We used the following four tobacco cultivars: Ambalema, secreting only the diterpenoid cis-abienol (C20H34O, see Fig. ), BYBA secreting α- and β-cembratrien-diols (CBTdiols, C20H34O2, see Fig. ), and Basma Drama, secreting all these compounds . The new 3H02 line does not exude diterpenoids at all (see Appendix ).

Seeds of the tobacco cultivars were obtained from the Leibniz Institute of Plant Biochemistry, Department of Cell and Metabolic Biology, Halle. The plants were grown in the green houses of the Institute of Ecology of the University of Innsbruck for 8–10 weeks in standard soil.

Before being used in the experiments the sample plants were allowed to adapt 1–4 weeks in the laboratory, obtaining light from the same true light lamp type as used during the measurements (see Sect. 2.2).

Plants were installed into the plant enclosure used for ozone fumigation the evening before the actual experiment, so they could adapt to the system and recover from possible stress during installation. The sample plants were well watered and in a good physiological condition and showed no visible signs of damage. At the beginning of the experiments, when no ozone was added, no significant stress signals in form of green leaf volatiles were detected.

In total, combined dark and light ozone fumigation experiments were conducted with five Ambalema, two Basma Drama, one BYBA and three 3H02 samples. Moreover, experiments under solely light conditions were conducted with eight Ambalema, four Basma Drama, four BYBA, and two 3H02 plants. Each sample plant was tested only once.

In the present ozone experiments we used only inert materials such as Teflon^®, PEEK^® or Duran^® glass in order to minimize artificial side-reactions of ozone with unsaturated compounds, present in, e.g. sealing materials like rubber. Moreover, special care was taken to avoid fingerprints, which could result in side reactions of ozone with skin oils . Ozone loss, estimated from measured ozone concentrations at the inlet and outlet of the empty plant enclosure, was typically less than 5 %.

For plant fumigation, synthetic air 5.0 grade was mixed with CO2 4.8 grade (both Messer Austria GmbH, Gumpoldskirchen, Austria). By bubbling the air in distilled water and passing it by a subsequent thermoelectric cooler (TEC) the relative humidity was set. Before entering the plant enclosure, the air was flushed through an ozone generator (UVP, Upland (CA), USA). The enclosure system consisted of a desiccator (Schott Duran^®) of 17.3 L volume, turned upside-down, and two end-matched PTFE^® ground plates. A central hole served as feed-through for the plant stem, possible gaps were sealed with Teflon^® tape. The (single-sided) leaf area enclosed was typically in the range of 250–850 cm2.

An ozone detector (Model 49i, Thermo Fisher Scientific Inc. Franklin (MA), USA) and an infra-red gas analyser (LI-840A CO2/H2O Analyzer, LI-COR^® inc., Lincoln (NE), USA) were sampling at 2 min intervals from either the inlet or outlet of the enclosure. Plant enclosure inlet ozone concentrations were typically kept constant throughout each experiment and were adjusted to obtain realistic ambient ozone concentrations at the enclosure outlet during light conditions (e.g. ∼60 ppbv in Fig. ). Relative humidity in the plant enclosure ranged from typically ∼ 55 % in dark experiments up to ∼95 % in light experiments.

VOC were quantitatively detected at the enclosure outlet by a Selective Reagent Ionization Time-of-Flight Mass Spectrometer (SRI-ToF-MS, see next section) which was switched every 6 min between H3O+ and NO+ reagent ion mode.

Sample plants were illuminated by a true light lamp (Dakar, MT/HQI-T/D, Lanzini Illuminazione, Brescia, Italy). Infra-red light was shielded off by a continuously flushed water bath in order to prevent heating of the plant enclosure. Photosynthetically active radiation (PAR) was measured with a sunshine sensor (model BF3, Delta T Devices Ltd, Cambridge, UK) and temperature on the outer plant enclosure surface with K-type thermocouples.

