Introduction

Stable isotopes of water (δ2H and δ18O) provide effective tools for tracing water fluxes within the soil–plant–atmosphere continuum (SPAC) and have been widely used for evapotranspiration (ET) partitioning (Moreira et al. 1997; Yepez et al. 2003; Williams et al. 2004; Wang et al. 2010; Wang et al. 2015; Xiao et al. 2018), as well as investigating the dynamics of water exchange at the interfaces (e.g., ground soil surfaces and leaf surfaces) between the land surface and atmosphere (Dongmann et al. 1974; Farquhar and Cernusak 2005; Lai et al. 2005; Cernusak et al. 2016). Plant transpiration (T) is the largest component of the terrestrial-atmospheric water/energy flux, accounting for somewhere between 50 and 90% of the annual global land surface water flux (Good et al. 2015; Jasechko et al. 2013; Maxwell and Condon 2016; Lian et al. 2018). Therefore, vegetation exerts strong feedback effects on the climate by modifying the energy, momentum, and hydrologic balances of the land surface (Twine et al. 2000; Arora 2002; Fatichi et al. 2016). Climatic feedback effects resulting from vegetation are mainly attributable to the physiological properties of vegetation, such as the leaf area index (LAI), stomatal resistance, rooting depth, albedo, and surface roughness and its effect on soil moisture (Arora 2002). The seasonal rhythm of ecosystem transpiration emerges from interactions among the climate, plant functional traits, and ecosystem structure. An increasing number of studies report a close relationship between the LAI and transpiration fraction (T/ET), indicating that leaf development dominates the water flux composition (e.g., Hu et al. 2009; Wang and Yamanaka 2014; Wang et al. 2014; Wei et al. 2015, 2017; Wang et al. 2018b). Higher correlations between the LAI and evapotranspiration ratio, lET/(Rn – G), where lET is the latent heat flux, Rn is the net radiation, and G is the ground heat flux, were also reported, illustrating that the energy flux composition is largely regulated by plant growth (e.g., Wilson et al. 2002; Li et al. 2005; Hammerle et al. 2008, Zhao et al. 2017). However, few isotopic investigations within the SPAC have been conducted to improve the understanding of the interactions between plant/leaf development and the isotopic dynamics of water fluxes (ET, T, and soil evaporation, E) along with the isotopic dynamics at the plant-atmosphere interface. In addition, the biophysical effects of vegetation on the climate have been addressed in a number of studies investigating the effects of deforestation or clear-cutting (Charney 1975; Dickinson and Henderson-Sellers 1988; Lean and Rowntree 1997; Zhou et al. 2015), while only a few investigations have been conducted to examine the biophysical effects of mowing from a water-isotopic perspective in grassland ecosystems.

The isotopic composition of ET (δET) provides valuable information for tracing the atmospheric water cycle (e.g., water sources and flow paths) and for diagnosing plant activities and their roles in the hydrological cycle. Currently, indirect techniques, including the Keeling plot approach (Good et al. 2014; Keeling 1958; Yakir and Wang 1996; Wang et al. 2010), flux-gradient method (Good et al. 2012; Hu et al. 2014; Wen et al. 2016; Wang et al. 2016), and direct eddy covariance measurements (Jelka et al. 2019), are widely used for estimating δET. As is clear, the isotopic composition of water vapor (δV) is an essential variable that must be determined to derive δET using the abovementioned methods. The traditional cold-trap methods for measuring δV are accurate but suffer from coarse time resolutions (Yakir and Wang 1996; Yakir and Sternberg 2000). Rapid developments in laser techniques have largely overcome this methodological constraint by enabling the online, near-continuous quantification of water vapor isotopes and, by extension, δET (Lee et al. 2005, Wen et al. 2008, Griffis 2013). However, the Keeling plot, gradient-based, and eddy covariance methods also involve many uncertainties (Good et al. 2012) and sometimes fail to reflect the true values of δET, particularly during periods of poor information retrieval or top-down diffusion (entrainment, which represents another end member). Nevertheless, these approaches focus on the measurement of δET, which lacks crucial information on the components of ET, hindering our understanding and ability to trace water transfers through the SPAC. The term δET represents the comprehensive result of the interactions among the climate, soil, and biosphere. Many land-surface models with isotopic tracers have been established (e.g., Braud et al. 2005; Henderson-Sellers et al. 2006; Riley et al. 2002; Yoshimura et al. 2006; Xiao et al. 2010). To date, there are still many challenges involved in capturing the isotopic signals of soil evaporation, transpiration, and their mixing process (Griffis 2013). The SPAC model is relatively simple but is still capable of incorporating separate water/energy controls on vegetation and soil as well as coupling isotopic fractionation with H2O exchange and the mixing processes in terrestrial ecosystems (Wang et al. 2016).

