant growth and plant nitrogen limitation (Oren et al., 2001;Norby et al., 2009;Zaehle et al., 2014).
We analyse the response of the Northern hemispheric carbon cycle to changes in climate, atmospheric CO 2 and O 3 as well as N deposition for the historical period  and two future scenarios , a high and a low climatechange mitigation scenario, RCP2.6 and RCP8.5 respectively. In a factorial analysis, we investigate the impact of the single drivers (O 3 , CO 2 and N deposition), as well as their interaction (specifically the interaction between O 3 and CO 2 , and O 3 and N deposition) on plant growth and terrestrial carbon storage. We employ a significantly enhanced version of the O-CN terrestrial biosphere model , which explicitly accounts for the O 3 transport and deposition from the free troposphere into the stomata, as well as O 3 removal by other processes (such as soil and leaf surface uptake) (Franz et al., 2017). This model has been evaluated against biomass damage relationships observed in a range of fumigation/filtration experiments with European tree species (Büker et al., 2015;Franz et al., 2018).

Methods
Simulations are conducted with the O-CN terrestrial biosphere model Franz et al., 2017), version 70 tun V C where O 3 damage is calculated based on injury functions to the maximum carboxylation capacity of the leaf V cmax (Franz et al., 2018). The tun V C injury functions were calibrated to reproduce observed biomass damage relationships of experiments with a range of European tree species in fumigation/filtration experiments (Franz et al., 2018;Büker et al., 2015).
The O-CN model includes an O 3 deposition scheme that explicitly accounts for the O 3 transport and deposition from the free troposphere into the stomata (Franz et al., 2017). Here, we use the ozone deposition scheme referred to as D-model in 75 Franz et al. (2017), contrary to Franz et al. (2018) where the O 3 deposition scheme was turned off.

The O-CN model
O-CN  is a further development of the land-surface-scheme ORCHIDEE (Krinner et al., 2005), and simulates the coupled terrestrial carbon (C), nitrogen (N) and water cycles for twelve plant functional types. The model accounts for the effects of nitrogen availability on growth, root:shoot allocation, litter and soil organic matter decay, and 80 represents a comprehensive nitrogen cycle including process-oriented formulations for nitrogen leaching and gas losses, and its ability to reproduce N fertilisation experiments has been evaluated by (Meyerholt and Zaehle, 2015). O-CN compares well to a range of regional to global terrestrial biosphere benchmarks (Quéré et al., 2018). O-CN is driven by climate data, N deposition, atmospheric composition including the atmospheric CO 2 and O 3 concentrations and land use information.
O-CN simulates a multi-layer canopy with up to 20 layers (each with a thickness of up to 0.5 leaf area index) where net 85 photosynthesis is calculated for shaded and sun-lit leaves with consideration of the light profiles of diffuse and direct radiation (Kull and Kruijt, 1998;Friend, 2001;. Stomatal conductance to water, CO 2 and O 3 is calculated coupled to net photosynthesis following a Ball-&-Berry-type formulation (see Sect. 2.2). Leaf nitrogen concentration and leaf area determine the photosynthetic capacity, which are both affected by ecosystem available N. The maximum carboxylation capacity (V cmax ) and electron transport capacity (J max ) of the leaf increase with an increased leaf nitrogen concentration, 90 leading to an increase in the maximum net photosynthesis and stomatal conductance per unit leaf area ). The highest leaf N content is simulated at the top of the canopy and exponentially decreases with increasing canopy depth (Friend, 2001;Niinemets et al., 2015). Following this, the net photosynthesis, stomatal conductance and O 3 uptake are generally highest in the top of the canopy and lowest in the bottom of the canopy. Changes in stomatal conductance affect transpiration rates and estimates of O 3 uptake and O 3 damage.

