the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 01 Feb 2023
HESS Opinions: Are soils overrated in hydrology?
Abstract. Traditional "physically-based" hydrological models are based on the assumption that soil is key in determining water's fate. According to these models, soil properties determine water movement in both saturated and unsaturated zones, described by matrix flow formulas known as the Darcy-Richards equations. Soil properties would also determine plant available moisture and thereby control transpiration. These models are data demanding, computationally intensive, parameter rich and, as we shall show, founded on a wrong assumption. Instead, we argue the reverse: it is the movement of the water through a porous medium, creating preferential patterns, that determines soil properties; while water movement is primarily controlled by the ecosystem's reaction to the climatic drivers. According to this assumption, soil properties are a “consequence”, rather than a “cause” of water movement. It is not the soil that is in control of hydrology, it is the ecosystem. An important and favourable consequence of this climate and ecosystem-driven approach is that models developed with this approach do not require soil information, are computationally cheap, and parsimonious. Our assumption is motivated by several arguments. Firstly, in well-developed soils the dominant flow mechanism is preferential, which is not particularly related to soil properties, such as pF curves. Secondly, we observe that it is the ecosystem, rather than the soil, that determines the land-surface water balance and hydrological processes. Top-down analysis by large-sample datasets reveal that soil properties are often a poor predictor of hydrological signatures. Bottom-up models do not directly use field measurements of the soil, but "rebalance" the observed soil texture and translate it to soil structure by vegetation indices. Thus, soil-based models may be appropriate at small spatio-temporal scale in non-vegetated and agricultural environments, but introduce unnecessary complexity and sub-optimal results in catchments with permanent vegetation. Progress in hydrology largely relies on abandoning the compartmentalized approach, putting ecosystem at the centre of hydrology, and moving to a landscape-based modelling approach. This change in perspective is needed to build more realistic and simpler hydrological models that go beyond current stationarity assumptions, but instead regard catchments as the result of ecosystems' coevolution with atmosphere, biosphere, hydrosphere, pedosphere, and lithosphere.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(754 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(754 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
CC1: 'A critical note on ET “reality”', Adriaan J. (Ryan) Teuling, 12 Feb 2023
Gao and co-authors provide an interesting perspective on the importance and use of soil information in hydrology. I generally agree with most of their arguments — in fact I believe the adaptive nature of vegetation and the dominance of preferential flow is one of the reasons that for instance simple power-law relations can accurately describe discharge dynamics at the catchment-scale (Kirchner, 2009) in spite of all the underlying variability and process complexity. However not all the arguments provided are robust. In particular the observed ET shown in their Figure 1 is, in my opinion, questionable. I am well aware of this figure, as in fact I use it in my lectures on evaporation. But rather than using it as an argument to show that models are bad, I use it as a warning to students that wrongly-tuned satellite based retrievals can come up with results that are not only physically implausible but also completely inconsistent with other more reliable local observations, for instance made by eddy covariance. This is illustrated by Figure 1, which is taken from a study in which I analysed eddy covariance data for forest and non-forest sites during heatwaves in the Netherlands also during July 2006.
Figure 1. Energy exchanges at the peak of the July 2006 heatwave for neighbouring flux towers over Cabauw (top, grassland site), Loobos (bottom, forest site). Distance between the sites is 60 km (see map inset). The solid lines are direct observations on 19 July 2006; the dashed lines indicate the baseline conditions in a normal year. Black: net radiation (Rn), blue: latent heat flux (λET), red: sensible heat flux (H). The arrows indicate maximum anomalies for λET (grassland site, upper panel), H (forest site, lower panel) and Rn. Figure taken from Teuling et al. (2010).
The authors refer to the top panels in their Figure 1 as “reality”. However analysis from eddy covariance data in July 2006 reveals a completely different picture. At the Cabauw grassland site, there is a strong shift in the energy balance partitioning towards more ET, in fact more than double that was observed during normal summers. At the Loobos forest site, the ET is similar to that observed in other years, and most of the additional energy is used for increasing the sensible heat flux. Compared the grassland site, the forest site has less than half the ET. And it should be noted that part of the footprint at Loobos contains trees that might have access to shallow groundwater, which is not the situation in most of the larger Veluwe forest region. So the available eddy covariance data does not support the claims made on high ET over forests on sandy soil. This does however not mean that this is necessarily because of a lack of soil moisture: also high atmospheric VPD might induce such reductions in ET (van Heerwaarden et al., 2014; Lansu et al., 2020). Interestingly, we found this signal across many sites in Europe that have very different soil and groundwater conditions. This, in a different way, actually supports the hypothesis by the authors that not soil, but vegetation might be the main determinant of hydrological response.
There is a danger in dismissing models simply because they don’t match a product that is wrongly seen as observation. At least the models shown are forced to obey the principle of mass conservation, and no such principle is behind the “observations” shown. The observations show a daily mean ET of up to 5 mm over coarse sandy soils over a month with hardly any rainfall. That is 5*31=155 mm total for July 2006, an incredible amount even for the best soils under constant well-watered conditions. In our continuous quest for a better understanding of the hydrological cycle, we should always stay critical of existing theories and methods, but not readily dismiss them based on anything else but extraordinary evidence.
References
van Heerwaarden, C. C. & A. J. Teuling (2014), Disentangling the response of forest and grassland energy exchange to heatwaves under idealized land-atmosphere coupling. Biogeosciences, 11, 6159–6171, doi:10.5194/bg-11-6159-2014.
Kirchner, J. W. (2009), Catchments as simple dynamical systems: Catchment characterization, rainfall-runoff modeling, and doing hydrology backward. Water Resour. Res., 45, W02429, doi:10.1029/2008WR006912.
Lansu, E.; C. C. van Heerwaarden; A. I. Stegehuis & A. J. Teuling (2020), Atmospheric aridity and apparent soil moisture drought in European forest during heatwaves. Geophysical Research Letters, 47(6), e2020GL087091, doi:10.1029/2020GL087091.
Teuling, A. J.; S. I. Seneviratne; R. Stöckli; M. Reichstein; E. Moors; P. Ciais; S. Luyssaert; B. van den Hurk; C. Ammann; C. Bernhofer; E. Dellwik; D. Gianelle; B. Gielen; T. Grünwald; K. Klumpp; L. Montagnani; C. Moureaux; M. Sottocornola & G. Wohlfahrt (2010), Contrasting response of European forest and grassland energy exchange to heatwaves. Nature Geoscience, 3(10), 722–727, doi:10.1038/ngeo950.
Citation: https://doi.org/10.5194/egusphere-2023-125-CC1 -
AC1: 'Reply on CC1', Hongkai Gao, 18 Feb 2023
Reply to Ryan Teuling of 12 Feb 2023,
We thank Ryan Teuling for his comments which support our main thesis that soil information leads to unnecessary complexity of hydrological models and even to soil-biased results.
We are also grateful for the additional information provided to further discuss Fig 1. of our paper. Teuling shows that point observations of actual evaporation indicate that our figure 1 is not fully accurate. However, we should be careful not to focus too much on the accuracy of what is called “observed” in the upper part of this figure, but rather focus on the inaccuracy of the soil-based model on the lower part of the figure.
We see in the upper part that the vegetation apparently has adapted to the climate variability by creating a large enough buffer to prevent drastic moisture constraints. As a result, the evaporation in the upper part is much more homogeneous than the evaporation in the lower part of the Figure, which mimics the variability of the soils. This homogeneity is the vegetation’s response to a high potential evaporation. The actual values may not be completely correct at point scale, as Teuling indicates, but as a pattern it is closer to reality. In the Netherlands the vegetation did not wilt during this extremely dry period, whereas the lower part of the figure indicates that the sandy parts would have suffered seriously and only the clayey parts would have flourished.
The lower part shows a very unrealistic result where the ecosystems on sandy soils have very low evaporation (less than 1 mm/d) due to moisture constraints, while ecosystems on clay evaporate at 5 mm/d. The figure presented by Teuling of evaporation on 19 July supports our argument that both forest and grassland evaporated well during this extremely dry month, independent of their soil condition. From our own rough calculation based on reading this graph, the average evaporation on 19 July amounted to about 5 mm/d on grassland and 2.9 mm/d on forest.
We also have observations of a similar site (Speulderbos), a Douglas Fir stand not too far from the Loobos site Teuling refers to, where eddy covariance observations were done during the heat wave of 2019 (from 14/8/2019-28 to 25/9/2019) (see: César Jiménez-Rodiguez, 2020). During that period, a total evaporation was observed of 154 mm, or 3.6 mm/d. This average evaporation during a hot summer month is much the same as the value in the top right figure colored in orange (3.5 to 4 mm/d in July 2006), which shows that the energy-balanced based estimate is not far off the mark. Of course, the 3.5 to 4 mm/d is an average over the month of July in which there also was some rain (42 mm/month was recorded in July 2006) of which the intercepted part would have evaporated as well. This implies that daily values could well have been less or more than the average. More importantly, the 3.5 mm/d in the top graph may be a bit too high, but the 1 mm/d in the lower graph is clearly too low, and the pattern is too heterogeneous.
Regarding the high evaporation in the grass plot of Cabauw, it should be noted that the grass pixel lies in a well-watered polder where the groundwater is shallow; as such, it behaves as irrigated grassland. In any case, even if the evaporation on 19 July in Speulderbos was only 2.9 mm/d, then this still is substantial evaporation, considerably higher than the 1 mm/d in the lower part of the graph.
Teuling is right that the term “observed” implies too high a claim of truth. The top part of the figure follows from an energy balance approach based on remote sensing information (making use of some point observations by eddy covariance and lysimeter) and is indeed not moisture constrained since the approach does not keep track of the moisture balance, as correctly argued by Teuling.
However, whether or not the remote sensing based “observed” graph is fully correct, even though it made use of ground observations, fact is that the spatial pattern of the soil-based model is far off the mark, which is the main argument of this paper.
In response to the comment, we do agree with Teuling that our term “observed” gives a wrong impression. Therefore, in the revised submission, we will change the description of Fig. 1 in line with Teuling’s comment and will change the term “observed” by “Remote Sensing derived”, which we hope puts Fig. 1 in a better perspective.
Reference:
César Dionisio Jiménez Rodríguez, 2020. Evaporation partitioning of forest stands; the role of forest structure. PhD thesis, TU Delft, https://repository.tudelft.nl/islandora/object/uuid%3A046a7cc0-1d66-4c7a-b294-feea556d7246?collection=research.
Citation: https://doi.org/10.5194/egusphere-2023-125-AC1
-
AC1: 'Reply on CC1', Hongkai Gao, 18 Feb 2023
-
CC2: 'Comment on egusphere-2023-125', Luca Brocca, 18 Feb 2023
I read the paper with great interest as I think the topic is very relevant for hydrology. I generally agree with the authors, but I think that one aspect should be clarified. I have one main comment that could help the authors to address the problem I found.
Before addressing this issue, I would like to emphasise that the results shown in Figure 1, comparing evaporation estimates from models and observations, deserve careful consideration. In our research activities (and also in several publications) we have found several times large discrepancies in the soil moisture spatial patterns between modelled and observed data (from in situ and satellite data). We note that the soil-, land use- or topography-driven spatial patterns that we have in modelling do not match the spatial patterns from observations (and also the spatial statistics, e.g., Cornelissen et al., 2014, doi: 10.1016/j.jhydrol.2014.01.060). Currently, high resolution (<1 km) satellite soil moisture products are available thanks to the Sentinel-1 satellites and these datasets are largely unexplored by the hydrological community. I believe that the comparison of modelling and satellite data in space (most - 95% - of the validation and comparison studies have made only temporal analysis) will provide important insights for improving our ability to model hydrological processes at large scales. Likely, part of this comment can be incorporated into the paper in section 3.2.
