the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 05 Nov 2025
Coastal Circulation and Dispersion of Passive Tracers in the Red River Plume Region: Unveiling Seasonal- and Intra-seasonal Variability
Abstract. Surface current velocity measurements by high-frequency radars (HFR) were analyzed to characterize the surface circulation and to investigate the dispersion of passive tracers in the Red River (RR) plume region within the Gulf of Tonkin (GoT) from August to December 2024. The coastal circulation in the region, found to be strongly influenced by winds, tidal forcing, riverine input, and coastal bathymetry, exhibited a large spatio-temporal variability during the analysis period with the occurrence of small-scale structures, i.e., permanent submesoscale eddies. The dispersion under varying forcing conditions and an extreme event – the typhoon Yagi – was analyzed by particle tracking and Lagrangian diagnostics. The results revealed that the dispersion within the RR plume region predominantly followed a Richardson super-diffusive regime after 24 hours of tracking. Under the influence of typhoon Yagi, the dispersion quickly followed a ballistic regime after 12 hours of tracking, with the spreading rate ten times faster than that during normal conditions. In addition, the presence of Lagrangian Coherent Structures (LCSs), i.e., eddies next to the river outflow jets, coastal plume fronts, and zones of surface current convergence and divergence in the vicinity of river outlets, significantly influenced the dispersion behavior of tracers in the RR plume region.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(4533 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(4533 KB) - Metadata XML
- BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-5203', Anonymous Referee #1, 09 Jan 2026
-
AC1: 'Reply on RC1', Kim-Cuong Nguyen, 18 Feb 2026
Dear Reviewer #1, we would like to express our gratitude for your insightful points regarding the impact of discharge on the accuracy of HFR measurements. Below, we would like to clarify in detail our responses to your comments, questions, and suggestions. The text in red (Lines XX-YY) indicates the modified lines/figures in the revised manuscript.
RC1#1
· The HFR data and drifter data generally agree. However, there is some question as to whether the accuracy of the HFR current velocity data remains the same even during periods of large river discharge.
· Although the paper makes no mention of it, large river discharge leads to lower salinity and reduced electrical conductivity of the seawater.
Since ocean radar waves propagate along the sea surface, low seawater electrical conductivity causes greater radiowave attenuation, reducing the signal-to-noise ratio (SNR) of the Doppler spectrum data.· I think this results in reduced current velocity accuracy.
· Unfortunately, we did not have a chance to release drifters during summer time due to some technical problems on field. It would have been a more comprehensive comparison, we totally agree.
We added some paragraphs to discuss on sources of uncertainties in the flow fields into Section 4.1, including the impact of river discharge on range reduction (lines 545-562). And for more detail clarification, we would like to break down to some points as below:
· We agree the low seawater electrical conductivity can reduce working range of HFR system, consequently leads to the reduce in range and effective SNR (due to loss of propagation). The SNR, affecting accuracy of radial current measurements, is influenced by sources of errors that were investigated and discussed by Essen et al., (2000); Paduan and Graber, (1997); Paduan and Washburn, (2013) including sea states, radio frequency interference (RFI), pointing errors and geometric dilution of precision (GDOP). In addition, the accuracy of surface current velocity fields can be degraded in grid cells near river outflow jets due to strong shear and eddy activity, which cause a spreading of the first-order peak in the Doppler spectra.
To investigate these effects, we selected three locations: in front of Van Uc (blue), off Van Uc (black), and in front of Tra Ly (red) (Figure below). We found that errors in surface currents (average for u and v components) within the Van Uc river plume (with a mean value of approximately 0.06 m s⁻¹) are generally larger than those observed at the other two locations (<0.04 m s⁻¹). However, these larger error values may also be partly attributed to high GDOP, as the area near Van Uc is located close to the baseline between the two radar stations. Furthermore, the accuracy of the surface flow fields does not appear to vary systematically with river discharge, as no clear correlation is observed between the temporal evolution of surface current velocity errors and the discharge time series (Figure 3 in the manuscript).
Figure: Accuracy of surface current velocity ( m s-1) at different locations: in front of Van Uc (blue), off Van Uc (black), in front of Tra Ly (red)
The constraint caused by baseline was overcome by the HF radar processing using MIO and 2DVar that has been described in the Method section 2.2.1 (lines 165-182, Figure 4).
· The low-salinity water can affect the uncertainties of the total current velocity at far range (overlapping area reduces), but not directly the accuracy of the radial currents, or so far no clear evidences on this argument has been proven. In addition, in the RR plume area, the influence of low-salinity water body does not extend more than 8km offshore and the salinity at the river mouth (Van Uc) is approx. 10-15‰ (Figure 3A, Sentchev et al., 2025), the HFR radars can work at this salinity level (Table 4, Mantovani et al., 2020).
The accuracy of the radial currents reduces linearly with the width of the first-order peaks and depends on the number of samples (Parks et al., 2009). We compared two Doppler spectra at a point in front of Van Uc River mouth on the 1 September and 5 December, at comparable night time (to neglect day-night effects on radar performance). For the current measurement, only first order peak is considered. Both peaks were well-defined (sharp peaks, two figures below) from the noise level, even the noise level in September (20dB) is two times smaller than that in December (40dB).