The UIBK Advanced SRI-ToF-MS (University of Innsbruck Advanced Selective Reagent Ionization Time-of-Flight Mass Spectrometer, ) combines the high mass resolution of PTR-ToF-MS with the capability to separate isomeric compounds having specific functional groups. For this purpose, the SRI-ToF-MS makes use of different chemical ionization pathways of a set of fast switchable primary ions (here: H3O+ and NO+). Moreover, the employment of different primary ions could help to differentiate molecules suffering from fragmentation onto the same mass to charge ratio in the standard H3O+ mode .

Examples of differentiable isomers are aldehydes and ketones. In the H3O+ reagent ion mode, aldehydes and ketones both exhibit proton transfer and thus, e.g. methyl vinyl ketone (MVK) and methacrolein (MACR) are both detected as C4H7O+ (m/z 71.050). In NO+ reagent ion mode, most aldehydes exhibit hydride ion transfer and ketones clustering reactions, comparable to the ionization mechanisms in a SIFT instrument . Thus MVK is detected as C4H6O ⚫ NO+ (m/z 100.040), whereas MACR is detected as C4H5O+ (m/z 69.034).

In addition to isomeric separation, the high flow through the drift tube (here: ∼500 mLmin-1 compared to 10–20 mLmin-1 in a standard instrument) allows for the first time the detection of semi-volatile compounds such as the diterpenoid cis-abienol (C20H34O).

The SRI-ToF-MS was operated under standard conditions, 60 ∘C drift tube temperature, 540 or 350 V drift voltage and 2.3 mbar drift pressure, corresponding to an E/N of 120 or 78 Td (E being the electric field strength and N the gas number density; 1Td=10-17 Vcm2) in H3O+ or NO+ reagent ion mode, respectively. The instrument was calibrated approximately once a week by dynamic dilution of VOC using two different gas standards (Apel Riemer Environmental Inc., Broomfield (CO), USA), containing ca. 30 different VOC of different functionality distributed over the mass range of 30–204 amu. Full SRI-ToF-MS mass spectra were recorded up to m/z 315 with a 1 s time resolution. Raw data analysis was performed using the PTR-ToF Data Analyzer v3.36 and v4.17 .

For the identification of cis-abienol a pure standard was acquired (Toronto Research Chemicals, Toronto, Canada). The powder was dissolved in n-hexane and applied on the surface of a glass container, which was put into the enclosure system and treated like the plant samples. In H3O+ reagent ion mode, the major cis-abienol derived signal was detected on m/z 273.258 (C20H33+); like many other alcohols, cis-abienol is losing H2O after the protonation reaction. Minor fragment signals in the range of a few percent were detected at m/z 191.180 (C14H23+), m/z 163.149 (C12H19+) and m/z 217.196 (C16H25+), respectively.

In NO+ reagent ion mode, the major cis-abienol derived signals were detected at m/z 272.250 (C20H32+) and m/z 178.172 (C13H22+). Minor signals were measured at m/z 163.149 (C12H19+) and m/z 134.101 (C10H14+), respectively.

Ozonolysis of the pure cis-abienol standard yielded the same primary ozonolysis products (see below) as in the case of Ambalema plants.

In order to relate the observed ozonolysis carbonyls to plant surface reactions, leaf exudates of untreated tobacco plants were stripped off by dipping leaves (of similar area) of untreated Ambalema, Basma Drama and 3H02 plants into n-hexane (∼100 mL for 1000 cm2 leaf area) for ∼1 min. The n-hexane – leaf exudate solution was then distributed as evenly as possible onto the inner surface of the empty desiccator serving as plant enclosure. n-hexane evaporated quickly and was further reduced by flushing the glass cuvette with pure synthetic air. Afterwards, ozone fumigation experiments were performed similar to the experiments with intact plants.

Non-volatile ozonolysis products and unreacted surface compounds were analysed by GC-MS (see also Supplement). Directly after the ozone fumigation experiments we extracted leaf exudates and low-volatility ozonolysis products from the fresh tobacco leaves (see Sect. 2.5). 1 µL portions of the samples were then injected directly into a GC-MS for analysis on a 6890 N gas chromatograph coupled to a 5973 N mass spectrometer (Agilent Technologies) according to the procedures described elsewhere .

Tobacco diterpenoids were identified on the basis of their mass spectra, as described in the literature .

