Youtube ziploc 1981 nov 18 2015

1 Introduction

Water is essential for life as it is a principal constituent of living organisms. Humanity’s dependence on water is evident as water is used in every human cell and is necessary for several basic processes and functions. Besides the direct use within the human body, water is vital for humanity due to its role in agriculture, household, industry, and environment. Even though water is a renewable resource, its availability is variable and limited (Pimentel et al. 1997). Almost every country in the world occasionally experiences water shortage (Gleick 1993). Water shortage can be caused by both natural hydrological variability and the effects of human actions. According to the most recent comprehensive global data (2010), agriculture accounted for approximately 69% of the global water use, with high regional variability, obviously (Gleick and Cooley 2021). Insufficient availability of water resources causes substantial problems to agriculture (Kang et al. 2017) and thus requires efforts to understand hydrological processes to effectively reduce agricultural water use.

Stable isotopes of hydrogen (δ2H) and oxygen (δ18O) in water can be used to understand hydrological processes and quantify water fluxes at the catchment scale (McGuire and McDonnell 2006, Kendall and McDonnell 2012; Hogan et al. 2020a) and the plot scale, via methods such as soil profiles or lysimeters (Hogan et al. 2020b; Maloszewski et al. 2006, Stumpp et al. 2009). They have also been used to investigate processes within the soil–vegetation–atmosphere continuum (Gehrels et al. 1998, Sprenger et al. 2016, 2019, Anderson et al. 2017, Stumpp et al. 2018). A major field of water stable isotope investigations within the soil–vegetation–atmosphere continuum and at the plot scale is the partitioning of evapotranspiration (ET) into soil evaporation (E) and plant transpiration (T) (Beyer et al. 2020, Rothfuss et al. 2020). Stable-isotope-based partitioning methods refer to the concept that ET changes both the water content and the water isotope distribution within a soil profile (Fig. 1). While ET decreases the soil water content, the isotopic composition changes due to isotope fractionation during evaporation only. Generally, fractionation is driven by vaporization, where water changes its phase to gaseous and lighter isotopes vaporize more readily than heavier isotopes. This leads to enrichment of heavier isotopes in the liquid phase and thus a change of the isotope ratio in the residual liquid. Regarding evaporation, this causes an accumulation of heavier isotopes in soil water within the soil surface layer (Allison et al. 1983). For transpiration, vaporization at plants’ stomata results in the accumulation of heavier isotopes in leaves and plant stems (Zimmermann et al. 1967) without relevant isotope fractionation and the accumulation of heavy isotopes in soil water during root water uptake (Ehleringer and Dawson 1992). Consequently, measurement of water and isotope distribution in combination with quantification of isotope fractionation allows calculation of the E and T ratios and rates by use of a water and stable isotope mass balance. The basic concept of the water and stable isotope mass balance is illustrated in Fig. 1.

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Figure 1. Soil water content (SWC) and isotopic ratio profile changes due to evapotranspiration (ET): Transpiration (T) is considered to reduce SWC without changing the soil water’s isotopic composition. Evaporation (E) causes isotope fractionation and shifts in both the SWC and its isotopic composition.

Figure 1. Soil water content (SWC) and isotopic ratio profile changes due to evapotranspiration (ET): Transpiration (T) is considered to reduce SWC without changing the soil water’s isotopic composition. Evaporation (E) causes isotope fractionation and shifts in both the SWC and its isotopic composition.

A number of recent studies question the assumption of fractionation-free root water uptake (De Deurwaerder et al. 2018, Barbeta et al. 2019, Oerter and Bowen 2019). They base this on the observation that individual stable isotope water molecules require more energy to disassociate from water aggregates for passing through the plasma membrane in the root endodermis than do the most abundant water molecules (Chacko et al. 2001). This is the case for the symplastic pathway for water movement from soil to the root xylem (Luo et al. 1991). The extent of symplastic movement depends on the characteristic of the radial root cell walls of the endodermis (Casparian strip). The form of the Casparian strips vary widely among plant species (Ellsworth and Williams 2007) and they are specifically developed in halophytes and xerophytes (Poljakoff-Mayber 1975). Consequently, (glycophytic and mesophytic) plants with less developed Casparian strips and predominant apoplastic water movement, where water aggregates can enter the root without dissociation into single water molecules, have no distinct fractionation (Ellsworth and Williams 2007). Due to the characteristic tripartite symbiosis of soybean with rhizobia and mycorrhizal fungi (Meng et al. 2015), it should further be mentioned that the presence of arbuscular mycorrhizal funghi increased isotope fractionation on xerophytes (Poca et al. 2019). However, the isotopic fractionation during root water uptake is considered to be negligible for (the mesophytic) soybean. Furthermore, all calculations for partitioning ET were based on oxygen isotope rates, as hydrogen is considered to be more reactive and involved in reactions not taken into consideration. This is advantageous, as the fractionation of isotopes during root water uptake by plants and mycorrhizal fungi is observed more distinctly for hydrogen than oxygen (Ellsworth and Williams 2007, Poca et al. 2019). Vargas et al. (2017) observed even 10 times more fractionation for hydrogen isotopes than for oxygen isotopes.

Several experimental and modelling approaches exist based on isotope composition of soil water and vapour to partition E and T fluxes (Soderberg et al. 2012, Wu et al. 2017, Rothfuss et al. 2020). The variety of experimental approaches comprises techniques to sample and analyse water vapour above or at the soil surface (Wang et al. 2012, Hu et al. 2014, Lu et al. 2017), or to sample and analyse soil water and soil water vapour from pore space directly affected by soil evaporation in the vadose zone (Gaj et al. 2016). Theoretical approaches to partition fluxes are often based on the free water evaporation Craig-Gordon model (Craig and Gordon 1965) or on analytical or numerical isotope transport models (Soderberg et al. 2012). Beside micrometeorological approaches, process-based models for catchment-scale are also commonly used for ET partitioning. These models consider the entire hydrological cycle in a watershed, including isotope compositions of runoff, vadose, ground-, stream, and xylem waters. Consequently, isotope-based ecohydrological models can be used for partitioning ET and determining further processes of the hydrological cycle, such as water storage and mixing (Ala-Aho et al. 2017, Knighton et al. 2020), water travel-time tracking (Ala-Aho et al. 2017, Kuppel et al. 2018), and investigation of runoff-generating processes (Birkel et al. 2014, Ala-Aho et al. 2017). Furthermore, the assessment of isotope data can serve for development and calibration of these hydrological models and reproduction of hydrological patterns (Birkel et al. 2010, Knighton et al. 2017).