RC1#2
This study performs 2D interpolation of current velocity vectors. When the number of measurement data used for the interpolation differs, the accuracy of the interpolated data also varies.
Particularly when river discharge is high, there may be many missing HFR radial velocity measurements. Consequently, the resulting current field is primarily based on 2D interpolation results and is represented as a uniform flow, which may affect the analysis results.
Radial velocity fields were post-processed using the MIO radar-processing software before being used to reconstruct surface current vectors by 2DVar. The authors have furtherly clarified the processing procedure and its results in lines 165-182 with a newly-added Figure 4.
RC1#3
The comparison with drifter data was conducted during November and December, when river outflow is small.
Therefore, the accuracy of the HFR velocity data is likely better than that during periods of large river discharge.We agree with the viewer’s comment. Please refer to our response to your comment in RC1#1
RC1#4
It is necessary to discuss how the number of Doppler spectra sufficient to determine the radial velocity differs between periods of low river discharge and periods of high discharge,
and how this difference affects the accuracy of the 2D interpolated current velocity vectors.
In the radar configuration, we used 2048 samples with 512 samples each bin and the setup were the same for the entire measurement period. In MIO software, we increase the integration time to 30 minutes for better identifying the Bragg peaks and also using antenna subspaces for a better statistical distribution of doppler spectra and sources. The method applied in MIO software was described in (Dumas et al., 2025; Dumas and Guérin, 2023).
We used the same configuration throughout the analysis period and projected the data onto a universal 1 km grid. The accuracy of the 2DVar interpolation is primarily affected by data availability within the interpolation domain—specifically, whether a given grid cell is constrained by observations from both radar sites or from only one site. This point is already addressed in our response to RC1#1 above.
RC1#5
(1) Around Line 165:
The explanation was unclear.
Is this simply calculating the distance between the actual buoy and the virtual buoy every 30 minutes for up to 48 hours after buoy deployment?
The explanation here seems different.
t_i is time and N is the number of buoys, but are they adding these in Equation 1?A clearer explanation has been provided in the revised manuscript (lines 207-217),together with a new figure A1 added in Appendix A.
RC1#6
(2) Figure 6: Where is the location? What about the confidence interval? Wouldn't the rotary spectrum be better?
We agree with the reviewer and calculated the rotary spectra. The authors replaced the old figure with a new figure (updated as Figure 7) showing mean CW and CCW rotary power spectra curve (of all points over the domain) with confidence interval at ±1 sigma, and modified the Section 3.2 accordingly.
-
AC1: 'Reply on RC1', Kim-Cuong Nguyen, 18 Feb 2026
-
RC2: 'Comment on egusphere-2025-5203', Anonymous Referee #2, 11 Jan 2026
The authors present a study of circulation and dispersion in the waters of the Red-River plume region, in Tonkin Gulf. They use surface currents obtained from High Frequency Radars (HFR) for different seasonal and wind conditions, including the passage of a typhoon. Mean circulation patterns are presented, and also Lagrangian tracking of virtual particles on the HFR velocity fields is performed and characterized in terms of absolute and relative dispersion. Real surface drifters are also released. Comparison of the trajectories of the virtual trajectories with the drifters gives generally good validation of the computed trajectories. Lagrangian results are presented in terms of the classical variance-based estimators of absolute and relative dispersion.
It is somehow surprising that, despite citing several papers (Bertin et al., 2024; d’Ovidio et al., 2004, 2009; Gough et al., 2016, etc.) employing advanced methods to characterize Lagrangian transport (for example those based on different types of Lyapunov exponents) which seem better suited to describe the filamental dispersion observed in the present data, they are not used here.
Despite this limitation, which perhaps needs some comment in a revised version, the paper is generally well written and the results sound. There are however some issues that need to be addressed before I can recommend publication:
- Some additional data are needed about the drifters used (sect. 2.2.2), in particular their size, if they have some type of subsurface drogue or not, and what proportion of their heigh is exposed to the wind. These data are needed to estimate how much do these drifters follow the surface currents measured by the radar and how much are they influenced by wind. Some brief discussion on this is missed in the paper. In addition, an indication of the duration of the tracking time of the drifters is needed in sect. 2.2.2, not restricting this information to just what can be seen in the figures.
- The definition of the Lagrangian error previous to Eq. (1) is not clear enough. For example, what is the average indicated by brackets <> there?
- The discussion of the different relative dispersion regimes (Richardson, ballistic, etc.) is somehow superficial. At least some statement should be done on the fact that Richardson is expected in the direct 3d cascade and in the inverse-cascade regime in 2d, and comment in which sense these may be appropriate for the spatial scales analyzed in this paper. Also, the comment on the case beta=-1 is inappropriate, since this is non-physical (infinite energy in the system) and also violates the locality assumption leading to the exponent 4/(3-beta) for the time exponent of delta^2.
- There is a square root missing in some terms of Eq. 3a. Also, R_i^2(t) are the semiaxes length, not the axes length.
