Review of Kern et al. ‘Assessing the Impact of Freshwater Fluxes from Major Rivers on the

Atlantic Ocean’

The manuscript describes assessing the water mass balance (WMB) method for estimating river discharge into the Atlantic basin, in which change terrestrial water storage calculated from satellite. River discharge from variations on this method are compared with in situ observations. The impact of the different estimates of river discharge on ocean model simulations of salinity and currents are further investigated. The manuscript is well-written, and both the potential and the limitations of the method are reasonably communicated. The manuscript should be a useful contribution to the literature. There are a couple of areas where the content could be expanded. Firstly, the authors remark how the method is sensitive to the reanalyses used -  it would good to get an idea of the spread possible with a range of reanalyses.  I don’t suggest repeating the analysis multiple times, but it might be possible to get a picture of the spread by calculating one diagnostic (say net evaporation over the amazon) for different reanalyses. Secondly, it would be good for the authors to unpack the confusing finding that WMB1 is better than WMB2 at  improving the representation of river discharge but worse at improving the representation of ocean salinity.

Further specific and technical comments.

Title: Suggest a title more like ‘Quantifying freshwater fluxes from Major rivers and assessing their impact on the Atlantic Ocean’

Abstract

Line 10. Suggest ‘riverine freshwater fluxes’.

Methods

Line 84. A little justification/ quantification required here. Which rivers were excluded? What was the discharge of the largest of these? What was the total annual discharge excluded (say as a percentage of the Amazon)?

Seven rivers without in-situ observations. What uncertainties does this bring to the result. Is there enough data to estimate variability of in those rivers?

Line 167-170. Consistency of fluxes. Are the ERA5 fluxes self-consistent? Specifically, globally does precipitation=evaporation on a reasonable time-scale? (I enquire because we know that most reanalyses are not close to producing a globally balanced net ocean heat flux).

Tsujino et al (2018) use a catchment basin river routing model driven by JRA reanalysis. Should n’t the fact the fact that this is driven by reanalysis mean the approach is not too dissimilar to the authors and that there might not be much quantitative difference?

Line 188. ‘Several experiments’. My understanding is the authors conducted 6 experiments (4 experiments and 2 ‘control’ simulations). Could the authors list these in a table, giving the control simulations an abbreviation?

Lines 200-202.  I find this unclear. Firstly NEMO is an ocean model, so it does not have river basins as such, merely river outlets. It makes sense to regrid or reassign the outlets on to the NEMO grid, but I don’t understand regridding the basin. Secondly, I cannot understand the phrase  ‘the local spread of runoff with the coastlines of pertinence of each river drainage basin, proportional in area to the monthly climatology’. Could the authors, break this down into smaller more easily comprehensible chunks?

Line 213. The correlation is high because of the strong seasonal cycle. The authors should correlate the time series with the seasonal cycle removed.

Lines 217 onwards.  Do you the authors understand the reason for the weaker agreement in the annual cycle they describe in WB2?

Lines 230- 234. Suggest this paragraph is replaced by something like. ‘Linear trends of the lines in Fig 3a were calculated and found to be statistically insignificant.’

Line 254. ‘and highlight’ suggest ‘but highlight’.

Line 296. ‘overestimated’. Just say it is larger in WMB2 by this amount. It cannot be described as an overestimate as WMB1 is not ground truth.

Line 330. ‘overestimated’ suggest ‘overestimated relative to the in-situ observations.’

Table 2/ 3 could the authors clarify the difference between Table 2 and Table 3.

Line 345.  ‘Fig. 6’ should be ‘Fig. 5’, I think.

Line 368. ‘Figure 7’ should be ‘Fig. 6’ ,

If I understand correctly Fig. 6a shows the MAE of A1 relative to EN4 minus the MAE of ref relative to EN4. I think this could be stated with greater clarity in the caption

Line 373.  I suggest the authors add ‘not shown’. i.e. ‘The REF experiment…West African coast (no shown). Alternatively, the authors could show The MAE of REF relative to EN4 in another panel.

Line 376. ‘Key Regions’. Because the MAE of Ref relative to EN4 is not shown we don’t know what the key regions are.

Line 384. Suggest ‘into the ocean model’

Please comment on the larger impact of the WMB method around the Gulf of St. Lawrence and Congo.

Also on Table 3 it would be of great interest to add the Dai and Trenberth Values as 2 columns to Table 3 if possible and comment on them. This is highly relevant since the authors compare WB forced ocean simulations with those forced by Dai and Trenberth.

Figure 7 and elsewhere. The authors do not include any caveats about the EN4 dataset. The data set is far from perfect and the key input from Argo floats are absent in the shelf regions around the coasts.

Line 422.  Fig. 8 ERA5 WMB1 is better than WMB2 at improving river discharge but worse (or no differerent) at simulating ocean salinity. How can this be explained?

In this paper, the authors produce 20-year long time series of runoff for the main rivers of the Atlantic Ocean (including the Amazon and Congo, world’s 1st and 2nd largest) using a water mass balance approach, combining freshwater fluxes products over river catchment areas. They validate their products with available in situ time series. Then they do test simulations with an ocean general circulation model forced by different kinds of runoff (daily or monthly, climatological or varying interannual) including different versions of their own runoff products, to evaluate the impact of these runoffs on the Atlantic Ocean, in terms of salinity (which is the benchmark to estimate the runoff products), stratification and currents.

