A top-down approach of surface carbonyl sulfide exchange by a Mediterranean oak forest ecosystem in Southern France

The role that soil, foliage and atmospheric dynami cs have on surface carbonyl sulfide (OCS) exchange i a 15 Mediterranean forest ecosystem in Southern France ( the Oak Observatory at the Observatoire de Haute Pr ovence, O3HP), was investigated in June of 2012 and 2013 with esse ntially a top-down approach. Atmospheric data demon strate that the requirements are fulfilled as that OCS uptake can b e used as a proxy of gross primary production. Firs tly, OCS and carbon dioxide (CO2) diurnal variations and vertical gradients show no net exchange of OCS during the night when the carb on fluxes are dominated by ecosystem respiration. This contrasts with other oak woodland ecosystems of a Mediterranean 20 climate, where nocturnal uptake of OCS by soil and/ or vegetation has been observed. Since temperature, the water and organic carbon content of soil at the O3HP should f avor the uptake of OCS, the lack of nocturnal net u ptake would indicate that its gross consumption in soil is compensated b y emission processes that remain to be characterize d. Secondly, the uptake of OCS during the photosynthetic period was charact e ized in two different ways. We measured ozone (O 3) deposition velocities and estimated the partitioning of O 3 deposition between stomatal and non-stomatal pathw ays before the start of a 25 joint survey of OCS and O 3 surface concentrations. We observed an increasing trend in the relative importance of the stomatal pathway during the morning hours and synch ronous steep drops of OCS (60-100 ppt) and O 3 (15-30 ppb) after sunrise and before the break-up of the nocturnal bo undary layer. The uptake of OCS by plants was chara terized from vertical profiles too. However, the time window for calculation of the ecosystem relative uptake (ERU) of OCS, which is a useful tool to partition measured net ecosystem exc hange, was limited in June 2012 to few hours after midday. This is due to 30 the disruption of the vertical distribution of OCS by entrainment of OCS rich tropospheric air in the morning, and as the vertical gradient of CO2 reverses when it is still light. Moreover, pollute d air masses (up to 700 ppt of OCS) produced dramatic variation in atmospheric OCS-to-CO 2 ratios during daytime in June 2013, further reduci ng the time window for ERU calculation. Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c © Author(s) 2016. CC-BY 3.0 License.


Introduction
Terrestrial ecosystems modulate the water balance over land and fix carbon dioxide (CO 2 ) from the atmosphere in the form of carbon rich materials. Experimental and modeling studies have shown that changes in atmospheric CO 2 concentration and changes in climate, induced by increasing anthropogenic emissions of greenhouse gases, impact on the fixation of 5 atmospheric CO 2 by plants (gross primary production, GPP), and on the release of CO 2 by terrestrial ecosystems (respiration, Reco) as modulated by temperature and water availability, and effects by fertilization (e.g. Arora and Boer, 2014). Large uncertainties in the determination in GPP and Reco fluxes and in the magnitude of effects induced by climate and fertilization remain. Further experimental and modeling studies should help to better constrain those fluxes. In this context, atmospheric carbonyl sulfide (OCS) displays a high potential as a tracer of GPP at various temporal and spatial scales since 10 Chromatography Data System. Calibration was performed as in Belviso et al. (2013) but the primary standard, drawn with a gas-tight syringe, was injected in a line flushed with OCS-free helium (He was passed through an empty stainless-steel trap immersed in liquid nitrogen) connected to the preconcentror inlet. Although the calibration gas (1.013 ppm of OCS in helium) commercialized by Air Products has a tolerance of 2.5%, we found an agreement better than 0.2% between this standard and a calibration gas provided by U. Seibt and K. Maseyk who purchased it from Air Liquide. Since the PFPD 5 response is quadratic, the calibration equation is obtained by plotting the natural logarithm of the peak area against the natural logarithm of OCS (picolitre or pL). Mixing ratios are calculated by dividing pL of OCS by volumes of air dried at -25°C, corrected to room temperature and pressure. Semi-continuous measurement repeatability is 1% (1 SD, n= 38 consecutive hourly analyses of atmospheric air from a compressed cylinder (target gas) containing 573 ppt of OCS).
Accuracy and long-term repeatability (LTR) were evaluated from periodic analyses of an atmospheric air standard prepared 10 and certified by NOAA-ESRL containing 448.6 ppt of OCS and were better than 2.5%.
In June 2013, air was analyzed continuously for OCS using a commercially available OCS, CO 2 , H 2 O, and CO off-axis integrated cavity output spectroscopy analyzer (Los Gatos Research, Enhanced Performance Model, California, USA). In early 2013 at the O3HP, the instrument was tested for the first time in the field. We calibrated the instrument with OCS measured by the GC (over a range of atmospheric concentrations of 439 ppt to 699 ppt). OCS data collected with a ½ Hz 15 frequency by the spectroscopy analyzer were subsequently reduced to 5-minute averages which correspond to the sampling time of the GC. The OCS signal varied by less than ± 2 ppt (standard error) in the 5 minute time window. GC and LGR data showed a linear and strong positive correlation (OCS GC = 1.14 OCS LGR + 12.3 ppt, R 2 =0.95, n=110). Absolute readings were regularly cross-checked with a NOAA-ESRL standard showing good stability throughout the campaign. OCS LGR data were essentially used to document OCS variations in between GC measurements. We thus scaled the OCS LGR to the OCS GC 20 data fitting dynamically by linear interpolation.