Isotopic variations in water fluxes and the isotopic enrichment process at the plant-atmosphere interface involve many factors that occur simultaneously but are usually studied in isolation (e.g., Wang et al. 2010; Xiao et al. 2012; Wen et al. 2016). Many models exist for simulating the isotopic composition of bulk leaf water (δL,b), such as the isotopic steady-state (ISS) model (Craig and Gordon 1965), the non-steady-state (NSS) model (Dongmann et al. 1974; Helliker and Ehleringer 2000; Cernusak et al. 2003; Barbour et al. 2000, 2004), and the NSS model with the Péclet effect (Farquhar and Cernusak 2005)). Additionally, simple two-pool mixing models (Leaney et al. 1985; Song et al. 2015) and models based on the Péclet effect (Farquhar et al. 1993; Cuntz et al. 2007) have been used to simulate δL,b and the isotopic distribution within a single leaf. Although many studies have been conducted to evaluate these models under natural conditions (Barnard et al. 2007; Kahmen et al. 2008; Xiao et al. 2012; Dubbert et al. 2014), only a few studies have included model-measurement comparisons for δET and δL,b at an ecosystem scale (Lee et al. 2007; Xiao et al. 2010). To date, comprehensive investigations of the isotopic dynamics of water fluxes (ET, T, and E) along with isotopic enrichment processes at the plant-atmosphere interface are still missing.

Land-surface models are promising tools for integrating isotopic fractionation processes with separate E and T in terrestrial ecosystems and have the advantages of long-term assessments of ET partitioning (Wang et al. 2015, 2016; Wei et al. 2018b) and isotopic enrichment at vegetation/soil-atmosphere interfaces across multiple temporal scales (Sprenger et al. 2016; Dubbert and Werner 2019; Wang et al. 2018a). Isotopic measurements (e.g., the Keeling plot and gradient-based approaches for δET, field sampling with vacuum extractions of leaf water in the laboratory to obtain δL,b) combined with modeling approaches have been demonstrated to be reliable in most cases (Wei et al. 2018a). In this study, isotopic measurements of various water pools (e.g., soil, plants, and the atmosphere) are combined with an isotope-enabled SPAC model (Iso-SPAC) to assess the seasonality of the isotopic compositions of soil evaporation (δE), transpiration (δT), evapotranspiration (δET), and bulk leaf water (δL,b) and their relations with vegetation dynamics (e.g., LAI development). The objectives of the present study were to (1) evaluate the seasonality of the isotope compositions of water fluxes (δET, δT, and δE) and bulk leaf water (δL,b) in a grassland ecosystem and (2) investigate the factors controlling seasonal variations in δET.

Materials and methods

Study site

We conducted field investigations at the Center for Research in Isotopes and Environmental Dynamics (CRiED) experimental site at the University of Tsukuba, Japan (36.1° N, 140.1° E), 27 m above sea level. Based on the observations collected between 1981 and 2005 at this site, the mean annual air temperature and the mean annual precipitation were approximately 14.1 °C and 1159 mm, respectively. The dominant plant species are Solidago altssaima, Miscanthus sinensis, and Imperata cylindrical. The study site, a flat circular field with a diameter of 160 m, is surrounded by pine forests and lawns. Mowing was performed on DOY 213 and 214 (August 1st and 2nd) in 2011. The soil surface was covered by litter made up of mown and dead plants from previous years. The top of the soil consists of a rich volcanic ash layer approximately 2 m thick, underlain by a clay layer.

Continuous micrometeorological and eddy covariance measurements of plant dynamics

Routine micrometeorological (e.g., downward solar radiation, air temperature, relative humidity, wind speed, and soil heat flux) and eddy covariance measurements (sensible heat flux and latent heat flux) were conducted in the CRiED experimental fields (Hiyama et al. 1993). More details of the instruments and the database used in this study are described in Ma et al. (2018). The fresh grass leaves from three randomly selected 1*1 m quadrats were harvested and LAI was measured by an automatic area meter in laboratory (AAM-7, Hayashi Denko, Tokyo, Japan) every week or every other week. In the periods without measurements, the LAI values were estimated by curve fitting: before mowing, LAI = 3.97 ln (DOY) - 18.70 (R2 = 0.99), and after mowing, LAI = 6.15 ln (DOY) - 32.85 (R2 = 0.95). At the same time, the total leaf water weight per unit ground surface (W, kg m−2) was measured by weighing fresh and dried leaves before and after the LAI measurements, respectively, were conducted on each day. The vegetation height (Zp) was measured every 7 to 15 days throughout the study period. A close relationship was found between the LAI and W, with a linear function of W = 320.0 LAI – 46.05 (R2 = 0.93).

