Evaluation of simulated biomass damage in forest ecosystems induced by ozone against observation-based estimates

Regional estimates of the effects of ozone pollution on forest growth depend on the availability of reliable injury functions that estimate a representative ecosystem response to ozone exposure. A number of such injury functions for forest tree species and forest functional types have recently been published and subsequently applied in terrestrial biosphere models to estimate regional or global effects of ozone on forest tree productivity and carbon storage in the living plant biomass. The resulting 5 impacts estimated by these biosphere models show large uncertainty in the magnitude of ozone effects predicted. To understand the role that these injury functions play in determining the variability of estimated ozone impacts, we use the O-CN biosphere model to provide a standardised modelling framework. We test four published injury functions describing the leaflevel, photosynthetic response to ozone exposure (targeting the maximum carboxylation capacity of Rubisco (Vcmax) or net photosynthesis) in terms of their simulated whole-tree biomass responses against field data from 23 ozone filtration/fumigation 10 experiments conducted with European tree species at sites across Europe with a range of climatic conditions. Our results show that none of these previously published injury functions lead to simulated whole-tree biomass reductions in agreement with the observed dose-response relationships derived from these field experiments, and instead lead to significant over/ or underestimations of the ozone effect. By re-parameterising these photosynthetic based injury functions we develop linear, plant


Introduction
Ozone is a phytotoxic air pollutant which enters plants mainly through the leaf stomata, where reactive oxygen species (ROS) are formed that can damage essential leaf functioning (Ainsworth et al., 2012). Ozone induced declines in net photosynthesis (Morgan et al., 2003;Wittig et al., 2007) have been observed as the result of damage of the photosynthetic apparatus, increased respiration rates caused by investments in repair of injury, as well as the production of defence compounds (Wieser and 5 Matyssek, 2007; Ainsworth et al., 2012). At the leaf-scale, ozone damage occurs and accumulates, when the instantaneous stomatal ozone uptake of leaves surpasses the ability of the leaf to detoxify ozone (Wieser and Matyssek, 2007). These effects are likely the primary cause for reduced rates of net photosynthesis and decreased supply of carbon and energy for growth and net primary production (NPP), which contributes to the commonly observed ozone-induced reductions in leaf area and plant biomass (Morgan et al., 2003;Lombardozzi et al., 2013;Wittig et al., 2009). Changes in tropospheric ozone abundance and 10 associated changes in ozone-induced damage thus have the potential to affect the ability of the terrestrial biosphere to sequester carbon (Harmens and Mills, 2012;Oliver et al., 2017). However, a quantitative understanding of the effect of ozone pollution on forest growth and carbon sequestration at the regional scale is still lacking. Terrestrial biosphere models can be used to obtain regional or global estimates of ozone damage based on an understanding of how ozone affects plant processes leading to C assimilation and growth. Modelling algorithms to estimate regional or global impacts of ozone on gross primary production 15 (GPP) have been developed for several of these terrestrial biosphere models (Sitch et al., 2007;Lombardozzi et al., 2012aLombardozzi et al., , 2015Franz et al., 2017;Oliver et al., 2017). However, simulated reductions in GPP due to ozone damage vary substantially between models and model versions.
This uncertainty is predominantly due to the different approaches that these models use to relate ozone uptake (or ozone exposure) to reductions in whole-tree biomass, and in the exact parameterisation of the dose-response relationship applied 20 Pleijel et al., 2004;Wittig et al., 2007;Lombardozzi et al., 2012aLombardozzi et al., , 2013. The dose-response relationships employed by current terrestrial biosphere models differ decidedly in their slope (i.e. the change in damage per unit of timeintegrated ozone uptake), intercept (ozone damage at zero time-integrated ozone uptake), and in their assumed threshold, below which the ozone uptake rate is considered sufficiently low that ozone will be detoxified before any damage occurs Pleijel et al., 2004;Lombardozzi et al., 2012a). For example, Sitch et al. (2007) relates the instantaneous ozone 25 uptake exceeding a flux threshold to net photosynthetic damage via an empirically derived factor. An alternative approach has been to relate ozone damage to net photosynthesis in response to the accumulated ozone uptake rather than to the instantaneous ozone uptake as in Sitch et al. (2007), e.g. by using the CU OY , which refers to the cumulative canopy O 3 uptake above a flux threshold of Y nmol m −2 s −1 (Wittig et al., 2007;Lombardozzi et al., 2012aLombardozzi et al., , 2013.
