the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 30 May 2024
Investigation of Different Roughness Approaches and Vegetation Height Effects on rain-induced overland flow
Abstract. Overland flow is a critical aspect of the hydrological cycle, and understanding its dynamics is crucial for managing water-related issues such as flooding and soil erosion. This paper investigates the impact of various roughness estimation methods on simulating overland flow during intense rain events, with a specific focus on the influence of vegetation. The study assesses various approaches to vary roughness as a function of water sheet thickness and vegetation height, including two different constant Manning's coefficients, a linear approach, an exponential function, a power law function, an empirical formula, and a physics-based approach. The investigation emphasizes the importance of accurate roughness estimation for improving the reliability of hydrological models and enhancing flood prediction capabilities. Experimental data from artificial rainfall experiments on 22 different natural hillslopes in Germany are used to calibrate the OpenLISEM hydrological model, adjusting parameters such as saturated hydraulic conductivity and soil suction at the wetting front. Subsequently, various Manning's coefficient estimation methods are applied, and the model's performance is evaluated numerically.
Preliminary results indicate satisfactory calibration outcomes, with NSE values ranging from 0.75 to 0.95 in most cases for various sites. To validate the models, 100 different experimental rainfall events are used for each roughness method. Validation findings suggest that the physics-based approach, the linear function, and constant Manning roughness, demonstrate the best performance based on NSE values. According to our results, areas with more vegetation coverage demonstrate higher saturated hydraulic conductivity value, indicating that, for two sites with the same soil type, the locations with dense vegetation exhibit higher infiltration parameters. Consequently, it is crucial to evaluate the influence of vegetation on runoff, considering not only its effects on Manning's coefficient but also on saturated hydraulic conductivity.
- Preprint
(1164 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-1276', Anonymous Referee #1, 22 Oct 2024
This manuscript compares performance of several overland flow models of previously published experimental data. The manuscript suffers from poor organization, and it is difficult to sift through details to get the broad picture, and, inversely, to find details that are sprinkled throughout. The focus on detailed reporting of statistical fit parameters throughout limits the utility of the manuscript because it comes at the expense of clarity in the reasons for the performances. The introduction does not separate detailed thoughts from main ideas, and it even contains methodological choices mixed in with the background. Two of the three objectives listed in bullet points too broad to be tested. The discussion is written mostly from the perspective of which models perform better practically, not on what modeling concepts best match the physical processes they are intended to mirror. This perspective is especially limiting given that the paper is focused mainly on one set of experimental results—no matter how good those experiments are. The conclusions oddly emphasize the importance of modeling infiltration, rather than on roughness as emphasized in the introduction.
Detailed comments
L28 citation please for constant velocity profile with emergent vegetation
L58 citation please for OpenLISEM. The Jetten (2002) ref is for LISEM. Maybe that’s being picky, but it’s important to be precise about the theory and methods.
References to the theoretical origins of LISEM would be much better. The model predates 2002. Also Jetten 2002 is difficult to obtain.
L100 the first two sentences are unrelated to Methods
L143 assumed by who and for what purposes?
L153 impact on what?
L159-165 are not study site.
Fig 3 is the overland flow mm/min? It would be better if the rain and flow were the same units, but at least the time basis for the flow must be specified.
The naming of the experiments is difficult to access. Why not use more intuitive names instead of number codes?
Percent bias is not defined anywhere. It is also inconsistently named, e.g., as “percentage of bias,” “bias percentage,” and “pBias.”
Saxton and Willey incomplete reference
L225 this paragraph is an example of how poor organization makes it difficult to follow this manuscript. In six sentences, there is a note on anomalously low NSE at one site, a comparison of runoff and rainfall rates at one site, a discussion of how runoff quantity affects model fits, a description of a figure that duplicates figure captions, discussion of differences in Ksat obtained by various methods, and a note that best-fit soil moisture at two sites was different than elsewhere. There is no way to assimilate this information to form a comprehensive picture, and there is no hope of later referring to these disorganized facts.
