Interannual variability of ozone fluxes in a broadleaf deciduous forest in Italy

Ozone (O₃) deposition on vegetation has been a topic of interest for the scientific community for 4 decades now (e.g., Wesely et al., 1978; Neumann and Den Hartog, 1985; Fan et al., 1990), as regards both the quantification of the O₃ budget in the troposphere and the evaluation of the impacts on vegetation. In fact, O₃ is a well-known threat to vegetation, being responsible for a series of negative effects ranging from the onset of visible leaf symptoms to significant reductions of plants photosynthetic performance and growth rate (Ashmore, 2005; Wittig et al., 2009; Marzuoli et al., 2019; Emberson, 2020).

Many efforts have been made in the past to measure the O₃ deposition on different terrestrial ecosystems, both in agricultural (e.g., Grantz et al., 1997; Gerosa et al., 2003; Lamaud et al., 2009) and forest sites (e.g., Rannik et al., 2012; Fares et al., 2014; Finco et al., 2017), and in particular to correctly quantify the stomatal dose absorbed by vegetation. Indeed, the negative effect of O₃ on plants is more related to the amount of pollutant taken up by leaves through stomata than the simple cumulative exposure. Ozone, in fact, can enter the leaf only through stomata, because deposition on the leaf cuticle and other plant surfaces tends to cause the disruption of the molecule (Cape et al., 2009). For this reason, the O₃ critical levels for the protection of vegetation have been recently redefined by the United Nations Economic Commission for Europe (UNECE) with the introduction of the Phytotoxic Ozone Dose (PODy), that is, the cumulative stomatal flux of O₃ above a detoxification threshold y (Convention on Long-Range Transboundary Air Pollution [CLRTAP], 2017). The O₃ dose absorbed by vegetation strongly depends on the environmental factors that influence the physiological regulation of stomata (Emberson et al., 2000), particularly on the soil water availability, which is directly correlated to the O₃ stomatal flux (Lin et al., 2020).

On the other hand, many studies have highlighted that the non-stomatal processes can play an important (Fowler et al., 2009) or even dominant (Gerosa et al., 2005; Gerosa et al., 2009a; Clifton et al., 2017) role in the total O₃ deposition on an ecosystem (i.e., the total amount of O₃ removed from the troposphere). These processes include the physical deposition processes and air chemistry reactions that lead to O₃ consumption, such as the deposition on dry and wet surfaces (Altimir et al., 2006; Cape et al., 2009), the deposition on soil (Stella et al., 2011) and chemical reactions with nitric oxide (NO) (Rummel et al., 2007) and biogenic volatile organic compounds (BVOCs) (Goldstein et al., 2004). A comprehensive review of the O₃ stomatal and non-stomatal deposition processes in forests, crops, water, and cryosphere can be found in Clifton et al. (2020), who also proposed a mechanistic formulation for the 2 main non-stomatal deposition pathways: deposition on dry and wet cuticles and deposition on soil. Those formulations, however, still need to be tested with new long-term measurements.

In fact, current knowledge of O₃ dry deposition is mainly based on a few O₃ flux datasets, often covering only short-time periods (Clifton et al., 2020; He et al., 2021). Only few multiannual series of O₃ flux measurements are available for forests (Mikkelsen et al., 2004; Neirynck et al., 2012; Rannik et al., 2012; Neirynck and Verstraeten, 2018), even though multiannual measurements are currently running at the Amazon Tall Tower Observatory (Andreae et al., 2015) and at the Virginia Forest Research Facility in the United States. In the temperate zone, some long-term O₃ flux measurements are available for evergreen, coniferous, and mixed forests (e.g., Lamaud et al., 2002; Gerosa et al., 2005; Hogg et al., 2007; Neirynck et al., 2012; Fares et al., 2014), but similar measurements for deciduous forest are still missing, except for a few short campaigns (Fuentes et al., 1992; Munger et al., 1996; Finkelstein et al., 2000; Michou et al., 2005; Le Morvan-Quéméner et al., 2018). A complete annual series of O₃ fluxes was reported by Zona et al. (2014) for a poplar plantation in Belgium. However, to our knowledge, the only multiannual O₃ flux series available for a deciduous forest so far is that reported by Clifton et al. (2017) for the Harvard Forest in Northeastern United States, although this ecosystem is not a pure deciduous forest, because it includes a minor percentage of conifers. Clifton et al. (2017) highlighted that for this site the interannual variability of O₃ fluxes could not be explained by meteorology and was largely attributable to non-stomatal deposition, as also concluded by Rannik et al. (2012) in their analysis of a long-term O₃ flux series taken in an evergreen coniferous forest in Hyytiälä (Finland).

Multiannual measurements of O₃ fluxes are therefore needed on the one hand to investigate the reason for the interannual variability, and on the other hand, to get solid data against which to evaluate the factors that influence both stomatal and non-stomatal deposition. Thus, our multiannual O₃ flux measurements in a pure deciduous temperate forest (i.e., without any conifer in their composition) aim to help close some of the knowledge gaps highlighted by Clifton et al. (2020) such as assessing the temporal variability of deposition processes and the contribution of stomatal uptake and chemistry of the canopy air to O₃ dry deposition. The studied forest is located in the Po Valley (northern Italy), within an aerological basin of great interest because of the high levels of air pollution and photooxidants in the summer season (Gerosa and Ballarin-Denti, 2003). In this forest, ecosystem-level O₃ flux measurements have been underway since 2012 as a follow-up to the European project ÉCLAIRE (2011–2015) aimed at evaluating the Effects of Climate Change on Air Pollution and Response Strategies for European Ecosystems (https://cordis.europa.eu/project/id/282910/reporting).

The objectives of this study are (1) to investigate the interannual, seasonal, and day-to-day variability of the O₃ fluxes in this forest type, (2) to assess the relative importance of stomatal and non-stomatal deposition, (3) to compare different partitioning approaches (based on H₂O and CO₂ fluxes) for the correct estimation of the canopy stomatal uptake, (4) to investigate the effect of the O₃ storage on the diurnal course of the O₃ deposition, (5) to identify the main environmental drivers of the temporal variability of the summertime fluxes, and (6) to assess the phytotoxic O₃ dose absorbed by the plants during the growing season and the related potential damage to plants. An additional purpose is to share a new multiannual database of O₃ fluxes that could help to improve and calibrate O₃ deposition models, to test models’ ability to reproduce interannual variations, and to verify the performances of different O₃ risk estimation metrics for forests on a multiannual basis.

The study site is a 235 hectare mature Oak-Hornbeam forest at the outskirts of the city of Mantua, in the middle of the Po Valley, Italy. It is a natural reserve named “Bosco della Fontana” declared by the European Union as Site of Community importance and Special Protection Zone of the network Natura 2000 (site code IT20B0011). The forest is one of the last remnants of the flood plain forests that covered the Po Valley in the past.

The dominant tree species are Carpinus betulus and Quercus robur, but Acer campestre, Prunus avium, Fraxinus ornus and oxycarpa, Ulmus minor, and Alnus glutinosa are also present. The canopy is closed and the mean canopy height is 26 m. The understorey is formed by Corylus avellana, Sambucus spp, Cornus mas, Crataegus oxyacantha and monogyna and Sorbus torminalis, with a shrubs layer of evergreen and semideciduous species (Ruscus aculeatus, Hedera helix, Rubus ulmifolius). A few individuals of alien species are also present in the forest: Quercus rubra, Juglans nigra, Platanus orientalis and Robinia pseudoacacia.