Carbon dioxide (CO 2 )
At the O3HP site, in June 2012, air was analyzed for CO 2 from two sampling lines (10 and 2 m height), alternatively (measurement interval duration was 30 min and data collected during the first 10 min were discarded), using a commercially available PICARRO cavity ring-down spectroscopy (CRDS) analyzer (Model G2401) placed next to the OCS gas 25 chromatograph. In addition to CO 2 , this instrument analyzes CH 4 and CO mixing ratios and applies corrections for water vapor levels. Precision and stability of the measurements performed with this instrument were investigated using the rigorous testing procedures described by Yver et al. (2015) and reported in Table 1 of this manuscript (see instrument G2401 with serial number CFKADS2022 and ICOS ID 108). For CO 2 , similar or better results in terms of continuous measurement repeatability (CMR) and LTR were obtained in the field as compared to the factory or to the test laboratory (i.e., 0.027 ppm 30 and 0.020 ppm), respectively (Yver et al., 2015). The CRDS analyzer was calibrated in the test laboratory following ICOS standard procedures, once before shipping and right after the one month deployment in the field.
In June 2013, air was analyzed continuously for CO 2 using the LGR Enhanced Performance instrument (see above). CO 2 measurements were not reported on a calibration scale.

Carbon monoxide (CO)
At the O3HP site, in June 2012, air was analyzed for CO using the PICARRO CRDS analyzer described above. Precision in terms of CMR and LTR measured in the field was not as good as in the factory or in the test laboratory (i.e., 6.8 ppb and 2.2 5 ppb), respectively (Yver et al., 2015). Data were calibrated as for CO 2 measurements. In June 2013, air was analyzed continuously for CO using the LGR instrument. CO measurements were not reported on a calibration scale.

Ozone (O 3 ), O 3 deposition velocity (V d O 3 ) and its partitioning
Ozone was measured at O3HP in June 2012 with an instrument based on ultra-violet absorption (model T-400 from API-10 Teledyne, San Diego, USA). This instrument, calibrated with an internal ozone generator (IZS, API) is operated with a flowrate of about 700 mL min -1 and delivers data every minute. In June 2013, ozone concentrations measured at a few hundred meters from the main O3HP site were downloaded from the regional Air quality network Air-Paca, France,

Stomatal conductance
Canopy stomatal conductance for water vapour (g s H 2 O) was estimated in 2012 from the latent (LE) and sensible (H) heat flux from the Penman Monteith method for relative humidity larger than 70%. Under wet conditions the stomatal 25 conductance was estimate following Lamaud et al. (2009) based on the proportionality between the assimilation of CO 2 and the conductance.
Leaf stomatal conductance was measured in June 2013 with a porometer (AP4, Delta-T Devices, Burwell UK). Before the measurement, it was calibrated in ambient conditions with the provided calibration plate, and recalibrated either three hours after the previous calibration or when average leaf temperatures were 3°C greater or lower than at the time of the most recent 30 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c Author(s) 2016. CC-BY 3.0 License.
calibration. Due to the unilateral distribution of stomata (hypostomatous leaf) only the abaxial sides of the leaf were measured using the 'slotted' configuration of the chamber. Five leaves were sampled per tree and cycle. Light was measured holding the sensor horizontally above the leaf.

