Bridging experimental and monitoring research for visible foliar injury as bio-indicator of ozone impacts on forests

ABSTRACT Tropospheric ozone (O3) is a phytotoxic air pollutant and the O3-induced visible foliar injury (O3 VFI) is a biomarker. A recently developed Free-air O3 eXposure (FO3X) is a promising facility to verify field-observed “O3-like” VFIs and to establish a flux-based threshold for the O3 VFI onset. The present study compared O3-like VFI registered in the southern European forest sites with actual O3 VFI observed in a FO3X experiment. The O3-like VFIs were evaluated by eye in forests and thus it was subjective. According to the imaging analysis, we firstly demonstrated that major parts of the colors were similar in the field and the FO3X. The color pallets for O3 VFI was species-specific and considered a advanced tool for the O3 VFI diagnosis. In addition, we calculated a flux-based threshold for the O3 VFI onset at the FO3X based on a Phytotoxic Ozone Dose (POD1), which ranged from 4.9 to 18.1 mmol m−2 POD1. This FO3X-derived threshold partly explained but did not necessarily match with the observation for several tree species in actual forests. The multivariate analysis showed that O3 VFI was decreased by the presence of various species and suggested the importance of continuous monitoring activities in the field for the further analysis.


Introduction
Tropospheric ozone (O 3 ) is a significant greenhouse gas and phytotoxic air pollutant, which has a deleterious effect on forest health (Watanabe et al. 2021). Therefore, monitoring forest health and functionality under climate change and air pollution is crucial for estimating forest vulnerability to the ongoing changes and for establishing mitigation measures (Paoletti et al. 2022). Furthermore, the recently revised National Emission Ceilings Directive of the European Union (2016/2284/EU, hereafter NEC Directive 2016 targets a 30% reduction in air pollutant emissions by 2030, compared with 2005 levels, and it is necessary to ensure the monitoring of air pollution impacts on ecosystems in EU member states as specified in Art. 9 of NEC Directive. Various O 3 metrics such as AOT40 (Accumulated exposure Over a Threshold of 40 ppb; abbreviations are listed in Table S1) have been proposed to examine forest responses to O 3 (Lefohn et al. 2018). However, scientific evidence has pointed out that O 3 damage is closely related to O 3 uptake through stomata (CLRTAP 2017). Therefore, Phytotoxic Ozone Dose above a stomatal-flux threshold Y (POD Y ), which is derived from a cumulative O 3 uptake over the growing season, is now included in the NEC Directive for the assessment of quantitative O 3 impacts on forest trees (de Marco et al. 2019).
When plants are exposed to O 3 , they often show specific O 3 -induced visible foliar injury (O 3 VFI), which has been used as a bio-marker to assess potential negative impacts of O 3 (Feng et al. 2014;Hoshika et al. 2018). Ozone-like VFIs have been found in various tree species across forest sites in Europe, indicating a possible phytotoxicity of the ambient level of O 3 (annual O 3 mean concentrations: 35 to 50 ppb) in forests (Paoletti et al. 2019;Vollenweider et al. 2019;Sicard et al. 2020).