The growing season ranges from April to October: leaves emerge around mid-April and last on the trees until November. In the autumn–winter months, only grasses and evergreen understorey vegetation are active. The canopy Leaf Area Index (LAI) was measured periodically in the forest acquiring hemispherical photos with a Sigma 8 mm fisheye lens mounted on a Canon 400D camera. The digital elaboration of the canopy images was made with Hemisfer 3.0 software (WSL Switzerland). The annual evolution of LAI (Figure 1) shows values between 2.5 and 3.0 m² m⁻² during the summer season, with a steep increment from 0.9 m² m⁻² in the first weeks after leaf emergence and a slow decrease starting from September.

For decades, the forest management has been oriented toward reaching “old growth forest” status (Peterken, 1996), with no significant management except for the eradication of some alien species introduced during the 1950s and the retention of the woody debris to preserve the arthropod and bird biodiversity of the reserve.

The aspect is flat (25 m a.s.l.) and the soil is a loamy skeletal, mixed, mesic Petrocalcic Palexeralf (Campanaro et al., 2007) according to the USDA classification. The soil depth is 1.5 m, but a petrocalcic hardened layer, formed after the gradual deepening of the water table, lies between 0.80 and 1 m depth.

The site climate is typical of the Po Valley, with cold winters and hot summers (Longo, 2004). The mean annual temperature is 13.2°C (period 1840–1997; Bellumè et al., 1998), with an average temperature of 1.3°C in the coldest month (January) and 24.6°C in the warmest month (July). The mean annual precipitation is 658 mm with 2 maxima in May and October, and 2 minima in February and July. A period of hydrological deficit generally occurs in July, due to the combination of temperature and precipitation conditions. However, plants can rely on the underground aquifer as the water table can as shallow as 1.5 m below ground. The prevailing winds are generally from the East (E) and North-East (NE), particularly in spring and summer.

Inside the natural reserve, a 41-m scaffolding tower was installed in 2012 (45°11′52.27” N, 10°44′32.27” E) as part of the research infrastructure of the European FP7 Project ÉCLAIRE. The distance of the tower from the forest edge ranges between a minimum of 390 m to the South (S) and a maximum of 1,440 m to the NE.

The walk-up tower is equipped with a sonic anemometer at the top (METEK USA-1, D), along with a LI-7500 (Li-Cor, USA) for CO₂ and H₂O flux measurements, and a chemiluminescence fast O₃ analyzer (COFA, Ecometrics, I), for the O₃ flux measurements. The latter is a clone of the GFAS sensor (Güsten et al., 1992) and required consumable disks of coumarin which have been replaced every 4 days. In order to correct the sensitivity decay of the coumarin disks, a conventional UV photometer (mod. O342 M, Environnement, France) was used as a reference for the chemiluminescence O₃ analyzers (for details, please refer to Finco et al., 2018). The conventional O₃ analyzer was calibrated every 6 months with a primary standard reference analyzer provided by Cosmic Technology s.r.l. (Italy).

In 2018, a second sonic anemometer (METEK USA-1, D), a second COFA and a fast hygrometer KH20 (Campbell sci., USA) were also installed below canopy at 8 m above ground level (a.g.l.) for about 1 year. Moreover, starting from 2018, hourly nitrogen oxides (NO_x) mole fractions were measured 1 m above the soil surface with an APNA 360 analyzer (Horiba, Japan).

At the top of the tower, sensors of incoming and reflected solar photosynthetically active radiation (PAR quantum sensors LI-190, LICOR, USA), air temperature and humidity (E2-ACT, Rotronic, Germany), air pressure (PTB101B, Vaisala, Finland), net radiation (NR-LITE, Kipp&Zonen, Netherlands), precipitation (rain gauge 52202, Young, USA), and leaf wetness (237-L, Campbell sci., USA) were also installed. A second Young tipping gauge was installed on the forest floor to measure precipitation below canopy. Along the vertical profile, temperature, and humidity probes (E2-ACT, Rotronic, Germany) were installed at 20 cm a.g.l., 1 m a.g.l., and then every 8 meters up to the top.

Soil temperature probes (PT100, GMR Strumenti, Italy) were placed in the soil at 5, 30, and 60 cm depth. Three Time Domain Reflectometry sensors (TDR CS616, Campbell sci., USA) were horizontally inserted in the soil to measure the soil volumetric water content, 2 of them at 30 cm depth and 1 at 100 cm depth. Three soil heat flux plates (HFP01-SC, Hukseflux, Netherlands) have been inserted at 5 cm depth to continuously detect soil heat fluxes.

The sonic anemometers and the other fast instruments were sampled at 20 Hz by a dedicated personal computer (PC), and the raw data were stored in 30-min-long files. Slow sensors were sampled every 30 s by two CR1000 dataloggers (Campbell sci., USA), located at the top and the bottom of the tower. Half-hour averages of the acquired data were automatically stored in a PC.

Additionally, 4 automatic static chambers for CO₂ flux measurements (SassFlux, Ecometrics, Italy), with their dedicated control system, were placed on the ground around the tower to measure the soil respiration (for details, please refer to Gerosa et al., 2014).

Measurements reported in this paper range from January 1, 2013, to December 31, 2020. However, 2016 was excluded from the elaboration and discussion, because of a series of instrumentation and power supply problems that affected the monitoring campaign.

Half-hour raw averages were processed following the eddy covariance technique (EC). First of all, they were despiked according to Vickers and Mahrt (1997), and the gaps were linearly interpolated following a spline procedure. Then, a double rotation was applied to the instantaneous wind vectors, the first rotation to align the reference system of the sonic anemometer to the mean horizontal wind stream direction, and the second rotation to zero the mean vertical wind component (Lee et al., 2004). Samples with tilt rotation that exceeded 5° were discarded.

The rotated data were linearly detrended (Aubinet et al., 2012) and the fluxes were calculated as the covariance between the fast vertical wind speed and the scalar of interest. For water and O₃ fluxes, the lag which maximized the covariance was chosen. Covariances with lag values that differed from the mode by more than 10% were discarded.

Cospectra were calculated to assess the high frequency flux losses of the fast O₃ sensor, and the correction factors were found by an Ogive-analysis (Rummel et al., 2007). The identified corrections ranged between 0 and 1.5% of the raw O₃ fluxes, according to the stability conditions, with a mean value of 1.02%.

The Schotanus correction (Schotanus et al., 1983) was applied to the sensible heat fluxes, and the WPL correction (Webb et al., 1980) was applied to H₂O, CO₂, and O₃ fluxes to account for density fluctuations.

The correction proposed by Boylan et al. (2014) to account for the sensitivity variations of the coumarin dye in fluctuating air humidity was found to be negligible (peak diurnal value of 0.28 nmol m⁻² s⁻¹, on average) and, as a consequence, was not applied.

The fulfillment of stationarity conditions was checked according to Foken and Wichura (1996) and nonstationary samples were discarded. Moreover, for the O₃ fluxes elaboration, samples for up to 2 h immediately after every change of the coumarin disks were excluded from the analysis.

