Current and future impacts of drought and ozone stress on Northern Hemisphere forests

Rising ozone (O3) concentrations, coupled with an increase in drought frequency due to climate change, pose a threat to plant growth and productivity which could negatively affect carbon sequestration capacity of Northern Hemisphere (NH) forests. Using long‐term observations of O3 mixing ratios and soil water content (SWC), we implemented empirical drought and O3 stress parameterizations in a coupled stomatal conductance–photosynthesis model to assess their impacts on plant gas exchange at three FLUXNET sites: Castelporziano, Blodgett and Hyytiälä. Model performance was evaluated by comparing model estimates of gross primary productivity (GPP) and latent heat fluxes (LE) against present‐day observations. CMIP5 GCM model output data were then used to investigate the potential impact of the two stressors on forests by the middle (2041–2050) and end (2091–2100) of the 21st century. We found drought stress was the more significant as it reduced model overestimation of GPP and LE by ~11%–25% compared to 1%–11% from O3 stress. However, the best model fit to observations at all the study sites was obtained with O3 and drought stress combined, such that the two stressors counteract the impact of each other. With the inclusion of drought and O3 stress, GPP at CPZ, BLO and HYY is projected to increase by 7%, 5% and 8%, respectively, by mid‐century and by 14%, 11% and 14% by 2091–2100 as atmospheric CO2 increases. Estimates were up to 21% and 4% higher when drought and O3 stress were neglected respectively. Drought stress will have a substantial impact on plant gas exchange and productivity, off‐setting and possibly negating CO2 fertilization gains in future, suggesting projected increases in the frequency and severity of droughts in the NH will play a significant role in forest productivity and carbon budgets in future.


Tropospheric ozone (O 3 ) concentrations have doubled in the
Northern Hemisphere (NH) since the pre-industrial period (Yeung et al., 2019) and are currently increasing at a rate of 0.5%-2% per year due to changes in the release of precursor compounds from industrial activities (Gaudel et al., 2018;Hartmann et al., 2013). By the end of this century, NH tropospheric O 3 could increase by as much as 18% (Young et al., 2013) and drought frequency by 50%-200% (Zhao & Dai, 2017). Surface O 3 is a powerful phytotoxin (Ainsworth, Yendrek, Sitch, Collins, & Emberson, 2012;Ashmore, 2005). It enters leaves through the stomata and damages cell membranes, proteins and DNA through oxidation reactions (Leisner & Ainsworth, 2012;Omasa & Takayama, 2002). O 3 damages the photosynthetic apparatus affecting leaf gas exchange, leading to reductions in plant productivity, growth and biomass accumulation (Ainsworth et al., 2012;Paoletti, 2009).
Plants can respond to O 3 -induced oxidative stress by closing stomata (an avoidance strategy), thus limiting water loss and stomatal O 3 flux, and by synthetizing antioxidants (a tolerance strategy) to regulate reactive oxygen species levels (Andersen, 2003;Pellegrini et al., 2019). Both tolerance and avoidance can be parameterized in vegetation models. The former assumes that plants can detoxify limited doses of O 3 , thus reducing the oxidative stress. Such a pathway has been extensively described by several authors in the phytotoxic O 3 dose POD y metric Emberson, Büker, & Ashmore, 2007;Mills, Hayes, et al., 2011;Mills, Pleijel, et al., 2011). In broad terms, the POD y represents the cumulative quota of O 3 that a plant is not able to detoxify, and that is consequently harmful to the plant's ecophysiological processes. This approach has been shown to perform well across a variety of ecosystems in modelling studies (Clark et al., 2011;Sitch, Cox, Collins, & Huntingford, 2007). The latter strategy assumes that plants regulate stomata by directly reducing the exposure of internal plant tissues to O 3 . It has been observed in many experiments that plants fumigated to high concentration of O 3 exhibit a general decrease in stomatal conductance (Wittig, Ainsworth, & Long, 2007). Hoshika, Watanabe, Inada, and Koike (2013) recently hypothesized that plants can optimize their stomatal behaviour to minimize O 3 influx and transpiration while maximizing carbon assimilation, and they reparameterized the optimal stomatal behaviour model developed by Medlyn et al. (2011).
This optimal stomatal behaviour theory has also been shown to improve model estimates of photosynthesis and stomatal conductance on different seedling species in field experiments (Hoshika, Watanabe, et al., 2013) but has not been widely applied.
The complexity of modelling O 3 and drought stress impacts on vegetation is compounded by the differing levels of sensitivity of different ecosystems. Mediterranean climates are characterized by high temperature, strong insolation and prolonged drought during the summer, conditions which promote photochemical tropospheric O 3 formation (Millán et al., 2000;Paoletti, 2006). These conditions are expected to increase in frequency and intensity in future (IPCC, 2013). Vegetation in this region has developed adaptations to such stresses, for example, leaf morphology, water conservation by reduced transpiration and the synthesis and emission of biogenic volatile organic compounds including powerful antioxidants and compatible solutes (Calfapietra, Fares, & Loreto, 2009;Nali et al., 2004;Paoletti, 2006), and may therefore be better able to tolerate such stressors. By contrast, Boreal climates have mild wet summers and cold winters, leading to generally low O 3 concentrations and infrequent droughts. Hence, Boreal forests have not developed strategies to avoid or tolerate either stress and may be more vulnerable to damage than Mediterranean forest ecosystems. These contrasting characteristics make Mediterranean and Boreal ecosystems ideal for testing the effect of droughts and O 3 on NH forests. As they also make up 9.4% (M'Hirit, 1999) and 17% (Kasischke, 2000) of the Earth's land surface area, respectively, changes in their productivity could have major implications for the global carbon cycle.
Vegetation models play an important role in predicting likely impacts of climate change on forest productivity, but confidence in future projections is dependent on their performance when evaluated against present-day observations. We test the skill of a one-di-