Established sampling and analysing approaches provide valuable information on E and T ratios and insights into the partitioning process. However, depending on the theoretical framework and set-up, every method has its inherent, unique limitations, such as being representative only for a short term or as a snapshot of current boundary conditions and plant development (e.g. Aouade et al. 2016). Some methods are constrained to laboratory conditions such as the determination of evaporation with high spatio-temporal resolution analysis of isotopes in pore water (Rothfuss et al. 2015). Others are based on sensitive parameters that may result in substantial uncertainty or systematic bias, if not determined accurately, as pointed out by Hsieh et al. (1998). Others involve the risk of neglecting influencing factors or being based on vague assumptions for deriving water balance components, such as steady-state transpiration (Rothfuss et al. 2010). Although ET is the key water balance component, some methods lack information about actual ET rates (Mahindawansha et al. 2020). Wenninger et al. (2010) combined weighing lysimeter measurements and stable isotope techniques in a laboratory experiment to partition ET based on a water and stable isotope mass balance. Small indoor weighing lysimeters were utilized to precisely quantify actual ET rates. Sutanto et al. (2012) enhanced the experimental set-up, also taking interception into consideration. In both studies, ET from a grass vegetation was partitioned into E and T under laboratory conditions. As the laboratory set-up of Sutanto et al. (2012) resulted in reliable E and T rates, it appears promising to further adapt the methodology for field application for investigating ET portioning of commodity crops under natural conditions.

The goal of this study was to develop a stable isotope and lysimeter based method to determine E and T rates of commodity crops under natural conditions. The method is based on a laboratory experimental set-up by Sutanto et al. (2012) and was tested in a case study on soybeans. The adaptions of the original laboratory approach included the modification of isotope sampling for field investigations and the improvement of assessing water and isotope balance components. The improved assessment of water and isotope balance components was achieved by advancing the approach of Sutanto et al. (2012) in three steps. First, the isotopic composition of the transpired water fraction was related to a crop-specific root water uptake profile. Second, the fractionation factors for each evaluation period were calculated from hourly weather data weighted with actual ET rates from lysimeter measurements instead of period-averaged weather data. Third, the determination of the isotope ratios of evaporated water was based on isotope ratios in the evaporation zone near the soil surface instead of on the soil column averages. This adaption required excavation of soil cores for isotope analysis instead of in situ extraction of water. As the permanently operated lysimeter facility did not allow destructive soil sampling, the isotopic profiles were assessed in the adjacent area. Consequently, it was necessary to verify the basic requirement that the lysimeter and its adjacent area were similar with respect to soil water conditions and crop development. Finally, the results of the adapted determination of E and T ratios were compared with ratios from a numerical simulation with ratio calculation based on soil cover fractions.

2 Material and methods

2.1 Experimental site with lysimeter facility

The experimental site with a weighing lysimeter was in Groß-Enzersdorf (48°12’N, 16°34’E; 157 m elevation a.s.l.), east of Vienna, Austria. The climate is sub-humid (10.5°C mean temperature, 550 mm annual precipitation). During the investigated vegetation period of soybean (25 April to 17 September 2019) the mean air temperature was 19.7°C, and 154 mm of precipitation were measured with the lysimeter. April and May were colder and wetter than the long-term average, which caused late planting of soybeans (66 plants/m2, var. GL Melanie) and delayed plant development; June was atypically dry and hot. The lysimeter and its adjacent area were irrigated several times as necessary to secure crop yield by use of a watering can and a water hose, respectively (152 mm of tap water in total). Weather data were recorded with a weather station that was placed within the soybean stand. Two pairs of temperature and relative humidity sensors (HMP35C, Campbell Scientific, Inc., USA) were mounted at approximately 2.5 m height and near the surface. Data were recorded at 15-minute intervals and transferred by a telemetry system (Adcon Telemetry GmbH Klosterneuburg, Austria) as described in Nolz et al. (2013a). Daily precipitation and irrigation, as well as daily averaged values of temperature and relative humidity, are shown in Fig. 2.

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Figure 2. Daily precipitation and irrigation in mm, daily averaged values of temperature in °C, and of relative humidity in %, during the vegetation period in Groß-Enzersdorf, 2019.

Figure 2. Daily precipitation and irrigation in mm, daily averaged values of temperature in °C, and of relative humidity in %, during the vegetation period in Groß-Enzersdorf, 2019.

Actual ET of the soybean stand was determined with a weighing lysimeter (A = 2.86 m2). The soil type was sandy loam (0–140 cm) over gravel (140–250 cm). Percolation water was collected in a tank at a free-draining bottom outlet. Mass changes of the lysimeter and the drainage tank were measured with load cells, logged at 10-minute intervals, and averaged to hourly means. The recorded values were converted to soil water content and percolation water in units of mm of water. Conversion and processing of data noise and outliers with smoothing operations were done as described in Nolz et al. (2013b). ET was determined by solving a water balance equation, calculated for each logged 10-minute time interval and summed to daily rates. The lysimeter and its adjacent area (approx. 60 m2) were cultivated alike (manually).

2.2 Soil water monitoring

Soil water content (SWC) was monitored using EnviroSCAN® and Diviner 2000® soil moisture probes (Sentek Pty Ltd., Stepney, Australia). These probes have a sphere of influence where 99% of the reading is taken within a 10 cm radius from the outside of the access tube. For both devices the respective standard calibration relationship for sandy loam (Sentek Pty Ltd 2009) was applied to convert sensor readings (frequency domain resonance principle) to SWC values in cm3 cm−3. An EnviroSCAN probe was permanently installed within a vertical access tube in the lysimeter. It comprised 16 sensors on a mounting rail, measuring SWC hourly every 10 cm from 10 to 160 cm down the soil profile. For the mass balance, only the SWC data from 10 to 80 cm at times of soil sampling were considered. Another access tube was installed in the adjacent area for measuring with the portable Diviner at the same depths. Measurements with the Diviner were taken manually on a weekly basis in the access tube in the lysimeter as well as in the access tube next to the lysimeter.

The SWC values of each soil profile depth in and next to the lysimeter were compared to statistically prove the hypothesis of equal water content distributions in both profiles. This is necessary as the isotope and water mass balance was calculated for the lysimeter (system), which must not be disturbed by soil sampling for determination of isotope profiles. The comparison was made by applying a two-sample t-test for independent samples of each sensor depth, as the measuring errors were assumed to be normally distributed. Therefore, SWC distribution was measured in each access tube at least six times. The number of minimum required observations was determined beforehand by requiring a significance level of .05 and a power of the t-test of 0.8 and therefore an acceptable type II error. The significance level for each two-sample t-test was adjusted by applying the Bonferroni correction to consider inflation of type I error due to multiple comparisons. The p value for each layer depth was lessened by dividing it by the number of compared sensor depths.