The issue to be addressed is the uncertainty in mapping soil properties, which is even more important in mapping soil hydraulic properties. At very small scales we can obtain quite accurate information about the soil, but when extrapolating over large areas it is very difficult to obtain reliable estimates. The problem raised by the authors, that soil properties are not a good variable to explain hydrological modelling parameters and patterns, is strongly influenced by this issue. We have found similar results in our studies, but the uncertainties in soil properties may be the reason for the poor correlations we have obtained. This aspect should be at least discussed in the paper.
I have also two minor comments:
- At line 123 Figure 1 should be Figure 3, but likely better the figures order should be changed.
- In the last part of the paper I have found typos that need to be corrected.
Citation: https://doi.org/10.5194/egusphere-2023-125-CC2 -
AC2: 'Reply on CC2', Hongkai Gao, 21 Feb 2023
Reply to Luca Brocca, 18 Feb 2023 Firstly, we are grateful for Luca Brocca’s generally supportive comments. We also thank Brocca to bring up the issue about the discrepancies between observed and remotely sensed soil moisture patterns. This gives us another chance to clarify what we mean by root zone storage, which we will incorporate in the revised manuscript. We agree that the surface soil moisture condition is important for agricultural studies and the land-atmosphere interaction in arid regions. However, it is important to note that remote sensors can only detect soil moisture within a few centimeters below the surface (2∼5cm, see Brocca et al., 2014). However, what is of key interest for modelling streamflow is the variability of the moisture content in the root zone. In most cases, the presence of vegetation prevents the observation of soil moisture and further deteriorates the remote sensing results. Avoiding the influence of vegetation in observing soil moisture (e.g. by SMOS or SMAP) is seen as a challenge in the remote sensing community. However, we found that a highly significant correlation exists between remotely sensed Normalized Difference Infrared Index (NDII, indicating the equivalent water thickness of leaves and canopy) and root zone moisture storage (Su), particularly during periods of moisture stress. This means vegetation, rather than a troublemaker to be avoided, is a good indicator of the dynamics of moisture storage in the root zone, (Sriwongsitanon et al., 2016). It is also worthwhile to note that The Netherlands, the subject of Fig.1 in our paper, is probably the most densely mapped area of the world, with an enormous number of soil probe analyses and very detailed mapping. Yet this enormous detail has not helped to improve the evaporation modelling based on the combination of the soil map and crop information. It indicates that more detail is not the way forward, but that there is a need for another perspective: the ecosystem perspective which manipulates the soil characteristics to its advantage. We admit that the term “observed” in Figure 1., as we also responsed to the comment of Ryan Teuling, may give the wrong impression. Therefore, in the revised submission, we will change the term “observed” by “Remote Sensing derived”, which we hope puts Fig. 1 in a better perspective. Regarding the issue of the uncertainty in mapping soil properties and soil hydraulic properties, we agree that these are highly uncertain. However, uncertainty characterizes any observation that we use in hydrological models. Yet, some of these observations appeared more useful than others in explaining observed patterns of hydrological variability. For example, lithology maps, although uncertain, have proven useful to explain baseflow partitioning. Commonly available soil data, instead, appeared generally less informative than other data to explain signatures of hydrological variability. Of course, this is just circumstantial evidence that soil information is not that necessary for hydrological modelling, and one may argue that more detail is needed, which is what the soil community has often been saying. And we agree that if we could observe anything, every pore space and preferential pathway, we could in principle also model it. But for what purpose? If the purpose is hydrological modelling at the scales and resolution that is commonly provided by hydrological models, that is, to partition precipitation between storage, evaporation, and runoff, then, as we argue in the paper, this is not needed. The reason is that the parameters we are more interested in that affect such partitioning are not determined by soil, but by vegetation. Understanding the principles by which vegetation operates, and incorporating them in hydrological models, provides the necessary information. Although in our perspective describing soil processes is sometimes useful for small scale hydrological modelling, such as in agriculture. we would argue that this should be done under the premise that the processes in the soil compartment are subjected to constraints determined by the vegetation. Reference: Brocca, L., L. Ciabatta, C. Massari, T. Moramarco, S. Hahn, S. Hasenauer, R. Kidd, W. Dorigo, W. Wagner, and V. Levizzani (2014), Soil as a natural rain gauge: Estimating global rainfall from satellite soil moisture data, J. Geophys. Res. Atmos., 119, 5128–5141, doi:10.1002/2014JD021489. Sriwongsitanon, N., Gao, H., Savenije, H. H. G., Maekan, E., Saengsawang, S., and Thianpopirug, S.: Comparing the Normalized Difference Infrared Index (NDII) with root zone storage in a lumped conceptual model, Hydrol. Earth Syst. Sci., 20, 3361–3377, https://doi.org/10.5194/hess-20-3361-2016, 2016.
Citation: https://doi.org/10.5194/egusphere-2023-125-AC2
-
RC1: 'Critical comments on egusphere-2023-125', Conrad Jackisch, 02 Mar 2023
Hongkai Gao and co-workers present an interesting contribution to the debate about key elements in the concepts of hydrological modelling. As an opinion paper, the authors argue that the affinity of hydrological model concepts to soil properties are more a relict than a substantial information basis. They propose to shift focus to the rootzone (as manifestation of the ecosystem) as alternative conceptual foundation.
I congratulate the authors for their work and I agree that our community has to keep challenging the conceptual assumptions and traditions. The role of soils in hydrology and land surface modelling is a particularly interesting debate. Recently Novick et al. (2022) have pointed to a “water potential information gap” in similar notion but opposite proposals. Discussing the role of pedotransfer functions (Looy et al. 2017, Vereecken et al. 2022) and soil hydraulic functions (Peters et al. 2023) together with structural adequacy of models (Gupta et al. 2012), perceptual model consistency (Wagener et al. 2021) and the data flow in model building (Gharari et al. 2021) and model analyses (Loritz et al. 2018) is in my view very important and promising. Hence I see the topic of this manuscript as worth an opinion paper.
However, I am not really convinced that the current arrangement of the arguments in the manuscript is really substantiating this timely debate. My main concern is that the authors use the word “soil” for different concepts and at different scales without much differentiation. The critical zone concept (Lin et al. 2006) was already much further than this. Also the debate about landscape organisation and hydrologic functioning (Jackisch et al. 2021) including a critical assessment of conceptual assumptions about processes and scaling is more advanced on the topic.
The authors do address such aspects in their manuscript and point to intertwined factors and sub-systems. However, the arguments are not really brought to consistently support the very fundamental claim of the manuscript. Without meaning to offend the authors, I would see many of the claims rather being rooted in conceptual limitations in the view of soil functions by the authors than in the lack of information or importance of soils in hydrological processes and models. I will substantiate this in the more detailed assessment.
In general, I do not really see, how the replacement of a soil-centred with a rootzone-centred concept deviated from the critical zone concepts (Lin et al. 2006). I also do not see, why the authors omit the main driving concept for fluxes (depletion of gradients) and thus the whole debate about potentials (Novick et al. 2022). I would have liked to see to which degree their arguments are essentially an expression of the conceptualisation of hydrological models i) as distributed and linked storages, ii) at a broader scale (in the sense of the scale triplet) and iii) with soils expressed by texture classes.
Moreover, I find many claims very strong and confrontative (e.g. L21f, L111) and not well-balanced. To really spark the debate (and not a battle) I would have liked a more balanced and substantiated formulation.
Detailed comments:
L43: I would argue that this is the debate throughout in pedology. At least not only recently.
L49: Why do you limit the perspective to abiotic boundary conditions when you actually argure for an ecosystem perspective? Biodiversity, niches, disturbances, stressors, carbon pools are not all determined by climate and geology. Moreover, at least most temperate soils do not develop directly from bedrock material but on deposited material from rather old geomorphological processes (which include path-dependendtdevelopment options). Pointing to this, soil degradation and soil loss too is an important and largely irreversible process with severe implications on regional hydrological and biogeochemical cycles.
L59: I think I have an idea what you intend to express, but since this simplistic/reductionist pedotransfer approach has a couple of implications which could be challenged. I suggest to clarify this sentence a little more and link to the debates in soil physics and the pedotransfer community.
L63 (opposite): I am not sure if I can follow. The argument before was that simple pedotransfer and soil hydraulic property models are an issue and that they become coupled to rooting depth.
Consequently, rooting depth has to be defined somehow to eventually assess plant-available soil water storage. Why does it matter if in this step plant-available soil water storage becomes the dependent variable of the other involved variables, if it is used as independent variable in the proceeding calculation steps? Isn’t this a question about the perceptual model underlying any form of conceptualisation and numerical expression?
L63 (root zone storage): Yes, but maybe at different time scales? Plants and ecosystem may adapt and co-evolve (within a range of their survival). So why should the debate be solved by exchanging the depending/independent variables? Could this actually be a scaling issue?
L65: This might depend on what exactly we see as detailed. As you open your argumentation with coevolution, maybe a broad idea about the general type of soil (not texture class) and biome (including its ecohydrological properties) could be sufficiently detailed? If so, remote sensing claims various solutions to gather such data…
L67: Again, i would see this as a scale issue: Ecosystem and climate are both terms referring to large scales (in the sense of the scale triplet). Hydrology is not referring to a specific scale.
L72f: This assumes that the ecosystem is somewhat in equilibrium with determinable drivers of its development. However, path-dependent trajectories, dynamic deviations from equilibrium more or less buffered by the ecosystem and any application for global changes (climate, land use, cohabitation…) but severe challenges to this view.
L78: Why do you refer exactly to these citations? I would think that e.g. the work of Gardner including his famous lab experiments have been far more important for propagating this perception.
L87ff: Yes, and this might be one of the actual issues to address here. Linear Darcy filter flow has been coupled to highly non-linear retention properties with the Richards equation and as a first-order diffusive flow model, it does an ok job for diffusive flow in somewhat well-defined porous media. However, especially infiltration (as initial soil water redistribution into the soil during rain events) is often not dominated by diffusive flow but by advection (Newtion Shear Flow equation (Germann, 2020), soil moisture velocity equation (Ogden et al. 2017), particle model (Jackisch and Zehe 2018), non-equilibrium flow (Vogel et al. 2023)). To my understanding, this deficit is rooting back to the very limited means to measure antecedent state-dependent infiltration and to use such data in hydrologic models. But why this is an argument for soils not being central in the question for one of their fundamental services to mediate the local soil water cycle is not clear to me. Especially because infiltration is state dependent, precipitation may not be retained after drought conditions, requiring vast amounts of light rains, slow snow melts or similar to replenish the water stocks, while storm events will simply lead to preferential flow and possibly erosion…
L93: Well, it is not dominant when it comes to storm events, yes. But these experiments use rather steep gradients with a lot of water. The debate about when and to what degree soil water flow is preferential is ongoing. If this was the full story, soils could hardly sustain the ecosystems.
L98: Partly yes. But preferential flow can also simplify our models. Anyways, I suggest to ease the dispute opened by this statement with a slightly more balanced view on achievements towards unifying forms of non-uniform infiltration.
L108: Yes. But this again can be seen as a scaling issue. At the hillslope- and plot-scale, these parameters/concepts have been very central. Only at the catchment-scale they could be easily subsumed as general soil property parameters not requiring for a dual domain definition. And this is true for the hindcast of our observations…
L111: Again a strong claim. I can agree that the pref flow debate has always struggled to connect to Darcy-scale soil physics. But fog? No progress? Is this claim really needed for your argument?