The effect of ozone on plant growth has been investigated by ozone filtration/fumigation experiments either at the indi-30 vidual experimental level or by pooling data from multiple experiments that have been conducted according to standardised experimental method. These experiments typically rely on small trees or saplings. A challenge in developing and testing process-based models of ozone damage from these ozone fumigation experiments is that often only the difference in biomass accumulation between plants grown in an ozone treatment and in ambient or charcoal-filtered air at the end of the experiment are reported. Data from these studies provide evidence for a linear, species-specific relationship between accumulated ozone uptake and reductions in plant biomass (Pleijel et al., 2004;Mills et al., 2011;Nunn et al., 2006, e.g.). Sitch et al. (2007) for instance calibrated their instantaneous leaf-level dose-response relationship between ozone uptake and photosynthesis by relating simulated annual net primary production and accumulated ozone uptake to observed biomass dose-response relationships developed by Karlsson et al. (2004) and Pleijel et al. (2004), where biomass/yield damage is related to the Phytotoxic 5 Ozone Dose (P ODy). The P ODy refers to the accumulated ozone uptake above a flux threshold of y nmol m −2 s −1 by the leaves representative of the upper canopy leaves of the plant. Such an approach applies biomass dose-response relationships of young trees to mature trees. However, the effects of ozone on leaf physiology (e.g. net photosynthesis and stomatal conductance) or plant carbon allocation may differ between juvenile and adult trees (Hanson et al., 1994;Samuelson and Kelly, 1996;Kolb and Matyssek, 2001;. Whether or not biomass dose-response relationships can be used to calibrate 10 dose-response functions for mature trees is uncertain. An alternative approach is to directly simulate ozone damage to photosynthesis, which may have been a major cause for the observed decline in plant biomass production (Ainsworth et al., 2012). Possible damage targets in the simulations can be for example the net photosynthesis or leaf-specific photosynthetic activity (such as represented by the maximum carboxylation capacity of Rubisco, V cmax ). For instance Lombardozzi et al. (2012a) based their dose-response relationships on an experi-15 mental study involving a single forest tree species, whereas more recent publications (e.g. Lombardozzi et al. (2015) andFranz et al. (2017)) have used dose-response relationships from meta-analyses of a far larger-set of filtration/fumigation studies.
Meta-analyses have attempted to summarise the responses of plant performance to ozone exposure across a wider range of experiments and vegetation types (Wittig et al., 2007;Lombardozzi et al., 2013;Feng and Kobayashi, 2009;Li et al., 2017;Wittig et al., 2009) and to develop damage functions for plant groups that might provide an estimate of mean plant group 20 responses to ozone. However, these meta-analyses suffer from a lack of consistency in the derivation of either plant damage or ozone exposure, and generally report a large amount of unexplained variance. A further complication in the meta-analyses of ozone damage (e.g. Wittig et al., 2007;Lombardozzi et al., 2013) is that they have to indirectly estimate the cumulative ozone uptake underlying the observed ozone damage based on a restricted amount of data, which causes uncertainty in the derived damage functions. 25 Büker et al. (2015) provides an independent data set of whole-tree biomass plant responses to ozone uptake which is independent of data sets that were used to describe damage functions by Wittig et al. (2007) and Lombardozzi et al. (2013).
This data set has been collected from experiments that follow a more standardised methodology to assess dose-responses and has associated meteorological and ozone data at a high time resolution that allow more accurate estimates of modelled ozone uptake to be made. These dose-response relationships describe whole-tree biomass reductions in tree seedlings derived from 30 standardised ozone filtration/fumigation methods for eight European tree species at ten locations across Europe (see Tab. A.2 for details Büker et al., 2015). These data thus provide an opportunity to evaluate simulations of biosphere models that use leaf level damage functions (describing the effect of ozone uptake on photosynthetic variables) to estimate C assimilation, growth and ultimately whole tree biomass against these robust empirical dose-response relationships that relate ozone exposure directly to whole tree biomass response. 35 Here we test four alternative, previously published ozone damage functions that target either net photosynthesis or the leaf carboxylation capacity (V cmax ), which have been included in state-of-the-art terrestrial biosphere models (Lombardozzi et al., 2012a(Lombardozzi et al., , 2015Franz et al., 2017) against these new biomass dose-response relationships by Büker et al. (2015). We incorporate these damage functions into a single modelling framework, the O-CN model (Zaehle and Friend, 2010;Franz et al., 2017).