Traditionally, the O 3 -like VFIs have been evaluated by eye and, thus, the approach was rather subjective, although much efforts have been made to harmonize the VFI monitoring through the inter-calibration courses by international organizations such as ICP-Forests (Schaub et al. 2016). A manipulative O 3 exposure experiment is therefore considered as one of the key approaches for the validation of O 3 -like VFI to diagnose and identify foliar symptoms caused by O 3 (Moura et al. 2018;Vollenweider et al. 2019). Previously, O 3 -like VFIs have been verified in a wide variety of settings, including greenhouses and chambers of various sizes and designs (e.g., Bussotti et al. 2003). Although those experiments provided an insight into potential mechanisms, a difference in meteorological factors such as light, air temperature, and wind speed between chambers and natural conditions may also affect a quantitative analysis of O 3 VFI relative to actual field conditions (Nussbaum and Fuhrer 2000). Free-air O 3 eXposure (FO 3 X) has been established in Italy since 2015, which is a unique facility in Mediterranean Europe to assess the effects of O 3 on vegetation within the framework of the European AnaEE (Analysis and Experimentation on Ecosystems) Research Infrastructure. As is the case with free-air CO 2 enrichment, experiments by free-air O 3 enrichment are considered as the best approach to provide realistic estimates of tree responses under real-world conditions (Paoletti et al. 2017). Symptoms of O 3 VFIs were reported in several tree species at the FO 3 X (Hoshika et al. 2018. However, there is a need for validation studies to confirm whether the O 3 -like VFIs observed in actual forests are similar enough to real O 3 VFIs produced in the FO 3 X. In addition, little is known whether the FO 3 X-derived thresholds for O 3 VFI can be applied to actual forest conditions, although such manipulative experiments have been used to develop standards for forest protection against O 3 (Büker et al. 2015). A flux-based threshold for the O 3 VFI onset is considered essential for the evaluation of regional O 3 pollution ). According to the forest monitoring data, Sicard et al. (2020) proposed flux-based thresholds for O 3 -like VFI of 5 and 12 mmol m −2 POD 1 in dominant conifers and broadleaved trees, respectively.
The present study compared O 3 VFI for forest tree species at the FO 3 X facility with the symptoms observed at actual forest sites with the main aim of validating the field observations. For this purpose, a dataset obtained at southern European forest sites (France, Italy, and Romania) was utilized thanks to the European project MOTTLES (MOnitoring ozone injury for seTTing new critical LevelS) (Paoletti et al. 2019). A novel color pallet approach was developed to analyze the color composition of symptoms for categorizing and standardizing the O 3 VFI. The onset of O 3 VFI was also analyzed on a stomatal flux basis using POD 1 . This study aimed to answer the following specific questions: 1) Are O 3 -like foliar symptoms observed in the field comparable to actual O 3 VFIs produced at the FO 3 X? and 2) Does the FO 3 X-derived flux-based threshold also explain the incidence of O 3 VFI at the forest sites?
The symptoms were carefully checked using a × 10 hand lens for the closer examination of injuries, and all leaves were targeted for each plant. Two-well experienced observers periodically assessed O 3 VFI to determine the date for first-symptom onset. The observers were involved in validation activities, attended field courses and performed annual inter-comparison exercises, organized by ICP-Forests. Foliar injury was compared with the reference picture atlas provided by the validation center for central Europe (www.wsl.ch) and identified according to the previous O 3 VFI studies in tree species (Calatayud et al. 2010;Hoshika et al. 2013;Schaub and Calatayud 2013).

MOTTLES forest sites and field observations
The study is based on data recorded by the MOTTLES O 3 monitoring network from 2017 to 2021 for Italian sites, and 2017-2019 for other countries. The network includes various biogeographical regions in southern European forest sites in three countries (France, Italy, and Romania). The target forest sites cover representative species in these regions, such as Fagus sylvatica in mountainous Alpine areas and Quercus ilex in Mediterranean forest areas. As a result, the MOTTLES network considered 7 broadleaved and 4 coniferous dominant tree species (Table 2).
Each site consisted of i) an open area (open field -OFD) to record remotely and continuously meteorological and O 3 values with active sensors, ii) a nearby Light Exposed Sampling Site -LESS (Schaub et al. 2016), located in the light-exposed forest edge closest to the OFD station (maximum radius of 500 m) LESS area, and iii) an "in the plot" (ITP) area, where soil moisture and forest-health indicators are recorded into the forest. Specific details on MOTTLES active O 3 monitoring sites are available in Paoletti et al. (2019). The field protocols for assessing O 3 VFI by MOTTLES surveyors followed the ICP-Forests manual (Schaub et al. 2016). Surveys were carried out by the same two people in each country from August to early-mid September, i.e., when O 3 VFI are more likely to be observed in the MOTTLES sites (Dalstein et al. 2005;Paoletti et al. 2010). Two campaigns of trans-country inter-calibration were carried out within MOTTLES, and the surveyors had taken part in the international crosscalibration courses organized by ICP Forests.
