Comparing the Runoff Decompositions of Small Experimental Catchments: End-Member Mixing Analysis (EMMA) vs. Hydrological Modelling

Round 1

Reviewer 1 Report

I found this paper of interest and highly recommdable

I suggest publication after minor changes

Comments for author File: Comments.pdf

Author Response

Dear Reviewer,

We're greatful a lot for your effort to review our manuscript. We revised the text according to your recommendations.

Author Response File: Author Response.docx

Reviewer 2 Report

Although this is an interesting study, it may be useful for hydrological modelling results validation. However, the article requires some revisions. Please address the comments to improve the quality of your article. 

1. Introduction needs improvement. For example, in introduction line number 55-57 authors mansion "...scale physical laws...", I recommend that the authors provide the following references: (a) Sarker et al. (2019), Critical Nodes in River Networks, Scientific Reports. https://www.nature.com/articles/s41598-019-47292-4, (b) Gao et al. (2022), Analyzing the critical locations in response of constructed and planned dams on the Mekong River Basin for environmental integrity, Environmental Research Communications, https://iopscience.iop.org/article/10.1088/2515-7620/ac9459.

2. Figure 1 should be revised; I would superimpose the river network on the Catchments elevation map. See the aforementioned manuscript.

3. What is the novelty of this study? Please explain more about the possible implication of this study.

4. The figures 2,5,6,7, and 8 are terrible! Using specialized software such as Python or MATLAB, I would generate high-quality figures.

Author Response

Dear Reviewer,

We're greatful a lot for your effort to review our manuscript. We revised the text according to your recommendations.

Author Response File: Author Response.docx

Reviewer 3 Report

The idea of the reviewed manuscript is to compare end-member mixing analysis results to the outputs of three well-established rainfall-runoff models, based on observed data from two neighbouring small catchments in the Pacific Russia region. This region is dominated by summer rainfall runoff due to strong monsoon circulation, and small catchments are less inertial, or – in the same sense – more reactive to rainfall events.

My general comments are given in the attached pdf file. The manuscript will benefit from adding a more quantitative touch to certain points of reasoning. E.g., a randomness test results will support the authors’ conclusions driven by the visual inspection of the residual plots (Figure 2). EMMA end members must be put in correspondence with RR model outputs, since by default they do not necessarily converge to one another. I.e., rainfall, soil water and groundwater EMs might largely stick to surface, lateral and return flows of the SWAT model – is it in fact the case in the sense that they indeed converge and have the same hydrological meaning?

Also, EMMA EMs must be explained in terms of water chemistry, in the sense that each tracer component might have a characteristic range in each end member, and their lack of overlap along the observed time series might support the end member isolation and decomposition results.

In the SWAT sub-section, I highly doubt the calibration output concerning GWQMN parameter, and potentially the REVAPMN parameter as well, for the Elovy catchment. In the combination, this parameter setup assumes that the authors, in their calibration efforts, struggled to repartition the water between the lateral and return flows, while constantly getting surface flow close to 0, leading to oversaturation of the root zone – then GWQMN is set to 0 to enhance return flow and drain the shallow aquifer so that it could receive more water from the root zone. At the same time, REVAPMN is also set to 0 to compensate the rapid draining of the root layer via both lateral and return flow, reintroducing the return flow to the root zone – also high GW-REVAP, which is tweaked to maximum, plays to this end, and high ESCO, enhancing root zone drainage via evapotranspiration. I really wonder if the differences in parameter set between the two catchments are supported by actual differences in physiography//water cycling.

I can not say anything about the other models, but for SWAT, I would suggest redoing the Cal/Val routine and sensitivity analysis using SWAT-CUP/SUFI2 software, and inspecting/interpreting the results carefully. As with most if not all models, SWAT ‘out-of-the-bundle’ solutions might not be the best ones. For the SWAT model not getting surface runoff correctly, I would check several things: first, verify that you have the “print log flow” box unchecked – if it is checked, the surface flow is printed in logarithmic and is indeed too low; second, play with SOL-AWC and ESCO to retain water in the root zone, and probably set a higher threshold for both REVAPMN and GWQMN to route the water toward the surface flow.

The same but in the other direction could be true for the Medvezhy catchment, with slightly shorter slope length, where most water is cycling through the root zone without getting either to the shallow aquifer or to the surface as surface runoff. Potentially, the HBV model finally suffers from the same or similar inconsistencies, as is supposed from data on Figure 8.

I also note that normally SWAT model requires a LULC map which is not presented in the manuscript, but should be provided upon revision.