Ozone injury calculation in O-CN
Leaf-level O 3 uptake is determined by stomatal conductance and atmospheric O 3 concentrations, as described in Franz et al. (2017). In contrast to Franz et al. (2017), the stomatal conductance g st is calculated based on the Ball and Berry formulation (Ball et al., 1987) as where RH is the atmospheric relative humidity, f (height l ) the water-transport limitation with canopy height, [CO 2 ] the atmospheric CO 2 concentration, A n,l the net photosynthesis, g 0 the residual conductance when A n approaches zero, and g 1 the stomatal-slope parameter as in Krinner et al. (2005). The index l indicates that g st and A n are calculated separately for each canopy layer. A n,l is calculated as described in  as a function of the leaf-internal partial pressure of CO 2 , absorbed photosynthetic photon flux density on shaded and sunlit leaves, leaf temperature, the nitrogen-specific rates 105 of maximum light harvesting, electron transport (J max ) and carboxylation rates (V cmax ).
The stomatal conductance to O 3 g O3 st,l is calculated as where the factor 1.51 accounts for the different diffusivity of O 3 from water vapour (Massman, 1998 and g st,l is calculated as where the leaf-internal O 3 concentration ([O 3 ] i ) is assumed to be zero (Laisk et al., 1989).
with R a the aerodynamic resistance, R b the canopy-scale quasi-laminar layer resistance and R c the compound surface resistance to O 3 deposition. R c is calculated as the sum of the canopy scale stomatal and the non-stomatal resistance to O 3 uptake (Franz et al., 2017). The non-stomatal resistance is defined by the O 3 destruction on the leaf surface, within-canopy resistance 120 to O 3 transport, and ground surface resistance (Franz et al., 2017).
Without the application of the O 3 deposition module, the O 3 uptake inside the leaves would be calculated based on the near surface O 3 concentrations from the forcing data without accounting for the turbulent transport between the lower troposphere and the leaves, as well as the deposition and destruction of ozone on other surfaces.
The accumulation of O 3 fluxes into the leaves above a threshold of X nmol m gives the CU OX l . Summing CU OX l over all canopy layers gives the canopy value CU OX (Franz et al., 2017). In this study, a flux threshold of 1 nmol m −2 s −1 , i.e. CUO1, is applied to account for the plants ability to detoxify part of the taken up O 3 (Franz et al., 2018;LRTAP-Convention, 2017;Büker et al., 2015).
where the slope of the injury function (b) is set to 0.075 for broadleaf species and 0.025 for needleleaf species (Franz et al., 2018). d O3 l is calculated separately for each canopy layer l according to the specific accumulated O 3 uptake of the respective canopy layer (CU O1 l ), and takes values between 0 and 1. Within-canopy gradients in stomatal conductance and photosynthetic capacity cause variations of CU O1 l and hence d O3 l between canopy layers.

140
The effect of O 3 injury on plant carbon uptake is calculated by with the maximum carboxylation capacity of the leaf in the respective canopy layer (V cmax,l ), which is used in the calculation of A n,l . J max,l is reduced in proportion to V cmax,l such that the ratio between both is maintained.
Ozone induced reductions in A n,l cause a decline in g st,l as both are tightly coupled. Lower values of g st,l diminish the O 3 145 uptake into the plant (f st,l ) and slow the increase in CU O1 l and hence O 3 induced injury.

Model forcing
The model is driven by climate model output of the Institute Pierre Simon Laplace (IPSL) general circulation model IPSL-CM5A-LR (Dufresne et al., 2013), bias-corrected according to the Inter-Sectoral Impact Model Intercomparison Project (Hempel et al., 2013). The applied meteorological forcing for near-surface conditions comprises daily data of specific humid-150 ity, incoming long wave radiation, incoming short wave radiation, cloudiness, wind speed, maximum temperature, minimum temperature and total precipitation, which are disaggregated to the 30 min time step of the model using a statistical weather generator (Krinner et al., 2005). Reduced and oxidised monthly mean nitrogen deposition in wet and dry form and monthly mean near surface O 3 concentrations are provided by CAM, the community atmosphere model (Lamarque et al., 2010;Cionni et al., 2011). Land cover, soil, and N fertiliser application are used as in Zaehle et al. (2011)  Figure 1 provides and overview over the scenarios applied in this study. Note that there are important regional patterns behind the changes in N deposition and near surface O 3 , which are shown in the Supplement Fig. S1 and S2.

