CAMELS-Chem: augmenting CAMELS (Catchment  Attributes and Meteorology for Large-sample Studies)  with atmospheric and stream water chemistry data

Sterle, Gary; Perdrial, Julia; Kincaid, Dustin W.; Underwood, Kristen L.; Rizzo, Donna M.; Haq, Ijaz Ul; Li, Li; Lee, Byung Suk; Adler, Thomas; Wen, Hang; Middleton, Helena; Harpold, Adrian A.

doi:https://doi.org/10.5194/hess-28-611-2024

Articles | Volume 28, issue 3

https://doi.org/10.5194/hess-28-611-2024

© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-28-611-2024

© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 28, issue 3

Research article

|

13 Feb 2024

Research article |

| 13 Feb 2024

CAMELS-Chem: augmenting CAMELS (Catchment Attributes and Meteorology for Large-sample Studies) with atmospheric and stream water chemistry data

Gary Sterle, Julia Perdrial, Dustin W. Kincaid, Kristen L. Underwood, Donna M. Rizzo, Ijaz Ul Haq, Li Li, Byung Suk Lee, Thomas Adler, Hang Wen, Helena Middleton, and Adrian A. Harpold

Download

Final revised paper (published on 13 Feb 2024)
Supplement to the final revised paper
Preprint (discussion started on 08 Mar 2022)
Supplement to the preprint

Interactive discussion

Status: closed

RC1:
'Comment on hess-2022-81', Anonymous Referee #1, 08 Apr 2022
Evaluating the overall quality of the preprint ("general comments"),

Sterle et al. present a compiled novel dataset of water quality solutes and atmospheric deposition inputs for the CAMEL catchments. Their work augmenting existing and widely used CAMELS datasets is needed for further research in analyzing spatial and temporal water quality trends in minimally impacted watersheds. Existing papers have done large-scale water quality analyses, but few have provided open-access datasets and the breadth of solutes.

At its core, this is a data paper. As such, I think the methods need to be expanded. My comments primarily pertain to the data and methods, as I see that this is the paper's novelty.

The CAMELS dataset is widely used, and the addition of water chemistry provides the opportunity for analysis. However, I think the dataset could be improved substantially. This paper by Sterle et al. provides an excellent resource for the community.

Methods and Results

Length of dataset

In the paper, the dataset is stated to end in 2014 (see Data Comments below because I'm not sure if this is accurate). However, in many cases, solute and discharge data are available until the at least end of the NADP reporting period. Soon 2014 will be 10-years ago, and I worry the data will not be quickly obsolete and not used to its fullest capacity. The value of this dataset would be exponential if the authors were able to harmonize data from various agencies.

The USGS has developed methods to generate the longest timeseries possible for watersheds. However, I do not see these methodologies applied to these CAMELS watersheds.

Three different approaches can be used alone or in concert to expand the dataset.

The following USGS report looked at the statistical differences between similar solutes. They were then able to merge solutes that were registered under different parameter codes but were statistically similar, allowing for a richer dataset.

Secondly, the USGS report also has a methodology for pairing gauges, and sampling locations are not co-located.

With dataRetrieval package in R you can pull USGS, state, tribal, and NGO water quality measurements. This can potentially expand the dataset. However a lot of manual data cleaning is required because of the poor metadata for non-USGS agencies (ex. nitrate can be reported as N or NO₃, and not explicitly stated in the metadata) (Sprague et al. 2017). With the data harmonization from multiple sources, I would expect a section on the methodology of data cleaning.

Oelsner, G. P., Sprague, L. A., Murphy, J. C., Zuellig, R. E., Johnson, H. M., Ryberg, K. R., Falcone, J. A., Stets, E. G., Vecchia, A. V., Riskin, M. L., Cicco, L. A. D., Mills, T. J., & Farmer, W. H. (2017). Water-Quality trends in the nation's rivers and streams , 1972 – 2012 — Data Preparation, Statistical Methods, and Trend Results Scientific Investigations Report 2017 – 5006. U.S. Geological Survey Scientific Investigations Report 2017–5006. https://doi.org/10.3133/sir20175006

Figures capture the various hydrological/biogeochemical metrics for select solutes in section 4. However, as a user, I find it challenging to evaluate whether the data would be sufficient for my use case. The authors have included some analysis of data coverage in section 3.1, however, since the strength of this paper lies in the data there could be more information to help users evaluate whether this dataset suits their needs.

