X Water Worlds and how to investigate them: A review and future perspective on in situ measurements of water stable isotopes in soils and plants

The number of ecohydrological studies involving water stable isotopes has been increasing steadily due to 10 technological (i.e. field deployable laser spectroscopy and cheaper instruments) and methodological (i.e. tracer approaches or improvements in root water uptake models) advances in recent years. This enables researchers from a broad scientific background to incorporate water isotope-based methods into their studies. Several isotope effects are currently not fully understood, but essential when investigating root water uptake depths of vegetation and disentangle isotope processes at the soil-vegetation-atmosphere continuum. In particular different viewpoints 15 exist on i) extraction methods for soil and plant water and methodological artefacts potentially introduced by them; ii) the pools of water measured with those methods and iii) spatiotemporal issues related with water stable isotope research. In situ methods have been proposed as an innovative and necessary way to address these issues and are required in order to disentangle isotope effects and take them into account when studying root water uptake depths of plants and for studying soilplant-atmosphere interaction based on water stable isotopes. 20 Herein, we review the current status of in situ measurements of water stable isotopes in soils and plants, point out current issues and highlight potential for future research. Moreover, we put a strong focus and incorporate practical aspects into this review. Finally, we propose an integrated methodology for measuring both soil and plant water isotopes when carrying out studies at the soil-vegetation-atmosphere interface. For all in situ methods, extreme care needs to be taken particularly during set-up in order to obtain reliable data. In situ methods 25 for soils are well established. For transpiration, reliable methods also exist but are not common in ecohydrological field studies due to the required effort. Little attention has been payed to in situ xylem water isotope measurements. Research needs to focus on improving and further developing those methods. There is a need for a consistent and combined (soils and plants) methodology for ecohydrological studies. Such systems should be designed and adapted to the environment to be studied. We further conclude that many studies currently might not rely on 30 https://doi.org/10.5194/hess-2019-600 Preprint. Discussion started: 28 November 2019 c © Author(s) 2019. CC BY 4.0 License.


Figure 1
In addition to the isotope effects summarized in Fig.1, there might be methodological alteration of water isotope values caused by different extraction methods extracting different water pools and organic contamination causing an offset of isotope values 80 when measured with laser spectroscopy (e.g. Barbeta et al., 2019;Martín-Gómez et al., 2015;Orlowski et al., 2016a).
One of the most important steps -and this is where the community seems to agree (Brantley et al., 2017;Dubbert et al., 2019b;Sprenger et al., 2016;Stumpp et al., 2018;Werner and Dubbert, 2016) -to investigate, disentangle, quantify and incorporate those isotope effects is by increasing the spatiotemporal resolution of water isotope measurements at the soil-vegetationatmopshere interface and to integrate measurements of sources and consumers into one framework. In other words, we need 85 combined in situ systems for measuring both soil and xylem water isotopes in a higher spatiotemporal resolution in order to achieve an integrated analysis of soils and plants using a the same methodology and ultimately, measuring the same water pools (Sprenger et al., 2016). While it might be possible to achieve a high temporal resolution by destructive sampling, a number of disadvantages are associated with that: For instance, the experimental plot is disturbed multiple times, small-scale heterogeneity might bias the outcomes and longer-term studies in a high temporal resolution are basically impossible. For 90 plants, a high frequency of destructive sampling might harm and even kil lthe plant. Lastly, time and costs are associated with destructive sampling and over longer periods this adds up to larger sample amounts. Hence, in situ methods are essential for the detection, analysis and interpretation of related isotope effects (see Fig.1).
This review aims on summarizing recent advances in in situ water isotope measurement techniques of for soils (depthdependent and bulk soil) and plants (xylem and transpiration via leaf chambers). We begin with an overview of in situ studies in the compartments soils and vegetation. From thereon, we focus in separate chapters on main issues emerging from the existing sudies, namely i) materials and measurement systems, ii) calibration, standardization and validation and iii) comparability with water extraction studies and measurement of natural abundances of water isotopes. We then conclude and propose ways forward in terms of a combined approach for a consistent, integrated method in order to study processes at the soil-vegetation-atmosphere interface. 100 2 Review: In situ approaches for measuring soil and plant water stable isotopes