In this scope, I would recommend a major revision of this manuscript. The modeling component is an important part of this research alongside EMMA caliculations, and as such it must be better substantiated. Or alternatively, as it seems that the ECOMAG model is behaving quite well as compared to EMMA, unlike others, then probably SWAT and HBV could be omitted. 

Comments for author File: Comments.pdf

Author Response

Dear Reviewer,

We're greatful a lot for your effort to review our manuscript. We revised the text according to your recommendations.

Author Response File: Author Response.docx

Reviewer 4 Report

I have read the manuscript “Comparing the runoff decompositions of small experimental catchments: end-member mixing analysis (EMMA) vs hydro-logical modelling” by Bugaets and co-authors. This manuscript compares results from an end-member-mixing-analysis (EMMA) with  the results from 3 different well known hydrological models. The results of this paper are interesting and show that the implementation of EMMA streamflow segmentation can be a valuable tool for model evaluation. Although the paper/manuscript provides an interesting perspective, there are still some issues that I think, could/ should be improved before publishing this article in Water. I have added my comments and suggestions in the main comments and detailed comments part of my review.    

One of my main concerns is the effect that manual calibration (and the general equifinality of such models) might have on the different model performances, which cannot be entirely disentangled from model calibration performance. I guess it would be a more fair comparison if all models were inversely optimized with the same data (although we do still deal with equifinality in that case). Although I think it is an elegant approach to not calibrate the models on their streamflow separation behaviour, performing such calibration would give more outcome on the difference in capability to fit both the data and the streamflow separation behaviour. I think the authors should keep this into account when analysing their results and making statements about differences in model performance.      

----------------------------------------------------------------------------------------------------------------------

Main comments:

1)    Methods structure: I would recommend the authors to consider putting section 2 and 3 together as a overall “Materials and Methods” part of the manuscript. I mainly suggest this, because parts described in section 3 relate to text in section 2 and having it as one overall section might make these parts of the manuscript more connected. One could still maintain the structure Methods: by keeping the current segments as sub-sections.

2)    In the methods section, sometimes details are missing that make the research at the moment not very reproducible. A lot of information about sensors, but also sometimes other method related details are missing. More details on this and suggestions for improvement(s) have been added to “detailed comments”. Making data and/or code available in future publications will help to make the publication even more reproducible, which I would also recommend to the authors. 

3)    In section 3 (Methods of Modelling), the models are described in quite some detail, but for the parameterization, the reader is referred to other modelling studies that were performed for the same catchments. I personally think it would be very important to shortly describe the parameterisation of the different models in this manuscript, as this helps the reader to compare the setup of the different models directly in the current manuscript. Currently, the author is only guided to the parameterisation (Appendix A-C) in the result section of the manuscript. I would recommend to add the parameterisation in the methodology and leave the description in the Appendix. If the authors wish to keep it in the current structure, I would at least recommend referring to the parameterisation in the methodology instead of in the results section. Part of the text in appendix also seems to repeat model description text in the model methods sections.

4)    EMMA: The method section about EMMA (3.1) reads more as a lit. review on what EMMA is and how it can be used, although it also has some practical equations in it.  From my point of view, this should include more information on  the application to the current case study – How is EMMA used in the current study? Which programming language was used to perform the EMMA? How is conservatism of the tracers guaranteed? Part of this is written down in the results section of the paper (4. Emma results) – I think part of the description of the analyses in here should be shifted to the methodology section. 

5)    Although the authors recognise that the model parameterisation can have a pronounced effect on the output of the models (see introduction), the current manuscript just compares the different models, without considering that they all have been manually set up.

6)    Layout & graph quality – sometimes Tables and Figures do not have an ideal layout in the manuscript yet (e.g. also alignment and size). As most of the layout is done by the authors for these type of journals (as far as I am aware), I propose to review the location/position and size of all tables and Figures. Sometimes the quality of the graphs on the pdf is not very high (e.g. figure text in Figure 2) and sometimes is cut off (e.g. years in Figure 6). I would recommend the authors to provide as high as possible figure quality, plus to avoid having any text cut off in the final version of the manuscript. 

7)    Structure of the results in the publication – Besides adjusting the methods section, I would recommend the authors to structure the paper more generally (Intro, Methods, Results and Discussion, Conclusions). First of all, the authors are discussing their results in the results sections without naming it in the section. Plus, all results can be combined in one results section with sub-headings. 