There should be summary figures for all solutes, so users can adequately assess whether the dataset is appropriate for their use. I think the paper would benefit tremendously if the dataset had more metadata and signatures/summary statistics. Specifically, in Addor et al. (2017), the authors summarized many indices and described the indices in great detail (see Table 2 and Table 3 in Addor et al). I suggest summarizing information about (1) missing data/data gaps, years of continuous data, (2) low/high flow distribution, (3) FDC the WQ spans (figure 7), (4) seasonality of hydrology and solutes, and other metrics that the authors deem useful.

Data Comments

These comments pertain to the files on the google drive.

Data provided has inconsistencies in the way the date is reported. Example from camel_chem_v3.

Gauge ID 14309500: sample_timestamp reads 8/15/67 15:00 whereas other dates are in 2010-12-11 format.

Gauge ID 6447000: sample_start_dt appear to indicate that the data from 1950s. If so, this falls outside of the time period listed in the paper.

It appears that the data available in camel_chem_q_v1 ends in 2018. Please update the manuscript with the correct dates if this data is available.

The original CAMELS dataset provides Shapefiles, however, to allow for seamless merging of data, the header used to identify which column the watershed IDs are the same as the name used by the original CAMELS dataset.

The dataset provided should be able to stand on its own without needing the other CAMELs dataset. The watershed metadata should thus be included (area, outlet latitude and longitude, USGS gage number with leading 0s).

Individual scientific questions/issues ("specific comments"),

Line 84-87: Authors state that they have WQ data from 1980-2014. Later the authors state that the data is "for the same time period" as NADP data (1985-2019). Dates should be consistent.

Line 103: Section 2.2 should be written in a high-level abstract way. I find it unclear how these frameworks are applied to this data in the way it is currently written. I would be better to see more specificity and names of the database (ex. NADP, NWIS).

Line 125: It is unclear whether "daily average discharge" means a continuous dataset or just discharge measurements for the data that there are solutes. The dataset provided suggests the latter, but I think there would be value in providing the daily discharge timeseries for the same timespan of the solute data.

Line 135: In Table 2, deposition units are reported as mg/L. However, NADP reports their deposition in both concentration and kg per hectare. Are the units in Table 2 a mistake? If not, can you add some detail on the methods used to convert concentration to an area normalized load?

Table 1: Consider added the NWIS parameter code. For example, is "Nitrate, water filtered" the nitrate plus nitrite (00631) or just nitrate (00618)? There are many parameters for slightly similar solutes and it would help with reproducibility if the parameters codes were included. Also, consider listing the difference between pH in the field and pH in the lab for users.

Table 1: Also consider adding more detail to units. For example, is nitrate mg-NO3/L or mg-N/L.

Line 204: EPA link is broken. I have had many issues with direct links where they are archived and become a dead end. I highly encourage the authors to find a paper with a DOI to support this sentence. As a starting point, you can consider:

Baumgardner, R. E., Lavery, T. F., Rogers, C. M., & Isil, S. S. (2002). Estimates of the Atmospheric Deposition of Sulfur and Nitrogen Species: Clean Air Status and Trends Network, 1990−2000. In Environmental Science & Technology (Vol. 36, Issue 12, pp. 2614–2629). https://doi.org/10.1021/es011146g

Lloret, J., & Valiela, I. (2016). Unprecedented decrease in deposition of nitrogen oxides over North America: the relative effects of emission controls and prevailing air-mass trajectories. Biogeochemistry, 129(1-2), 165–180.

Technical corrections ("technical corrections": typing errors, etc.).

Subscripts for solutes should be consistent throughout the manuscript.

Table 3 formatting caused solutes to be cut off.

Line 18 and 149-151: 18 solutes listed in the abstract, 17 listed in Table 1 and in Line 149, and 16 listed in the text. Make them all consistent.

Line 57: Remove (?).

Line 60: Remove “CITE”

Line 168, 174 and others: When referencing figure (ex Figure 2), please add the panel letters (a,b,c, etc.).

Figure 2: Panel a, why does daily average discharge only have 393 watersheds while the original CAMELS dataset uses USGS discharge in the original CAMELS dataset?

