Review of Wei et al. BG 2021-164

It is always very interesting to test the knowledge we have about natural systems by translating them into equations in mathematical models. This allows first to test the validity of the equated processes and second to build scenarios according to environmental changes. This manuscript presents results of nutrient retention in estuaries of the French Atlantic coast obtained with a mathematical model entitled C-GEM.

Macrotidal estuaries are complex transitional environments. Modeling nutrients in these systems is a challenge, because the biogeochemical processes are numerous and complex and the physics of estuaries alone is the subject of complex 3D models. Here the choice was made to build a simplified one-dimensional physical model coupled with biogeochemical reactions involving dissolved N, P and Si compounds, as well as suspended particles. This model is applied to 7 estuaries of different sizes and the model outputs are compared to the available data for the tested period. The model could be criticized for being too simple (1-D), but the biogeochemical part is relatively complete and in the end the results are promising. This 1-D approach is a sufficiently precise step to obtain interesting results. I have a few remarks that concern the validation of the model, the tidal cycles, and points of detail described below.

Line 40: It is conventional to write that estuaries, and more generally water bodies, are facing increasing anthropogenic impacts. For European estuaries, it would have been fair to write this in the late 1990s. Since then, there have been major efforts to improve water quality. Nitrate and phosphate levels are much lower today than 30 years ago. It would be more accurate to write something like "despite efforts to improve natural water quality since the 2000s, estuaries remain receptacles for nutrients and contaminants..."

Table 1. I am a little confused to see that the Dordogne watershed is included in the third row of the table and not in the second. The same goes for the Nive, I suppose, in the Adour column. In general, this 3rd row (Estuary basin area) should be better explained or its name should be changed

Section 2.3.1: In this manuscript, the model is a black box and it is necessary to consult the articles of Volta et al. to know the details of the modeled processes. For example, I tried to see how the interactions between SPM and phosphates were accounted for, but I could not find this information. Wouldn't it be appropriate to put the model elements in an appendix?

Line 177. The time resolution of the model is hourly and daily depending the parameter. However, the input data to the model does not have this resolution. Here, it is stated that the data was linearly interpolated to obtain the temporal resolution of the model. Could you give some indication of the frequency of acquisition of the measurements of the Naiades data set?

Table 3 : concentrations unit in µmol/L, and in Table 4, units in mg/L

Table 4 : the value of SPM concentration of the Dordogne is very high : it is not representative of the Dordogne river. It is most likely a value from a station located in the tidal estuarine part of the Dordogne river. The Dordogne is a river with many upstream dams: SPM concentrations are low until the tidally influenced zone is reached. It would be interesting for readers familiar with these estuarine systems to give the names of the stations used to define the river mixing end-members.

page 12 and Fig 3. The model results are compared to values measured along the estuaries at two contrasting dates, one during a period of high tributary discharge and the other during a period of low discharge. However, one date, for example January 15, 2015, is a 24-hour period when there were 2 tidal cycles. For these macrotidal estuaries, it is likely that the timing of the tide plays an major role in the spatial distribution of the different compounds described here. However, I did not see when this tidal cycle effect was discussed. Do the model outputs correspond to a time of high tide? low tide? For an estuary the size of the Gironde, a given time corresponds to different moments in the tidal cycle upstream and downstream. This would need to be discussed.

Line 295 “…donwstream of the confluence with the Dordogne tributary". This sentence implies that the Dordogne is a source of material for the Gironde estuary at the confluence. This ignores the reality of the environment. The Dordogne and the Garonne meet in the estuarine zone. At the confluence, the waters of the two rivers are very efficiently mixed by the tide: mixed waters of the Garonne and the Dordogne rise largely upstream of the confluence, up to the dynamic tidal limit, located on both rivers more than 70 km from the confluence (e.g. Parra et al., Continent Shelf Res 19, 135-150, 1999)

For some parameters shown in Figure 3 (and Fig S1), only the data from the river mixing end-member and the data from the marine mixing end-member are shown. There is no data in the estuarine part: we can therefore not talk about model validation here (e.g. the bottom graphs for the Loire). For the Adour, there is only one control point. Is it a point taken at high tide or low tide?