8)    The authors describe that “Calibration was performed manually using a trial-and-error approach”. As far as I understand this means that all three models have been calibrated manually. Considering that in all cases, not the “best” parameter set is used to forecast discharge conditions, differences in performance cannot only be coupled to the model structure, but can theoretically also be related to the way the models are manually calibrated. I think that especially this aspect needs to be covered in the manuscript as a discussion point. Considering the equifinality point made in the introduction, this is also an important consideration. Furthermore, considering that none of the models have been calibrated (inversely optimized) based on the mixing and purely based on manual calibration, the difference in performance can also be partly resulting from the subjective choice of the final model parameterisation.

9)    Results structure: Although this paper presents sufficient results, the results could be interpreted in more detail I think. Especially related to the process based understanding of the catchments. First of all, it would be great if the authors could very clearly define if EMMA explained the differences in hydrological behavior between the catchments. Furthermore, it would be great if the authors could explain why a certain model would potentially be better at predicting certain components of the discharge curve. What are the mechanisms captured in the specific model that cause that model to outperform. Maybe this could also beforehand be brought up as a hypothesis in the methodology and then tested with the obtained results.

10)    Overall Text: Although I generally had little problems reading the text of the manuscript, and I think that the text was generally clear, I did find some small text issues. In the detailed comments, I have also added some comments related to sentences that could potentially be improved (because I might have not entirely understood them/ because a word might have been missing). I have made sometimes suggestions for improvements here.

-------------------------------------------------------------------------------------------------------------------

Detailed comments:

1 – Introduction

1)    In line 36, the authors state “The small catchments offer a possibility to get less noise in the data compared with a macro-scale data”. It would be great if the authors explain this in an additional sentence. Plus, it would be great if the authors could give a source/ citation that supports this statement.
2)    In line 55/56, the authors introduce HRU-based models: “The main reason to hang on this empiric knowledge is the scale-dependency of HRU-based models, where they are more appropriate than any small-scale physical laws [18,19].”  I noticed that in section 3, the abbreviation is explained, but this should (from my point of view) already be explained in the introduction.
3)    In Line 59 – 61, the authors state: “Model structure uncertainty is as impactive as parameter uncertainty, if mathematical formulation is not based on fundamental laws but adopts a set of lumped functions relating impact to the response [18].” This sentence is slightly unclear to me. A comparison is made here – model uncertainty has as much impact as parameter uncertainty. Here it is unclear to what subject the impact refers – impact on what? The outcome of the model? 
4)    Line 81 – 86 – This sentence is quite long. I would propose to create single smaller sentences here, as that would improve the readability of this last section of the introduction.

