High spatial resolution ozone risk-assessment for Asian forests

Background tropospheric ozone (O3) is increasing particularly over China and India, and becomes a major threat to Asian forests. By using the coupled WRF-Chem model at high spatial resolution (8 km) over Asia in 2015, we showed that both standards AOT40 (European) and W126 (United States) underestimated the O3 risk to deciduous forests and overestimated it to evergreen forests compared to the biologically based metric POD1. Both metrics AOT40 and W126 showed different spatial distribution and exceedance extent with respect to POD1. We found very high potential of O3 impacts on deciduous forest growth in Asia, while potential O3 impacts on evergreen forest types were lower. The most limiting factors were light availability, soil water content and air temperature (65%, 29% and 6%, respectively), making this region of the globe at high O3 risk for deciduous species and at medium O3 risk for evergreen species. For the first time, the O3 risk to Asian forests was quantified at high spatial resolution; and our results suggested: (i) a relevant overestimation of O3 risk to evergreen forests when using AOT40 and W126 relative to the more biologically based POD1 metric; and (ii) a significant underestimation of O3 risk to the boreal deciduous forests when using AOT40 and W126 relative to POD1 because of stomatal aperture permissive condition.


Introduction
Tropospheric ozone (O 3 ) is a seriously damaging air pollutant affecting human health (World Health Organisation 2013, Fu and Tai 2015, Cohen et al 2017, materials (Screpanti and De Marco 2009) and vegetation (Lu et al 2018, Mills et al 2018. Despite effective control efforts and legislation to reduce O 3 precursors emissions, such as nitrogen oxides (NO x ) and non-methanic volatile organic compounds (NMVOCs), surface O 3 pollution is still a major air quality issue over large regions of the globe , Gaudel et al 2018, and it is expected to increase in the future because of increasing methane emissions  and climate change (Anav et al, 2019). The high O 3 concentrations found over China and India megacities pose a major threat to food production and other ecosystem services in Asia (Tai et  Several studies reported the decrease of surface O 3 mean concentrations in United States (Lefohn et al 2010, Cooper et al 2012, Lin et al 2017 and Western Europe (Sicard et al 2013, Paoletti et al 2014, EEA 2016, Yan et al 2019. In contrast, in the last decades Asia became the world's largest emitter of O 3 precursors (Hoesly et al 2018). In particular, China emits 30% and 19% of the global emissions of NO x and NMVOCs, respectively, followed by India with 13% and 11% (Hoesly et al 2018). Some region of East Asia have experienced decreases of O 3 precursors emission in recent years such as Beijing, the Pearl River Delta, Taiwan and Japan, and additional work is required to understand the response of surface O 3 (Duncan et al 2016, Krotkov et al 2016, Liu andWang 2020). Recent analyses of Chinese O 3 monitoring stations (for the years 2015 and 2016) showed that O 3 levels were well above the threshold set to protect forests (Lu et al 2018, Feng et al 2019. All these considerations suggest that O 3 impacts on vegetation is a relevant issue in Asia and thus warrant more investigations (Oksanen et al 2013, Feng et al 2015.
Due to the lack of observations with sufficient spatial and temporal coverage, especially in South Asia, many studies have used global and regional scale models to supplement the missing information from in situ measurements (Kunhikrishnan et al 2006, Engardt 2008, Sheel et al 2010. In the past decades, the important role of numerical models has been increasingly recognized and numerous air dispersion or air quality models were developed at various scales to assist in understanding, predicting and controlling air pollution (Lamarque et al 2013, Miranda et al 2015. These models were successfully applied to air pollution investigation and management in populated cities and regions worldwide (e.g. Reis et al 2005, Calori et al 2006, Haase et al 2014, Anav et al 2016. The current O 3 pollution levels may lead to adverse effects on forest trees in East and Southeast Asia (Lu et al 2018) where high species richness is present (Kier et al 2009). Although some information on the O 3 effects on plants species in East Asia is available, the pollution situation of most Asian countries is not yet well clarified (Koike et al 2013). Main reason is the need of high-resolution regional model, to provide better estimates of air pollution with a lower bias and the second one is the low availability of tropospheric air quality measurements, needed for model validation. Indeed mostly of the modelling information available for Asia are from global models at horizontal resolution of e.g.