L113 (a priori assumption): Well, they are ABOUT the description of soil water flow. If they are key for describing hydrological processes is part of the actual model conceptualisation, its numerics and the respective regimes under study. Again, I would argue that this is no other “a priori assumption” as most other parts of the perceptual model. And since its actual effect in the model can be and is challenged (Glaser et al. 2018), I would rather see it as a positive example for advancing hydrologic models.
L118: So far soil variability has not been motivated. This is especially difficult, because the effect of soil variability is again a matter of scale (including the respective range of processes). After reading subsection 3.1, I can think of quite a number of papers, providing good evidence for the opposite: When you have the average soil right, you can easily reproduce observed hydrologic patterns (e.g. Loritz et al. 2017).
L123f: Ok. Known and well established. Maybe citing some of the many studies would be nice.
L131f: This argument is not really sound. Studies fully agree that plants and ecosystems strongly moderate the net ET flux of a stand. But without soil as the part of the ecosystem which can actually store water for weeks and beyond, this percentage cannot be reached. We exactly see this in data based on Budyko-like assessments that more draining locations (sandy, karstic) have very little ETact simply because precipitation is largely drained.
L137f: Yes, difficult but steeply advancing. Please see Peters et al. (2023) and Hohenbrink et al. (submitted to ESSD) for examples. The most critical part might be the reduction of such data to van Genuchten/Mualem SHP model parameters and the weakly informative relation to the broad texture classes, BD and Corg. But the issue of pedotransfer models is a discussion on its own, and which is currently gaining momentum.
L144: The issue here might be that soil mapping is not particularly done for hydrological purposes. On the one hand, pedological classes are not always directly convertible to hydrological properties. On the other hand, soil stratification and the respective hydrological properties are rarely conveyed into land surface models with sufficient degree of vertical resolution. Moreover, the uncertainty about the hydrological properties of the mapped soil classes is largely unknown and very different from region to region. Given all of these points, I am not quite sure if “interpolation and upscaling” is the core issue here. Maybe it is more a disconnection between soil mappers and hydrologic modellers?
L148: I fully agree that unnatural lab conditions are a fundamental difficulty. However, many measurements are conducted using “undisturbed” samples for soil hydraulic property analyses. It is unnatural because the samples are extracted from their capillary context, exposed to free evaporation at the surface and a no flow boundary at the bottom (for the standard HYPROP protocol). However pedotransfer functions are then correlating lab measurements (soil hydraulic properties) to lab measurements (texture) and the scaling and transfer involved in its application to field conditions remain hidden.
L152: I fully agree, but again this is an issue with quite a bit of literature from hydropedology to cite here.
L153ff: I do not get this point. The discussion about parameter regionalisation has a long standing in hydrology. E.g. mHM (Samaniego et al. 2017) exactly works because it modifies the initial lab scale parameters to match its distributed effects on fluxes in the landscape. Showing one odd model result can have so many reasons that I find it very difficult to support your argument through it.
L157: Which is a nice example for model extrapolation and the shift in parameter sensitivity under climate change (Melsen and Guse 2021).
L177: Again, a difficult claim. They test if texture classes and soil depth is informative. However Novick et al. (2022) point nicely to soil water potential being most informative and often omitted in LSMs. The model you are referring to are not particularly strong in soil physics as they conceptualise soils as stores instead of any framework of potentials as drivers. So your assessment might actually pinpoint that soil hydrology based on a storage concept is not very informative? As stated in the general section, I find it very difficult that you do not discern between weak conceptualisations of soils and the actual physical properties and dynamics linked to soils.
L188: These intertwined factors mostly manifest at “soil scales”, which are not necessarily very small.
L194f: Again, I would argue that the concept of infiltration capacity as rigid site property maybe the root of the issue here? Infiltration capacity to my understanding does not necessarily entail a constant or any specific model (e.g. Horton which is subsuming site properties and antecedent condition into an exponential decay function for infiltration rate or Green and Ampt which indeed is rarely proven in natural soils). Since infiltration is the passage of water into the soil domain, I would argue that soil structures (draining macropores and storing finer pores) facilitate it and that antecedent conditions plus the rainfall supply dynamics govern the individual initial (non-uniform) soil water redistribution (see comment to L87ff). The ecosystem modifies the boundary conditions, state dynamics and structure formation in the long run (Lange et al. 2015 and other publications from the Jena experiment).
L203ff: I agree and I admire the authors for their very nice contributions to these examples. However, this comparison is not fair since the intended applications of more complex models are often more than rainfall-runoff modelling. Especially when models are used to analyse effects of changes in land use , climate regime, management etc. the stationarity assumption collapses and we require parameters and submodels with physical meaning. Once we have a good understanding about how the modified hydrologic system can be conceptualised, the simple models are much more efficient and maybe even less error-prone again. But the transition (in system characteristics or scale) remains very challenging for these kind of models.
L210f and Fig. 2: I do not find it a logical proof of your argument that some models can succeed without soil information. If soil information is only texture class and porosity maybe it is more telling that these properties are not very informative for hydropedological characteristics and that the variable for the most frequent antecedent conditions (aridity) has far more influence because it is more informative for hydrological functioning? Hohenbrink et al. (submitted to ESSD, soon at https://doi.org/10.5194/essd-2023-74) show very nicely how these standard properties and texture-based soil classes do not inform hydropedologic functioning.
L217ff: I find it difficult to discern your “ecosystem”/“rootzone” approach from the hydropedology concepts (Lin et al. 2006).
L226ff: Within the lines of arguments, I think you are jumping through different scales here (with concepts and properties which are known NOT to be scale-invariant). The assumption that the ecosystem will be able to become the dominant driver is only true if the system has sufficient degrees of freedom to do so. Mediterranean basins have been deforested long ago, soil has been lost and there is no sign of spontaneous ecosystem replenishing under the current climate conditions. Badlands, crusts, long-term unstable debris are examples contradicting your claim. Hence a more differentiated analysis would be more insightful?
L232: I fully agree that water can bypass the rootzone but is not necessarily reaching groundwater. In many soil systems of the mid latitudes we find laterally conductive layers formed by more distant ice ages leading to relatively quick drainage or even interflow. Your FLEX approach has nicely shown this for the Ardennes…
L235ff: With having FLEX in mind I can understand your reasoning but I find your PERCEPTUAL model rather inflexible in the first place. The notion to simplify as much as possible is fully legit but deterministic concepts are in my understanding rather a thing of the past when we were limited in computational powers. And I find that this stiffness weakens your argumentation.
L245 (and the paragraphs before and after): I do not see why this is an argument against the importance of soils. Just because modellers use non-informative variables about soils and just because they have not found laws to scale the scale-dependent concepts/models does not mean that soils are not important. If these observations are biased, this does actually point to a misconception of the soil system rather than serving as an argument for omitting soils altogether. I would claim that this only shows that soil function cannot be described by texture classes (alone).
L251: I find it very difficult to agree to your arguments at this too general level of characterisation of somewhat arbitrarily selected model examples. I suggest to build the arguments based on the state of the art about structural adequacy and model conceptualisation (see general comments)
L282ff (and the whole subsection): You are proposing a new conceptualisation in which you omit various central properties governing water retention and drainage, which are not only governed by vegetation alone. With most of the terrestrial surface of our planet being actively managed by humans and a massively changing climate and biosphere, I find it not very helpful from a physical and system perspective. Moreover, your concept does not evade the scale issues. Quite to the contrary the active rootzone is not a static thing (at many scales). When we look at root water uptake alone, the sourcing depth of water within the root zone is dynamic over the year and very different from site to site (with the very same tree species and ages) (Jackisch et al. 2020). Giving reference to ERA5 data for this is maybe a little too large scale to substantiate your arguments with?
L295ff: From a (soil and hydrologic) physics perspective the main fundament might be that fluxes are driven by gradient depletion and that the degrees of freedom for these fluxes are state dependent (including subscale properties subsumed as hysteresis). The fill-and-spill concept (McDonnell et al. 2021) is a very powerful description of dynamic connectivity and threshold behaviour resulting from the strong non-linearities in soils. However, the depletion of gradients is largely omitted in such models. You might argue (L298f?) that storage-based models do not require an explicit treatment of gradients since it is all implicitly covered by the individual storage and transfer functions. However, this is not an argument against the importance of soils nor does it solve the standing issue to be capable to convey changing landscape properties into the required storage characteristics.
L308ff: Why do you jump from the debate about the concepts back to the debate about available data (which has so far not been really opened)?
L320ff: Since I read your manuscript as a strong claim for a simplified hydropedologic perceptual model, I find the argument with Occams razor very problematic. I would claim that we are in a situation with plenty of data to challenge our perceptual models and we have the tools to do this (e.g. Höge et al. 2020, Guthke 2017). Occams razor is a perceptual assumption, too.
Again, I sincerely thank the authors for raising this debate. I hope that my review can contribute to sharpening the arguments and to raise awareness about the many aspects that might have fallen a little too short in preparing this manuscript.
Bibliography
Germann, P.: Viscosity Controls Rapid Infiltration and Drainage, Not the Macropores, Water, 12, 337–15, https://doi.org/10.3390/w12020337, 2020.
Gharari, S., Gupta, H. V., Clark, M. P., Hrachowitz, M., Fenicia, F., Matgen, P., and Savenije, H. H. G.: Understanding the Information Content in the Hierarchy of Model Development Decisions: Learning From Data, Water Resour Res, 57, https://doi.org/10.1029/2020wr027948, 2021.
Glaser, B., Jackisch, C., Hopp, L., and Klaus, J.: How Meaningful are Plot-Scale Observations and Simulations of Preferential Flow for Catchment Models?, Vadose Zone J, 18, 0–18, https://doi.org/10.2136/vzj2018.08.0146, 2019.
Gupta, H., Clark, M. P., Vrugt, J. A., Abramowitz, G., and Ye, M.: Towards a Comprehensive Assessment of Model Structural Adequacy, Water Resour Res, 48, 1–40, https://doi.org/10.1029/2011wr011044, 2012.
Guthke, A.: Defensible Model Complexity: A Call for Data-Based and Goal-Oriented Model Choice, Groundwater, 55, 646–650, https://doi.org/10.1111/gwat.12554, 2017.
Höge, M., Guthke, A., and Nowak, W.: Bayesian Model Weighting: The Many Faces of Model Averaging, Water-sui, Water, 12, 309, https://doi.org/10.3390/w12020309, 2020.
Jackisch, C., Hassler, S. K., Hohenbrink, T. L., Blume, T., Laudon, H., McMillan, H., Saco, P., and Schaik, L. van: Preface: Linking landscape organisation and hydrological functioning: from hypotheses and observations to concepts, models and understanding, Hydrol Earth Syst Sc, 25, 5277–5285, https://doi.org/10.5194/hess-25-5277-2021, 2021.
Jackisch, C., Knoblauch, S., Blume, T., Zehe, E., and Hassler, S. K.: Estimates of tree root water uptake from soil moisture profile dynamics, Biogeosciences, 17, 5787–5808, https://doi.org/10.5194/bg-17-5787-2020, 2020.
Jackisch, C. and Zehe, E.: Ecohydrological particle model based on representative domains, Hydrol Earth Syst Sc, 22, 3639–3662, https://doi.org/10.5194/hess-22-3639-2018, 2018.
Lange, M., Eisenhauer, N., Sierra, C. A., Beßler, H., Engels, C., Griffiths, R. I., Mellado-Vázquez, P. G., Malik, A. A., Roy, J., Scheu, S., Steinbeiss, S., Thomson, B. C., Trumbore, S. E., and Gleixner, G.: Plant diversity increases soil microbial activity and soil carbon storage, Nature Communications, 6, 1–8, https://doi.org/10.1038/ncomms7707, 2015.