Isotopic measurements

Field investigations of various water pools (water vapor, soil water, plant stem water, and leaf water) were conducted between April and September 2011. The isotopic measurements of water vapor were conducted by the traditional cold-trap method during the period from mid-day to early afternoon (1300–1500 JST) every week for laboratory analysis. The water vapor sampling system used in this study was described by Yamanaka and Tsunakawa (2007). The isotopic sampling of atmospheric water vapor combined with air temperature and air humidity measurements were conducted at three levels, and the sampling heights were adjusted with the development of canopy. In the early growing season (DOY 117~DOY 124) and a period after mowing (DOY 215~DOY 236), the three sampling heights were160 cm, 50 cm, and 5 cm, respectively. In the rapid growing season, the three sampling heights were 200 cm, 70 cm, and 5 cm, respectively. At the same time, leaf water, stem water, and soil water samples were collected. Leaf samples from different species (Solidago altssaima, Miscanthus sinensis, and Imperata cylindrica) were collected separately. Leaf and stem water samples were extracted in triplicate in the laboratory by cryogenic vacuum distillation (Dawson and Ehleringer 1993; West et al. 2006). Soil water samples were collected at five depths (5 cm, 10 cm, 20 cm, 40 cm, and 80 cm) on each sampling day using suction lysimeters with ceramic cups attached on their tips (DIK-8390, Daiki, Japan) and 100-ml syringes (Zhao et al. 2013). Precipitation samples were collected every 2–3 days using a high-density polyethylene tank with a funnel.

The isotopic ratios of the water samples were analyzed at the CRiED. The hydrogen and oxygen stable isotopic compositions of the water samples were determined using a liquid water isotope analyzer (L1102-i, Picarro, Santa Clara, CA, USA). The isotopic ratios of 18O/16O and D/H were expressed as δ18O and δD, respectively, relative to Vienna Standard Mean Ocean Water. The analytical errors were 0.1‰ for δ18O and 1‰ for δD (Yamanaka and Onda 2011).

Processed-based modeling approach

The Iso-SPAC model developed by Wang et al. (2015) was used in this study, and the details of the model and relevant equations are provided in Method S1. Briefly, Iso-SPAC is an energy-balanced two-source model (soil surface and canopy interface) used to simulate day-to-day energy (Rn, lET, H, and G), ET, E, and T fluxes (Wang and Yamanaka 2014). Based on the outputs of resistance, surface temperature, and water fluxes at the soil surface and canopy interface, an isotopic budget is coupled by integrating the isotopic fractionation processes with E and T processes. This model was selected for several reasons. First, this model treats water fluxes and the isotopic compositions of plant transpiration and soil evaporation separately and simultaneously integrates them to determine the isotopic composition of the ET flux. Second, the Iso-SPAC model, which includes three submodels with a range of complexities, namely, the ISS (Craig and Gordon 1965), NSS-D74 (Dongmann et al. 1974), and NSS-F05 (Farquhar and Cernusak 2005), was previously used at the leaf scale. The submodules were adopted here for the δL,b and δT determinations. The Iso-SPAC model can be validated not only by physical components (e.g., ET, T, E, leaf temperature, and energy flux) but also by isotopic components (δET, δT, δE, δL,b), which are expected to be more robust and reliable. Third, the radiation/energy balances in both the vegetative canopy and at the ground surface were solved by the Newton-Raphson scheme, enabling us to estimate the vegetation canopy temperature (TL) and ground temperature (TG) separately. The scheme can easily integrate isotopic fractionation processes with H2O exchanges, and it has the advantage of enabling the long-term assessment of ET compositions as well as isotopic enrichment at the plant-atmosphere interface. Based on biophysical measurements (e.g., radiation, air temperature, humidity, LAI) and field observations of the isotopic compositions of soil water, stem water, and water vapor, the Iso-SPAC model can output the seasonal variations in δT, δE, δET, and δL,b (Table S1). The root mean square error (RMSD), the I index (Willmott et al. 1985), and R2 (R is the correlation coefficient) were used to evaluate the model performance.

Determination of the transpiration fraction and evapotranspiration ratio

The transpiration fraction (T/ET) was determined by both the Iso-SPAC model (Method S1, heat and water flux part of the Iso-SPAC model) and by direct isotope measurements (Method S2). The evapotranspiration ratio (lET/(Rn – G)) was determined by both eddy covariance system measurements in the field and simulations using the Iso-SPAC model (Method S1).