At each site, the assessment of O 3 VFI was conducted every year within the ITP and the LESS. In ITP, the evaluation of O 3 VFI were performed at each plot on five trees randomly selected. For each tree, five light-exposed branches with ≥30 needles/leaves per branch or needle age class were removed from the upper crown. For deciduous species, current year (C) leaves/needles were assessed. For evergreen species, C, one -year-old (C + 1) and two -year-old (C + 2) leaves/needles were assessed. On each leaf/needle, the extent of O 3 VFI (as a percentage of the total leaf area) was visually scored by using the actual percentage, and results were then averaged for the five branches, resulting in one value per tree; finally, the tree values were averaged per plot. In the MOTTLES network, the LESS is 50 m long and divided into 25 × 2 m 2 nonoverlapping quadrats. In 20 of the 25 quadrants, which were randomly chosen, all plant species were listed, and the presence or absence of O 3 VFI was recorded on the same day in which the ITP survey was carried out.
For the species with validated O 3 VFI, the available literature (Innes, Skelly, and Schaub 2001) and atlas provided by the validation center for central Europe (www.wsl.ch) were used to support the observations. Here, O 3 VFI datasets at MOTTLES sites during 2017-2021 were used for the analysis.

Calculation of ozone indices
Both exposure-and flux-based O 3 metrics were already calculated at MOTTLES sites over the study period from measured parameters . In addition, the metrics, i.e., AOT40 and species-specific POD 1 , were calculated by using meteorological and O 3 concentration data measured at the FO 3 X facility. Accumulated exposure over a threshold of 40 ppb (AOT40) was calculated using the hourly O 3 data during daytime (solar radiation >50 W m −2 , CLRTAP, 2017) measured at each site: where [O 3 ] is the hourly concentration of O 3 (ppb), and dt is the time step (1 h). Phytotoxic Ozone Dose above a detoxification threshold of 1 nmol m −2 s −1 (POD 1 ) was calculated according to the standard methodology suggested by CLRTAP (2017).
where F st is the hourly stomatal uptake of O 3 (nmol m −2 s −1 ), which is derived from leaf surface resistance (r c ) and boundary layer resistance (r b ) (CLRTAP, 2017). F st is thus given by: where g s is the stomatal conductance (mmol O 3 m −2 projected leaf area [PLA] s −1 ). r b is calculated by wind speed and leaf dimension, and r c is defined as 1/(g s +g ext ) (s m −1 ). g ext is the cuticular conductance (0.0004 s m −1 ). For the details of the calculation of r b and r c , see CLRTAP (2017). A simplified formula of the multiplicative stomatal conductance model was applied because all plants were grown under well-watered conditions at the FO 3 X . It is given by: where g max is the maximum stomatal conductance to O 3 expressed on a total leaf surface area (mmol O 3 m −2 PLA s −1 ), f min is the minimum conductance, f light , f temp, and f VPD indicate the stomatal response functions to photosynthetic photon flux density (PPFD), air temperature (T) and vapor pressure deficit (VPD), respectively. For several species (A. glutinosa, Q. ilex, Q. pubescens, Q. robus, S. aucuparia, V. myrtillus), species-specific stomatal conductance parameters were already reported in our previous studies (Hoshika et al. 2018. Stomatal conductance for the other species was parameterized according to the measurements with various meteorological conditions using a portable leaf gas exchange measurement system (CIRAS-2 PP Systems, Herts, UK). Measurements were carried out at least once a month (7 to 19 days in total for each species) throughout the experimental period (Table 1). Pooled data (A. unedo: 128 data, P. angustifolia: 374 data, P. halepensis: 174 data, P. pinaster: 303 data, P. pinea: 164 data, R. ulmifolius: 249 data) were used for the species-specific parameterization according to the boundary line technique (Elvira et al. 2004;Büker et al. 2015;Hoshika, Paoletti, and Omasa 2012). To draw the boundary line (i.e., upper limits of point g s data in a scatter diagram) for each environmental variable, the data were divided into the following stepwise classes: PPFD: 200 μmol m −2 s −1 steps (when PPFD <200 μmol m −2 s −1 , 50 μmol m −2 s −1 steps were applied), T: 2°C steps, VPD: 0.2 kPa steps. Each model function was then fitted according to 95 th percentile values per each stepwise class of environmental factors. g max and f min were estimated as the 95 th and 5 th percentile, respectively (Bičárová et al. 2019;. For the detailed formula of f light , f temp, and f VPD functions, see CLRTAP (2017).

