Effects of grass leaf anatomy, development and light/dark alternation on the triple oxygen isotope signature of leaf water and phytoliths: insights for a new proxy of continental atmospheric humidity

Continental relative humidity (RH) is a key-climate parameter. However, there is a lack of quantitative RH 15 proxies suitable for climate model-data comparisons. Recently, a combination of climate chamber and natural transect calibrations laid the groundwork for examining the robustness of the triple oxygen isotope composition (d18O, d17O) of phytoliths as a new proxy for past changes in RH. However, it was recommended that besides RH, additional factors that may impact d18O and d17O of plant water and phytoliths be examined. Here, the effects of leaf anatomy, leaf development stage and day/night alternations are addressed from the 20 growth of the grass species F. arundinacea in climate chambers. Plant water and phytoliths are analyzed in d18O and d17O. Silicification patterns are examined using light and scanning electron observation of phytoliths. The isotope data show the increasing contribution of evaporated epidermal water to the bulk leaf water, from sheath to proximal and apical leaf blade. However, despite this isotope heterogeneity, d18O and d17O of the bulk leaf water can be predicted by the Craig and Gordon model, in the given experimental conditions (high 25 RH). Regarding phytoliths, their forming water (mainly epidermal) is, as expected, more impacted by evaporation than the bulk leaf water. This discrepancy increases from sheath to proximal and apical blade and can be explained by the steepening of the radial concentration gradient of evaporated water along the leaf. However, we show that because most of silica polymerizes in epidermal long cells of the apical blade of the leaves, the d18O and d17O of bulk grass phytoliths should not be impacted by the diversity in grass anatomy. 30 The data additionally show that most of silica polymerizes at the end of the leaf elongation stage and at the transition towards leaf senescence. Thus, climate conditions at that time should be considered when interpreting d18O and d17O of phytoliths from the natural environment. At least, no light/dark effect was detected on the d18O and d17O signature of plant water and phytoliths of F. arundinacea. However, when day/night alternations are characterized by significant changes in RH, the lowest RH conditions favoring 35 evaporation and silica polymerization should be considered when calibrating the phytolith proxy. This study contributes to the identification of the parameters driving the d18O and d17O of bulk grass phytoliths. It additionally brings elements to further understand and model the d18O and d17O of grass leaf water, which influences the isotope signal of several processes at the soil/plant/atmosphere interface. 40