2 – The study objects, data and measurements methods description

5)    Line 88 - “The studied territory relates to the Pacific Russia boreal forests” – I assume “relate” is not the correct word here. Propose rephrasing.
6)    Line 89 – 91: Here for two small river catchments (Elovy (3.5 km2) and Medvezhy  (7.6 km2) creeks) belonging to the Upper-Ussuri Biocenological Experimental Station (45 90 km2, 44°02' N, 134°11' E) runoff modelling and hydrograph separation were performed.” – Propose to change order of sentence: “Here, runoff modelling and …. Were performed for two small catchment… that belong to the ….”
7)    Line 92: “moderately steep (locally very steep) hillslopes” – can this statement be improved by adding steepness values – e.g. slope gradients? A reference to the table with slopes?
8)    “Air temperature goes through strong variability from year to year” – consider rephrasing.
9)    Line 98 - 99: “Unstable intra-annual and long-term precipitation regimes of the territory define runoff formation conditions of the investigated catchments.” Perhaps the authors can elaborate a little bit more on this. What different runoff conditions are found in this catchments during different precipitation regimes?
10)    “Resources of groundwater are not significant due to fractured rocks.” It would be great if the authors explain the significance to what – to the water balance/ to runoff generation?
11)    Line 103: “The range of maximal daily heavy rains is 100–200 mm.” This seems to be a huge amount of rainfall, right (ca. 14 – 40% of the total yearly P, depending on real amount and total yearly P)? It would be great if the authors could explain how frequent this amount of daily P is happening? Do they have examples of such extreme events in the basin? What is the source for this statement? Also for the other discharge/P statements, it would be great if the authors could supplement that with a citation/source (can of course also just be the measured Q and P).
12)    Line 106 - 107: “Two weather stations were mounted at 650 (Elovy catchment) and 750 m a.s.l. (Medvezhy catchment) to record air temperature and humidity, wind characteristics and precipitation 108 volume (see Table 1)”. It would be great if the authors could state which exact equipment was used – brand and model of the sensors installed for the two weather stations. This could be either added to the text or added to Table 1. The same accounts for the hydrostatic water level loggers. These details are provided in [34] and should be also added to this publication!
13)    Line 113: “Rain, stream and soil waters were sampled” .Water in an uncountable noun and does not exist in plural. I would propose to rewrite the sentence, e.g. “Rain, stream and soul water samples were taken … “. Besides that, the authors do not describe in detail how the samples where chemically analysed on site (equipment used, brand, specific sensor(s)). Also, too little information is provided about the way the samples were analysed in the lab. Again, here the lab equipment (ICP-MS?) used for the analysis is not mentioned and a detailed description of the sampling handling (e.g. dilution or other things that were done with the samples before the chemical analysis with lab equipment X was performed).
14)    Line 118 mentions “fair profiles number” – It would be great if the authors could specify what they mean by this. I even downloaded source [34] and searched for “fair” but could not find this in the text. Do the authors mean that they have analysed some soil profiles and have made a digital map based on these “representative” soil profiles? It would be great if the authors could explain this a little more in the text.
15)    “…and the digital soil map was developed (see Figure 1).” After looking at reference [34], I understand that the authors refer to Figure 1 of that article. However, this is not clear when you are reading  the current manuscript. I also think that adding the soil types distribution to this manuscript in Figure 1 would be beneficial! Therefore I recommend adding this information to Figure 1 of the current manuscript.
16)    Figure 1 – in this figure height information is provided – It would be great if the authors mention the units of the shown height somewhere in the figure (legend?)/ caption if possible. On the figure also a “laboratory is shown. I was curious to know why the authors choose to display this information? Is this the laboratory where certain samples were analysed? If this is the case, the authors should mention this more clearly in the manuscript. If this information has no clear relevance to the current study, I propose to remove such information.
17)    In lines 120 – 121, the authors state “ Brown forest soils here are characterized by a high rock fragments content (greater than 80% of profiles volume), which leads to quick infiltration rates.” If the soils have such a high rock content, are they then not classified as “Leptosols” which have “parent material: Various kinds of continuous rock or of unconsolidated materials with less than 20 percent (by volume) fine earth” – WRB, ISIC 2015? Or is the soil still weathered sufficiently to call it a cambisol? Just a question, because I was unsure about the official criteria/ classification here.
18)    Line 124: “Soil water samples were taken with the vacuum lysimeters installed at the depth of 0.35–0.80 m.” – add also here which lysimeters were installed – brand + model if possible.

3 - Methods of modelling
19)    Line 155: “Overall, this could be interpreted as a sophisticated way of the runoff sources observation.” Feels like something is missing in this sentence, consider rephrasing.

4 - EMMA results
20)    Line 318: “The GW is related to streamflow samples taken during the low flow period and represents the baseflow component of the runoff.” To illustrate this easier, these samples could be colored in Figure 3. Alternative , the flow conditions could be color-coded in Figure 3 to generally show how the flow regime relates to the mixing.     

5 - Hydrological simulation results

21)    The authors provide graphs of the simulated and observed discharge in Figure 5. First of all, I think it would be great if the authors could mark the observations more clearly, so that we can see at which points in time the measurements were taken (either using markers e.g. x to show the locations of the observations and using the line for the interpolated discharge curve; or just showing the observations with markers and leaving the line cure out for the observations). Secondly, it would be great if the authors also show the non-log distributed curves, so that we can also see how good the higher discharge peaks are reproduced by the different models. Furthermore, I wonder if the authors can also show the precip input on the top of the panel?   It would also be interesting how similar/ different cumulative Q curves look between observed and modeled data.

6 - Hydrograph separation results and discussion

22)    Figure 6. I generally really like the color-based source segmentation for the figures. That’s very clear! Some small remarks: the years in the figure are slightly cut off. Plus, I would probably add the name of the catchment in the title (where the year is now). I also recommend the authors to number the sub-figures (a-x). This makes it directly clear what catchment we are talking about in which sub-figure.
23)    Figure 6: An option would be to add a cumulative discharge curve with component distribution as well.
24)    Section 6 starts with a lot of discussion points – what to take into consideration when looking at the results. I propose to first discuss the results and then the limitations here.
25)    “Next, monthly data…” Consider starting a new paragraph here!
Conclusions
26)    “Also, it is necessary both the observation techniques improvement and modelling algorithms, and the accumulation of an extensive dataset on hydrograph separation” – Do not really understand this sentence. Consider rephrasing.

Author Response

Dear Reviewer,

We're greatful a lot for your effort to review our manuscript. We revised the text according to your recommendations.

Author Response File: Author Response.docx

Round 2

Reviewer 2 Report

Thanks for the revision. 

Reviewer 4 Report

I had read the revised version of the manuscript and the replies by the authors to the comments/suggestions. I think that the manuscript can be published in its current form.