To reduce model-data mismatch, we test the functions in simulations that mimic to the extend possible the conditions of 5 each of the experiments in the Büker et al. (2015) data-set, in particular the young age, such that we can directly compare the simulated to the observed whole-tree biomass reductions of the empirically derived dose-response relationships. This allows us to identify the contribution of these alternative damage function formulations on the simulated whole-tree biomass response. The simulated biomass dose-response relationships are then compared to the data from the experiments to evaluate the capability of the different model versions to reproduce observed dose-response relationships. Based on these comparisons 10 we use a similar approach to that of Sitch et al. (2007) and develop alternative parameterisations of the damage functions to improve the capability of the O-CN model to simulate the whole-tree biomass responses observed in the field experiments, with the notable exception that we explicitly simulate in-fumigation experiments and the approximate age of the trees. Finally, we explore whether or not there is a substantial difference in the biomass response to ozone of young or mature trees by using a sequence of model simulations and comparing the response both in terms of whole tree biomass as well as net primary 15 production.

Methods
We use the O-CN terrestrial biosphere model (Zaehle and Friend, 2010), which is an extension of the ORCHIDEE model (Krinner et al., 2005) to simulate conditions of the ozone fumigation experiments described in Büker et al. (2015). The O-CN model simulates the terrestrial coupled carbon (C), nitrogen (N) and water cycles for up to twelve plant functional types and is 20 driven by climate data and atmospheric composition.
O-CN simulates a multi-layer canopy with up to 20 layers with a thickness of up to 0.5 leaf area index each. Net photosynthesis is calculated according to a modified Farquhar-scheme for shaded and sun-lit leaves considering the light profiles of diffuse and direct radiation (Zaehle and Friend, 2010). Leaf nitrogen concentration and leaf area determine the photosynthetic capacity. Increases of the leaf nitrogen content increase V cmax and J max (nitrogen specific rates of maximum light harvest-25 ing, electron transport) and hence maximum net photosynthesis and stomatal conductance per leaf area. The leaf N content is highest at the top of the canopy and exponentially decreases with increasing canopy depth. Following this net photosynthesis, stomatal conductance and ozone uptake are generally highest in the top canopy and decrease with increasing canopy depth.
Canopy-integrated assimilated carbon enters a labile non-structural carbon pool, which can either be used to fuel maintenance respiration (a function of tissue nitrogen), storage (for seasonal leaf and fine root replacement and buffer of inter-annual 30 variability of assimilation) or biomass growth. After accounting for reproductive production (flowers and fruits), biomass growth is partitioned into leaves, fine roots, and sapwood according to a modified pipe-model (Zaehle and Friend, 2010), accounting for the costs of biomass formation (growth respiration). In other words, changes in leaf-level productivity affect the build-up of plant pools and storage, and thereby feed back on the ability of plants to acquire C through photosynthesis, or nutrients through fine root uptake.

Ozone damage calculation in O-CN
Leaf-level ozone uptake is determined by stomatal conductance and atmospheric O 3 concentrations, as described in  (Ball et al., 1987) as where net photosynthesis (A n,l ) is calculated as described in Zaehle and Friend (2010) as a function of leaf nitrogen and nitrogen specific rates of maximum light harvesting, electron transport (J max ) and carboxylation rates (V cmax ). RH is the atmospheric relative humidity, f (height l ) the water-transport limitation with canopy height, C a the atmospheric CO 2 concentration, g 0 is the residual conductance when A n approaches zero, and g 1 is the stomatal-slope parameter as in Krinner et al. (2005). The index l indicates that g st is calculated separately for each canopy layer.

15
The O 3 stomatal flux (f st,l , nmol m −2 (leaf area) s −1 ) is calculated from the atmospheric O 3 concentration the plants in the field experiments were fumigated with (χ O3 atm ) and g st,l as where the leaf internal O 3 concentration (χ O3 i ) is assumed to be zero (Laisk et al., 1989). The accumulation of ozone fluxes above a threshold of Y nmol m −2 (leaf area) gives the CU OY l . The canopy value of CU OY is calculated by summing CU OY l over all canopy layers (see Franz et al. (2017) for details).