2.3 Sampling and isotope analysis

2.3.1 Water samples

Precipitation and percolation water samples for liquid analysis were taken at the beginning and end of every (weekly) evaluation period. Precipitation water was collected with a rainwater collector for isotope sampling (Rain Sampler RX1, Palmex Itd, Croatia), which was positioned next to the lysimeter. Percolation water was taken from the collection tank at the bottom of the lysimeter. Irrigation water samples for liquid analysis were taken before each application directly from the tap.

2.3.2 Pore water sampling (vapour analysis)

Isotope ratios of soil pore water were determined from soil samples based on the water-vapour equilibration method by Wassenaar et al. (2008). Pre-tests were made initially to assess the minimum feasible water content that still allowed accurate isotope analysis. Furthermore, pre-tests were made to determine the number of soil samples required for a composite soil sample that ensured representative and repeatable isotope profiles. For determination of the minimum feasible water content, soil from the experimental site was oven-dried, saturated with isotopically known waters to SWCs from 0.050 cm3 cm−3 to 0.150 cm3 cm−3, and analysed (~ 50 g soil/sample). For determination of the required number of soil samples contributing to a representative composite soil sample, the deviations of composite samples with different numbers of subsamples from the same soil depth were assessed. Results of the pre-tests showed that all samples down to the driest tested soil water content (0.050 cm3 cm−3) had excellent accuracy, with isotope deviation of δ18O = −17.42 ± 0.14 and δ2H = −137.13 ± 0.37 (H20 ppm = 34 200 ± 560), compared to saturated soil samples. Composite samples mixed from three subsamples showed satisfactory results with deviations between 0.02–0.08‰ for δ18O and 0.05–0.22‰ for δ2H, compared to other composite samples from three subsamples. Based on the pre-test, soil cores were sampled weekly with an auger from 5 cm down to 85 cm, and directly placed into airtight plastic freezer bags with double zipper seals in the field (Ziploc, 17.7 × 19.5 cm). Samples of every 10 cm depth interval were collected containing soil within ± 5 cm at the corresponding depth of the soil water sensors. For every increment, three samples of the same depths from representative places were mixed. Surface layer soil samples from 0–5 cm depth were taken with a core cutter. All composite soil samples were immediately put into airtight plastic freezer bags with double zipper seals (double-bagged), which were carefully deflated, and stored in the fridge (max. 14 days) before preparation for analysis.

Additional to the weekly sampling across the soil profile for the mass balance calculation, the top layer was analysed at a higher resolution (0.5 cm increments; 16 September 2020) after a dry period of 10 days without precipitation for isotope analysis of pore water from 0.5 to 8.5 cm depth. This singular assessment was done at a plane homogeneous area with a mason’s trowel, a measuring tape, and a water level. Therefore, a rectangular soil column (approx. 25 × 15 cm) was bared in advance with a small trench and subsequently sampled layer by layer.

2.3.3 Analytical methods

Oxygen and hydrogen isotope ratios of all liquid and vapour samples were analysed with a laser-based isotope analyser (Picarro L2140-i (liquid samples) and L2130-i (vapour samples)). A two-point calibration was used with laboratory reference material (deionized Baltic Sea water and tap water) that was calibrated twice a year against international standards (U.S. Geological Survey (USGS) standards: USGS 46, USGS 47, USGS 50, and Vienna Standard Mean Ocean Water, VSMOW). Considering that the Picarro manual recommends three secondary standards for routine calibration (only as an intermediate point on the calibration line for verification), a third secondary standard (snow from an alpine region) was only used for samples exceeding the range of values on the calibration line spanned by the standards. Measured values were normalized to the internal standards and reported in delta notation δ referenced to the Vienna Standard Mean Ocean Water-Standard Light Antarctic Precipitation scale (Craig 1961). For analysis of liquid samples, each sample was measured up to seven times. Precision of the analysis was better than 0.1‰ and 0.5‰ for δ18O and for δ2H, respectively.

For vapour samples, procedures of the water-vapour equilibration method described in Wassenaar et al. (2008) and Gralher et al. (2018) were followed. Soil samples and samples with isotopic calibrated standard waters were prepared identically. Approximately 120 mg of soil samples from the auger (except surface layer samples with 240 mg) or 5 mL of standard water were put into airtight plastic freezer bags with double zipper seals (Ziploc, 17.7 × 19.5 cm) in the field and in the laboratory, respectively, and inflated with dry air in the laboratory. Before inflation, soil aggregates were manually and carefully broken within the sealed bags to allow optimum conditions of reaching water-vapour and isotope equilibrium within the closed bags. To reach the equilibrium, inflated bags were stored for three days before analysis. Airtightness of the bags was checked manually and by weighing the prepared samples after preparation and before analysis. Additionally, all prepared samples were double-bagged to prevent transient liquid-vapour isotope fractionation. In general, sample preparation was done immediately after sampling or with minimum storage of closed, double-bagged samples in the fridge. Analysis of vapour samples (soil core samples) was done manually by puncturing the sampling bags with a needle and enabling continuous flow to the analyser chamber of the Picarro L2130-i for vapour samples at a fixed suction rate. Before and after every three soil samples, a standard was measured for data normalization and instrumental drift correction. A measuring period of five minutes allowed the instrument to reach a constant measured value for at least two minutes, where the characteristic sample value was determined from a 1-minute mean value. The averaged standard deviations of all 104 soil samples were 27.4 ppm for H2O vapour concentration, 0.182‰ for δ18O, and 0.374‰ for δ2H. These deviations were considered as constant readings (Gralher et al. 2018) and as acceptable measurement uncertainties.

2.3.4 Isotopic reference systems

Dual-isotope plots were used to analyse the alteration of δ2H/δ18O ratios of the residual soil water due to non-equilibrium fractionation processes during evaporation. The latter causes a shift of the ratios from a local meteoric water line (LMWL) that describes the isotope composition of its original meteoric source waters. The LMWL for this study was based on long-term (1960–2018) precipitation data in Vienna from the International Atomic Energy Agency (IAEA) database (WISER [Water Isotope System for Data Analysis, Visualization and Electronic Retrieval]; accessed on 22 December 2020). An LMWL with a slope of 7.54 and an intercept of 2.36 was determined based on a reduced major axis (RMA) regression (R2 = 0.97). The influence depth of isotope fractionation was further assessed with the line-conditioned excess (lc-excess) (Landwehr and Coplen 2006): (1) lc−excess =δ2H−a δ18O−b(1)

with a as slope and b as intercept of the LMWL. The lc-excess expresses the deviation from the LMWL, where negative lc-excess values indicate the effect of evaporation fractionation processes.