160
The model is run at a spatial resolution of 1 • × 1 • and operates on a half hourly time step. As the injury functions developed by To investigate the impact of the O 3 deposition scheme on the simulation results, the factorial runs are repeated with a model version where the O 3 deposition scheme is turned off (see ATM model version in (Franz et al., 2017). In simulations where the   S1 1850-2099 1901-1930 1850 1850 S2 1850-2099 1901-1930 1850 1850-2099 S3 1850-2099 1901-2099 1850 1850-2099 S4 1850-2099 1901-2099 1850-2099 1850 S5 1850-2099 1901-2099 1850-2099 1850-2099 O 3 deposition module is turned off the canopy O 3 concentration equals the O 3 concentration at 45 m above the surface which 175 is the hight of the lowest level of the forcing data.

Factorial analysis
The impact of a single forcing driver on the simulation results can be approximated by subtracting the simulation results of suitable combination of factorial runs from one another (see Tab.2). In the following, the term 'forcing driver' is used to refer to the input variables of the conducted simulations and 'single driver' refers to the approximated impact of a single forcing 180 Table 2. Calculation of the single driver effects (CO2, climate, nitrogen deposition, O3) from the conducted simulations. S1 ref refers to the mean of the years 1850 to 1859 of the S1 simulation. See Tab. 1 for info on the forcing setting of the factorial runs S1-S5.
Attributed single driver Simulations driver on the simulation results. The described approach is an approximation of the impact of the single drivers and assumes that the effects on the analysed output variables are additive. The assumption of additive effects is a necessary simplification to restrict the number of simulations and computation time . For O 3 , a main driver of interest, two different approaches to calculate the single driver were realised. In one approach, the O 3 impact is calculated from the two factorial runs with only one/ two transient drivers (S1 and S2), and a second time from the factorial runs where all and all but one driver 185 (S5 and S4 respectively) are simulated transient. The comparison of these two approaches to calculate the single driver might indicate the extent of impact of interacting forcing drivers on the estimate of the O 3 single driver. The relative changes between two simulation runs SX and SY are calculated as (SX − SY )/SY .

Results
The simulations show a strong increase in gross primary production (GPP) in the Northern Hemisphere (≥ 30 • N) between the O 3 concentrations (Fig.3a). The decline in peak uptake rates is the consequence of the reduced ratio of stomatal conductance to net photosynthesis under high atmospheric CO 2 .

205
We next decompose the simulations into the effects of the different model drivers with a special focus on the effects of O 3 and nitrogen deposition. In all five factorial runs the simulated GPP increases strongly between 1850 and 2099 and approximately doubles for the run S5 based on RCP8.5 (see Fig. 2a). The primary cause for this simulated increase is the CO 2 fertilisation effect induced by increasing atmospheric CO 2 concentrations (see Fig. 4c and Tab. 3). Climate change is the second most important factor for the simulated increase, whereas the positive effect of N deposition is less pronounced. Ozone injury causes 210 a modest decrease in productivity, which manifests strongest during the 1990s. During the 20 th century the decline gradually reverses. The land carbon sink strongly responds to elevated levels of CO 2 (see Fig. 4f), whereas climate change induces a varying impact on the land carbon sink. During the second half of the 21 st century, the effect of climate mainly causes a reduction in the simulated land carbon sink.  Figure 4. Single drivers obtained by subtracting factorial runs for selected output variables. Displayed are the results for simulated regional mean O3 uptake (Fst), regional mean cumulative canopy O3 uptake above a flux threshold of 1 nmol m −2 s −1 (CUO1), regional summed GPP, regional summed stocks of total carbon biomass (vegetation-C), soil organic matter carbon (soil-C), and summed land carbon flux (land  The two different approaches to assess the contribution of O 3 to the simulated trends in the carbon cycle based on analysing 250 alternative combinations of model drivers (see Tab. 2) yield similar but not identical results (see Fig. 7). Typically, the differences between the two approaches do not exceed 1 % except for CUO1, where larger relative changes occur for small absolute changes (see Fig. 7b).
effects of O 3 on GPP nearly completely counteract the positive effect of rising CO 2 concentrations (see Fig. 4c). The negative impact of O 3 on GPP shows a maximum approximately in the 1990s at approximately -1.5 PgC yr −1 (4 %) compared to preindustrial values (see Fig. 7c, Supplement Fig. S5 and Tab. 4). In the subsequent decades, the simulated O 3 induced reduction in GPP declines to 1 % by the end of the 21 st century for RCP8.5 and to close to zero for RCP2.6.
Due to the stabilisation of atmospheric CO 2 in the RCP2.6 scenario, the increase in GPP levels off at 2040s levels, but 260 continues to rise under RCP 8.5 with increasing CO 2 . The growth-stimulating effect of N-deposition is smaller than the negative impact induced by O 3 during the 20 th century (see Fig. 4c). This pattern is reversed during the course of the 21 st century (see Section 3.4).
The O 3 effect on GPP propagates to vegetation and thus considerably affects the simulated total carbon biomass in vegetation