Generally speaking, most of the control points are located either at the level of the marine end-member or in the zone where the salinity is close to zero. There are not many control points in the salinity gradient area. This makes it difficult to claim that the model calibration is robust.

line 344: data expressed in % saturation would allow a more direct visualisation of O2 consumption

Figure 4. Simulated and measurement of salinity are important information that is missing here. Indeed, it is stated in line 316 that the stations were chosen in such a way that they are not too influenced by the marine boundary. But these stations should not be strongly influenced by the river boundary as well. Indeed, according to figure 3, all stations in figure 4 are in the area where the salinity is close to 0.

Here again the question of the tide comes up: for the selected stations what is the variation of the concentrations during a tidal cycle in summer, in winter, during spring tides or neap tides? This aspect is not discussed here. Measurements on tidal cycles probably do not exist, but can the model simulate them? If not, this role of the tide should still be discussed.

section 3.2.1: in this paragraph I found it difficult to know whether reference was made to flows from estuaries or river flows into estuaries.

In line 361 and in table 8, the flow results are given with 2 decimal places. Is this level of precision justified?

Line 373: “They accounted for about 80% of the total water discharge from all the estuaries on the French Atlantic coast and 83% of the total watershed areas on the French Atlantic coast, and hence a similar proportion of the nutrient fluxes.”: The missing 20% are represented by small rivers with small estuaries. As a result, this 20% certainly has a lower nutrient retention rate than the average of the estuaries studied here. Thus, the contribution to the nutrient flow to the Atlantic coast of this 20% is probably higher than 20%.

Are the nutrient import data in Table 8 and Figure 7 direct outputs from C-GEM? It is not clear to me

Table 7; TP values do not take into account the P associated with inorganic particles. However, some of this P is desorbed in the salinity gradient, so that an estuary can become a source of DIP (e.g. Deborde et al. 2007 L&O 52, 862-872). Is this reaction taken into account in the model?

Fig. 7: This figure shows average calculations of retention rates and residence times at the annual scale. For me, an annual average is meaningless and does not explain the relationship between the two parameters. It would be more interesting to compare these two properties in flood and low water periods. The relationship should certainly be better and the processes that go with it should be easier to explain.

Line 472 : simulations do not resolve the tidal cycles.

line 481 to 486: I have the impression that we are going in circles in this paragraph.

line 503: perhaps it should be recalled here that the link between water residence time and nutrient retention is a known phenomenon for lakes, wetlands or dams and that it is this principle that leads to the restoration or construction of wetlands

The first paragraph of section 4.3 should be included in the introduction

I hope that my comments and considerations will make it possible to better highlight the quality of the results of this modelling, which has interested me greatly

Review of

Nutrient transport and transformation in macrotidal estuaries of the French Atlantic coast: a modelling approach using C-GEM

By Xi Wei et al.

In this manuscript Wei et al apply an existing model to 7 macrotidal estuaries along the French coast for the years 2014 – 2016 to simulate impacts of estuarine characteristics on  riverine nutrient fluxes to the coastal Atlantic. The model used has been tested widely across different systems and modeled parameters compare well with observed values. The paper is well written, and the main conclusions that large estuaries have higher retention rates presumably due to higher residence times is clearly supported by the results.

In my view, two aspects need attention: 1) Riverine particulate (and organically bound) nutrient input and 2) impact of import particulate (organic )matter from the coastal Atlantic.

L85ff: Estuarine circulation is driving the accumulation of marine particulate matter and instrumental in the formation of the Turbidity Maximum (Burchard et al., 2018). How is this solved in a 1-D model? Import of marine organic matter can be an important source of nutrients in the estuary and how is this accounted for in your  model set-up?

L129ff: The data used include inorganic nutrients. However, for a nutrient budget it is important to estimate total nutrient loads as many nutrients may be bound to organic matter (either dissolved or particulate and in case of P, Fe-bound PO4 may be important). Especially in case of riverine phytoplankton blooms, particulate loads may be significant. Living phytoplankton may only capture a small part of the total particulate nutrient load (e.g.Hillebrand et al., 2018)  due to a substantial fraction of detritus. How do you account for this?

