The experiment was established in the climate chamber facilities at the University of Bayreuth comprising seven species, including two herbaceous species (Plantago lanceolata L., Oxalis triangularis A. St.- Hil.), two shrubby species (Acacia longifolia (Andr.) Willd., Coffea arabica L.), and three tree species (Quercus suber L., Fagus sylvatica L., Picea abies (L.) H. Karst.) (see Table 1). This variety of species was chosen to cover different growth forms, leaf types, biomes, and layers of an ecosystem (understory vs. overstory). All plants, except F. sylvatica and P. abies, were grown in greenhouses. Fagus sylvatica and P. abies were obtained from a local nursery (Bavaria State Forest Enterprise, Bayreuth). At least 4 weeks before the experiment started, plants were placed in a climate chamber adjusted to an air temperature of 20°C, a relative air humidity (rH) of 60%, and a 12 h photoperiod with an average of 400–600 μmol m⁻² s⁻¹ incident at the upper leaf level (depending on the species height). CO₂-concentrations were not controlled and ranged from 300 to 500 ppm. Plants were kept well watered throughout the entire experiment, were regularly rotated to ensure similar light conditions, and fertilized with Hoagland solution (Peperkorn et al., 2005) and WUXAL Super (Manna, Wilhelm Haug, Ammerbuch-Pfäffingen, Germany). Measurements were conducted starting at least 3 h after start of the light period to ensure that the plants were not a in a transition stage and lasted not longer than 4 h. Isotopic variation of irrigation water was minimized using water from the same storage tank throughout the whole experiment. The isotopic signature of irrigation water was repeatedly determined showing mean δ¹⁸ O of −8.79 ± 0.53 ‰ (n = 10).

Stomatal density (mm⁻²) and guard cell length (μm) were determined (n = 15) from stomatal imprints by randomly covering leaf surfaces of both leaf sides with a thin layer of transparent nail polish (Kardel et al., 2010). High resolution pictures of imprints were taken with a light microscope (Zeiss Alitplano, Oberkochen, Germany) connected to a camera (Sony Nex-5, Tokio, Japan). Stomates were counted and guard cell length measured. This method could not be applied to Q. suber, hence values for stomatal density and guard cell length were taken from literature (Molinas, 1991). Three to five leaves of each individual were taken to determine leaf water contents. For P. abies three times five needles (n = 3–5) were collected while five needles were treated as one sample.

The first to develop an equation describing isotopic fractionation associated with evaporation of water were Craig and Gordon (1965). Accordingly, the isotopic ratio of evaporation R_E is linked to the isotopic ratios of water at the evaporating sites R_e and ambient vapor R_a (Craig and Gordon, 1965): (2)RE=1αkα+(1-h)(Re-α+hRa) with α_k and α⁺ being the kinetic and equilibrium fractionation factors (>1), respectively and h the relative humidity corrected for leaf temperature (see Table 2 for abbreviations). The Craig and Gordon steady-state model requires that the isotopic composition of vapor departing from the leaf must be the same as the isotopic composition of incoming water: R_E = R_s. This leads to: (3)Rc=αkα+(1-h)Rs+α+hRa where R_C is the isotopic composition of leaf water at the evaporating site in steady state.

The non-steady-state isotopic composition of leaf water at the evaporating site R_l can be written in an iterative form, if leaf water volume W [mol (H₂O) m⁻²] is assumed constant (Dongmann et al., 1974; Farquhar and Cernusak, 2005; Cuntz et al., 2007): (4)Rl(t+dt)=RC+(Rl(t)-RC)e-dtτ where R_l at a time t + dt is calculated from R_l at an earlier time t with constant environmental conditions during the time step dt. g_s is the leaf conductance for water vapor from the stomatal cavity to the point of observation, and w_i the humidity in the stomatal cavity, i.e., vapor saturation at leaf temperature expressed as mole fraction [mol(H₂O) mol(air)⁻¹].

Following Farquhar and Cernusak (2005) the time constant (τ), can be described as: (5)τ=(1-f)Wgswiαkα+ with α_k and α⁺ ≈ 1: (6)τ=(1-f)Wgswi

Recently, Song et al. (2015) suggested a modified formulation for τ specifically suitable for modeling isotopic signatures of leaf water and transpirative fluxes in a cuvette environment, substituting g_s × w_i with E and accounting for the proportional difference between w_in and w_i. In a cuvette scenario with the ingoing airstream (w_in) often being dry air (as in Song et al., 2015), this effectively corrects for the strong influence of E on isotopic signatures of vapor (δ_a).