- In line 204: ‘… how large particles …’. Why ‘large’?
- Some indication on the duration of trajectories is needed in the caption of Fig. 4.
- The sentence starting at the end of line 235 seems unfinished.
- Paragraph around line 240, and Table 1: minimum velocity values are mentioned, but not presented in Table 1. In addition, a definition should be given for the maximum velocity values: in turbulent flows this can depend on the time step on which velocities are computed, so this time step should be stated. Also, in Table 1, please define ‘mean’ and ‘max’ value of ‘Sep. dist’.
- Please state in section 3.3 or 3.4 how particles are initially distributed, instead of restricting this information to what can be seen in the figures.
- In Fig. 8, what does it mean ‘Circles illustrate final positions’?
- In many parts of the text statements such as ‘… followed by a Richardson’s super-diffusive …’ or ‘… dispersed with an exponential growth …’ and similar ones appear. I do not see perfect power laws or exponential in most of the referred plots, as is expected from the lack of isotropy, homogeneity, stationarity, etc. of the studied flow. I suggest to somehow weaken this type of claims as for example ‘…approaching a Richardson …’, ‘…approximately exponential …’ at least in the more evident places. Also, in most figures the statement ‘black dashed lines are fitting curves’ appears. But these lines do not appear to be fitting the data. Are they fits or guides to the eyes? If they are fits, please specify how are they calculated.
Citation: https://doi.org/10.5194/egusphere-2025-5203-RC2 -
AC3: 'Reply on RC2', Kim-Cuong Nguyen, 18 Feb 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5203/egusphere-2025-5203-AC3-supplement.pdf
-
AC3: 'Reply on RC2', Kim-Cuong Nguyen, 18 Feb 2026
-
RC3: 'Comment on egusphere-2025-5203', Anonymous Referee #3, 16 Jan 2026
This study describes the circulation and dispersion (transport) within the Red River Plume Region of the Gulf of Tonkin at novel spatial and temporal resolution. Results expand upon historical knowledge of this region by employing in situ and simulated lagrangian drifters, characterizing dispersion across seasons and during a typhoon event. The resultant manuscript will contribute greatly to the community, especially regarding the high resolution description of the local circulation, which has not been done before.
The manuscript applies valid methods to a substantial new dataset. I think this manuscript could benefit from some restructuring, a greater emphasis on the impact of the results, and potentially some further analysis or mention of future directions. Overall, I think this is a novel study and I look forward to seeing this manuscript in publication.
Abstract:
- I suggest changing the introductory sentence to something that describes the “why” of the study rather than jumping right into the methods. Why is this novel? Why is this important?
- The second sentence starts with describing how much variability is in the system and ends with “permanent submesoscale eddies”. I suggest breaking this into two sentences and/or clarifying what you mean by “permanent”. I don’t think “permanent” is used elsewhere in the manuscript to describe these eddies.
- The last sentence of the abstract should include context as to what results from this study mean for the open questions in this region. How is this novel?
Introduction
- Line 50: resolution is presented as both a strength and a weakness of satellite data? I would leave resolution as only a weakness of satellite data, especially in comparison to HFR.
- Line 60: “cannot be fully resolved by numerical models”. I’m not sure if this is true. Perhaps this is true for the current state of models in the GoT? This is also not a great segway into the next paragraph, which doesn’t mention the resolution capable of HFR observations and/or how they capture finer scale features than models.
- Line 61-65: I think it is worth going a bit deeper into how the HFR work, such as the doppler-shifted backscattering radio waves, that two radars are needed because each one gives a radial measurement, etc.
- Line 85: I think the references here need to be broken up into several different claims. I agree with RC2 that it is strange that many advanced methods to characterize Lagrangian transport were cited but not employed (ex: FSLE in d’Ovidio et al 2004 and Hernandez-Carrasco et al 2011 FTLE in Gough et al., 2016). I suggest this sentence is broken into several sentences that outline the literature like below:
- Claim 1: Lagrangian diagnostics have been widely employed and they provide information that cannot be seen from Eulerian techniques (d’Ovidio et al 2009 and maybe a few more)
- Claim 2: These techniques can be deployed to HFR (Gough et al 2016, Veatch et al., 2024, McKee et al 2025, …https://academic.oup.com/icesjms/article/81/4/760/7633546, https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2024JC022101
- Claim 3: Past studies have used Lagrangian techniques in GoT with lower resolution (Sentchev et al 2025, Tran et al 2022)
Materials and Methods:
- Line 101: is this a upwelling scenario? Explain for clarity and readability
- Line 105: downwelling scenario?
- Line 108: add Son Tay station to figure 1?
- Line 113: are dams released when there is lots of runoff (large rain event?) or are the dam releases and the typhoon unrelated?
- Figure 1: I suggest adding an inset of a map of the larger region for context. Were the C1 and C2 release locations in that area for a scientific reason? Was this an operational constraint? I think that is worth mentioning somewhere. Also, I recommend you add a kilometer scale to the map (key).
- Figure 2: Why was ERA5 data used and not data from the Hon Dau meteorological station? How does ERA5 compare to observed data (if available)?