In situ soil water isotope depth profiles
A number of early semi-in situ studies (pre laser-spectroscopy) exist, where researchers tried to collect soil water vapour. For the sake of completeness and acknowledging these pioneering efforts, those will be summarized briefly. Thoma et al. (1979) directed water from up to 25m through a molecular sieve, vacuum extracted this water and determined δ 2 H. The determined 105 isotope ratios agreed well with samples extracted from the soil core. With the similar technique, Saxena and Dressie (1984) analysed δ 18 O from soil water vapor in profiles up to 4m depth. Allison et al. (1987) extracted soil water vapor in glass jars and extracted it using dry ice and an alcohol bath. Though the shape of the isotope depth profiles for δ 18 O and δ 2 H showed a similar shape, the values did not match with those obtained by cryogenic extraction. Izbicki et al. (2000) used a similar technique and achieved a better agreement compared to distilled core samples. 110 With the introduction of laser spectroscopy, rapid progress in terms of field deployable systems started. While a gas permeable membrane system was introduced by Munksgaard, Wurster, & Bird (2011) for liquid water samples, the first reported in situ measurement in soils was reported by Herbstritt, Gralher, & Weiler (2012). A microporous hydrophobic membrane contactor was combined with an isotope laser spectrometer system and tested for both liquid and soil water isotopes with a soil column experiment. Though not strictly measuring soil water isotopes, this study created the foundation for the membrane-based 115 techniques, which are the method-of-choice at present. The concepts tested in Herbstritt et al. (2012) therefore can be seen as a baseline for all subsequent in situ soil water isotope studies. The authors stated that 'in order to be able to observe real-time processes in the liquid phase, one needs to continuously transfer the isotopic signal from liquid water to the vapor phase' (Herbstritt et al., 2012). They first tested the applicability of the gas permeable membrane with liquid water samples, determined fractionation factors and isotopic equilibrium fractionation factors for a range of temperatures. The latter, being a 120 crucial for any liquid-vapor equilibrium-based isotope measurement, were determined by fitting the empirical factors a, b and c to the type-1 model of Clark & Fritz (1997, Eq. where α is the isotopic fractionation factor, Tk is the temperature (in K), and a, b, and c are empirical parameters. Membraneinduced deviations from Majoube's (1971) prediction ranging from 0.27% to 0.64% for δ 18 O and from 1.0% to 3.9% for δ 2 H 125 (Majoube, 1971) were reported. In addition, a vapor concentration correction similar to Schmidt et al. (2010) was conducted and recommended by Herbstritt et al. (2012).
In the same year, Soderberg, Good, Wang, & Caylor (2012) presented the first 'real' (i.e. measured in the field) in situ data set from a semiarid environment. The authors aimed at investigating the validity of the Craig-and-Gordon (CG) model and incorporating the effect of soil water potential on kinetic fractionation into the model, which they argue improves the model 130 fit for very dry conditions. Their dataset was tested on a single profile of in situ measured soil water isotope data in a semiarid environment in Kenya. Soil air was drawn from several depths (5, 10, 20, and 30 cm) and directed to the laser water vapor isotope analyser via buried Teflon tubing, with the final 10 cm of each tube perforated and packed with glass wool. Each depth was measured for 90 s. Soil temperature and soil water potential were measured at several depths. Though the in situ field data does not match with the model predicted CG data, the authors support the assumption of liquid-vapor isotopic equilibrium and 135 propose a correction factor to account for dry soils in order to normalize 'free atmosphere' humidity to the soil evaporation.
Most likely, the non-matching isotope values reported in the study were due to the dilution with a high amount of ambient air (400 ml/min) causing a contamination of the drawn soil air.
Volkmann & Weiler (2014) presented an automated system for non-destructive, high-resolution monitoring of soil water isotopes. The authors developed specific probes for the purpose of sampling soil water vapor. The main elements of these 140 probes are a microporous membrane (Porex, Aachen, Germany), a mixing chamber and a sampling, dilution andoptionala throughflow line. The principle of operating the probes is based on drawing soil water vapor into the water isotope analyzer via the sampling line (30 -35 ml/min). The system can be operated in two modes, advection-dilution sampling (ADS) and diffusion-dilution sampling (DDS). In ADS mode, air is simply drawn into the sampling line, and dry air supplied at a lower rate via the dilution line, which causes i) soil water vapor to actively move into the tube (because of the slightly lower pressure 145 inside the probe) and ii) lowering of the water vapor concentration of the whole system. In DDS mode, a throughflow line ending at the lower end (tip) of the probe is added to the system. This throughflow line allows to supply dry gas (N2) to the system at a rate that is the difference of water drawn by the sampling line and supplied by the dilution line. Hence, the pressure difference between soil water vapor outside and inside the probe diminishes, and isotopic exchange occurs only via diffusion. Volkmann & Weiler (2014) were also the first to present validated natural isotope soil depth profiles (i.e. via bag equilibration; 150 Hendry et al., 2015;Wassenaar, Hendry, Chostner, & Lis, 2008). An acceptable agreement was achieved with their system.
For the soil depth profiles, the authors report 95% limits of agreement of +1 ‰ (upper) and -1 ‰ (lower) for δ 18 O and +6 ‰ (upper) and -6 ‰ (lower) for δ 2 H, both for ADS and DDS sampling methods compared to destructive laboratory-based isotope measurements. The range of measured isotope values for δ 18 O and δ 2 H was further in the range of antecedent rainfall isotope ratios. 155 Rothfuss, Vereecken, & Brüggemann (2013) and Rothfuss et al. (2015) tested precision and accuracy of membrane-based in situ in laboratory experiments. Rothfuss et al. (2013) set up an airtight acrylic vessel filled with fine sand, where a custommade throughflow system with a gas-permeable polypropylene membrane was installed. Synthetic dry air was directed through the system, which during the passage would equilibrate with the water isotopic composition of the surrounding sand. The https://doi.org/10.5194/hess-2019-600 Preprint. Discussion started: 28 November 2019 c Author(s) 2019. CC BY 4.0 License. authors used a mass flow controller to subsequently dilute the sample's water vapor concentration to 17.000 ppm, which 160 eliminates the effect of water concentration dependence (Schmidt et al., 2010). They further investigated the effects of soil temperature, sand water content and dry air flow rate as well as fast changes of isotope ratios on δ 2 Hvap and δ 18 Ovap. Although the measured sand δ 2 Hvap showed a significant enrichment relative to those determined at equilibrium according to (Majoube, 1971) at all tested temperatures, this could be linearly corrected for. Rothfuss et al. (2015) proofed that the gas permeable membranes are capable of delivering reliable isotope data over long time periods under laboratory conditions (in the experiment 165 290 days). The same group presented another study, where they measured both soil water and atmospheric oxygen isotopologues in carbon dioxide with their membrane-based system (Gangi et al., 2015). It was shown again that the PP membranes used (Accurel PP V8/2HF, Membrana GmbH; 0.2-µm porosity, 0.155 cm wall thickness, 0.55 cm inner diameter, 0.86 cm outer diameter) did not lead to any isotopic fractionation and was suitable for combined measurements of δ 18 . The experimental results were further modelled using MuSICA (Ogee et al., 2003). The authors proofed that it is possible to 170 simultaneously study oxygen isotope exchange between soil water and CO2 in natural soils which has an immense potential for constraining the atmospheric CO2 budget. However, they state explicitly that further testing of the method is required.
The first in situ study in a semi-arid environment was published by Gaj et al. (2016). Inspired by the system of Volkmann & Weiler (2014), this study represents a proof-of-concept. In northern Namibia, the authors used commercially available polypropylene membranes (BGL-30, Umweltmesssysteme, Munich) and automated their system for measuring of soil water 175 isotope-depth profiles up to 50 cm depth over multiple campaigns, different land covers (bare soil vs. vegetated) and different climatic conditions (dry and post-rain event) . Further, they are the first to study spatiotemporal differences in isotope depthprofiles with their in situ system.The profiles were compared to those obtained by cryogenic vaccuum distillation. While the shape of the isotope depth-profiles were in agreement and the precision of the in situ approach was good (0.8 and 2.5‰ for δ 18 O and δ 2 H, respectively), there were partially large differences between the in situ data and the results based on cryogenic 180 distillation: Between 15 and 50 cm, the RMSE was 3.9‰ for δ 18 O and 9.2‰ for δ 2 H. For the shallow depths RMSE was as high as 7.0 ‰ for δ 18 O and 43.4‰ for δ 2 H. Gaj et al. (2016) interpret these differences potentially caused by an incomplete soil water extraction, the time-lag between sampling soil water vapor and destructive sampling (day vs. night) , rayleigh fractionation caused by the uptake of air from the in situ measurement, natural processes (e.g. hydraulic redistribution), or natural heterogeneity. From today's perspective (author personal comment), it seems very likely that depleted atmospheric 185 water vapor was drawn into the system for the upper soil depths causing the depletion of the in situ measurements. Further, condensation inside of the capillaries or contamination with organic substances could explain the observed differences. An important finding of this study is that for the medium sand the authors studied, the standard deviation of for δ 18 O and δ 2 H was inversely correlated to the soil water content, i.e. the lower the water content of the soil, the higher the standard deviation and

viceversa. 190
A study on the effects of materials and methods for in situ water isotope measurements was presented by Pratt, Lu, Lee Barbour, & Jim Hendry (2016). While the first part deals with the optimization of the vapor bag sampling method (bag type, tubing, relative humidity) and is not scope of this review, the authors also compared in situ analyzed soil water vapor from https://doi.org/10.5194/hess-2019-600 Preprint. Discussion started: 28 November 2019 c Author(s) 2019. CC BY 4.0 License.
depths of up to 180 m. The already existing HDPE tube of two waste sites studied were screened with a 50 mm stainless steel mesh filter and the vapor drawn into the isotope analyzer. The results of the in situ part of the study show partially large 195 differences for depth profiles when comparing the field measurements with the ones obtained by the bag sampling method (up to ±30‰ δ 2 H and ±10‰ δ 18 O). In some depths, the agreement is better (±5‰ δ 2 H and ±1‰ δ 18 O). Pratt et al. (2016) conclude that the in situ results of their study suggest that establishing natural, stable water isotope depth profiles for pore water in thick unsaturated mine waste is challenging.
Another group from the United States developed a system for in situ measurements of soils and has apllied this methodology 200 since in several studies Bowen, 2017, 2019). Their systemup to dateprobably constitutes the most complete in terms of field deployability, calibration and the results reflect that (in particular see . The authors further present a novel approach for correcting their samples by including water and clay content (see chapter calibration). In a primer,  used a vapor-permeable membrane technique and measured soil water isotopes in situ at four sites in North America and validated the water vapor probe method with a vapor direct equilibration 205 method, and vacuum extraction with liquid water analysis. The authors found that the accuracy of the three compared methods in their study is equivalent, with increased ease of use in its application, and sample throughput rates of 7 samples per hour by using the vapor probes. In fact, RMSE of the vapor probe method for δ 2 Hliq values is lower than for direct vapor equilibration (bag sampling method) in matching the δ 2 Hliq values of vacuum-extracted soil water (1.7‰ for δ 2 Hliq values and 0.62‰ for δ 18 Oliq). Hence, trueness for the vapor probe method in their application was greater than for the bag equilibration method. The 210 analyzed profiles were used to investigate the effect of soil texture and pedogenic soil horizons control the shape of the isotope profiles, which are reflective of local evaporation conditions in the soils.