Introduction
Recently, a combination of growth chamber and natural transect calibrations laid the groundwork for examining the robustness of the triple oxygen isotope composition of phytoliths as a new proxy for past changes in continental atmospheric relative humidity (RH) (Alexandre et al., 2018).Continental RH is a key-Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-73Manuscript under review for journal Biogeosciences Discussion started: 27 March 2019 c Author(s) 2019.CC BY 4.0 License.climate parameter.When combined with atmospheric temperature, it can be used to estimate the concentration 45 of atmospheric water vapor, one of the main components of the global water cycle.However, global climate models have difficulties to properly capture continental RH (Sherwood et al., 2010;Risi et al., 2012;Fischer and Knutti, 2013) and there is a lack of RH proxies suitable for model-data comparisons.
Phytoliths are micrometric particles of hydrous amorphous silica (SiO2 (H2O)n) that form in and between living plant cells.Silica polymerizes from the sap that contains dissolved silicon (among other nutrients) absorbed 50 by the plant roots from the soil.Phytoliths, can take the shape of the cells they form in, which gives them morphological taxonomic properties.Hence, phytolith morphological assemblages extracted from buried soils, loess and sediments are commonly used for paleoenvironmental reconstructions (e.g.Miyabuchi and Sugiyama, 2015;Nogué et al., 2017;Woodburn et al., 2017).In grasses, silica that can represents several percent of the dry weight (d.w.) (Alexandre et al., 2011), polymerizes mainly in the stem and leaf epidermis 55 from where the plant water evaporate during transpiration (Kumar et al., 2016).This polymerization is assumed to occur in isotope equilibrium with the plant water (Alexandre et al., 2018;Shahack-Gross et al., 1996;Webb and Longstaffe, 2000).
Variations in the triple oxygen isotope composition (d 18 O, d 17 O) of natural surficial waters, and consequently of minerals formed in isotope equilibrium with these waters (Gázquez et al., 2015;Herwartz et al., 2017; 60 Passey et al., 2014) are mainly driven by the extent of kinetic fractionation during evaporation.To the contrary, they are only weakly sensitive to distillation processes (Angert et al., 2004;Barkan and Luz, 2007;Landais et al., 2008;Uemura et al., 2010;Steig et al., 2014) and not significantly affected by temperature (Barkan and Luz, 2005;Uemura et al., 2010), in contrast to variations in the deuterium-excess (d-excess = δ 2 H -8.0 x δ 18 O).This makes the triple oxygen isotope composition of surface waters a powerful tool for tracing 65 evaporative conditions (Luz and Barkan, 2010;Surma et al., 2018Surma et al., , 2015) ) that are mainly dependent on atmospheric RH (Cernusak et al., 2016;Craig and Gordon, 1965).In plant water, changes in the triple oxygen isotope composition due to evaporation during transpiration, can exceed most of the variations identified so far in seawater, surface water, rainfall and ice (Landais et al., 2006;Li et al., 2017;Sharp et al., 2018) In Alexandre et al. (2018) the triple oxygen isotope composition of phytoliths from a grass species (Festuca 70 arundinacea) grown under controlled conditions of RH, the other climate parameters being set constant, was examined.A linear relationship with RH was demonstrated.The triple oxygen isotope composition of soil phytoliths collected in West Africa along a vegetation and RH transect where soil water availability and climate parameters other than RH changed simultaneously, also showed a linear regression with RH, close to the correlation obtained from the growth chamber.This linear regression allowed for the prediction of RH 75 with a standard error of 5.6%.However, it was recommended that besides RH, additional factors that may impact the isotope composition of bulk phytolith samples be examined, before using this relationship for paleoenvironmental reconstruction.
In particular, in nature, the biomass of grass stem, sheath and blade, as well as the geometry of leaf blade is highly species-dependant.Previous studies showed that for grasses the water δ 18 O increases from stem to blade 80 and from the bottom to the tip of the blade (e.g.Helliker and Ehleringer, 2000;Farquhar and Gan, 2003;Webb and Longstaffe, 2003;Cernusak et al., 2016).Meanwhile, the 17 O-excess decreases (Landais et al., 2006).This raises the question of the effect of grass anatomy diversity on the triple oxygen isotope signal of bulk phytolith samples.Moreover, climate parameters (e.g.mean annual RH or mean RH of the growing season; mean daily RH or RH at 12:00 UTC…, etc) to be considered when interpreting the triple oxygen isotope signature of bulk 85 grass phytoliths recovered from soils or sediments depends on when most of the silica deposition occurs in plants.In grasses, silica deposition can be either metabolically controlled or passive, i.e. depending mainly on silica saturation during cell dehydration when the leaf water evaporates (Kumar et al., 2017 andreferences therein, Kumar et al., 2019).The contribution of evaporated water to the bulk leaf water is expected to vary with transpiration (Cernusak et al., 2016) which decreases from day to night (Caird et al., 2007) and when 90 senescence occurs (Norton et al., 2014).This raises the question of the impact of senescence and day/night alternations on the triple oxygen isotope composition of bulk leaf water and phytoliths.In order to address these issues, a first growth chamber experiment was set up to explore the spatial heterogeneity in d 18 O and d 17 O of water and phytoliths in F. arundinacea.Bulk water and phytoliths were extracted from sheaths, stems, proximal and apical leaves, young, mature and senescent bulk leaves.Silica 95 concentration was measured and phytolith morphological assemblages, that give information on the type of tissue and cells that are silicified, were examined in the same samples.A second experiment was set up to explore light/dark and day/night alternations on d 18 O and d 17 O of plant water and phytoliths of F. arundinacea.Implications for the calibration of the triple oxygen isotope composition of phytoliths as a RH proxy are discussed in the light of the newly acquired results.100

The triple oxygen isotope composition of waters and minerals: notations
In the oxygen triple isotope system (d 18 O, d 17 O), the fractionation factors ( 17 a and 18 a) are related by the exponent q where 17 a = 18 a q or q = ln 17 a / ln 18 a.q can also be formulated as q = D' notations are expressed in ‰ vs VSMOW.For atmospheric temperature conditions, when isotope equilibrium 105 is reached between water and silica or between liquid and vapor, q stays quasi-constant.For water-silica, qsilicawater equals 0. 524 for the 5-35°C range (Sharp et al., 2016).For water liquid-water vapour qequil equals 0.529 for the 11-41°C range (Barkan and Luz, 2005).When evaporation occurs, a fractionation due to the vapour diffusion in air is added to the equilibrium fractionation, as conceptualized by the Craig and Gordon model (Craig and Gordon, 1965;Gat, 1996).qdiff associated with this diffusion fractionation equals 0.518 (Barkan 110 and Luz, 2007).When RH decreases, amplitude of the diffusion fractionation governed by qdiff increases.q is linked to a particular physical process.When several processes occur at the same time, the resulting fractionation exponent is rather named l.In the d'  Barkan, 2010) which, in the d' 17 O vs d' 18 O space, is the departure from a reference line with the slope l of 0.528.This reference line is the trend along which meteoric waters were initially shown to plot (Luz and Barkan, 2010).As it is close to the liquid-vapor equilibrium exponent qequil (0.529), its use makes the 17 O-excess very convenient to highlight kinetic processes that result from evaporation.As the magnitudes of changes of the 120 triple oxygen isotope composition in the water cycle and terrestrial minerals are very small, the 17 O-excess is expressed in per meg (per meg = 10 -3 ‰).Here, we additionally use the difference in 17 O-excessA-B associated with a chain of processes from A to B ( 17 O-excessA-B = D' 17 OA-B -0.528 x D' 18 OA-B) (Hayles et al., 2018).