Results

Observed seasonal variations in meteorology, flux, and leaf development

The air temperature increased progressively from DOY 110 to DOY 213, with a marked increase after grass cutting occurred during DOY 213–214, and then decreased gradually. The relative humidity exhibited much larger variations, with an obvious decrease after grass cutting (Fig. 1a). Table 1 summarizes the observed monthly meteorological factors. The latent heat flux (lET) dominated the energy flux during the growing season. With grass cutting, the lET dropped sharply, and the sensible heat flux (H) increased correspondingly, which illustrated that vegetation growth has obvious effects on energy flux partitioning (Fig. 1b). The observed seasonal variations in volumetric soil water and the LAI are shown in Fig. 1c. The soil water content (0–20 cm) varied substantially during the growing season. The mean LAI increased from 0.16 to 2.49 m2 m−2 before grass cutting. After grass cutting, the plant LAI increased again from 0.76 to 2.06 m2 m−2.

Fig. 1
figure 1

Seasonal variations in the (a) air temperature (Ta) and relative humidity (ha); (b) energy fluxes (net radiation (Rn), sensible heat flux (H), latent heat flux (lET), and ground heat flux (G)) measured by the EC system; and (c) volumetric soil water (%) and leaf area index (LAI). LAI-a and LAI-b represent the stages of plant growth before and after grass cutting. Dotted gray lines show the days of mowing

Full size image
Table 1 Summary of monthly meteorological and vegetation properties
Full size table

Observed isotopic signatures of water pools and fluxes during the growing season

A dual isotope-based approach (δ18O and δ2H) was used to examine the linkages between various water pools and water fluxes. From April to October 2011, water isotopic signals were monitored in five soil layers, groundwater, plant leaves, and stems from three species, rainfall, and water vapor from three layers, and δET was estimated by the Keeling plot approach for a grassland ecosystem (Fig. 2). The linear fits of δ18O and δ2H in local rainfall were very close to the GMWL (global meteoric water line). The δL,b values for the three species showed a wide range of isotopic fractionation and high values due to transpiration. The isotopic signal of the stem water of each species was close to those of precipitation and deep soil water. The isotopic composition of surface soil water (5 cm) varied over time because of precipitation and evaporation. Fluctuations in deep soil water (20, 40, 60, and 80 cm) decreased with depth and remained small below 40 cm. The δV values at the three levels were similar and showed low isotopic values. The δET values estimated by the Keeling plot approach varied from −10.71 (−65.55) to −20.24 (−137.18) ‰ for 18O/16O (D/H) during the growing season; these results are summarized in Method S2.

Fig. 2
figure 2

Dual-isotope plot for all samples (soil water, groundwater, plant leaf and stem water, rainfall, water vapor, and evapotranspiration) collected, analyzed, and estimated in a grassland ecosystem. Soil water collected from five layers (5cm, 10cm, 20cm, 40, 80cm). Leaf and stem water were collected from the top, middle, and base of Miscanthus sinensis, as well as the whole Solidago altissima, and Imperata cylindrical. Water vapor collected from three heights, and isotope composition in evapotranspiration (ET) estimated by a Keeling plot approach. Isotopes were collected during the growing season, from DOY 110 to 300. GMWL is the global meteoric water line (solid line, δ2H=10 + 8 δ18O). The dashed line is a linear fit of all rainfall samples collected during the growing season (dashed line, δ2H=10.2 + 8.2 δ18O)

Full size image

Model performance

The model performance in representing energy fluxes, leaf temperature, and canopy conductance is summarized in Table S2. Very good agreement between the observations and models can be found for all energy fluxes, with small RMSD (14.7–36.8 W m−2) values and high I indices (0.56–0.92). Similar results were obtained for the canopy leaf temperature and leaf stomatal conductance, with RMSD values of 2.08 °C and 1.97 mm s−1 and I indices of 0.99 and 0.79, respectively. In addition, the model performances in a range of isotopic scenarios (e.g., ISS and NSS) can be assessed from the observed isotopic signatures (δET, δL,b). Field sampling and measurements of leaf-water isotopic compositions for each species and the representative canopy-scale bulk leaf water measurements are summarized in Table S3. The agreement between the measured (Table S3 and Table S4) and simulated values demonstrates the accuracy of the proposed model for reconstructing day-to-day variations in both δET and δL,b (Table 2). Among the four submodules, ISS had the best performance in reconstructing day-to-day variations in δET and δL,b at the ecosystem scale, with smaller RMSD (1.69 and 2.08, respectively) values and high I indices (0.91 and 0.40, respectively). The results illustrated that ISS can be used to obtain a reasonable assumption for the selected grassland ecosystem because of its fast transpiration and rapid turnover of water in leaves, resulting in a transpiration signature similar to that of the source water.