Soil chambers for measuring Evaporation water isotopes (δE)
Isotopic signatures of evaporated soil water vapor (δ 18 OE) can equally be measured in situ, using laser spectrometers coupled 215 to chamber systems. There are both flow-through steady-state (Dubbert et al., 2013) and closed chamber systems (Keeling, 1958) plot approach, see Wang et al., 2013). In a closed chamber the amount of water vapor will, upon closure of the chamber, increase over time, while the isotopic composition of water vapor will change due to the ongoing input of evaporated vapor.
This method was first applied to measure isotopic signatures of respired CO2 (Keeling, 1958 and later adapted to determine the isotopic signature of water vapor (evaporation or also evapotranspiration, see Walker and Brunel, 1990). The Keeling 220 (1958) plot approach is based on two assumptions. First, the isotopic compositions of the source and background air are constant over the measuring period. Second, there is no loss of water vapor from the ecosystem, e.g. during dewfall despite by turbulent mixing. In an open chamber system, ingoing (ambient background) and outgoing (mixed air inside the chamber) air are measured alternately and the isotopic signature of evaporation (or transpiration or evapotranspiration) can be calculated by a mass-balance equation (von Caemmerer and Farquhar, 1981). Isotopic signatures of soil evaporated water vapor have been predominantly used to achieve better understanding of the dynamics of hydrological processes (Braud et al., 2005b(Braud et al., , 2005a(Braud et al., , 2009b(Braud et al., , 2009aHaverd et al., 2011) and mainly to partition ecosystem water fluxes into their components: soil evaporation and plant transpiration (e.g. Dubbert et al., 2014bDubbert et al., , 2013Haverd et al., 2011;Rothfuss et al., 2012Rothfuss et al., , 2010Williams et al., 2004;Yepez et al., 2007Yepez et al., , 2003. Prior to the development of laser spectroscopy there were few studies measuring δ 18 OE directly using cold trapping methods under controlled conditions and -230 to the best of our knowledge -no observations whatsoever under natural conditions. Instead, researchers relied on the Craig and Gordon model, predicting the isotopic signature of evaporated vapor based on isotopic signatures of source (=soil) water isotopic signatures (e.g. Brunel et al., 1997;Wang and Yakir, 2000). Therefore, first approaches to combine soil gas-exchange chambers and laser spectrometers, concentrated on conducting sensitivity analysis of the Craig and Gordon model towards its input parameter (relative humidity, soil temperature, soil water 235 isotopic signature, soil water content and atmospheric vapor isotopic signature (Braud et al., 2005b(Braud et al., , 2005a(Braud et al., , 2009b(Braud et al., , 2009aDubbert et al., 2013;Haverd et al., 2011;Rothfuss et al., 2010Rothfuss et al., , 2012. Conclusively, the correct estimation of the evaporating front is particularly important. Usually the soil layer with the highest isotopic composition is associated with the location of the evaporation front but therefore the spatial resolution of the soil layers should be as small as possible (preferably layer thickness 2 cm or smaller). Sensitivity analysis revealed that precise parameterization of the environmental conditions at the 240 evaporating front, which may diverge tremendously within a few cm of soil depth (up to 8 ‰ from the soil surface to 5 cm depth), is pivotal for correct predictions of δ 18 OE (Dubbert et al., 2013), particularly in arid regions, where dry periods without any precipitation can last several months and the evaporating front can be located in deeper soil layers (Dubbert et al., 2013;Gaj et al., 2016). This does not only highlight the value of direct in situ estimates of δ 18 OE, but also the deployment of in situ soil water isotope measurement set ups when using the Craig and Gordon model. Only spatially highly resolved continuous in 245 situ soil water observations meet the desired requirements necessary to resolve the evaporating front.
Equally important for a correct estimation of δ 18 OE is the choice of formulation for the kinetic fractionation factor αK. Braud et al. (2009a) modelled αK-values during long-term laboratory experiments using SiSPAT-Isotope (soil vegetation atmosphere model) and found αK -values to decrease with soil drying. This is in contrast to the findings of Mathieu and Bariac (1996) who derived an empiric model predicting αK -values to increase from saturated to unsaturated soil conditions. Similar results were 250 obtained from Rothfuss et al. (2015) during a long-term soil column laboratory experiment. Dubbert et al., (2013) achieved a very good predictability of in situ measured δ 18 OE in the field using the Craig and model and the formulation for αK proposed by Mathieu and Bariac (1996) for bare soil and soils with root ingrowth over an entire growing season.
In any case, direct in situ measurements of soil evaporation are mostly limited to laboratory studies conducting sensitivity analysis of the Craig and Gordon model and its input parameters (for a recent paper see Quade et al., 2018). It is often not 255 technically possible to observe δ 18 OE of undisturbed vegetated soil in the field. However, the isotopic signature of evaporated vapor from bare soil patches differs significantly from the isotopic signatures of evaporation from soil with vegetation cover (particularly in grasslands, see Dubbert et al., 2013). Therefore, even today in ecosystem studies aiming i.e. at partitioning evapotranspiration the isotopic signature of evaporation is mostly modelled using the Craig and Gordon equation.