For describing the isotope profile, the term evaporation front was used for the vapour-liquid interface and not the maximum depth of the kinetic fractionation signal of soil water, according to Barnes and Allison (1988). For the vapour-liquid interface, upward water vapour flux exceeds the liquid flux (Braud et al. 2005), and soil evaporation (and fractionation) mainly occurs. Then, the evaporation front can be observed as a peak (of isotopic enrichment) in the isotopic profile (Braud et al. 2005, Beyer and Dubbert 2019). The desiccated zone above the front (inter-phase zone) was considered to contain vapour in equilibrium with the liquid phase and a steady concentration profile, where transport takes place by molecular diffusion (Gat et al. 2001).

2.4 Water and stable isotope mass balance

The mass balance of a parcel of soil (column) for calculating the E and T fraction over a time interval was calculated as (2) mtotal =mi+mp=me+mf+mt+ml(2)

where m represents a mass of water per unit volume of soil; the indices i and f represent the initial and final soil water content, respectively; p indicates precipitation plus irrigation; e is evaporation; t is transpiration; and l is percolation (incoming components are positive, outgoing components are negative). Analogously, each component of the mass balance was computed as product of the stable isotope ratios δ (e.g. δ18O) and the fraction of water in that component x (as xj = mj/mtotal). The components δi, xi, δp, xp, δf, xf, δt, δl, xl (δi and δf as weighted averages for the soil water column) were measured, and δe was calculated from atmospheric conditions and fractionation factors (Sutanto et al. 2012). Based on conversions and assumptions described by Sutanto et al. (2012), xt and xe were determined as a residue of the balance calculation as (3) xt=xp +xi−xe−xf−xl(3)

and (4) xe=x iδi+xpδp−xfδf−xpδt−xiδt+ xfδt+xlδt−xlδl−εtotal(4)

where εtotal is the sum of equilibrium and kinetic fractionation.

2.4.1 Adapted determination of δt and δe

The isotope composition of transpired water, δt, was determined from a weighted arithmetic mean of isotope soil water ratios and SWC down to the effective root depth. Water uptake within the root zone was calculated for each time step and soil depth based on the actual SWC and weighted with a factor depending on the root density, which was assumed to linearly decrease from its maximum at the surface down to its minimum at the maximum root depth. For calculating the isotope ratio of the water taken up by the roots, each layer was further weighted with the actual isotope ratio at this depth. (5) δt= ∑j=1nδjSWCjHjn+1−j∑j=1nSWCjHjn+1−j(5)

To better approximate the actual fractionation factors based on measured high-resolution weather data, the total isotope fractionation factor εtotal (comprising equilibrium and kinetic fractionation) was calculated on an hourly basis. Hourly values of εtotal were weighted with measured ETact data to better reflect actual conditions during times of high evaporation. (6) εtotal =∑i=1mETactiεtotali ∑i=1mεtotali(6)

where m was the number of hourly time steps, ETact i was ETact in time step i, and εtotal i was the total isotope fractionation factor εtotal in time step i. Furthermore, δe was calculated based on δsurface(7) δe=δs urface−εtotal(7)

instead of δt, which also considered deeper soil layers not affected by E.

2.4.2 Determination of sampling depth

It was necessary to determine an appropriate sampling depth – as a lower system boundary – that comprised the effects of evaporation and root water uptake throughout the vegetation period. Therefore, the maximum depth of the kinetic fractionation signal was detected with lc-excess profiles. Furthermore, root length was measured, and vertical water flux was numerically modelled to ensure the assessment of all relevant fluxes in the observed system. The choice of the lower system boundary was in addition checked by calculating the closure of the water mass balance.

2.5 Crop monitoring

The crop parameters plant height, phenological stage, root length, leaf area index (LAI), and soil cover were determined once or twice a week. Plant height and root length were measured with a measuring tape. Phenological development was monitored using the BBCH (Biologische Bundesanstalt für Land- und Forstwirtschaft, Bundessortenamt und CHemische Industrie) scale for uniform coding of development stages (Meier 2018). Canopy development was described by LAI, measured with an AccuPAR PAR/LAI Ceptometer Model LP-80 (Metre Group Inc., USA). LAI measurements on the lysimeter and on the adjacent area were repeated five times and checked for consistency with a two-sample t-test for independent samples (with a type I error rate α of 0.05). Additionally, soil cover fractions (residues, stones, and living plant material) were acquired on a weekly basis with the use of an image analysis tool (Bauer and Strauss 2014). Yield data were derived from manually harvested plots of 2 m2 each at the lysimeter and the adjacent area. Evaluation of yield data comprised number of plants, average number of husks per plant, weight per thousand grains (thousand seed weight, TSW), grain yield, and average number of seeds per plant and husk. For yield data comparison, the 2 m2 standard sampling areas did not allow replications on the lysimeter and thus statistical evaluation.

2.6 Numerical simulation with HYDRUS-1D

The HYDRUS-1D model (Šimůnek et al. 1998a) was set up according to the lysimeter soil profile in seven layers with known physical soil properties (Neuwirth and Mottl 1983). Calibration, validation, and simulation were performed using the single-porosity van Genuchten-Mualem model (Van Genuchten 1980) without hysteresis and in daily time steps. Upper boundary conditions were set as atmospheric fluxes with a max. height of 5 cm of surface water layer (lysimeter ring height). Precipitation and irrigation were obtained from lysimeter data. The daily potential evapotranspiration (ETpot) was calculated from weather and crop data – reference evapotranspiration (ET0) with the standardized ASCE (American Society of Civil Engineers) Penman-Monteith equation and crop evapotranspiration (ETc) with the dual crop coefficient approach (Allen et al. 1998). These daily ETpot values were partitioned in HYDRUS-1D according to Beer’s law to Epot and Tpot, which splits ETpot complementary based on a soil cover fraction (8) SCF=1−e− 0.463 LAI (8)