Material
In three growth chambers, the grass species F. arundinacea was sown and grown in commercial potting soil 125 in a 35 L container (53 x 35 x 22 cm LxWxD).Ten days after germination, agar-agar was spread on the soil surface around the seedlings, to prevent any evaporation from the soil as described in Alexandre et al. (2018).Ambient RH was kept constant in the growth chamber by combining a flow of dry air and an ultrasonic humidifier that produces vapour without any isotope fractionation.The vapour and the soil irrigation water (IW) came from the same source and their triple oxygen isotope composition was similar (-5.6 ± 0.0 ‰ and 130 26 ± 5 per meg for d' 18 O and 17 O-excess, respectively).
Experiment 1.This experiment was designed to examine the grass leaf water and phytolith isotope signatures in different parts of the leaf and at different stages of the leaf development.Briefly, the stages taken into account were i) young leaf, where only the end of the blade is visible as it emerges from the sheath of the preceding leaf, ii) adult leaf where the blade is fully developed, the ligule visible and the sheath is well formed, 135 and iii) yellow and desiccated senescent leaf.
F. arundinacea was grown for 39 days, in a climate chamber where light, air temperature and RH were set constant at 290 µmol/m 2 /sec, 20°C and 73% respectively.On day 28, irrigation was stopped to force senescence of the leaves.Eleven days later, a total of 197g of biomass was collected.From this biomass, young leaf (visible end of the blade), adult leaf and senescent leaf (blade only) were separated.Adult leaves were 140 sectioned into three parts: sheath, proximal part of the blade (10 cm long) and apical part of the blade.Five samples resulted (Table 1).For all samples except the senescent leaves, three to five g of biomass were put in gastight glass vials and kept frozen for bulk leaf water extraction.Senescent leaves were too dry for water extraction.The rest of the biomass (between 10 and 70 g depending on the sample, (Table 1) was dried for phytolith extraction.

145
Experiment 2.a Light triggers the opening of plant stomata with, as an inevitable consequence, an increase in water loss through these stomata.At night, however, stomata often do not close totally.Night transpiration is often 5 to 15 % of day transpiration (Caird et al., 2007).In F. arundinacea, stomatal conductance at night can be as high as 30 % of conductance during the day (Pitcairn et al., 1986).Together with difference in air RH between day and night, this could affect isotope enrichment of leaves (Barbour et al., 2005).This experiment 150 was thus designed to assess whether light/dark alternation may impact the isotope signature of F. arundinacea leaf water.
In a growth chamber, F. arundinacea was grown for 22 days with constant light.Then, a 12h light/12h dark alternation was introduced.Temperature and RH were kept constant at 25°C and 60% respectively.Half of the biomass was harvested at the end of day 22 (constant light).The second half was harvested at the end of 155 the alternation period on day 26.In order to consider potential spatial heterogeneity, leaf blades (both young and adult leaves) were collected from four different places in the culture for each harvest.The eight resulting samples (Table 2) were put in gastight glass vials and kept frozen for bulk leaf blade water extraction.
Experiment 2.b In natural conditions, day/night alternations imply changes in temperature and RH in addition to changes in light intensity.This experiment was designed to assess whether over a period of several day/night 160 alternations, changes in RH during the night impacted the mean isotope signature of grass leaf phytoliths.
For this experiment, the leaf water was not analyzed as it only gives a snapshot of its isotope composition.F. arundinacea was grown in two growth chambers.In the first chamber, light, temperature and RH were kept constant (290 mmol/m2/sec, 25°C and 60%, respectively).In the second chamber, 12H day/12H night alternations were set.During the day (light 290 µmol/m2/sec), temperature and RH were set to 25°C and 60%, 165 respectively, whereas during the night (no light), they were set to 20°C and 80%, respectively.The leaf blades (both young and adult leaves) were harvested after a first growth of 16 days and a second growth of 18 days (Table 2) and dried for phytoliths extraction.

Phytolith chemical extraction, counting and analysis 170
Phytoliths were extracted using a high purity protocol with HCl, H2SO4, H2O2, HNO3, KClO3 and KOH at 70 °C following Corbineau et al. ( 2013) and Alexandre et al. (2018).Phytoliths were weighed and their mass reported to the initial leaf dry weight (d.w.).To account for leaf mass loss during senescence, a mass loss correction factor of 0.7, previously estimated for graminoids (Vergutz et al., 2012) was applied to the phytolith concentration in senescent leaves (Table 1a).