For the calculation of the gas exchange parameters we followed well-established procedures by and . Transpiration rate E, assimilation rate A, total ozone flux Ftot,O3 and total water vapour conductance gl,H2O were calculated from E=ues⋅wo-we1-wo×10-3,[mmolm-2s-1]A=ues⋅ce-1-we×10-31-wo×10-3⋅co,[µmolm-2s-1]Ftot,O3=ues⋅oe-1-we×10-31-wo×10-3⋅oo,[nmolm-2s-1]gl,H2O=103⋅E1-wo+wi2×103wi-wo,mmolm-2s-1, with ue the molar flow of air entering the enclosure in [mols-1], s the leaf area in [m2], we/ce/oe and wo/co/oo the mole fraction of water vapour/CO2/ozone entering respectively leaving the plant enclosure in [mmolmol-1], [µmolmol-1] and [nmolmol-1], respectively. wi is the mole fraction of water vapour inside the leaf in [mmolmol-1] and is typically assumed to be the saturation mole fraction at leaf temperature .

For the calculation of the total ozone conductance we applied a ternary diffusion model as has been proposed by . Thereby, pairwise interactions between ozone, water vapour and air are considered (for the sake of simplicity we neglected interactions with CO2). Interactions of ozone molecules with water vapour are important only for that portion of ozone, which is entering the stomatal pores and not for that lost in reactions at the leaf surface. However, in the latter case the consideration of binary diffusion between ozone and water leads to an overestimation of the total ozone conductance in the range of <1%.

Total ozone conductance gl,O3 is then defined by gl,O3=-103⋅Ftot,O3+oa+oi2⋅Eoa-oi,[mmolm-2s-1], with oi and oa the mole fractions of ozone inside the leaf (at the leaf surface for reactive leaf surfaces) and in the surrounding air, respectively. oa equals the ozone mole fraction oo measured at the outlet of the plant enclosure. Typically, we consider oi≈0 and therefore Eq. () simplifies further to gl,O3=-103⋅Ftot,O3+oa2⋅Eoa.

In our fumigation experiments the ozone concentrations in the plant enclosure varied between the different experiments and within experiments switching from light to dark conditions. In order to compare the ozone depletion capability (i.e. surface plus stomatal sinks) of different plants or of the same plant under dark and light conditions, it is therefore important to use a  concentration independent measure. As for a given ozone conductance the ozone flux increases with the ambient ozone concentration (cf. Eqs.  and ), we follow others see e.g. and use the ozone conductance values instead. In experiments with plants having an ozone reactive surface, the total ozone conductance gl,O3 (Eq. ) comprises boundary layer conductance, stomatal conductance, and cuticular conductance. Stomatal and boundary layer ozone conductances can be calculated from those of water vapour by correcting for the different diffusivities of the two gases. The boundary layer water vapour conductance could be determined by measuring temperature and evaporation rate from leaf models made of chromatography paper (see ). However, in our experiments this was not really practical for all sample plants which were all complexly and differently shaped. Consequently, also the stomatal water vapour and ozone conductances could not be inferred from the calculated total water vapour conductance (Eq. ).

As we show in the Supplement, even if stomatal and boundary layer ozone conductances are known, for semi-reactive leaf surfaces the calculation of stomatal and non-stomatal parts of the total ozone flux is not feasible.

For these reasons we report here only total ozone conductance values (Eq. ), normalized to the single-sided leaf area or to the area of the enclosure covered with leaf exudates in experiments with pure leaf surface compounds (see Sect. ).

Data (gl,O3, A, gl,H2O) were tested for statistically significant differences between dark and light experiments (using the same variety) and between different tobacco varieties (in either dark or light experiments), respectively, using the Wilcoxon-Mann-Whitney test in Matlab^®. Due to the partially small sample size, probabilities p < 0.1 are reported as marginally significant. Lacking replicates of dark experiments with BYBA plants, in the statistical analysis this type of experiment was omitted.

In order to visualize the ozone concentration gradients caused by plant ozone uptake, two idealized setups were simulated: a macroscopic plant model in an ambient air flow and a microscopic model for the stomatal gas exchange. The simulations were done using the open-source CFD code OpenFOAM (www.openfoam.com).