~50 km (Engardt 2008). Several air quality or chemical transport models (CTMs) have been developed and ran to represent the complex mechanisms involved in transport, transformation and deposition processes in East Asia (Han 2007). The CTMs were applied to study air pollution for China as a whole (Hu et al 2016, Li et al 2016 and for several Chinese regions , Liao et al 2015. Hu et al (2016) applied the Community Multi-scale Air Quality (CMAQ) and Weather Research Forecasting (WRF) modelling system to predict air pollutant concentrations over China. The results showed an overestimation of 1 h and 8 h O 3 averages, probably due to the coarse horizontal resolution (36 km). A modified WRF-CMAQ modelling system was used to simulate O 3 concentrations in winter (December 2014-February 2015 and summer (June-August 2015) for the Sichuan Basin . Most of the basin was found to exceed the World Health Organisation (WHO) guidelines for 8 h O 3 on >70% of winter days and >40% of summer days. The 1 h and 8 h O 3 averages were both greatly over-predicted in winter, but the model performance was acceptable in summer, when the photochemical production of O 3 due to anthropogenic emissions should be strongest in the basin .
The availability of station data used to validation chemistry models was a main issue in this area due to the scarcity of monitored information (Li et al 2007). Recently, few authors have validated CTMs results with in situ and satellite observations (e.g. Kumar et al 2012, Sharma et al 2016, Sicard et al 2020. As an example, the WRF-CMAQ model was used in India with different spatial resolution for emissions and meteorological inputs (e.g. 36 km) to assess source and species sensitivities of ground-level O 3 concentrations (Chatani et al 2014, Sharma et al 2016. Simulations of O 3 and its precursors have been conducted using the updated version-2 (HTAP-v2) emission inventory and the offline global chemistry transport model MOZART-4 (Surendran et al 2015), showing reasonable model performance, but some disagreement in O 3 concentrations and seasonal variation over South Asia were still evident (Surendran et al 2015).
Forests in Asia are important for carbon sequestration (Fang et al 2001, Yu et al 2014 and biodiversity conservation (Myers et al 2000). China has been implementing the most ambitious afforestation programs in the world (Zhang et al 2017). Thus, it is important to estimate O 3 impacts on forest ecosystems in an area characterized by many different climatic conditions and plant species (Xu et al 2010).
Different criteria have been developed to define O 3 risk assessment for forests (Lefohn et al 2018). A concentration-based metric, i.e. AOT40 defined as the sum of hourly O 3 concentration exceeding 40 ppb across daily and seasonal time windows, is currently used in the European legislation for O 3 risk assessment (CLRTAP 2017). The second metric, i.e. PODY, developed more recently in Europe, is based on phytotoxic O 3 dose entering the leaves, depending on the stomatal aperture (Paoletti and Manning 2007), with an hourly threshold Y that is set to 1 nmol m −2 s −1 for forests (CLRTAP 2017). The two metrics showed inconsistent spatial (Anav et al 2016) and temporal patterns (De Marco et al 2015) from local to regional scales over Europe. The current standard recommended in United States for forest protection is W126 (US Federal Register 2015), defined as the sum of the hourly O 3 concentrations during the growing season, and each concentration is weighted by a sigmoidal function to assign greater emphasis to the highest concentrations (Lefohn et al 2018). Ozone critical levels (CLs) were developed for the three metrics, intended as dose (POD1) or concentration below which no effect on forests is expected (Büker et al 2015), that showed a different sensitivity between evergreen (lower) and deciduous species (higher) (Sicard et al 2016). Information on PODY in Asia is still limited to specific countries by modelling approaches (e.g. Japan, Hoshika wheat production loss in China and India using fluxbased methods (POD6) and compared different O 3 dose metrics (AOT40 and POD6) with a resolution of 40 km. At the moment, similar risk assessment is not available for forests in the same region. Therefore, there is an urgent need to perform regional simulations of POD1 to provide high spatial resolution inputs for a more realistic O 3 risk assessment for forests over Asia.
The main aims of the present study were to (i) evaluate magnitude and distribution of O3 risk to Asian forests at high spatial resolution, by comparing concentration-based (AOT40 and W126) and uptake-based (POD1) metrics; (ii) assess the spatial consistency between metrics in order to identify areas where they disagree; (iii) quantify the percentage of the Asian domain exposed to O3 levels exceeding the thresholds of protection for evergreen and deciduous forests in northern, continental and (sub)tropical climates; and (iv) identify the most important climate constraints affecting the stomatal uptake of O3 by Asian forests. We hypothesized that AOT40, W126 and POD1 are uncoupled, and that deciduous forests are at higher risk than evergreen forests, as suggested by a meta-analysis of experimental results in Asia .