Lin, H., Bouma, J., Pachepsky, Y., Western, A., Thompson, J., Genuchten, R. van, Vogel, H.-J., and Lilly, A.: Hydropedology: Synergistic integration of pedology and hydrology, Water Resour Res, 42, 2509–13, https://doi.org/10.1029/2005wr004085, 2006.
Looy, K. V., Bouma, J., Herbst, M., Koestel, J., Minasny, B., Mishra, U., Montzka, C., Nemes, A., Pachepsky, Y. A., Padarian, J., Schaap, M. G., Tóth, B., Verhoef, A., Vanderborght, J., Ploeg, M. J., Weihermüller, L., Zacharias, S., Zhang, Y., and Vereecken, H.: Pedotransfer Functions in Earth System Science: Challenges and Perspectives, Rev Geophys, 55, 1199–1256, https://doi.org/10.1002/2017rg000581, 2017.
Loritz, R., Hassler, S. K., Jackisch, C., Allroggen, N., Schaik, L. van, and Wienhöfer, J.: Picturing and modeling catchments by representative hillslopes, 21, 1225–1249, https://doi.org/10.5194/hess-21-1225-2017, 2017.
Loritz, R., Gupta, H., Jackisch, C., Westhoff, M., Kleidon, A., Ehret, U., and Zehe, E.: On the dynamic nature of hydrological similarity, HESS 22, 3663–3684, https://doi.org/10.5194/hess-22-3663-2018, 2018.
McDonnell, J. J., Spence, C., Karran, D. J., Meerveld, H. J. (Ilja) van, and Harman, C. J.: Fill-and-Spill: A Process Description of Runoff Generation at the Scale of the Beholder, Water Resour Res, 57, https://doi.org/10.1029/2020wr027514, 2021.
Melsen, L. A. and Guse, B.: Climate change impacts model parameter sensitivity – implications for calibration strategy and model diagnostic evaluation, Hydrol Earth Syst Sc, 25, 1307–1332, https://doi.org/10.5194/hess-25-1307-2021, 2021.
Novick, K. A., Ficklin, D. L., Baldocchi, D., Davis, K. J., Ghezzehei, T. A., Konings, A. G., MacBean, N., Raoult, N., Scott, R. L., Shi, Y., Sulman, B. N., and Wood, J. D.: Confronting the water potential information gap, Nat Geosci, 15, 158–164, https://doi.org/10.1038/s41561-022-00909-2, 2022.
Ogden, F. L., Allen, M. B., Lai, W., Zhu, J., Seo, M., Douglas, C. C., and Talbot, C. A.: The soil moisture velocity equation, Journal of Advances in Modeling Earth Systems, 9, 1473–1487, https://doi.org/10.1002/2017ms000931, 2017.
Peters, A., Hohenbrink, T. L., Iden, S. C., Genuchten, M. Th. van, and Durner, W.: Prediction of the absolute hydraulic conductivity function from soil water retention data, Hydrology Earth Syst Sci Discuss, 2023, 1–32, https://doi.org/10.5194/hess-2022-431, 2023.
Samaniego, L., Kumar, R., Thober, S., Rakovec, O., Zink, M., Wanders, N., Eisner, S., Schmied, H. M., Sutanudjaja, E. H., Warrach-Sagi, K., and Attinger, S.: Toward seamless hydrologic predictions across spatial scales, Hydrol Earth Syst Sc, 21, 4323–4346, https://doi.org/10.5194/hess-21-4323-2017, 2017.
Vereecken, H., Amelung, W., Bauke, S. L., Bogena, H., Brüggemann, N., Montzka, C., Vanderborght, J., Bechtold, M., Blöschl, G., Carminati, A., Javaux, M., Konings, A. G., Kusche, J., Neuweiler, I., Or, D., Steele-Dunne, S., Verhoef, A., Young, M., and Zhang, Y.: Soil hydrology in the Earth system, Nat Rev Earth Environ, 3, 573–587, https://doi.org/10.1038/s43017-022-00324-6, 2022.
Vogel, H., Gerke, H. H., Mietrach, R., Zahl, R., and Wöhling, T.: Soil hydraulic conductivity in the state of nonequilibrium, Vadose Zone J, https://doi.org/10.1002/vzj2.20238, 2023.
Wagener, T., Gleeson, T., Coxon, G., Hartmann, A., Howden, N., Pianosi, F., Rahman, M., Rosolem, R., Stein, L., and Woods, R.: On doing hydrology with dragons: Realizing the value of perceptual models and knowledge accumulation, Wiley Interdiscip Rev Water, 8, https://doi.org/10.1002/wat2.1550, 2021.
Citation: https://doi.org/10.5194/egusphere-2023-125-RC1 - AC3: 'Reply on RC1', Hongkai Gao, 14 Mar 2023
-
RC2: 'Reviewer comment on egusphere-2023-125', Anonymous Referee #2, 16 Apr 2023
This opinion paper makes interesting and bold claims about the importance of soil properties for hydrology. I agree with many of the statements for natural soils and mature ecosystems. However, the majority of our earth is no longer a natural mature ecosystem. We have changed the surface cover drastically and very large areas are under agriculture or are so badly degraded and not in a mature “steady” state that the ecosystem perspective that is advocated in this paper is possibly no longer applicable. I think that this has to be mentioned in the text and that the reader needs to be reminded more frequently that these statements are made for mature natural ecosystems.
That soil is important is clear in situations where severely degraded ecosystems are restored. It is the restoration of the soil that leads to the very large changes in the flow pathways (from overland flow to subsurface flow) and thus streamflow responses. Indeed, it is the ecosystem that changes the soil properties that lead to the changes in the hydrological flow pathways and runoff responses, but this does not mean that the soil itself is not important at all. It means that the ecosystem has such a large effect on soil that the ecosystem would be a better predictor to be used in models (because ecosystem and soil properties become correlated as the ecosystem matures and the ecosystem is easier to observe), but it does not mean that soil is not important at all, especially not when one wants to understand processes. I think that some of the statements about soil not being important therefore require a bit more nuance. In particular, the model perspective (rather than process perspective) for some of the claims should be made clearer.
One of the confusing parts of the paper is that the authors state that the rooting zone is important but that soil is not important. This seems to suggest that they think that the rooting zone is not part of the soil. I think that what they mean is that soil texture is not important. To me it seems that most of the time when the authors say that soil is not important, they mean that soil texture is not important. For example when the authors refer to soil in the top down approach of catchment comparisons (Section 3.3), they actually refer to texture, not soil hydraulic properties. I urge the authors to more explicitly state that they focus on the soil texture. A better description of what parts of the soil they think are not important would be really helpful. It will also help if they give their definition of soil early in the paper.
The authors should point out much more clearly (and explicitly) that a major problem is that we use texture in pedotransfer functions to derive the soil characteristics that are related to water flow and storage, especially because these pedotransfer functions were developed for agricultural soils. The sand or silt content of a soil do not affect water flow or storage. We only attribute such an effect when we use pedotransfer functions to derive properties related to water flow and storage based on the texture. Because the pedotransfer functions were largely derived for agricultural soils, they do not take the effects of structure (and preferential flow) into account.
The writing of the manuscript could be a bit sharper. At several places, the authors make a good argument for why the ecosystem is important and then conclude that the soil is not important. I think that these sections need to be improved for two reasons. First, reasons are given for why the ecosystem is important but not for why the soil is not important. In particular, no references are given for this second part. In other words, the authors provide arguments for the first part (the ecosystem is important) but not for the second part (soil is not important). Thus either the second part (soil is not important) has to be taken out or arguments and references need to be included for the second part as well. Second, ecosystem and soil are interconnected. It is the ecosystem that changes the soil properties. So one can not directly argue that because the ecosystem is important, the soil is not important. It is still important but the ecosystem is perhaps the better predicting variable to be used in models because it is easier to observe and has a large effect on the soil properties that actually affect how water moves through the soil.
Other parts of the writing could also be improved. In several sentences words are missing and some other sentences are not clear and should be reformulated. The structure of the paper and individual sections was sometimes unclear to me. For example, section 4.1 consists of four paragraphs. Paragraph one highlights the importance of ET and states that hydrologists focus on discharge instead (but this point was already made on L128). The second paragraph then describes that ecosystems maximize storage and drainage. This section is interesting and fits the caption of this section. One would expect the next paragraph to get deeper into this but the third paragraph describes that the numbers for soil properties used in models don’t match the actual measurement values, and the fourth paragraph describes the rebalancing of soil properties that needs to be done in models. While the first two paragraphs sort of fit together and the last two paragraphs as well, the link between the first two and last two is not obvious. It also means that the second paragraph ends abruptly and this line of thinking could use some more elaboration. In addition, the part on the soil properties and the rebalancing starts abruptly without an introduction. The latter two paragraphs would probably better fit in a separate section on the problematic part of using pedotransfer functions based on texture (see comments above). This is just one example, there are other sections where the flow was unclear and I expect other readers to also wonder how the paragraphs are connected. I made some suggestions in the annotated pdf but there are more places where text could be reordered for a better flow. I don’t request that the authors use the suggested order but I do recommend that they carefully read through the paper to see if the order is logical for a reader.
Oter specific points:
- L56/139: I think that the problematic part of the use of pedotransfer functions based on texture to derive properties about pores should be described in more detail. Especially knowing that these pedotransfer functions were developed based on cores from agricultural fields and that texture does not really influence the hydraulic conducitivity (e.g., Jarvis et al., 2013; Gupta et al., 2021). See also comments above.
- Section 3.1: I don’t think that anyone claims that soil affects the long term water balance more than climate and vegetation. So, I think that it is fine to use this section to highlight that the ecosystem and climate are the main factors that determine the long term water balance but it makes less sense to use this as an argument that soils are not important.
- L129-132: Yes, land use change (if severe) alters runoff generation, exactly because of the large effect it has on soils. So, I don’t think that you can use this argument here to say that soils don’t matter. You can use it to make the argument that vegetation has a large effect on the soil properties that actually matter for water flow and storage. Also, it would be good to reference some field studies here (not only model studies).
- L158: But the comparison is basically between a model and a model with more data. I don’t think that one should call this observations.
- L162: But it also mimics the depth to the groundwater – maybe this has a different effect in the two models?
- L181: The problem is in part that we use texture here. Texture does not describe the soil pores that are important for storage or flow of water. The problem is that we use pedotransfer functions that are largely based on data from agricultural soils and are not appropriate for forested systems. See also the comments above. Furthermore, soil depth data is usually very rough and not very reliable. Maps of soil properties that actually describe water flow and storage are rarely available. Thus, one could also argue that the big problem is that we don’t have soil maps with sufficient information on the properties that actually matter and are related to water flow, and that instead we rely too much on texture and pedotransfer functions.
- L188: I agree that all these processes are intertwined or connected. Therefore, I think that the opinion paper should use more careful wording. It is OK to say that for hydrological modeling it is more useful to look at the ecosystem because the soil properties that matter for hydrology are highly correlated with land cover, and ecosystem properties are much easier to observe or measure. However, if we want to actually understand processes and the factors that affect these processes, it is important to look at the processes. In other words, then we have to look at the partitioning of rainfall into infiltration, overland flow, deeper drainage, etc. and soils are important. I think that this distinction between model application and process understanding should be made more clearly throughout the text.
- Section 4.2: I am sorry but I don’t understand what these ERA5 storage volumes contribute to the arguments of the opinion paper. The volume is one thing, the total flux from repeated filling and emptying is another. Certainly, I agree that the total storage is highest in the root zone but I consider the root zone to be part of the soil. So why is the root zone important but soil not? The paragraph on 277-284 goes some way into explaining this but it could have been added to section 4.1. It would be good if the authors give a definition of soil early in the paper. I have the feeling that often the authors mean soil texture instead of the soil itself.
- Several minor comments and suggestions are given in the annotated pdf.