For the CO₂, H₂O, O₃, and sensible heat fluxes, the storage flux below the measurement height was calculated according to Rummel et al. (2007), and the fluxes were corrected accordingly. More details on this procedure can be found in Finco et al. (2018). Finally, missing data were filled using the Mean Diurnal Variation method as proposed by Aubinet et al. (2012) and Moffat et al. (2007).

Considering all the missing data and all the data exclusions in the different steps above described, the resulting dataset of flux measurements (i.e., not-gapfilled data) amounted to 53.12% of the total 7 years data.

The measured O₃ deposition fluxes in dry canopy conditions were partitioned between the stomatal and non-stomatal components following 3 different approaches.

The first approach derived the canopy scale stomatal resistance from the measurement of water fluxes and by inverting the classical Penman–Monteith equation for evapotranspiration. This approach is based on the energy balance and implicitly assumes that in dry canopies all the observed water fluxes are due to stomatal transpiration. Once the stomatal resistance to O₃ R_StomO3 is derived, the non-stomatal resistance R_NonStomO3 is obtained by considering a resistive deposition network where R_StomO3 and R_NonStomO3 are set in parallel and their summation is equal to the surface resistance R_C which, in turn, is obtained from the total O₃ deposition resistance minus the atmospheric resistance R_a and the sublaminar resistance R_b. Details can be found in Gerosa et al. (2005).

The second approach derived the canopy scale stomatal resistance from the CO₂ fluxes, as described by Lamaud et al. (2009). This approach is based on the observation that regardless if the canopy is dry or wet, there is a relationship between the stomatal conductance to O₃ g_StomO3 = 1/R_StomO3 and the CO₂ assimilation flux (A). The latter can be obtained from the direct measurements of the net ecosystem exchange of CO₂ (NEE) by subtracting the heterotrophic soil respiration flux of CO₂ (R_H). Since the photosynthetic assimilation occurs only with the daylight, g_StomO3 can be estimated only for diurnal hours.

The third approach was a modification of the first one, where the assumption that all the H₂O fluxes are due to stomatal transpiration was modified. The amount of water transpired by the canopy was estimated by subtracting the H₂O evaporated directly from the soil below the canopy. The latter was taken from the measurements of H₂O fluxes made with EC below canopy in 2018. Then the canopy transpiration obtained was used in the inverted Penman–Monteith equation instead of the latent heat flux (LE), to yield the R_Stom. The resistive network, the geometries (h_canopy = 26.0 m, z_m = 41.0 m, z₀ = 2.6 m, d = 17.4 m), as well as the calculation of R_a and R_b were the same for all the 3 approaches. The big-leaf equivalent to the forest canopy was assumed as a hypostomatous leaf (Monteith and Unsworth, 2014). Air density (ρ), air specific heat (c_p), and the latent heat of vaporization (λ) were adjusted to the half-hourly conditions of pressure (P), temperature (T), and specific humidity (q). In addition, the diffusion coefficients in air of O₃, CO₂, and H₂O were not considered constant but corrected to the actual T and P conditions following Massman (1998), as for the Schmidt numbers of O₃, CO₂, and H₂O.

Based on the results obtained from the comparison of the 3 abovementioned approaches (see Results), all the presented stomatal and non-stomatal fluxes have been calculated with the third approach.

The calculation of the PODy index for each year of the experiment was possible thanks to the availability of the O₃ stomatal flux measured. The instantaneous O₃ stomatal fluxes were integrated for the time interval April 1 to September 30 (the growing season for deciduous plants), with the application of an O₃ detoxifying threshold of 1 nmol m⁻² s⁻¹ to account for the detoxification capacity of forest trees (Buüker et al., 2015; CLRTAP, 2017). The estimation of the effects in terms of biomass growth reduction was based on the available dose–response relationships described in the UNECE Mapping Manual (CLRTAP, 2017) which defines the critical levels for the protection of various forest plants species in different geographical contexts. In particular the potential growth reductions due to the POD1 values measured at Bosco Fontana were assessed using the dose–response relationships for the Mediterranean and for the continental broadleaf deciduous forests.

The interannual variability of total O₃ fluxes (F_TotO3) in response to stomatal (g_StomO3) and non-stomatal (g_NonStomO3) conductance, and to O₃ mole fraction, was evaluated with a 1-way analysis of variance (ANOVA) on the highest monthly mean of F_TotO3 recorded in each year always in July. Stomatal and non-stomatal conductance, and O₃ mole fraction were considered as fixed factors in 3 separate ANOVA tests and assigned at 2 levels (“high” and “low”) by comparing the monthly means recorded in correspondence of the F_TotO3 peak to the median of all the monthly mean values considered. The identification of the factors responsible for short-time variability (day-to-day and hourly variability) of F_TotO3, g_StomO3, and g_NonStomO3 was made with a backward stepwise regression on the daily means and half-hourly values (not-gapfilled data) of the summer seasons. All the analysis was performed with Statistica 12.0 software (Statsoft Inc., USA).

Table S1 describes the climatic conditions of the different seasons during the 7 years of measurements. These data will be a useful reference when discussing the results.

The year 2013 was characterized by the coldest, most humid and rainy spring; 2014 exhibited the coldest and most rainy summer; 2015 was dry and arid, in both spring and summer, with the lowest soil water content (SWC) values; 2017, as 2015, had a quite arid spring, but it was followed by a cool and humid summer; 2018 had the summer with the highest soil water availability and a spring with the lowest O₃ mole fractions; 2019, after a spring with the highest levels of soil water availability, presented the hottest and most arid summer, which was also characterized by the highest O₃ mole fractions; finally, 2020 showed the hottest and most arid spring, followed by a summer with the lowest mean O₃ mole fraction among the 7 considered summers.

An overview of all the deposition fluxes, as well as their partition into stomatal (F_StomO3) and non-stomatal (F_NonStomO3) components, is reported in Figure 2. The ANOVA test on the monthly averages of F_TotO3 for all the 7 years (Figure 2a) highlighted a certain interannual variability of the total O₃ deposition when the year is used as a grouping factor for the test (P = 0.088, n.s.). However, the difference in the monthly average of F_TotO3 between the two most contrasting years, that is, the years with the highest and the lowest annual peak of F_TotO3 (2018 and 2020), was significant at a paired t test (P = 0.004), although the sample size was relatively small (n = 24). This variability was mainly driven by the stomatal component of the flux: when the annual peak of the stomatal flux was high, the peak of the total flux was also high (e.g., year 2018: F_StomO3 = 7.51 ± 0.26 and F_TotO3 = 11.92 ± 0.38 nmolO₃ m⁻² s⁻¹), and when the annual peak of the stomatal flux was low, the annual peak of the total flux was also low (e.g., year 2020: F_StomO3 = 3.48 ± 0.13 and F_TotO3 = 6.36 ± 0.20 nmolO₃ m⁻² s⁻¹). On the contrary, the annual peak of the F_NonStomO3 showed a much lower variation over the years, ranging between 2.87 ± 0.12 nmolO₃ m⁻² s⁻¹ of July 2020 and 4.41 ± 0.23 nmolO₃ m⁻² s⁻¹ of July 2018 (Figure 2a). The average diurnal cycle of the F_StomO3 showed a marked seasonal variation (Figure 2b), with the average daily peak ranging from a minimum of 0.93 ± 0.09 nmolO₃ m⁻² s⁻¹ in January (1:00 pm) to a maximum of 8.89 ± 0.33 nmolO₃ m⁻² s⁻¹ in July (1:00 pm), while the non-stomatal flux showed a lower diurnal variation in the different months, with the average daily peak ranging from a minimum of 2.7 ± 0.24 nmolO₃ m⁻² s⁻¹ in December to a maximum of 4.43 ± 0.36 nmolO₃ m⁻² s⁻¹ in July.