WRF-Chem model
We used the Weather Research and Forecasting model (WRF-Chem, v3.9), a coupled climate-chemistry model (Grell et al 2005, Skamarock andKlemp 2008), to reproduce the regional climate and surface O 3 concentrations over South-Eastern Asia. The model domain (figure 1) is projected on a lambert conformal map with a horizontal resolution of 8 km, which allows to simulate atmospheric chemical and physical processes at fine spatial scale. The entire year 2015 was simulated, with a spin up of 1 month (1st-31st December 2014).
The initial and boundary conditions for meteorology, updated every 3 h, were retrieved from the European Centre for Medium-range Weather Forecast ERA5 product (C3S-ERA5 2017), whose outputs are available with a horizontal resolution of~31 km. Similarly, chemical boundary conditions were provided from MOZART-4 (Model for OZone And Related chemical Tracers which is available on a grid of~10 × 10 km for the year 2010. Ozone concentrations obtained by the model were considered as top of the forest canopy, because the first layer of WRF/Chem is around 30 m height. The model validation is not showed in this manuscript, but is fully reported by Sicard et al (2020).

Definition of the forest types
The dominant forest distribution (figure 1) was obtained by merging the USGS landcover distribution and the Koppen climate, following the methodology proposed by Anav et al (2016). This allows using the vegetation definition and the parameterizations described in chapter 3 of the Mapping Manual (CLRTAP 2017) and computing the POD1 and its CL derivation. The six categories of forests identified are boreal deciduous (BD), boreal evergreen (BE), continental deciduous (CD), continental evergreen (CE), (sub)tropical deciduous (TD), and (sub)tropical evergreen (TE) species. For boreal and continental types, we used the parameterization developed by CLRTAP (2017) for beech/birch (deciduous) and Norway spruce (evergreen). Due to the lack of specific tropical and sub-tropical forest parameterization in CLRTAP (2017) and in the scientific literature, we decided to join tropical and subtropical forest species in a single category and approximate the (sub)tropical species to Mediterranean species. Indeed, the Koppen classification includes both sub-tropical and Mediterranean climates into the same category C of temperate climates (Kottek et al 2006). For the (sub)tropical type, we thus selected the parameterization suggested for Mediterranean conditions (CLRTAP 2017) for deciduous oaks (TD) and evergreen oak (TE).

Estimation of AOT40, W126 and POD1 metrics
AOT40 (expressed in ppm h) was computed according to the following formulation (CLRTAP 2017), i.e. as the sum of the hourly exceedances above 40 ppb over the time window between start date of the growing season (SGS) and end date of the growing season (EGS) according to figure 2, during daylight hours: where [O 3 ] is the hourly O 3 concentration (ppb) and dt is the time step (1 h). The function 'maximum' ensures that only values exceeding 40 ppb are included. The W126 exposure index (expressed as ppm h) is a non-threshold index that is described as the sigmoidal weighting sum of hourly O 3 concentrations recorded during specified daily and seasonal time windows, where each hourly O 3 value is  given a weight that increases from zero to one with increasing value (Lefohn et al 2018) and is defined as follows: where M = 4403, A = 126, and C i is the hourly average O 3 mixing ratio in ppb. Further details about the index are available in Lefohn et al (2018). For consistency W126 and ATO40 were cumulated during the same daylight hours and growing season.
For the PODY, we applied a threshold Y of 1 nmol m −2 s −1 for consistency with the approach recommended by CLRTAP (2017) for forest protection; the POD1 was computed as follows , Büker et al 2015, CLRTAP 2017: where dt is 1 h, [O 3 ] is the hourly O 3 concentrations (ppb), R b is the quasi-laminar resistance (s m −1 ), R c is the leaf surface resistance (s m −1 ), and g sto is the hourly value of stomatal conductance to O 3 (mmol O 3 m −2 PLA s −1 , where PLA is the Projected Leaf Area) computed as following.
where g sto is the actual stomatal conductance and g max is the maximum stomatal conductance of a plant species (mmol O 3 m −2 PLA s −1 ) expressed on a projected total leaf surface area. The functions f light , f temp , f VPD , and are the variation in the maximum speciesspecific stomatal conductance g max with photosynthetically flux density at the leaf surface (PPFD µmol photons m −2 s −1 ), surface air temperature (T, • C), vapor pressure deficit (VPD, kPa), and volumetric soil water content (SWC, m 3 m −3 ), respectively. The function f min is the minimum stomatal conductance. These species-specific functions vary between 0 and 1, with 1 meaning no limitation to g sto , and are expressed as CLRTAP (2017): where light a is a dimensionless constant, PPFD is hourly photosynthetic photon flux density, T opt , T min , and T max represent the optimum, minimum, and maximum temperature for g sto , VPD min and VPD max are minimum and maximum VPD for g sto , and WP and FC are the soil water content at wilting point and field capacity, respectively (Anav et al 2016, CLRTAP 2017. In addition, f min is the species-specific fraction of g max and f phen , i.e. the phenology of vegetation, is used to compute the duration of the growing season during which plants can uptake O 3 . In detail, we used the seasonal variation of third generation Leaf Area Index, i.e. bi-weekly LAI3g data with~8 km of spatial resolution  to define the start of the growing season (SGS) and end of the growing season (EGS) (figure 2), as described by Anav et al (2018).