Table 2 Model performance statistics for the isotope compositions of evapotranspiration flux (δET) and bulk leaf water (δL,b)
Full size table

Reasonable agreement (R2 =0.45 and 0.30 for transpiration fraction and evapotranspiration ratio, respectively) was obtained between the observed and modeled mean daily ratios of both the transpiration fraction (T/ET) and evapotranspiration ratio (lET/Rn – G) (Fig. 3). The modeled evapotranspiration ratio agreed reasonably well (R2 =0.30) with the ratios observed by the EC system, but the modeled values were mostly underestimated when the measured ratio was over 0.7. The differences between the isotopic measurements and numerical simulations ranged from 1 to 26% during the growing season, with an average of 6%. Considering the uncertainty of isotope measurements, the reasonable consistency between the two methods indicates the reliability of the water flux partitioning results. Therefore, it is expected that the partitioning results are robust and reliable for quantifying the water/energy flux components with high LAI (e.g., LAI >1) values. The simulated T (E) flux ranged from 0 to 7.24 mm day−1 (0 to 1.65 mm day−1), with an average value of 2.05 ± 1.56 mm day−1 (0.52 ± 0.37 mm day−1) during the growing season.

Fig. 3
figure 3

Comparison of measured and modeled daily mean flux ratios; T is transpiration, ET is total evapotranspiration, lET is latent heat flux, Rn is net radiation, and G is soil heat flux

Full size image

Temporal dynamics of the isotopic compositions of water fluxes and bulk leaf water

Figure 4 shows the seasonal variations in the LAI, δET, δT, δE, and δL,b obtained by both field measurements and the Iso-SPAC model outputs. The δT values exhibited the smallest seasonal variations, whereas δE showed the largest variations. At the ecosystem scale, if the transpiration process was at an isotopic steady state (ISS) at a daily scale, then δT would equal that of the water used by plants (Flanagan et al. 1991). In contrast, the water evaporated from soil surfaces was strongly fractionated and depleted in heavy isotopes (Craig and Gordon 1965; Gat 1996). There was a significant difference between δT and δE. The day-to-day δET values increased with leaf development, showed a marked decrease during grass cutting, and increased gradually with plant regrowth. The daily δL,b values decreased gradually with leaf development during the fast growing period and showed an increase after grass cutting.

Fig. 4
figure 4

Seasonal variations in the leaf area index (LAI), the oxygen isotopic composition of measured water vapor (δV), the evapotranspiration (δET) values estimated by the Keeling plot approach and the Iso-SPAC model, and the oxygen isotopic compositions of transpiration (δT) and soil evaporation (δE) simulated by the Iso-SPAC model. The canopy-scale oxygen isotopic composition of bulk leaf water (δL,b) was simulated by the Iso-SPAC model, and the measured δL,b values were obtained during the growing season. Error bars represent the standard deviations of the leaf samples within and among species

Full size image

Relationships between leaf development and the isotopic signatures of water fluxes and canopy leaf water

Figure 5a shows the relationship between the LAI and δET obtained by Keeling plot measurements and the Iso-SPAC model output with ISS, which was chosen because of its superior performance. A good logarithmic relationship was found between the LAI and δET, with an R2 of 0.83 for the modeled dataset. The Keeling plot approach showed a similar relationship. Figure 5b shows the relationship between the LAI and isotopic signals of canopy leaf water (δL,b) obtained from both field measurements and the Iso-SPAC model simulation. The field-measured δL,b had a similar but smaller slope compared with that of the Iso-SPAC model-simulated value, which may be caused by the coarse time resolution of the field-measured data. With leaf development, δET gradually became enriched and approached δT (Fig. 5a), whereas δL,b lost isotopic enrichment (Fig. 5b). These results illustrate that leaf development greatly affects seasonal variations in δET and δL,b.