Meteorological conditions and soil climate 5
The two campaigns took place in June of 2012 and 2013. The cumulated precipitations before the campaigns were about 400 mm and 500 mm since the beginning of the year, respectively (Fig. 1a). As few precipitation events of small intensity took place during the campaigns, the volumetric soil water content (measured at 5 cm depth) was in a decreasing phase from about 0.3 m 3 m -3 during the wet season to about 0.1 m 3 m -3 during the dry season (Fig. 1b). Soil temperatures went the opposite way ( Fig. 1b) and were in the range 14-19°C and 14-17°C during the 2012 and 2013 campaigns, respectively ( Fig.  10 1c,d).

Diel variations in the canopy (2m)
CO 2 presented a clear and reproducible diurnal cycle with a maximum during the night (Fig. 2c). This maximum, an increase of 10-20 ppm, is correlated with the decrease of global radiation (Fig. 2a). This increase occurred between the period of maximum atmospheric turbulence (u * > 0.4 m s -1 , Fig. 2b), a few hours after the maximum solar radiation (Fig. 2a), and the 15 nocturnal period when atmospheric turbulence is reduced (u * < 0.2 m s -1 , Fig. 2b) and strong temperature gradients above ground level form (~ -0.5 °C m -1 , Fig. 2a). The temperature gradient is a proxy of low atmospheric mixing and boundary layer stability. During this period, the range in OCS was relatively low as compared to CO 2 (10 ppt at the most). The  Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c Author(s) 2016. CC-BY 3.0 License. amplitudes in the range of 150-250 ppt (Fig. 3b), were confirmed by independent measurements carried out with the LGR CO 2 /OCS/CO/H 2 O analyzer which was running in parallel (Fig. 3b). The concomitant decrease of OCS and O 3 in the early morning hours was confirmed in the 2013 records (Fig. 3b). Further, the richest air masses in O 3 , which were transported over O3HP by strong winds in the late afternoon, were not the richest in OCS throughout the campaign (Fig. 3b).

Vertical gradients 5
Diel variations in near-surface OCS and CO 2 vertical gradients were documented twice in June 2012 from data collected alternatingly at 2 m and 10 m (Fig. 5). Both time series show no apparent OCS gradient during the night whereas CO 2 data showed strong vertical gradients with CO 2 at 2 m being higher by approximately 5 ppm than at 10 m. During the day, the CO 2 gradient reversed, CO 2 mixing ratios being lower at 2 m than at 10 m, with a back-reversal of the CO 2 gradient occurring in the late afternoon at 17:00-18:00 UTC. During the day, OCS mixing ratios were systematically lower at 2 m 10 than at 10 m by a few ppt in the morning and up to 10-20 ppt in the afternoon. Hence, CO 2 and OCS were consistently lower at 2 m than at 10 m during the day, during the night however, CO 2 had a gradient in line with the respiratory production of CO 2 , whereas OCS showed no measurable gradient.