Calculation of CL exceedances and biomass losses
CLs are defined as 'concentrations, cumulative exposure or cumulative stomatal flux of atmospheric pollutants above which direct adverse effects on sensitive vegetation may occur according to present knowledge' (CLRTAP 2017). The parameter usually evaluated for estimating such adverse effect on forests is a 4% reduction in biomass except for evergreen species where the biomass reduction is set to 2% (CLRTAP 2017). Exceedances of O 3 CL were calculated for the three metrics AOT40, W126 and POD1 as recommended by CLRTAP (2017), i.e. as difference between the estimated value in each grid cell and the CL obtained by literature data. We used different CLs depending on the metric, i.e.: for AOT40 we applied the European CL set to 5 ppm h to protect all forests types (CLRTAP 2015); for W126, we used 7 ppm h or 21 ppm h, as recommended by Environmental Protection Agency (EPA 2007) to protect the most sensitive tree species or any kind of vegetation; for POD1, we used 5.2 mmol m −2 for BD and CD, 9.2 mmol m −2 for BE and CE, 14.0 mmol m −2 for TD and 47.3 mmol m −2 for TE, according to CLRTAP (2017).
After performing a point-wise calculation of the exceedances over the model domain, the nonattainment area of the target value (in %) was calculated relative to the total domain covered by either deciduous or evergreen forests. The biomass losses for each forest type were estimated based on the dose-response functions derived by CLRTAP (2017), as indicated in table S1 (available online at https://stacks.iop.org/ERL/15/104095/mmedia).