Figure 6: Regarding NO3, if arid and humid sites are a subset of all sites, I am unsure how the slope for all sites can be larger than both arid and humid.

Citation

Sprague, L. A., Oelsner, G. P., & Argue, D. M. (2017). Challenges with secondary use of multi-source water-quality data in the United States. Water Research, 110, 252–261.
Citation: https://doi.org/10.5194/hess-2022-81-RC1
- AC1: 'Reply on RC1', Adrian Harpold, 21 Jul 2022
  
  We greatly appreciate the review comments of reviewer 1, particularly about tangible ways we can improve the dataet. We agree with points 1 and 2 raised about improving USGS datasets, but we do not wish to expand the dataset beyond USGS and initiate additional harmonizing efforts. We appreciate you going through the posted dataset and will address inconsistencies. We are unclear about the comment on summar figures for all solutes, as this was done between the main document (Figure 2 and 3) and the supplemental (Figure S1 and S2). We plan to incorporate the suggestions into our final manuscript.
  
  Citation: https://doi.org/10.5194/hess-2022-81-AC1
RC2:
'Comment on hess-2022-81', Anonymous Referee #2, 02 Jun 2022

Overall comment:

This paper describes efforts to amend the known and quite successful US-wide CAMELS dataset with hydrochemical and deposition data. This effort is overdue and would greatly enhance the usage of the original CAMELS data as well. Having stated this I have to admit that the manuscript is not convincing to me. It fails to state what data (e.g., constituent codes in the USGS data) was used in what exact way (missing description of data evaluation, conversion, filtering). It visualizes data coverage but does not state number of observations and number of stations in a consistent way. So overall the reader is left rather unclear about the whole data handling process and the outcome. While I like the idea of three examples what this data could be used for, this leaves me rather puzzled. This is submitted as a research paper but does not really come up with research. For me this manuscript would rather fit the purpose of ESSD (Environmental Science System Data) as a dataset description than HESS as a research paper. For the latter it would have an appropriate structure and depths but still would need to acknowledge the details comments below.

Specific comments:

Abstract

The abstract needs to transport the content and motivation in a better way. It misses quite some information e.g. resolution of provided data. It is not clear what is really provided in this new data set and what is already available from CAMELS.

Line 18: The linkage to the original US CAMELS dataset remains unclear in the abstract. Is this an addition to the original one as induced by title and text here or is this something completely new.

Line 24: This is odd. Ether give the exact number or >. Rounding to something easier to read would probably best.

Introduction

From my point of view the introduction should state some examples of existing water quality databases. There are recent advances here such as: GRQA: Global River Water Quality Archive (Virro et al., ESSD), GLORICH (Hartmann, J., Lauerwald, R., and Moosdorf, N.: A Brief Overview of the GLObal RIver Chemistry Database, GLORICH, Proced. Earth Plan. Sc., 10, 23–27, https://doi.org/10.1016/J.PROEPS.2014.08.005, 2014. a, b) or QUADICA (https://doi.org/10.5194/essd-2022-6)

Line 31: There are quite a number of examples of nation-wide to continental scale water quality studies using more than a single catchment. I disagree here that availability of datasets does not go hand in hand with usage of this.

Line 35: Check citation formatting here.

Line 35: Can you be more specific on the "issues" mentioned here?

Line 42: Year in citation missing.

Line 43: Use the complete name of CAMELS here.

Line 56: Check the question mark here.

Line 60: Check citation here.

Line 83: I think you should better introduce the idea to provide atmospheric deposition data.

Line 86: Can you specify why to stop in 2014? Because the original CAMELS data also have that time frame?

Material and Methods

Line 94: I would expect a citation to the NWIS data sources mentioned here.

Line 97: What is meant with the "two datasets"?

Line 97: "for each observation" seems unnecessary here.

Line 97: The match with the CAMELS station does not explain the promised "geographical coverage". Please adjust this sentence.

An option could be a reference to a figure showing the distribution of stations, maybe including the ones from the CAMELS data that are not amended with chemical data.

Line 102: From my point of view the 2.2 chapter on the methods is not informative and needs some revision. The reader does learn some fancy names but not what exactly was done with the data. The linkage of methods to the provided text-files remains unclear. This does not fit the previously stated wish to make everything reproducible by providing scripts. See also details below.

Line 103: I cannot follow this argumentation.