This leads to: (7)τ=(1-f)WE(1-winwi)

We used this updated formulation for cuvette scenarios to calculate δ_E and hence also τ.

The factor f in Equation (5–7) denotes a Péclet term, with (8)f=1-e-℘m℘m with the Péclet number ℘m=ELCD where C = 10⁶/18 = 55.6·10³ mol m⁻³ is the molar water concentration, D (m² s⁻¹) is the tracer diffusivity in liquid water and L (m) is the effective length of water movement in the leaf mesophyll. E is the transpiration rate in mol m⁻² s⁻¹.

The isoflux of transpiration is expressed as the product of E and δ_E. In this experiment we calculated the isoflux as the product of E and Δ_E, as our main interest was in the impact of non-steady-state δ_E (Δ_E = δ_E − δ_X).

Kruskal-Wallis tests were used to test for species specific differences in stomatal density, size, leaf water content, E, g_s, and τ. The same test was used to test for species-specific differences in the magnitude of decrease/increase in the isotopic signature of transpiration, τ, and mean isofluxes of the three experimental stages in response to changes in rH. We performed a t-test to compare the isotopic signatures of transpiration with that of xylem water at the initial phase of the experiment. Non-linear correlations were used to relate stomatal conductance with stomatal density and guard cell length, stomatal density to guard cell length and leaf water content with stomatal traits. Power functions were used to relate species-specific differences in τ and isofluxes with their differences in g_s/W and stomatal density/W. In case significant relationships could be detected (p < 0.05), R² and p-values are given. Statistical analyses were carried out with the software STATISTICA 6.0 (StatSoft Inc., Tulsa, OK, USA).

Significant differences in stomatal density and guard cell length could be observed between species (Figure 1; see also Figure S2). Picea abies and O. triangularis had lowest stomatal densities of 68 ± 16 and 69 ± 14 mm⁻², respectively. Highest stomatal densities were observed in P. lanceolata, with 542 ± 86 mm⁻². Notably, three of the studied species, P. abies, A. longifolia, and P. lanceolata, possess amphistomatous leaves, hence stomatal densities are expressed here as total stomatal densities of both leaf sides. In addition, differences in guard cell length could be found, with a tendency toward smaller guard cell length at higher stomatal density (R² = 0.44; p = 0.08; ranging between 17 ± 3 μm in F. sylvatica and 40 ± 10 μm in P. abies). Moreover, together stomatal density and guard cell length (i.e., stomatal density ^* guard cell length) explained species-specific differences in stomatal conductance quite well (R² = 0.96, p < 0.01). Leaf water content (W) also differed considerably between species, but differences were not related to stomatal traits (Figure 1). Three species contained high leaf water contents between 11.7 ± 0.8 and 13.8 ± 2.8 mol m⁻² (P. abies, A. longifolia, P lanceolata), while medium leaf water contents could be found in O. triangularis, C. arabica, and Q. suber (~8.5 mol m⁻²). Fagus sylvatica had by far the lowest leaf water content of 2.5 ± 0.2 mol m⁻².

During the first initial stage of the measurement period, most species exhibited transpiration rates and stomatal conductance of around 0.6–1.2 and 70–90 mmol m⁻² s⁻¹, respectively (Figure 2). Only C. arabica and P. lanceolata showed significantly smaller (E = 0.41 and g_s = 37 mmol m⁻² s⁻¹) and higher (2.3 and 278 mmol m⁻² s⁻¹) E and g_s, respectively.

Similarly, the leaf water time constant τ (i.e., Equation 8) in the initial experimental phase varied between species, which accordingly can be divided into 3 main groups (Figure 2): (I) τ < 50 min (P. lanceolata and F. sylvatica); (II) τ > 50 but < 100 min (A. longifolia, Q. suber, O. triangularis) and III) τ > 100 min (C. arabica, P. abies). It has to be noted that O. triangularis faced a significant increase in τ following the first change in rH, thereafter belonging to group III (Figure 2). Mean values of measured bulk and modeled evaporating site leaf isotopic signatures is given in Table S1.