175
Most grass phytoliths have a morphology characteristic of their cell of origin.Phytolith morphological assemblages were thus determined to follow the spatial evolution over time of the leaf silicification.Phytoliths assemblages from experiment 1 were mounted on microscope slides in Canada Balsam, and counted in light microscopy at a 600X magnification.More than 200 phytoliths with a dimension greater than 5 µm and with a characteristic morphology were counted.Phytolith types were named using the International Code for Phytolith 2017).In addition, thin silica particles with uncharacteristic shape and with a refractive index too low to be accurately described using light microscopy were also counted.Abundance of the phytolith categories are expressed in % of the sum of counted particles.Three repeated counts usually give an error lower than ± 5 % (SD).The phytolith assemblages were further observed with a Scanning Electron Microscope (FEG-SEM, HITACHI 190 SV6600, accelerating voltage of 3KV, 15degree tilt, working distance of 14mm and probe current of a few pA to avoid charging issues), after carbon coating.
Phytoliths triple oxygen isotope analysis was performed as described in details in Alexandre et al. (2018).The IR Laser-Heating Fluorination Technique (Alexandre et al., 2006, Crespin et al., 2008;Suavet et al., 2010) was used to extract the O2 gas after a dehydration and dehydroxylation under a flow of N2 (Chapligin et al., 195 2010).The purified O2 gas was analysed by dual-inlet isotope ratio mass spectrometry (IRMS, ThermoQuest Finnigan Delta Plus) against a working O2 standard calibrated against VSMOW.Each analysis consisted in two runs of eight dual inlet measurements with an integration time of 26s.The sample isotope compositions were corrected on a daily basis using a quartz laboratory standard (Boulangé).During the measurement period, Boulangé reproducibility (SD) was ± 0.13 ‰, ± 0.07 ‰ and ± 11 per meg for δ 18 O, δ 17 O and 17 O-excess 200 respectively (n = 9).For a given sample two phytoliths aliquots were analyzed.

Leaf water extraction and analysis
Leaf water was extracted during 6 hours using a distillation line.Then a fluorination line was used to convert water to oxygen using CoF3.Oxygen was analyzed by dual inlet IRMS (ThermoQuest Finnigan MAT 253) against a working O2 standard calibrated against VSMOW.The detailed procedure was previously described 205 in Landais et al. (2006) and Alexandre et al. (2018).The reproducibility (2 replicates) was 0.015 ‰ for d 17 O, 0.010 ‰ for d 18 O and 5 per meg for 17 O-excess.

Irrigation and vaporization water analysis
The irrigation and vaporization waters were analyzed with an isotope laser analyzer (Picarro L2140i) operated in 17 O-excess mode using an auto-sampler and a high precision vaporizer as described in detail in Alexandre

Phytolith concentrations, assemblages and origins in grass leaf (experiment 1)
In adult leaves, the phytolith concentration that ranges from 0.7 to 0.8% dry weight (d.w.) in sheaths and 215 proximal blades increases to 2.1% d.w. in apical blades (Table 1a, fig.2a).Thus, close to 60% of leaf phytoliths precipitate in apical blades.From young to adult and senescent blade, the phytolith content increases sharply from 0.8 to 1.4 and 3.0 % d.w.This means that close to 60% of blade phytoliths precipitate in the late period of leaf development, at the transition between adult and senescent stages.Short cell phytoliths (fig. 1) are found in all samples but their relative abundance decreases from young to 220 adult and senescent leaf blade and in adult leaf from sheath to proximal and apical blade (table.1b).Conversely, long cell phytolith (fig. 1) abundance increases with phytolith concentration from young to adult and senescent leaf blade (fig.2a).Long cell phytoliths are absent from the sheath but reach 14% in proximal blade and 48% in apical blade.Parallelepipedal bulliform and Acicular hair phytoliths can be observed but in small amount (<2%) in young leaves and senescent leaf blade samples.All phytolith assemblages contain thin 225 silica particles with low refractive index, difficult to count with accuracy in light microscopy.SEM observation shows they are composed of multi-cellular silica sheets (mostly silicified cell walls and a few silicified stomata complexes) (fig.1).Their abundance ranges from 16 to 33% in young and adult leaf blade and increases up to 52% in senescent leaf blades (Table 1b).In summary, silicification occurs mainly in the epidermis, starting with short cells and continuing with long cells.Cell walls also silicify (silica sheets), in low proportion when 230 the leaf is young or adult, in higher proportion when the leaf is close to senescence.Because these silica sheet particles are very thin, their weight contribution to the isotope signature of bulk phytolith assemblages is expected to be significantly lower than their number.2b and 2c).Water from the young leaf blade plots on this line, close to water from the adult upper leaf blade (fig.2b).