Line 109: Is ETL an established method? If so, please reference this. If not: Is there really a need for a fancy name for a standard data transformation method needed that uses a scripting language in a reproducible way?

Line 112: I do not see the link from this method to the resulting data provided, which are easy-to-use but simple text-files. This paragraph seems to be overly complicated.

Line 116: Similar as before. Either reference this as established methods or leave it. These abbreviations are not used further in the manuscript - so why introducing them here?

Line 124: Is type of discharge match (paired daily, instantaneous...) indicated in the dataset?

L131: "once per day" at the same time? Is this what is meant here?

Line 135: Again a reference to the dataset would makes sense for me. It would be, moreover, helpful for the readers to know about the absolute number of stations the interpolation is based on.

L 137: I do not understand what "align" means in this context. Please be more specific.

Line 139: Somehow strange to mention Table 2 first in the text while Table 2 comes second.

Table 1: It would be helpful, if not done elsewhere, to state the USGS parameter codes as this often causes confusion. The caption is not fully matching the column names, e.g. is abbreviation=attribute? Also it would be good if this is also consistent with or clearly linked to the column names in the provided data files, which seems to be not the case.

Line 145: The "+" seems unnecessary.

Dataset desription

Line 151: This description says all what is needed. No further need for a table here.

Line 151: Please use subscript in the constituent's abbreviations similar to the table where the constituents are introduced. This applies to the entire manuscript.

Table 3: I do not get the meaning and idea of Table 3.

Line 164: This chapter would profit from a table (maybe combined with table 1) that lists number of stations and number of observations and maybe median number of observations per station for each of the constituents.

Line 171: I was not aware of this varying foci. Is there a reference for that?

Line 175: This sentence needs to be checked - the logic is not clear.

Line 186: While the previous chapter states reasons for different data density and distribution. Here, however, suddenly spatial pattern in the deposition is described. This belongs elsewhere. The chapter should describe the meta-information only.

Trends in...

Line 226: "not" missing?

Line 234: This reads quite strange. You compute cq relationships to check if there are cq relationships? Consider revising.

Line 250: I do not get the meaning of "overlapping solutes" here.

Line 251: Isn't this part of the A (attributes) in the CAMELS dataset?

Line 127: I very much like the analyis of FDC coverage. For me this is, however, not just an addition to the cq analysis but rather part of the dataset description in the former chapters.

Line 269: "response of the relationship"? Consider revising.

Line 270: What are uneven collection dates?

Line 299: This statement seems to be unrelated to stream chemistry and thus not directly a result of CAMELS-Chem.

Line 302: This sentence is very long. What exactly is meant by large-scale response? The observed response is always local, isn't it?

Line 404: This sentence is redundant, please revise.

Code availability:

I think the authors may already give an idea which repository they are aiming at.

Citation: https://doi.org/10.5194/hess-2022-81-RC2
- AC2: 'Reply on RC2', Adrian Harpold, 21 Jul 2022
  
  We greatly appreciate the review comments of reviewer 2, particularly about better ways to frame the study and increase consistency and clarity. We understand your point about HESS versus ESSD, but would contend that the CAMELS database development has a track record in HESS (https://hess.copernicus.org/articles/19/209/2015/ and https://hess.copernicus.org/articles/21/5293/2017/) making it a better fit. Moreover, we demonstrate several new, analyses that illustrate continental-scale patterns and link stream and atmospheric chemistry data, which are more apporparte for HESS. We can address this based on comments form the handling editor. We plan to address the detailed comments of the reviewer in our resubmission.
  
  Citation: https://doi.org/10.5194/hess-2022-81-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

ED: Reconsider after major revisions (further review by editor and referees) (27 Jul 2022) by Genevieve Ali