In concert with transpiration, isotopic signatures of leaf transpiration (δ_E) were measured first at environmental conditions of the climate chamber (rH = 60%), and following a decrease (to 30%) and increase (to 60%) in rH (Figure 3). In addition, δ_E was modeled using Equations (2) and (4), using the formulation for the time constant (Equation 7) as described by Song et al. (2015). During the initial stage, plants transpired at isotopic steady state, i.e., no significant deviation from source water isotopic signature (Figure 3) and a high agreement between measured and modeled isotopic signatures of transpiration (δ_E) was found. This can be expected here, as all plants were kept under stable environmental conditions regarding temperature and relative humidity within the climate chamber environment at all times. Further, during the initial phase of the experiment the conditions of the air entering the cuvette matched those within the climate chamber.

In response to environmental changes, (i.e., decreasing and increasing rH) a direct strong decrease/increase in δ_E was observed in all species. Notably, the magnitude of this immediate response of δ_E directly following a change in rH did not differ between species and was −8.27 ± 0.67 and 11.35 ± 0.9‰, respectively after decrease and increase in rH. In contrast, the studied species showed different behavior in the subsequent approach to isotopic steady state and three distinct groups could be identified, in accordance to their time constant. Only species of group I (P. lanceolata and F. sylvatica) reached isotopic-steady-state transpiration after 120 min of stable environmental conditions. Group III (P. abies, O. triangularis, and C. arabica) did not reach isotopic steady state even after 120 min of unchanging environmental conditions, while group II (A. longifolia and Q. suber) approached steady state but did not fully reach it (Figure 3).

Given the good agreement between modeling results and observations, the time constant (Equations 2, 4, and 7) described the approach of δ_E to isotopic steady state following environmental perturbation rather well (Figure 3). Species-specific differences in the time constant (τ) could be very well described by the term W/E and W/g_s (R² = 0.98 and 0.95; p < 0.01, Figures 4A,B). Moreover, although showing a weaker relationship, species specific differences in τ were also significantly related to the term W/stomatal density (providing a rough estimate of g_max; R² = 0.3, p < 0.05, Figure 4C).

In addition to the quantification of the isotopic non-steady-state effect of δ_E in response to changes in rH, the isoflux of the transpired vapor was determined for all three stages of the experiment. Here, transpiration isofluxes were calculated relative to xylem/source water δ¹⁸ O because despite equal irrigation source for all species, averaged soil water δ¹⁸ O differed between species, most probably due to evaporative enrichment of soils without dense soil cover, leading to different water residence times in the soil (data not shown).

As expected, isofluxes were around zero during the initial stage, as plants were transpiring in or near to isotopic steady state (i.e., Δ_E ≈ 0; Figure 5A). Following step changes, mean isofluxes over the measurement duration after changes in rH differed between species (Figures 5B,C), varying between 3.2 and 12.9‰ mmol m⁻² s⁻¹ after the drop in rH and between 0.8 and 8.8‰ mmol m⁻² s⁻¹ after increasing rH. Again, significant relationships were found between mean isofluxes during the second and third experimental stage and the terms E × W, g_s × W and stomatal density × W (Figure 6). Finally, we modeled the temporal development of the transpirational isoflux of Plantago lanceolata assuming 2 different values for E and W. Original values of E and W of P. lanceolata were therefore divided by 4 (Figure 7), representing the range for E and W observed in this study across all species (Figures 1, 2). The impact of changing E and W on δ_E (Figure 7A) are as can be expected from theory (i.e., Equations 2 and 4). However, the impact of changes in E and W on the transpirational isoflux is strongly depending on the time of observation/integration (Figure 7B). Up to 10 min following the drop in rH, the smallest isoflux can be observed for the model run with small E and the run with small E and W (on average 6.3 and 6.1‰ mmol⁻² m⁻² s⁻¹ compared to 21.5‰ mmol⁻² m⁻² s⁻¹ considering high E and W, see also Table 3). However, the longer the time of integration is, the higher becomes the relative influence of W. Integrating over the full 120 min before the next step change, mean isofluxes were smallest considering small E and small W and E (2.3 and 2.6 ‰ mmol⁻² m⁻² s⁻¹, respectively, see Table 3).