| FORCAsT model
FORCAsT is a 1D model of biosphere-atmosphere chemical exchange which has previously been used to study canopy structure and mixing (Bryan et al., 2012, stomatal regulation and atmospheric chemistry within and above forest canopies (Ashworth et al., , 2016 and the impact of drought stress on biogenic volatile organic compound emissions and forest gas exchange (Otu-Larbi, Bolas, et al., 2020). A full description of the FORCAsT model can be found in Ashworth et al. (2015). Three different coupled photosynthesis-stomatal conductance (A-g s ) models have since been incorporated into FORCAsT giving users the flexibility to select the most appropriate for the ecosystem of interest and the meteorological and physiological observations available (see Otu-Larbi, Conte, et al., 2020 [in preparation] for full details).
Here, we describe the parameterizations of drought and O 3 stress used in this study. We apply the Medlyn et al. (2011) optimal stomatal behaviour modification of the Farquhar, Von Caemmerer, and Berry (1980) photosynthesis model in which photosynthesis rate (A; μmol m −2 s −1 ) is the minimum of two limiting factors: electron transport and carboxylation rate. Stomatal conductance (g s ) is modelled assuming that stomatal aperture is regulated to maximize carbon gain while simultaneously minimizing water loss (Medlyn et al., 2011): where g o (mol m −2 s -1 ) is the residual stomatal conductance when A approaches zero and g 1 is a fitted parameter representing the sensitivity of g s to A. The values of g o and g 1 are determined at the speciesor plant functional type (PFT)-specific level from experimental data.
Here, we use values obtained from Lin et al. (2015) and De-Kauwe et al. (2015), respectively, as indicated in Table S1. D (kPa) is the vapour pressure deficit calculated by FORCAsT and C s (μmol/mol) is the CO 2 concentration at the leaf surface. LE (W/m 2 ) is estimated following Lhomme, Elguero, Chehbouni, and Boulet (1998) where ρ (kg/m 3 ) is the air density, C p (J kg −1 K −1 ) is the specific heat capacity of air at constant pressure, γ (kPa/K) is the psychrometric constant (the ratio of C p to latent heat of vaporization of water), e s and e a (kPa) are the saturated vapour pressure at leaf temperature and the air water vapour pressure, respectively, and ge v (m/s) is an equivalent conductance for horizontal vapour transfer estimated as: where LAI i (m 2 /m 2 ) is the leaf area index at model layer i, g bw (mol m −2 s −1 ) and g sw (mol m −2 s −1 ) are the leaf boundary layer and stomatal conductance to water respectively.