References:
Jarvis, N., Koestel, J., Messing, I., Moeys, J., and Lindahl, A.: Influence of soil, land use and climatic factors on the hydraulic conductivity of soil, Hydrol. Earth Syst. Sci., 17, 5185–5195, https://doi.org/10.5194/hess-17-5185-2013, 2013.
Gupta, S., Hengl, T., Lehmann, P., Bonetti, S., and Or, D.: SoilKsatDB: global database of soil saturated hydraulic conductivity measurements for geoscience applications, Earth Syst. Sci. Data, 13, 1593–1612, https://doi.org/10.5194/essd-13-1593-2021, 2021.
- AC4: 'Reply on RC2', Hongkai Gao, 01 May 2023
- AC5: 'Reply on RC2', Hongkai Gao, 05 May 2023
Interactive discussion
Status: closed
-
CC1: 'A critical note on ET “reality”', Adriaan J. (Ryan) Teuling, 12 Feb 2023
Gao and co-authors provide an interesting perspective on the importance and use of soil information in hydrology. I generally agree with most of their arguments — in fact I believe the adaptive nature of vegetation and the dominance of preferential flow is one of the reasons that for instance simple power-law relations can accurately describe discharge dynamics at the catchment-scale (Kirchner, 2009) in spite of all the underlying variability and process complexity. However not all the arguments provided are robust. In particular the observed ET shown in their Figure 1 is, in my opinion, questionable. I am well aware of this figure, as in fact I use it in my lectures on evaporation. But rather than using it as an argument to show that models are bad, I use it as a warning to students that wrongly-tuned satellite based retrievals can come up with results that are not only physically implausible but also completely inconsistent with other more reliable local observations, for instance made by eddy covariance. This is illustrated by Figure 1, which is taken from a study in which I analysed eddy covariance data for forest and non-forest sites during heatwaves in the Netherlands also during July 2006.
Figure 1. Energy exchanges at the peak of the July 2006 heatwave for neighbouring flux towers over Cabauw (top, grassland site), Loobos (bottom, forest site). Distance between the sites is 60 km (see map inset). The solid lines are direct observations on 19 July 2006; the dashed lines indicate the baseline conditions in a normal year. Black: net radiation (Rn), blue: latent heat flux (λET), red: sensible heat flux (H). The arrows indicate maximum anomalies for λET (grassland site, upper panel), H (forest site, lower panel) and Rn. Figure taken from Teuling et al. (2010).
The authors refer to the top panels in their Figure 1 as “reality”. However analysis from eddy covariance data in July 2006 reveals a completely different picture. At the Cabauw grassland site, there is a strong shift in the energy balance partitioning towards more ET, in fact more than double that was observed during normal summers. At the Loobos forest site, the ET is similar to that observed in other years, and most of the additional energy is used for increasing the sensible heat flux. Compared the grassland site, the forest site has less than half the ET. And it should be noted that part of the footprint at Loobos contains trees that might have access to shallow groundwater, which is not the situation in most of the larger Veluwe forest region. So the available eddy covariance data does not support the claims made on high ET over forests on sandy soil. This does however not mean that this is necessarily because of a lack of soil moisture: also high atmospheric VPD might induce such reductions in ET (van Heerwaarden et al., 2014; Lansu et al., 2020). Interestingly, we found this signal across many sites in Europe that have very different soil and groundwater conditions. This, in a different way, actually supports the hypothesis by the authors that not soil, but vegetation might be the main determinant of hydrological response.
There is a danger in dismissing models simply because they don’t match a product that is wrongly seen as observation. At least the models shown are forced to obey the principle of mass conservation, and no such principle is behind the “observations” shown. The observations show a daily mean ET of up to 5 mm over coarse sandy soils over a month with hardly any rainfall. That is 5*31=155 mm total for July 2006, an incredible amount even for the best soils under constant well-watered conditions. In our continuous quest for a better understanding of the hydrological cycle, we should always stay critical of existing theories and methods, but not readily dismiss them based on anything else but extraordinary evidence.
References
van Heerwaarden, C. C. & A. J. Teuling (2014), Disentangling the response of forest and grassland energy exchange to heatwaves under idealized land-atmosphere coupling. Biogeosciences, 11, 6159–6171, doi:10.5194/bg-11-6159-2014.
Kirchner, J. W. (2009), Catchments as simple dynamical systems: Catchment characterization, rainfall-runoff modeling, and doing hydrology backward. Water Resour. Res., 45, W02429, doi:10.1029/2008WR006912.
Lansu, E.; C. C. van Heerwaarden; A. I. Stegehuis & A. J. Teuling (2020), Atmospheric aridity and apparent soil moisture drought in European forest during heatwaves. Geophysical Research Letters, 47(6), e2020GL087091, doi:10.1029/2020GL087091.
Teuling, A. J.; S. I. Seneviratne; R. Stöckli; M. Reichstein; E. Moors; P. Ciais; S. Luyssaert; B. van den Hurk; C. Ammann; C. Bernhofer; E. Dellwik; D. Gianelle; B. Gielen; T. Grünwald; K. Klumpp; L. Montagnani; C. Moureaux; M. Sottocornola & G. Wohlfahrt (2010), Contrasting response of European forest and grassland energy exchange to heatwaves. Nature Geoscience, 3(10), 722–727, doi:10.1038/ngeo950.
Citation: https://doi.org/10.5194/egusphere-2023-125-CC1 -
AC1: 'Reply on CC1', Hongkai Gao, 18 Feb 2023
Reply to Ryan Teuling of 12 Feb 2023,
We thank Ryan Teuling for his comments which support our main thesis that soil information leads to unnecessary complexity of hydrological models and even to soil-biased results.
We are also grateful for the additional information provided to further discuss Fig 1. of our paper. Teuling shows that point observations of actual evaporation indicate that our figure 1 is not fully accurate. However, we should be careful not to focus too much on the accuracy of what is called “observed” in the upper part of this figure, but rather focus on the inaccuracy of the soil-based model on the lower part of the figure.
We see in the upper part that the vegetation apparently has adapted to the climate variability by creating a large enough buffer to prevent drastic moisture constraints. As a result, the evaporation in the upper part is much more homogeneous than the evaporation in the lower part of the Figure, which mimics the variability of the soils. This homogeneity is the vegetation’s response to a high potential evaporation. The actual values may not be completely correct at point scale, as Teuling indicates, but as a pattern it is closer to reality. In the Netherlands the vegetation did not wilt during this extremely dry period, whereas the lower part of the figure indicates that the sandy parts would have suffered seriously and only the clayey parts would have flourished.
The lower part shows a very unrealistic result where the ecosystems on sandy soils have very low evaporation (less than 1 mm/d) due to moisture constraints, while ecosystems on clay evaporate at 5 mm/d. The figure presented by Teuling of evaporation on 19 July supports our argument that both forest and grassland evaporated well during this extremely dry month, independent of their soil condition. From our own rough calculation based on reading this graph, the average evaporation on 19 July amounted to about 5 mm/d on grassland and 2.9 mm/d on forest.
We also have observations of a similar site (Speulderbos), a Douglas Fir stand not too far from the Loobos site Teuling refers to, where eddy covariance observations were done during the heat wave of 2019 (from 14/8/2019-28 to 25/9/2019) (see: César Jiménez-Rodiguez, 2020). During that period, a total evaporation was observed of 154 mm, or 3.6 mm/d. This average evaporation during a hot summer month is much the same as the value in the top right figure colored in orange (3.5 to 4 mm/d in July 2006), which shows that the energy-balanced based estimate is not far off the mark. Of course, the 3.5 to 4 mm/d is an average over the month of July in which there also was some rain (42 mm/month was recorded in July 2006) of which the intercepted part would have evaporated as well. This implies that daily values could well have been less or more than the average. More importantly, the 3.5 mm/d in the top graph may be a bit too high, but the 1 mm/d in the lower graph is clearly too low, and the pattern is too heterogeneous.
Regarding the high evaporation in the grass plot of Cabauw, it should be noted that the grass pixel lies in a well-watered polder where the groundwater is shallow; as such, it behaves as irrigated grassland. In any case, even if the evaporation on 19 July in Speulderbos was only 2.9 mm/d, then this still is substantial evaporation, considerably higher than the 1 mm/d in the lower part of the graph.
Teuling is right that the term “observed” implies too high a claim of truth. The top part of the figure follows from an energy balance approach based on remote sensing information (making use of some point observations by eddy covariance and lysimeter) and is indeed not moisture constrained since the approach does not keep track of the moisture balance, as correctly argued by Teuling.
However, whether or not the remote sensing based “observed” graph is fully correct, even though it made use of ground observations, fact is that the spatial pattern of the soil-based model is far off the mark, which is the main argument of this paper.
In response to the comment, we do agree with Teuling that our term “observed” gives a wrong impression. Therefore, in the revised submission, we will change the description of Fig. 1 in line with Teuling’s comment and will change the term “observed” by “Remote Sensing derived”, which we hope puts Fig. 1 in a better perspective.
Reference:
César Dionisio Jiménez Rodríguez, 2020. Evaporation partitioning of forest stands; the role of forest structure. PhD thesis, TU Delft, https://repository.tudelft.nl/islandora/object/uuid%3A046a7cc0-1d66-4c7a-b294-feea556d7246?collection=research.
Citation: https://doi.org/10.5194/egusphere-2023-125-AC1
-
AC1: 'Reply on CC1', Hongkai Gao, 18 Feb 2023
-
CC2: 'Comment on egusphere-2023-125', Luca Brocca, 18 Feb 2023
I read the paper with great interest as I think the topic is very relevant for hydrology. I generally agree with the authors, but I think that one aspect should be clarified. I have one main comment that could help the authors to address the problem I found.
Before addressing this issue, I would like to emphasise that the results shown in Figure 1, comparing evaporation estimates from models and observations, deserve careful consideration. In our research activities (and also in several publications) we have found several times large discrepancies in the soil moisture spatial patterns between modelled and observed data (from in situ and satellite data). We note that the soil-, land use- or topography-driven spatial patterns that we have in modelling do not match the spatial patterns from observations (and also the spatial statistics, e.g., Cornelissen et al., 2014, doi: 10.1016/j.jhydrol.2014.01.060). Currently, high resolution (<1 km) satellite soil moisture products are available thanks to the Sentinel-1 satellites and these datasets are largely unexplored by the hydrological community. I believe that the comparison of modelling and satellite data in space (most - 95% - of the validation and comparison studies have made only temporal analysis) will provide important insights for improving our ability to model hydrological processes at large scales. Likely, part of this comment can be incorporated into the paper in section 3.2.
The issue to be addressed is the uncertainty in mapping soil properties, which is even more important in mapping soil hydraulic properties. At very small scales we can obtain quite accurate information about the soil, but when extrapolating over large areas it is very difficult to obtain reliable estimates. The problem raised by the authors, that soil properties are not a good variable to explain hydrological modelling parameters and patterns, is strongly influenced by this issue. We have found similar results in our studies, but the uncertainties in soil properties may be the reason for the poor correlations we have obtained. This aspect should be at least discussed in the paper.
I have also two minor comments:
- At line 123 Figure 1 should be Figure 3, but likely better the figures order should be changed.
- In the last part of the paper I have found typos that need to be corrected.