Data analysis
For three species, Rubus ulmifolius, Vaccinium myrtillus, and Sorbus aucuparia, an O 3 VFI image analysis was developed to compare the symptoms observed in the field with those developed at the FO 3 X. Three pictures with minimum 180 dpi (RGB -red, green, blue color mode) of leaves presenting O 3 VFI of each species were selected from the field and FO 3 X conditions. First, the mid-part of each leaf was cut off from the original picture with the same size for each species. Next, each cut of the sample picture was transformed in Indexed Color mode, and a Local (Perceptual) and an 8-color pallet was generated. Next, the color pallets were merged, forming a general 8-color pallet for the O 3 VFI pictures obtained from MOTTLES or FO 3 X conditions. The final 8-color pallets, which characterize the color composition of O 3 VFIs, were then compared between field and FO 3 X conditions, considering the percentage of the same colors appearing in both conditions. We surveyed the significant O 3 VFI colors of the final color pallet (50% of the total colors range) as the ones better describing the O 3 VFI (O 3 VFI/Color), while the other colors were related to leaf typical chlorophyll pigments (LTCP). All analyses were conducted using Adobe Photoshop CC 2017, and the color names were verified in the Hex dictionary (https://www.hexdictionary.com/). All pictures analyzed with its, respectively, 8-color pallet are available as supplementary material (Fig. S1).
In order to detect main predictors for symptoms in the natural environment, we correlated symptom variables with site variables. We applied two linear models by using the linear model (lm) function of R software (Team R Development Core, 2018): in the first model we considered the percentage of symptomatic species within the LESS as response variable; in the second model we considered the O 3 VFI (merging ITP and LESS data) as response variable. The following site variables were included in the model as predictors: year, country, slope aspect, dominant species, and biogeographical region (as categorical), elevation (as continuous), and the total number of species within the LESS (as count variable); all interactions among site variables were also considered in the model. Selection of the optimal model, among those generated by all possible combinations, was based on the Akaike Information Criterion (AIC) values to assess the quality of the model (Barton et al. 2012). For the best model selected, we also calculated R 2 and parameter-specific p-values for each predictor level (e.g., value).

Ozone visible injury at MOTTLES forest sites
The O 3 VFI occurring in the MOTTLES sites were characterized by the following categories: Homogeneously distributed interveinal reddish (Rd), Reddish interveinal stippling (RdSt), Homogeneously distributed interveinal brownish (Brw), Dark brownish interveinal stippling (BrwSt), Bronzing (Br), and chlorosis (Cl) ( Table 3). The 23 symptomatic species presented, in most of the cases, a combination of these symptoms. The most common O 3 VFIs were Cl and BrwSt, occurring in 43% of the species, followed by RdSt, detected in 35% of the species. Rd, Br, and Brw occurred in 26%, 13%, and 9% of the species, respectively. Some pictures were selected to show each of the categorical O 3 VFI occurring in symptomatic species at MOTTLES sites (Figure 1.).
In MOTTLES forest sites, merging ITP and LESS data, O 3 VFI was observed on 23 species, consisting of 14 trees, 8 shrubs and 1 liana (Clematis vitalba) (Table 3). Fagus sylvatica was the species found to be most frequently symptomatic (21 times across all sites and years), followed by the shrubs Rubus ulmifolius and Corylus avellana, recorded 12 and 11 times, respectively. Most of the symptomatic species were deciduous angiosperms, while only two conifers, i.e., Pinus cembra and Picea abies, were found to be symptomatic 2 times and once, respectively. Regarding the forest LESS, linear models showed that the percentage of symptomatic species, despite a significant influence of the country (Table 4; Figure 2.), decreased with the increment in the total number of tree and shrub species within the LESS (Figure 1.). The same trend was observed for O 3 VFI (Figure 3.); however, in this case, species richness in the LESS was the unique driver among the selected site characteristics.

