EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Simulating soil-plant-atmosphere interactions for sub-daily in situ observations of stable isotopes in soil and xylem water to assess two-pore domain model hypothesis

Fabian Bernhard1, Stefan Seeger2, Markus Weiler2, Arthur Gessler1,3, and Katrin Meusburger1
Fabian Bernhard et al.
  • 1Swiss Federal Research Institute WSL, Switzerland
  • 2University of Freiburg, Freiburg, Germany
  • 3ETH Zürich, Zurich, Switzerland

Recent advances in stable isotope measurements within the soil-plant-atmosphere continuum have paved the way to high-resolution sub-daily observations of plant water supply (Stumpp et al. 2018, Volkmann et al. 2016a, 2016b). It seems time is ripe for in-depth assessments of long-standing yet much-debated assumptions such as complete, homogenous mixing of water in the vadose zone (“one water world” versus "two water world") or absence of fractionation during root water uptake and vascular transport in plants.

Information on the nature of these processes contained in high-resolution data sets needs to be exploited. One way to test hypotheses and thereby advance our understanding of soil-plant water interactions is by analysing observations with numerical simulations of the system dynamics – a method also known as inverse modelling. By evaluating the model performance and parameter identifiability of different model structures, conclusions can be drawn regarding the relevance of the modelled processes for reproduction of the observations. Testing two different models allows thus to assess the impact of the difference.

We develop a framework for numerical simulation and model-based analysis of observations from soil-plant-atmosphere systems with a focus on isotopic fractionation. A central objective is to facilitate the evaluation of different model structures and thus test model hypotheses. This can assist development of models specifically tailored to the intended purpose and available data. The framework will first be tested with the "SWIS" model presented by Sprenger et al. (2018).

As an illustration of the framework, we will test the model performance on a dataset of continuous, in situ observations of stable isotopes in xylem water of beech trees and soil water in four depths combined with observations of soil water content. The model assumes one-dimensional soil water flow taking place in one or two separate flow domains for tightly and weakly bound pore water. These two water pools are separated by a matrix potential threshold and isotopic exchange is modelled only through the vapour phase. Root water uptake is parametrised using the Feddes-Jarvis model. First results allow to assess the relevance of the two-pore domain hypothesis for the different soil depths and xylem water.


Sprenger, M., D. Tetzlaff, J. Buttle, H. Laudon, H. Leistert, C.P.J. Mitchell, J. Snelgrove, M. Weiler, and C. Soulsby. 2018. Measuring and modeling stable isotopes of mobile and bulk soil water. Vadose Zone J. 17:170149. doi:10.2136/vzj2017.08.0149

Stumpp, C., N. Brüggemann, and L. Wingate. 2018. Stable isotope approaches in vadose zone research. Vadose Zone J. 17:180096. Doi: 10.2136/vzj2018.05.0096

Volkmann, T.H., K. Haberer, A. Gessler, and M. Weiler. 2016a. High‐resolution isotope measurements resolve rapid ecohydrological dynamics at the soil–plant interface. New Phytologist, 210(3), 839-849.

Volkmann, T.H., K. Haberer, A. Gessler, and M. Weiler. 2016a. High‐resolution isotope measurements resolve rapid ecohydrological dynamics at the soil–plant interface. New Phytologist, 210(3), 839-849.

How to cite: Bernhard, F., Seeger, S., Weiler, M., Gessler, A., and Meusburger, K.: Simulating soil-plant-atmosphere interactions for sub-daily in situ observations of stable isotopes in soil and xylem water to assess two-pore domain model hypothesis, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-17975,, 2020

Display materials

Display file

Comments on the display material

AC: Author Comment | CC: Community Comment | Report abuse

Display material version 1 – uploaded on 04 May 2020
  • CC1: Comment on EGU2020-17975, Valentin Couvreur, 06 May 2020

    Hi Fabian,
    Thanks again for your question in the chat room. Actually, I had in mind slide 5 when replying. You are right, in slide 3, each point of d18O_tiller (red) and d18O_leaf (green) was taken from 3 different plants and had to be merged to obtain enough water. We were hoping that population variability of xylem water d18O would be sufficiently low to get a representative signal in time, which is why points are connected like in a time series. Our simulated results suggest that population variability of xylem water d18O dominated by far the temporal variability. This is due to a combination of very steep gradient of labeled water at the bottom of the profile, and slight differences of rooting depth (standard deviation less than 10 cm).