where measured LAI values were given in tabular form, and the original extinction coefficient (0.463) was used (Šimůnek et al. 2012b). Eact was calculated from Epot by limiting the E rate to the condition that the pressure head at the soil surface was larger than a minimum allowed pressure head based on relative humidity, air temperature, physical constants, and the molecular weight of water. Tpot was reduced to Tact by applying the stress response function by Feddes et al. (1978) and default parameters of the respective crop. The lower boundary condition was set as free drainage at a bottom outlet according to the lysimeter design. The soil hydraulic parameters were calibrated with a Levenberg-Marquardt type parameter estimation technique for inverse estimation of selected soil hydraulic parameters and observed SWC data from 2005 (220 days) (Šimůnek et al. 2012a, 2012b). The objective function as defined in Šimůnek et al. (1998a, 1998b) was minimized in the parameter estimation process based on a Levenberg-Marquardt nonlinear minimization method. Initial soil hydraulic properties such as saturated hydraulic conductivity were obtained experimentally in the lab. SWC data originated from the EnviroSCAN sensors at selected depths of 10, 20, 40, 60, and 80 cm to calibrate the soil hydraulic properties, in particular in the section of the investigated soil column. The saturated hydraulic conductivity parameters were calibrated for each layer with up to 10 iterations without predefined bounds for parameters’ space. The model requirements (independence, normal distribution, homoscedasticity) were checked with and autocorrelation analysis and a Shapiro-Wilk test. Validation was performed with observations from maize vegetation in 2011 (183 days, 915 observations, R2 = 0.89, and a mean weighted absolute error of 0.00759). The simulation of soybean vegetation in 2019 was based on measured initial soil water conditions and observed root length, plant height, and LAI. A graphical comparison of measured and simulated time series is shown in section 3.2.3.

3 Results and discussion

3.1 Adaptation of the method to field application

3.1.1 Assessing the similarity of conditions in and next to the lysimeter

SWC measurements were generally equal in and next to the lysimeter during the measuring period (). Only comparisons at day 10 after seeding indicated rejection of the hypothesis of similar soil water distributions.

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Table 1. Comparison of soil water distributions in the lysimeter (L) and its adjacent area (A): Measurements at different crop development stages for given days after seeding (DAS) on 25 April 2019. Averaged SWC values in % (with standard deviation in parentheses) for each sensor depth (PD) in cm. Acceptance/rejection of H0 (equal means): xP < .05 (P* < .00625); xxP < .01 (P** < .00125) (P* as adjusted p values for multiple comparisons, Bonferroni correction). X means H0 (equal means) rejected

Plant development at the lysimeter and the adjacent area was assessed based on BBCH development stages (). The identification of phenological development stages showed that all development steps emerged simultaneously for the lysimeter and the adjacent area within a range of three days.

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Table 2. Comparison of observed BBCH development stages according to Meier (2018) on the lysimeter (L) and its adjacent area (A). Dates given as day after seeding (DAS) on 25 April 2019

Additionally, monitored plant heights, LAI values, and soil cover fractions demonstrated similar plant development in and next to the lysimeter (). In general, the sub-optimal weather conditions during the vegetative phase and the characteristic low growth height of the soybean strain are responsible for a low extent of canopy. During the first few development stages, plant heights showed little difference due to slightly faster plant development in the adjacent area. With ongoing plant development, differences disappeared within the inaccuracy of height measurement. Statistical evaluation of all LAI values indicated that all observations from both areas had equal means.

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Table 3. Plant heights (PH) and averaged LAI values (plus standard deviation in parentheses) on the lysimeter (L) and its adjacent area (A), root length (RL) only on A. PH and RL are given in cm, LAI in m2 m−2. Dates are given as days after seeding (DAS)

Similar to LAI measurements, estimations of soil cover fractions showed good agreement (Fig. 3). Small differences were associated with differing weed infestation or faulty attribution of shadow. These differences disappeared with increasing soil cover by living plants.

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Figure 3. Progress of soil cover at the lysimeter (upper two rows) and adjacent area (lower two rows) for selected dates in days after seeding (DAS). Soil surface cover images (SSC) above corresponding classification results (CR), respectively.

Figure 3. Progress of soil cover at the lysimeter (upper two rows) and adjacent area (lower two rows) for selected dates in days after seeding (DAS). Soil surface cover images (SSC) above corresponding classification results (CR), respectively.

A posterior validation of conformity of plant growth at the lysimeter and the adjacent area was the comparison of yield parameters, which showed consistency. For both areas, about two thirds of the seeds emerged and developed. Differences of all yield parameters between the lysimeter and the adjacent area were less than 5%.

All considered parameters (SWC distribution, BBCH, LAI, plant height, soil cover fractions, and crop parameters) verified similar conditions in the lysimeter and the adjacent area. Furthermore, the applied control measurements proved suitable for monitoring growing conditions and plant development at the lysimeter experiments. In the case of only monitoring with selected control measurements, it seems practicable to apply a combination of methods that are statistically provable (e.g. SWC distribution) and such that indicate unequal conditions instantly (e.g. soil cover fractions).

3.1.2 Assessing the chosen monitoring and sampling depth

The dual-isotope plot of soil waters and meteoric water lines shows the importance of enrichment due to evaporation for soil water samples (Fig. 4). Soil water, that partly evaporated and thus experienced fractionation, was enriched in heavy isotopes. Consequently, samples from pore water deviated from the original position and mainly plotted below the LMWL. δ values of irrigation samples (empty circles) were smaller than δ values of soil water (coloured circles) and close to the intersection of LMWL and the soil water regression line (soil water isotopic trend line). Isotope ratios of the pore water sampled throughout the entire vegetation period indicated a distribution along a linear regression line (soil water isotopic trend line) with a slope of 5.1 (for all samples between surface and 80 cm; slope is 5.0 for samples between surface and 40 cm).

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Figure 4. Dual-isotope plot of soil water (with soil water isotopic trend line), precipitation and irrigation samples. GMWL: global meteoric water line; LMWL: local meteoric water line based on long-term precipitation data from Vienna (IAEA database WISER). It shows clusters of soil water samples for each profile depth along the soil water regression line.

Figure 4. Dual-isotope plot of soil water (with soil water isotopic trend line), precipitation and irrigation samples. GMWL: global meteoric water line; LMWL: local meteoric water line based on long-term precipitation data from Vienna (IAEA database WISER). It shows clusters of soil water samples for each profile depth along the soil water regression line.

The isotope trend line is comparable with trend lines from other studies, which show similar y-intercepts and slopes between 2.0 and 5.0 (Wenninger et al. 2010, Kendall and McDonnell 2012, Sutanto et al. 2012, Clark and Fritz 2013, Gaj et al. 2016). The steep slope of the trend line (Fig. 5) – in particular compared to laboratory experiments by Sutanto et al. (2012) (slope of 3.7) – may be due to the higher relative humidity (Dincer et al. 1974) or the fact that it is caused by fractionation of variable isotope source water under natural conditions (Benettin et al. 2018) compared to uniformly spiked irrigation water. Samples plotted below the trend line originated from dry and warmer days (day after seeding, DAS 60, 68, 75, 91, 117). Samples near the trend line originated from intermediate weather conditions (DAS 103, 124). Samples above the trend line originated from cooler and wetter days (DAS 139, 145), which were additionally influenced by prior heavy rainfall events (DAS 111, 131). This was another reason for the steeper slope of 5.1, especially compared to laboratory experiments.