Plant chambers for measuring Transpiration water isotopes (δT)
Direct estimation of stable isotopic signature of plant water fluxes (δ 18 OT) has been difficult prior to the development of laser spectroscopy. Nevertheless, some studies used cold trapping methods to estimate isotopic signatures of plant transpiration (e.g. Harwood et al., 1998) via gas exchange systems and in particular ecosystem evapotranspiration (e.g. Yepez et al., 2007Yepez et al., , 2003 via keeling plots (Keeling, 1958). The main limitation of these early studiessimilar to those for soil water isotopeswas the 265 spatiotemporal resolution. With the advent of laser technology, studies multiplied coupling laser spectrometer to gas-exchange systems of different scales (i.e leaf to canopy level) to assess the isotopic signature of ecosystem water fluxes (i.e. transpiration, evaporation and evapotranspiration, see Wang et al. (2012)  In situ observations of isotopic signatures of transpiration had an immense impact on ecosystem partitioning studies, as they have the advantage of directly measuring the transpiration signature, while destructive sampling techniques observe xylem or leaf isotopic signatures, essentially involving a modelling step to obtain δ 18 OT. A number of ecosystem partitioning studies (e.g., Griffis et al., 2010;Williams et al., 2004;Yepez et al., 2003) even simplified by assuming isotopic steady-state (isotope values of xylem=transpiration), although there is growing evidence that plants rarely reach isotopic steady-state throughout 275 the day (Dubbert et al., 2014aSimonin et al., 2013). Therefore, assuming isotopic steady-state for the purpose of ET partitioning will largely depend on the desired temporal scale (considering non steady-state definitely necessary at sub-diurnal to diurnal scale but unimportant at larger time scales, i.e. weeks or months). In case non steady-state is likely to occur, δ 18 O of transpiration can be modeled using a Dongmann style version of the Craig and Gordon equation (Dongmann et al., 1974).
However, this complicates the partitioning approach tremendously in comparison to direct chamber measurements of 280 transpiration water isotope values, as a large number of additional observations are necessary (in particular, stomatal conductance and Transpiration amounts). Another important consideration in regard to the method of choice (in situ transpiration measurements vs. modelling) is the possibility to sample unfractionated xylem isotopic signatures. For example, herbaceous and grass or agricultural species do not have suberized stems and destructive sampling would rely on leaf water sampling or sampling the plant culm belowground, which is highly destructive and not possible on normal plot sizes. Moreover, 285 while the majority of studies still provide evidence for an unfractionated uptake and transport of xylem water through plants, there is growing evidence of fractionation of xylem water during times of limited transpiration rate (drought conditions, for deciduous species, see e.g. Martin-Gomez et al. 2017).
Similar to in situ soil evaporation isotope observations, in situ observations of isotopic signatures of transpiration have further been used to advance our understanding on water isotope fractionation (e.g. Dubbert et al., 2017;Piayda et al., 2017;Simonin 290 et al., 2013;Song et al., 2015aSong et al., , 2015bSong et al., , 2013. In particular, the leaf water turn-over time has gained research focus. The leaf water turn-over time can effectively be described by stomatal conductance and leaf water volume, is extremely species-specific https://doi.org/10.5194/hess-2019-600 Preprint. Discussion started: 28 November 2019 c Author(s) 2019. CC BY 4.0 License.
spanning from several minutes to several hours (Song et al., 2015a). As the leaf water turn-over time describes the necessary time for a leaf to reach isotopic steady-state (see Simonin et al., 2013;Song et al., 2015a), isotopic steady-state can either be observed for large parts of the day (e.g. in many herbaceous species) or not at all (e.g. in plant species strongly controlling 295 their stomatal conductance, see Dubbert et al., 2017Dubbert et al., , 2014a for an overview).
Direct in situ estimates of transpiration isotopes have also been used to derive root water uptake proportions (Kühnhammer et al., 2019a;Volkmann et al., 2016b) by assuming isotopic steady state and substituting δ 18 OX (xylem) with δ 18 OT. However, given the often very likely violation of steady state assumptions under natural field conditions, this can be recommended only under very tightly controlled laboratory conditions and knowing the leaf water turn-over time. 300

In situ measurements of plant xylem water isotopes
For the direct measurement of plant xylem water isotopes, only two studies are reported up to date. Volkmann et al. (2016) present field observations of xylem water isotope ratios of two adult field maple trees (Acer campestre L.) obtained over several days during a labelled irrigation event using and laser spectroscopy. The obtained in situ data was compared against results from destructive sampling with cryogenic distillation and mass spectrometric analysis. Similar to their in situ soil 305 measurements, (Volkmann et al., 2016a) used the same membrane system to infer the isotope ratios of xylem water. Several holes were drilled into each of the target trees and the gas permeable membranes inserted into those. In order to prevent the intrusion of atmospheric air the outside was sealed with silicone glue. Similar to the soil studies, dry gas (here N2) is provided by a throughflow line and directed to the laser spectrometer via the suction of its vacuum pump.  With the obtained data Volkmann et al. (2016) demonstrated that temporal changes as well as spatial patterns of integration in xylem water isotope composition can be resolved through direct measurement. In both studied trees, diurnal cycles of xylem water isotopes were found. However, the authors could not prove whether this is a true diurnal cycle or introduced through imperfect accounting for temperature-dependent liquid-vapour fractionation at the probe interface. The authors achieved a median precision of 1.1‰ for δ 2 H and 0.29‰ for δ 18 O values (1σ) for an integration period of 120 s. When comparing the in 315 situ measured xylem isotopes with the results obtained by destructive sampling (measured with mass spectrometer) a significant correlation was found for both water isotopes (δ 2 HIRIS=1.26×δ 2 HIRMS+14.51, r 2 =0.86, P<0.0001, δ 18 OIRIS=0.91×δ 18 OIRMS-4.87, r 2 =0.46, P<0.001, robust BSquare-weighted M-regression). However, when taking a closer look at the agreement of in situ and destructive data, partially high uncertainties are apparent (see Fig.2, reprinted with permission).
In addition, the uncertainty (especially of the in situ data) is up to 20 permille for δ 2 H and up to 3 permille for δ 18 O. 320 Marshall et al. (2019) tested an alternative method for monitoring the isotope values of tree xylem and showed that both natural abundances and highly enriched isotope values (labelling experiment) can be monitored over more than two months in situ.
Their approach is based on drilling a hole (which the authors refer to as 'stem borehole') laterally through the complete trunk of a tree and connecting both ends with tight fittings to the CRDS-manifold system. The temperatures within the boreholes were monitored using thermocouples and later used for vapor-liquid conversion of the isotope data. The authors tested their system on two occasions on pine trees: i) in a cut-stem experiment and ii) in a whole-root experiment. They further developed a model to test the feasibility and limits of the borehole method. This included the estimation of the time constants for diffusion of water vapor to and from the borehole wall, and for the flowing airstream in the centre of the borehole (i.e. isotope exchange during the passage of air through the borehole) as well as the prediction of isotope values. For both experiments, Marshall et al. (2019) found close agreement of the source water isotope values provided for the trees, the ones measured in the 'stem 330 boreholes' and the ones predicted by the model. In the cut-stem experiment, it took several hours after a change in water source before this agreement was reached. In the case of the intact-root experiment, it took almost two weeks until source and measured water isotope values agreed. In this experiment, the authors further tested equipping the tree with two stem boreholes.
For the bottom borehole, the deviations to source water isotope values were nearly zero, meaning that the derived xylem water isotopic composition agreed with the source water values for both natural abundance (δ 18 O = -0.1 ± 0.6 (SD) ‰, δ 2 H = 1.8 ± 335 2.3 ‰) and the label phase (δ 18 O = -0.25 ± 0.22 ‰, δ 2 H = 0.09 ± 7.8 ‰). In contrast, the top borehole showed systematic deviations from source water values for both δ 18 O and δ 2 H. δ 18 O xylem values were depleted in 18 O in relation to source water by -2.8 ± 1.5 ‰ and -3.9 ± 0.3 ‰ for the natural abundance and label phase, respectively. In contrast, δ 2 H xylem values were enriched in 2 H as compared to source water by 5.3 ± 3.0 ‰ and 1.9 ± 8.5 ‰. Figure 3 shows the results from their intact-root experiment. 340 With the additional measures taken and the developed model, Marshall et al. (2019) proofed that this deviation was due to non-equilibrium conditions in the borehole due to the small diameter of the upper borehole (relative humidity was 98 ± 2 % for the bottom borehole and 88 ± 3 % for the top borehole).
Finally, the authors measured the sap flow velocity, which was 0.97 + 0.4 cm/hr based on the sap probes (heat-ratio method). 345 Comparing this to the sap flow estimates derived from the stem borehole isotope measurements yields 1.08 cm/hr on average, which is in very good agreement to the sap flow estimates.
In Supplement 1 information on all reviewed studies, details on the setup main findings as well as advantages and disadvantages of the applied methodologies are compiled.