The results are wordy, overdetailed, and repetitive. For example, most of p 12 is not needed: Table 4 is not needed, L275-280 not needed. L285-288 not needed. Most instances of the word “value” are not needed.
Fig 8 and 9 lack Y axis labels and their meaning is obscure unless the reader remembers how many models there are. Overall, these figures are of low value.
Fig 10 why are some cells green?
L321 if these sites should not be compared to the others, they should not be presented in a figure comparing the sites.
Discussion on the importance of antecedent conditions ought not lead with a citation to a catchment modeling text. Catchment responses are not the subject of this work.
The frequent citation of Feldman et al. (2023) indicates insufficient original content in this manuscipt.
The discussion is just as disorganized as the rest of the manuscript. For example, section 5.3 on vegetation coverage is mainly focused on Ksat and infiltration. I thought this paper was about roughness?
Citation: https://doi.org/10.5194/egusphere-2024-1276-RC1 -
AC1: 'Reply on RC1', Azam Masoodi, 06 Nov 2024
We appreciate the reviewer’s insightful and constructive feedback and intend to revise the paper accordingly. The misalignment between the introduction and conclusion may have resulted from an evolving focus in our study. Initially, our aim was to evaluate how different roughness estimation methods affect the accuracy of model predictions when compared to physical measurements. However, our results revealed that, in the presence of vegetation, saturated hydraulic conductivity plays a critical role that cannot be ignored. This insight led us to emphasize infiltration parameters in the conclusions, which may have created some confusion for readers.
To address this, we will revwrite the abstract, introduction, and discussion to clarify the role of infiltration in our study and refine the stated objectives to provide a clearer framework for readers.
Additionally, we compared our results with a related study by Feldman et al. (2023), as both studies share the same experimental data set and a similar objective but use different approaches. Feldman et al. extrapolated a nearly constant infiltration rate for the falling limb of the hydrograph, which allowed them to separate the impact of roughness on hydrograph shape from infiltration effects. Our approach, however, considers infiltration across the entire hydrograph. To highlight the differences between these approaches and the importance of considering infiltration effects, we referenced to this paper frequently. However, we did not emphasize the relation to the paper enough in the introduction, which will be changed in a revised version.
Below are our responses to the detailed comments, along with our plan to improve the manuscript’s clarity and organization.
Detailed Comments
Comment L28: The sentence has been modified for better clarity, and a relevant citation has been added at this point in the text.
Comment L58: OpenLISEM is the open-source version of LISEM. We explained about it in L58 and updated the citation in L60 and 63.
Comment L100: these sentences were deleted.
Comment L143: This text “Nevertheless, they do not provide information on the ℎ₀ value. Consequently, this parameter is assumed to be 5 times the plant height for each experimental site” was changed to “As Feldmann et al., (2023) did not provide information on the ℎ₀ value, we assumed ℎ₀ to be five times the plant height to apply the Kadlec’s method in our study.”
Comment L153: “on overland flow” was inserted to the text.
Comment L159-165: We changed the title of this section to “Study site and experimental setup”
Comment on Figure 3: The picture was modified.
Comment on Experiment Naming: A table describing each experiment was added in Section 2.3 for easier access.
Comment on Percentage of Bias: It was modified in the manuscript.
Comment on Saxton and Willey Reference: The reference was completed.
Comment on Figures 8 and 9: These figures are modified.
Comment on Figure 10: An explanation for the green cells is added below the figure.
Comment on L321: We removed these sites from the Figure. 11
Comment on antecedent conditions: We have modified the discussion on antecedent conditions and removed the citations related to catchment modeling.
Citation: https://doi.org/10.5194/egusphere-2024-1276-AC1
-
AC1: 'Reply on RC1', Azam Masoodi, 06 Nov 2024
-
RC2: 'Comment on egusphere-2024-1276', Anonymous Referee #2, 22 Nov 2024
Especially in light of the increase in heavy rainfall events with enormous damage potential and the associated great efforts of the federal states to create high-resolution heavy rainfall hazard maps, the further improvement of hydrological and hydraulic modeling is of great importance. Particularly with regard to the quantification resp. modeling of the influence of different vegetation conditions on runoff formation and flow parameters of overland flow, I also see an urgent need for research. In this respect, the authors’ commitment is very welcome, and the results can provide valuable input both from a scientific perspective and for practical application. In the publication of the research results, I still see potential for improvement overall, in order to present the findings more transparently and comprehensibly for the professional community.