The comparison of the 3 different flux-partition approaches was made on the dry canopy data for 2018, when H₂O flux measurements below canopy were fully available. Moreover, 2018 was the year that showed the highest stomatal fluxes of both O₃ and H₂O. Below canopy measurement of LE fluxes are useful for estimating the fraction of water evaporated from the soil, a fraction that should be subtracted from the LE measured above the canopy to avoid attributing this amount of water to stomatal transpiration, thus leading to an overestimation of the stomatal fluxes in the partition process. The application of the approach developed by Lamaud et al. (2009) resulted in a good linear relationship between the stomatal conductance to O₃ (g_StomO3) and the CO₂ assimilation flux (A), g_StomO3 = α·A, with α = 19 cm s⁻¹/mmol m⁻²s⁻¹ and R² = 0.93. The g_StomO3 calculated from nonnegative A values resulted in the stomatal flux reported in Figure 3a (blue line), which in summer months was lower than the same flux calculated with the classic Penman–Monteith approach (green line). This difference reflects the fact that the Penman–Monteith approach may not completely exclude the non-stomatal water-vapor flux, that is, the water flux from soil, while the Lamaud’s approach explicitly excludes it because it is based on the analysis of the CO₂ fluxes. Furthermore, it should be remembered that the Lamaud’s methodology was developed for maize canopies where the effect of soil evaporation is less relevant than in forests, because maize canopy is denser and more closed. Nonetheless, during the rest of the year, the 2 approaches were fairly in agreement.

For the application of the third approach, we started from the observation that regardless of the season, the below-canopy water fluxes were well correlated with the above-canopy water fluxes (LE_below = β LE_above), with a proportionality coefficient β = 0.1991 and R² = 0.86. Hence, we assumed that (1 − β)·LE_above was the LE quota due to the canopy transpiration, and we used it also for the partition process of all the other years. The resulting stomatal flux for the comparison year is shown in Figure 3a (red line). The estimation with this method is in agreement with that obtained with the approach of Lamaud et al. (2009) for most of the year (+1.2% on average, with respect to the Lamaud’s approach), even though it showed slightly higher differences in the midday hours of July (+17.1%) and August (−17.2%).

Figure 3b shows that in the daylight hours, the stomatal fraction of the total deposition was on average 16.4% lower than the fraction estimated with the Penman–Monteith inversion (the first approach), and it ranged between 20% and 60% of the total O₃ deposition, with an average value of 42.6% for the whole year and 52% for the spring and summer seasons. It is worth noting that despite the high interannual variability of the F_TotO3 shown in Figure 2a, the stomatal fraction was similar throughout the 7 years. In fact, the maximum fraction of O₃ absorbed by the canopy crowns was on average 60% of the total O₃ deposition on the forest (Figure 3b), regardless the total amount of O₃ received by the ecosystem in the different years (compare for example 2018 vs. 2020, 2 years with very different total O₃ fluxes).

The O₃ storage represents a temporary accumulation of O₃ in the air column below the EC measuring point or a temporary O₃ release out of the same air column, and it does not necessary represent a true O₃ removal or production at the target surface. The effect of the storage is reported in Figure 4, which presents a comparison between the total deposition fluxes obtained by considering (blue line) or not considering (black line) the storage flux. The effect of the storage caused an increase of the evening deposition fluxes during the vegetative months that could be due to the decrease of the O₃ mole fractions occurring in those hours, because O₃ production was no longer supported by photochemistry, and physical deposition or chemical consumption processes on surfaces tended to prevail. However, in the morning hours, when O₃ mole fractions increase and nitric oxide (NO) mole fractions below canopy are relatively high (Figure 4, yellow line), the storage correction could not fully represent the real deposition flux occurring in those hours, since for a reactive species like O₃ some of the stored gas may be destroyed within the canopy by NO titration and potentially by reaction with BVOCs, slowing down the O₃ mole fraction increase observed in the morning.

The deposition velocities (Vdep_O3) illustrated in Figure 5a and b showed a similar mean diurnal course in spring, summer, and autumn, with a peak between 5.2 ± 0.2 mm s⁻¹ (summer) and 6.7 ± 0.4 mm s⁻¹ (autumn) at 10:00 am. On the other hand, in winter Vdep_O3 was on average almost doubled, and the mean diurnal course showed a peak of 17.8 ± 4.4 mm s⁻¹ in the early morning hours (07:00 am).

The diurnal course of g_StomO3 (Figure 5c), averaged over the 7 years, was surprisingly similar in all the 4 seasons, with a maximum value around 3.5 mm s⁻¹. Only a slightly different diurnal shape and a greater uncertainty in autumn and winter can be observed. Nonetheless, the summer diurnal course showed an interannual variability that matches the interannual variability of both total and stomatal O₃ fluxes (Figure 5d). The highest and lowest values were recorded in 2018 and 2015, respectively. These 2 years were also characterized by the highest (year 2018) and lowest (year 2015) SWC in summer.

The non-stomatal conductance to O₃ (g_NonStomO3) is reported in Figure 5e and f. The diurnal cycle showed an early morning peak both in summer and winter, even though in winter the peak was about one order of magnitude greater than in summer (Figure 5e). Then, in the afternoon, g_NonStomO3 declined to low values in all the seasons. The high mole fractions of NO in winter (cfr Figure 4, yellow line), particularly in the morning, suggest that the non-stomatal deposition in winter was linked to O₃ titration by the NO advected in the canopy without leaves.

The temporal alignment of the morning peak of g_NonStomO3 with the morning peak of NO mole fractions in the summer season suggests a similar process even for the growing season months, but with NO originated by soil as well described by Finco et al. (2018) for the same site. On the contrary, the g_NonStomO3 in the evening and in the night hours during the vegetative period showed values which are comparable with the typical values of the leaf cuticular resistance (0.4 mm s⁻¹, R_NonStom = 2,500 s m⁻¹ as from Emberson et al., 2000 and Simpson et al., 2012), once they have been rescaled according to the LAI, that is, between 0.1 mm s⁻¹ and 0.58 mm s⁻¹ (R_NonStom between 1,000 and 1,700 s m⁻¹). Higher g_NonStomO3 values during the night were due to high humidity values or dew deposition (see next section). A summer interannual variability was also evident in Figure 5f. For example, in 2018, the summer mean diurnal cycles of g_NonStomO3 peaked at 5.5 ± 1.8 mm s⁻¹ in the first hours of the morning, while in 2020, the peak diurnal value was halved, 2.5 ± 0.5 mm s⁻¹.

