This paper details the Water Survey of Canada’s standard operating procedures in estimating discharge values from stage values. The paper addresses an important issue that is often not documented and has critical impacts on uncertainties in discharge time series. Generally the paper is well written and the figures are well presented, with lots of interesting examples of different types of rating curves. However, the paper is long with a lot of figures and as a result, the key message of the paper gets lost. I recommend shortening the paper (moving more material to supplementary information) and better clarifying the key aims and messages of the paper in the introduction, conclusions and abstract.

Abstract – I don’t think the abstract is a clear summary of the work that has been conducted and the key messages of the paper. I would recommend revising it to better synthesise the outcomes from the paper.

L51-52. ‘River discharge or streamflow has significant importance for planning, impact and sustainability assessment’ – this is very generic and could apply to planning, impact and sustainability assessment of anything! This needs to be more specific to water resources.

Aims L99-104 – I find the aims of “the study” quite confusing as it is not clear whether “the study” relates directly to this paper or to a wider project? Please revise this section and more clearly state what your core aims and objectives of this paper are.

L152-154. What is “discharge activity”? The estimated discharge may then be used to correct what? These sentences are not clear.

Table 1 and 2. I think you can place these in supplementary information. Many of these terms are described in the text already.

Figures 3-5. These are very nice but could you combine these into one figure?

L316-317. It would be good to add a sentence here on why you are developing an independent Python workflow.

L371-374. This sentence isn’t clear and needs re-writing.

L405. The Environment Agency for England does not use this method. They use this method: https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/290629/sw6-058-tr-e-e.pdf

L411-418. The text on observed stage-discharge records is out of place here. It could be removed.

Figure 6 and 7 – you could move some of these examples to the supplementary information and combine these different examples of rating curves?

Figure 10. I like this figure a lot and really interesting to see the regional differences.

L488-494. The description of the figure can be moved into the figure caption.

L496. “significantly lower” – can you quantify this? How much lower?

Discussion and Conclusions – I would recommend splitting these and having a separate conclusions section where you turn your bullet points in L735-761 into a conclusions section.

Data availability. I appreciate that the streamflow data would need to be requested from the WSC but are there any other outputs from your extensive analysis that could be made available to users?  For example, could you release the fraction of the discharge within 5% of reported discharge values for each station, or the number of days with a temporary shift for each station, or the fraction of time higher than the maximum observed stage? These outputs could be valuable for researchers conducting large-sample studies in Canada and could be used as a (admittedly crude) way of filtering out stations with more/less robust data.

This article presents a systematic study of the hydrometric data production process across 1800 active stations operated by Water Survey Canada. An independent (Python-based) approach intended to reconstruct the archived discharge times series based on information and data available from the Aquarius operational software. Interestingly, only 67% of the data could be reproduced (within 5%) from the stage series, the rating curves and the rating shift curves, the other differences being explained by a significant use of temporary shifts and “overrides”. This exercise is valuable as it quantifies the frequency of operational practices that are more complex to reproduce than the simple application of rating curves (and permanent shifts). In particular, the need for suitable uncertainty computation methods is rightly emphasized.

The paper is generally very well written and well illustrated, however I fear that its length may discourage some readers less passionate about hydrometry (including data users!) and reduce its impact. I would recommend shortening the paper (20 pages max and 10 figures max). Some technical details (eg multiple data examples) could be cut or moved to Annexes or Supplementary materials.

Technical comments

• 69-70 the method (IVE) introduced by Cohn et al. 2013 does not relate to rating curves. Not sure about Whalley et al. 2001 and Huang 2018. Please check and remove if need be.
• 405 I’m not sure the method presented by Coxon et al. 2015 is actually applied systematically by UK Env agency to establish their rating curves. I don’t think so. Kiang et al. 2018 compared 7 methods for rating curve uncertainty and only the NVE method (in Norway) and the Baratin method (in France) were applied by national hydrological services.
• What about (seasonal) aquatic vegetation? Is it a problem for Canadian stations (as it is in Europe for instance) and is it managed through temporary shifts? I assume that beaver dams are another issue…
• As stated end of 2.5 and elsewhere (Tab. 3), the central issue is the traceability, reproducibility of the data production process. However, reproducibility and repeatability are different things, and this could be made clearer in the paper. A first step is that discharge computation can be repeated (exactly) using available data and already established rating curves, shifts and overrides (from Aquarius especially): this doesn’t seem to be the case as some important information is missing (or not easily retrieved through API), which would be a first issue of incomplete traceability (am I correct?). Another step is that discharge computation can be reproduced (with some permissible variability) from scratch by another equivalent expert: this should be OK thanks to established SOPs and well-trained operators, hopefully, but this statement is not evaluated in this study (through some comparison exercise, for instance). Actually, the problem seems to arise because the assumptions and decisions made by the hydrographers for establishing rating curves, rating shifts, temporary shifts and overrides are not available in a formal way. I think that beyond the statistical technique chosen for uncertainty estimation, this is the key issue: each data production process must be ‘modelled’ in a reproducible way, even expert-based operations. I agree that some solutions have been published that apply to rating curves and (partially) shifts but not to temporary shifts and overrides. This paper is a first step towards modelling these operations but much more work looks necessary to write mathematical models, especially for ‘override’ operations, which refer to multiple situations and data estimation techniques. The discussion and comments in the paper could elaborate on this issue more clearly.
• Another obstacle stated in the paper is the deterministic approach: the uncertainty of stage-discharge measurements must be accounted for, as well as the uncertainty of the input data (stage) and of the rating curves (and more generally the “discharge models”). It looks difficult to quantify the uncertainty of data that have been produced in a fully deterministic approach without reprocessing them. The ideal way to go is to reproduce the data in a probabilistic framework, hence the need for reproducibility…

Editorial comments

• 146 Aquatic?
• 176 include?
• 230 curve
• 558 To investigate what?
• 738 Pytho