Furthermore, I suggest to use moles throughout the text.

Further smaller comments:

L 67-70: This sentence is difficult to understand (…despite mixing curves have (having?) been useful… Furthermore, I suggest to elaborate a bit on the limitation of using mixing curves. Wouldn’t they have been useful to constraint he presented budgets?

L185ff: The Dordogne has exceptionally high SPM (>40 * other rivers). What is the reason and what is the impact of this on the model outcome of the Gironde?

L212ff: How would an import of SPM from the coastal Atlantic influence the parameters listed in Table 5? By which modelled processes is the SPM-max generated?

L240: Again, how is dealt with particulate/organically bound nutrients?

L273ff: Are the high Gironde values for SPM driven by the high Dordogne values?  Is the SPM in the SPM max mainly riverine or marine?

L285ff: Desorption of PO4 is mentioned but from the model description I notice that this is not implemented in the present model. Please explain.

L289: I suggest to use moles instead of grams to indicate nutrient concentrations. Please note that phosphorus is misspelled as phosphorous at several instances throughout the text and in tables.

L361ff: If riverine TN is derived from DIN + N in phytoplankton, the TN load may be underestimated as riverine particulate matter may consist of phytoplankton detritus (see e.g. Hillebrand et al., 2018). Also DON and DOP are not accounted for. The potential underestimation of TN and TP loads should be discussed.

L 381ff: What drives the differences in TN/TP ratio?

L441: Unclear sentence: …often unavailable data sets (needed for?)

L507: which elimination process could be responsible (other than denitrification and sedimentation?)

L524: what is meant by diatom outburst under osmotic pressure?

Literature cited

Burchard et al. (2018) Sediment Trapping in Estuaries. Annual Review of Marine Science. Vol. 10:371-395

Hillebrand et al., (2018) Dynamics of total suspended matter and phytoplankton loads in the river Elbe Journal of Soils and Sediments (2018) 18:3104–3113

Review of “Nutrient transport and transformation in macrotidal estuaries of the French Atlantic coast: a modelling approach” by Xi Wei et al.

The manuscript describes a model that simulates transport and biogeochemistry in various estuaries in France. The 1-D grid is oriented along the flow path, which is constrained at the marine side by results from an ocean model and at the upstream river side by time-series data. The calibration of the model parameterization is based on data from 1 year and then validated based on the ability to reproduce data from the previous and following year. For most locations the model captures the trend of various biogeochemical parameters well, despite major simplifications, which include 1) a 1-D grid that cannot account for the depth-dependency of biogeochemical processes and cannot fully resolve the spatial variability in residence times; and 2) the biogeochemistry in sediments is ignored (it only accounts for organic matter burial).

My major criticism is that the role of biogeochemical processes is not well teased out. The reaction rates that the model explicitly resolves are not shown in any figure and the paper would benefit from a more rigorous comparison of the simulated rates with those reported in literature. It is unclear to what extent the trends in the modeled output are driven by biogeochemical processes and to what extent by the constantly changing boundary conditions. The influence of the boundary conditions could be especially large for sites where the location for model calibration/validation is near the inlet or outlet (such as in smaller estuaries). The reported retention rates are often low, meaning that the influx and outflux are nearly equal. This could indicate that the respective chemicals do not react much (making model results more trivial), but there could also be a dynamic balance between sinks and sources. The authors may want to improve the manuscript by describing in more detail the cycling of elements within the model domain. A sensitivity analysis would be useful to show to which biogeochemical rate constants the model results are most sensitive. Additionally, a simulation with the biogeochemical reactions turned off would be informative. Without these analyses it is hard to assess if the model captures the biogeochemical dynamics adequately.