For comparison to observations, the Phytotoxic Ozone Dose (P OD, mmol m −2 ) can be diagnosed by the accumulation 25 of f st,l for the top canopy layer (l = 1). The accumulation of ozone fluxes of the top canopy layer above a threshold of y nmol m −2 (leaf area) s −1 gives the P ODy. The estimates of P ODy (both P OD2 and P OD3) can be used off-line to re-construct dose-response relationships equivalent to those described in Büker et al. (2015). These modelled dose-response relationships can then be compared with the empirically derived dose-response relationships to assess the ability of the model to estimate damage. As such, the P OD2 and P OD3 used for the formation of these modelled dose-response rela- Ozone damage, i.e. the fractional loss of carbon uptake associated with ozone uptake d O3 l , is calculated as a linear function of the cumulative leaf-level uptake of ozone above a threshold of Y nmol m −2 (leaf area) s −1 (CU OY l ) where a is the intercept and b is the slope of the damage function. The damage fraction (d O3 l ) is calculated separately 10 for each canopy layer l based on the specific accumulated ozone uptake of the respective canopy layer (CU OY l ), and takes values between 0 and 1. The magnitude of d O3 l in Eq. 4 varies between the canopy layers because CU OY l varies driven by within-canopy gradients in stomatal conductance and photosynthetic capacity.
The effect of ozone damage on plant carbon uptake is calculated by 15 where x is either leaf-level net photosynthesis A n,l or the maximum photosynthetic capacity (J max,l and V cmax,l ), which is used in the calculation of A n,l . J max,l and V cmax,l are reduced in proportion such that the ratio between the two is not altered.
While there is some evidence that ozone can affect the ration between J max and V cmax , we believe that for the purpose of this paper is is justifiable to assume a fixed ratio between them.
Reductions in A n , l cause a decline in stomatal conductance (g st,l ) due to the tight coupling between both. Other stress 20 factors that impact g st,l are accounted for in the preceding calculation of the g st,l undamaged by ozone (see Eq. 1). Reductions in g st,l decrease the O 3 uptake into the plant (f st,l ) and slow the increase in CU OY l and thus ozone damage.

Model set-up
Four published damage functions were applied within the O-CN model (see Tab. 1 for the respective slopes, intercepts and flux thresholds). As these did not match well with the observed biomass dose-response relationships by Büker et al. (2015), 25 we calibrated two additional damage relationships, one each for A n or V cmax , based on the data presented in Büker et al. instantaneous ozone uptake). As described above, in all model versions, ozone damage is calculated independently for each canopy layer based on the accumulated O 3 uptake (CU OY l ) in that layer, above a specific flux threshold of Y nmol m −2 (leaf area) s −1 for the respective damage function (see Tab. 1). Simulations were run for each fumigation experiment using meteorological input from the daily CRU-NCEP climate data set

Calculation of the biomass damage relationships
At each experiment site and for all treatments the annual reduction in biomass due to ozone (RB) is calculated as in Büker 30 et al. (2015): Separate biomass dose-response relationships were estimated by grouping site data for broadleaved and needleleaved species.
The biomass dose-response relationships are obtained from the simulation output by fitting a linear model to the simulated val-20 ues of RB and P ODy dr (with flux thresholds of 2 and 3 nmol m −2 (leaf area) s −1 for needleleaved and broadleaved species, respectively), where the regression line is forced through one at zero P ODy dr . Büker et al. (2015) report two alternative doseresponse relationships for their data set, the simple and the standard model, B SI and B S T respectively. We evaluate our different model versions regarding their ability to reach, with the biomass-dose-response relationships computed from their output, the area between those two functions (target area). The tuned damage relationships tun P S and tun V C were obtained by adjust-

25
ing the slope b in Eq. 4 such that the corresponding biomass dose-response relationships fits the target area. The intercept of the damage relationships are forced to 1 to simulate zero ozone damage at ozone fluxes lower than 1 nmol m −2 (leaf area) s −1 .

Testing published damage functions
None of the versions where ozone damage is calculated based on previously published damage functions fit the observations well. Some versions strongly overestimate the simulated biomass dose-response relationship and others strongly underestimate it (see Fig. 1) compared to the dose-response relationships developed by Büker et al. (2015). In the W07 P S simulations, where damage is calculated based on the damage function by Wittig et al. (2007), biomass damage is strongly underestimated compared to the estimates from Büker et al. (2015). Ozone damage estimates are mainly driven by the intercept of the relationship, which assumes a reduction of net photosynthesis by 6.16% at zero ozone uptake.