In the microscopic model the air flow was neglected and a pure diffusion process was simulated. Stomata were modelled as 100 µm long and 40 µm wide eye-shaped openings recessed 20 µm deep into the leaf surface. The simulation domain with 500 000 cells covered an area of 300 µm square around the stoma and extended 2 mm from the leaf surface into the surrounding gas. A single stoma with cyclic boundaries was used to represent a whole leaf with stomata spread repeatedly over its surface. The ozone-reactive bottom of the stomata was modelled as 100 % efficient sink with a constant ozone concentration of zero, while the side walls of the stomata were assumed not to absorb ozone and set to zero gradient. The top of the measurement domain acting as ozone inlet from the surrounding was set to one. The leaf surface around the stomata was set to zero gradient or to a fixed concentration of zero, representing two idealized plant types with either non-reactive or reactive leaf surface. “scalarTransportFoam” was run on this grid with a uniform zero velocity field until a steady state was reached.

For the macroscopic model (see Supplement) a laminar flow around the plant was simulated using the steady-state Reynolds averaged Navier–Stokes solver “simpleFoam”, the transport of ozone in the resulting flow velocity field was studied using the “scalarTransportFoam” solver. The simulated gas volume consisted of a cube with 20 cm edge length with the shape of an exemplary tobacco plant cut out of its volume (see Fig. S3). The resulting simulation domain was divided into a hexahedron-dominant grid of 3.7 million cells with the finest granularity around the stomata and the leaf surfaces with the OpenFOAM tool “snappyHexMesh”. The domain was divided into eight subdomains for parallel computation. Stomata were represented by small patches spread equally over the leaf surfaces, covering 10 % of the total leaf area. The boundary conditions for the gas flow simulation consisted of an inlet with 2 mms-1 velocity entering on one face of the cube and a constant pressure boundary condition outlet on the opposite face. The gas velocity on the plant surface was set to zero. Initial conditions for the flow simulation were calculated with “potentialFoam” to speed up convergence of the “simpleFoam” solver. The simulation was run until the flow velocity field reached a steady state. For the diffusion calculations a relative initial concentration of ozone was set to one at the inlet and to zero on the stomata patches. Like in the microscopic model calculations, the leaf surface was either a zero concentration gradient boundary (for an idealized 3H02 plant type) or a fixed concentration value of zero (for an idealized Ambalema plant type). In the previously calculated velocity field the ozone transport was simulated until a steady state was reached, too.

Apart from the 3H02 variety, the investigated tobacco varieties secrete different unsaturated diterpenoids (see Sect. ). According to the Criegee mechanism , ozone attacks the carbon double bonds of alkenes forming primary carboyls and so-called Criegee Intermediates (see Supplement). Criegee Intermediates are, however, expected to be too short-lived to be detected directly by the instruments used in our experiments (see Supplement). We were therefore interested primarily in the stable, volatile ozonolysis carbonyls, which could be detected in real-time by our SRI-ToF-MS.

For the semi-volatile diterpenoid cis-abienol with two exocyclic double bonds, exuded by the Ambalema and Basma Drama varieties, we expected the formation of formaldehyde (HCHO) and methyl vinyl ketone (MVK, C4H6O, see Fig. ).

In the case of the ring structured CBTdiols with three endocyclic double bonds, produced by the Basma Drama and BYBA plants, at least two ozonolysis steps are needed to form volatile carbonyls. The three smallest carbonyl products are shown in Fig. , whereby 4-oxopentanal (C5H8O2) is expected to be the most volatile one .

In order to relate a release of carbonyls to surface chemistry only and to exclude stimulated emissions caused, e.g. by the plant ozone defence system, we investigated ozone reactions with pure leaf surface extracts. Leaf surface compounds were extracted with n-hexane and subsequently applied onto the inner surface of an empty plant enclosure and fumigated with ozone (see Sect. 2.5).