| Soil moisture stress
Accounting for drought stress impacts on plants in vegetation models is challenging. The response depends on soil characteristics, climatic conditions and PFT. Metrics based on SWC, soil water potential and predawn leaf water potential have all been developed to assess plant water status (e.g. see Keenan et al., 2010;Zhou et al., 2014). Predawn leaf water potential provides the best measure of plant water status, but the lack of long-term observations makes these metrics difficult to apply in modelling studies. In contrast, SWC, while not as robust, is measured at most forest sites and can also be derived from satellite data making it easier to use in model parameterizations and simulations.
In this study, the effect drought stress on A and g s is assumed to be the result of biochemical and stomatal limitations as demonstrated in previous studies (e.g. see Egea et al., 2011). A soil moisture stress function was incorporated into the photosynthesis module in FORCAsT as described by Otu-Larbi, Conte, et al. (2020). The stress function, β, ranges between 1 (in the absence of drought stress) and 0 (at wilting point) and is calculated from: where θ (m 3 /m 3 ) is the volumetric soil moisture, θ w is the wilting point (m 3 /m 3 ) and θ c is a critical soil moisture content above which drought stress is found not to affect plant-atmosphere gas exchange (Egea et al., 2011;Keenan et al., 2010). q is a site-specific empirical factor describing the non-linearity of the effects of soil drought stress on tree physiological processes. θ c , θ w and q were calculated from soil texture data (i.e. sand, clay and silt fractions) or calibrated using long-term soil moisture observations at each site as detailed in Otu-Larbi, Conte, et al., 2020 and provided in Table S1.
The water-stressed values of carboxylation (V cmax* ) and electron transport (J max* ) rate are then calculated from the maximum rates (V cmax and J max ) as: and these values are applied to calculate the impact of soil moisture deficit on photosynthesis. The stomatal conductance then becomes:

| Incorporating O 3 damage
The reduction in photosynthesis and plant productivity due to O 3 cellular damage is incorporated into FORCAsT following two assumed strategies.

| O 3 avoidance (AVD)
O 3 avoidance (stomatal closure) follows Hoshika, Watanabe, et al. In the optimal stomatal behaviour theory, the control of leaf gas exchange may be considered optimal when it maximizes carbon gain while simultaneously minimizing water loss. Assuming stomata act to minimize O 3 damage in a similar manner, then the optimal stomatal conductance can be found from a modification of Equation (

| O 3 tolerance (TLR)
Plants' strategy to tolerate O 3 consists of enzymatic processes and chemical reactions to detoxify photooxidants. O 3 -tolerant trees (e.g. Pinus strobus) have been shown to have higher glutathione reductase and ascorbate peroxidase than O 3 -sensitive species (Chevone, 1991).
This prevents oxidative damage to the photosystem, enabling plants to maintain photosynthesis at higher doses of O 3 . Here, we assume that the instantaneous uptake of O 3 by plants only leads to an immediate suppression of leaf photosynthesis above a critical stomatal O 3 flux threshold. The decrease in leaf photosynthesis from its potential maximum is therefore proportional to the flux above that critical flux.
The reduction factor, F, is calculated following Pleijel et al. (2004) as: where A and g s are the (potential) photosynthesis rate and stomatal conductance in the absence of O 3 .