Setup, Calibration and Validation of in situ measurements of soil and plant water isotopes 350
Apparent from the review of studies is that in situ measurements are still in development stage; hence, applied methods and approaches vary greatly. In this chapter, we pick out key aspects that need to be considered and propose a way towards more comparable and homogeneous setups. The biggest and most critical issues emerging from the existing studies are i) the materials and approaches used for sampling the air; ii) the calibration of the system iii) the avoidance of condensation and iv) how to validate the in situ data compared to other methods and how to interpret it best. We focus in this chapter on methods 355 for obtaining in situ depth profiles of soil water isotopes and the measurement of xylem water isotopes due to the fact that https://doi.org/10.5194/hess-2019-600 Preprint. Discussion started: 28 November 2019 c Author(s) 2019. CC BY 4.0 License. methods for monitoring bulk soil evaporation and transpiration at the leaf level have been reported previously in detail (Soderberg et al., 2012, Song et al., 2015.

Materials and approaches for sampling soil water vapor
Most of the reviewed studies used gas permeable membranes (e.g. Accurel PP V8/2HF, Membrana GmbH; 0.2-µm porosity, 360 0.155 cm wall thickness, 0.55 cm inner diameter, 0.86 cm outer diameter) with an inlet and outlet (e.g. Gaj et al., 2016;Rothfuss et al., 2013;Volkmann and Weiler, 2014). Some groups built the probes themselves Rothfuss et al., 2013) others used factory-made probes (Gaj et al., 2016;Volkmann and Weiler, 2014). The important point is that with all of these membrane systems it was shown that no isotopic fractionation occurs due to the membranes; hence this type of probes is suitable in general. Self-made membranes are much cheaper and can be adjusted to the application (i.e. length 365 of exchange path, number and position of in-and outlets, size of capillary/tubing connected).
The number of in-and outlets of the probes depends on the measurement approach. In general, two of these exist: i) a pullonly system (e.g. Volkman & Weiler, 2014), where water is drawn simply through the gas-permeable membrane by the force of the vaccuum pump of the laser spectrometer. Such a system in fact requires only one capillary and thus is the simplest of the setups. However, it should be considered that a notable amount of air is drawn from the media to be measured (soil/plant) 370 and might cause a dry-out when applied extensively (e.g. measuring one soil depth all the time). This could be especially relvant for applications tree xylem as it might increase the risk of cavitation and hence, damage to the plant. The extracted volume of soil water vapor can be easily calculated by multiplying the flow rate with the measurement time. Because of this, most studies use ii) probes with two capillaries: one in-and one outlet (e.g. Rothfuss et al., 2013). This changes the approach drastically, because now dry air is now pushed through the inlet (via a dry gas supply) entering the 375 membrane from one side and leaving it at outlet. During the passage of the dry air, water from the (saturated) soil air diffuses into the membrane and exchanges isotopically through the gas-permeable membrane. Ideally, saturation is reached while the air is passing through the membrane. Only in this case equilibrium fractionation can be assumed. With the push-through method, the system is under overpressure, which has two consequences: First, one needs to get rid of this overpressure before it enters the laser spectrometer to avoid damage. This is commonly achieved by an access tube just before the analyzer inlet. 380 Second, due to this overpressure, the chance of external air entering the stream is excluded, which is a clear advantage over the push-only method, where it needs to be assured that all connections are tight.
The pull-only system can also be operated with an additional inlet capillary/tube connected to a reservoir with drying agent. Doing so, atmospheric or dry air (via passage through a drying agent) is drawn into the gas permeable probe and equilibrated therein during the passage. Flow rates, however, are not adjustable using this approach. 385 It needs to be carefully decided which approach to use and, ultimately, this depends on the application (e.g. tracer test, measuring natural abundances, long-vs. short-term measurements) A pull-only system is technically much easier to build, install and maintain and also cheaper, but it it is critical to avoid external air to enter the system at any of the connections. The push-through approach is more flexible and flow rates can be adjusted, but it requires more maintenance, connections (for https://doi.org/10.5194/hess-2019-600 Preprint. Discussion started: 28 November 2019 c Author(s) 2019. CC BY 4.0 License.
provision and control of dry air at the inlet), and valves. Due to the fact that the system is operated at overpressure, the risk of 390 drawing outside air (via non-tight connections) into the lines is minimized. Figure 4 depicts a schematic of an in situ soil water isotope system (reprinted with permission from Oerter and Bowen 2017).

Saturation of water vapor, condensation and dilution
Condensation (or better: avoidance of it) is the most critical practical issue for all in situ approaches, regardless if soils or 395 plants are measured. If condensation occurs inside of the tubing or inside the chamber, the isotope values measured will be subjected to Raleigh fractionation and hence, do not represent the isotope value of the medium that is subject to measurement.
Hence, it needs to be assured that the water vapor pressure in the sampling line is never exceeding the saturation water vapor pressure. In the reviewed studies, this issue is resolved in three different ways:

i)
Dilution with dry air directly in the membrane system (Volkmann et al., 2016a;Volkmann and Weiler, 2014) 400 or shortly after Bowen, 2017, 2019;Rothfuss et al., 2013Rothfuss et al., , 2015. This way, the water vapor concentration of the system is lowered and condensation less likely.

ii)
Heating of the tubing (suggested by Gaj et al., 2016). Assuring that the temperature of the transport line is always warmer than the temperature at the location where isotopic equilibration occurs will avoid condensation to occur. Even in warm climates this might be necessary as solely the temperature difference between the 405 location where water vapor is equilibrated (i.e. inside of the gas permeable probe) and the sampling line is decisive if condensation occurs or not (refer to section recommendations for further elabortion on this issue). iii) Flushing the system with dry air prior to the measurement (Kühnhammer et al., 2019, under review;Volkmann and Weiler, 2014). Flushing the tubing system with dry air prior to each measurement cycle will remove previously condensed water from the transport line. 410 Condensation is very likely to occur (whenever temperature outside of the in situ system is higher than inside) and will affect the water isotope data tremendously. Hence, it is of utmost importance to avoid it at all costs. However, it is not always easy to identify. For this reason, we present three examples of (raw) isotope measurements which depict i) a 'good' measurement cycle; ii) a measurement cycle initially influenced by condensation, but then turning into a clean measurement once the condensation dissappears and iii)/iv) bad measurement cycles with condensation affecting the complete data in Fig. 5. 415

Calibration protocols
The calibration of water isotope values is a crucial point, and it is more complex and error-prone when measuring water vapor isotopes in situ compared to liquid water samples. It is generally comprised of the following steps: i) Standard preparation;
Normalization to VSMOW scale. There is a great variety in the existing studies on how (and even if) each of these steps were addressed in the reviewed papers. The subsequent section highlights the key points based on the reviewed papers in terms of calibration procedures. We then put a special focus on the approaches presented by  and , who developed a novel, innovative method for the calibration of in situ measurements of soils. 425