Ambalema leaf extracts showed a weak signal of cis- abienol (we refer to Sect. 2.4 for the identification of this compound), which disappeared during ozone fumigation while MVK and formaldehyde were prominently observed. These carbonyls were produced by surface-assisted ozonolysis of cis-abienol (see Fig. ). MVK was detected at m/z 71.050 (C4H7O+) and m/z 100.040 (C4H6O ⚫ NO+) in the H3O+ and NO+ reagent ion mode of the SRI-ToF-MS, respectively. Formaldehyde was detected only using H3O+ as reagent ion at m/z 31.018 (CH3O+), taking into account the humidity dependent sensitivity . In the NO+ reagent ion mode formaldehyde cannot be ionized , consequently we detected no signal.

In the ozone fumigation experiments using Basma Drama leaf extracts, besides MVK and formaldehyde as ozonolysis products of cis-abienol, also the most volatile CBTdiol ozonolysis product – 4-oxopentanal – was detected in the gas phase by SRI-ToF-MS. 4-oxopentanal was detected at m/z 101.060 (C5H9O2+) in H3O+ and m/z 99.045 (C5H7O2+) in NO+ reagent ion mode, respectively.

No significant amount of volatile carbonyls was observed from ozonolysis of 3H02 leaf extracts. Consistently, the total ozone conductance was far less than in experiments with extracts from diterpenoid-exuding tobacco varieties (see Fig. ). This is in line with the results from the corresponding experiments with intact plants (see below). The ozone depletion efficiency of the 3H02 exudates was decreasing fast, while the presence of cis-abienol in Ambalema leaf exudates kept the ozone conductance at elevated levels for many hours (cf. Fig. ).

Also in experiments with intact plants we observed a prompt release of volatile carbonyls as soon as the tobacco leaves were fumigated with ozone. The Ambalema and Basma Drama varieties released MVK and formaldehyde. In addition, we detected sclaral, a non-volatile compound, in surface extracts obtained from ozone fumigated plants of the same varieties (see Sect. 2 and Supplement). Sclaral is an isomerization product of the C16 carbonyl formed in cis-abienol ozonolysis (cf. Fig. ). All these compounds can therefore be attributed again to surface-assisted ozonolysis of cis-abienol (see Fig. ).

In experiments using Basma Drama and BYBA plants we detected the CBTdiol ozonolysis product 4-oxopentanal, similar to the ozone fumigation experiments with leaf surface extracts (see previous section).

Figure shows a typical result of an ozonolysis experiment using Ambalema plants. Immediately after starting the ozone fumigation, the cis-abienol signal decreased, while initial bursts of MVK and formaldehyde were detected. These initial bursts can be attributed to surface ozonolysis of cis-abienol deposited on all surfaces (i.e. surfaces of the whole plant, the enclosure and the enclosure outlet tubing) during plant acclimatization under ozone free conditions lasting > 12 h (see Sect.  and Supplement).

In plant experiments using diterpenoid exuding tobacco varieties, the carbonyl emission and consequently the total ozone conductance and flux (under constant light) eventually reached a steady state, when the diterpenoid production by the trichomes (leading to a permanent deposition of those onto the plant surface) and plant surface reactions were in equilibrium (cf. Fig. ). This is in contrast to experiments with pure leaf surface compounds, in which the diterpenoids were slowly consumed as ozone fumigation progressed (see Sect. ).

Simulating diurnal ozone variations over 2 days in experiments with Ambalema and Basma Drama plants, we could show that the reactive layer at the plant surface is a large pool and not quickly consumed (see Supplement and Fig. S2). We therefore assume that the diterpenoids released are likely to represent a long-term ozone protection for these varieties.

In further experiments we investigated the ozone depletion by different tobacco varieties under dark and light conditions.

In dark experiments, when stomatal pores are almost closed, the Ambalema variety showed the highest total ozone conductance under steady-state conditions (cf. Fig. , top panel). This is a direct indication for the high ozone depletion capacity of the surface of this variety.

Due to the lack of reactive diterpenoids on the leaf surface of 3H02 plants, the surface ozone sink plays a minor role for this tobacco line. However, we cannot totally exclude the presence of other unsaturated compounds at the surface of this variety.