The main drivers of the stomatal and non-stomatal deposition are showed in Figures 6 and 7, which are based on the analysis of binned half-hourly data. The stomatal uptake of O₃ was mainly driven by the SWC and by all those factors that affect the stomatal behavior (Figure 6). The increase of g_StomO3 was directly related to soil water increase (Figure 6a) and inversely related to air dryness increase (Figure 6b), in both summer and winter. In summer, g_StomO3 was around its average minimum value until SWC exceeds the threshold of 0.14 m³_H2O m⁻³_soil. In winter, the increase of g_StomO3 required a higher SWC because the evergreen species of the understorey vegetation has a less developed rooting system compared to mature deciduous trees. The functional relationship between the SWC and g_StomO3 proposed by the UNECE Mapping Manual (CLRTAP, 2017) for a broadleaf deciduous forest, and adopted by the DO₃SE model (as f_SWC), does not seem to describe adequately what was observed in our forest. However, a direct comparison between our data and the DO₃SE parameterization should be taken with caution, since our measurements come from a whole ecosystem with mostly adult trees, while the DO₃SE parameterization is based on leaf-level measurements mainly taken on young trees in mesocosms (e.g., Open Top Chambers). Moreover, the parameterization of f_SWC strongly depends on the soil type, and it should not be surprising that the thresholds of SWC below which g_StomO3 starts to decrease (SWC_max) and reaches its minimum (SWC_min), vary among the different ecosystems. Nonetheless, in our forest ecosystem and with our soil type, the functional relationship of the DO₃SE model fits better for the summer months if the values 0.13 m³ m⁻³ and 0.23 m³ m⁻³ are assigned to SWC_min and SWC_max, respectively.

Regarding the relationship with vapor pressure deficit (VPD), in summer g_StomO3 decreased exponentially as VPD increased above 1.2 kPa and remained around a constant value below this threshold, until VPD got close to zero, when air was close to saturated, and g_StomO3 started to increase more than exponentially. The threshold of 1.2 kPa for VPD is similar to the value of 1.1 kPa suggested by the UNECE Mapping Manual (CLRTAP, 2017) for the beginning of stomatal closure in modeling exercises for this kind of vegetation. However, above a VPD of 3.1 kPa g_StomO3 reached a minimum value of 1.26 mm s⁻¹ (36% of g_max), a value which is greater than the value of 0.35 mm s⁻¹ (10% of g_max) indicated by the UNECE. The g_StomO3 showed a typical parabolic behavior, increasing with temperatures between 13°C and 18°C in summer (Figure 6c), showing an optimum around 23°C, and decreasing at higher values. The value of the optimum temperature is very close to that adopted by the DO₃SE model for Mediterranean broadleaf deciduous forest (22°C), as suggested by the UNECE (CLRTAP, 2017), but the minimum temperature of DO₃SE (0°C) appears to be appropriate only for the spring season in our case. Instead, the summer data show a better fit with the DO₃SE parameterization for the stomatal response to air temperature if the minimum temperature is set to 11°C, while keeping the optimum and maximum temperature to 23°C and 39°C, respectively.

Regarding the effect of O₃ (Figure 6d), in all the seasons g_StomO3 tended to decline as O₃ mole fraction increased, but this was likely due to the co-occurrence of conditions that favor the stomatal closure. However, the negative effect of O₃ on stomata cannot be excluded, as shown by many authors in controlled experiments (Hoshika et al., 2012; Lombardozzi et al., 2012).

For the non-stomatal deposition (Figure 7), the negative relationship with mechanical (Figure 7a) and thermal turbulence (Figure 7b), as well as wind speed (data not shown), would exclude a deposition process based on the disruption of O₃ on leaf surface, which could be expected to increase at increasing wind speed and turbulence. The remarkable negative relationship with air temperature (Figure 7c) is probably the result of a temporal misalignment between the g_NonStomO3 daily cycle (Figure 5e) and the T daily cycle. The non-stomatal deposition, in fact, was higher when T was low and vice versa. This was related to the fact that in the first hours of the morning (characterized by low temperatures) NO mole fractions were high and O₃ removal by the chemical sink was also high. Thus, the negative relationship between g_NonStomO3 and air T does not exclude the thermal decomposition of O₃ on cuticles, barks, and nonliving surfaces.

Instead, the non-stomatal deposition was driven by air humidity because g_NonStomO3 tended to rise with relative humidity (RH, Figure 7d) and exponentially decreased at increasing VPD (Figure 7e). The link between non-stomatal deposition and air humidity has been shown also by Altimir et al. (2004) and Lamaud et al. (2002). Even water on leaf and bark surfaces plays a role in non-stomatal deposition. Figure 8 shows that the surface conductance to O₃ was higher when the canopy was wet compared to when it was dry (no dew on leaves and air temperature not higher than 2°C above dew temperature). This is particularly evident in autumn and winter. While in summer and spring the presence of water on the leaves reduced the cuticular conductance to O₃, as proposed by Zhang et al. (2002), in the other months (when broadleaf trees have senescent or no leaves at all), it is likely that the O₃ decomposition in water films is enhanced by the presence of soluble organic compounds in water, coming from senescent or dead vegetal material (Potier et al., 2017). On the contrary, the SWC did not influence the non-stomatal deposition in summer (Figure 7f) because in this season the canopy is closed and the soil has less influence on this deposition, while in winter soil water leads to a decrease of g_NonStomO3 because, when water fills the soil pores, the internal soil surface available for O₃ deposition decreases, as shown by Fumagalli et al. (2016).

Regarding the chemical sink, there was a general tendency for an increase of non-stomatal deposition at increasing NO mole fractions (Figure 7g), thus confirming what was already explained in the section about the effect of the storage (Figure 3a). The evidence of an O₃ chemical sink due to reaction with NO can also be seen in the relationship between the non-stomatal deposition and O₃ (Figure 7h). Two distinct regimes are evident in the graph, both in winter and in summer: a linear relationship below 20 ppb of O₃ mole fraction, and an exponential relationship above 20 ppb. The linear decline of g_NonStomO3 at increasing O₃ mole fractions below 20 ppb can be associated with O₃ reaction with NO, which occurs stoichiometrically 1:1. When O₃ mole fraction is above 20 ppb, the excess of O₃ leads to a slower exponential decline of g_NonStomO3 with increasing O₃. Again, looking at Figure 7e, it is worth noting that below VPD of 0.9 kPa, in summer nights (cross symbols), the non-stomatal deposition decreased when air gets closer to water saturation. This suggests that the greater non-stomatal deposition observed in daylight hours with VPD values below 0.9 kPa, that is, when stomata are fully open, is related to some chemical reactions with substances emitted by the stomata, likely BVOC and isoprene among others, as reported by Schallhart et al. (2016) for the same site.

The analysis of the flux variability and its determinants at different time scales was focused only on the summer season, the season in which O₃ fluxes are maximum and the deciduous forest has uniform conditions of LAI. This could give more emphasis to the role of the stomatal component, as non-stomatal processes seem to dominate in other periods of the year.

As shown in Figure 2, F_TotO3 presented the maximum monthly mean value always in July for all the 7 years of measurements, regardless its magnitude. The 1-way ANOVA test between July monthly means and the corresponding g_StomO3 (as fixed factor) monthly means subdivided in 2 dimensional classes (high values and low values) confirmed that the F_TotO3 peak in July (i.e., the maximum monthly value for each year) is significantly influenced by g_StomO3 (P = 0.026, n = 7). The same analysis performed on g_NonStomO3 or O₃ mole fraction as fixed factors (both at 2 levels) didn’t show any statistically significant effect on the maximum annual peak of F_TotO3 (P = 0.222 for g_NonStomO3, P = 0.781 for O₃ mole fraction, n = 7). In any case, the small sample size used for the test suggests some caution in generalizing these results until more multiannual measurements will became available.