Isotope Standards
Ideally, isotope standards are prepared in the same medium that is measured (personal comment, P. Koeniger). That means, use soil standards when measuring soil water isotopes, use water standards when measuring liquid water samples. The clear advantage of this is that effects such as mineral-mediated isotope effects will be incorporated into the calibration procedure in a way that the standard will be affected in the same way as the measured sample. The disadvantage is that the preparation of 430 soil standards requires more practical effort. Soil from the site of interest needs to be collected, oven-dried and placed in suitable standard bags or containers. Subsequently this soil needs to be spiked with the isotope standards (Gaj et al., 2016;Rothfuss et al., 2015). Ideally, soil from different horizons is used for that as well, because the soil texture and, hence, isotope effects might change throughout the soil profile . In addition, a range of water contents should be covered in the calibration process. This makes the 435 calibration using soil standards labor-intensive and multiplies the number of standards to be measured (different soil horizons x different standards x different water contents). In contrast, using water standards for calibration is rather straightforward, as only different water vapor concentrations need to be injected for the calibration. This can be done either using a system for vapor injection (e.g. a standards delivery module or nebulizer) or simply placing the water standards in bags or containers and measuring the headspace. In the latter case, calibration of water vapor concentrations needs to be controlled via diluting the 440 headspace with dry air to obtain lower H2O ppm values. The big disadvantage of using water standards is that soil induced isotope effects are not incorporated at all and this can lead to notable errors in the corrected isotope values later on. Hence, for best isotope data we recommend soil standards when measuring soil water isotopes (depth profiles and evaporation) and water standards when measuring in situ plant water isotopes (transpiration and xylem) or atmospheric water vapor. In regard to chamber based measurements, correction has mostly been applied liquid injected standards in the past, however, we 445 recommend to, when integrating chambers in a larger in situ framework, to use water equilibration standards instead. Obviously the background dry-gas is of major importance here, as the air matrix of the standard should be the same as that of the sample.

Correction for water vapor concentration
Because of the influence of different water vapor concentrations on measured isotope ratios (Picarro, 2015;Schmidt et al., 2010), a correction needs to be performed. A linear best-fit equation can be fitted if a standard of known isotope composition 450 is measured at different water vapor concentrations. The slope and intercept of the best-fit line through these points are the two values that are used to post-process vapor delta values with variable water concentration (Picarro, 2015). Furthermore, the water vapor concentration is affected by the temperature of the media-of-interest but also soil moisture and relative humidity. Indirectly, the flow rate chosen by the user also affects the latter (if flow rates are too high, saturation will not be reached). The interplay of those factors is complex and no trivial to account for (refer to discussion for elaborations on 465 this issue).

Other corrections (mineral mediated fractionation, organic contamination, carrier gas and biogenic matrix effects)
Recent research has shown that especially in clay-rich soils, an offset can be produced due to tightly bound water (Gaj et al., 2017;Newberry et al., 2017;Oerter et al., 2014). This creates a real challenge for any soil water isotope measurements and 470 was discussed heavily (Orlowski et al., 2013(Orlowski et al., , 2016b(Orlowski et al., , 2016aSprenger et al., 2016). Up to date, it is not clear how to best handle these additional factors. A stated above, a preparation of isotope standards in the same soil that is to be measured seems to be the most promising approach, and  provide an innovative procedure to calibrate their data (see section down below).
In addition, spectral contamination of laser spectroscopy caused by organic compounts has been discussed frequently and was 475 recognized as a major source of error when extracting water from plant tissues (Barbeta et al., 2019;Brantley et al., 2017;Martín-Gómez et al., 2015;Millar et al., 2018;Newberry et al., 2017;Penna et al., 2018;West et al., 2010;. It is not known up to date, if this plays a role for in situ approaches (refer to discussion). Volkmann et al. (2016) speculated in their study that organic contamination might be one of the reasons for the observed discrepancies in their dataset. For liquid water samples, a method for correcting for the influence of organic substances exists (Lin et al., 2019;Schultz et al., 2011;Wu et al., 480 2013). Thereby, deionized water is spiked with varying amounts of methanol and ethanol to create correction curves for δ 18 O and δ 2 H. An adaptation of this method is theoretically feasible for water vapor measurements, but has not been tested until today (personal communication, M. Hofmann, Picarro). It should be noted, however, that methanol and ethanol are not the only possible contaminants and others might influence the absorption spectra. Generally, it is advisable to perform a check if organic contamination for the particular set of samples is an issue using the pertinent software (e.g. Chemcorrect). If this is the 485 Finally, the issues of carrier gases and biogenic matrix effects have been raised recently. However, those can be corrected for rather straightforward (Gralher et al., 2016(Gralher et al., , 2018.

Drift correction 490
As for the measurement of liquid water samples, it is recommended to always use a drift standard that can be measured either after each run (e.g. after measuring one soil profile or a set of tree replicates) or after a certain time. A linear correction similar to the regression for water concentration can than be performed.

Conversion of vapor to liquid values
All of the presented studies are based on isotopic exchange between the air outside and inside of the gas permeable probe. 495 Ideally, equilibrium fractionation is achieved during the passage of the air through the membrane. The isotope value of water (soil or xylem) can then be calculated applying the well-established equations for equilibrium fractionation (see Clark and Fritz, 1997;Horita and Wesolowski, 1994;Majoube, 1971): where α is the fractionation factor, T is the temperature in Kelvin and δ the isotope ratio of water vapor and liquid water, respectively. The empirical factors a, b and c are tabulated in the above cited literature and commonly used as a= 28.844, b= -76.248, c=52.612 for δ 2 H and a=1.137, b=-0.4156, c=2.0667 for δ 18 O. As per equations 2 and 3, the temperature is needed for 505 this conversion. Hence, it needs to be measured at the location of exchange (at the gas permeable probe). A conversion of vapor to liquid values is also possible when the water vapor is not saturated (via equal treatment principle of isotope standards), but is not recommended because, for soils, for example, the isotope standards would be needed to be prepared with the exact soil moisture and temperature as the sample to be measured. This becomes very laborious because soil water contents are highly variable with depth and time. 510 The final step issimilar to liquid water isotope measurementsthe Normalization to VSMOW scale: (we spare the equation here is this procedure is widely known and sufficiently documented).