Little additional ozone damage occurs due to the accumulation of ozone uptake. As a consequence, the ozone treatments and reference simulations differ little in their simulated biomass. Similarly, the Lombardozzi et al. (2013) damage function (L13 P S ) 10 calculates ozone damage as a fixed reduction of net photosynthesis independent of the actual accumulated ozone uptake. The reference simulations with zero atmospheric ozone thus equals the simulations with ozone treatments and results in an identical simulated biomass. We tested accounting for effects of ozone on stomatal conductance besides net photosynthesis as suggested by Lombardozzi et al. (2013). However, this additional direct damage to stomatal conductance yielded a minimal decrease in simulated biomass accumulation in needle-leaved trees, but did not qualitatively change the results (results not shown). These We investigated the cause for this at the example of the Pinus halepensis stand in the Ebro Delta with a high ozone treatment 10 as shown in Fig. 2. The CU OY quickly increases after the onset of fumigation (Fig. 2a) and is paralleled by a rapid decline in canopy integrated net photosynthesis (A can n , see Fig. 2b). Once all canopy layers accumulated more than 5 mmol O 3 m −2 , the canopy photosynthesis is fully reduced, and A can n becomes negative as a consequence of ongoing leaf maintenance respiration. Thereafter, leaf and total biomass steadily decline (Fig. 2c,d), and the plants are kept alive only by the consumption of stored non-structural carbon reserves. Despite the 100 % reduction in gross photosynthesis, the biomass compared to a 15 control simulation (relative biomass, RB) reaches only values of approximately 0.7 (Fig. 2e), because of the remaining woody and root tissues (see Eq. 6 for the calculation of RB).

Tuned damage relationships
We next tested whether a linear damage function is in principle able to reproduce the observed biomass dose-response relationships. Simulations conducted with our tuned damage relationships produce biomass dose-response relationships which fit the target area defined by the B SI and B ST dose-response relationships by Büker et al. (2015) (see Fig. 3 and Tab. A.5, A.6).
For the calibrated relationships used in these simulations, we chose a flux threshold value of 1 nmol m −2 (leaf area) s −1 , as 5 suggested by LRTAP-Convention (2017). We forced the intercept (a) of these relationships through 1, to simulate zero ozone damage at ozone fluxes lower than 1 nmol m −2 (leaf area) s −1 . The resulting slope of the tun P S function for broadleaved PFTs is approximately 30 times higher compared to the slope suggested by Wittig et al. (2007) and a fourth of the slope by Lombardozzi et al. (2012a). For the needle-leaved PFT, the tuned slope (tun P S ) is approximately 10 times higher (lower) than the slopes by Wittig et al. (2007) and Lombardozzi et al. (2012a), respectively. Notably, we did not observe any difference in 10 the model performance irrespective of whether net photosynthesis or photosynthetic capacity (V cmax ) was reduced.

Ozone damage to mature trees
The simulation of young trees (simulated as in the previous section) compared to adult trees with the same model version reveals a distinct difference between the simulated versus observed dose-response relationship when expressed as reduction of biomass. Ozone damage causes a much shallower simulated biomass dose-response relationship for adult trees (tun mature V C in 15 Fig. 4a,b) compared to young trees (tun young V C in Fig. 4a,b), both for broadleaved and needle-leaved species. It is worth noting that this is primarily the consequence of the higher initial biomass of the adult trees before ozone fumigation starts (tun mature V C ).