- Line 134: How were the subperiods decided? It is not clear to me if these three subperiods correspond to those mentioned later in the text. Can you mark these subperiods on Figure 2 or 3?
- Second paragraph is great! Nice description and clever technique.
- Line 145 needs more info. How big were the deployed drifters? How much surface expression/draft of the buoy?
- Line 152: see comment on Figure 2.
- Line 165: Awesome!
- Line 175: How the buoy differs from simulated, massless particles is not mentioned here (ex. Friction, more influence of wind with surface expression, etc.)
- Line 180: should be “2.3.2”
- Line 182: Are particle pairs those that are initially neighboring? Or particles from the whole cluster?
- Line 185: Were particles released in a uniform grid? How many particles per cluster? What is the initial separation of particles? Are these the same clusters that were described in 2.3.1?
- Line 187: is the method described here the same that is used in the referenced manuscripts?
- Line 194: This is a slightly confusing opening sentence. Does this theta(t) and R21 R22 give info on the direction of the dispersion? Which isn’t given by delta squared(t)?
- Line 210: “cluster” – all metrics calculated with the same release clusters?
Results:
- Line 225: Move this to discussion and elaborate more?
- Figure 4: were releases during similar points in the tidal cycle?
- Figure 5: move title to y-axis label
- Line 258-260 feels out of place. Plot tidal amplitude spatially to elaborate on this point? Or omit? (Spatial tidal amplitude example: Figure 7A Veatch et al 2025, https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2025JC022447)
- Line 265: do these time periods correspond with those mentioned in the methods?
- Line 315: move to discussion and elaborate? Or re-visit in discussion?
- Line323: how many particles in each cluster?
- Figure 8: using green and red together is not very colorblind friendly. Could you separate these two panels into four and not use so many colors?
- Line 363: tie bathymetry back to 20m isobath delineation shown in Figure 7? I think it would be nice to have a discussion section on the influence of bathymetry.
- Figure 10: change “20241205TZ000” to “December 5th 2024 0:00” to be more readable
- Line 409: add August 28 to supplementary material?
- Figure 11: see comment from Figure 10. Also, there are many colorblind friendly and perceptually uniform colormaps available, such as those in cmocean (https://matplotlib.org/cmocean/)
- Line 432: where are the “two coherent structures (eddies)”? one is deeper than 20m and one is shallower rotating from TraLy to ThaiBinh?
- Figure 13: change dates and colormap to be more readable (see above suggestions)
Discussion:
- Section 4.1 is really interesting!
- Line 489: I think exploring more reasons why there is a 19% difference would be really useful here. We don’t expect in situ buoys to behave exactly like virtual, masses particles. RC1 brings up a good point about the reduction in electrical conductivity of the seawater during larger river discharge events which could have a negative impact on the quality of HFR observations. I don’t believe there is any current work-around to this problem, but it is worth discussing.
- Line 496: explicitly stating that there is offshore flow in this scenario might make this section easier to follow.
- Line 501-503: this type of discussion is missing from the downwelling scenario a few lines above.
- Line 520: are these the same eddies referred to in line 432?
- Line 526-531: restating results? Start with a conceptual take-away.
- Line 535: tie back to this study. Are these features resolved by the HFR?
- Line 545: How does this fit into the current understanding of flow in the GoT?
- Line 545: How does this effect results?
I think the discussion would benefit from moving some “discussion” text from the Results to this section and expanding on a few topics such as (1) how the 20m isobath acts as a border between two different types of flows (2) forcings that create the eddy in Figure 11b and (3) the influence of the Typhoon on local circulation and dispersion.
Conclusion:
- Line 581: “structure with sizes of 10-30km” this information is missing from the discussion.
I think further analysis with uniform grid releases over the whole domain could tell us more about the boundaries between the different oceanographic regimes that were suggested by the current results and the coherent structures (eddies) mentioned throughout the text. This further analysis could be as simple as the residence time experiments done in Kohut et al. 2018 (https://royalsocietypublishing.org/rsta/article/376/2122/20170165/115543/Variability-in-summer-surface-residence-time) or the more complex Lyapunov experiments that are in many of the citations for this paper (Veatch et al 2024: https://academic.oup.com/icesjms/article/81/4/760/7633546). If this further analysis is not included in the final paper, I suggest that it is outlined in the Line 590 as a "future study".
Overall, this study uses valid methods on a novel dataset and I believe results will be widely used by the community.
Citation: https://doi.org/10.5194/egusphere-2025-5203-RC3 -
AC2: 'Reply on RC3', Kim-Cuong Nguyen, 18 Feb 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5203/egusphere-2025-5203-AC2-supplement.pdf
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-5203', Anonymous Referee #1, 09 Jan 2026
This study analyzes surface current velocity data obtained by high-frequency (HFR) ocean radar in the Red River (RR) plume region within the Gulf of Tonkin, characterizing surface circulation and investigating the dispersion of passive tracers.
It describes the dispersion processes in this sea area in detail for several cases, which is an interesting study.The HFR data and drifter data generally agree. However, there is some question as to whether the accuracy of the HFR current velocity data remains the same even during periods of large river discharge.