The paper showcases new runoff products that are freely available and will be useful to the community, as in situ discharge time series have been discontinued or are simply not available for a lot of rivers, The tests clearly show that using these interannual runoff, rather than another interannual runoff product (JRA55) or climatological runoff, results in better oceanic simulations.

Therefore, I think the paper is worth publishing after the relatively minor comments/revisions below are adressed :

L23 After reading the paper, I would rather suggest « associated weakening in vertical stratification, changes in upper ocean circulation and basin scale meridional transports » or « associated changes in vertical stratification, upper ocean circulation and basin scale meridional transports »

L28 Freshwater inputs (« these » would refer to freshwater fluxes that also include evaporation)

L36 land-to-ocean water fluxes (exchanges would be two-way)

L67 « such as EN4 » : this suggests that EN4 is one of several SSS products used but I later found that it is the only one. It would have been to use another for comparison, like the satellite CCI product available from 2010 onward.

L113 Then why the HYBAM database, which main focus is obviously the Amazon, is not mentioned as a source of in situ measurements for this river in table 1 ?

L147-148 GIA, DDK : could you define these acronyms ?

L176 « a horizontal resolution of 1° with enhanced resolution in the tropical regions » : Please precise, if 1° is the lower resolution at high latitudes, what is the higher resolution in the tropics ?

L188-198 If I understand correctly, 6 NEMO simulations are used. A table summarising the main differences between them would be welcome, especially because the name of the sensitivity experiments are not very explicit.

L195 & L198 I understand from L167-168 that simulations A1 and S1 use WMB2 while A2 and S2 use WMB1 for river discharges. This is very misleading and I strongly recommend to switch the names.

L200-202 Very unclear. How could river basins that cover the continent be remapped on the NEMO grid that cover the ocean ? I had to check the reference to understand that this means the runoff enters the ocean at the surface as rain. Please rephrase and include this key point.

L212 Again, why is the HYBAM database, which main focus is obviously the Amazon, is not used besides GRDC as another in situ reference ?

L237-L249 I think this part where you define metrics should belong to a sub-section in the data and methods section.

L368 rather Figure 6, also I think it would be good to define the MAE in the data and methods section with other metrics.

L369 The EN4 observational dataset should be presented in the data and methods section.

L373 Rather than only giving a vague description of it, It would be better to show the MAE map for the REF simulation in Figure 6, where there is space for an additional plot. This would help to visually estimate the relative improvement with the new river discharge products. This should even be the first plot, and more generally this would be easier to follow if plots are in the same order on the figure as they are discussed in the text.

L380 The sentence should be splitted in 2 parts where A1/A2 results and S1/S2 results are discussed separately. Also, does the MAE reduction values given in the text correspond to the spatial average of the MAE differences on the Atlantic basin ? It is not explicitly said so the reader can only guess. To be fair, it should also be noted that the use of WMB discharges locally increases MAE in A1/A2 in some regions, particularly along the US east coast and in some parts of the Gulf of Guinea.

L387 « it appears that both systematically increase sea surface salinity, obviously due to a decrease the North Atlantic-wide freshwater input »

L398 I guess the RMSE for the simulations is here relative to EN4, but I think it should be explicitly said. Also EN4 being based on hydrographic profiles that are relatively rare over the continental shelves, this product probably does not capture very well the freshening and overestimates SSS close to the coast. At least it is the case for the ISAS product (also based on profiles), compared to satellite SSS (Houndegnonto et al., 2021). Thermosalinograph transects are the reference observations for coastal-to-offshore salinity gradients (Boutin et al., 2018 ;Dossa et al., 2021) but do not provide homogeneous coverage. It would be very relevant to use satellite SSS here.

L403 « the skill score rapidly worsens towards the open ocean » for J5 : I do not see that in the middle panel of figure 7, rather it does not improve compared to REF. Please correct.

Figure 7 It seems there are more colors in the legend than colored lines on the plots, please correct.

L411-414 This part is quite obscure to me. I am not sure to understand what is the bias vs random error. Is the bias associated with the representation of the mean SSS, and the random error with the SSS variability ? But then why would the variability be considered as random ? Please clarify. Also I do not think that there is an experiment called ALL ? L431 Aroucha et al (2025) actually found that changes in stratification do not affect SST in the Congo plume, but mention that it has been showed elsewhere notably in the Niger plume (Topé et al., 2023). Please correct.

L436-451 While there are obviously changes in the currents, the discussion here is not very convincing to me and should be reorganised a bit. As you mention other river mouths than the Amazon, can you rely on the A1-REF difference in Figure 6 to highlight some general patterns, which are not obvious to me ? If not, I would suggest to look at the S1-REF differences in current to concentrate more on the Amazon region. I think the North Brazil Current/retroflection/NECC current system (no need to mention the Atlantic Cold Tongue that is not at the same latitude as the NECC) should be quickly introduced with the help of the left plot in figure 9 (maybe with schematics ?) before discussing the influence of discharge on this system, and before mentioning the Atlantic meridional transport which is more general. Some references should be added, for example to support the claim « in general less freshwater-driven stratification allows for larger net meridional exchanges ». Also the experiment A1 in the figure is called ALL in the text, which should be corrected.

Figure 10 I would suggest to use the same correspondence between colors and experiments as in figures 7 & 8.

L496 basin-scale meridional transports L510 As future satellite missions are mentioned, although it differs from the WMB approach, I think that the SWOT potential to estimate river discharges should also be mentioned somewhere.