Comparing the dose-response relationship of young and mature trees based on the annual net biomass production (NPP) shows nearly identical slopes for needle-leaved species (Fig. 4d and Tab. 3), whereas the slopes for broadleaved tree species (Fig. 4c 12 Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-358 Manuscript under review for journal Biogeosciences Discussion started: 27 July 2018 c Author(s) 2018. CC BY 4.0 License. and Tab. 2) suggests only a slightly lower reduction in NPP in mature compared to young trees, likely related to the larger amount of non-structural reserves that increases the resilience of mature versus young trees.  Damage functions that relate accumulated ozone uptake to fundamental plant processes such as photosynthesis are a key component for models that aim to estimate the potential impacts of ozone pollution on forest productivity, growth and carbon sequestration. We tested four published damage functions for net photosynthesis and V cmax within the framework of the O-CN model to assess their ability to reproduce the empirical whole tree biomass dose-response relationships derived by Büker et al. Meta-analyses (Wittig et al., 2007;Lombardozzi et al., 2013) are designed to minimise the effect of species-specific ozone sensitivities and provide estimates of the average species response. However, we found that the relationships derived by these meta-analyses substantially underestimate biomass damage. Technically, the reasons for this are a weak or non-existent increase of the ozone damage with increased ozone uptake (shallow or non-existent slopes) and/or high ozone damage at zero accumulated ozone uptake (intercept lower than one). Apparently, the diversity of species responses and experimental settings that are assembled in the meta-analyses by Wittig et al. (2007) and Lombardozzi et al. (2013), together with uncertainties in precisely estimating accumulated ozone uptake in these databases preclude the identification of damage functions that are consistent with the damage estimates by Büker et al. (2015). The high intercepts in the meta-analyses by Wittig et al. (2007) and Lombardozzi et al. (2013), which assume a considerable damage fraction even when no ozone is taken up at all, seem to 5 be ecologically illogical and suggest that an alternative approach is necessary to simulate ozone damage. As a consequence of these points, the Europe-wide GPP reduction estimates by Franz et al. (2017), which has been based on the damage function by Wittig et al. (2007), may substantially underestimate actual GPP reduction. Similarly, global estimates as well as spatial variability of ozone damage to GPP by Lombardozzi et al. (2015), based on Lombardozzi et al. (2013), are virtually independent of actual ozone concentrations or uptake for all tree plant functional types and should be interpreted with caution.

10
A crucial aspect in forming dose-response relationships is the calculation of the accumulated ozone uptake (e.g. P ODy or CU OY ). The calculation of accumulated ozone uptake is realised in different ways in the meta-analyses and the study by Büker et al. (2015) as well as in our approach here. Experiments synthesised in the meta-analyses generally do not have access to stomatal conductance values at high resolution measured throughout the experiment, which impedes precise determination of O 3 uptake. The uncertainty in the necessary approximations of accumulated ozone uptake can be assumed to be considerable, 15 and it is thus highly recommendable to measure and report required observations in future ozone fumigation experiments. Büker et al. (2015) use the DO 3 SE model to simulate ozone uptake and accumulation similar as done in our model here. These modelled values for ozone uptake and accumulation can assumed to be more reliable since both models simulate processes that determine ozone uptake continuously for the entire experiment length at high temporal resolution. They account for diurnal changes in stomatal conductance as well as climate factors restricting stomatal conductance and hence ozone uptake. However, 20 both models vary in their complexity of the simulated plants, carbon assimilation, and growth processes, which will also impact the estimate of ozone accumulation.
The meta-analyses do not account for non-stomatal ozone deposition (e.g. to the leaf cuticle or soil), which imposes a bias towards overestimating ozone uptake and accumulation contrary to the DO 3 SE model used by Büker et al. (2015), which accounts for this. The O-CN model in principle can simulate non-stomatal ozone deposition from the free atmosphere to 25 ground level (see Franz et al. (2017)). The leaf boundary layer is implicitly included into the calculation of the aerodynamic resistance of O-CN and included in Franz et al. (2017). However, for the simulation of the chamber experiments we used the observed chamber O 3 concentrations, rather than estimating the canopy-level O 3 concentration based on the free atmosphere (approximately 45 m above the surface) and atmospheric turbulence. This required not accounting for aerodynamic resistance and therefore the leaf-boundary layer resistance as well as it prevented the calculation of the non-stomatal deposition, which 30 may lead to a slight overestimation of ozone uptake and accumulation in our simulations.