| Scaling up to the canopy
GPP is estimated as: where A n (μmol m −2 s −1 ) is the net photosynthesis (including the effects of drought and O 3 stress) and R d (μmol m −2 s −1 ) is the canopy dark respiration which is estimated by the model. Leaf-level A n , GPP and LE in each layer of the canopy (i) were scaled by LAI at each model level (LAI i ) and summed over all model layers (n) to obtain canopy-scale (c) estimates of A, GPP and LE.  Sorooshian, Li, Hsu, & Gao, 2012). These sites are part of the FLUXNET network (Pastorello et al., 2017). Full details of the sites, and the data and model parameters used are provided in Table S1.

Observations of photosynthetically active radiation (PAR;
µmol m −2 s −1 ), air temperature (K), CO 2 concentration (ppm), volumetric SWC (m 3 /m 3 ), wind speed (m/s) and direction (degrees clockwise from North), relative humidity (RH; %) and atmospheric pressure (Pa) were obtained for each site from the FLUXNET-2015 data set at a temporal resolution of 30 min. O 3 data were obtained directly from site lead investigators. The number of years for which data are available at each site is given in Table S1.  (Köppen, 1923). Precipitation mainly occurs in autumn and winter with little or none in the summer, resulting in annual droughts. Average soil moisture ( Figure 1) drops from 0.20 m 3 /m 3 in the winter and spring to ~0.10 m 3 /m 3 during

| Impact of future changes in SWC and O 3 concentrations
We investigate the potential impacts of climate change on GPP and see https://portal.enes.org/data/enes-model -data/cmip5/ resol ution for a list of models and their characteristics). Only the seven models (from five modelling centres) that provide both O 3 mixing ratios and SWC were selected. Details of these are provided in Table S2.
Variables were obtained from historical GCM simulations for 1850-2005 and GCM future simulations for 2006-2100 following RCP8.5, a scenario in which emissions of CO 2 follow an exponential growth trajectory throughout the century (Riahi et al., 2011), with concentrations increasing to 936 ppm and nominal anthropogenic forcing to 8.5 W/m 2 by 2100 (IPCC, 2014).
Comparing historical model output and observations shows systematic (but differing) biases in all seven models (see Figure S1). We

| Model configurations and experiments
We evaluate FORCAsT performance and determine the most suit- for the analysis because 2006 was a drought year (Gao et al., 2016) and therefore allows for assessment of drought impact. An evaluation of FORCAsT performance at HYY for the entire 1997-2014 period is shown in Figure S5.

| RE SULTS
Droughts occur almost annually at BLO and CPZ but rarely at HYY, as shown in Figure Table 1 and Figure 2 show the annual average observed and simulated GPP and LE for each site calculated for each 2-year simulation period. Under present-day conditions, CPZ and BLO are more productive than HYY; observed GPP at HYY was about half of that observed at CPZ and ~70% of that at BLO. LE at BLO was approximately 35% and 60% higher than the observed values at CPZ and HYY respectively.

| Current impacts of drought and O 3 on GPP and LE
In general, FORCAsT overestimated GPP and LE across all three sites in CTR simulations when the effects of stress were excluded.
Model overestimation was higher when drought stress was excluded (CTR) than O 3 stress irrespective of whether TLR or AVD was assumed.
Drought stress has a greater impact on model estimates of GPP and LE at CPZ and BLO than at HYY due to the lower SWC and frequent drought at these sites. At CPZ and BLO, the inclusion of drought stress alone in FORCAsT (CTR + Dr) led to a 20% average reduction in model overestimation of GPP and LE but only a 10% reduction at HYY. For example, while CTR + Dr and TLR led to 22% and 11% reductions in GPP, respectively, their combined effect (TLR + Dr) was ~5% less (a 28% reduction). Similar results were obtained for all sites for both TLR and AVD parameterizations.