Comparative impact of N deposition and O 3
The magnitude of the O 3 induced damage on GPP exceeded the growth stimulating effect induced by nitrogen deposition over large parts of the 20th century and until the beginning of the 21 st century (see Fig. 4c). In contrast to the near surface O 3 concentrations, the regional mean nitrogen deposition does not decline during the 21 st century but slightly increases in RCP8.5 310 and RCP2.6. The growth stimulating effect on GPP induced by nitrogen deposition becomes higher in magnitude during the 21 st century compared to the detrimental effect of O 3 (see Fig. 4c, Supplement Fig. S8 and Tabs. 4 and 4).
The growth stimulating effect of nitrogen deposition on vegetation-C remains lower in magnitude compared to the detrimental effects of O 3 for both emission scenarios throughout the entire simulation period (see Fig. 4d and Tab 2013), while the regional means including many areas with low O 3 exposure, results in lower average O 3 damage than estimated by these meta-analyses.

395
Several process-based models estimated O 3 induced damage to GPP or net primary production (NPP) on global or regional scale: a mean global O 3 induced reduction in NPP of 0.8-2.9 % from 1989 to 1993 is estimated by the Terrestrial Ecosystem Model (Felzer et al., 2005). to charcoal filtered air and a 17 % reduction for trees grown in elevated O 3 concentrations compared to charcoal filtered air (Wittig et al., 2009). In a meta-analyses by Li et al. (2017) a 14 % reduction in total biomass is calculated for trees grown in elevated O 3 concentrations (mean of 116 ppb) compared to controls grown in a mean O 3 concentration of 21 ppb. The simulated regional mean estimate of O 3 induced damage to vegetation-C is higher compared to the estimate of trees grown in ambient vs. charcoal filtered air by Wittig et al. (2009) and lower compared to trees grown in elevated O 3 vs. charcoal filtered 420 air or a mean of 21 ppb O 3 (Wittig et al., 2009;Li et al., 2017). Simulated damage values in the hotspots are higher compared to the estimates by the meta-analyses.
Our simulated declines in O 3 induced damage to GPP and vegetation-C during the 21 st century generally agree with simulated reductions in potential threat to vegetation by Klingberg et al. (2014). Klingberg et al. (2014) (Klingberg et al., 2014).

430
An ensemble of six global atmospheric chemistry transport models project improvements of the AOT40 index in the Northern Hemisphere by 2099 under the RCP2.6 and RCP4.5, while critical levels continue to be exceeded over many areas (Sicard et al., 2017). By 2099 the potential impact of O 3 on photosynthesis and carbon assimilation is projected to decline by 61 % under the RCP2.6 scenario, by 47 % under RCP4.5 and increase by 70 % under the RCP8.5 scenario compared to the early 2000s (Sicard et al., 2017). Elevated levels of CO 2 (eCO 2 ) have the potential to induce stomatal closure (Paoletti and Grulke, 2005) what might limit O 3 uptake and damage. Contradictory evidence exists showing that either eCO 2 ameliorated the negative effects of O 3 on plants ( Barnes and Pfirrmann, 1992;Broadmeadow and Jackson, 2000;Isebrands et al., 2001;Riikonen et al., 2004) or that there was little interaction between both gases and the stimulating effect of eCO 2 on NPP persisted (Talhelm et al., 2014;Zak et al., 440 2011). Results from the Aspen FACE indicate that stomatal conductance and O 3 uptake were not reduced by eCO 2 in their experiment (Uddling et al., 2010), and that O 3 fumigation completely offset the growth enhancement observed in the eCO 2 treatment for O 3 sensitive and tolerant clones (Karnosky et al., 2003).
Several studies find species specific positive or negative impacts of eCO 2 and elevated levels of O 3 (eO 3 ) on photosynthesis (Noormets et al., 2001), growth  and biomass (King et al., 2005). An amplification of the negative effects of O 3 under eCO 2 on leaf chlorophyll content, nitrogen content and electron transport capacity (J max ) was observed in O 3 sensitive and tolerant aspen clones (Noormets et al., 2010). A possible reason for the amplification of O 3 induced negative effects under eCO 2 is a possible down regulation or suppression of antioxidant production under eCO 2 and hence increased injury (Wustman et al., 2001;Karnosky et al., 2003). All in all, a clear picture of the joint effects of eCO 2 +eO 3 on plants or plant groups is still lacking.