Fig. 5
figure 5

Relationships between the leaf area index (LAI) and (a) the oxygen isotopic compositions of ET (δET) obtained by both the Keeling plot approach and the simulation using the Iso-SPAC model incorporating an isotopic steady state (ISS) for the transpiration flux and (b) bulk leaf water (δL,b) at the canopy scale obtained by field measurements and numerical modeling. The cross symbols represent the isotopic compositions of the transpiration flux (δT)

Full size image

Discussion

Biotic control of seasonal variations in δET

Identifying the factors that control δET is vital for tracing the atmospheric water cycle. However, this remains a great challenge because δET is interactively affected by multiple factors, including the climate (e.g., air temperature, relative humidity, wind speed, solar radiation, the isotopic compositions of rainfall, and water vapor), soil (e.g., soil moisture, the isotopic composition of soil water, soil texture, and temperature), and the biosphere (e.g., LAI, Zp, canopy structure, the isotopic composition of source water used by plants). Combining isotopic measurements (e.g., Keeling plots, gradient-based approach) with a modeling approach for estimating δET is more reliable than using a single approach in most cases (Wang et al. 2015). The agreement between the model simulations and observations demonstrates that using the Iso-SPAC model with ISS to determine T can reproduce reasonable day-to-day variations in all surface energy balance components, as well as δET. Sensitivity analyses of both the physical flux part and the isotope part suggested that T/ET is relatively insensitive to uncertainties or errors in the assigned model parameters and measured input variables, which illustrates the robustness of flux partitioning (Wang et al. 2015). Few studies have investigated the dominant factors controlling δET. The results obtained here reveal a good nonlinear relationship between the LAI and δET, as obtained by both the Iso-SPAC simulation and Keeling plot measurements. As shown in Fig. 4, the isotopic composition of rainfall (δrainfall) affects the δET values through the regulation of soil water dynamics (e.g., the isotopic composition of soil water, δS) and plant uptake processes (e.g., the isotopic composition of stem water, δX). The δS and δX, terms, when used as origin sources during soil evaporation and transpiration analyses were well represented in Craig and Gordon (1965) as well as in ISS and non-steady state models (Method S1). Generally, dynamic variations in δET can be explained by δE, δT and their mixing ratio (e.g., T/ET or E/ET) measured at daily intervals, as shown in Eq. S27. This study found that T/ET has a close relationship with the LAI (Fig. S1), but no significant relationship was found for δE and δT during the growing season (Fig. S2), which indicates that δrainfall has less of an effect on δET. Therefore, the LAI exerts a dominant effect on the day-to-day variations in δET within the lower (δE, range from −24.76 to −86.75‰) and upper (δT, range from −6.28 to −10.57‰) limits by controlling T/ET. Early in the growing season, δET showed isotopic depletion, indicating that E dominated the total ET flux, and δET was subsequently more affected by δE. With ongoing leaf development, both T/ET and lET/(Rn-G) increased, illustrating the increasing biotic control on the ecosystem water/energy flux partitioning (Fig. S1). As the LAI increased, more water was lost by T, and δET became enriched due to the increased T and was consequently better reflected by δT.

Towards a better understanding of ecosystem behavior using isotopic tracers

Figure 6 shows a set of box-and-whisker plots presenting the minimum, maximum, median, lower quartile, and upper quartile of all water pools and fluxes in the studied temperate grassland ecosystem. The seasonal variations in the isotopic composition of soil water decreased in amplitude with soil depth. The δs value at a depth of 80 cm hardly showed any change. In contrast, the δs values at shallower depths varied over time with isotopic inputs from precipitation and soil evaporation. Under field conditions, the leaf-water isotopic composition showed large deviations both within and among species, even for single leaves (Table S3). For canopy leaf water, a previous study showed that the isotope fractionation factors acting at the leaf scale can behave differently at the canopy scale (Lee et al. 2009), and a distinction should be made between leaf- and canopy/ecosystem-scale processes. By incorporating the environmental conditions outside the plant community in the atmospheric surface layer, the appropriate modeling framework could be applied at the ecosystem scale, reflecting the community behavior and ecosystem properties. Therefore, despite the heterogeneity caused by the biodiversity and position of leaves, δL,b can be measured and predicted for multispecies grassland ecosystems at a daily time scale (Fig. 4). A comparison between the field measurements and the simulations performed by the four submodels indicated that ISS can be achieved at the ecosystem scale (Table 2). Isotope measurements combined with Iso-SPAC modeling are a promising tool that can be used to reproduce day-to-day variations in all surface energy balance components, as well as δET and δL,b. The measured and simulated results revealed that leaf development had a significant influence on day-to-day variations in δET (P<0.01) but a small influence on δL,b (P = 0.35) at the canopy scale (Fig. 5).