Dear authors,

Two referees have read your manuscript and provided comments on the Interactive Discussion page. While both referees think that enhancing the CAMELS dataset with hydrochemical and deposition data is greatly needed and a commendable effort on your part, they are both rather critical of your manuscript in its current form. Both referees describe your manuscript as a “data paper”: while data papers are increasingly frequent and highly valuable to the community, expectations regarding their formatting vary depending on the target journal. One referee suggested that your paper be better suited for publication in a journal like Environmental Science System Data because they found it more akin to “a dataset description than […] a research paper”. The other referee rather stated that “the methods need to be expanded” and that their “comments primarily pertain to the data and methods, as [they] see that this is the paper's novelty”. These comments highlight that HESS readers and referees expect so-called “data papers” to be comprehensive in both technical dataset construction details and original research questions: revisions would need to be made to your manuscript in that regard. Both referees also question why CAMELS-Chem only includes data up to 2014, and I echo that question given that water quality data availability is increasing (and therefore considering more recent years would enrich CAMELS-Chem). Data papers are also especially useful when they provide details that allow readers and potential users to assess whether the dataset may be suitable to address research questions of their own: however, in its current form, the manuscript does not provide enough summary figures, tables and statistics, or metadata, to allow such assessments to be made by readers (in my opinion). I believe that a similar point was also raised by at least one of the referees. For the aforementioned reasons and others, one referee recommended rejection, and the other recommended that your manuscript be only reconsidered after major revisions have been made.

I have read your short responses to reviewer comments. I agree that CAMELS-related data papers have been published in HESS before; yours would also be in scope, as long as it is comprehensive in both technical dataset construction details and original research questions (as I mentioned before). Regarding original research questions, in its current form your manuscript *mainly* states that analyses which have been conducted before (e.g., c-Q analyses) could be expanded to larger scales thanks to CAMELS-Chem. I think that your manuscript would be stronger if you could articulate questions or analyses that would NOT be possible to address at all if a product like CAMELS-Chem did not exist (and by that, I mean questions that go beyond applying existing methods to hundreds of gauges rather than dozens of gauges)… While section 4 of your manuscript provides interesting examples of what could be done with CAMELS-Chem, the information (text, tables, figures) included in sections 2 and 3 could be expanded to allow readers (me included) to better assess what *other* research questions we may be able to address with the dataset. For instance, Figure 2 provides information about the length of record (in years) and the number of observations available for each gauge for different constituents, which is great. However, other additional information that would help me assess the suitability of the dataset for some of my research includes: minimum (and maximum, and maybe average) number of observations per year, distribution of observations by season (to assess whether any sampling timing bias exists for some constituents or some gauges), number of observations associated with low, intermediate or high flow conditions, availability of sub-daily observations, length of longest data gap, etc.

I am returning your manuscript for major revisions. I invite you to consider all reviewer comments carefully and provide detailed answers to each of them. Should you decide to submit a revision, your manuscript and response document will be sent out for another round of review.

With best wishes,

Genevieve Ali

Hide

AR by Adrian Harpold on behalf of the Authors (11 Apr 2023) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (24 Apr 2023) by Genevieve Ali

RR by Anonymous Referee #2 (02 May 2023)

Suggestions for revision or reasons for rejection

General Comments
The authors describe an extension of the well-known CAMELS data set with water chemistry and deposition data. I was reviewing a first version of the manuscript already and still think that this is a super-valuable data compilation. However, the manuscript still has, from my point of view, some serious shortcomings. My expectation is that the manuscript shows how data was merged and gives an overview on the dataset. The method is, to my surprise, not reproducible and remains very vague. The results are mostly displayed in a visual way - I very much miss a quantitative overview on the basic descriptive statistic, on the sample frequency per year. I also miss a brief overview on what information are delivered with the original CAMELS data set. So, it is hard for the reader to say if the dataset actually fits his or her needs. Finally, I have problems with the embedded two examples. Motivation and methods are given in chapter 4 and are not aligned with the introduction and methods of the entire manuscript. Results of that examples are not much more than two plots and the statement that others may interpret that. So, overall I think this manuscript needs some serious revision.

Specific Comments
Abstract
L20: Potentially remove the first "and" here.
L21ff: I cannot fully follow which constituent is given with full name and which is not - consider to homogenize that. Nitrate is given first as a full name, then as the short chemical form and then again as full name.
Introduction
L45: GLORICH (among other water quality databases) is part of GRQA. Not sure how to address this here but I encourage you to make this more clear in this sentence.
L50-55: You mix water quantity data efforts such as CAMELS with water quality AND water quantity data efforts such as Ebeling et al. here. I think this need better separation and clarification. Why should an atmospheric deposition dataset be relevant for water quantity? Note that atmospheric deposition of nitrogen is part of Ebeling et al.
L75: There is something wrong here. Please check the whole sentence.
L85: Be precise - do you mean stream "flow" data?
L87: This is a bit unclear. Why 2018, when CAMELS ends in 2014?
L90ff: As already stated in the first review round I am not convinced to publish this as a research paper if the manuscript does not attempt to set itself a scientific goal. I think the two examples needs a motivation within the introduction section already and should be mentioned explicitly. The examples are not totally randomly selected as written later in chapter 4!