| Future impacts of drought and O 3 stress
To assess how closely FORCAsT was able to reproduce observed GPP and LE driven by meteorological and O 3 data from each GCM, a test simulation was conducted for each site using bias-corrected 'historical' data for the period 1996-2005 ( Figure S2). Figure 4 shows that although there were differences in the GPP and LE estimated from each individual GCM, the ensemble means closely matched estimates made using observed meteorology. The good performance of the historical GCM driving data relative to the observed driving data is further confirmed by low RMSEs, high correlation coefficients and low SDs (see Taylor diagrams in Figure S6), lending confidence in our use of ensemble mean driving data for future simulations.   Figures S9 and S10, there is uncertainty about the projected GPP and LE fluxes in future as individual GCM ensemble members provide diverse estimates.

| Changes in GPP and LE in future
The uncertainty is higher between 2091 and 2100 ( Figure S10) than 2041-2050 ( Figure S9). The projected decrease in ensemble mean LE at BLO is due to lower LE estimated by several individual GCM ensemble members as shown on Figure S9. HYY and CPZ are expected to experience higher percentage increases in productivity between the middle and end of the century than BLO although the overall productivity level at HYY will remain lower than those at CPZ and BLO. The higher productivity projected for CPZ and HYY could be due to bigger increase in projected winter and spring temperatures at the two sites ( Figures S3 and S4), which is likely to extend the length of the growing season at these sites.  There is negligible difference between the impacts of drought stress on either GPP or LE at the end and middle of the century at HYY.

| Impacts of drought and O 3 through the 21st century
The addition of O 3 stress based on the tolerance parameterization (FUT + O 3 ) reduced estimated GPP and LE at all three sites compared to FUT, although the reduction was more pronounced at CPZ and BLO than at HYY and for 2041-2050 than 2091-2100. GPP could be reduced by 3%-4% due to O 3 damage by mid-century but only 2%-3% (1% less) by the end of the century, with similar impacts seen on LE across all sites. Figure 6 shows that the combined effect of drought and O 3 stress leads to bigger decreases, but there are differences in the F I G U R E 4 Estimates of gross primary productivity (GPP; a-f) and latent heat flux (LE; d-e) at CPZ, BLO and HYY respectively using bias-corrected historical (1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005) Figures S3 and S4). By mid-century, drought and O 3 stress could lead to reductions in GPP and LE at CPZ from 16 to 13 g C m −2 month −1 and 38-30 W m −2 month −1 , decreases of 18% and 20% respectively. The combined impact of drought and O 3 stress on GPP increases to a reduction of ~22% by the end of the century, although their impact on LE remains unchanged. Reductions in GPP and LE are also projected for BLO and HYY as shown on Figure 6.
For the Mediterranean sites, there is a marginal difference (~1% lower) between the sum of drought and O 3 impacts on GPP and LE when applied separately than when the two stressors are applied together while no difference is observed at HYY. This is smaller than the 5% difference seen under present-day conditions, which suggested that the two stresses interact and could compensate for each other.