Ozone visible injury at the FO 3 X
Pictures of the O 3 VFI for the symptomatic species at FO 3 X are shown in Figure 1.. Ozone exposure caused O 3 VFI in all target deciduous species, while only two of six evergreen species were symptomatic, i.e., a Mediterranean shrub, A. unedo, and a Mediterranean pine, P. halepensis (Table 5). Among the symptomatic species, A. glutinosa, S. aucuparia, and V. myrtillus showed first symptoms even under real-world O 3 conditions (18 May to 21 June), while O 3 VFI was found only later for A. unedo and Q. pubescens (3 to 25 September).
Following the same classification used to identify O 3 VFI in the MOTTLES sites, a similar combination of symptoms was observed in the eight symptomatic species in the FO 3 X facility ( Table 4). The most common O 3 VFI observed was BrwSt occurring in 50% of the species, followed by BrwSt and RdSt, detected in 38% of the species. Rd occurred in 25%, while Cl was found in 13% of the species. So far, Br has not been observed in the FO 3 X. The Cl mottling appeared on needles of P. halepensis, while the other evergreen species, A. unedo, presented RdSt occurring between the veins on the upper leaf surface. A similar symptom was found in V. myrtillus in which Rd was also observed. The deciduous

Comparing ozone visible injury at the MOTTLES and FO 3 X sites: A color analysis
For R. ulmifolius, 62.5% of colors selected in the final 8-color pallets were common colors presented in both MOTTLES and FO 3 X samples ( Figure 4A-D). Considering only the O 3 VFI/Color, 75% of the colors were common between the MOTTLES and FO 3 X samples (Coral Reef − 512E10, Peat -C7BAA1, and Pesto − 6C6133). One specific O 3 VFI/Color was not identified at FO 3 X samples and was recorded only at the MOTTLES site (Bracken − 512E10). Furthermore, 50% of the LTCP colors were common between the MOTTLES and FO 3 X samples.
For S. aucuparia, 50% of colors selected in the final 8-color pallets were present in both MOTTLES and FO 3 X samples ( Figure 4B-C). Considering only the O 3 VFI/ Color, 25% of the colors were common between the MOTTLES and FO 3 X sample (Yellow Metal − 6C6133). One O 3 VFI/Color was recorded only at the MOTTLES site (Saddle −472D22), whereas two O 3 VFI/Color was recorded at the FO 3 X (Sepia Skin -B5A42, and Tan -D7B690). Furthermore, 75% of the LTCP colors were common between the MOTTLES and FO 3 X samples.
For V. myrtillus, however, not all the considered O 3 VFI in the MOTTLES sites were similar to the FO 3 X O 3 VFI pictures (Fig. S1), 75% of colors selected in the final 8-color pallets were present in both samples ( Figure 4C,D). Considering only the O 3 VFI/Color, 75% of the colors were common between the MOTTLES and FO 3 X samples (Leather − 8B674F, West Coast − 695328 and Mongoose -A3926A). One specific O 3 VFI/Color, was not identified at FO 3 X samples and was recorded only at the MOTTLES site (Beaver -A97E6D). Furthermore, 75% of the LTCP colors were common between the MOTTLES and FO 3 X samples.