Materials and Methods
L105: That is a nice statement but I wonder why it is made here? What is the consequence?
L116: "Water resources data" is not entirely clear for me.
L96-114: This chapter leaves me a bit puzzled. This is not about "data sources and description". Would there be a more fitting header? Where are actual multiple data sources described?
L126f: I thought discharge data is the key of CAMELS (see also line 47)? Why was discharge not available in all 671 catchments anymore while it was available when CAMELS was published?
L128: How many catchments had data before and after 1980-2018? Or is this coverage part of the result section?
L121ff: The methods should describe you workflow in a way that it is reproducible for the readers. This is not the case. Also matching fig. S1 is not of big help here. I think you need to state what exactly you did to the data. Fig. S1 lists aggregation, filtering, cleansing, profiling, joining and sorting. At least the first three processes are not described in the text.

Results
L141: I would not call discharge a general water quality parameter.
L143: Also give full names of Cl and the cations.
Table 1: You state for some constituents "Water, filtered and total" - is that visible in the data base? Why is magnesium measured in the suspended sediment? This makes it hard to be analysed similar to all other cations. Why is not unit and database abbreviation given for discharge? Why is alkalinity not stated here?
L146: Mention the fig. S5 here specifically. It takes a while finding the right figure when just referring to "supplementary materials"
L148f: Something went wrong here. 325,477 catchments?
L155f: Interesting that you start with the spatial distribution. For sure very important. However, does it make sense to start with descriptive statistics on the data? A table such as table 1 but station median, percentiles, number of observations...? Especially as all catchments aim at pristine areas and do not incorporate too much direct human impacts.
Fig. 1: Given the different catchment size, I recommend to give discharge per area (e.g., mm/a) in 1a. Otherwise this is not very informative. I struggle with the colors vs. size. More intuitive for me would be the value represented by colors and the CV by symbol size. However, this is up to you. For pH it does not work well - but here you can maybe scale the symbol size between min and max similar to the other plots.
Fig. 2: Why is this coming after fig. 1? Seem to be a better fit for the methods or an earlier point in the results.
L158: These are maybe not the best examples as DO is having strong physical constraints and pH is a log-unit.
L183ff: From point of view of the USGS these are impressing numbers. However, the reader would be more interested in the average number of samples per station per year.
L188: Can you give some of the published coverage of different hydroclimates from the original CAMELS references? What range do you really cover here?
L225: Further above you stated that you had a large number of chemistry samples what do not have a matching instantaneous discharge observation. How do these numbers (here >90%) fit together?
L240-245: This section mostly contains data source, interpolation methods, GIS-matching. All this belongs to the method section and not to the results.

Example analyses
L275ff: All this needs a proper method description in the methodology chapter. I strongly miss a proper description of the actual results (only a weak verbal description and a link to Fig. 7) and a scientific interpretation of the results. Just stating that more sophisticated methods may reveal underlying controls is not enough here.
L278: Maybe it slipped through my read but what is FDC?
L303ff: All what I stated for the first example is also valid for the second one.

Hide

RR by Anonymous Referee #3 (30 May 2023)

Suggestions for revision or reasons for rejection

To Authors:

The CAMELS-Chem dataset will be a useful tool for catchment research, and I look forward to seeing this paper published. What appears to be the most novel and intensive part of this work is the addition of atmospheric deposition inputs to the CAMELS catchments. Mirroring comments from previous reviewers, I think that presenting this work as a data paper is the most appropriate path. As such, I also still have some comments about increasing transparency of the methods.

As someone familiar with accessing large amounts of USGS flow and water quality data from the NWIS, I think that the paper should make it more clear how the dataset they present is not just a compilation of flow and water quality for the same group of catchments. Be very explicit and detailed with what was done for data cleaning, interpolation, gap filling, etc. I find the language about ETL in Section 2.2 confusing, mainly in that the end product that I saw (three CSV files on Hydroshare) seems fairly straightforward, whereas the described end product (a relational database using PostgreSQL) sounds very fancy. I fully acknowledge my ignorance of the ETL methodology, but as someone who would be very excited to use CAMELS-Chem, I’d like this section to paint a picture of what exactly is going into the production of your dataset/database, and what I would be downloading to use it.