The low surface reactivity of the Basma Drama and BYBA varieties correlates with the lower amount of detected ozonolysis carbonyls compared to that of the Ambalema variety in dark conditions. This might be related to a lower diterpenoid surface coverage of these two varieties and the expected lower reactivity of the CBTdiols having endocyclic double bonds .

The Ambalema variety also shows a higher gl,H2O and dark respiration than the other varieties (cf. Fig. , middle and bottom panels). gl,H2O linearly correlates with the stomatal water vapour conductance and therefore also with the stomatal ozone conductance. However, higher stomatal conductance during dark conditions cannot explain the large differences in gl,O3 between the plant types. While gl,H2O of the Ambalema variety in dark conditions is about twice as high as that of the 3H02 variety, the corresponding gl,O3 is four times as high.

When switching from dark to light conditions we assume cuticular conductance not to change significantly and thus an increase in the calculated gl,O3 is attributable mainly to an increasing stomatal ozone conductance. In the case of Ambalema, switching the light on increased the total conductance by ∼55 % (see Fig. , top panel). In contrast, in the 3H02 case, switching on the light triggered a substantial increase in the total ozone conductance by ∼340 % (cf. Fig. , top panel).

During light conditions the total ozone conductances of the different tobacco varieties were in a comparable range; slightly higher values were observed for the diterpenoid exuding lines Ambalema, Basma Drama, and BYBA.

Statistical analysis confirmed the observed tendencies of the total ozone conductance: only for the Ambalema variety was gl,O3 under light conditions not significantly different from the values measured under dark conditions (p > 0.1). Conversely, gl,O3 calculated for the Ambalema variety was significantly higher than that of the other tobacco lines under dark conditions (p < 0.1, see Fig. , top panel).

In the ozone fumigation experiments the yield of volatile ozonolysis products was generally in the low percentage range, e.g. for Ambalema plants ∼7 % under dark and ∼5 % under light conditions considering the major volatile ozonolysis products MVK and formaldehyde quantified by SRI-TOF-MS. The slight change from ∼7 to ∼5 % when switching from dark to light conditions can be explained by the effect of the open stomata. Open stomata offer an alternative sink for ozone and for volatile carbonyls produced in surface assisted reactions . The reason why only a small percentage of the consumed ozone is detected as volatile products indicates that most of the ozonolysis products are not volatile enough to leave the plant surface (cf. Fig.  and Supplement). The fate of the Criegee Intermediates in surface ozonolysis is discussed in detail in the Supplement.

In order to qualify the measured total ozone fluxes for the calculation of gl,O3 values, we had to take into account the possibility of homogeneous gas phase ozonolysis of the semi-volatile diterpenoids exuded by the tobacco varieties.

To assess the significance of gas phase ozonolysis to our results, we connected the plant enclosure containing a diterpenoid emitting tobacco plant with a second empty enclosure downstream and added ozone only to the second enclosure. Only negligible carbonyl signals were observed once the initial burst from deposited diterpenoids faded away (see Supplement and Fig. S1). This result indicates that with our setup gas-phase reactions of the diterpenoids were not significant.

This observation can be explained theoretically, too. The air in our enclosure system was exchanged every ∼5 min. Therefore, only extremely fast gas phase ozone–alkene reactions have to be considered. For an ozone concentration of 100 ppbv, a reaction rate of 1.35×10-15 cm3s-1 results in an alkene ozonolysis lifetime of 5 min. Such fast ozonolysis rates have only been measured for a few very reactive sesquiterpenes . We found no reaction rates of cis-abienol and CBTdiols with ozone in the literature to exclude the possibility of a gas phase contribution to total ozone loss in our experiments a priori. Nonetheless, taking into account the estimated vapour pressures of cis-abienol (∼10-9 bar) and CBTdiol (∼10-12 bar) we can state that the bulk of the exuded diterpenoids stayed at the leaf surface and that other surfaces (e.g. the inner surface of the plant enclosure and the tubing system) were very slowly covered by condensed diterpenoids. This is also the explanation for the bursts of volatile ozonolysis products at the beginning of every ozone fumigation (see e.g. Fig. ). We therefore assume that gas phase reactions are unlikely to have played a major role in our experiments.