Diel variations of fluxes and deposition velocities
The latent heat and CO 2 fluxes (GPP and NEE) followed a clear diurnal cycle well correlated with global radiation, 15 indicating that there was no significant water stress which would tend to lower the flux in the afternoon (Fig. 6a,b).
However, the latent heat flux was significantly higher on June 13 than for later days (Fig. 6a). Higher water fluxes were also measured June 11 and 12 which were likely due to the evaporation of precipitations of low intensity (2 mm at the most) that occurred June 10, 11 and 12 as well as the water that was deposited as dew the nights of June 11 and 12 which was clearly shown by the air temperature reaching the dew point temperature and the sensible heat flux being highly negative at night 20 (data not shown). Significant positive isoprene fluxes were only observed during daytime, following diel cycles with midday maxima ranging from 10 to 35 nmol m −2 h −1 (Fig. 6c redrawn from Kalogridis et al., 2014).
Unfortunately, the fast-O 3 sensor that was used to assess the O 3 deposition velocity had some sporadic down times which occurred frequently during the June 12 to 18 sampling period. During that period, the analyser only performed well during one night. Good quality data, however, were recorded continuously from May 29 to June 3 and, and from June 7 to 9 ( . The shape of these diel cycles provides another indication that the canopy was never under water stress and the gsO 3 mostly light-driven. The ozone deposition velocity (V d O 3 ) exhibited diurnal variations with, in general, larger deposition before mid-day (Fig. 7a). Since the 30 stomatal conductance showed a much more symmetrical feature during daytime (Fig. 7b), it indicates that non-stomatal Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c Author(s) 2016. CC-BY 3.0 License. ozone deposition occurred preferentially during the morning. However, estimates of gnsO 3 were less numerous in the afternoon than in the morning because of inconsistencies between gcO 3 and gsO 3 values noticed during the afternoons of May 29-31 and June 9, where gsO 3 was higher than gcO 3 (Fig. 7b). Nevertheless, in five cases out of six, a peak in gnsO 3 was observed during the period between May 29 and June 3. Data show a shift in the relative importance of both pathways since from June 7 the ozone deposition in the morning in all cases was predominantly through the stomatal pathway. 5 Unfortunately, we have no indication about ozone deposition pathways during the periods where OCS was monitored in the atmosphere. However, the shift towards higher O 3 deposition through the stomatal pathway during the second week of June ( Fig. 7b) and the strong similarities between OCS and O 3 diurnal patterns in June 2012 (Fig. 3a), suggest that the nonstomatal pathway lost importance throughout the month of June.

Role of atmospheric dynamics on OCS exchange
OCS diel variations presented here (Fig. 3) resemble those reported by Berkelhammer et al. (2014) at two sites of central North America where steep rises in OCS also occurred after sunrise (see their Fig. 7b and supplementary Fig. 11). The authors suggested that this morning rise was related to boundary layer dynamics when air from above, richer in OCS than the air from the nocturnal boundary layer, is entrained downwards. This is likely also the case at O3HP. However, diurnal 15 variations with amplitudes over 200 ppt as observed at the O3HP in June 2013 were never reported before. This raises the question of the origin of air masses so rich in OCS advected over O3HP in mid-June 2013. It is highly unlikely that longrange transport of biomass burning gases and aerosols between North America and the Mediterranean region was responsible for OCS contamination because the transport of biomass burning material occurred in late June 2013 so after the end of our OCS surveys (see Fig.4 in Ancellet et al, 2016). As the O 3 rich air masses reaching the O3HP in the late afternoon are 20 lagging those rich in OCS by ~ 4 hours (Fig. 3b), it is clear that the OCS and O 3 peaks have distinct origins. Backward trajectories at 300 m above ground level ending at 12 UTC (Stein et al., 2015), when OCS levels at the O3HP in June 2013 were over 600 ppt (Fig. 3b), show that the circulation of the air masses during both periods was at low altitude (below about 500 m a.g.l., i.e. below 1100 m a.s.l.), thus generally in the boundary layer. The back trajectories show that the air masses were in closer contact with the continent in June 2013 than in June 2012, and that the transport in June 2013 was from the 25 N/NW so along the Rhône Valley (Fig. S1). South of the city of Lyon, the Rhône Valley is highly industrialized and it is therefore likely that the O3HP site is impacted by anthropogenic direct or indirect (i.e. from the oxidation of CS 2 ) emissions of OCS. In the afternoon, polluted air from the metropolitan area of Marseille is transported by the sea breeze thus leading to an increase of ozone at elevated layers above the convective boundary layer. The highest ozone concentrations above 100 ppb can be found about 50 km further downwind north and northeast of Marseille both on the mountainous areas of Luberon 30 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c Author(s) 2016. CC-BY 3.0 License. and above (Kalthoff et al., 2005; see Fig. 6 of that manuscript). We can therefore conclude that the photosmog of the city of Marseille is not a source of OCS.