Regarding the deposition data, that section is also scant on details. What is the resolution of the input data, and what resolution are you reporting the data? The data was also not available for me to review in your Hydroshare link, so I was unable to comment on its formatting.

Again, from my perspective, doing a quick merge of USGS flow and water quality data would be quite easy for someone using the R package “dataRetrieval”, and so I think you need to make clear that your dataset is value-added: mainly from the addition of deposition data, but also through your data processing/cleaning procedures and the construction of the relational database. Make sure to clearly sell CAMELS-Chem to the audience! I look forward to seeing this work progress.

Comments:
1. Line 18. Maybe in the abstract say how many sites are in CAMELS-Chem? See Comment Line 97.
2. Line 24. “Annual deposition loads and concentrations” are promised, but they are not included in the included database, nor are they mentioned in Methods section. See comment from line 239.
3. Line 75. “Thus, we now have There are opportunities”. This text should be fixed.
4. Line 97. Cite the number of sites from CAMELS, and then CAMELS-Chem (516 sites cited at Line 84)
5. Line 101. “most of its sites are drawn come from” USGS HBN. Remove the word “come”. Also, how many sites from HBN?
6. Line 103. How many sites from NWIS or HCDN?
7. Line 116. You list the challenges: missing data, mis-matched sample times, inconsistent parameter names, or varying units of measure. Of these, later in the paragraph I only see sample times and units of measure discussed. Did you fix parameter names? Or fill missing data? Please elaborate.
8. Line 127. How do the 506 and 488 become 516 sites? More detail please.
9. Line 130. “impute”. Could you use a more colloquial form for this sentence to improve readability?
10. Line 131. Forgive my ignorance, but does the Hydroshare upload represent your final data repository, using PostgreSQL? Having looked at your CSV files, which I found well organized (see some minor comments later about the dataset), I’m not sure if there’s something I’m missing.
11. Line 148. Could you include the number of catchments and number of measurements for all the parameters, as you’ve done for Si and DOC, into Table 1?
12. Table 1. Units and abbreviation and USGS code are missing for discharge
13. Table 1. In a response to a previous reviewer comment about merging solutes, you justify using unfiltered and filtered parameter codes (e.g. chloride, sulfate, nitrate). This justification needs to be in the manuscript, explaining why there are multiple parameter codes in the last column of Table 1. See also my comment on the database, about my confusion about the column “total_no3” and others.
14. Line 225. This sentence needs to be corrected (referring to right-most column or bottom-most row). This does not reflect the information in the table. Should it be second column, and then second row? I might advise moving Q to first row/column before Temp.
15. Table 3. As mentioned elsewhere, leading zeros should be present for USGS parameter codes.
16. Line 239. Section 3.5. Atmospheric deposition data. This paragraph should be included under Section 2 Methods. In describing the NADP data, what is the data resolution (temporal, spatial)?
17. Table 4. Table 4 should probably be in Methods, rather than Results?
18. Figure 1: The inset histograms referenced in the figure caption (and shown in manuscript version 1) are no longer present. They should be added to the figure.
19. Comments on the database. Mirroring previous reviewer comments, the data should be able to stand on their own, and so the following comments are requesting context and explanation to improve readability.
19.1. The paper discusses atmospheric deposition data (measured concentrations and annual flux estimates), but those data do not appear anywhere in the dataset.
19.2. In the “metrics” file, the following columns have no explanation: q_inst, q_15, q_derived, q_derived_note, q_daily_note, q_daily_cd, q_inst_cd, q_15_cd, measure_unit_code, sample_start_dt, sample_start_time, sample_timestamp, q_inst_ts, q_15_ts, inserted_ts, updated_ts, gauge_id2
19.3. In the “dataset” file, there is a second column “gauge_id2”. This column is not described in the “metrics” file, or how it is different from “gauge_id”
19.4. Leading zeros. The gauge IDs should have the proper number of leading zeros. The USGS parameters should also have leading zeros. This applies to all CSV files in your database. This is critical for users not familiar with USGS standards, and might be looking for “940” instead of “00940” (for example).
19.5. The columns total_cl, total_no3, and total_so4 are confusing. My guess is that these columns are filtered and unfiltered values, harmonized? These columns are not mentioned in the manuscript Table 1, nor are the methods for calculating the column values in the manuscript.