Before I list my specific comments and questions in detail, please allow me to address some general critical-constructive comments:
The title raised my expectation of an examination of the physical (and particularly the hydraulic) processes of surface runoff. However, the methodological focus is exclusively on data-driven analyses with little to no engagement with the actual processes. The influence of vegetation height as an input parameter is also only evaluated in a rudimentary or purely model-based manner. Therefore, a title such as “Investigation of Different Roughness Approaches in Hydrological Run-off Modelling” would be more appropriate.
Regardless of the chosen title, I would have fundamentally wished for a deeper engagement with the physical phenomena and their modeling representation when evaluating the predictive capability of the investigated model approaches. This is especially relevant in terms of comparing the simulated and measured hydrographs, where differences between model and measurement set-up could also influence the results. The possible influence of the microscale surface structure, which together with the vegetation determines the flow resistance (friction and drag), should also be mentioned more visibly.
The presentation of the literature review on vegetation hydraulics in the introduction does not differentiate between studies on channel/river hydraulics and those on overland flow (thin-layer surface runoff) due to heavy rainfall. However, the boundary conditions of the mostly highly empirical studies are completely different in terms of slope, flow depths, vegetation situations, and thus only allow limited mutual conclusions.
Since the present contribution exclusively uses field experiments with very low flow depths of a few centimeters for the validation of the approaches, and there is (presumably) hardly any significant submergence of the vegetation, the presentation of the modeling approaches and third-party research results could perhaps be limited to the investigated ranges to avoid misunderstandings (in the sense of “less is more”).
I am somewhat uncertain to what extent the NSA value is sufficient as a key parameter for evaluating the model quality. Especially when it is balanced over the entire hydrograph, the advantages and disadvantages of the methods and, above all, the reasons for deviations are difficult or impossible to understand. Therefore, the red thread for comparative evaluation and the conclusions were only partially comprehensible to me. More comparative illustrations as shown in Figure 3 and a detailed process-related discussion of agreements and deviations might be helpful.
Line 1: Titel misleading (see comments above)
Lines 19-21: Is this conclusion a new insight (sounds like) or just the confirmation of an expected dependency? Is it really the result of your model applications or rather of the measured data on the plot (Accounting inflow/outflow on field plot)? Are there possible other effects (e.g. increasing pressure due to higher waterlevels) or model-wise dependencies which have an impact on infiltration?
Introduction: Many sources refer to studies of river hydraulics (not overland flow) with other boundary conditions (e.g. slope, water depth). This should be highlighted or even reduced on relevant approaches (see comments above). A compact overview can be found in DWA-M 524.
Line 35: Oberle et al. (2021) presents a comprehensive literature study on hydraulic flow resistance of overland flow. In addition, a recommendation for depth-dependent roughness values is derived from different laboratory experiments with artificial grass (Yörük, Karantounias, Ruiz Rodriguez).
Line 42: “approximated” instead of “effectively characterized”.
Line 70: This was not the main focus of the paper. Investigation of vegetation effects… reference to modelling is missing.
Chapter 2: General it would be easier to read if the model and the study site would be explained one after the other. Like e.g.: Presentation of Study site and its boundary conditions -> leads to this model -> leads to different approaches which have been tested -> leads to sequence of modelling.
Or the other way around: These are approaches from literature which we want to examine. For that we build this model based on this study site…
Chapter 2.1: The model looks diagonally inclined although it seems that the study sites of Ries et al. are sloped orthogonal. If it´s like that the deviation of the model from the original (different specific discharge distributed over the plot) and impact on results needs to be discussed. Furthermore, the discharge of the model at the outlet is measured instant in contrast to the set-up from Ries where the discharge measurement is from my understanding 10-20 m (drainage tubes) after the plot. This should lead to distorted results later in the evaluation and needs to be discussed. Also, a rough overview of ranges of discharge, velocities and water depth on the plot should be given. Maybe some maximum values for a better understanding.