Ecosystem relative uptake (ERU)
At the O3HP, OCS concentration gradients showing lower concentrations at 2 m than at 10 m were observed during daytime ( Fig. 5), especially during the afternoon so when turbulent mixing was strongest (Fig. 1b). Gradients were inexistent during 5 the night. This implies that the forest ecosystem was essentially a net sink of OCS. Measured CO 2 vertical gradients indicate that the forest ecosystem was a net sink of CO 2 during daytime and a net source during the night, features that were confirmed by the eddy covariance data showing NEE to range between -15 and -20 µmol m -2 s -1 around midday and 0-5 µmol m -2 s -1 during the night (Fig. 6). However, the sharp rise in OCS concentrations between 6 am and 12 am UTC (Fig. 2) and the reversal of the CO 2 gradients at 5-6 pm UTC (Fig. 3) reduce the time window to few hours in the afternoon where 10 the ecosystem relative uptake of OCS (ERU), which is the ratio of the relative vertical gradients of OCS and CO 2 , can be assessed. ERU is an important parameter since it is proportional to GPP/NEE scaled by the ratio of relative leaf exchange rates (LRU) following Eq. 2:

GPP/NEE = ERU/LRU
(2) (Campbell et al., 2008;Blonquist et al., 2011). Therefore, we anticipate that this approach to partition measured NEE will 15 hardly be applicable at O3HP not only because the amplitude of the diurnal variations in LRU is a general unknown, but also because OCS vertical gradients cannot be calculated from measurements carried out throughout the whole period of illumination. In 2012, only data collected in the afternoon were exploitable. In June 2013, polluted air masses produced dramatic variation in atmospheric OCS-to-CO 2 ratios in the morning and the afternoon, leaving no time window for ERU calculation. These air masses were not related with urban photosmog episodes since there was a gap of ~ 4 hours between 20 the peaks of OCS (up to 700 ppt) and O 3 (up to 85 ppb). With these caveats in mind, the ratio of the mean relative vertical gradients of OCS and CO 2 (calculated from linear OCS profiles) was equal to 4.3 for both the afternoons of June 6 and 17 with, however, large relative error (≥ 50%), and was consistent with ERUs reported by Blonquist et al. (2011) at the Harvard Forest AmeriFlux site in summer-autumn 2006.