Fig. 6
figure 6

Box-and-whisker plots showing the minimum, maximum, median, lower quartile, and upper quartile δ18O values of soil water, groundwater, leaf water, and the stem water of the species Miscanthus sinensis measured from the top, middle, and base of the plants (A), Solidago altissima (B), and Imperata cylindrical (C), rainwater, water vapor, the evapotranspiration fluxes estimated both by the Keeling plot approach (δET-K) and by the Iso-SPAC model (δET-S), and the soil evaporation flux (δE) simulated by the Iso-SPAC model. Here, S-5, S-10, S-20, S-40, and S-80 indicate soil water for species at depths of 5, 10, 20, 40, 60, and 80 cm, respectively. GW is groundwater; LA1, LA2, and LA3 indicate leaf water for the species Miscanthus sinensis measured from the top, middle, and base, respectively; LB4 and LC5 denote leaf water collected from Solidago altissima and Imperata cylindrical, respectively; XA1, XA2, and XA3 indicate leaf water for the species Miscanthus sinensis measured from the top, middle, and base of the plant; XB4 and XC5 denote leaf water collected from Solidago altissima and Imperata cylindrical, respectively; and V-200, V-50, and V-5 indicate water vapor measured at the top (200 or 170 cm), middle (70 or 50 cm), and bottom (5 cm) levels, respectively

Full size image

The isotopic composition of soil evaporation, with its much-depleted isotope signature, showed a distinct pattern in the dual-isotope plots (Fig. S3). As shown in Fig. 7, the water was initially isotopically depleted in δET during the initial growing season. Gradually, δET lost its depleted signal with leaf development in the fast-growing season, accompanied by a rapid increase in T/ET, which led to increasingly enriched isotopic signals. There were smaller variations observed in δET during the stable growing season. In a similar pattern, the initially isotopically enriched canopy leaf water lost its fractionation signal, leading to more depleted isotopic signals in the canopy leaf water, which may be explained by the increasing relative humidity in the canopy layer due to increasing amounts of water lost by transpiration (Wang et al. 2018a). Smaller variations in δET were observed during the stable growing season. The results illustrated that canopy leaf water has well-behaved properties that were predictable for a multispecies grassland ecosystem at a daily time scale and emphasized that ecosystem properties (canopy-scale leaf water, ET flux) and behavior (seasonality) can be simulated and traced by many isotopic tracers (e.g., δ2H and δ18O).

Fig. 7
figure 7

Conceptualization of the processes affecting the water-stable isotopic compositions at the land–atmosphere interface during the growing season in a temperate grassland ecosystem. The plus symbol indicates an isotopic fractionation process leading to enrichment in heavy isotopes, the minus symbol represents depletion in heavy isotopes, and the zero symbol is a sign of a non-fractionating processes. The text indicates the labels of the closest arrows (up to two). The white arrows indicate evapotranspiration processes that are dominated by soil evaporation during the initial growing period (ETI), and the green arrows indicate evapotranspiration processes that are dominated by plant transpiration during the fast-growing period (ETF) and stable growing period (ETS). The terms L,bI, L,bF, and L,bs represent the canopy leaf water during the initial, fast, and stable growing periods, respectively

Full size image

Implications of vegetation impacts on atmospheric moisture

Water vapor isotopes are used increasingly often to trace the impacts of vegetation on atmospheric moisture transport from regional to continental scales (e.g., Gat 1996; Yakir and Sternberg 2000). To quantify the impacts of the δ18O values in ET, T, and E (δET, δT, and δE) on the δ18O value of atmospheric water vapor (δV), the ET, T, and E isoforcings, also called isofluxes (IET, IT, and IE), were defined as the products of ET, T, and E flux and the deviation of their isotopic ratios from that of near-surface atmospheric water vapor, respectively (Dansgaard 1953; Huang and Wen 2014). The formulas used to calculate IET, IT, and IE are given as follows:

$$ {\mathrm{I}}_{\mathrm{ET}}=\mathrm{ET}\ast \left({\updelta}_{\mathrm{ET}}-{\updelta}_{\mathrm{V}}\right) $$
(1)
$$ {\mathrm{I}}_{\mathrm{T}}=\mathrm{T}\ast \left({\updelta}_{\mathrm{T}}-{\updelta}_{\mathrm{V}}\right) $$
(2)
$$ {\mathrm{I}}_{\mathrm{E}}=\mathrm{E}\ast \left({\updelta}_{\mathrm{E}}-{\updelta}_{\mathrm{V}}\right) $$
(3)