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Figure 5. (a) isotopic soil profiles, (b) soil water profiles, and (c) lc-excess soil profiles for sampling dates.

Figure 5. (a) isotopic soil profiles, (b) soil water profiles, and (c) lc-excess soil profiles for sampling dates.

The isotope ratios of samples from the same depth throughout the vegetation period showed clusters along the soil water regression line (Fig. 4). Illustrated in a profile plot, the shape of the line describing the isotopic distribution (δ18O) across the soil profile for different dates was remarkably constant throughout the entire vegetation season (Fig. 5a). The similarity of all sampled curves indicates that the high evaporativity in the summer caused a rapid reshape of the characteristic isotope profile curve after rain events (Fig. 2), e.g. in the soil profile from 16 July (DAS 82) after a rain event on 13 July (14.5 mm). In general, the constancy of the isotope profiles agrees with findings of Allison et al. (1983), who reported a steady exponential decrease of isotope values, and thus an influence of isotope enrichment, down the profile under evaporative conditions. Accordingly, in the present study, isotopes in pore water from surface layer samples (0–30 cm) were most enriched by fractionation during evaporation (Fig. 4), and the impact of evaporation decreased substantially down to a depth of approx. 50–60 cm, where a minimum was observed (Fig. 5a). Below, isotope ratios of pore water samples stayed constant or slightly increased up to 1.6‰ for δ18O. Samples from 80 cm depth already had similar isotope ratios to percolation water, which indicated little effect of evaporation (Fig. 5). The largest isotope ratios were consistently within the top 5 cm. This isotope enrichment measured within the top 5 cm surface layer samples (Fig. 5a) and the lc-excess (Fig. 5c) indicates that soil evaporation mainly occurred in these uppermost 5 cm and that the vapour-liquid interface (evaporation front) was close to the surface.

The detected maximum depth of the kinetic fractionation signal at approx. 50–60 cm depth was noticeably lower than that reported in literature. This applied to all settings, comprising laboratory experiments (Wenninger et al. 2010, Sutanto et al. 2012) and field measurements (Zhang et al. 2011, Dubbert et al. 2013), which aimed to partition ET based on stable isotope balance or investigated evaporation dynamics (Sprenger et al. 2017), or unproductive water loss (Mahindawansha et al. 2020). In these studies, the fractionation signal, which indicated fractionation caused by evaporation, disappeared within the first 25 cm or less. A review by Sprenger et al. (2016) showed that evaporation fractionation could reach significant depths in warmer climate regions such as in arid climates, which may explain the observations found in our study.

The location of the largest isotope ratios within the top 5 cm agreed with Wenninger et al. (2010), but differed from the results of Sutanto et al. (2012), who located the largest isotope ratios mainly in 20 cm depth. To identify the particular depth of maximum isotope ratios and the layer with typical lower isotope ratios very close to the surface – which was isotopically depleted by vapour diffusion due to exchange with atmospheric air (Beyer and Dubbert 2019) – soil samples were analysed from 0.5 to 8.5 cm depth in 0.5 cm increments (Fig. 6). The maximum isotope ratios were close to the surface (in 0–0.5 cm depth for δ2H and in 0.5–1.0 cm depth for δ18O), which agrees with results from numerical simulation for non-steady state conditions in the literature (Braud et al. 2005). This showed that the desiccated zone – where water movement predominantly occurs as vapour diffusion – was very shallow and that back-diffusion of atmospheric water vapour had apparently no impact below 1 cm. According to the closeness of the maximum isotope ratio to the surface, resistance of vapour movement rapidly became determinant of the evaporation rate (instead of atmospheric evaporativity) and slowed the downward movement of the evaporation front. For conditions where the evaporation front is at greater depths (e.g. high evaporativity and coarser texture), the desiccated zone above the front (inter-phase zone) could gain importance, though. An increase of the inter-phase zone can decrease the impact of kinetic fractionation and raise the relevance of transport isotope fractionation due to diffusive processes (Gat et al. 2001). Consequently, that would require adaptions in the evaluation (e.g. additional consideration of transport isotope fractionation).

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Figure 6. Soil surface profiles after 10 days without precipitation. (a) δ2H, (b) δ18O, and (c) d-excess.

Figure 6. Soil surface profiles after 10 days without precipitation. (a) δ2H, (b) δ18O, and (c) d-excess.

The shallow location of the evaporation front within the upper 5 cm differed from the range of 0.1–0.5 m assumed by Sutanto et al. (2014). This could be attributed to different soil textures. The top layer in the lysimeter used by Sutanto et al. (2012) had sand/silt/clay fractions of 77/7/16, compared to 29/50/21 in this study.

Beside the importance of locating the evaporation front near the surface layer, the maximum depth of the kinetic fractionation signal of soil water is of interest as it is one of the indicators that affected the choice for setting the lower system boundary. For the studied period, evaporation caused enrichment of the pore water isotopes down to depths of 50 cm (Fig. 5a, c). The exponential decrease of fractionation over the soil profile in the summer months was consistent with other observations in arid climates (Allison et al. 1983). Diffusion of isotopes in both the liquid and the vapour phase broadened the isotope profile (Barnes and Turner 1998). Altogether, the isotope and lc-excess profile require an assessment of at least 50 cm (Fig. 5). A further criterion for determination of the lower system boundary – mass balance closure – indicated the need for assessment down to a lower depth. As there were still relevant changes of SWC below 60 cm soil depth (Fig. 5b), the lowest residues of water balance closure (13 mm for the vegetation period) were reached with a soil column of 80 cm depth. At the observation point of this depth, the numerical simulations with HYDRUS-1D indicated no relevant fluxes (percolation or capillary rise). Considering all relevant factors (lc-excess signal profile, maximum root length, vertical water flux at lower system boundary, and minimization of water balance residues), the lower mass balance boundary (and sampling depth) was set to 80 cm depth.

For future experiments, the lc-excess (Fig. 5c) and the water mass balance closure could be convenient references for the determination of the mass balance system depth. A first orientation for choosing proper sampling and monitoring depths for other soils and conditions could be found in the review by Sprenger et al. (2016), discussing lc-excess with regard to several boundary conditions, especially climate. In general, determination of isotope ratios in pore water of soil profiles from core samples proved to be suitable, even for dry soil samples. Analysis with the water-vapour-equilibration method considered the isotopic composition of total soil water comprising both mobile and immobile water. Therefore, the sampling approach also considered strongly bound water with longer residence time. This allowed the consideration of the precipitation water in its entirety, independent of rain event size and its capability to infiltrate into different pore sizes, where it mixed with pre-event water (Sprenger et al. 2015, 2016). Another benefit of soil sampling, especially from the surface with core cutters, was the potential validation of SWC in exactly definable soil depths.