Citation: https://doi.org/10.5194/egusphere-2023-125-CC2 -
AC2: 'Reply on CC2', Hongkai Gao, 21 Feb 2023
Reply to Luca Brocca, 18 Feb 2023 Firstly, we are grateful for Luca Brocca’s generally supportive comments. We also thank Brocca to bring up the issue about the discrepancies between observed and remotely sensed soil moisture patterns. This gives us another chance to clarify what we mean by root zone storage, which we will incorporate in the revised manuscript. We agree that the surface soil moisture condition is important for agricultural studies and the land-atmosphere interaction in arid regions. However, it is important to note that remote sensors can only detect soil moisture within a few centimeters below the surface (2∼5cm, see Brocca et al., 2014). However, what is of key interest for modelling streamflow is the variability of the moisture content in the root zone. In most cases, the presence of vegetation prevents the observation of soil moisture and further deteriorates the remote sensing results. Avoiding the influence of vegetation in observing soil moisture (e.g. by SMOS or SMAP) is seen as a challenge in the remote sensing community. However, we found that a highly significant correlation exists between remotely sensed Normalized Difference Infrared Index (NDII, indicating the equivalent water thickness of leaves and canopy) and root zone moisture storage (Su), particularly during periods of moisture stress. This means vegetation, rather than a troublemaker to be avoided, is a good indicator of the dynamics of moisture storage in the root zone, (Sriwongsitanon et al., 2016). It is also worthwhile to note that The Netherlands, the subject of Fig.1 in our paper, is probably the most densely mapped area of the world, with an enormous number of soil probe analyses and very detailed mapping. Yet this enormous detail has not helped to improve the evaporation modelling based on the combination of the soil map and crop information. It indicates that more detail is not the way forward, but that there is a need for another perspective: the ecosystem perspective which manipulates the soil characteristics to its advantage. We admit that the term “observed” in Figure 1., as we also responsed to the comment of Ryan Teuling, may give the wrong impression. Therefore, in the revised submission, we will change the term “observed” by “Remote Sensing derived”, which we hope puts Fig. 1 in a better perspective. Regarding the issue of the uncertainty in mapping soil properties and soil hydraulic properties, we agree that these are highly uncertain. However, uncertainty characterizes any observation that we use in hydrological models. Yet, some of these observations appeared more useful than others in explaining observed patterns of hydrological variability. For example, lithology maps, although uncertain, have proven useful to explain baseflow partitioning. Commonly available soil data, instead, appeared generally less informative than other data to explain signatures of hydrological variability. Of course, this is just circumstantial evidence that soil information is not that necessary for hydrological modelling, and one may argue that more detail is needed, which is what the soil community has often been saying. And we agree that if we could observe anything, every pore space and preferential pathway, we could in principle also model it. But for what purpose? If the purpose is hydrological modelling at the scales and resolution that is commonly provided by hydrological models, that is, to partition precipitation between storage, evaporation, and runoff, then, as we argue in the paper, this is not needed. The reason is that the parameters we are more interested in that affect such partitioning are not determined by soil, but by vegetation. Understanding the principles by which vegetation operates, and incorporating them in hydrological models, provides the necessary information. Although in our perspective describing soil processes is sometimes useful for small scale hydrological modelling, such as in agriculture. we would argue that this should be done under the premise that the processes in the soil compartment are subjected to constraints determined by the vegetation. Reference: Brocca, L., L. Ciabatta, C. Massari, T. Moramarco, S. Hahn, S. Hasenauer, R. Kidd, W. Dorigo, W. Wagner, and V. Levizzani (2014), Soil as a natural rain gauge: Estimating global rainfall from satellite soil moisture data, J. Geophys. Res. Atmos., 119, 5128–5141, doi:10.1002/2014JD021489. Sriwongsitanon, N., Gao, H., Savenije, H. H. G., Maekan, E., Saengsawang, S., and Thianpopirug, S.: Comparing the Normalized Difference Infrared Index (NDII) with root zone storage in a lumped conceptual model, Hydrol. Earth Syst. Sci., 20, 3361–3377, https://doi.org/10.5194/hess-20-3361-2016, 2016.
Citation: https://doi.org/10.5194/egusphere-2023-125-AC2
-
RC1: 'Critical comments on egusphere-2023-125', Conrad Jackisch, 02 Mar 2023
Hongkai Gao and co-workers present an interesting contribution to the debate about key elements in the concepts of hydrological modelling. As an opinion paper, the authors argue that the affinity of hydrological model concepts to soil properties are more a relict than a substantial information basis. They propose to shift focus to the rootzone (as manifestation of the ecosystem) as alternative conceptual foundation.
I congratulate the authors for their work and I agree that our community has to keep challenging the conceptual assumptions and traditions. The role of soils in hydrology and land surface modelling is a particularly interesting debate. Recently Novick et al. (2022) have pointed to a “water potential information gap” in similar notion but opposite proposals. Discussing the role of pedotransfer functions (Looy et al. 2017, Vereecken et al. 2022) and soil hydraulic functions (Peters et al. 2023) together with structural adequacy of models (Gupta et al. 2012), perceptual model consistency (Wagener et al. 2021) and the data flow in model building (Gharari et al. 2021) and model analyses (Loritz et al. 2018) is in my view very important and promising. Hence I see the topic of this manuscript as worth an opinion paper.
However, I am not really convinced that the current arrangement of the arguments in the manuscript is really substantiating this timely debate. My main concern is that the authors use the word “soil” for different concepts and at different scales without much differentiation. The critical zone concept (Lin et al. 2006) was already much further than this. Also the debate about landscape organisation and hydrologic functioning (Jackisch et al. 2021) including a critical assessment of conceptual assumptions about processes and scaling is more advanced on the topic.
The authors do address such aspects in their manuscript and point to intertwined factors and sub-systems. However, the arguments are not really brought to consistently support the very fundamental claim of the manuscript. Without meaning to offend the authors, I would see many of the claims rather being rooted in conceptual limitations in the view of soil functions by the authors than in the lack of information or importance of soils in hydrological processes and models. I will substantiate this in the more detailed assessment.
In general, I do not really see, how the replacement of a soil-centred with a rootzone-centred concept deviated from the critical zone concepts (Lin et al. 2006). I also do not see, why the authors omit the main driving concept for fluxes (depletion of gradients) and thus the whole debate about potentials (Novick et al. 2022). I would have liked to see to which degree their arguments are essentially an expression of the conceptualisation of hydrological models i) as distributed and linked storages, ii) at a broader scale (in the sense of the scale triplet) and iii) with soils expressed by texture classes.
Moreover, I find many claims very strong and confrontative (e.g. L21f, L111) and not well-balanced. To really spark the debate (and not a battle) I would have liked a more balanced and substantiated formulation.
Detailed comments:
L43: I would argue that this is the debate throughout in pedology. At least not only recently.
L49: Why do you limit the perspective to abiotic boundary conditions when you actually argure for an ecosystem perspective? Biodiversity, niches, disturbances, stressors, carbon pools are not all determined by climate and geology. Moreover, at least most temperate soils do not develop directly from bedrock material but on deposited material from rather old geomorphological processes (which include path-dependendtdevelopment options). Pointing to this, soil degradation and soil loss too is an important and largely irreversible process with severe implications on regional hydrological and biogeochemical cycles.
L59: I think I have an idea what you intend to express, but since this simplistic/reductionist pedotransfer approach has a couple of implications which could be challenged. I suggest to clarify this sentence a little more and link to the debates in soil physics and the pedotransfer community.
L63 (opposite): I am not sure if I can follow. The argument before was that simple pedotransfer and soil hydraulic property models are an issue and that they become coupled to rooting depth.
Consequently, rooting depth has to be defined somehow to eventually assess plant-available soil water storage. Why does it matter if in this step plant-available soil water storage becomes the dependent variable of the other involved variables, if it is used as independent variable in the proceeding calculation steps? Isn’t this a question about the perceptual model underlying any form of conceptualisation and numerical expression?
L63 (root zone storage): Yes, but maybe at different time scales? Plants and ecosystem may adapt and co-evolve (within a range of their survival). So why should the debate be solved by exchanging the depending/independent variables? Could this actually be a scaling issue?
L65: This might depend on what exactly we see as detailed. As you open your argumentation with coevolution, maybe a broad idea about the general type of soil (not texture class) and biome (including its ecohydrological properties) could be sufficiently detailed? If so, remote sensing claims various solutions to gather such data…
L67: Again, i would see this as a scale issue: Ecosystem and climate are both terms referring to large scales (in the sense of the scale triplet). Hydrology is not referring to a specific scale.
L72f: This assumes that the ecosystem is somewhat in equilibrium with determinable drivers of its development. However, path-dependent trajectories, dynamic deviations from equilibrium more or less buffered by the ecosystem and any application for global changes (climate, land use, cohabitation…) but severe challenges to this view.
L78: Why do you refer exactly to these citations? I would think that e.g. the work of Gardner including his famous lab experiments have been far more important for propagating this perception.
L87ff: Yes, and this might be one of the actual issues to address here. Linear Darcy filter flow has been coupled to highly non-linear retention properties with the Richards equation and as a first-order diffusive flow model, it does an ok job for diffusive flow in somewhat well-defined porous media. However, especially infiltration (as initial soil water redistribution into the soil during rain events) is often not dominated by diffusive flow but by advection (Newtion Shear Flow equation (Germann, 2020), soil moisture velocity equation (Ogden et al. 2017), particle model (Jackisch and Zehe 2018), non-equilibrium flow (Vogel et al. 2023)). To my understanding, this deficit is rooting back to the very limited means to measure antecedent state-dependent infiltration and to use such data in hydrologic models. But why this is an argument for soils not being central in the question for one of their fundamental services to mediate the local soil water cycle is not clear to me. Especially because infiltration is state dependent, precipitation may not be retained after drought conditions, requiring vast amounts of light rains, slow snow melts or similar to replenish the water stocks, while storm events will simply lead to preferential flow and possibly erosion…
L93: Well, it is not dominant when it comes to storm events, yes. But these experiments use rather steep gradients with a lot of water. The debate about when and to what degree soil water flow is preferential is ongoing. If this was the full story, soils could hardly sustain the ecosystems.
L98: Partly yes. But preferential flow can also simplify our models. Anyways, I suggest to ease the dispute opened by this statement with a slightly more balanced view on achievements towards unifying forms of non-uniform infiltration.
L108: Yes. But this again can be seen as a scaling issue. At the hillslope- and plot-scale, these parameters/concepts have been very central. Only at the catchment-scale they could be easily subsumed as general soil property parameters not requiring for a dual domain definition. And this is true for the hindcast of our observations…
L111: Again a strong claim. I can agree that the pref flow debate has always struggled to connect to Darcy-scale soil physics. But fog? No progress? Is this claim really needed for your argument?
L113 (a priori assumption): Well, they are ABOUT the description of soil water flow. If they are key for describing hydrological processes is part of the actual model conceptualisation, its numerics and the respective regimes under study. Again, I would argue that this is no other “a priori assumption” as most other parts of the perceptual model. And since its actual effect in the model can be and is challenged (Glaser et al. 2018), I would rather see it as a positive example for advancing hydrologic models.
L118: So far soil variability has not been motivated. This is especially difficult, because the effect of soil variability is again a matter of scale (including the respective range of processes). After reading subsection 3.1, I can think of quite a number of papers, providing good evidence for the opposite: When you have the average soil right, you can easily reproduce observed hydrologic patterns (e.g. Loritz et al. 2017).
L123f: Ok. Known and well established. Maybe citing some of the many studies would be nice.
L131f: This argument is not really sound. Studies fully agree that plants and ecosystems strongly moderate the net ET flux of a stand. But without soil as the part of the ecosystem which can actually store water for weeks and beyond, this percentage cannot be reached. We exactly see this in data based on Budyko-like assessments that more draining locations (sandy, karstic) have very little ETact simply because precipitation is largely drained.
L137f: Yes, difficult but steeply advancing. Please see Peters et al. (2023) and Hohenbrink et al. (submitted to ESSD) for examples. The most critical part might be the reduction of such data to van Genuchten/Mualem SHP model parameters and the weakly informative relation to the broad texture classes, BD and Corg. But the issue of pedotransfer models is a discussion on its own, and which is currently gaining momentum.
