2 Methodology

2.1 Experimental Design

For our experiment, we chose two well-studied (see e.g.(Orlowski et al., 2016b, 2018b)) physico-chemically different soil types – a clayey loam (LUFA 2.4) and a silty sand (LUFA 2.1) – from the German State Research Institute for Agriculture (LUFA Speyer, 2015). For a detailed description of the soil properties, the reader is referred to Orlowski et al. (2018). Soils were sieved (2 mm) and oven-dried (48 h, 200°C). We chose two experimental approaches to test the effect of different pressure levels on the extracted soil water isotopic composition. Experiment 1 aimed at sampling soil water along the pF curve and during experiment 2 soil water was sampled sequentially at the highest pressure level (15 bar) over 7 days. We further tested whether there is an isotopic exchange between the ceramic plate water of the extractor and the water to be extracted from the soil samples sitting on these ceramic plates during extraction. Both experiments are based on spiking approaches with two different waters of known isotopic composition (see e.g.,(Orlowski et al., 2016b, 2018b)).
For the soil water extractions and the determination of water retention curves, we used pressure extractors (Soilmoisture Equipment Corp., USA; Figure 1). For errors associated with this method, the reader is referred to Solone et al. (2012). Each pressure extractor cell contains a porous ceramic plate covered on one side by a thin Neoprene diaphragm sealed to the edges of the plate (Figure 1). Soil water is extracted via air pressure under controlled conditions. Once air pressure inside the pressure extractor cell is raised above atmospheric pressure, the higher pressure inside the extractor forces excess water through the microscopic pores in the ceramic plate. An internal screen between the ceramic plate and a diaphragm further allows the extracted water to exit the pressure plate cell via an outlet tube running through the plate, which connects this water passage. However, the high pressure air will not flow through the pores in the ceramic plate since the pores are filled with water and the surface tension of the water, at the gas-liquid interface at each of the pores, supports the pressure much the same as a flexible rubber diaphragm (Soilmoisture, 2008). Thus, the pressurized air will not isotopically interact with the soil water.

2.1.1 Experiment 1

The null hypothesis guiding experiment 1 was that soil water sampled along the pF curve has the same isotopic composition over different pF values, which also does not differ from the isotopic label used for spiking the soils.
For rehydration, disturbed oven-dried soil was packed (silty sand: 105 g; clayey loam 121 g) into 100 mL open-bottom, stainless steel cylinders (N=4 for each pF level). The bottom of the cylinders were covered with a sterile polypropylene mesh to allow for water uptake but to prevent a loss of soil material. Cylinders were placed in a water bath, filled with distilled water (DIW) of known isotopic composition (δ2H: −58.4±0.2‰, δ18O: −8.6±0.1‰). The water bath was covered and sealed with a gas-tight lid to prevent evaporation. Soils were left to saturate for two days. The ceramic plates of the pressure extractors were likewise placed in a water bath containing the same type of water and were also left to saturate for two days (following the technical description by Soilmoisture (2008)). Thielemann et al. (2019) showed that for spiking experiments with water of known isotopic composition, 94% of the isotopic change is already manifested after 1 day of equilibration. Afterwards, cylinders including the saturated soils were placed in the pressure extractors and increasing pressure levels were applied (pF: 1.4–4.2) (Appendix A). The water being extracted at each pressure level was directed via an outlet tube (consisting of Swagelok® fittings; Swagelok Company, Solon, OH, US) into a sampling flask (Figure 1). Before applying the next pressure level, ports and tubing were dried with compressed air and a new sampling flask was attached to the outlet tube of the extractor. After each pressure level, the amount of sampled water was determined and soil samples were weighed. After two pressure stages (pF 3 and 4.2), eight samples of each soil type were transferred into glass vials for cryogenic vacuum extraction to remove any remaining water. Cryogenic vacuum extraction was performed using the facility described in Orlowski et al. (2013). Following Orlowski et al. (2018), clayey loam samples were extracted for 240 min and silty sand soils for 45 min at a temperature of 98°C and a baseline pressure of 0.1 Pa. After cryogenic extraction, soils were oven-dried (24 h, 105°C) and weighed again with no significant additional weight loss indicating that the water extraction process was complete.
For isotope analysis, the extracted soil waters were filtered on 0.45 μm disk filters, transferred to 2 mL amber glass vials covered by solid silicone septa, and tightly sealed with Parafilm®.