Hide

ED: Reconsider after major revisions (further review by editor and referees) (05 Jun 2023) by Genevieve Ali

Dear authors,

Two reviewers (including one from the previous round) have evaluated your revised manuscript and provided detailed comments. While they both agree that significant improvements have been made compared to the original version, they still find the manuscript has some shortcomings. Both reviewers noted a lack of details in the Methods and Results sections (mostly sections 2 and 3), which led them to comment (negatively) in different ways: one found that the lack of details made your work not reproducible, while the other felt that your dataset, as described, may appear as being no more than a “merge of USGS flow and water quality data”, which is something that can be done using an existing R package (and hence is not really novel or value-added). Obviously, the CAMELS-Chem dataset is (or has the potential to be) *much more than a simple data-merging exercise*, but the manuscript in its current state is underselling that potential (I do share the reviewers’ opinions on that). One of the reviewers sees your paper as mostly a data paper while the other thinks that submitting your paper to HESS (rather than Earth System Science Data (ESSD), for instance) means that the goal of the research should go beyond just the presentation of the dataset. Personally, I think that your paper would be most impactful as a hybrid one between those two formats, but this makes it tougher to write because you need to articulate not only technical details pertaining to the creation of the dataset, but also hydrological and earth system process details. Both reviewers have included numerous comments for your consideration, so I invite you to carefully look at them and see how you could address them. I do look forward to receiving and reading your revised manuscript and response document, which will be sent out for another round of review.

With best wishes,

Genevieve Ali

Hide

AR by Adrian Harpold on behalf of the Authors (24 Jul 2023) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (15 Aug 2023) by Genevieve Ali

RR by Anonymous Referee #3 (04 Oct 2023)

RR by Anonymous Referee #4 (13 Oct 2023)

Suggestions for revision or reasons for rejection

Line 51: Missing a closing parenthesis “)” following “… Addor et al., (2017)”.
125: Strike “National” from “USGS National NWIS”; it’s already represented in the acronym.
136: Change “obtained NADP” to “obtained from NADP”.
229: It appears that the lowercase references to Figure 3 sub-panels are offset by one; for example SO4 and Ca are subpanels “l” and “m” in Figure 3, not “m” and “n” as represented in the document body. This offset needs to be corrected for this and all following references to Figure 3.
238: Replace “…temperature is measurement relatively…” with “…temperature is generally measured…”
252: Cumbersome and repetitive sentence beginning with “The range…”. I suggest replacing the entire sentence with “The range of hydrological and meteorological conditions represented is nearly identical between CAMELS and CAMELS-Chem catchments.”
268: replace “…covered of the percent of the FDC during…” with “…percent of the FDC covered…”
276 and 278: Table 4, not 5.
283: Table 5 not 4.
297: Suggest replacing this sentence with “In many cases these patterns are consistent with patterns in stream chemistry; for example, patterns of NO3 deposition (Figure 7c) compare closely with the corresponding pattern in chemistry (Figure 1j).”
337: You appear to be missing a citation in the sentence “In another example use of CAMELS-Chem, … used DIC stream chemistry…”
369: Replace “…to develop new hypothesis…” with “…to develop a new hypothesis…” if only one hypothesis, or “…to develop new hypotheses…” if multiple.
389: Replace “coincident” with “coincidence”.
570 and beyond: several of your references are listed twice.

Hide

ED: Publish subject to minor revisions (review by editor) (25 Oct 2023) by Genevieve Ali

AR by Adrian Harpold on behalf of the Authors (13 Nov 2023) Author's response Author's tracked changes Manuscript

ED: Publish as is (17 Nov 2023) by Genevieve Ali

AR by Adrian Harpold on behalf of the Authors (28 Nov 2023) Author's response Manuscript

Short summary

We develop stream water chemistry to pair with the existing CAMELS (Catchment Attributes and Meteorology for Large-sample Studies) dataset. The newly developed dataset, termed CAMELS-Chem, includes common stream water chemistry constituents and wet deposition chemistry in 516 catchments. Examples show the value of CAMELS-Chem to trend and spatial analyses, as well as its limitations in sampling length and consistency.