Heterogeneity in the triple oxygen isotope composition of leaf silica and leaf silica forming water
Polymerization of silica is supposed to occur in isotope equilibrium with the formation water, and, therefore, 245 its isotope composition should only be governed by temperature and the isotope composition of the formation water (FW) (Alexandre et al., 2018;Dodd and Sharp, 2010;Sharp et al., 2016).The d' 18 O of this formation water can be estimated from the temperature-dependent equation established by Dodd and Sharp (2010) from measurement of lake diatoms and lake water.At 20.4°C, the fractionation (D' 18 O) between water and silica accounts for 31.07‰.Then, equation 10 from Sharp et al. (2016) can be used to calculate qsilica-water (qsilica-water 250 = 0.5242) for the same temperature.Finally, the enrichment in 18 O and the difference in 17 O-excess between phytolith forming water and irrigation water (respectively D' 18 OFW-IW and 17 O-excessFW-IW) can be estimated (Table 1a).
When plotted in the 17 O-excess vs D' 18 O space (fig.2c), the phytolith forming waters from adult leaf show a clear evaporative fractionation trend from the sheath to the proximal and apical blade, similarly to the bulk 255 leaf water.However, as early as in the sheath the phytolith forming water appears more impacted by evaporation than the bulk water.This discrepancy increases from the sheath to the proximal and apical blade (fig.2c).This is also illustrated by the lFW-IW values (ranging from 0.516 to 0.517) that are significantly lower than the lLW-IW values (ranging from 0. 524 to 0.526) (table 1a).Variations in lFW-IW or lLW-IW values are too small to be interpreted.Interestingly, the discrepancy between 17 O-excessLW-IW and 17 O-excessFW-IW is the same 260 for the apical blade as for the bulk leaf (-119 and -129 per meg, respectively), in link with the dominance of apical blade phytoliths in the leaf phytolith assemblages (Table 1a).
The young leaf blades and senescence leaf blades phytolith forming waters plot on or close to the line in fig.2c, and show similar 17 O-excessPhyto-IW and D' 18 OPhyto-IW values than phytolith forming water in the adult leaf blade, given the measurement precision.

265
In adult leaves, silicified short cell and silica sheets abundances decrease when 17 O-excessPhyto-IW or 17 O-excessFW-IW decrease from sheath to the proximal and apical leaf blade (r 2 > 0.8).The inverse occurs for silicified long cells (fig.2d, r 2 = 0.9).As previously noted, young, adult and senescent leaf blade show similar 17 O-excessFW-IW although the senescent leaf blade is characterized by an amount of silica sheets (52% of counted particles) much higher than in young and adult leaf blades (24 and 18% of counted particles, 270 respectively).

Effect of light/dark alternation on the triple oxygen isotope composition of leaf water and leaf
silica.
Plant water isotope data from experiment 2a where light/dark alternations were set without changing RH, are presented in  In experiment 2b, transpiration and leaf blade phytolith concentrations are close when day/night alternation or constant day conditions are set (Table 2a).Phytolith concentrations range between 2.2 and 3.1% d.w, i.e. slightly higher than the adult leaf blade concentration obtained from experiment 1 (1.4% d.w.).A previous experiment showed that silicification increases with the duration of growth and transpiration (Alexandre et al., 285 2018).In experiment 2b, the period of growth is shorter than in experiment 1 (16 vs 39 days) but RH is lower during daytime (60 vs 73%).Isotope data from experiment 2b (Table 2) shows that D' 18 OPhyto-IW and 17 O-excessPhyto-IW values obtained for the same tank are very close when constant day conditions of light, temperature and RH or when day/night alternation conditions are set.lPhyto-IW ranges from 0.520 to 0.521.

Despite leaf water heterogeneity bulk leaf water triple oxygen isotope composition is predictable
In the conditions set for experiment 1, evaporative fractionation appears effective as soon as the water taken up by the roots reaches the sheath.This was expected as 18 O-enrichment of several per mil relatively to the soil water was previously measured in the water of grass sheaths (Webb and Longstaffe, 2003).Indeed, stomata are few but still present in the sheath (e.g.Chaffey, 1985) allowing weak transpiration (Martre et al., 2001).

295
Here, the low amplitude of the 17 O-excessLW-IW value (-11 per meg) supports that evaporated water weakly contributes to the water circulating in the sheath.
The strong evaporative 18 O-enrichment observed from experiment 1 in the adult leaf water from the proximal to the apical blade (fig.2b and 2c) was also expected.This has been commonly observed in monocot leaves and was modelled using Péclet numbers (synthesis in Cernusak et al., 2016;Farquhar and Gan, 2003;Helliker 300 and Ehleringer, 2000;Ogée et al., 2007).Briefly, after stomatal evaporation (evaporation model of Craig and Gordon, 1965), two main hydraulic processes combine in the leaf.The first one is a radial back-diffusion of evaporated 18 O-enriched water from the epidermis to the mesophyll and into the veins, which leads to the second one, i.e. the progressive enrichment of the vein water from the base to the tip of the leaf, such as in the case of a string of lakes (Gat and Bowser, 1991).In grasses, leaf blade is long and tapering, which influence 305 interveinal distances (decreasing from bottom to tip).While stomatal densities and the capacity of water movement within the xylem remain relatively constant along the blade (Martre et al., 2001), the diffusional path length from vascular bundle to stomata decreases.Consequently, the stomatal conductance increases from bottom to tip (Ocheltree et al., 2012;Affek et al., 2006).This likely triggers increasing radial back-diffusion of epidermal evaporated water which adds to the stem water already increasingly impacted by evaporation.