2.1.2 Experiment 2

With experiment 2, we tested whether soil water collected sequentially over a period of 7 days under the highest pressure level (15 bar) would differ isotopically from the introduced isotopic label. Additionally, we checked whether there is an isotopic exchange between the ceramic plate water and the water to be extracted from the saturated soil samples. Thus, a different rehydration and extraction approach was used. This time, soil samples (N=30 per soil type, 2 per time step) were rehydrated in the same manner as for experiment 1 but with “Lauretana” water (LW, commercial sparkling water; δ2H: −64.6±0.6‰, δ18O: −9.8±0.1‰) and the ceramic plates of the pressure extractors were rehydrated with the same distilled water as in experiment 1. Both waters differ isotopically. Mean water content of the clayey loam samples after rehydration was 44.2±1.2 Vol-% and for the silty sand samples 25.5±0.6 Vol-%. For water extraction, the highest pressure level (15 bar) was directly applied to the pressure extractors and a sequential soil water extraction was performed. Therefore, water was sampled after 10-80 min (every 10 min), 105, 115, 125, 150, 180, 240 min, 1 day, 2, 5 and 7 days. After pressure extraction, cryogenic vacuum extraction was performed on all soil samples (as in experiment 1). We further crushed the ceramic plates used for this experiment and cryogenically extracted the bound water (in the same manner as the clayey loam). Pre- and post-extraction (pressure extractor and cryogenic vacuum extraction) weights and oven-drying weights of soil samples were determined. There was no significant additional weight loss after oven-drying of the cryogenically extracted soil samples (mean±SD weight loss for the clayey loam and silty sand samples, respectively: 0.07±0.03g, 0.03±0.02g).

2.2 Isotope analysis

δ2H and δ18O compositions of extracted soil water samples were measured at the Institute for Landscape Ecology and Resources Management (Justus Liebig University Giessen, DE) on a L2130-i isotope analyzer (Picarro Inc., US). The accuracy of the isotope analyses was ±0.2/±0.8‰ for δ18O/δ2H (determined via repeated measurements of the same sample). All isotope ratios are reported in per mil (‰) relative to Vienna Standard Mean OceanWater (VSMOW) (δ2H or δ18O=(Rsample/Rstandard-1) x 1000 ‰), where R is the isotope ratio of the sample and the known reference (i.e., VSMOW) (Craig, 1961). In-house standards, were run as samples to allow the results to be reported against VSMOW (Nelson, 2000). Isotope data of soil water extracts were checked for spectral interferences (caused by potentially co-extracted organics such as methanol or ethanol) using ChemCorrect™. This software attempts to identify contaminations in water samples both through fitting to a known library of spectral features, and by examining changes in baseline, slope, line-broadening and residual noise of the spectra (West et al., 2011). Further information about this approach is available from the manufacturer (Picarro, 2010). No sample was found to be contaminated by organics.

2.3 Statistical Analysis and Evaluation

We used R for statistical analyses (R version 3.6.3; R Core Team (2014)). All data were tested for normality using the Shapiro-Wilk test. Homoscedasticity was tested using either the Levene’s test for normally distributed data or the Fligner-Killeen test for non-normally distributed data. Cluster analysis based on the furthest neighbor approach using the Euclidean distance as measure was performed in order to identify outliers. This method is more robust for non-normally distributed data. Depending on the type of data (normally distributed and homoscedastic), either Kruskal-Wallis rank sum tests or Analyses of Variances (ANOVAs) were applied and posthoc tests (e.g., Tukey-HSD tests (equality of variances) or Dunnett-T3 tests (non-equality of variances)) were run to determine which groups were significantly different (p≤0.05). Dual isotope (δ18O vs. δ2H) graphical representation was used to compare water extracts from different pF levels (experiment 1) and times of extraction (experiment 2). Statistically significant (p≤0.05) linear regressions were added to dual isotope plots as well as the Global Meteoric Water Line (GMWL: δ2H=8.2×δ18O+11.3‰, as defined by Rozanski et al. (1993)).
Many models for fitting the SWRCs have been proposed (Leong and Rahardjo, 1997). We used the well accepted and widely applied van Genuchten model (Eq. 1) (van Genuchten, 1980) to fit water retention curves to our measured soil water retention data of the two soil types:
\(S=\frac{\theta-\theta r}{\text{θs}-\theta r}=\left(1+{(\alpha h)}^{n}\right)^{-m}\)\(m=1-\frac{1}{n}\) [Eq. 1]
where S is effective saturation (dimensionless), θ is the soil moisture content (Vol., %), while θ s andθ r are the saturated and residual soil water content (Vol., %), respectively. h denotes the soil water potential or pressure head (hPa, also written as cm (H2O)), α is the scaling parameter which reciprocal can be rated as the air entry pressure (cm−1), m and n are dimensionless parameters related to the curve shape. Best fit parameters (α ,m and n ) were estimated for the van Genuchten model of SWRC using “SWRC Fit” (SWRC Fit, 2020) and checked against literature values from Carsel and Parrish (1988). It is, however, well known that the SWRC relationship may vary substantially even for the same soil texture class due to the variation in fitting parameter (α ) and pore size distribution parameter (n ) (Tuller and Or, 2005). To measure the goodness of fit between the measured and the predicted data, coefficients of determination (R2) were obtained for each dataset.