At a shorter time level, such as day-to-day and half-hourly scale, the number of the available summer data (June–July–August) enables using a stepwise regression method, which allows the simultaneous analysis of a large number of variables and the identification of the predictors subset that most adequately describes the variability at that level. This analysis highlighted that at a daily level, g_StomO3 was a significant driver of the O₃ flux variability (while g_NonStomO3 was not), as evidenced by the backward stepwise regression (R² = 0.42), even though this method identifies also the O₃ mole fraction as a significant factor at this level (Table 1a). In turn, the day-to-day variability of g_StomO3 was confirmed to be governed by SWC and the daily mean value of VPD, with the addition of PAR, although the predictivity of the obtained regression model was low (R² = 0.28).

Instead, the determinants of the day-to-day variability of g_NonStomO3 were the mean air temperature and turbulence (u*), both inversely correlated to g_NonStomO3, as well as the soil temperature which can be considered as a descriptor of the seasonal climatic evolution because it varies gradually during the season. However, when the backward stepwise regression included the daily average NO mole fraction (available only from 2018), the model identified this latter factor as the only determinant of g_NonStomO3 (Table 1a, last column), although the overall predictivity was still very low (R² = 0.15).

At a shorter time scale (Table 1b), both g_StomO3 and g_NonStomO3 together with O₃ mole fraction were selected by the stepwise procedure as significant descriptors of the half-hourly O₃ flux in the summer season (R² = 0.35). The predictors of the half-hourly g_StomO3 variability were the same already identified for the daily time scale, with the addition, however, of RH, T_air, and u* (R² = 0.32). Regarding the half-hourly g_NonStomO3 variability, air temperature and humidity (T_air, RH) were added to the only predictor selected for the daily scale, that is, the NO hourly mole fraction (Table 1a, last column).

It is worth noticing that the stepwise regression models confirmed, for the selected drivers, the directions (positive or negative) of the relationships with g_StomO3 and g_NonStomO3 already found in the previous section, both on the half-hourly and daily time scale. The only exception was the VPD, which on a half-hourly scale was positively correlated with g_StomO3 (Table 1b) while, instead, g_StomO3 is expected to decrease at increasing VPD. This apparent contradiction is explained by the fact that, on the half-hourly time scale, the diurnal course of VPD is the same as the course of g_StomO3, that is, VPD is high around midday exactly when g_StomO3 tends to be maximum, and vice versa. In fact, when the half-hourly variability was removed from the analysis and only the daily averages were considered, VPD was inversely related to g_StomO3 (Table 1a) as expected.

The phytotoxic O₃ dose above a threshold of 1 nmolO₃ m⁻² s⁻¹ for the whole ecosystem (POD1_ECOSYS; Table S1) showed a remarkable interannual variability, ranging from a minimum of 17.77 mmolO₃ m⁻² (year 2015) to a maximum of 37.08 mmolO₃ m⁻² (year 2018), with an annual average of 26.37 mmolO₃ m⁻². Considering on average a canopy LAI of 2.5 m² m⁻² between April and October (Figure 1), the abovementioned annual average is equivalent to a POD1 of 10.55 mmol m⁻², for m² of projected leaf area (PLA; Table 2). Since the dose–response relationships available for forest trees are based on POD1 values referred to LAI = 1, the annual values of POD1 can be used to estimate an annual growth rate reduction of the total biomass ranging between 1.83% and 4.30% (2.93% on average over the 7 years) if this forest is considered as a Mediterranean deciduous oak forest, or between 4.57% and 9.99% (6.98% on average) if it is considered as a continental forest (Table 2).

Few papers have dealt with O₃ deposition over a temperate deciduous forest such as Bosco Fontana. The summer daily course of Vdep_O3 observed at Bosco Fontana (Figure 5a) was slightly lower than those reported by Clifton et al. (2017), who showed the average daily courses of Vdep_O3, measured in summer from 1990 to 2000 at the Harvard Forest (USA). Clifton et al. (2017) reported summer daily peaks of Vdep_O3 between 5.0 ± 0.4 mm s⁻¹ and 10.0 ± 1.0 mm s⁻¹ while at Bosco Fontana the daily peaks were between 4.3 ± 0.4 mm s⁻¹ and 7.0 ± 0.4 mm s⁻¹ (Figure 5b). Furthermore, Finkelstein et al. (2000) and Visser et al. (2021) reported average daily peaks of the Vdep_O3 between 8.0 ± 0.2 mm s⁻¹ and 10.0 ± 0.1 mm s⁻¹ for mixed temperate forests. Deposition velocities in winter were remarkably higher than in summer (Figure 5a) with a peak value of 17.8 ± 4.4 mm s⁻¹ in the early hours of the day. The winter enhancement of deposition velocity is related to an increase of NO mole fractions, air humidity, and surface wetness in that season, all factors that favor O₃ removal by non-stomatal processes, as described below.

The average daily peak of canopy conductance (g_cO3) observed at Bosco Fontana in summer (Figure 8) was 7.1 ± 1.1 mm s⁻¹ for the case of a dry canopy, and 11.4 ± 1.7 mm s⁻¹ for the case of a wet canopy. These values are comparable to the value of 10.0 ± 0.3 mm s⁻¹ reported by Clifton et al. (2017) for the average daily peak of g_cO3, without any distinction between dry and wet canopy. A g_cO3 daily peak of 12.0 ± 0.5 mm s⁻¹ (averaged over 3 years of measurements) was instead reported by Visser et al. (2021) for a deciduous forest in northern Italy.

The average daily peak of g_StomO3 recorded in summer at Bosco Fontana (3.6 ± 0.2 mm s⁻¹; Figure 5c) was lower than those reported by Hogg et al. (2007), 4.5 ± 0.2 mm s⁻¹ for a mixed temperate forest, and Visser et al. (2021), 7.0 ± 2.0 mm s⁻¹ for a deciduous forest. Nonetheless, these differences are not surprising since stomatal conductance is strongly dependent on climatic conditions and particularly on the water availability in the soil. At Bosco Fontana, in fact, the g_StomO3 was governed mainly by the SWC, but also by the air dryness and temperature. The relationships found between g_StomO3 and these factors are quite similar to those adopted and parameterized in the CLRTAP (2017) for the stomatal conductance models used to calculate the POD in the O₃ risk assessment procedures. Nonetheless, considering these results, some adjustments of the threshold values adopted for the parameterization of temperate deciduous forests seem necessary.

The presence of nonnegligible stomatal flux at night and in the winter seems surprising. However, a significant nighttime g_StomO3 was reported from direct leaf measurements in a Californian forest (Grulke at al., 2004), from sap-flow measurements in a Mediterranean ecosystem (Mereu et al., 2009), and from measurements made with soil impermeabilized lysimeter in crops (Resco de Dios et al., 2015). As highlighted by Fowler et al. (2009) and Cailleret et al. (2018) in their comprehensive reviews, these results complicate the simplistic view of day–night differences in O₃ sinks in vegetation, with implications for O₃ risk assessment for vegetation. The detection of winter stomatal flux, on the other hand, was probably due to the presence of a dense understorey layer composed largely of evergreen and semideciduous species, with a remarkably high LAI (2.5 m² m⁻²) as reported by George et al. (2021) in their comparative study performed across 29 different forest sites in Europe. The understorey species were active also in autumn and winter during the central part of the day, as confirmed by the significant CO₂ assimilation fluxes recorded during those seasons (data not shown).