    Leaf water potentials were also taken on individual plants, and we conclude that the signal was rather uniform within the population, so that the time series would be representative of temporal variability.
    Transpiration was measured for the whole setup, based on its weight.
    Does that answer your question?

    • AC1: Reply to CC1, Fabian Bernhard, 06 May 2020

      Thanks Valentin for your follow-up to your display. Yes that answers my question from the chat-room.

      Have you identified these plants with IDs? Were these the same plants for tiller and leaf at the same point in time? And also for the leaf water potentials? This might allow to use a population approach for modelling, with each individual having its own parameter set. Although the data might be too sparse for that. Was it possible to re-sample the same individual?

      Otherwise, have you tested to trade in temporal resolution in favour of bigger groups for the computation of δ18O_tiller and δ18O_leaf?

      • CC2: Reply to AC1, Valentin Couvreur, 06 May 2020

        I don't think there were ID's but the same plants were sampled for tiller and leaf at the same time, indeed. This is why d18O_tiller and d18O_leaf are so correlated. For the leaf water potential, we'd have to ask Youri.
        I doubt the same combination of 3 individuals was sampled several times as there were about 1500 plants with no ID.
        Trading temporal resolution in favour of bigger groups would have been an excellent choice, and would have allowed confirming that there was so much "population variability" of xylem water d18O. I would definitely opt in favour of that in a future experiment.

  • CC3: Calibration of bucket model?, Matthias Sprenger, 07 May 2020

    Hi Fabian,

    Great poster!!

    Do you have an idea or suggestions how one would use a bucket-type model when only bulk soil water isotope data from field samples are available? Such samples (e.g., from destructive sampling and then cryogenic extraction or direct equilibrium method for isotope analysis) are usually representing resident concentration and not flux concentrations (as you would have in lysimeter studies). However, from my understanding, bucket type models would need flux concentrations (as seen in Christine Stumpp's work for example).


    • AC2: Same as distributed models, Fabian Bernhard, 11 May 2020

      Hi Matthias, Thank you for your comment and question. I'm not sure I understand your question correctly. In my view, bucket type models can be applied in the same way as distributed models:

      Conceptually, bucket type models are not different from spatially distributed models (e.g. Richards and transport equation) in that they connect fluxes with storages as a dynamic system. Here, I think of bucket type models as described e.g. by [1].

      As such, resident bulk soil water concentrations are time-integrated measures of the net fluxes. In that sense, these concentrations, when combined with information on temporally varying precipitation, can contain information on the turnover of soil water, i.e. how quickly the storage achieves equilibrium. Thereby, these concentrations can constrain bucket model parameters related to that. Similar information could be extracted using an end-member mixing approach that analyses relative importance of isotopically different precipitation events.

      You're right that this needs additional information on the isotopic concentrations of precipitation or throughfall fluxes.

      From what I understand, lumped parameter modelling as applied in some papers of Christine Stumpp (e.g. [2]) is different from the bucket type models as described in my display. By using the convolution integral, the system is assumed to be linear and storage is completely characterized by the transit time distribution that links in- to outflow. I understand the focus here is really on the tracer concentrations. Resident concentration measurements could possibly help to constrain the transit time distributions in that approach...

      [1]: Rodriguez-Iturbe, I., Porporato, A., Laio, F., & Ridolfi, L. (2001). Plants in water-controlled ecosystems: Active role in hydrologic processes and response to water stress: I. Scope and general outline. Advances in Water Resources, 24(7), 695–705. [2]: Stumpp, C., Maloszewski, P., Stichler, W., & Fank, J. (2009). Environmental isotope (δ18O) and hydrological data to assess water flow in unsaturated soils planted with different crops: Case study lysimeter station “Wagna” (Austria). Journal of Hydrology, 369(1), 198–208.