310
The linear relationship obtained when plotting 17 O-excessLW-IW vs D' 18 OLW-IW for sheath, proximal and apical adult leaf blades of F. arundinacea (fig.2c) agrees with this pattern.In the 17 O-excessLW-IW vs D' 18 OLW-IW space, water of young blades plots close to water of the apical adult leaf blades as both blades are narrow with small diffusional path length to stomata favoring back-diffusion of evaporated water.
As an exercise, we estimated D' 18 O LW-IW, D' 17 O LW-IW, 17 O-excessLW-IW and lLW-IW for the bulk leaf blade water 315 from experiment 1 (Table S1).For that purpose we used the Craig and Gordon model from Farquhar et al. (2007) (spreadsheet provided in Cernusak et al., 2016).The 17 O-excessLW-IW and lLW-IW values were calculated using equilibrium and kinetic fractionation factors (respectively 17 aeq and 17 ak) of 0.529 and 0.518.We assumed that the grass transpiration had reached a steady state as climatic conditions were set constant during the 39 days of growth.We additionally assumed that irrigation water and water vapour had the same isotope 320 composition since i) there is no soil evaporation, ii) transpiration should not fractionate (e.g.Welp et al., 2008) and iii) the vaporised water comes from the same source as the irrigation water and is not fractionated by the vaporiser.Last, we measured the temperature of adult leaf of F. arundinacea grown under conditions similar to those of experiment 1 (Table S2).No significant temperature difference was detected between the sheath,  S1).When the observed temperature difference was included in the calculation, the predicted values of D' 18 OLW-IW, D' 17 OLW-IW, 17 O-excessLW-IW and lLW-IW were similar to the observed values 330 (Table S1).This result supports that under the experimental conditions of high RH, despite heterogeneity in the triple oxygen isotope composition of the water along the narrow blade, the bulk leaf blade water presents an averaged isotope signature that can be predicted by the single-water-source Craig and Gordon model.This result is also a reminder that, in addition to RH and the isotope signature of the source water, the temperature difference between leaf and air is an important parameter to consider.335

Bulk leaf water vs phytolith forming water triple oxygen isotope compositions
Experiment 1 shows that the epidermal water from which silica polymerizes is more impacted by evaporative isotope fractionation than the bulk water.This discrepancy occurs as early as in the sheath and increases from proximal to apical blade.This suggests that from bottom to tip the concentration gradient of evaporated water that occurs radially in the leaf increases faster than the vertical concentration gradient in the veins.The oxygen 340 isotope composition of evaporated water is very dependent on the concentration and isotope composition of the atmospheric water vapor around the leaf.Thus, experimentations including vapor triple oxygen isotope measurements are necessary to further assess which model accurately describes changes in the composition of epidermal water on the one hand and in bulk leaf water on the other hand.

345
The phytolith content and assemblages obtained from experiment 1 can be discussed in light of previous studies investigating silica deposition in grasses.
In a growing grass leaf, the cells are produced at the base of the leaf and are pushed towards the tip through the elongation zone during the growth (Kavanová et al., 2006;Skinner and Nelson, 1995).Silica precipitate mainly in the epidermis.At the cell level, silicification, which is a rapid process taking a few hours (Kumar 350 and Elbaum, 2017), initiates either in the extra-membranous space or in the cell wall and proceeds centripetally until the cell lumen is filled up (Bauer et al., 2011).During cell lumen silicification, some cells are still viable and transfer their content to each other before their full silicification (Kumar and Elbaum, 2017).Cell wall silicification, not followed by cell lumen full filling, has been also frequently observed, both in the epidermis (Kumar et al., 2017) and in bundle sheath parenchyma cells surrounding the veins (Motomura, 2004).Over 355 the course of leaf development, short cells are the first to silicify.Then, when the leaves become mature, long cell silicification takes over (Motomura, 2004;Kumar et al., 2016).Results from experiment 1, i.e. increase of long cell and decrease of short cell abundance from young to adult and senescent leaf blade, confirm this pattern.Experiment 1 additionally show that cell wall silicification (not followed by cell lumen filling) occurs at all development stages and increases when the leaf reaches or is close to senescence.