Stomatal ozone uptake for visible injury onset at the FO 3 X
New stomatal conductance model parameters for species (A. unedo, P. angustifolia, P. halepensis, P. pinaster, P. pinea, R. ulmifolius) are shown in Table S2    (P. halepensis, and R. ulmifolius). The stomatal light response function (f light ) followed a typical lightsaturation curve with a saturation point above 1000 µmol m −2 s −1 . The g s response to temperature (f temp ) had a bell-shaped curve where g s reached the maximum at the optimal temperature (22-26 °C). The f VPD function indicated that stomatal closure was caused by VPD higher than 1.2 to 1.5 kPa regardless of the species. However, the VPD attaining minimum g s (VPD min ) was relatively low in P. halepensis compared to the other species. Two representative O 3 indices, i.e., POD 1 and AOT40, were calculated to find the values corresponding to the onset of the first O 3 VFI in the 12 species at the FO 3 X experiment (Table 6). The AOT40 threshold for the onset of O 3 VFI was different among the symptomatic species (3.9 to 50.9 ppm h AOT40). The Mediterranean evergreen A. unedo required 50.9 ppm h AOT40 to show the first O 3 VFI, while the first O 3 VFI appeared on both deciduous A. glutinosa and S. aucuparia from low AOT40 values (3.9 to 6.5 ppm h AOT40). On the other hand, the onset of O 3 VFI was observed from 5 mmol m −2 POD 1 for A. glutinosa and V. myrtillus, and from 18.1 mmol m −2 POD 1 for Q. pubescens.
We then applied these FO 3 X-derived thresholds to the MOTTLES conditions. In theory, O 3 VFI should be present when POD 1 exceeds the threshold for the onset of O 3 VFI, whereas no O 3 VFI should be found when POD 1 is lower than this threshold. In fact, in the MOTTLES sites, the results were in line with this theory for A. glutinosa and Q. pubescens ( Figure 5). However, plants were often asymptomatic for Q. robur, S. aucuparia and V. myrtillus even though POD 1 was higher than this threshold. On the contrary, R. ulmifolius was symptomatic even in most cases when POD 1 was lower than the threshold.

Color analysis and comparisons of the foliar symptoms between FO 3 X and actual forests
This first attempt to diagnose O 3 VFI by comparing actual field and manipulative experimental conditions based on the symptom color composition appeared promising, especially for two species where 75% of the O 3 VFI/Color were shared between MOTTLES and FO 3 X samples. A leaf color chart is defined as a series of color swatches used to identify leaf physiological status and characteristics (Takebe and Yoneyama 0000). Color composition in terms of the color chart has been used in several studies, especially to estimate leaf and canopy chlorophyll content and their variation in field conditions (Nguy-Robertson et al. 2015) and for adjusting nitrogen (N) fertilization rate, thus improving N management in field conditions on the basis of leaf color.
Comparing the O 3 VFI between the field and experimental conditions using the color composition was here demonstrated to be a potential tool for the O 3 VFI diagnosis. For the three species compared in the presented study, 50% to 75% of the colors in the final 8-color pallets were commonly present in the field and under manipulative experimental conditions. This  indicates that the O 3 VFI/Color are similar between field and FO 3 X conditions, suggesting that it is possible to make successful validation of field-observed O 3 symptoms by using manipulative fumigation studies. However, it is important to follow standard procedures for the shooting of pictures used in the image analysis, as lighting conditions and picture resolution may affect the colors. The ICP-forest manual (Schaub et al. 2016) listed a series of indications for taking photographs of O 3 VFI, that must be followed for the quality assurance.  The O 3 VFI provides clear indications of O 3 -induced oxidative stress and is reliable when verified by a combination of approaches, once the symptoms expression can be similar even when the stress factors are different (Vollenweider and Gunthardtgoerg 2006;Guerrero et al. 2013;Alves et al. 2016;Moura et al. 2018;Vollenweider et al. 2019). For instance, the adaxial St can appear following several other sources of oxidative stress (Vollenweider and Gunthardtgoerg 2006), and Cl is a common symptom that needs careful validation to be diagnosed as O 3 VFI (Vollenweider et al. 2019). Furthermore, usually observed O 3 VFI in field surveys are aspecific and difficult to interpret (Bussotti et al. 2006) because leaf colors may also be affected by the plant phenological nutritional and physiological status and other abiotic causes (Bussotti and Ferretti 2009). However, it should be noted that such a field-specific O 3 VFI/Color was found to be always lower than 25% of color composition.