Oxygen isotope signatures of transpiration are increasingly used to partition evapotranspiration differing at spatial (plant to global) and temporal (minute to annual) scales. The present results emphasize that assuming plant transpiration to be in isotopic steady state can have a large impact i.e. for ecosystem partitioning studies (Yakir and Sternberg, 2000; Yepez et al., 2003; Williams et al., 2004; Zhang et al., 2011; Dubbert et al., 2013, 2014b; Hu et al., 2014). Current approaches modeling the isotopic signature of leaf water allowing for isotopic non-steady state are often based on the assumptions of one unenriched water pool in a leaf (i.e., vein water) and enrichment occurring at the evaporative sites within the mesophyll. They predict that the time needed for the isotopic signature of leaf water at the evaporating sites (δ_e) to approach isotopic steady state under stable environmental conditions is dependent on the time constant (τ; Dongmann et al., 1974; Farquhar and Cernusak, 2005; Cuntz et al., 2007). The deviation of the isotopic signature of transpired vapor from steady state (Δ_E) is tightly linked to Δ_e, however deviations of δ_e from isotopic steady state are amplified dependent on relative humidity (Δ_E = (Δ_e − Δ_C)/(α+αk(1 − h)); see Dubbert et al., 2014a). Consequently, analyzing the temporal development of δ_E and its deviation from isotopic steady-state (Δ_E), one needs to separate environmental from leaf trait related drivers, g_s (E under cuvette scenarios) and W that are influencing δ_E via the leaf water time constant term in equation 4, i.e., δ_e.

Responses to changes in rH were analyzed regarding two distinct aspects: the immediate response following a step-change in rH and the subsequent approach to isotopic steady state under stable environmental conditions. While changes in rH directly influence δ_E, changes in other environmental parameters like variations in PPFD or c_a, as analyzed in previous studies (Simonin et al., 2013), indirectly influence δ_E via their control on g_s and E. Notably, the magnitude of the direct response of δ_E to step changes did not differ between species. Considering Δ_E to be the deviation of Δ_e from Δ_C amplified by 1-h, it becomes apparent the that the magnitude of the immediate response after a step change is directly related to the magnitude of the step-change in rH and its influence on 1-h as well as δ_C (see Equation 3).

In contrast, we found significant differences in the approach to isotopic steady state subsequent to changes in rH. Notably, the Dongmann-style model (see Equation 4) as used here predicted the observed temporal dynamics in δ_E following step-changes in rH reasonably well (Figure 3; see also Simonin et al., 2013). We can clearly see that τ can be nicely predicted by E or g_s and leaf water content (Farquhar and Cernusak, 2005; Song et al., 2015). Interestingly, neither of these two parameters alone was able to predict changes in τ and hence non-steady-state δ_E pooled across all species. The species-specific reaction to the decrease of rH led to significant differences in E and hence τ before and after changes in rH. Particularly P. abies and O. triangularis strongly decreased g_s and hence E following the step change from 60 to 30% rH. Therefore, the responsiveness of the stomatal aperture to changing environmental conditions seems to be an importing factor characterizing τ and thus non-steady-state effects of δ_E. However, the still significant relationship between τ and W/stomatal density, seem worthwhile for further investigation, potentially providing very easy to quantify means for a priory assumptions of the impact of isotopic non-steady state in ecosystem with distinct plant functional groups. This could be beneficial especially for approaches aiming at partitioning evapotranspiration by use of stable isotopes. We believe that the possible impact of assuming transpiration to be in isotopic steady state on the final partitioning approach could be thus estimated a priori.

Song et al. (2015) demonstrated that in a cuvette scenario, τ is not influenced by the gross flux of water (g_s× w_i) but rather by the net flux of water from the leaf (E) as well as the impact of E on cuvette vapor. Consequently they adapted the model of τ by Farquhar and Cernusak (2005). This “leaf cuvette effect” leads to higher τ values than expected for open field scenarios. In fact, with w_in being dry air, Song et al. (2015) showed that τ in a cuvette is twice as high as under field conditions. While this effect will be smaller when w_in is not dry air as in this study (see Figure S3), it is still very important to consider this effect in studies involving cuvette-based measurements. Moreover, it should also be kept in mind that τ under field conditions will be smaller and hence the total impact of non-steady-state transpiration following changes in environmental conditions will be somewhat lower than predicted by cuvette measurements.