360
Silicification processes can be metabolically controlled such as in the epidermal short cells, or passive, i.e. depending mainly on silica saturation during cell dehydration when the leaf transpires, such as in the epidermal long cells, bulliform cells, or hair cells (Kumar et al., 2017 andreferences therein, Kumar et al., 2019).The decrease in silicified short cell and silica sheet abundance from sheath to proximal and apical leaves when 17 O-excessPhyto-IW or 17 O-excessFW-IW decreases (i.e. when evaporation increases) corroborates that evaporation (and 365 induced transpiration) is not the main driver of these types of silicification.To the contrary, the strong negative relationship obtained between silicified long cells abundance and 17 O-excessPhyto-IW confirms a strong control of evaporation on long cell silicification.This result has implications for paleoclimate reconstruction: in fossil phytolith assemblages, provided that the phytolith type relative abundances are not biased by taphonomical processes, increasing long cell vs short cell phytoliths should covary with decreasing 17 O-excessPhyto and 370 increasing evaporative conditions.
6.4.Highlights for interpreting the triple oxygen isotope signature of grass phytolith assemblages In nature, the biomass of grass sheath, stem, lower and upper blade, with different 17 O-excessPhyto signatures, is highly species-dependant.However, if a large amount of silica polymerizes in the apical blades (52% of adult leaf silica in the case of F. arundinacea, Table 1), one may expect the RH dependant-evaporative 375 fractionation impacting the apical blades to imprint the 17 O-excessPhyto of all grass adult leaves.
Leaf senescence is a stress-induced or age-related developmental aging during which transpiration decreases to minimal level but is still efficient (Norton et al., 2014), epidermal conductance progressively prevailing over stomatal conductance (Smith et al., 2006).If the cells already contain dissolved silica, evaporation, not balanced by water input due to decreasing transpiration, may lead to silica saturation.Such a process likely 380 occurs at the transition between the end of the elongation stage and the beginning of the senescence stage.Thus, if more than half of the silica, and of the silicified long cells polymerizes at that time (tables 1a and 1b), the isotope composition of the bulk silica sample is expected to be impacted by the soil water isotope composition and the climate conditions (RH, isotope composition of the water vapour, leaf vs atmosphere temperature difference) occurring during this transition.In experiment 1, climate conditions are constant when 385 senescence occurs.In nature, seasonal climate change such as a drastic decrease of RH in the tropical and Mediterranean areas, are responsible for senescence, and may imprint the averaged isotope signature of grass phytoliths.Mere experimental monitoring of stress-induced (including changes in RH) senescence effect on 17 O-excessPhyto could determine whether RH averaged over both elongation and senescence time lapses should be considered when interpreting 17 O-excessPhyto.

390
The results obtained from experiment 2a showed very close isotope composition of bulk leaf water during light and dark periods.The constancy of atmospheric relative humidity and temperature, as well as the shortness of experiment 2a (4 day/night light alternations after more than 2 months of constant day light) may have played against the closure of the stomata at night.A previous study on elongating leaves of F. arundinacea showed that spatial distribution of water content within the elongation zone can stay almost 395 constant during the dark and light period (Schnyder and Nelson, 1988), supporting that dark/light alternations do not always impact the stomata openness.Assuming from experiment 2a that light/dark alternation has no impact on stomatal conductance, experiment 2b suggests that in case of RH variation from day to night, silica precipitation mainly occur when RH is the lowest (60% during the day instead of 80% during the night in experiment 2a), i.e. when evaporation and thus silica saturation are the highest.In the natural environment, 400 nights are often characterized by lower temperature and higher RH than days, which should favor higher evaporation and silicification rates during the day.In these conditions, the triple oxygen isotope composition of grass leaf phytoliths should correlate more closely to daytime RH conditions than to day and night averaged RH conditions.

405
Increasing contribution of evaporated epidermal water to the bulk leaf water, from sheath to proximal and apical leaf blade, that has been previously observed in grass leaves, is well recorded by the triple oxygen isotope composition of F. arundinacea leaf water.Despite this heterogeneity, in the given experimental conditions (high RH), the bulk leaf blade water presents an averaged isotope signature that can be predicted by the single-water-source Craig and Gordon model.