The successful assessments of O 3 VFI are mainly dependent on the observer's experience, and knowledge and color images such as photo guides of O 3 VFI are considered important tools, especially in field conditions where observation is required for several species and a large number of individuals. The validation of O 3 VFI is therefore crucial and must be conducted under experimental conditions and using microscopy observation (Vollenweider, Ottiger, and Günthardt-Goerg 2003;Bussotti et al. 2005Bussotti et al. , 2006Moura et al. 2018) to ascertain an O 3 VFI diagnosis by providing a mechanistic understanding of O 3 effect (Günthardt-Goerg and Vollenweider 2007;Faoro and Iriti 2009). However, the new methodology proposed here based on color composition can informatize welltrained observer's experience and digitalizes their visual information and knowledge of how to identify O 3 VFI. For example, the arithmetic manipulation of the RGB color channels has been used to detect a wide variety of plant disease symptoms (Barbedo 2017). Although further analyses for the specificity of O 3 VFI are still needed, a proper visual information for the O 3 VFI can be a simple informatic tool to establish an automated O 3 VFI identifier using a RGB image alternative to the traditional method of a time-consuming  Table 6 for a detail). validation by microscopy, thus improving the assessment and identification of O 3 VFI.

Comparison of flux-based threshold for the ozone visible injury onset between the FO 3 X and actual forests
In the 2000s, several field studies reported that O 3 VFI was shown in forest trees when AOT40 reached 10 ppm h (Vollenweider et al. 2019;Vanderheyden et al. 2001). However, as Vanderheyden et al. (2001) pointed out for forest tree species in Switzerland, the AOT40 threshold showing O 3 VFI varied from year to year. In fact, O 3 damage is closely related to stomatal O 3 uptake rather than O 3 exposure only (CLRTAP, 2017;Watanabe et al. 2021). Therefore, a consensus has increased in the scientific community to recommend the flux-based O 3 risk assessment on forest trees (Paoletti et al. 2022) suggesting that an equivalent stomatal O 3 dose results in a similar O 3 damage over various species with a different sensitivity to O 3 (Reich 1987;Feng et al. 2018). In fact, the AOT40-based threshold for the onset of O 3 VFI for the O 3 resistant evergreen species, A. unedo, was 10-fold higher than that for the O 3 sensitive deciduous species, A. glutinosa. Interestingly, the flux-based threshold corresponding to the first symptom onset for A. unedo was rather similar to that for A. glutinosa. In fact, A. unedo, which has a low g max limiting stomatal O 3 uptake, showed O 3 VFI only in the autumn season, while A. glutinosa showed O 3 VFI in early summer even under ambient O 3 conditions because it shows a very high stomatal conductance and stomatal O 3 uptake easily exceeds the critical range of O 3 dose that can be detoxified.
According to the field observations in MOTTLES sites, Sicard et al. (2020) proposed flux-based thresholds for O 3 VFI of 5 and 12 mmol m −2 POD 1 in dominant conifers and broadleaved trees, respectively, while 11 mmol m −2 POD 1 was required for the presence of O 3 VFI in various sensitive tree species present within the LESS . Although categorizing plant types is useful for setting the critical standard for forest protection, plant responses to O 3 are rather species-specific, as confirmed by the high fluxbased threshold reported for Pinus halepensis (e.g., 8.2 mmol m −2 POD 1 ) in Southeastern France (Sicard et al. 2016). At the FO 3 X, the flux-based threshold for the O 3 VFI onset was ranged from 4.9 to 18.1 mmol m −2 POD 1 . It seems that the FO 3 X-derived threshold also explained well the presence of O 3 VFI for A. glutinosa and Q. pubescens in the MOTTLES conditions. However, the threshold observed for the injury onset at the FO 3 X did not often match the presence of O 3 VFI in actual forests. The incidence of the VFI was lower for S. aucuparia and V. myrtillus in actual forests compared to the FO 3 X condition. Indeed, S. aucuparia, and V. myrtillus showed O 3 VFI in early summer at FO 3 X even under AA conditions, while they were not always found symptomatic in actual forests. Also, for the less sensitive species such as Q. robur, very limited O 3 VFIs were found under the MOTTLES conditions even though a relatively high