3.2 E and T ratio determination for soybean

3.2.1 E and T ratios based on isotope mass balance calculation

The results from the isotope mass balance calculation based on the adapted determination of fractionation factors are shown in Fig. 7b. The weekly periods cover the plant development stages from vegetative development (LAI of 0.5) to maturation (harvest). The results of the adapted isotope balance calculation (Fig. 6b) are compared to results from the original calculation procedure (Fig. 6a) and a numerical simulation (Fig. 6c), and illustrate the differences of ratios and their progress.

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Figure 7. Daily Evaporation (E) and Transpiration (T) rates for the evaluation periods from the (a) original isotope mass balance, (b) adapted isotope mass balance calculation, and (c) HYDRUS-1D simulation. Start and end dates of each period given in days after seeding (DAS).

Figure 7. Daily Evaporation (E) and Transpiration (T) rates for the evaluation periods from the (a) original isotope mass balance, (b) adapted isotope mass balance calculation, and (c) HYDRUS-1D simulation. Start and end dates of each period given in days after seeding (DAS).

In general, the E and T rates determined for soybean were reasonable, and the partitioning ratios (43% to 85% from blossom to beginning of maturation) were within the range given in the literature (Sakuratani 1987, Brisson et al. 1998, Sauer et al. 2007, Singer et al. 2010, Zhou et al. 2016). Within this range, the measured weekly E and T ratios show remarkable variation. In the first evaluation period (which equals the vegetative development stage with formation of blossoms), the fraction of transpiration was larger than expected. A reason might be negligible precipitation prior to this period (DAS 43–63, Fig. 2); thus, soil near the surface was dried out (SWC was approx. 0.06 cm3 cm−3), limiting evaporation, while deeper layers still provided plant available water (SWC was approx. 0.25 cm3 cm−3 in around 40 cm depth, Fig. 8). Root length was already nearly fully developed (35–40 cm, ), whereas canopy had not yet covered the soil surface (LAI of 0.65, ). The large T fraction calculated with the isotope mass balance also explains the large absolute ET rates after such a dry period. With increasing frequency of rain events (Fig. 2), surface layer wetness also increased, and the evaporation fraction rose. Nonetheless, transpiration was dominant with a proportion of 0.79 to 0.85 during times of maximum soil cover (LAI > 2.2, , Fig. 7b). At the end stage of crop development (finished seed maturation/dieback), the measured T fraction did not decrease to zero. Considering the late maturation and low total ET, the remaining foliage and ground covering by weeds could have been responsible for this relatively large T fraction (, Fig. 7b).

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Figure 8. Observed and simulated daily soil water content (SWC) in selected profile depths at the lysimeter station Groß-Enzersdorf from day after seeding (DAS) 0 to 145.

Figure 8. Observed and simulated daily soil water content (SWC) in selected profile depths at the lysimeter station Groß-Enzersdorf from day after seeding (DAS) 0 to 145.

3.2.2 Impacts of adapted determination of δe and δt

Measured relative humidity values within the vegetation period ranged, according to averaged measurements taken at 15-minute intervals, from 20 to 100%, and temperature ranged from 2.1 to 36.3°C. Figure 2 shows the mean daily values, with relative humidity ranging from 44.4 to 94.4% and temperature ranging from 5.3 to 28.2°C. For comparison, the experiments of Sutanto et al. (2012) covered values between 22 and 36% (relative humidity) and 3 and 15°C (temperature). Consequently, the main influencing factors of isotopic fractionation processes were far more variable in the field compared to the experiment in the laboratory, emphasizing the need for adaption for field conditions. This large variation increased the effect of the adaptions in the calculations compared to the original approaches for δe and δt as described by Sutanto et al. (2012).

δe: The higher temporal resolution determination of fractionation factors (Equations 4 and 5) based on 10-minute lysimeter data and 15-minute weather data resulted in an increase of the average of the weekly total isotope fractionation factor εtotal from 14.0 to 16.2‰ compared to the non-weighted calculation. Weekly εtotal factors increased by between 1.7 and 2.8‰. This caused a higher estimation of the transpiration fraction during the vegetation period, of approximately 5% compared to the non-weighted calculation (Fig. 7a). This increase was due to the weighting of fractionation factors, which excluded or reduced phases of low evaporation and transpiration, such as night-times and rain events, and emphasized phases of high evapotranspiration. For the entire observed period, the weighted effects of equilibrium fractionation (11.5‰) were on average about twice as high as those from kinetic fractionation (5.2‰). However, the variance of the weighted kinetic fractionation (7.3‰), which was calculated based on relative humidity, was considerably higher than that of equilibrium fractionation (0.2‰), which was calculated based on temperature. Therefore, the kinetic fractionation process also had a considerable effect in the weighted determination of εtotal.

The second adaption, the calculation of δe based on δsurface instead of δt, resulted in 5% lower estimates of the transpiration fraction during the vegetation. Consequently, this adaption (−5%) and the weighted determination of fractionation factors (+5%) cancelled each other out over the entire vegetation period. However, the effects of the two adaptions were not uniformly distributed across the vegetation period and therefore still shifted the ratios between the individual sampling periods.

δt: A more substantial impact on ET fraction calculation was obtained with the adapted determination of δt. Compared to averaged δ values from the entire surveyed soil column, δt values shifted from a range of −8.2 to −5.7‰ to a range of −4.9 to −2.5‰. This meant that the altered assumptions regarding root distribution increased the evaporation fraction by about 20% compared to the calculation as described in Sutanto et al. (2012). For future experiments, isotopic composition of transpired water δt should be further determined with adjusted weightings down to the effective root depth. The choice of weight distribution could be alternatively determined and validated with numerical modelling and calibration results (Šimůnek and Hopmans 2009), or with a preferred and suitable root water uptake model (Jarvis 1989). Further, the measuring methods could be optimized (e.g. root water uptake based on matric potentials or root density distribution assessment with photogrammetry).

The adaptions of the evaluation (weighting of fractionation factors, δe calculation based on δsurface, and weighted root water uptake) decreased the T ratio compared to the basic evaluation described by Sutanto et al. (2012). As isotope approaches were supposed to generally overestimate T ratios (Sutanto et al. 2014), the proposed adaptions could contribute to better represent actual E and T ratios and their temporal variability.