Results
The lowest AOT40 levels were found in South Asia (i.e. Vietnam, Laos and Thailand) dominated by moist and dry broadleaf forest types (figure 1) and with lower O 3 concentrations (figure 3). The area with highest AOT40 values was observed in central China, characterized by high O 3 concentrations, especially during growing season (Sicard et al 2020) and in the Indo-Gangetic Plain region, characterized mainly by high-elevation forests e.g. Himalaya (figure 1). The spatial distribution of W126 was similar to AOT40, but the highest and lowest values were amplified due to the nature of the metric. Peaks of W126 were located mainly in Northern and Northeastern India (figure 3).
In contrast, the spatial pattern of POD1 was different from that of AOT40 and W126 (figure 3), with the highest absorbed dose for TD and TE forests and the lowest for BE forests (figure 4). A different spatial distribution of hot-spots was observed in the southern region, with the highest POD1 values in southern China rather than in India (figure 3). The average spatial correlation coefficient was 0.96 between AOT40 and W126 and 0.20 between AOT40 and POD1 (data not shown).
The selected metrics split in terms of dominant vegetation are displayed in figure 4; our results indicate that, in 2015 the lowest AOT40 and W126 values were found in the northern region (figure 3) dominated by boreal or continental forest species (figure 1), while the highest AOT40 and W126 values were observed in the areas where tropical and subtropical forests grow ( figure 4).
Considering the CLs, our results indicate that the AOT40 CL was exceeded over 53%, 93%, 74%, 86%, 98% and 97% of the areas covered by BD, BE, CD, CE, TD and TE forest types, respectively (figure 5). The W126 CL recommended for protecting sensitive plant species was exceeded over 31%, 87%, 57%, 73%, 98% and 98% of the areas covered by BD, BE, CD, CE, TD and TE forest types, respectively, while the exceedances of the W126 CL for the protection of all species covered 5%, 59%, 28%, 43%, 93% and 94% of the areas covered by BD, BE, CD, CE, TD and TE forest types, respectively. For both AOT40 and W126, the main attainment areas (i.e. achieving the air-quality standard) were in boreal and continental climates, with deciduous species showing lower risk than evergreen species. Regarding the POD1, 99%, 12%, 65%, 18%, 93% and 46% of the areas covered by BD, BE, CD, CE, TD and TE forest types exceeded the respective CLs (figure 5). The POD1-based O 3 risk was higher for deciduous forests than evergreen forests, despite the shorter duration of the growing season.
To link the POD1 to the forests biomass loss, we applied the dose-response function to the different forests type. The average POD1-driven biomass loss for the six forest types is shown in figure 6. The  most POD1-affected forest types were the deciduous species, in particular BD and TD, with a respective biomass loss of 16% and 17%. The CD showed a biomass reduction of 7%, while evergreen species showed lower biomass reductions, even if the accumulation period was longer. Indeed, a biomass loss of 1.5%, 1.6% and 4.4% was estimated for BE, CE and TE species, respectively. When all forests were averaged, the POD1-estimated biomass loss was 7%.
The spatial distribution of the limiting functions f, i.e. the functions regulating the stomatal opening and consequently the O3 uptake by leaves, is shown in figure 7.
Considering the temperature, we found a relevant limitation to stomatal opening only in the mountainous region around the Tibetan plateau, while in the remaining area of the domain the air temperature is not significantly limiting O3 uptake. In contrast, the maximum f light limitation was observed in the south-eastern area of the domain, and this function was relatively low (i.e. high limitation) over the whole domain. f VPD was generally not limiting for stomatal conductance in Asia, except in the central-western part of the domain, while f SWC was strongly limiting for stomatal conductance over almost all India and the central part of the domain. The average values of the 4 limitation functions in the six forest types is shown in figure 8. As expected f temp is mostly limiting for stomatal conductance in boreal forests (0.62 and 0.44 for BD and BE, respectively), and f light is mostly limiting stomatal conductance in tropical forests (0.46 and 0.48 for TD and TE, respectively), while intermediate values were found in the continental forests for both CD and CE, where f SWC limitation seems to be more relevant (0.56 and 0.51 for CD and CE, respectively) By selecting the most limiting function in each cell grid, we observed that the most distributed limiting function over Asia was f light (over 65% of the domain), followed by f SWC (29%) and f temp (6%) (figure 9). Just in few grid points (<1%) the most limiting function was f VPD .