| D ISCUSS I ON
We investigated the current and future impacts of drought and O 3 stress on gas exchange and forest productivity in three NH forests: Mediterranean forests at BLO and CPZ and a Boreal forest at HYY. We found that all three become more productive over time with GPP projected to increase by 7%, 5% and 8% at CPZ, BLO and Under RCP8.5, CO 2 concentrations are projected to increase rapidly from current values of ~380 to 936 ppm by 2100, and average global temperature by 4.5°C with some areas experiencing even higher temperature increases as shown by Figures S3 and S4.
Warmer temperatures could lead to an earlier onset of the growing season (Menzel et al., 2006) leading to increased plant productivity early in the season (Keenan, Chin, & Whorf, 2014). Increased atmospheric CO 2 is also expected to provide additional atmospheric CO 2 for photosynthesis, and the resultant CO 2 fertilization (a phenom- Nitrogen (Norby et al., 2010) and phosphorus (Cleveland et al., 2013) availability are particularly crucial to terrestrial carbon storage as they regulate plant productivity throughout the terrestrial biosphere (Cleveland et al., 2013). Wieder, Cleveland, Smith, and Todd-Brown (2015) have shown that accounting for nitrogen and nitrogen-phosphorus limitation could lower model-projected primary productivity substantially, highlighting the important role that these two nutrients could play in the ability of plants to sequester CO 2 in future. However, it is not currently understood how soil nutrient availability will change in future and we have not explicitly considered that here.
Plant response to increasing atmospheric CO 2 is also modulated by drought and temperature (Gray et al., 2016;Manderscheid, Erbs, & Weigel, 2014), factors which could become even more relevant in the warmer drier climate projected under RCP8.5. Other factors that could limit the CO 2 fertilization effect in forests include tree species migration (Midgley, Thuiller, & Higgins, 2007;Scheller & Mladenoff, 2005) and forest management practices which could affect the structure, density and tree diversity in these forests, and hence the impacts estimated here. Therefore, our simulations are intended to investigate specific ecosystems (three managed forests) and do not attempt to predict responses for broad PFTs. By using driving data from a range of GCMs, the impact of future changes in drought and temperature and their associated uncertainties have been implicitly accounted for in our estimate of changes in GPP and LE. However, we have not explored the impacts of the availability of soil nutrients or tree age on future GPP and LE. This presents an uncertainty in the projected increases in plant productivity for middle and end of the century and the impact these will have on carbon uptake at the study sites. ing a decoupling effect between the two processes as described by Lombardozzi, Levis, Bonan, Hess, and Sparks (2015). Therefore, the application of a correction factor derived from the response of A to O 3 uptake (as in Equation 10) led to an underestimation of the impact of O 3 on gs and consequently LE. The avoidance method (Hoshika, Watanabe, et al., 2013) assumes that only g s is directly affected as stomatal O 3 flux increases, with only an indirect impact on photosynthetic rate. Plant transpiration rates, and hence LE, however, are controlled only by gs, resulting in a greater impact on LE.
When comparing the O 3 stress strategies alone (when drought stress function β was set to 1), we observed that the best performances were provided by applying the tolerance strategy in the study sites characterized by a Mediterranean climate. We hypothesize that in these sites, drought-induced stomatal control dominates over the O 3 -induced stomatal control protecting plants from both the stressors (Löw et al., 2006) and that their characteristic O 3 -induced antioxidants production (Nali et al., 2004;Paoletti, 2006) Mills, Hayes, et al., 2011;Mills, Pleijel, et al., 2011), the key determinant of stomatal O 3 flux (Emberson et al., 2018). Elevated CO 2 has been observed to significantly decrease O 3 damage in several plant species (Fiscus, Reid, Miller, & Heagle, 1997;Harmens, Mills, Emberson, & Ashmore, 2007;Mills, Hayes, et al., 2011;Mills, Pleijel, et al., 2011). Our results show a decrease in stomatal conductance in future relative to the present day which is likely to reduce stomatal O 3 flux and hence its impact. Second, the decreasing impact of O 3 on plants could also be due to the interactive effects of drought and O 3 stress on plants as drought stress reduces stomatal conductance (e.g. Basu et al., 2016;Farooq et al., 2009). In FORCAsT, as most coupled stomatal conductance-photosynthesis models, drought stress directly downregulates both stomatal conductance and photosynthesis rates (e.g. Clark et al., 2011;De Kauwe et al., 2015;Egea et al., 2011;Keenan et al., 2010). In present-day simulations, we found that the combined effects of the two stresses were up to 5% lower than the sum of the impacts of the two stresses acting individually. A similar but less pronounced interaction between the two stresses is also seen in future simulations ( Figure 6). We therefore conclude that the decreasing impacts of O 3 stress in future climates are partly due to the decrease in stomatal conductance as a result of increasing frequency and severity in drought stress projected for future climates (Dai, 2011;IPCC, 2014). This conclusion is supported by recent findings that future stomatal O 3 uptake in plants will decrease under drought stress (e.g. see Fuhrer, 2009;Lin et al., 2020).
We found that drought stress had a greater effect on estimated GPP and LE than O 3 stress at all sites across all time periods and was more pronounced at the Mediterranean sites (CPZ and BLO). We hypothesized that when water availability is limited, Mediterranean vegetation is more responsive to drought stress than to O 3 exposure (Löw et al., 2006), so stomatal regulation induced by drought stress indirectly acts as O 3 response, by reducing the O 3 stomatal flux together with the water loss, explaining also the reduced predictive ability of the model when both stressors are combined (Figure 6).
Although a general rapid reduction of stomatal aperture in response to short-term exposure to O 3 was observed (Wittig et al., 2007), the chronic exposure to O 3 may induce a phenomenon known as 'stomatal sluggishness', that is, a reduction of plant's ability to regulate stomata (Carriero et al., 2015;Emberson et al., 2009;Hoshika, De Marco, Materassi, & Paoletti, 2015;Hoshika et al., 2016;Hoshika, Watanabe, Carrari, Paoletti, & Koike., 2018). This is a serious prob- In present-day simulations, the inclusion of drought stress alone led to ~20% decrease in estimated GPP and LE at CPZ and BLO, but at HYY, the reduction was only 13% for GPP and 10% for LE. This is a surprising result considering that drought is an annual occurrence at CPZ and BLO, and accounting for drought stress has been shown to improve model fit to observations of photosynthesis in Mediterranean ecosystems (Fares et al., 2019;Keenan et al., 2010). This indicates that although plants in Mediterranean ecosystems have adapted to drought stress (Calfapietra et al., 2009;Paoletti, 2006), their growth and productivity is still likely to be negatively impacted by any further decrease in SWC. The results for HYY over the 1997-2014 period ( Figure S5) and 2005-