L144: The issue here might be that soil mapping is not particularly done for hydrological purposes. On the one hand, pedological classes are not always directly convertible to hydrological properties. On the other hand, soil stratification and the respective hydrological properties are rarely conveyed into land surface models with sufficient degree of vertical resolution. Moreover, the uncertainty about the hydrological properties of the mapped soil classes is largely unknown and very different from region to region. Given all of these points, I am not quite sure if “interpolation and upscaling” is the core issue here. Maybe it is more a disconnection between soil mappers and hydrologic modellers?
L148: I fully agree that unnatural lab conditions are a fundamental difficulty. However, many measurements are conducted using “undisturbed” samples for soil hydraulic property analyses. It is unnatural because the samples are extracted from their capillary context, exposed to free evaporation at the surface and a no flow boundary at the bottom (for the standard HYPROP protocol). However pedotransfer functions are then correlating lab measurements (soil hydraulic properties) to lab measurements (texture) and the scaling and transfer involved in its application to field conditions remain hidden.
L152: I fully agree, but again this is an issue with quite a bit of literature from hydropedology to cite here.
L153ff: I do not get this point. The discussion about parameter regionalisation has a long standing in hydrology. E.g. mHM (Samaniego et al. 2017) exactly works because it modifies the initial lab scale parameters to match its distributed effects on fluxes in the landscape. Showing one odd model result can have so many reasons that I find it very difficult to support your argument through it.
L157: Which is a nice example for model extrapolation and the shift in parameter sensitivity under climate change (Melsen and Guse 2021).
L177: Again, a difficult claim. They test if texture classes and soil depth is informative. However Novick et al. (2022) point nicely to soil water potential being most informative and often omitted in LSMs. The model you are referring to are not particularly strong in soil physics as they conceptualise soils as stores instead of any framework of potentials as drivers. So your assessment might actually pinpoint that soil hydrology based on a storage concept is not very informative? As stated in the general section, I find it very difficult that you do not discern between weak conceptualisations of soils and the actual physical properties and dynamics linked to soils.
L188: These intertwined factors mostly manifest at “soil scales”, which are not necessarily very small.
L194f: Again, I would argue that the concept of infiltration capacity as rigid site property maybe the root of the issue here? Infiltration capacity to my understanding does not necessarily entail a constant or any specific model (e.g. Horton which is subsuming site properties and antecedent condition into an exponential decay function for infiltration rate or Green and Ampt which indeed is rarely proven in natural soils). Since infiltration is the passage of water into the soil domain, I would argue that soil structures (draining macropores and storing finer pores) facilitate it and that antecedent conditions plus the rainfall supply dynamics govern the individual initial (non-uniform) soil water redistribution (see comment to L87ff). The ecosystem modifies the boundary conditions, state dynamics and structure formation in the long run (Lange et al. 2015 and other publications from the Jena experiment).
L203ff: I agree and I admire the authors for their very nice contributions to these examples. However, this comparison is not fair since the intended applications of more complex models are often more than rainfall-runoff modelling. Especially when models are used to analyse effects of changes in land use , climate regime, management etc. the stationarity assumption collapses and we require parameters and submodels with physical meaning. Once we have a good understanding about how the modified hydrologic system can be conceptualised, the simple models are much more efficient and maybe even less error-prone again. But the transition (in system characteristics or scale) remains very challenging for these kind of models.
L210f and Fig. 2: I do not find it a logical proof of your argument that some models can succeed without soil information. If soil information is only texture class and porosity maybe it is more telling that these properties are not very informative for hydropedological characteristics and that the variable for the most frequent antecedent conditions (aridity) has far more influence because it is more informative for hydrological functioning? Hohenbrink et al. (submitted to ESSD, soon at https://doi.org/10.5194/essd-2023-74) show very nicely how these standard properties and texture-based soil classes do not inform hydropedologic functioning.
L217ff: I find it difficult to discern your “ecosystem”/“rootzone” approach from the hydropedology concepts (Lin et al. 2006).
L226ff: Within the lines of arguments, I think you are jumping through different scales here (with concepts and properties which are known NOT to be scale-invariant). The assumption that the ecosystem will be able to become the dominant driver is only true if the system has sufficient degrees of freedom to do so. Mediterranean basins have been deforested long ago, soil has been lost and there is no sign of spontaneous ecosystem replenishing under the current climate conditions. Badlands, crusts, long-term unstable debris are examples contradicting your claim. Hence a more differentiated analysis would be more insightful?
L232: I fully agree that water can bypass the rootzone but is not necessarily reaching groundwater. In many soil systems of the mid latitudes we find laterally conductive layers formed by more distant ice ages leading to relatively quick drainage or even interflow. Your FLEX approach has nicely shown this for the Ardennes…
L235ff: With having FLEX in mind I can understand your reasoning but I find your PERCEPTUAL model rather inflexible in the first place. The notion to simplify as much as possible is fully legit but deterministic concepts are in my understanding rather a thing of the past when we were limited in computational powers. And I find that this stiffness weakens your argumentation.
L245 (and the paragraphs before and after): I do not see why this is an argument against the importance of soils. Just because modellers use non-informative variables about soils and just because they have not found laws to scale the scale-dependent concepts/models does not mean that soils are not important. If these observations are biased, this does actually point to a misconception of the soil system rather than serving as an argument for omitting soils altogether. I would claim that this only shows that soil function cannot be described by texture classes (alone).
L251: I find it very difficult to agree to your arguments at this too general level of characterisation of somewhat arbitrarily selected model examples. I suggest to build the arguments based on the state of the art about structural adequacy and model conceptualisation (see general comments)
L282ff (and the whole subsection): You are proposing a new conceptualisation in which you omit various central properties governing water retention and drainage, which are not only governed by vegetation alone. With most of the terrestrial surface of our planet being actively managed by humans and a massively changing climate and biosphere, I find it not very helpful from a physical and system perspective. Moreover, your concept does not evade the scale issues. Quite to the contrary the active rootzone is not a static thing (at many scales). When we look at root water uptake alone, the sourcing depth of water within the root zone is dynamic over the year and very different from site to site (with the very same tree species and ages) (Jackisch et al. 2020). Giving reference to ERA5 data for this is maybe a little too large scale to substantiate your arguments with?
L295ff: From a (soil and hydrologic) physics perspective the main fundament might be that fluxes are driven by gradient depletion and that the degrees of freedom for these fluxes are state dependent (including subscale properties subsumed as hysteresis). The fill-and-spill concept (McDonnell et al. 2021) is a very powerful description of dynamic connectivity and threshold behaviour resulting from the strong non-linearities in soils. However, the depletion of gradients is largely omitted in such models. You might argue (L298f?) that storage-based models do not require an explicit treatment of gradients since it is all implicitly covered by the individual storage and transfer functions. However, this is not an argument against the importance of soils nor does it solve the standing issue to be capable to convey changing landscape properties into the required storage characteristics.
L308ff: Why do you jump from the debate about the concepts back to the debate about available data (which has so far not been really opened)?
L320ff: Since I read your manuscript as a strong claim for a simplified hydropedologic perceptual model, I find the argument with Occams razor very problematic. I would claim that we are in a situation with plenty of data to challenge our perceptual models and we have the tools to do this (e.g. Höge et al. 2020, Guthke 2017). Occams razor is a perceptual assumption, too.
Again, I sincerely thank the authors for raising this debate. I hope that my review can contribute to sharpening the arguments and to raise awareness about the many aspects that might have fallen a little too short in preparing this manuscript.
Bibliography
Germann, P.: Viscosity Controls Rapid Infiltration and Drainage, Not the Macropores, Water, 12, 337–15, https://doi.org/10.3390/w12020337, 2020.
Gharari, S., Gupta, H. V., Clark, M. P., Hrachowitz, M., Fenicia, F., Matgen, P., and Savenije, H. H. G.: Understanding the Information Content in the Hierarchy of Model Development Decisions: Learning From Data, Water Resour Res, 57, https://doi.org/10.1029/2020wr027948, 2021.
Glaser, B., Jackisch, C., Hopp, L., and Klaus, J.: How Meaningful are Plot-Scale Observations and Simulations of Preferential Flow for Catchment Models?, Vadose Zone J, 18, 0–18, https://doi.org/10.2136/vzj2018.08.0146, 2019.
Gupta, H., Clark, M. P., Vrugt, J. A., Abramowitz, G., and Ye, M.: Towards a Comprehensive Assessment of Model Structural Adequacy, Water Resour Res, 48, 1–40, https://doi.org/10.1029/2011wr011044, 2012.
Guthke, A.: Defensible Model Complexity: A Call for Data-Based and Goal-Oriented Model Choice, Groundwater, 55, 646–650, https://doi.org/10.1111/gwat.12554, 2017.
Höge, M., Guthke, A., and Nowak, W.: Bayesian Model Weighting: The Many Faces of Model Averaging, Water-sui, Water, 12, 309, https://doi.org/10.3390/w12020309, 2020.
Jackisch, C., Hassler, S. K., Hohenbrink, T. L., Blume, T., Laudon, H., McMillan, H., Saco, P., and Schaik, L. van: Preface: Linking landscape organisation and hydrological functioning: from hypotheses and observations to concepts, models and understanding, Hydrol Earth Syst Sc, 25, 5277–5285, https://doi.org/10.5194/hess-25-5277-2021, 2021.
Jackisch, C., Knoblauch, S., Blume, T., Zehe, E., and Hassler, S. K.: Estimates of tree root water uptake from soil moisture profile dynamics, Biogeosciences, 17, 5787–5808, https://doi.org/10.5194/bg-17-5787-2020, 2020.
Jackisch, C. and Zehe, E.: Ecohydrological particle model based on representative domains, Hydrol Earth Syst Sc, 22, 3639–3662, https://doi.org/10.5194/hess-22-3639-2018, 2018.
Lange, M., Eisenhauer, N., Sierra, C. A., Beßler, H., Engels, C., Griffiths, R. I., Mellado-Vázquez, P. G., Malik, A. A., Roy, J., Scheu, S., Steinbeiss, S., Thomson, B. C., Trumbore, S. E., and Gleixner, G.: Plant diversity increases soil microbial activity and soil carbon storage, Nature Communications, 6, 1–8, https://doi.org/10.1038/ncomms7707, 2015.
Lin, H., Bouma, J., Pachepsky, Y., Western, A., Thompson, J., Genuchten, R. van, Vogel, H.-J., and Lilly, A.: Hydropedology: Synergistic integration of pedology and hydrology, Water Resour Res, 42, 2509–13, https://doi.org/10.1029/2005wr004085, 2006.
Looy, K. V., Bouma, J., Herbst, M., Koestel, J., Minasny, B., Mishra, U., Montzka, C., Nemes, A., Pachepsky, Y. A., Padarian, J., Schaap, M. G., Tóth, B., Verhoef, A., Vanderborght, J., Ploeg, M. J., Weihermüller, L., Zacharias, S., Zhang, Y., and Vereecken, H.: Pedotransfer Functions in Earth System Science: Challenges and Perspectives, Rev Geophys, 55, 1199–1256, https://doi.org/10.1002/2017rg000581, 2017.
Loritz, R., Hassler, S. K., Jackisch, C., Allroggen, N., Schaik, L. van, and Wienhöfer, J.: Picturing and modeling catchments by representative hillslopes, 21, 1225–1249, https://doi.org/10.5194/hess-21-1225-2017, 2017.