The calibration of damage functions to net photosynthesis and V cmax shows that in principle, the linear structure of Eq. 4 is sufficient to simulate biomass dose-response relationships comparable to Büker et al. (2015) in O-CN. An advantage of the damage functions derived here compared to previously published damage functions (Wittig et al., 2007;Lombardozzi et al., 2012aLombardozzi et al., , 2013 is the intercept of one, implying that simulated ozone damage is zero at zero accumulated O 3 and steadily Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-358 Manuscript under review for journal Biogeosciences Discussion started: 27 July 2018 c Author(s) 2018. CC BY 4.0 License. increases with increased ozone accumulation. The flux threshold used in the simulations is 1 nmol m −2 (leaf area) s −1 as suggested by the LRTAP-Convention (2017). Since the tuned damage functions are structurally identical to previously published damage functions based on accumulated ozone uptake they can be directly compared to them. Slopes of the tuned damage functions lie in between the values proposed by Wittig et al. (2007) and Lombardozzi et al. (2012a) and thus take values in an expected range. We did not find any significant difference in simulated biomass responses between the use of net photosynthesis 5 or leaf-specific photosynthetic capacity (V cmax ) as a target for the ozone damage function, although we do note that the slopes were slightly lower for the net photosynthesis based functions. The simulation of ozone effects on leaf-specific photosynthetic capacity (V cmax ) seems preferable over the adjustment of net photosynthesis, because V cmax and J max are parameters in the calculation of net photosynthesis, and thus are likely easier transferable between models. Models with different approaches to simulate net photosynthesis might obtain better comparable results by using damage relationships that target V cmax instead of 10 net photosynthesis.
All damage functions included into the O-CN model base damage calculations on the damage index CU OY (canopy value) rather than P ODy, as used by some other models, e.g. the DO 3 SE model (Emberson et al., 2000). We tested the effect of basing the damage calculation on P OD1 rather than CU O1, and found that these produced comparable biomass dose-response relationships as the damage relationships based on CU O1 presented in Fig. 3 (results not shown). The slopes of damage 15 functions based on P OD1 are approximately two thirds and half compared to the slopes based on CU O1 for broadleaved and needle-leaved species respectively. The difference in the slope values associated with P OD1 and CU O1 results from the different calculation and application of them. The P OD1 is calculated in the top canopy layer and the respective damage fraction is applied for all canopy layers, the CU OY though is calculated separately in each canopy layer as well as the respective damage fraction. Higher frequency data on the ozone damage incurred by plants are required to disentangle whether 20 an ozone damage parameterisation based on instantaneous or accumulated ozone uptake results in more accurate simulation of the seasonal effects and more analysis of the differential effect of ozone damage within deep canopies are required to evaluate whether the scaling of top-of-the-canopy damage to whole canopy damage is appropriate.
Further aspects that determine ozone sensitivity and damage to carbon gain of plants like leaf morphology (Calatayud et al., 2011;Bussotti, 2008), different sensitivity of sunlit and shaded leafs (Tjoelker et al., 1995;Wieser et al., 2002), early 25 senescence (Gielen et al., 2007;Ainsworth et al., 2012) and costs for detoxification of ozone and/or repair of ozone damage that likely increases the plant's respiration costs (Dizengremel, 2001;Wieser and Matyssek, 2007) are not considered by either approach. Marzuoli et al. (2016) observed an ozone induced reduction of biomass but no significant reduction in physiological parameters like V cmax . They suggest that the reduced growth is caused by higher energy investments and reducing power for the detoxification of ozone whereas the photosynthetic apparatus remained undamaged (Marzuoli et al., 2016).
Some studies have found that ozone-affected stomata respond much slower to environmental stimuli than unaffected cells (Paoletti and Grulke, 2005), which can delay closure and trigger, stomatal sluggishness, an uncoupling of stomatal conductance and photosynthesis (Reich, 1987;Tjoelker et al., 1995;Lombardozzi et al., 2012b) and thus impact transpiration rates (Mills et al., 2009;Paoletti and Grulke, 2010;Lombardozzi et al., 2012b) and the plant's water use efficiency (Wittig et al., 2007;Mills et al., 2009;Lombardozzi et al., 2012b). The O-CN model is able to directly impair stomatal conductance, by uncoupling  (Lombardozzi et al., 2013), have a negligible impact on biomass production compared to not accounting for direct damage to the stomata (results not shown). However, our above mentioned concerns regarding the structure of the damage 5 relationships by Lombardozzi et al. (2013) should be taken into account when considering this result.
A key challenge for the use of fumigation experiments to parameterise ozone-damage in models is that trees (as opposed to grasses fumigated from seeds) typically possess a certain amount of biomass at the beginning of the fumigation experiment.