450
Terrestrial biosphere models often assume a tight coupling between net photosynthesis and stomatal conductance which induces stomatal closure in case of simulated eCO 2 and restricts O 3 uptake and damage (Felzer et al., 2004(Felzer et al., , 2005Sitch et al., 2007;Oliver et al., 2018;Yue and Unger, 2014). For example Sitch et al. (2007)  concentration rose quickly (Oliver et al., 2018). During the 21th century simulated O 3 concentrations changed less and the simulated elevated levels of CO 2 restricted O 3 uptake and induced damage (Oliver et al., 2018). This agrees well with our findings here that O 3 induced damage increases from pre-industrial times until the end of the 20 th century (GPP) or beginning 460 of the 21 st century (vegetation-C) and afterwards decreases again (see Fig. 7).
However, the simulation of reduced O 3 uptake and incurred damage induced by eCO 2 does not mirror all the effects observed in field experiments (Wustman et al., 2001;Karnosky et al., 2003;Noormets et al., 2010). Similar to other terrestrial biosphere models, O-CN does not account for observed effects like an exacerbation of O 3 induced damage due to eCO 2 (Wustman et al., 2001;Karnosky et al., 2003)

Limitations of comparisons between publications
When interpreting the comparison of our results and previously published simulation results one has to keep in mind that the 470 different modelling set-ups and approaches differ in several aspects that considerably affect the damage estimate. Models often apply different injury functions which relate O 3 uptake to plant damage (Lombardozzi et al., 2012(Lombardozzi et al., , 2015Franz et al., 2017;Oliver et al., 2018). However, injury functions have the potential to induce considerable over-or underestimation of simulated biomass damage compared to measured damage values (Franz et al., 2018). Simulations differ in the time period covered, e.g. (2018) simulate changing CO 2 concentrations and a partly fixed land-cover, but no effect of N deposition. Furthermore damage 480 estimates are calculated based on different reference periods and conditions. Damage might be given as the difference between a simulation accounting for O 3 damage compared to a reference simulation not accounting for O 3 damage (Lombardozzi et al., 2015;Franz et al., 2017). Another approach is to report the damage simulated between a specific time period. Sitch et al. (2007) calculate O 3 induced damage between 1901-2100 and Oliver et al. (2018Oliver et al. ( ) between 1901Oliver et al. ( -2001Oliver et al. ( and 2001Oliver et al. ( -2050 Different modelling studies apply differing emission scenarios, e.g. IPCC SRES (Sitch et al., 2007) and the RCP scenarios 485 used here, which might impact simulated O 3 uptake and incurred damage. The application of the IPCC SRES scenarios, which assume a large increase in O 3 precursor emissions, implies an increase in annual global mean surface O 3 concentrations by 4-6 ppb (Wild et al., 2012). Contrary to this, the application of the RCP scenarios (Moss et al., 2010;van Vuuren et al., 2011) in 14 global chemistry transport models results in the projection of declining annual global mean surface O 3 concentrations of as much as 2 ppb by 2050 in most regions of the globe except South Asia where increases are simulated (Wild et al., 2012).