Relative role of plants and soil on OCS exchange 25
Observations at O3HP and especially (1) the lack of nocturnal vertical gradients in OCS (Fig. 5), (2) a much higher decrease in OCS concentration after sunrise than during the night ( Fig. 2c and Fig. 4) and (3) the ERUs measured in the afternoons suggest that plant OCS uptake is the only relevant biospheric flux. Moreover, the early morning drop of OCS coincides with a rise of GPP (Fig. 6b), isoprene emissions (Fig. 6c, isoprene being a metabolic product of currently fixed carbon), latent heat (Fig. 6a), stomatal conductance (Fig. S2), and a drop of CO 2 (Fig. 2c). However, the latter is followed by a secondary 30 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c Author(s) 2016. CC-BY 3.0 License. CO 2 maximum due to the mixing of air from below the canopy (richer in CO 2 but not poorer in OCS since O3HP soils and plants do not take up this gas at night) with air from above. The lack of net uptake of OCS during the night is a specific feature to the O3HP site that is not shared by other open oak woodlands characterized by a Mediterranean climate (Kuhn et al., 1999;Sun et al., 2015). The study of Kuhn et al. (1999) Sun et al., 2015).
OCS fluxes at Stunt Ranch exhibited clear diurnal variations with higher uptakes during the night than during the day (Sun et 10 al., 2015). Unfortunately, the signature of these fluxes in the nocturnal boundary layer in terms of nocturnal drop in OCS mixing ratio where not reported in that manuscript. To give an illustration of what might be the atmospheric signature during stable nocturnal conditions of OCS uptake events of such intensity, we extracted data from a set of observations where the role that soil, leaf and atmospheric dynamics have on surface OCS exchange is investigated from OCS diurnal cycles (as at O3HP) and nocturnal fluxes calculated using the Radon-Tracer Method (Belviso et al., 2013). Figure S3 shows an eight day 15 time series of ambient mixing ratios of OCS, CO 2 , CO and O 3 carried out in mid-April 2015 (after bud break and almost complete leaf expansion) in a suburban area of the Saclay Plateau (Paris region), in relation to incoming global radiation, thermal stratification and wind speed (as at O3HP). Periods of low atmospheric turbulence over the Saclay Plateau were evaluated using 222 Rn accumulations. In April 2015, hourly variations show night-time and early morning decreases of OCS mixing ratios (Fig. S3c) and corresponding 222 Rn increases (Fig. S3b). The amplitude of OCS diurnal variations is in the 40-20 80 ppt range. OCS minima coincide with calm meteorological conditions with wind velocities lower than 6 km h -1 (Fig. S3b) which are favorable to thermal stratification (Fig. S3a), with CO 2 maxima sometime up to ~ 480 ppm (Fig. S3c) and with O 3 minima down to few ppb (Fig. S3d). However, it is worth noting here that the amplitudes of CO 2 and O 3 nocturnal variations over the Saclay Plateau in early spring are higher than those at O3HP due to anthropogenic emissions of CO 2 , which can be traced using CO mixing ratios (Fig. S3d), and to NOx emissions which accelerate the chemical removal of O 3 (O 3 reacts with 25 NO, data not shown). OCS fluxes calculated using the Radon-Tracer Method during stable nocturnal conditions ranged from -4.8 pmol m -2 s -1 (night of the 14 th ) to -14.2 pmol m -2 s -1 (night of the 11 th , Fig. S3c). They fall in the upper range of fluxes reported by Kuhn et al. (1999) and Sun et al. (2015) but the comparison should be made with caution because three different methods were used to estimate the OCS fluxes (i.e., a boundary layer model, soil chambers and the Radon-Tracer Method).
Qualitatively, it is clear that uptake rates of several pmol m -2 s -1 lead to drops in the OCS ambient mixing ratio by several 30 tens of ppt during periods of low atmospheric turbulence. Hence, a major difference between the three open oak woodlands investigated so far during springtime is that soil of the Mediterranean forest ecosystem of Southern France is not a net sink of OCS.
Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c Author(s) 2016. CC-BY 3.0 License. Soil OCS uptake has been shown to be dependent on soil physical properties like soil structure, water content, water-filled pore space and temperature (Van Diest and Kesselmeier, 2008;Ogée et al., 2016) but also on soil biological properties like microbial activity (Kato et al, 2008;Ogawa et al., 2013), active roots density (Maseyk et al., 2014) or the presence of a litter layer (Berkelhammer et al., 2014;Sun et al., 2015). Away from a range of optimum uptake, which varies between soils, changes in soil water content and temperature can markedly reduce OCS uptake by soils (Van Diest and Kesselmeier, 2008). 5 However, the soil temperature and water content at O3HP (Fig. 1c,d) are typically in the range of optimum uptake published by Van Diest and Kesselmeier (2008). A limitation of OCS uptake by soils due to a poor OCS diffusion is moreover unlikely considering that the soils from the O3HP are strongly structured and are far from being water saturated. Finally, the only physical property of soil differing among the three open oak woodlands is the soil texture with a fine clayey texture at O3HP but a coarse sandy loam texture at Hastings Reservation (Kuhn et al., 1999) and at stunt Ranch (Sun et al., 2015). If OCS 10 uptake by fine-textured soils have already been reported (Maseyk et al., 2014), this result pointed out the need for measurements of OCS uptake for a greater diversity of soils. Concerning the biological soil properties, the soil at O3HP is covered by a relatively thick litter layer that may induce a change from OCS uptake to OCS emission (Berkelhammer et al., 2014). Sun et al. (2015) however measured at Stunt Ranch that the litter was responsible for OCS uptake. The surface horizons at O3HP showed organic carbon contents ranging from 167 to 43 g.kg -1 in the surface soil horizons (Table 1) but  15 only 24 g.kg -1 at Hastings Reservation (no data on soil organic carbon are available for Stunt Ranch). Being richer in organic carbon, soils at O3HP show very likely higher microbial activity, factor which should stimulate uptake of OCS by soils but apparently do not. If the capacity of soils to consume OCS is more related to specific enzymatic activities (carbonic anhydrases (CA) and OCS hydrolases) than to the general variables presented above, our observations would highlight deficiencies in these enzymatic activities in calcium carbonate rich soils of O3HP. However, this hypothesis is not consistent 20 with the suggestion that CA performs essential role for microbial organisms to survive periods of osmotic stress such as drought at the surface of Mediterranean soils (Wingate et al., 2008). Finally, as roots and associated rhizosphere have been found to produce OCS, a greater abundance of roots in the surface soils at O3HP by comparison to the two other oak woodlands may explain why the soils at O3HP are not a sink of OCS. In other words, the lack of nocturnal net uptake of OCS would indicate that gross consumption of this gas in soil is compensated by emission processes that remain to be 25 characterized. No data on roots abundance are however available at Hastings Reservation or Stunt Ranch to confirm such hypothesis.