Even at lethal ozone doses, the relative biomass thus can not decline to zero, and tree death may occur at values of a relative biomass greater than zero. The relative biomass is positive even if carbon fixation is fully reduced and the plants survive due to 10 the use of stored carbon. The higher the initial biomass and the slower the annual biomass growth rate of the tree is, the harder it is to obtain low values of RB. When comparing RB values obtained from trees with substantial different initial biomass and tree species with different growth rates proportionate damage rates thus can not directly be inferred. This indicates that the explanatory value of the relative biomass between a control and a treatment to estimate long-term plant damage at a given O 3 concentration is limited. This is particularly the case when evaluating the damage of more mature forests. The simulated 15 biomass dose-response relationships of adult trees are much more shallow than dose-response relationships of young trees (see Fig. 4), because of the high initial biomass prior to fumigation. This suggests that the use of biomass damage functions derived from experiments with young trees to parameterise the biomass loss of adult trees, as done in Sitch et al. (2007), will likely lead to an overestimation of plant damage and loss of carbon storage. Dose-response relationships based on biomass increments or growth rates might be better transferable between saplings and mature trees and hence better suitable to be used 20 for parameterising global terrestrial biosphere models.
Our approach to overcome this challenge was to alter the vegetation model to simulate the ozone damage of small trees, where we could directly compare simulated biomass reductions to observations. Since we used damage relationships that are based on the calculation of leaf-level photosynthesis, we are able to apply the calibrated model also for mature stands. Our simulations have demonstrated that despite the different sizes of young and mature trees, and associated changes in the wood 25 growth rate and the available amount of non-structural carbon reserves to repair incurred damage, the simulated effect of ozone on the net annual biomass production (NPP) was very similar, when using a damage function associated with leaf-level photosynthesis. Overall our findings support the idea that the photosynthesis-based damage relationships developed here and evaluated against fumigation experiments of young trees, might be useful to estimate effect on forest production of older trees.
Monitoring approaches that are either capable of measuring the actual increment of biomass, or quantify at the leaf and canopy 30 level the change in net photosynthesis over the growing season, would allow to develop damage estimates that could be more readily translated into modelling frameworks.
Terrestrial biosphere models in general assume that plant growth is primarily determined by carbon uptake. However, an alternative concept proposes that plant growth is more limited by direct environmental controls (temperature, water and nutrient availability) than by carbon uptake and photosynthesis (Fatichi et al., 2014). The O-CN model provides a first step into this 35 Biogeosciences Discuss., https://doi.org/10.5194/bg-2018-358 Manuscript under review for journal Biogeosciences Discussion started: 27 July 2018 c Author(s) 2018. CC BY 4.0 License. direction because it separates the step of carbon acquisition from biomass production, both in terms of a non-structural carbon buffer, as well as a stoichiometric nutrient limitation on growth independent of the current photosynthetic rate. This would in principle allow to account for ozone effects on the carbon sink dynamics within plants. However, it is not clear that data readily exist to parameterise such effects. Given the availability of suitable data to parameterise a large scale model, ozone damage might be better simulated by targeting biomass growth rates or processes that limit these e.g. stomatal conductance, which 5 impacts the plants water balance compared to our approach here, which targets net photosynthesis.
All in all a multitude of aspects that impact ozone damage to plants is not yet incorporated into global terrestrial biosphere models. The ongoing discussion which processes are major drivers for observed damages, how they interact and impact different species and plant types plus the lack of suitable data needed to parameterise a global model are reasons why the simulation of ozone damage up to now focuses only on a few aspects where suitable data are available as presented in our study.

Conclusion
The inclusion of previously published dose-response relationships into the terrestrial biosphere model O-CN led to a strong over-or underestimation of simulated biomass damage compared to the biomass dose-response relationship by Büker et al. (2015). Dose-response relationships which are used as damage functions in terrestrial biosphere models are a key aspect in the simulation of ozone damage and have a great impact on the estimated damage. The calibration of damage functions performed 15 in this study provide the advantage to calculate ozone damage close to where the actual physiological damage might occur (photosynthetic apparatus) and simultaneously reproduce observed biomass damage relationships for a range of European forest species used by Büker et al. (2015). The inclusion of these damage functions into models that estimate regional or global ozone damage might improve damage estimates compared to previously published damage functions and might lead to better estimates of terrestrial carbon sequestration. The comparison of simulated biomass dose-response relationships of young and 20 mature trees shows strongly different slopes. This suggests that observed biomass damage relationships from young trees might not be suitable to estimate biomass damage of mature trees. The comparison of simulated NPP dose-response relationships of young and mature trees show similar relationships and suggests that they might more readily be transferred between trees differing in age.