490
Lower projected ozone-induced damage in our study compared to Sitch et al. (2007) is therefore also a consequence of the assumed scenario. Turnock et al. (2020) found that the CMIP6 models overestimate observed surface O 3 concentrations by up to 16 ppb across most regions of the globe. This will likely lead to a general overestimation of simulated O 3 damage by terrestrial biosphere models. However, the ozone deposition scheme included into O-CN has the potential to ameliorate this observed discrepancy. and others are forced by monthly diurnal mean values (e.g. Sitch et al. (2007) and the simulations here). As the formation of 500 O 3 shows a pronounced diurnal cycle (Sanz et al., 2007), the use of monthly mean O 3 concentrations probably impacts the simulated estimates of O 3 uptake. However, to which extent the omission of a diurnal cycle impacts O 3 uptake, accumulation and damage estimates is yet uncertain.

Limits to the parameterisation of O 3 damage in O-CN
Plants can activate defence mechanism and physiological pathways to produce protective compounds like ascorbate and 505 polyamines which can detoxify at least part of the ozone taken up (Kangasjärvi et al., 1994;Kronfuß et al., 1998;Tausz et al., 2007). In the simulations conducted here we account for detoxification by introducing a flux threshold but do not account for the cost to produce protective compounds like antioxidants due to the lack of suitable data. This could potentially introduce a bias bias towards underestimating damage to GPP if the leaf-injury parameterisations are based on leaf-level data.
Ozone sensitivity is known to differ between plant groups, plant species and between genotypes (Wittig et al., 2007;Lom-510 bardozzi et al., 2013;Li et al., 2017;Hayes et al., 2007;Karnosky et al., 2003). The assumed injury function is a key aspect of the simulation of O 3 damage and has a large impact on the extent of the estimated damage (Franz et al., 2018). However, the scarcity of suitable data restricts the possibility to parameterise injury functions for all simulated PFTs (e.g. 12 PFTs in O-CN), let alone a variation of the O 3 -sensitivity within PFTs. Furthermore it restricts the evaluation of O 3 -submodels and the included injury functions. The injury functions used for the simulations here are tuned to reproduce observed biomass damage 515 from filtration/fumigation experiments of broadleaved and needle-leaved tree species (Franz et al., 2018). The simulations are restricted to the Northern Hemisphere ≥ 30 • N to limit the domain of simulation to temperate/boreal forests and thus similar species as used for the tuning of the injury functions. Due to the lack of suitable damage functions for grass species we here applied the damage functions developed to match damage to trees. This induces a bias in the damage estimates and will likely results in an underestimation of simulated damage for example for the crop plant functional types.

520
The biomass damage experiments used to parameterise the injury function in O-CN were conducted with young trees grown in monocultures. The common attempt to estimate responses of adult trees grown under natural conditions by the extrapolation of results from short-term experiments with young trees is subject to several issues, e.g. due to the differing environmental conditions and changing O 3 sensitivities with increasing tree size or age (Schaub et al., 2005;Cailleret et al., 2018;Franz et al., 2018). It is yet uncertain if the simulation of injury to photosynthesis based on experiments with young trees can be 525 transferred to adult trees to obtain realistic biomass damage estimates.
Differing O 3 sensitivities can induce changes in community composition (Barbo et al., 1998;Kubiske et al., 2007;Zak et al., 2011) as well as the interactive effects of changed CO 2 and O 3 concentrations (Karnosky et al., 2003). The simulations conducted here are run offline and following this atmosphere and biosphere do not feedback on one another.
Forcing variables like O 3 concentrations and nitrogen deposition are provided by a different model than the climate. This imposes an inconsistency between the biosphere, climate and the abundance of the air pollutants whose formation depends on 550 climate variables. This contributes to unavoidable inconsistencies between the atmospheric forcing and the land fluxes when making offline simulations compared to a simulation with a fully coupled Earth System Model. However, these limitations, do not invalidate the simulated sensitivity of the land carbon cycle simulation to the forcing applied. Mankovska, B., Heilman, W., and Isebrands, J. G.: Tropospheric O3 moderates responses of temperate hardwood forests to elevated CO2:

Conclusion
a synthesis of molecular to ecosystem results from the Aspen FACE project, Functional Ecology, 17, 289-304, 2003.