Potential use of OCS to partition ozone decay near the ground
Data show strong similarities during the night and early morning hours between OCS and O 3 diel variations at the O3HP suggesting a similar sink during that period (Fig. 3). At the O3HP, volatile organic compounds (VOCs) produced by the 30 vegetation are essentially in the form of isoprene (Kalogridis et al., 2014;Genard-Zielinski et al., 2014). Isoprene is oxidized in the atmosphere by the hydroxyl radical (OH), O 3 and the nitrate radical (NO 3 ), but in-canopy chemical Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c Author(s) 2016. CC-BY 3.0 License. oxidation of isoprene at O3HP was found to be weak and did not seem to have a significant impact on isoprene concentrations and fluxes above the canopy (Kalogridis et al., 2014). Hence, ozone deposition at the O3HP was essentially through leaf uptake via stomata and surface deposition, without a strong contribution from chemical reactions. In late May and early June 2012, the non-stomatal contribution to the ozone flux was in general markedly higher than the stomatal one in the morning hours (before 10:00 UTC) but became much less significant in the afternoon (Fig. 7b). During the second week 5 of June, however, although there were still signs of non-stomatal loss of ozone in the morning, the major contribution to ozone deposition was through the stomatal pathway (Fig. 7b). The analogy with OCS during nighttime and early morning suggests that soil did not contribute much to the O 3 flux and that the deposition flux of O 3 in mid-June was essentially the result of leaf uptake. It is however difficult to evaluate the soil ozone pathways without turbulence measurements inside the canopy. It would be worth looking further on how OCS could be used to partition ozone fluxes near the ground between soil 10 and leaf deposition processes. The applicability of OCS to characterize the strength of ozone sinks would be reduced in situations where NOx would significantly impact on the chemical production or destruction of ozone in the canopy or when background air is contaminated by primary or secondary anthropogenic sources of OCS (Fig. 3b).

Conclusions and perspectives
The requirements for OCS uptake to be used as a proxy of GPP are fulfilled at O3HP at least during springtime. Indeed, diel 15 changes in OCS mixing ratio and in its vertical distribution show that the soil uptake of OCS is negligible compared to the uptake of this gas through the stomata, a feature which is not shared by other oak woodland ecosystems characterized by a Mediterranean climate. Hence, O3HP would be the adequate place to support the installation in the Mediterranean region of a monitoring station of OCS uptake by plants from eddy covariance measurements. However, the assessment of GPP from measured OCS fluxes remains tributary of our poor knowledge of the magnitude of the LRU diel variations which requires 20 further examination. The second method to estimate GPP requires NEE measurements, OCS and CO 2 vertical gradient and LRU measurements during daytime. Unfortunately, the time window for calculation of the Ecosystem Relative Uptake of OCS was found to be restricted at O3HP to few hours after midday (1) because the vertical distribution of OCS is disrupted by entrainment in the morning of OCS rich tropospheric air sometimes contaminated by anthropogenic emissions, and (2) because the CO 2 vertical gradient reverses when it is still light. So at the O3HP, uncertainties both in ERU and LRU during 25 the light period are hindering the suitability of OCS to infer GPP from NEE. In the framework of the European infrastructure Integrated Carbon Observation System (ICOS), an atmospheric measurement station (100m height tower) has been set up at OHP in the year 2014 to determine multi-year records of greenhouse gases. This site being suitable to perform continuous and high precision vertical profiles of OCS using quantum cascade laser spectrometry, improvements in the determination of the ERU are foreseen. 30 Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c Author(s) 2016. CC-BY 3.0 License.  Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys.  Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c Author(s) 2016. CC-BY 3.0 License. Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c Author(s) 2016. CC-BY 3.0 License. Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys.   Atmos. Chem. Phys. Discuss., doi:10.5194/acp-2016-525, 2016 Manuscript under review for journal Atmos. Chem. Phys. Published: 5 July 2016 c Author(s) 2016. CC-BY 3.0 License.