Validationcomparing apples and pears?
As shown in the succeeding chapter, calibration protocols for addressing the abovementioned steps vary greatly. Not always all the steps are addressedeither because it was not relevant for the particular investigation or because it was simply neglected. 515 Thus, it is necessary to introduce a way of assessing the measurements. Across studies, trueness, precision and reproducibility of in situ methods are generally good. For an evaluation of accuracy, the reviewed publications compared the obtained isotope values either with cryogenically extracted samples (Gaj et al., 2016;Soderberg et al., 2012;Volkmann et al., 2016a;Volkmann and Weiler, 2014), results from direct equilibration (bag) methods (Pratt et al., 2016) or both . Further, theoretical approaches (mass balance calculations and modelling) have been applied to reproduce the in situ measurements 520 (Rothfuss et al., 2013(Rothfuss et al., , 2015Soderberg et al., 2012). The agreement of soil profiles extracted with vacuum extraction at deeper soil layers is generally better. In the upper soil layers, however, partially large differences (> 10 ‰ in δ 2 H) are encountered.
Possible reasons include contamination with organic compounds (cryogenic distillation) or interference with atmospheric air when using a pull-only system. In the light of recent findings suggesting that water from cryogenic vacuum extraction and in situ approaches represent different water pools (Orlowski et al., 2016b;Sprenger et al., 2016), this way of validation might not 525 be suitable. Instead, validating in situ data with the established direct vapor equilibration method by deploying aluminium or plastic bags and measuring the headspace air, should deliver true means of comparing the data.
The validation of the only in situ study of xylem water isotopes so far (Volkmann et al., 2016a) yielded in good results in terms of precision (median of 1.1‰ for δ 2 H and 0.29‰ for δ 18 O) and reproducibility (median of 2.8‰ for δ 2 H and 0.33‰ for δ 18 O).
Diurnal variations in both isotopes did not correlate with those of temperature estimates for the different probes; hence, the 530 authors suggest measuring the temperature inside of the probe in the future. They further state that when comparing the values obtained in situ with cryogenic extractions and subsequent measurement using IRMS, a significant correlation between the two exists. For a short period of their measurements, they achieved a good agreement and little systematic difference for δ 2 H (pre-irrigation, 0.9 ±1.8‰). For δ 18 O, a clear inter-method bias of -4.3± 0.7‰ was found. The discrepancy in their other data was hypothetically attributed to contamination by volatile organic compounds (VOC's), lateral mixing (through intervessel 535 pits), axial dispersion and the time lag between irrigation water arrival at the twig/crown versus trunk level.
The closest agreement of the reviewed manuscripts when comparing in situ derived data with other methods was achieved in the study of . Both in terms of measurement and data handling, their methodology appears to be the most complete at present. In addition, the authors propose a novel, innovative way of calibrating in situ data of soil water isotopes.
A reprint of their isotope depth-profiles determined with gas-permeable soil gas probes, direct equilibration and vacuum-540 extracted profiles are shown in Figure 6. ii) add different amounts of isotope standard with known δ 2 H and δ 18 O values to obtain a range of water contents (e.g. 5%, 10%, 20% water content x 2 standard solutions x 3 depth intervals = 18 calibration bags) iii) add soil temperature sensors to standard bags/containers iv) measure standard preparations under a range of temperatures (e.g. 0 -35°C) 550 v) perform multi-linear regression analysis (e.g. nlme package in R) in order to estimate theoretical liquid water standard values using the parameters Measured Vapor Isotope Value (δ 18 Ovap and δ 2 Hvap), soil moisture content (GWC) and temperature (TEMP); other parameters such as clay content or water vapor concentration might be added vi) selection of best fit equation for estimation of δ 2 Hliq and δ 18 Oliq of the isotope standards (in Oerter et al., 2017: 555 δ 18 Oliq = 9.954 -0.163×TEMP + 0.002×TEMP 2 + 13.386 × GWC + 1.051×δ 18 Ovap; δ 2 Hliq = 120.128−1.255×TEMP + 0.008 × TEMP 2 + 1.138×δ 2 Hvap) vii) statistical analysis: Goodness of estimation? Which parameters explain variation in estimated liquid isotope values best? viii) Application of final equations to dataset, consequent check of isotope standards throughout measurement campaign 560 using derived equations A procedure like this has several advantages: First, it uses additional information that might have influence on the measurements, such as clay contents, soil temperature and water content. Second, it incorporates these information into one procedure, namely a multi-linear regression. Third, the vapor-liquid conversion that exist in several forms can be avoided. It also might be used as comparison to liquid isotope values calculated via the statistical approach. Finally, the derived 565 relationships can be objectively assessed using goodness-of-fit measures, tested throughout the measurement period, and, if required, adopted later. Thus, we highly recommend this way of calibration and derivation of liquid water isotope values for future studies.