Line 85: Size 1x1. Noticeably coarse resolution in hydraulic. Have cell size sensitivities been checked?
Line 86: Possibly misleading. Micro depressions where the water is retained and infiltrating after a rain event are not accounted for in the roughness factor. Only water that can later drain away is considered as well as small geometrical variances (drag force) additionally to frictional resistance.
Fig. 1: A picture of the experimental setup of Ries et al. (2020) next to the model would be helpful for the reader.
Lines 93 – 96: The model-related influence of Manning on infiltration rate should be discussed. How are the parameters related to each other in the model?
Line 105: Resistance factors for river hydraulics (different slope, water depth). Needs to be mentioned.
Line 110: Feldmann's approach/procedure should be explained in more detail, as comparison to him is a bigger part in the paper.
Table 1. Site 19. n Chow. Format not in line with others.
Line 115: Oberle et al. (2021) recommends a roughness spectrum based on experimental experiments to simulate overland flow in 2D hydrodynamic models. Investigations had been done by different authors.
Eq. 1: Should be 1 / nmanning = … ?
Line 121: Are higher flow depths in the model even achieved or is the linear method also a “constant method”, since only flow depths smaller than the vegetation height occur?
Line 123: Strickler k -> kStr
Line 126: It´s not clear whether the actual vegetation footage from Reis is considered using these equations. If it´s like that more information is needed how it´s done and what blockage factors have been calculated. What value was used for Cd?
Line 136: Explanation c and d. One sentence how Feldmann calculated them and what is the value from them.
Line 145: Oberle at al. (2021) did not confirm factor 5. As no vegetation height was considered in the study at that time. Short side note: New measurement results based on experiments on a natural plot (grassland, different vegetation conditions) will soon be published (see also Oberle et al. 2024 Dresden Wasserbaukolloqium).
Line 147: plant basal cover can be much lower than canopy cover depending on vegetation type. It´s to be expected that plant coverage of Ries is canopy cover? Needs to be addressed that there are uncertainties.
Chapter 2.3: Also, if possible a range of expected velocities, water depth or at least the measured specific discharge in the experiments would add benefit to the readers. Would it make sense to separate the sites into the ones with arable land (with and without vegetation) and grassland? Maybe you find a trend by looking at them isolated rather than all mixed up.
Chapter 3: How does the start value of Manning effects the calibration of ksat and Psi? Have the final ksat and psi been checked for plausibility? In addition to the NSE, an isolated consideration of the specific characteristics of the hydrograph seems beneficial. (see comments above)
Fig. 3: unit? Also, more of these graphs would give greater insight. Moreover, differences in rising and falling limb could be influenced by differences between model and experimental set-up and should be discussed. Also, on the rising limb micro depressions could have an impact (smooth model in contrast to nature). Fundamental discussion (instead of only focusing on NSE & bias) against the background of the methodology used would be desirable.
Fig. 4 & 5: Systematic differences are noticeable. These should be discussed and evaluated. Site seems to have a greater influence on the parameters than the different methods.
Fig. 10: Please check if the unit is correct? Discharge for the whole plot or specific discharge? Why is 3 parameters so bad? Discussion should be more profound.
Fig. 11: Show also psi per site? See also comments to Lines 19-21
Line 352: As mentioned above: Are you sure that submerged vegetation has been investigated or are the water depths shallower?
Chapter 5.3: Site (soil type?) was more dominant than vegetation and should be put into relation.
Line 412: Is this new insight (see comments above)?
Line 416: e.g. RoGer-Modell?
Citation: https://doi.org/10.5194/egusphere-2024-1276-RC2
Model code and software
modified_manning_console Philipp Kraft https://github.com/philippkraft/openlisem/tree/modified_manning_console
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 248 | 114 | 23 | 385 | 16 | 23 |
- HTML: 248
- PDF: 114
- XML: 23
- Total: 385
- BibTeX: 16
- EndNote: 23
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1