In summer, the average daily peak of g_NonStomO3 at Bosco Fontana was 3.9 ± 0.7 mm s⁻¹, lower than the values reported by various authors for other forest ecosystems. For example, Hogg et al. (2007) found an average daily maximum around 14.5 ± 2.0 mm s⁻¹ and Visser et al. (2021) around 5.0 ± 2.0 mm s⁻¹. In winter instead, g_NonStomO3 values were almost one order of magnitude higher than those detected in the other seasons (Figure 5e), providing evidence for a strong intra-annual variation of the non-stomatal deposition processes. Moreover, g_{NonStomO3 was} always greater than g_StomO3 in the evening hours, when the deposition on cuticles was prevailing, or at night when there was a reaction of O₃ with the NO emitted from the soil, or, again, in the early hours of the morning when the O₃ dissolved on the dew deposited on the canopy. The importance of the O₃ scavenging by the NO emitted below canopy at our site has already been shown by Finco et al. (2018), who evaluated the magnitude of this non-stomatal deposition process in a 1-month long measurement campaign in summer. Similar results were also reported by other authors for other forest ecosystems (e.g., Walton et al., 1997; Dorsey et al., 2004). The role of high air humidity, dew, and surface wetness in the non-stomatal processes has been repeatedly described by Fuentes et al. (1992), Lamaud et al. (2002), Gerosa et al. (2009a), and by Altimir et al. (2004; 2006) and Potier et al. (2015) who also proposed an appropriate parameterization. The positive influence of air humidity and dew on the g_NonStomO3 has also been experimentally demonstrated by Sun et al. (2016) who, by artificially inducing stomata closure, showed that O₃ deposition on leaves increased at increasing RH.

Unlike Lamaud et al. (2002), who found that non-stomatal deposition was linked to impaction processes and air penetration into the canopy, no positive relationship with atmospheric turbulence was found in our case. Similar findings were reported by Gerosa et al. (2009a), who observed a slightly negative correlation between g_NonStomO3 and u* in a maquis ecosystem. At Bosco Fontana, this result could be due to the co-occurrence of several factors that tend to favor O₃ destruction in conditions of low turbulence, such as the high levels of NO during the early daytime hours when u* was low.

The role of temperature with respect to g_NonStomO3 contrasted with what reported by Coyle et al. (2009), Hogg et al. (2007), and Fowler et al. (2001), who showed a positive relationship between the g_NonStomO3 and air temperature. Those authors concluded that a process of thermal decomposition of O₃ took place on the leaf surfaces facilitated by epicuticle waxes, confirming some of the results obtained by Cape et al. (2009) in lab experiments with simulated leaf surfaces. However, at our site the relationship between air temperature and g_NonStomO3 was negative (Figure 7c), suggesting that this process could have been masked by other processes that contributed to reducing the g_NonStomO3 as the temperature increased.

The great intra-annual variation of the g_NonStomO3 can be explained as the result of the combined effect of the seasonal variation of NO mole fraction and air humidity in the trunk space. NO mole fraction, in fact, rises in winter due to the increase of residential heating and vehicular traffic, particularly in the early hours of the morning. Moreover, in winter, NO is more easily advected in the intra-canopy space from the nearby urban areas because of the absence of leaves on the trees, thus enhancing the O₃ scavenging process. Air humidity also rises in winter with values of RH always above 85% and often close to 100% because of the presence of fog. These conditions, together with the enhanced presence of wet surfaces, lead to an increase of the non-stomatal removal of O₃ (Altimir et al., 2004; Altimir et al., 2006; Gerosa et al., 2009a), even though O₃ mole fractions and vertical fluxes in this season are low.

The interannual variability of O₃ fluxes could be generally attributed both to the high mole fractions of O₃ in some years compared to others, and to the variability of g_StomO3 and g_NonStomO3. However, the ANOVA test performed on the F_TotO3 means of July showed that the interannual variability of the summer O₃ fluxes at our site was significantly influenced by the g_StomO3 alone, and in turn to its drivers. This is particularly evident by comparing the year 2018 with the other years. These results are partially in contrast to what Clifton et al. (2017) found for the Harvard Forest, where non-stomatal deposition seemed to drive the interannual variability of O₃ fluxes. Moreover, Clifton et al. (2017) and Rannik et al. (2012) did not find any strong relationships with meteorology at hourly or daily scales and concluded that the unexplained interannual variability was likely due to processes that may act at a longer temporal scales (seasonal and multiannual) which may escape their analysis. At Bosco Fontana, where g_StomO3 seems to dominate the deposition in the spring and summer months, the long-term variability seems related to the seasonal variation of air temperature and SWC (Table S1). These findings underline once more the importance of multiannual measurements, as remarked by Clifton et al. (2020), since only this kind of measurement can capture features that could be missed at shorter time periods.

On a daily scale, the most significant drivers of the summer F_TotO3 were g_StomO3 and O₃ mole fraction (Table 1), partially confirming the results of Clifton et al. (2019), who also found that the stomatal deposition significantly guided the day-to-day variability of summer O₃ deposition velocity at the Harvard Forest. However, the same authors reported that also the dry–cuticular deposition (a non-stomatal process) contributed to the day-to-day variability, while in our site the g_NonStomO3 component showed to be a significant driver only at half-hourly time scale (Table 1b).

The average stomatal fraction of the O₃ flux at our site (42.6% for the whole year and 52% in summer) was higher than those detected in our previous works in other ecosystems, both agricultural (e.g., Gerosa et al., 2003; Gerosa et al., 2004) and forest (e.g., Gerosa et al., 2005; Gerosa et al., 2009a; Gerosa et al., 2009b). However, it is comparable to the values reported by Clifton et al. (2017) (41% over 24 h, with daytime maxima of 82%) and by Zona et al. (2014) (42% over 24 h, with daytime maxima of 77% in the summer period). Our summer values are also consistent with the stomatal fraction reported by Zhang et al. (2006) for the Kane deciduous forest (USA) in summer (between 50% and 55%), but well below the 65% of the temperate deciduous forest of Borden (USA), as it can be deduced from Fuentes et al. (1992). The stomatal fraction (F_StomO3/F_TotO3) was similar over the different years despite the significant interannual variability of the total O₃ fluxes, a result that was also highlighted by Rannik et al. (2012). Considering that this ratio did not vary too much during the 7 years, we may try to better characterize the non-stomatal processes partially investigated by Finco et al. (2018) and Nemitz et al. (2013) at our site. According to what was reported by those authors, 18% of the O₃ deposition was due to chemical reaction with intra-canopy NO in dry canopy conditions, while no more than 3% was due to reactions of O₃ with VOCs. Therefore, since on average 52% of the O₃ deposition occurs by stomatal uptake in summer, the remaining 27% of the non-stomatal deposition can be attributed, in dry canopy conditions, to the deposition on leaf cuticles, soil, and plant surface. However, this value could rise when air humidity is high, and the surfaces are wet.