Summary and Outlook
The goal of this review was to summarize the current state of in situ approaches for measuring the isotope ratios of soil water, 570 evaporation and transpiration (in both leaves and xylem) and point out current issues and challenges. Based on this, we propose toward future applications of combined in situ investigations at the soil-plant-atmosphere continuum.
In situ measurements are an inevitable step for any holistic study within the critical zone. The current design of many ecohydrological studies is still based on destructive sampling at discrete points in time and space. The number of artefacts (potential isotope effects) and methodological constraints (limited spatiotemporal resolution, issues of measuring different 575 water pools with different extraction methods) associated with that (refer to introduction) is increasingly questioning Another example is the high temporal resolution that can be achieved with such measurements which resolves the issue of lag time and enables the investigation of non-steady state conditions. Hence, in situ methods will be highly useful for any study involving rapid changes of environmental conditions, e.g. root water uptake studies, water partitioning, night-time transpiration, etc. They will also benefit long-term studies, such as monitoring combined reaction of soils and plants to droughts or extreme events. Moreover, high frequency in situ monitoring can elevate tracing of the water cycle via isotopic labelling 585 ( 2 H2O or H2 18 O) to a new level and will lead to improved parameterization for a novel generation of models (we spare modelling approaches in this review, but refer to recent developments, e.g. Braud et al., 2013Braud et al., , 2009Dubbert et al., 2014;Hirl et al., 2019;Kuppel et al., 2018). The same is true for isotopic mixing models. Another aspect that can be studied in much greater detail than before is the process of hydraulic redistribution (Burgess et al., 1998), to name one. Combined with labelling approaches, it might be able to quantify its relevance and impact on a much greater spatiotemporal scale. Let alone these 590 examples, online methods comprise immense potential for future applications.
Having that said, it always should be carefully evaluated, if an in situ approach is required for the purpose of study -or if destructive sampling is sufficient. When carrying out in situ studies, the aim of the study determines the design of the system to be used and a good starting point would be to clarify the following aspects: i) Is the particular study a long-term study (weeks to months) or rather short-term (days)? 595 ii) Is the goal to obtain data in a high temporal spatial resolution (or both)?
This aspects aims to define if the system needs to be portable or rather stationary.
iii) Is it a tracer experiment or is the goal to obtain natural abundances of soil/plant water isotopes?
The setup of any in situ system is neither simple nor easy at present. The complicated technical setup and calibration process might be a reason why only a few research groups have employed in situ studies so far. We hope to shed light on some of the 600 technical aspects involved and clarify those through this review.
In situ approaches for monitoring depth-dependent soil water isotopes employing gas permeable probes have advanced tremendously in recent years. It now seems feasible to obtain measurements of natural abundances of soil water isotopes in a high temporal frequency. For monitoring the isotopic ratios of xylem water continuously, on the other hand, there is only one existing study employing isotopic labelling (Volkmann et al., 2016a) Future efforts should be directed towards testing and 605 improving the methods suggested therein and develop novel approaches with the ultimate goal to measure natural abundances of plant water isotopes in situ (Beyer et al., 2019;Kühnhammer et al., 2019b;Marshall et al., 2019). Subsequently, continuous soil and plant water isotope measures should be combined. In situ monitoring of isotopic signatures of transpired and evaporated water vapor are advanced and have mainly been employed in frameworks focusing on partitioning of ecosystem evapotranspiration (Dubbert et al., 2013(Dubbert et al., , 2014a(Dubbert et al., , 2014bRothfuss et al., 2012)or studying isotopic fractionation during soil 610 evaporation and leaf water isotope signatures (Cernusak et al., 2016;Song et al., 2013Song et al., , 2015aWu et al., 2013).
They have also been used in ecohydrological studies tackling questions, such as root water uptake depths (e.g. Volkmann et al., 2016a). However, given the critical and under natural conditions mostly violated assumptions of isotopic steady-state of https://doi.org/10.5194/hess-2019-600 Preprint. Discussion started: 28 November 2019 c Author(s) 2019. CC BY 4.0 License. transpiration (i.e. isotopic signature of transpired vapor not equal to that of xylem water; see e.g. (Dubbert et al., 2014b;Piayda et al., 2017;Simonin et al., 2013)), this can be critical at least under natural ambient conditions. 615 Despite the great advances in monitoring depth-dependent soil water isotopes in situ , there is no generally accepted calibration protocol existing yet (such as van Geldern and Barth (2012) for water samples). Hence, homogenization of calibration protocols and validation are required. We propose here to make such a development based on the ideas of , which is -in the authors' opinion -the most complete of all currently existing approaches. It also provides an objective way of handling the data (via statistical measures) and is very flexible in including/excluding additional factors that might be 620 relevant (e.g. mineral-mediated fractionation). In terms of calibration, we further suggest that laboratory standards are provided using the same media that is to be measured (e.g. use standards prepared and measured in soils when measuring soils in situ) in order to fulfil the assumption of identical treatment principle, which has been violated in a number of studies. We confronted the developers of the bag equilibration method (Wassenaar et al., 2008) with this question and received the following response: 'We and others have wrestled with this, and you are correct our original publication is technically not fully an identical 625 treatment. I suppose the real question is how much does either approach matter in practice vs its convenienceare we talking only 10th's of a permil difference (not an issue) or a lot more (worrisome)?' (L. Wassenaar, personal communication).
However, the response also states that using soil standards also has pitfalls. 'It is also not identical treatment if you dry and wet soils or sand with lab standard waters, some soils may more potential for bound residual water or isotope exchange with clay particles, for example, or have soil properties that differ a lot from the field samples.' (L. Wassenaar,personal 630 communication). For this reason, an ideal preparation of soil standards is not existing at present. However, running pre-in situ laboratory tests using soil from different depths (e.g. A and B-horizon) from the site to be measured, oven-dry it, spike it with different water contents and measure it over a range of temperatures and water vapor concentrations will give a sound baseline for calibrating the on-site data. For the field calibration, soil standards (e.g. two to three) for each soil horizon should then be prepared and measured for each sequence in the field. We further propose to install TDR probes in each of the standard bags 635 to keep track of the water content and temperature which is needed for the calibration.
For validation, it has been shown that a comparison of cryogenically extracted samples, although this has been the standard method for decades, with equilibration methods is not feasible for soil samples because different water pools are measured with the two approaches. The same might be true for plant samples. There is an urgent need to develop alternative ideas. For soils, a comparison of in situ data with destructive sampling and using the bag equilibration method might be a way. However, 640 the issue of spatial heterogeneity between the two measures remains. For plants, the bag equilibration method might also be feasible but has not been tested thoroughly.
For both soils (e.g. the upper soil layers) and plants, the effect of organic contaminants (such as volatile organic compounds -VOC) on in situ measurements needs to be evaluated and measures developed to correct for (post-correction). Such might be included into the multi-step procedure suggested by . As stated, a method for correcting liquid water samples 645 for the influence of organic substances already exists (Lin et al., 2019;Schultz et al., 2011;Wu et al., 2013) and could be easily https://doi.org/10.5194/hess-2019-600 Preprint. Discussion started: 28 November 2019 c Author(s) 2019. CC BY 4.0 License. adopted to vapor-phase measurements. However, it needs to be determined before if contamination even plays a role for in situ obtained data.
Another point related to calibration we recommend is the establishment of a way of evaluation if equilibrium conditions prevailed at the site of isotope exchange during the in situ measurement (e.g. inside of the gas permeable soil/tree probe). All 650 reviewed studies presented herein use some sort of equilibrium water -vapor conversion (e.g. Horita et al., 2008;Majoube, 1971). Only one of them (Marshall et al., 2019) evaluated if this assumption actually was true for their particular setup (flow rate, exchange length, etc.).
Relative humidity corrected to borehole temperature can be defined as the ratio of "water vapor concentration" (in ppmv) measured by the cavity-ringdown spectrometer compared to the saturated specific humidity at stem temperature T (measured 655 using a thermocouple or PT100 sensor) to obtain an estimate of relative humidity in the borehole. If these two (roughly) match, it is likely that the chosen parameters of the (physical) system are suitable to reach saturation under the given conditions. It also reveals potential condensation under the given environmental conditions. Ideally, relative humidity h should range between 0.9 and 1.0. When h is substantially lower than 0.9, the dry air pushed into the borehole did not approach saturation; hence, the flow rate would need to be adjusted downward. We recommend for any system to check h for evaluating if the 660 defined settings of the physical setup are suitable. This concept is applicable to both push-through and pull-only setups.
One might argue that via equal treatment principal, saturation is theoretically not necessary because it can be calibrated.
However, this would require, for instance for soil samples, a preparation of soil standards with the exact same conditions as the measured soil depth (water content, temperature), which is practically not feasible.
In the final chapter, we propose a combined soil-plant in situ monitoring system which -in the authors' opinion -represents a 665 holistic way of investigating ecohydrological processes at the interfaces of soil, vegetation and atmosphere.

One system, one methodology -A call for combined in situ studies
The authors of this study have been involved into the development of in situ for nearly a decade. As a combined conclusion this literature review and their own experiences, an 'ideal' system is presented in Fig. 7.  The system depicted in Fig 6. combines measurements of all compartments covered in this review. A setup like this would enable one to monitor the complete cycling of water through the soil-plant system: i) gas permeable probes for measuring depth-dependent soil water isotope ratios (supported by soil moisture/temp. sensors for the equilibrium calculations); ii) soil chambers for monitoring the isotope composition of evaporation; iii) stem probes or stem boreholes (supported by thermocouples for the equilibrium calculations); iv) leaf chambers for monitoring the isotope composition of transpiration and 675 finally the monitoring of atmospheric water vapor. Ideally, all these fluxes are all controlled by one valve/manifold system.
Through the inlet of each measurement stream, dry air with the required flow rate (MFC 1) can be directed through the probes/chambers. At the same time, it can be used to flush the systems prior to the measurement sequence with dry air (diving air, synthetic air or N2 for removal of condensed water in the lines). The equilibrated water vapor then is send back through the manifold and to the water isotope analyser. A second mass flow controller (MFC 2) offers the opportunity to dilute the sampling air if the water vapor concentrations are too high (less precise values above 30.000 ppm). Any connection is an opportunity for leaks. The system is therefore limited to as little connection pieces as possible (i.e. one piece of Teflon tubing or stainlesssteel capillary from the probe/chamber to valve system/manifold). The excess tube avoids the possibility of overpressure at the analyser inlet. The calibration unit consists of a user-selected number of soil standards for the soil measurements and water standards for the plant water isotope measurements. Additional (optional) components might include a higher number of 685 monitored trees and/or soil profiles (heterogeneity), sap flow probes, stem water content sensors and for the soils, matric potential sensors. Though the depicted setup is constructed as push-through system (dry air is pushed through the compartments to be measured and equilibrated therein), it can be operated in pull-only mode as well.
When reading through this explanation, the reader probably gets the impression that this is very complicated and complex.
Admittedly, it is. It is, thus, the task of the community to further simplify in situ measurements. However, before this happens, 690 several holistic studies need to be carried out and evaluated. Lastly, we encourage the community to carry out and test in situ systems. The increased technical effort for setup is often compensated by far with the higher spatial (if using probes as mobile version) and temporal resolution.
Lastly, it needs to be clear to everybody applying in situ methods that a higher uncertainty has to be expected when working with such methods. While it should be certainly be decreased as much as possible, it is equally as important to communicate 695 those uncertainties. Many of the 'old' studies are employing a very low number of samples, for instance for plant source water studies. Many of those publications end up with clear statements, but completely neglect the dynamic character of natural systems. Thus, only a part of the story is reported. In order to improve the understanding of ecohydrological processes it is inevitable to develop ready-to-use in situ monitoring systems; it is crucial for the community to further develop such methods and make them available for a larger group of researchers and practitioners in the near future. 700