(Figures 2 and 3), and observed effects in Boreal forests in
Canada (Kljun, Sabate, & Gracia, 2007;Krishnan, 2006), Finland (Gao et al., 2016) and across Europe (Ciais et al., 2005) show that even for a well-watered forest, anomalous drought events can have a big impact on plant productivity. The Boreal region, extending across North America, Europe and Asia, constitutes the second largest forested biome after tropical forests (Landsberg & Gower, 1997) and therefore plays an important role in the global carbon cycle (Keeling, Chin, & Whorf, 1996). As global climate changes, productivity in Boreal ecosystems will be at a risk from drought stress although this effect could be mitigated by longer growing seasons which would potentially increase productivity as has been seen in other regions (e.g. Dragoni et al., 2011;White, Running, & Thornton, 1999).
One of the main challenges hindering accurate quantification of drought and O 3 stress impacts is the lack of long-term measurements at an appropriate spatial and temporal resolution for model parameterization, calibration and evaluation (see review by Emberson et al., 2018). In this study, we use half-hourly measurements of SWC and O 3 and empirical equations that relate these stresses to plant productivity and gas exchange. Presentday simulations show that incorporating both drought and O 3 stress gives the best model fit to observed GPP and LE, and that these two stresses counteract each other. Productivity increases in our Mediterranean and Boreal forest sites, with GPP (and potentially carbon sequestration) increasing by between 11% and 14%. Although we have not investigated future changes for other ecosystems, if our results were scaled to the regional level, the projected increase in GPP could be significant for the global carbon budget. Finally, we would like to thank Paul Young of Lancaster Environment

ACK N OWLED G EM ENTS
Centre for providing quality controlled CMIP5 O 3 data used in this study.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

AUTH O R CO NTR I B UTI O N
All authors designed the experiment and contributed to writing the manuscript. F.O.-L. and K.A. carried out the modelling work and analysed the output data.