Understanding the contribution of each non-stomatal deposition pathways remains an uncertain aspect in the characterization of O₃ removal dynamics in ecosystems (Clifton et al., 2020). However, our estimates represent an attempt to provide the relative weight of the non-stomatal deposition related to dry deposition on surfaces (soil and vegetation) and O₃ chemical reaction with other trace gases (NO_x and VOCs). Moreover, the application of flux correction for the O₃ storage effect was found to be important for the interpretation of the measurements, and in particular to carefully evaluate the exchange processes on the surfaces. Nonetheless, for reactive species such as O₃, this correction is still an approximation because it does not consider the chemical reactions that may occur in the intra-canopy space, which remains unsolved. For a more precise assessment of the influence of chemistry on storage, it is necessary to employ a model that simultaneously solves both the chemistry and the O₃ exchange. However, despite these drawbacks, the storage correction allowed us to separate the process of deposition on the cuticles from the other non-stomatal deposition processes.

If on one hand our measurements confirmed that the contribution of non-stomatal processes to O₃ deposition on forests cannot be considered negligible, on the other hand they showed that the correct estimation of the stomatal deposition is still affected by the flux partition procedure adopted. In their extensive review, Clifton et al. (2020) remarked that the correct quantification of the stomatal uptake with the EC technique is strictly related to the availability of canopy skin temperature and reliable estimates of the canopy transpiration and water evaporation from vegetation and soil. In this regard, the comparison of different flux partition approaches performed in this study has shown that the simple approach based on the subtraction of a percentage of H₂O flux attributed to the soil, as also used by Clifton et al. (2017), can be interesting and easy to apply, provided that reliable measurements of evapotranspiration below canopy are available for a sufficiently long period to estimate the value of that percentage. In our study, this latter approach gave results not dissimilar to those obtained with the CO₂-based approach by Lamaud et al. (2009), and it was in agreement with other methods to estimate forest transpiration based on seasonal LAI evolution, such as those proposed by Wang et al. (2014) and Wei et al. (2017). The consideration of these aspects is crucial to avoid overestimations of the O₃ stomatal fraction and the O₃ effect on vegetation. In any case, increasing the set of measurements to include below canopy measurements, at least of H₂O fluxes, is encouraged.

Although the quantification of the stomatal uptake is important for estimating the O₃ removal from atmosphere, it also provides important information for the assessment of plant response to O₃ stress.

The stomatal uptake, in fact, determined the interannual variability of the POD1 values, since the POD1 was not always related to the mean O₃ mole fraction recorded during the growing seasons (spring and summer; Table S1). The marked interannual variability of POD1 (e.g., 2015 vs. 2018; Table S1) also highlights the need for multiannual measurements in order to have a more reliable O₃ risk assessment for forest ecosystems. The annual average value of 10.55 mmolO₃ m⁻² PLA was lower than the critical level for total biomass established by the UNECE for the protection of Mediterranean deciduous forests (14 mmolO₃ m⁻² PLA for total biomass; CLRTAP, 2017) but slightly higher than the critical level for significant negative effects on roots biomass growth (10.3 mmolO₃ m⁻² PLA; CLRTAP, 2017). If, on the other hand, we consider the critical level established for continental forests (5.7 mmolO₃ m⁻² PLA for total biomass; CLRTAP, 2017), this limit was almost twofold exceeded. Although the effect of O₃ on total biomass growth may seem small (Table 2), these estimations confirm that O₃ is a persistent stress factor for this forest ecosystem. Furthermore, the growth rate reduction of root biomass resulting from lower translocation of photosynthates from the leaves to the roots transposes the negative effects on the vegetative restart of the following year, as also highlighted by Grulke and Heath (2020). We may predict that this abiotic stress affects the overall carbon balance and the NEE of the forest; however, this hypothesis could be confirmed with future data analysis.

A new multiannual dataset of O₃ flux measurements is now available to test and improve the parametrization of O₃ deposition models, and for the estimations of the interannual variability of tropospheric O₃ budget in global models. The long-term monitoring campaign performed on the temperate deciduous broadleaf forest of Bosco Fontana revealed a marked interannual and intra-annual variability of O₃ deposition that seemed to be dominated by the stomatal component. Therefore, the drivers that influence g_StomO3 (mainly soil water availability and air dryness) were responsible for the variability of the total O₃ flux at day-to-day and interannual time scales. The measurements confirmed the relationships used in SVAT models (Soil-Vegetation-Atmosphere Transfer models), such as DO₃SE (CLRTAP, 2017), to describe the dependence of the g_StomO3 on various environmental drivers and provide insights for adapting those parameterizations to deciduous temperate forests. The non-stomatal O₃ deposition was mainly driven by air humidity, surface wetness, and chemical sinks such as reaction of O₃ with NO, the latter being emitted by the forest soil in summer or transported into the trunk space from outside the forest in winter.

The assessment of the potential O₃ effects on plants, performed through the calculation of the phytotoxic stomatal dose (POD1) for the 7 years, suggests that O₃ may reduce the plants growth rate by 3% yr⁻¹, on average, with a serious threat to the carbon balance at ecosystem level. The marked interannual variability of POD1 highlighted the need for multiannual measurements for reliable O₃ risk assessment.

This work has shown the importance of using canopy transpiration for the estimates of g_StomO3 in the O₃ flux partition process, instead of the raw evapotraspiration measurements (LE) performed above the forest canopy. It also showed the benefit of correcting the EC data for storage in order to identify the different deposition phenomena, in particular the deposition on leaf cuticles that occurred in the evening hours.

The following datasets were generated:

• – Dataset-1 containing the monthly data of O₃ fluxes and deposition velocity (storage corrected).

• – Dataset-2 containing the hourly means of O₃ fluxes (storage corrected and not) averaged per each month over all years.

• – Dataset-3 containing the hourly means of deposition velocity, conductances, and main meteorological parameters, averaged over the summer months per each year.

The datasets are uploaded in a single supplemental file.

The supplemental files for this article can be found as follows:

Datasets 1–3. xlsx

Description of Supplementary Data. Docx

Table S1. Docx

The authors acknowledge the European FP7 Project ÉCLAIRE (Effects of Climate Change on Air Pollution Impacts and Response Strategies for European Ecosystems, Grant Number 282910).

The present research was co-funded by the Catholic University of the Sacred Heart in the framework of the D1 and D3.1 calls for proposals (Years 2013–2020).

The authors are grateful to the administration and the personnel of the State National Reserve of Bosco Fontana for their availability and continuous support.

No competing interests were identified.

Conception and design: GG.

Acquisition of data: GG, AF.

Analysis and interpretation of data: GG, RM.

Drafted and/or revised the article: All authors.

Approved the submitted version for publication: All authors.

How to cite this article: Gerosa, GA, Marzuoli, R, Finco, A. 2022. Interannual variability of ozone fluxes in a broadleaf deciduous forest in Italy. Elementa: Science of the Anthropocene 10(1). DOI: https://doi.org/10.1525/elementa.2021.00105

Domain Editor-in-Chief: Detlev Helmig, Boulder AIR LLC, Boulder, CO, USA

Associate Editor: Daniel Liptzin, Soil Health Institute, Morrisville, NC, USA

Knowledge Domain: Atmospheric Science