Although the paper makes no mention of it, large river discharge leads to lower salinity and reduced electrical conductivity of the seawater.
Since ocean radar waves propagate along the sea surface, low seawater electrical conductivity causes greater radiowave attenuation, reducing the signal-to-noise ratio (SNR) of the Doppler spectrum data.
I think this results in reduced current velocity accuracy.
The comparison with drifter data was conducted during November and December, when river outflow is small.
Therefore, the accuracy of the HFR velocity data is likely better than that during periods of large river discharge.This study performs 2D interpolation of current velocity vectors. When the number of measurement data used for the interpolation differs, the accuracy of the interpolated data also varies.
Particularly when river discharge is high, there may be many missing HFR radial velocity measurements. Consequently, the resulting current field is primarily based on 2D interpolation results and is represented as a uniform flow, which may affect the analysis results.
It is necessary to discuss how the number of Doppler spectra sufficient to determine the radial velocity differs between periods of low river discharge and periods of high discharge, and how this difference affects the accuracy of the 2D interpolated current velocity vectors.
Other comments:(1) Around Line 165:
The explanation was unclear.
Is this simply calculating the distance between the actual buoy and the virtual buoy every 30 minutes for up to 48 hours after buoy deployment?
The explanation here seems different.
t_i is time and N is the number of buoys, but are they adding these in Equation 1?(2) Figure 6: Where is the location? What about the confidence interval? Wouldn't the rotary spectrum be better?
Citation: https://doi.org/10.5194/egusphere-2025-5203-RC1 -
AC1: 'Reply on RC1', Kim-Cuong Nguyen, 18 Feb 2026
Dear Reviewer #1, we would like to express our gratitude for your insightful points regarding the impact of discharge on the accuracy of HFR measurements. Below, we would like to clarify in detail our responses to your comments, questions, and suggestions. The text in red (Lines XX-YY) indicates the modified lines/figures in the revised manuscript.
RC1#1
· The HFR data and drifter data generally agree. However, there is some question as to whether the accuracy of the HFR current velocity data remains the same even during periods of large river discharge.
· Although the paper makes no mention of it, large river discharge leads to lower salinity and reduced electrical conductivity of the seawater.
Since ocean radar waves propagate along the sea surface, low seawater electrical conductivity causes greater radiowave attenuation, reducing the signal-to-noise ratio (SNR) of the Doppler spectrum data.· I think this results in reduced current velocity accuracy.
· Unfortunately, we did not have a chance to release drifters during summer time due to some technical problems on field. It would have been a more comprehensive comparison, we totally agree.
We added some paragraphs to discuss on sources of uncertainties in the flow fields into Section 4.1, including the impact of river discharge on range reduction (lines 545-562). And for more detail clarification, we would like to break down to some points as below:
· We agree the low seawater electrical conductivity can reduce working range of HFR system, consequently leads to the reduce in range and effective SNR (due to loss of propagation). The SNR, affecting accuracy of radial current measurements, is influenced by sources of errors that were investigated and discussed by Essen et al., (2000); Paduan and Graber, (1997); Paduan and Washburn, (2013) including sea states, radio frequency interference (RFI), pointing errors and geometric dilution of precision (GDOP). In addition, the accuracy of surface current velocity fields can be degraded in grid cells near river outflow jets due to strong shear and eddy activity, which cause a spreading of the first-order peak in the Doppler spectra.
To investigate these effects, we selected three locations: in front of Van Uc (blue), off Van Uc (black), and in front of Tra Ly (red) (Figure below). We found that errors in surface currents (average for u and v components) within the Van Uc river plume (with a mean value of approximately 0.06 m s⁻¹) are generally larger than those observed at the other two locations (<0.04 m s⁻¹). However, these larger error values may also be partly attributed to high GDOP, as the area near Van Uc is located close to the baseline between the two radar stations. Furthermore, the accuracy of the surface flow fields does not appear to vary systematically with river discharge, as no clear correlation is observed between the temporal evolution of surface current velocity errors and the discharge time series (Figure 3 in the manuscript).
Figure: Accuracy of surface current velocity ( m s-1) at different locations: in front of Van Uc (blue), off Van Uc (black), in front of Tra Ly (red)
The constraint caused by baseline was overcome by the HF radar processing using MIO and 2DVar that has been described in the Method section 2.2.1 (lines 165-182, Figure 4).
· The low-salinity water can affect the uncertainties of the total current velocity at far range (overlapping area reduces), but not directly the accuracy of the radial currents, or so far no clear evidences on this argument has been proven. In addition, in the RR plume area, the influence of low-salinity water body does not extend more than 8km offshore and the salinity at the river mouth (Van Uc) is approx. 10-15‰ (Figure 3A, Sentchev et al., 2025), the HFR radars can work at this salinity level (Table 4, Mantovani et al., 2020).