POD 1 was observed (>20 mmol m −2 POD 1 ). There is a possibility that g s of field plants would be different with that of pot plants. In fact, Beikircher et al. (2021) suggested a lower g s in field-grown mature Acer pseudoplatanus trees than in potted-seedlings. On contrary, Samuelson and Kelly (1997) indicate a greater g s in mature Quercus rubra trees than seedlings. Actually, a site-specific g s parameter may provide a more precise estimation of POD 1 in the MOTTLES sites. In addition, in a manipulative experiment, plants are usually potted, well-watered, isolated, and totally exposed to the atmospheric O 3 conditions; therefore, the higher incidence of symptoms can be related to the lack of stress compensation by an interaction of leaves within a canopy crown (Löw et al. 2006). Otherwise, a complex structure of forests with species mixtures may decrease herbivory, disease and other abiotic stresses, with increasing nutrient supply rates over the long term (Tilman, Isbell, and Cowles 2014). Interestingly, according to the multivariate analysis on MOTTLES datasets, O 3 VFI decreased with an increasing number of species in the LESS. LESS is a forest edge, and therefore it is generally characterized by a high number of species with different growth forms determining a multi-layered structure (Ranney, Bruner, and Levenson 0000;Łuczaj and Sadowska 1997). As a vertical O 3 gradient can be observed in a forest canopy (Ollinger, Aber, and Reich 1997), the presence of nonsensitive species in the upper layer can establish a protective mechanism for more sensitive species in the bottom layer. Such a hypothesis is in line with the assumption that mixed stands are more resilient to environmental stress than monospecific ones (Grossiord et al. 2014). Eichhorn et al. (2005) and Pollastrini et al. (2016) identified species diversity as a relevant factor that positively influences the crown conditions (i.e., reduced defoliation) in North European and Mediterranean forests.
Only R. ulmifolius showed the opposite behavior, i.e., it was symptomatic in the field even with a relatively low POD 1 below the level in which the VFI was observed at the FO 3 X. Such differences can be explained by the strong competitive habit of the species for light (Gaudio, Balandier, and Marquier 2008). R. ulmifolius shrubs tend to overwhelm other species and, therefore, are more exposed to atmospheric conditions.

Conclusions
The results confirmed that a new generation free-air O 3 fumigation facility such as FO 3 X is an appropriate tool for the validation activities of field-observed O 3 VFI. New imaging analysis of color composition for O 3 VFI indicates that a major part of the colors were similar between the field and the FO 3 X. Such digitized information will provide an advancement in the approach for O 3 VFI assessment from the bio-informatic point of view such as a mobile application as diagnostic tool for field scientists. A further calibration for estimating O 3 VFI by color composition (i.e., differences between O 3 VFI and other visible injury, color distribution in the leaves) is recommended for a more extensive range of O 3 -sensitive species because other abiotic stress factors may sometimes mask the O 3 VFI.
The flux-based threshold for the O 3 VFI onset at the FO 3 X was ranged from 4.9 to 18.1 mmol m −2 POD 1 . Although this FO 3 X-derived threshold also explained the presence of O 3 VFI for A. glutinosa and Q. pubescens under real-world conditions, it did not always match with the observation for the other species. Q. robur, S. aucuparia and V. myrtillus were relatively resistant in actual forests rather than the FO 3 X experiment, given that they were not visibly affected in the MOTTLES sites even though POD 1 exceeded the FO 3 X-derived threshold. On the other hand, R. ulmifolius exhibited the O 3 VFI even at a relatively low POD 1 . Although the mechanisms are still unknown, the multivariate analysis indicated an interaction of the presence of various species on O 3 VFI and suggested the importance of biodiversity and continuous monitoring activities in the field.
European forests are considerable areas of the world terrestrial biodiversity. In MOTTLES, we investigated 23 tree species. The ozone FACE (Free Air Controlled Exposure) experiments need to be expanded to more species to realize a proper and representative assessment of forest health under O 3 pollution.