In larger estuaries the retention of nutrients is higher, which indicates that biogeochemical processes play a more important role. At these locations early diagenetic processes could also potentially have a larger effect on the water quality. The model contains denitrification in the water-column, but as the water remains oxygenated, it is probably unimportant. Benthic denitrification is likely more important, but not modeled/parameterized. Sediments could also act as a source of nitrogen by releasing DIN derived from remineralized organic matter. The model does neither account for benthic PO4 dynamics nor benthic O2 consumption. The manuscript does not make a compelling argument for why these benthic processes can be ignored.

Overall I find the manuscript well written, but it could be further improved. The results section contains many interpretations, which do not belong in this section.

Specific Comments:

Line 240-246: Sinks and sources in the model are not only related to the inlet and the outlet, but also to exchanges with the sediment and atmosphere. It is specified for Flux_out that it refers only to the outlet. For Flux_in it is stated that it accounts for all inputs. Does this refer only to the river inlet or also inputs from the atmosphere?  The calculated residence time of water is only based on river discharge, but does not account for inputs from the ocean. For the theoretical case that the river influx goes to 0, the water residence time would approach infinity. Obviously, the real residence time of water also depends on the exchange with the ocean and the reported residence times are less meaningful near the ocean boundary. It may be good to point this out. The estuaries are described to be macrotidal, but I believe the text does not describe the vertical stratification. Perhaps the vertical stratification can be ignored in these macrotidal estuaries. Readers may appreciate a better characterization of the flow in these estuaries to have a better idea of the implications of the 1-D approach.

Line 410-412: The regulation of outflow by the dam is an interesting point. Is this regulation explicitly modeled or averaged over time?

Line 451-453: The text states that the grid resolution (2 km) may be too coarse for small estuaries (< 30) km. If the model grid contains less than 15 grid cells, simulations can probably be run quickly. Why did the authors not run simulations with a finer grid resolution?

Line 457-460: “In the present… denitrification for nitrogen”. The text ignores that nutrients are released during remineralization and can be transported back into the overlying water. The statement that denitrification is “the only other potential nutrient removal term” missing in the model is wrong. There are other benthic processes that can act as a sink for nitrogen, sediments can sequester PO4, also Si can be taken up and released by various benthic processes, and there are many other nutrients (e.g. iron) not accounted for in the model. Also, the model does not account at all for the uptake of O2 by the sediment. These benthic processes are often parameterized in ocean circulation models. The benthic remineralization rate could be estimated based on the current burial flux and then be used to approximate the exchange fluxes of O2, P and N. The authors should either implement these processes or explain why these fluxes can be ignored.

The fits for the Charente estuary are not so good. Remarkably the trends in simulated PO4 and DO values are opposite to those in the measurements in the calibrated year. Also NH4 concentrations are generally too low. After reading the manuscript it is not clear to me why the model could not reproduce these trends. If the model fails to reproduce the data, should the results be presented in the manuscript?

Many statements in the results section contain an opinion or assessment:

Line 305: “These comparisons… available observations”

Line 335: “seasonal trends are properly captured…”

Line 342-344: “DO showed… high summer mineralization”: Please point to results that back up this statement about the relative importance of the effect of solubility and mineralization. Figure 4 only shows peaks and troughs.

Line 344: “Conversely, phytoplankton… time scale”. Please, just describe the model output.

Line 354-360: “Considering the suitable agreement… retention rates.” The first sentence is an assessment and the remainder of the paragraph are methods.

Line 412-414: “Most of the… retention itself”

There are other examples where the tone is not neutral or the text is not limited to results. The authors could improve the text by revising the results section.

Minor comments:

Line 155: “Numerical Schemes”. This paragraph mentions already existing models that have been used and describes the conceptual model, but not so much the numerical implementation. Therefore, consider changing the heading.

Line 173 (Figure 2b): What process does the arrow from “Denitrification” to “Aerobic Degradation” represent?

Line 204: “spin-up” instead of “warm-up”

Line 204-205: “repeating the… steady-state conditions.” Here I got lost.

Line 209: “Calibration was implemented based on 2015, as average..” Consider rephrasing.

Line 337-338: “showed a seasonal decrease from winter to autumn”… probably from autumn to winter is meant.