3.2.3 Comparison to results from a soil cover-based numerical approach

The results of the calibrated and validated HYDRUS-1D model indicated good agreement between measured and simulated SWC (Fig. 8) and ET rates (Fig. 9). In general, ET rates proceeded according to plant development. As long as the soil surface was mainly bare until around DAS 50, daily ET rates did not show a trend and had large fluctuations caused by varying evaporativity and wetness of the soil surface (Figs 2, 8). After DAS 50, ET rates not only fluctuated corresponding to actual evaporativity but also proceeded according to plant development (). From the HYDRUS-1D output information, E and T fractions were calculated that were compared to results from the isotope mass balance (Fig. 7c).

Partitioning evapotranspiration using water stable isotopes and information from lysimeter experiments

Published online:

21 February 2022

Figure 9. Observed and simulated evapotranspiration (ET) at the lysimeter station Groß-Enzersdorf from DAS 0 to 145.

Figure 9. Observed and simulated evapotranspiration (ET) at the lysimeter station Groß-Enzersdorf from DAS 0 to 145.

The comparison of results from the water and isotope mass balance and the partitioning within the HYDRUS-1D numerical simulation showed the differences of the partitioning approaches. Basically, total ET from the measurement (lysimeter) and the simulation (based on the Penman-Monteith equation) throughout the evaluated season agreed well (Fig. 9). Regarding the ET partitioning, the isotope mass balance produced larger T ratios (0.61 ± 0.18) than the numerical approach (0.52 ± 0.16). In particular, the first period with a remarkably high ET exhibited differing ratios (Fig. 7). The numerical simulation calculated a lower T ratio compared to the isotope approach, which stayed stable until the LAI exceeded 1.0 (Fig. 7c, ). After the rain events around DAS 80 (at blossom), the T rate nearly doubled (similar to the isotope mass balance results) and was stable at its second plateau until beginning maturation. During these periods of full plant development, the T ratio was below the rates determined with the stable isotope balance. Afterwards, it declined with decreasing LAI (Fig. 7c).

The different ratios could be explained with a joint consideration of the isotope and numerical model approach and their response to varying weather conditions (Fig. 2) and soil water conditions (Fig. 8). The isotope approach was less prone to variations of micrometeorological boundary conditions such as temperature, relative humidity, and atmospheric turbulences near the soil surface. The reason was that measuring at the beginning and the end of each period (except for δe, which was calculated) represented the cumulated effects of evaporation and transpiration processes within the period. With this approach, variations of the micrometeorological conditions below the canopy have less impact. This is an advantage, as they are known to be difficult to measure, but may strongly influence the separation and total fluxes of E and T (Monteith 1981, Jarvis 1985). The partitioning within the numerical modelling approach, however, depended on several calculated or simulated processes, each afflicted with uncertainties. This mainly affected the reductions of potential to actual E and T rates. The reduction of potential T to actual T with the applied stress response function, for example, considered all relevant factors of crop response. Nonetheless, actual factors such as atmospheric turbulences and micrometeorological gradients below the canopy were only roughly considered by the model’s algorithm. Additionally, the initial partition of ETpot based on the soil cover fractions could have constrained the detection of E and T ratio variations. This constriction becomes increasingly effective with a wide range of conditions within the surface layer that affect the process of vaporization. The described inherent restrictions of both approaches may be identified at the ratio progression.

For the first period with partial soil cover, the isotope mass balance indicated a considerable T fraction, which could be attributed to dry soil at the surface and plant available water at greater depths for well-developed roots (Fig. 8). In the numerical modelling, such a high T rate was excluded by the low LAI at this stage (, Fig. 7c). The evapotranspiration rate of the two approaches was still similar, because the larger evaporation rate in the numerical simulation compensated for the difference. The high ability of the local soil to deliver water (evaporability) assumed by the numerical simulation caused, in general, larger evaporation rates than those determined with the isotope mass balance. With progressing plant cover, the sensitivity to actual conditions was less restricted by the ETpot separation based on the soil cover fractions. This was indicated by the large E rate in the third period, which deviated from the general ratio progression in the numerical simulation. Thus, the large E ratio, caused by several rain events that frequently wetted the soil surface, was quantified by both approaches (Figs 2, 7c). The impact of the initial partition based on soil cover fractions was nevertheless still observable, since T from the numerical simulation remained constant at the second plateau while T from the isotope mass balance responded more to atmospheric conditions. Additionally, the partition based on soil cover fractions (from measured LAI) neglected ground covering weeds or fallen leaves, as shown in the periods after starting maturation. In general, the procedure of partitioning evaporation and transpiration rates within HYDRUS-1D still requires further validation (Kool et al. 2014). Direct improvement can be expected from additional measurement of micrometeorological conditions between soil surface and canopy. Another contribution to improve the limiting process of potential fluxes would be a more detailed determination of evaporability of the surface layer, which is mainly affected by evaporation.

4 Conclusions

4.1 Adaptation of the method to field application

Profile water contents inside and outside the lysimeter were similar, and plants developed similarly on the lysimeter and in the adjacent area. Hence, the precondition for the applied method was fulfilled. Data analysis and soil water simulations resulted in a profile depth of 80 cm for the mass balance determination and, thus, sampling depth. Under the given conditions, this profile depth represents the root water uptake zone of soybeans and is recommended for further studies. Furthermore, investigations of the evaporation front close to the surface have proven to be of particular importance. In this regard, the proposed procedure of sampling soil from the uppermost few centimetres and analysing soil water instead of drainage water worked well even for dry soil conditions (soil water content of 0.05 cm3 cm−3). Consequently, the adapted isotope mass balance approach is suitable for application in dry agricultural production areas.

4.2 Determination of E and T ratio for soybean

E and T rates were calculated based on actual ET lysimeter data, which were partitioned with an adapted isotope mass balance. In general, the determined E and T rates of soybean were reasonable, and the partitioning ratios were within the range given in the literature. A distinct interpretation of the sub-processes was possible based on the weekly sampling and analysis intervals.

The adapted mass balance evaluation comprised the determination of fractionation factors in greater detail and took the distribution of root water uptake into account. The adaptions in determination of fractionation factors had little effect on the averaged E and T ratios over the entire vegetation period. A bigger effect was observed from a more detailed calculation of isotopic composition during root water uptake. The consideration of isotope profiles and water uptake distribution caused a higher estimation of the evaporation fraction throughout the whole vegetation period, of about 20%. Therefore, the proposed adaptions improved the determination of actual E and T ratios and decreased the effect of a supposed general overestimation of the T fraction by previous isotope approaches.

In general, the adapted experimental set-up and analysis procedure of the isotope mass balance approach allowed the quantification of E and T rates for the entire vegetation period of a commodity plant, comprising a realistic range of weather and field conditions. Therefore, the method may be important and useful for evaluating management actions, particularly in regions where water availability is a critical factor in crop production.