The accuracy of the radial currents reduces linearly with the width of the first-order peaks and depends on the number of samples (Parks et al., 2009). We compared two Doppler spectra at a point in front of Van Uc River mouth on the 1 September and 5 December, at comparable night time (to neglect day-night effects on radar performance). For the current measurement, only first order peak is considered. Both peaks were well-defined (sharp peaks, two figures below) from the noise level, even the noise level in September (20dB) is two times smaller than that in December (40dB).
RC1#2
This study performs 2D interpolation of current velocity vectors. When the number of measurement data used for the interpolation differs, the accuracy of the interpolated data also varies.
Particularly when river discharge is high, there may be many missing HFR radial velocity measurements. Consequently, the resulting current field is primarily based on 2D interpolation results and is represented as a uniform flow, which may affect the analysis results.
Radial velocity fields were post-processed using the MIO radar-processing software before being used to reconstruct surface current vectors by 2DVar. The authors have furtherly clarified the processing procedure and its results in lines 165-182 with a newly-added Figure 4.
RC1#3
The comparison with drifter data was conducted during November and December, when river outflow is small.
Therefore, the accuracy of the HFR velocity data is likely better than that during periods of large river discharge.We agree with the viewer’s comment. Please refer to our response to your comment in RC1#1
RC1#4
It is necessary to discuss how the number of Doppler spectra sufficient to determine the radial velocity differs between periods of low river discharge and periods of high discharge,
and how this difference affects the accuracy of the 2D interpolated current velocity vectors.
In the radar configuration, we used 2048 samples with 512 samples each bin and the setup were the same for the entire measurement period. In MIO software, we increase the integration time to 30 minutes for better identifying the Bragg peaks and also using antenna subspaces for a better statistical distribution of doppler spectra and sources. The method applied in MIO software was described in (Dumas et al., 2025; Dumas and Guérin, 2023).
We used the same configuration throughout the analysis period and projected the data onto a universal 1 km grid. The accuracy of the 2DVar interpolation is primarily affected by data availability within the interpolation domain—specifically, whether a given grid cell is constrained by observations from both radar sites or from only one site. This point is already addressed in our response to RC1#1 above.
RC1#5
(1) Around Line 165:
The explanation was unclear.
Is this simply calculating the distance between the actual buoy and the virtual buoy every 30 minutes for up to 48 hours after buoy deployment?
The explanation here seems different.
t_i is time and N is the number of buoys, but are they adding these in Equation 1?A clearer explanation has been provided in the revised manuscript (lines 207-217),together with a new figure A1 added in Appendix A.
RC1#6
(2) Figure 6: Where is the location? What about the confidence interval? Wouldn't the rotary spectrum be better?
We agree with the reviewer and calculated the rotary spectra. The authors replaced the old figure with a new figure (updated as Figure 7) showing mean CW and CCW rotary power spectra curve (of all points over the domain) with confidence interval at ±1 sigma, and modified the Section 3.2 accordingly.
-
AC1: 'Reply on RC1', Kim-Cuong Nguyen, 18 Feb 2026
-
RC2: 'Comment on egusphere-2025-5203', Anonymous Referee #2, 11 Jan 2026
The authors present a study of circulation and dispersion in the waters of the Red-River plume region, in Tonkin Gulf. They use surface currents obtained from High Frequency Radars (HFR) for different seasonal and wind conditions, including the passage of a typhoon. Mean circulation patterns are presented, and also Lagrangian tracking of virtual particles on the HFR velocity fields is performed and characterized in terms of absolute and relative dispersion. Real surface drifters are also released. Comparison of the trajectories of the virtual trajectories with the drifters gives generally good validation of the computed trajectories. Lagrangian results are presented in terms of the classical variance-based estimators of absolute and relative dispersion.
It is somehow surprising that, despite citing several papers (Bertin et al., 2024; d’Ovidio et al., 2004, 2009; Gough et al., 2016, etc.) employing advanced methods to characterize Lagrangian transport (for example those based on different types of Lyapunov exponents) which seem better suited to describe the filamental dispersion observed in the present data, they are not used here.
Despite this limitation, which perhaps needs some comment in a revised version, the paper is generally well written and the results sound. There are however some issues that need to be addressed before I can recommend publication:
- Some additional data are needed about the drifters used (sect. 2.2.2), in particular their size, if they have some type of subsurface drogue or not, and what proportion of their heigh is exposed to the wind. These data are needed to estimate how much do these drifters follow the surface currents measured by the radar and how much are they influenced by wind. Some brief discussion on this is missed in the paper. In addition, an indication of the duration of the tracking time of the drifters is needed in sect. 2.2.2, not restricting this information to just what can be seen in the figures.
- The definition of the Lagrangian error previous to Eq. (1) is not clear enough. For example, what is the average indicated by brackets <> there?
- The discussion of the different relative dispersion regimes (Richardson, ballistic, etc.) is somehow superficial. At least some statement should be done on the fact that Richardson is expected in the direct 3d cascade and in the inverse-cascade regime in 2d, and comment in which sense these may be appropriate for the spatial scales analyzed in this paper. Also, the comment on the case beta=-1 is inappropriate, since this is non-physical (infinite energy in the system) and also violates the locality assumption leading to the exponent 4/(3-beta) for the time exponent of delta^2.