Discussion
East and South Asia has recently experienced rapid economic growth, during which anthropogenic emissions have increased and deteriorated air quality (Kurokawa and Ohara 2019). Air pollution, especially surface O 3 in East and Southeast Asia, is more serious than in Europe and North America (Koike et al 2013, Mills et al 2018 and is still expected to increase by 2100 (Sicard et al 2017). Thus, the use of air quality models has also increased in this region to better understand the spatial and temporal distributions of air pollutants and to examine the impact of the increased anthropogenic emissions on air quality degradation for Asian countries  and consequently the impacts on forests (Feng et al  2019).
In this study, we used WRF-Chem with a high horizontal resolution (8 km) over a domain covering all India, China, part of southern Asia, and 8  reaching southern Siberia. The same model was previously used to simulate ground-level O 3 over a smaller domain in East Asia , and emphasized the importance of the resolution in the performance of the model. For this reason, we selected a fine spatial resolution to have more realistic results. Our simulation was validated against insitu measurements from monitoring stations across China and satellite data (Sicard et al 2020); the comparison with measurements suggests that the model well reproduces the spatial pattern of meteorological variables and surface O 3 concentration. Indeed, the WRF-Chem model simulated well the spatial distribution and seasonal variation of O 3 . Compared to IASI-GOME2 satellite retrievals, a good spatial agreement is noticed in summer, with a spatial correlation of 0.99, and a lower correlation is observed during spring and winter (0.61 and 0.71, respectively). Compared to ground observations data (from 1500 air quality monitoring network across China), a mean annual bias of 5 ppb is observed in 2015. This bias is in line with the ones showed by the European regional models described in Colette et al (2011). Looking at the surface air temperature, one of the most important parameter affecting POD1 entity, WRF captures well the spatial pattern with a decreasing south-north gradient and a cold area over the Tibetan plateau (Sicard et al 2020).
Interestingly, the POD1 values were consistent with the results shown for Europe (Anav et al 2016), while AOT40 values were around 100%-200% higher in Asia than in Europe. The latter result is consistent with Lu et al (2018), that found significantly higher AOT40 and W126 levels in China, i.e. 35%-100% and 50%-170% higher than in Europe and United States, respectively.
Asian studies implementing PODY are fewer for trees than for crop species (Agathokleous et al 2018), and have normally provided experimental observations in very local plots and for a few species. POD1 is the index recommended by CLRTAP (2017) for protecting trees against O 3 -induced biomass loss. In an experiment with poplars in China, AOT40 and POD1 dose-response relationships indicated a 5% biomass loss at 12.0 ppm h and 6.1 mmol m −2 , respectively (Hu et al 2015). To calculate the exceedance, we used the thresholds, dose-responses and CLs set in Europe (CLRTAP 2017) because the only PODY-based doseresponse relationships for Asia are for hybrid poplar and thus insufficient to represent the variety of forest types in Asia. Our results showed different sensitivity of forest types to O 3 concentrations potential exposure (AOT40 and W126) or O 3 fluxes (POD1), in particular with respect to the non-attainment area for forest protection. Most of evergreen forests (86%-97%) were potentially exposed to O 3 concentrations exceeding the limits for forest protection while the percentage of evergreen forests exposed to POD1 above the CLs was lower (12%-46%). Asian studies should focus on flux-based O 3 metrics to provide relevant bases for developing proper standards. However, given the technical requirements in calculating flux-based O 3 metrics, which can be an important limitation in developing countries, cumulative exposure indices like AOT40 should always accompany flux-based indices (Agathokleous et al 2018).
A different spatial distribution of AOT40 and POD1 in East Asia was previously described by Hoshika et al (2011) andby De Marco et al (2015) and Anav et al (2016) in Europe. In particular in Asia POD0 values ranging from 10 to 48 mmol m −2 were reported (Hoshika et al 2011), while in our results we obtained values of similar magnitude, but with an hourly threshold Y of 1 nmol m −2 s −1 . Our spatial variability was in line with the spatial distribution obtained by Hoshika et al (2011). We obtained similar results with no limitation due to VPD on the major part of the domain, excluding the central north-western part of India. One possible explanation is that the Tibetan Plateau (mean elevation over 4000 m a.s.l.) acting as a strong heat source in summer, generates upward airflow motions over its eastern flank that, combined with large amounts of moisture from the tropics, result in strong monsoons and wet climate in East Asia (Ding and Chan 2005).
The limiting functions are important to determine the role of climatic conditions on stomatal conductance . Physiological responses to changes in climate are highly dependent on the limiting factors of a particular site to forest growth. The eastern part of the domain did not show SWC as most limiting factor, while the western part of the domain, including India, with either limited rainfall or high temperature showed some areas characterized by limiting soil moisture, a very important function in many dry area of the globe (De Marco et al 2016). We showed that the most limiting function to O 3 uptake in eastern Asia was f light , in agreement with Nemani et al (2003) that investigated geographic distribution of potential climatic constraints to plant growth derived from long-term climate statistics. As the exchange of gases between atmosphere and terrestrial vegetation is regulated by stomata opening, air pollutants may take advantage of the stomatal aperture to enter leaves, suggesting that the temporal evolution of O 3 and CO 2 uptake are consistent. Indeed Melillo et al (1993) found that the predicted NPP decreases for tropical evergreen forest, may be related to increased temperature and cloudiness. It is important to note that f light estimation included nighttime hours, and thus is affected by the duration of daylight hours.

Conclusions
The lack of information still presents in Asia in terms of stomatal O 3 uptake by the forests, both for modelling and in-situ measurements, warrant more intense studies in this region of the globe. To bridge this gap of knowledge we performed, for the first time, a risk assessment on Asian forests using a high spatial resolution model in order to estimate the phytotoxic O 3 uptake (POD1) into the tree leaves for six forest types, highlighting its spatial distribution compare to concentration-based metrics. We found very high potential of O 3 impacts on deciduous forest growth in Asia, while potential O 3 impacts on evergreen forest types were lower. In particular, the limiting conditions of light, soil water content and temperature in a context of climate change, make this region of the globe at high O 3 risk for deciduous species and medium O 3 risk for evergreen species.

Acknowledgments
This work was carried out with the contribution of:

Data availability statement
The data that support the findings of this study are available upon reasonable request from the authors.