Loritz, R., Gupta, H., Jackisch, C., Westhoff, M., Kleidon, A., Ehret, U., and Zehe, E.: On the dynamic nature of hydrological similarity, HESS 22, 3663–3684, https://doi.org/10.5194/hess-22-3663-2018, 2018.
McDonnell, J. J., Spence, C., Karran, D. J., Meerveld, H. J. (Ilja) van, and Harman, C. J.: Fill-and-Spill: A Process Description of Runoff Generation at the Scale of the Beholder, Water Resour Res, 57, https://doi.org/10.1029/2020wr027514, 2021.
Melsen, L. A. and Guse, B.: Climate change impacts model parameter sensitivity – implications for calibration strategy and model diagnostic evaluation, Hydrol Earth Syst Sc, 25, 1307–1332, https://doi.org/10.5194/hess-25-1307-2021, 2021.
Novick, K. A., Ficklin, D. L., Baldocchi, D., Davis, K. J., Ghezzehei, T. A., Konings, A. G., MacBean, N., Raoult, N., Scott, R. L., Shi, Y., Sulman, B. N., and Wood, J. D.: Confronting the water potential information gap, Nat Geosci, 15, 158–164, https://doi.org/10.1038/s41561-022-00909-2, 2022.
Ogden, F. L., Allen, M. B., Lai, W., Zhu, J., Seo, M., Douglas, C. C., and Talbot, C. A.: The soil moisture velocity equation, Journal of Advances in Modeling Earth Systems, 9, 1473–1487, https://doi.org/10.1002/2017ms000931, 2017.
Peters, A., Hohenbrink, T. L., Iden, S. C., Genuchten, M. Th. van, and Durner, W.: Prediction of the absolute hydraulic conductivity function from soil water retention data, Hydrology Earth Syst Sci Discuss, 2023, 1–32, https://doi.org/10.5194/hess-2022-431, 2023.
Samaniego, L., Kumar, R., Thober, S., Rakovec, O., Zink, M., Wanders, N., Eisner, S., Schmied, H. M., Sutanudjaja, E. H., Warrach-Sagi, K., and Attinger, S.: Toward seamless hydrologic predictions across spatial scales, Hydrol Earth Syst Sc, 21, 4323–4346, https://doi.org/10.5194/hess-21-4323-2017, 2017.
Vereecken, H., Amelung, W., Bauke, S. L., Bogena, H., Brüggemann, N., Montzka, C., Vanderborght, J., Bechtold, M., Blöschl, G., Carminati, A., Javaux, M., Konings, A. G., Kusche, J., Neuweiler, I., Or, D., Steele-Dunne, S., Verhoef, A., Young, M., and Zhang, Y.: Soil hydrology in the Earth system, Nat Rev Earth Environ, 3, 573–587, https://doi.org/10.1038/s43017-022-00324-6, 2022.
Vogel, H., Gerke, H. H., Mietrach, R., Zahl, R., and Wöhling, T.: Soil hydraulic conductivity in the state of nonequilibrium, Vadose Zone J, https://doi.org/10.1002/vzj2.20238, 2023.
Wagener, T., Gleeson, T., Coxon, G., Hartmann, A., Howden, N., Pianosi, F., Rahman, M., Rosolem, R., Stein, L., and Woods, R.: On doing hydrology with dragons: Realizing the value of perceptual models and knowledge accumulation, Wiley Interdiscip Rev Water, 8, https://doi.org/10.1002/wat2.1550, 2021.
Citation: https://doi.org/10.5194/egusphere-2023-125-RC1 - AC3: 'Reply on RC1', Hongkai Gao, 14 Mar 2023
-
RC2: 'Reviewer comment on egusphere-2023-125', Anonymous Referee #2, 16 Apr 2023
This opinion paper makes interesting and bold claims about the importance of soil properties for hydrology. I agree with many of the statements for natural soils and mature ecosystems. However, the majority of our earth is no longer a natural mature ecosystem. We have changed the surface cover drastically and very large areas are under agriculture or are so badly degraded and not in a mature “steady” state that the ecosystem perspective that is advocated in this paper is possibly no longer applicable. I think that this has to be mentioned in the text and that the reader needs to be reminded more frequently that these statements are made for mature natural ecosystems.
That soil is important is clear in situations where severely degraded ecosystems are restored. It is the restoration of the soil that leads to the very large changes in the flow pathways (from overland flow to subsurface flow) and thus streamflow responses. Indeed, it is the ecosystem that changes the soil properties that lead to the changes in the hydrological flow pathways and runoff responses, but this does not mean that the soil itself is not important at all. It means that the ecosystem has such a large effect on soil that the ecosystem would be a better predictor to be used in models (because ecosystem and soil properties become correlated as the ecosystem matures and the ecosystem is easier to observe), but it does not mean that soil is not important at all, especially not when one wants to understand processes. I think that some of the statements about soil not being important therefore require a bit more nuance. In particular, the model perspective (rather than process perspective) for some of the claims should be made clearer.
One of the confusing parts of the paper is that the authors state that the rooting zone is important but that soil is not important. This seems to suggest that they think that the rooting zone is not part of the soil. I think that what they mean is that soil texture is not important. To me it seems that most of the time when the authors say that soil is not important, they mean that soil texture is not important. For example when the authors refer to soil in the top down approach of catchment comparisons (Section 3.3), they actually refer to texture, not soil hydraulic properties. I urge the authors to more explicitly state that they focus on the soil texture. A better description of what parts of the soil they think are not important would be really helpful. It will also help if they give their definition of soil early in the paper.
The authors should point out much more clearly (and explicitly) that a major problem is that we use texture in pedotransfer functions to derive the soil characteristics that are related to water flow and storage, especially because these pedotransfer functions were developed for agricultural soils. The sand or silt content of a soil do not affect water flow or storage. We only attribute such an effect when we use pedotransfer functions to derive properties related to water flow and storage based on the texture. Because the pedotransfer functions were largely derived for agricultural soils, they do not take the effects of structure (and preferential flow) into account.
The writing of the manuscript could be a bit sharper. At several places, the authors make a good argument for why the ecosystem is important and then conclude that the soil is not important. I think that these sections need to be improved for two reasons. First, reasons are given for why the ecosystem is important but not for why the soil is not important. In particular, no references are given for this second part. In other words, the authors provide arguments for the first part (the ecosystem is important) but not for the second part (soil is not important). Thus either the second part (soil is not important) has to be taken out or arguments and references need to be included for the second part as well. Second, ecosystem and soil are interconnected. It is the ecosystem that changes the soil properties. So one can not directly argue that because the ecosystem is important, the soil is not important. It is still important but the ecosystem is perhaps the better predicting variable to be used in models because it is easier to observe and has a large effect on the soil properties that actually affect how water moves through the soil.
Other parts of the writing could also be improved. In several sentences words are missing and some other sentences are not clear and should be reformulated. The structure of the paper and individual sections was sometimes unclear to me. For example, section 4.1 consists of four paragraphs. Paragraph one highlights the importance of ET and states that hydrologists focus on discharge instead (but this point was already made on L128). The second paragraph then describes that ecosystems maximize storage and drainage. This section is interesting and fits the caption of this section. One would expect the next paragraph to get deeper into this but the third paragraph describes that the numbers for soil properties used in models don’t match the actual measurement values, and the fourth paragraph describes the rebalancing of soil properties that needs to be done in models. While the first two paragraphs sort of fit together and the last two paragraphs as well, the link between the first two and last two is not obvious. It also means that the second paragraph ends abruptly and this line of thinking could use some more elaboration. In addition, the part on the soil properties and the rebalancing starts abruptly without an introduction. The latter two paragraphs would probably better fit in a separate section on the problematic part of using pedotransfer functions based on texture (see comments above). This is just one example, there are other sections where the flow was unclear and I expect other readers to also wonder how the paragraphs are connected. I made some suggestions in the annotated pdf but there are more places where text could be reordered for a better flow. I don’t request that the authors use the suggested order but I do recommend that they carefully read through the paper to see if the order is logical for a reader.
Oter specific points:
- L56/139: I think that the problematic part of the use of pedotransfer functions based on texture to derive properties about pores should be described in more detail. Especially knowing that these pedotransfer functions were developed based on cores from agricultural fields and that texture does not really influence the hydraulic conducitivity (e.g., Jarvis et al., 2013; Gupta et al., 2021). See also comments above.
- Section 3.1: I don’t think that anyone claims that soil affects the long term water balance more than climate and vegetation. So, I think that it is fine to use this section to highlight that the ecosystem and climate are the main factors that determine the long term water balance but it makes less sense to use this as an argument that soils are not important.
- L129-132: Yes, land use change (if severe) alters runoff generation, exactly because of the large effect it has on soils. So, I don’t think that you can use this argument here to say that soils don’t matter. You can use it to make the argument that vegetation has a large effect on the soil properties that actually matter for water flow and storage. Also, it would be good to reference some field studies here (not only model studies).
- L158: But the comparison is basically between a model and a model with more data. I don’t think that one should call this observations.
- L162: But it also mimics the depth to the groundwater – maybe this has a different effect in the two models?
- L181: The problem is in part that we use texture here. Texture does not describe the soil pores that are important for storage or flow of water. The problem is that we use pedotransfer functions that are largely based on data from agricultural soils and are not appropriate for forested systems. See also the comments above. Furthermore, soil depth data is usually very rough and not very reliable. Maps of soil properties that actually describe water flow and storage are rarely available. Thus, one could also argue that the big problem is that we don’t have soil maps with sufficient information on the properties that actually matter and are related to water flow, and that instead we rely too much on texture and pedotransfer functions.
- L188: I agree that all these processes are intertwined or connected. Therefore, I think that the opinion paper should use more careful wording. It is OK to say that for hydrological modeling it is more useful to look at the ecosystem because the soil properties that matter for hydrology are highly correlated with land cover, and ecosystem properties are much easier to observe or measure. However, if we want to actually understand processes and the factors that affect these processes, it is important to look at the processes. In other words, then we have to look at the partitioning of rainfall into infiltration, overland flow, deeper drainage, etc. and soils are important. I think that this distinction between model application and process understanding should be made more clearly throughout the text.
- Section 4.2: I am sorry but I don’t understand what these ERA5 storage volumes contribute to the arguments of the opinion paper. The volume is one thing, the total flux from repeated filling and emptying is another. Certainly, I agree that the total storage is highest in the root zone but I consider the root zone to be part of the soil. So why is the root zone important but soil not? The paragraph on 277-284 goes some way into explaining this but it could have been added to section 4.1. It would be good if the authors give a definition of soil early in the paper. I have the feeling that often the authors mean soil texture instead of the soil itself.
- Several minor comments and suggestions are given in the annotated pdf.
References:
Jarvis, N., Koestel, J., Messing, I., Moeys, J., and Lindahl, A.: Influence of soil, land use and climatic factors on the hydraulic conductivity of soil, Hydrol. Earth Syst. Sci., 17, 5185–5195, https://doi.org/10.5194/hess-17-5185-2013, 2013.
Gupta, S., Hengl, T., Lehmann, P., Bonetti, S., and Or, D.: SoilKsatDB: global database of soil saturated hydraulic conductivity measurements for geoscience applications, Earth Syst. Sci. Data, 13, 1593–1612, https://doi.org/10.5194/essd-13-1593-2021, 2021.
- AC4: 'Reply on RC2', Hongkai Gao, 01 May 2023
- AC5: 'Reply on RC2', Hongkai Gao, 05 May 2023
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 1,633 | 551 | 37 | 2,221 | 13 | 25 |
- HTML: 1,633
- PDF: 551
- XML: 37
- Total: 2,221
- BibTeX: 13
- EndNote: 25
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Cited
1 citations as recorded by crossref.
Fabrizio Fenicia
Hubert H. G. Savenije
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(754 KB) - Metadata XML