- There is a square root missing in some terms of Eq. 3a. Also, R_i^2(t) are the semiaxes length, not the axes length.
- In line 204: ‘… how large particles …’. Why ‘large’?
- Some indication on the duration of trajectories is needed in the caption of Fig. 4.
- The sentence starting at the end of line 235 seems unfinished.
- Paragraph around line 240, and Table 1: minimum velocity values are mentioned, but not presented in Table 1. In addition, a definition should be given for the maximum velocity values: in turbulent flows this can depend on the time step on which velocities are computed, so this time step should be stated. Also, in Table 1, please define ‘mean’ and ‘max’ value of ‘Sep. dist’.
- Please state in section 3.3 or 3.4 how particles are initially distributed, instead of restricting this information to what can be seen in the figures.
- In Fig. 8, what does it mean ‘Circles illustrate final positions’?
- In many parts of the text statements such as ‘… followed by a Richardson’s super-diffusive …’ or ‘… dispersed with an exponential growth …’ and similar ones appear. I do not see perfect power laws or exponential in most of the referred plots, as is expected from the lack of isotropy, homogeneity, stationarity, etc. of the studied flow. I suggest to somehow weaken this type of claims as for example ‘…approaching a Richardson …’, ‘…approximately exponential …’ at least in the more evident places. Also, in most figures the statement ‘black dashed lines are fitting curves’ appears. But these lines do not appear to be fitting the data. Are they fits or guides to the eyes? If they are fits, please specify how are they calculated.
Citation: https://doi.org/10.5194/egusphere-2025-5203-RC2 -
AC3: 'Reply on RC2', Kim-Cuong Nguyen, 18 Feb 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5203/egusphere-2025-5203-AC3-supplement.pdf
-
AC3: 'Reply on RC2', Kim-Cuong Nguyen, 18 Feb 2026
-
RC3: 'Comment on egusphere-2025-5203', Anonymous Referee #3, 16 Jan 2026
This study describes the circulation and dispersion (transport) within the Red River Plume Region of the Gulf of Tonkin at novel spatial and temporal resolution. Results expand upon historical knowledge of this region by employing in situ and simulated lagrangian drifters, characterizing dispersion across seasons and during a typhoon event. The resultant manuscript will contribute greatly to the community, especially regarding the high resolution description of the local circulation, which has not been done before.
The manuscript applies valid methods to a substantial new dataset. I think this manuscript could benefit from some restructuring, a greater emphasis on the impact of the results, and potentially some further analysis or mention of future directions. Overall, I think this is a novel study and I look forward to seeing this manuscript in publication.
Abstract:
- I suggest changing the introductory sentence to something that describes the “why” of the study rather than jumping right into the methods. Why is this novel? Why is this important?
- The second sentence starts with describing how much variability is in the system and ends with “permanent submesoscale eddies”. I suggest breaking this into two sentences and/or clarifying what you mean by “permanent”. I don’t think “permanent” is used elsewhere in the manuscript to describe these eddies.
- The last sentence of the abstract should include context as to what results from this study mean for the open questions in this region. How is this novel?
Introduction
- Line 50: resolution is presented as both a strength and a weakness of satellite data? I would leave resolution as only a weakness of satellite data, especially in comparison to HFR.
- Line 60: “cannot be fully resolved by numerical models”. I’m not sure if this is true. Perhaps this is true for the current state of models in the GoT? This is also not a great segway into the next paragraph, which doesn’t mention the resolution capable of HFR observations and/or how they capture finer scale features than models.
- Line 61-65: I think it is worth going a bit deeper into how the HFR work, such as the doppler-shifted backscattering radio waves, that two radars are needed because each one gives a radial measurement, etc.
- Line 85: I think the references here need to be broken up into several different claims. I agree with RC2 that it is strange that many advanced methods to characterize Lagrangian transport were cited but not employed (ex: FSLE in d’Ovidio et al 2004 and Hernandez-Carrasco et al 2011 FTLE in Gough et al., 2016). I suggest this sentence is broken into several sentences that outline the literature like below:
- Claim 1: Lagrangian diagnostics have been widely employed and they provide information that cannot be seen from Eulerian techniques (d’Ovidio et al 2009 and maybe a few more)
- Claim 2: These techniques can be deployed to HFR (Gough et al 2016, Veatch et al., 2024, McKee et al 2025, …https://academic.oup.com/icesjms/article/81/4/760/7633546, https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2024JC022101
- Claim 3: Past studies have used Lagrangian techniques in GoT with lower resolution (Sentchev et al 2025, Tran et al 2022)
Materials and Methods:
- Line 101: is this a upwelling scenario? Explain for clarity and readability
- Line 105: downwelling scenario?
- Line 108: add Son Tay station to figure 1?
- Line 113: are dams released when there is lots of runoff (large rain event?) or are the dam releases and the typhoon unrelated?
- Figure 1: I suggest adding an inset of a map of the larger region for context. Were the C1 and C2 release locations in that area for a scientific reason? Was this an operational constraint? I think that is worth mentioning somewhere. Also, I recommend you add a kilometer scale to the map (key).