Oxygen isotope signatures of transpiration are used to trace water flows through ecosystems, but also to assess the impact of plant transpiration on atmospheric water vapor (Xiao et al., 2010, 2012; Lee et al., 2012). In this regard, it is important to understand the impact of environmental and plant physiological controls on the impact of isotopic non-steady-state effects of δ_E on atmospheric vapor. The impact of isotopic non-steady-state δ_E on atmospheric vapor (isoforcing), however, is not only dependent on the δ_E signature, but on the product of E and δ_E (i.e., the isoflux of the transpired vapor). Furthermore, the actual isoforcing (Lee et al., 2009) of transpired vapor under natural conditions will also be influenced by changes in the H₂O concentration and isotopic signature of atmospheric vapor.

Dubbert et al. (2014a) showed that under natural field conditions the transpirational isoforcing of Mediterranean Q. suber trees significantly deviated from the steady-state assumption at the least on short time scales (i.e., less than 24 h). In this experimental study, we calculated isofluxes as a deviation from the steady-state isoflux (i.e., –EΔ_E). Isofluxes of transpired vapor strongly differed between species and differed significantly from isotopic steady-state isofluxes after changes in rH, highlighting the potential forcing of isotopic non-steady-state transpiration on atmospheric vapor (Figure 6). Previous studies suggested, that isotopic non-steady-state effects of leaf transpiration will have a significant impact on atmospheric vapor, when the time constant (τ) is long, i.e., low E (or g_s in non-cuvette scenarios) and/or high W (Lai et al., 2006; Simonin et al., 2013; Dubbert et al., 2014a). Here, species-specific differences in τ and hence in δ_E can originate from either differences in E or W, while the observed differences in transpiration rate (E) are not influenced by W. The impact of W on the non-steady-state transpirational isoflux can be clearly predicted: high W will lead to a higher time constant and hence to a slower return to isotopic steady-state transpiration. This is more complex for E because higher E will lead to a smaller time constant and hence a faster return to isotopic steady-state transpiration, but also directly increases the term −EΔ_E (under a given rH).

Figure 7 clearly shows that within the first 10–30 min following a step change the direct impact of E on −EΔ_E dominates the development of the isotopic non-steady-state isoflux. Differences in W and its influence on τ and hence δ_E are getting increasingly important though if the time of integration becomes greater than 30 min (Figure 7B). Hence, if variation in the time constant is driven by changes in E, it has a bigger short-term effect on the resulting isoflux than if the source of variation is W, due to the direct effect of E on −EΔ_E. Similarly, a small time constant given by high E and high W results in a larger isoflux than a large time constant achieved via large W and small E, at least for time intervals from 10 to 30 min following environmental perturbation. Only in case of longer time periods with stable environmental conditions (>60 min), changes in W do become increasingly important for the non-steady-state isoflux, and it is questionable whether they are relevant to address under natural conditions.

Concluding, our results demonstrate the usefulness of coupled gas-exchange laser spectrometer set-ups to analyze water isotope fractionation processes at the leaf scale. The overall very good agreement between observations and modeling results in seven species with distinct leaf traits supports current modeling approaches regarding leaf water isotopic enrichment. Our results provided for the first time a species survey on the impact of distinct leaf traits on the temporal dynamics of the deviation of δ_E from isotopic steady state. We demonstrate that isoforcing of δ_E on atmospheric vapor is strongly driven by control of E over −EΔ_E, even though high E also leads to small time constants. By contrast, impact of W on the time constant only influences the non-steady-state isoflux on time scales probably not relevant under natural conditions. The significant relationships between E, g_s, and even stomatal density (as a measure of g_max) and W on the one hand and τ and the isotopic non-steady-state isoflux on the other hand are quite promising. Particularly morphological leaf traits are relatively easy to sample and measure under field conditions, and their use as a priori information of τ and isoflux deviation from the steady-state assumption would be highly useful. Transpirational isotopes provide the basis for environmental tracer/paleo-climatic studies and are used as tracers from leaf to atmospheric scale. Hence, our findings have wide reaching consequences across large temporal and spatial scales.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.