Microscopic fluid dynamic model calculations (see Materials and methods) revealed the principles responsible for the strong variety-dependent partitioning between stomatal and non-stomatal ozone loss (see Sect. ). The mixed convective and diffusive ozone transport from the surrounding atmosphere to the plant surface and into the stomata was simulated for two idealized plant types under light conditions when the leaf stomata are open. The stomatal pores were exemplarily modelled as small patches uniformly spread over the entire leaf surface. For one model plant we assumed stomatal ozone uptake only, corresponding to an idealized 3H02 variety plant lacking any reactive surface compounds. The second model plant was representing an idealized Ambalema variety. The surface acted as a perfect ozone sink with every ozone molecule reaching it being lost, either on the leaf surface or through the stomata.

Figure a and b show the resistance schemes used to describe the ozone flux to the leaves in the two scenarios, which were the basis for our simulations. Ambient ozone has to overcome the boundary layer resistance Rb and the stomatal resistance Rs before being destroyed in the stomatal cavity (for the sake of simplicity we neglected here the mesophyll resistance, which comprises diffusion through inner air spaces and dissolution of the gas in the cell wall water, followed by losses in the aqueous phase, penetration of plasmalemma or chemical reactions in the cell, cf. ). In the case of a non-reactive leaf surface, ozone depletion within the stomata is the sole ozone sink (see Fig. a).

In the case of an ozone-reactive leaf surface, an additional surface chemical resistance Rsc has to be introduced, which is parallel to the stomatal resistance (see Fig. b). Rsc inversely correlates with the reactive uptake coefficient of ozone at the leaf surface. In the case of the model plant having a non-reactive surface, Rsc is very large (Rsc→∞) and ozone flux to the leaf surface can be omitted. Conversely, Rsc is small for reactive surfaces.

The porous leaf surface architecture has special relevance for the gas uptake of plants. For gases having a negligible leaf surface sink (or source) – e.g. CO2 – steep concentration gradients parallel and perpendicular to the surface develop in close proximity to the stomata. These gradients enhance the gas transport in the diffusive leaf boundary layer towards the pores. This effect is extensively described in the literature as the “paradox of pores” (see, e.g. ). It enables plants to effectively harvest CO2 for photosynthesis, but in the same manner also “funnels” phytotoxic ozone through the stomata into the plant leaves (see Fig. c).

In the case of an ozone-reactive leaf surface, Rsc is small compared to Rs and only surface-parallel ozone concentration isosurfaces develop (black lines in Fig. d). Concentration gradients (white lines) close to the stomata are exclusively perpendicular to the surface. Consequently, the ozone transport in the diffusive leaf boundary layer is equally distributed over the whole leaf surface and the ozone concentration in this layer is strongly reduced (see Fig. d). Similarly, also macroscopic model calculations show that this effect broadens the space of reduced ozone concentrations surrounding a plant with opened stomata (see Supplement and Fig. S3).

The surface-parallel concentration isosurfaces are the reason why we can use the same reference concentration cb, r for both the stomatal and the surface chemical resistance, (cf. Fig. b). However, this approach does only hold if the leaf surface is a complete ozone sink (see Supplement and Fig. S5).

The different ozone concentration patterns in the two modelled scenarios have important implications for the stomatal ozone uptake. Typically, the stomatal conductance of ozone gs,O3 is estimated from that of water gs,H2O, by correcting for the different diffusivity of the two gases (see e.g. ). The stomatal ozone flux Fs,O3 can then be calculated with the following formula: Fs,O3=gs,O3⋅(ci,O3-cb,O3), with ci,O3 being the ozone concentration in the leaf intercellular space and cb,O3 the ozone concentration in the leaf boundary layer. For high ambient ozone concentrations ci,O3 was found to be positive , but typically it is assumed to be close to zero . Therefore, Eq. () simplifies to Fs,O3=-gs,O3⋅cb,O3. If now surface reactions drastically reduce cb,O3 (cf. Fig. b and d), the effective stomatal ozone flux (see Supplement) and with that the effective ozone dose are also reduced, which eventually determine the phytotoxic effects of ozone to plants . At this point, it is important to note that the uptake of non surface-reactive gases such as CO2 is not affected by the altered ozone gradients.