410
Regarding silica, its forming water is mainly epidermal and is thus more impacted by evaporation than the bulk leaf water.This translates into lower lFW-IW than lLW-IW values.This discrepancy increases from sheath to proximal and apical blade and is likely driven by the steepening of the radial concentration gradient of evaporated water along the leaf.However, results from experiment 1 show that most of silica polymerizes in epidermal long cells of the apical blade.Thus, the triple oxygen isotope composition of apical blade phytoliths 415 should imprint the isotope signal of bulk grass leaf phytoliths in nature, whatever is the diversity in grass anatomy.No light/dark effect was detected on F. arundinacea transpiration and 17 O-excessPhyto-IW in the frame of experiment 2. However, when day/night alternations are characterized by changes in RH, the lowest RH 425 conditions will likely favor higher stomatal conductance and more evaporation leading to silica polymerization.
Both experiments contribute to a more precise identification of the parameters to take into consideration when interpreting the 17 O-excessPhyto.They also brings elements to further understand and model the d 18 O and d 17 O of grass leaf water, which influences the isotope signal of several processes at the soil/plant/atmosphere 430 interface.
Irrigation and leaf water (IW and LW, respectively) d' 18 O, d' 17 O, and 17 O-excess values obtained from 235 experiment 1 are presented in Table 1a and figure 2b.As expected, the lowest d' 18 O and d' 17 O values occur in the adult leaf sheath which is supposed to transpire less than the leaf blade (Webb and Longstaffe, 2003, 2000).The values of d' 18 O and d' 17 O are however higher than the irrigation water by 7.1 and 3.7 ‰ respectively, showing that transpiration still occurs.The 17 O-excess of the leaf sheath water ( 17 O-excessLW) is only slightly lower (by 11 per meg) than 17 O-excessIW.In adult leaf waters, a clear evaporative fractionation trend occurs 240 from the sheath to the proximal and apical blade.It is expressed by a decrease in 17 O-excessLW (or 17 O-excessLW-IW) linearly correlated with an increase in D' 18 OLW (or D' 18 OLW-IW) (r 2 = 0.99, fig. Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-73Manuscript under review for journal Biogeosciences Discussion started: 27 March 2019 c Author(s) 2019.CC BY 4.0 License.considering the margins of error, the averaged values of D' 18 OLW-IW, 17 O-excessLW-IW and lLW-IW obtained after the dark period are similar to the ones obtained after the light period, suggesting that the contribution of evaporated water to the bulk leaf water stayed constant under constant RH.It was not possible to measure the 280 night and day transpiration flows or the stomatal conductance during the experiment.
Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-73Manuscript under review for journal Biogeosciences Discussion started: 27 March 2019 c Author(s) 2019.CC BY 4.0 License.lower and upper leaf blade because the leaf temperature was probably more controlled by light incidence than 325 by evaporation.However, the leaf was systematically 2°C cooler than the surrounding air.In the model, when the leaf blade temperature was set similar to the air temperature, the predicted and observed values of D' 18 OLW-IW, D' 17 OLW-IW, 17 O-excessLW-IW and lLW-IW were significantly different (predicted D' 18 OLW-IW being higher than measured D' 18 OLW-IW) (Table Biogeosciences Discuss., https://doi.org/10.5194/bg-2019-73Manuscript under review for journal Biogeosciences Discussion started: 27 March 2019 c Author(s) 2019.CC BY 4.0 License.Experiment 1 additionally show that most of silica polymerizes at the end of the elongation stage, close to senescence.Thus, climate conditions (RH, isotope composition of the water vapour, leaf vs atmosphere temperature difference) and soil water isotope composition occurring at the end of the elongation stage and at 420 the transition towards senescence should be considered when interpreting 17 O-excessPhyto.In nature, RH conditions leading to generalized leaf senescence should be considered, in addition to RH conditions prevailing during the grass leaf elongation, when interpreting 17 O-excessPhyto.
and l) between leaf water and irrigation water (LW-IW), phytoliths and irrigation water (Phyto-IW), phytoliths and leaf water (Phyto-LW) and phytolith-forming water/irrigation water (FW-IW) usingD 18Osilica-water calculated afterDodd and Sharp (2010) and qsilica-water calculated after Sharp et al.(2016, eq.10), are presented.For irrigation water (IW), 5.59‰, -and -2.92‰, respectively.b) Results of phytolith counting.Samples are named according to the climate chamber # they were collected in (P4), the set relative humidity (75%), the date of sampling (dd/mm/yy) and their anatomical origin."Adult leaves: bulk blade av." is the weighted average of values obtained for proximal and apical blade samples.Av : average ; n : number of replicates ; SD : standard deviation calculated on the replicates; Phyto Conc.(% d.w.) and Phyto proportion (%) stands for phytolith concentration expressed in % of the dry weight and phytolith proportion in % of adult leaf phytoliths.A mass loss correction factor of 0.7 was applied to the phytolith concentration in senescent leaves (see text for explanation).

-
, https://doi.org/10.5194/bg-2019-73Manuscript under review for journal Biogeosciences Discussion started: 27 March 2019 c Author(s) 2019.CC BY 4.0 License.Table 2. Growth chamber experiment 2a and 2b.Experimental set-up, phytolith content and triple oxygen isotope data obtained for phytoliths (Phyto), leaf water (LW) and irrigation water (IW).Isotope fractionation (and l) between leaf water and irrigation water (LW-IW) and phytoliths and irrigation water (Phyto-IW) are presented.Samples are named according to the climate chamber # they were collected in (e.g.F4), the date of sampling (dd/mm/yy) and the sampling after day or night (Day vs Night in experiment 2a) or after constant climate conditions or day/night alternation (Cst vs DN in experiment 2b).n : number of replicates ; SD : standard deviation calculated on the replicates; Phyto Conc.(% d.w.) stands for phytolith concentration expressed in % of the dry weight.