Combining online observations of δT or δET with isotopic values of atmospheric vapor and land surface models can inform researchers on the interplay of water exchanges between vegetation or soils and the atmosphere (Wei et al. 2016). In a previous study, the local ET acted to increase the δD and δ18O of δV, with isofluxes of 102.5 and 23.50 mmol m−2 s−1 ‰, respectively, in an agricultural ecosystem (Huang and Wen 2014) during the growing season. However, there are few processed-based modeling studies that have examined the impacts of vegetation on atmospheric moisture. Vegetation, through transpiration, impacts atmospheric moisture. The day-to-day variations in water fluxes (ET, T, and E) and isofluxes (IET, IT, and IE) are shown in Fig. 8. The transpiration isoflux (IT), which describes the isotopic imprint of transpiration on atmospheric vapor, always showed positive values ranging from 3.47 to 51.82 mm day−1 ‰, while the evaporation isoflux (IE) was negative, ranging from −46.01 to−0.37 mm day−1 ‰ during the growing season. The results indicate that the δT values with enriched isotopic signatures (from −6.28 to −10.57‰) acted to increase the δ18O of δV (from −11.22 to −19.59‰). In contrast, the δE values with depleted isotopic signatures (from −24.76 to −86.75‰) acted to decrease the δ18O of δV (from −11.218 to −19.59‰). The IET added dynamic effects (from −40.37 to 38.19 mm day−1 ‰) on the δ18O of δV, which depended on the seasonal tradeoff between IT and IE. IET acted to decrease the δ18O of δV during the initial growing period and during the period after grass cutting when soil evaporation dominated the ET flux; IET also acted to increase the δ18O of δV during the transpiration-dominated period. The impacts of vegetation on atmospheric moisture were also well captured by grass mowing in our study, which caused sensible heat flux and air temperature to increase while lET and relative humidity decreased (Fig. 1). The isotopic signals of the ET flux were also depleted due to increased soil evaporation (Fig. 4). The isotopic composition of the bulk leaf water increased because of the decrease in relative humidity (Fig. 4). The results obtained here emphasize that water fluxes and isotopic water flow paths (i.e., transpiration and evaporation) from temperate grasslands to the atmosphere are strongly controlled by the physiological properties of vegetation growth at a seasonal scale.

Fig. 8
figure 8

Seasonal variations in (a) evapotranspiration, transpiration, and soil evaporation and (b) the evapotranspiration, transpiration, and soil evaporation isoforcings (IET, IE, IT) during the growing season

Full size image

In this study, the water fluxes were partitioned into plant transpiration and soil evaporation to discuss their possible effects on atmospheric moisture. Nevertheless, Dubbert and Werner (2019) mentioned that there are still important open questions regarding how different plant functional types modulate water flows to the atmosphere. The results of our study support a general rule of thumb that water vapor from large water storage supplies (e.g., soil evaporation from large soil water pools) has abundant light isotope water supplies, which can support the continuous isotopic fractionation process, characterized by a slow turnover time with very depleted isotopic signatures that act to decrease δV. Water vapor from small water pools (e.g., grass leaves) has limited light isotope water supplies that can impair the isotopic fractionation process, characterized by a fast turnover time, and can reach ISS carrying the same isotopic signal as that of the original water plant uptakes (e.g., δX), which tends to increase the δV. Clearly, in the future, coupling different plant functional types with their water storages and turnover times to partition the water fluxes and couple the isotopic processes across multiple scales (e.g., from plot to regional scales) will be a promising way to improve our understanding of various vegetation feedbacks to climate systems that may buffer or enhance the susceptibility of these ecosystems.

Conclusions

In this study, in situ water isotopic measurements combined with process-based modeling were used to assess ecosystem water/energy flux partitioning, changes in the isotopic compositions of water fluxes (ET, soil evaporation, and transpiration), and canopy-scale leaf water enrichment as well as the relationships between water fluxes and vegetation growth during the growing season in a temperate grassland ecosystem. The seasonal changes in δET could be explained well by a logarithmic function of the LAI. The day-to-day variations in T/ET and the proportion of latent heat flux to available energy could also be represented well by a logarithmic function of the LAI. These results demonstrate that leaf development controlled the isotopic seasonal variations in water fluxes by regulating the flux compositions. With the progress of leaf development, an increasing amount of water is lost by transpiration, leading to “cooling effects” and enhanced relative humidity at the canopy–atmosphere interface, which subsequently results in the decreased isotopic enrichment of canopy leaves. Transpiration/evaporation acted to increase/decrease the δ18O of water vapor, and the local isoflux impacts on the δ18O of water vapor were then determined by analyzing the seasonal tradeoffs between them. This study demonstrates the biotic controls on the day-to-day variations in water/energy flux partitioning using isotopic evidence at the canopy–atmosphere interface. It also reemphasizes that the isotope tracer approach is a useful tool for quantitative evaluations of the relationships among ecohydrological components in ecosystems.