Thus, whenever surface loss plays a role, both surface and stomatal ozone uptake by plants have to be considered together. Previous studies might therefore have overestimated stomatal ozone uptake (e.g. ). Hence, their reported stomatal ozone flux values should be considered as upper limits.

In future studies investigating the ozone depositions to vegetation, it might be worth to analyse also the surface composition of the plants. If the surfaces are covered with substantial amounts of unsaturated organic compounds, surface loss has to be considered right from the beginning in order not to overestimate stomatal ozone uptake. Due to the fact that surface reactions reduce ozone concentrations in the leaf boundary layer, it is not correct to calculate stomatal ozone loss applying the resistance scheme shown in Fig. a and to eventually define the surface loss of ozone as that portion of the total loss which is not explainable by gas phase reactions and stomatal uptake.

For real plants the altered ozone gradient profile shown in Fig. d is less pronounced depending on stomata depth, which reduces the total stomatal uptake, and reactive surface compounds, which show smaller surface reaction rates than assumed for the idealized 100 % efficient ozone depleting surface (see Supplement). In the case of such semi-reactive leaf surfaces a more sophisticated resistance scheme has to be used, which strongly complicates the calculation of stomatal and non-stomatal ozone fluxes (see Supplement and Fig. S5). Nonetheless, the simulations explain the experimentally observed behaviour of different tobacco plants very well.

Over the last decade, several studies have shown discrepancies between measured and expected ozone deposition fluxes. Large downward ozone fluxes and high levels of oxidized VOC have been taken as evidence for “unconventional in-canopy chemistry” in a Ponderosa pine plantation, the Blodgett forest site. Measured ozone deposition fluxes could not be explained by modelled stomatal and known non-stomatal sinks, such as reactions with measured VOC in the gas phase . The same observation was made by in a Scots pine dominated field site in Hyytiälä. All these studies assume the presence of yet unmeasured highly reactive semi- or low-volatile compounds, which have a similar temperature-dependent emission pattern as mono- and sesquiterpenes.

assumed that the unmeasured reactive compounds might be unsaturated, cyclic terpenoids. Due to their low vapour pressure, the measurement of semi- or low-volatile compounds represents a challenge, since these substances strongly partition into the condensed phase and are therefore easily lost in the inlet systems of most current VOC instrumentation. However, were able to identify several different sesquiterpenes in ambient air and in branch enclosure experiments at the Blodgett forest site.

A large number of compounds with diterpenoid backbones were recently observed for the first time in a different Ponderosa pine forest site during the BEACHON-RoMBAS campaign 2011 . These unsaturated diterpenoids contain the same backbone as abietic acid, a primary component of resin acids. The observed temporal variations in concentrations were similar to those of sesquiterpernoids, suggesting they are directly emitted from the local vegetation.

Most recently, have shown that semi-and intermediate volatility organic compounds measured for the first time at the same site with a novel thermal desorption electron impact mass spectrometer (TD-EIMS) could likely close the gap between observed and expected secondary aerosol growth, estimated from gas-phase concentrations of the most abundant measured VOC (mono- and sesquiterpenes, toluene/p-cymene, isoprene). We therefore speculate that the high ozone deposition fluxes in such forest sites could be a result of not only gas-phase reactions, but to a certain extent also of ozone reactions with semi-volatiles emitted or redeposited onto the vegetation surfaces.

Possible sources of the measured and unmeasured higher terpenoids are – among others – constitutive plant emissions or resins, which are known to contain high amounts of sesqui-, di- and triterpenoids . Resins can be released during mechanical stress, e.g. in the event of hail storms and could eventually evaporate depending on their vapour pressure (and therefore ambient temperature).

Di- and triterpenoids are also known constituents of surface waxes . Moreover, it is estimated that about 30 % of vascular plants have glandular trichomes, which often exude higher terpenoid compounds, too .

Clearly, additional experiments are needed to better quantify the amount of semi-volatiles deposited onto vegetation surfaces and their impact on atmospheric chemistry.