- Figure 2: Why was ERA5 data used and not data from the Hon Dau meteorological station? How does ERA5 compare to observed data (if available)?
- Line 134: How were the subperiods decided? It is not clear to me if these three subperiods correspond to those mentioned later in the text. Can you mark these subperiods on Figure 2 or 3?
- Second paragraph is great! Nice description and clever technique.
- Line 145 needs more info. How big were the deployed drifters? How much surface expression/draft of the buoy?
- Line 152: see comment on Figure 2.
- Line 165: Awesome!
- Line 175: How the buoy differs from simulated, massless particles is not mentioned here (ex. Friction, more influence of wind with surface expression, etc.)
- Line 180: should be “2.3.2”
- Line 182: Are particle pairs those that are initially neighboring? Or particles from the whole cluster?
- Line 185: Were particles released in a uniform grid? How many particles per cluster? What is the initial separation of particles? Are these the same clusters that were described in 2.3.1?
- Line 187: is the method described here the same that is used in the referenced manuscripts?
- Line 194: This is a slightly confusing opening sentence. Does this theta(t) and R21 R22 give info on the direction of the dispersion? Which isn’t given by delta squared(t)?
- Line 210: “cluster” – all metrics calculated with the same release clusters?
Results:
- Line 225: Move this to discussion and elaborate more?
- Figure 4: were releases during similar points in the tidal cycle?
- Figure 5: move title to y-axis label
- Line 258-260 feels out of place. Plot tidal amplitude spatially to elaborate on this point? Or omit? (Spatial tidal amplitude example: Figure 7A Veatch et al 2025, https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2025JC022447)
- Line 265: do these time periods correspond with those mentioned in the methods?
- Line 315: move to discussion and elaborate? Or re-visit in discussion?
- Line323: how many particles in each cluster?
- Figure 8: using green and red together is not very colorblind friendly. Could you separate these two panels into four and not use so many colors?
- Line 363: tie bathymetry back to 20m isobath delineation shown in Figure 7? I think it would be nice to have a discussion section on the influence of bathymetry.
- Figure 10: change “20241205TZ000” to “December 5th 2024 0:00” to be more readable
- Line 409: add August 28 to supplementary material?
- Figure 11: see comment from Figure 10. Also, there are many colorblind friendly and perceptually uniform colormaps available, such as those in cmocean (https://matplotlib.org/cmocean/)
- Line 432: where are the “two coherent structures (eddies)”? one is deeper than 20m and one is shallower rotating from TraLy to ThaiBinh?
- Figure 13: change dates and colormap to be more readable (see above suggestions)
Discussion:
- Section 4.1 is really interesting!
- Line 489: I think exploring more reasons why there is a 19% difference would be really useful here. We don’t expect in situ buoys to behave exactly like virtual, masses particles. RC1 brings up a good point about the reduction in electrical conductivity of the seawater during larger river discharge events which could have a negative impact on the quality of HFR observations. I don’t believe there is any current work-around to this problem, but it is worth discussing.
- Line 496: explicitly stating that there is offshore flow in this scenario might make this section easier to follow.
- Line 501-503: this type of discussion is missing from the downwelling scenario a few lines above.
- Line 520: are these the same eddies referred to in line 432?
- Line 526-531: restating results? Start with a conceptual take-away.
- Line 535: tie back to this study. Are these features resolved by the HFR?
- Line 545: How does this fit into the current understanding of flow in the GoT?
- Line 545: How does this effect results?
I think the discussion would benefit from moving some “discussion” text from the Results to this section and expanding on a few topics such as (1) how the 20m isobath acts as a border between two different types of flows (2) forcings that create the eddy in Figure 11b and (3) the influence of the Typhoon on local circulation and dispersion.
Conclusion:
- Line 581: “structure with sizes of 10-30km” this information is missing from the discussion.
I think further analysis with uniform grid releases over the whole domain could tell us more about the boundaries between the different oceanographic regimes that were suggested by the current results and the coherent structures (eddies) mentioned throughout the text. This further analysis could be as simple as the residence time experiments done in Kohut et al. 2018 (https://royalsocietypublishing.org/rsta/article/376/2122/20170165/115543/Variability-in-summer-surface-residence-time) or the more complex Lyapunov experiments that are in many of the citations for this paper (Veatch et al 2024: https://academic.oup.com/icesjms/article/81/4/760/7633546). If this further analysis is not included in the final paper, I suggest that it is outlined in the Line 590 as a "future study".
Overall, this study uses valid methods on a novel dataset and I believe results will be widely used by the community.
Citation: https://doi.org/10.5194/egusphere-2025-5203-RC3 -
AC2: 'Reply on RC3', Kim-Cuong Nguyen, 18 Feb 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5203/egusphere-2025-5203-AC2-supplement.pdf
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 296 | 130 | 34 | 460 | 22 | 27 |
- HTML: 296
- PDF: 130
- XML: 34
- Total: 460
- BibTeX: 22
- EndNote: 27
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Thanh Huyen Tran
Alexei Sentchev
Dylan Dumas
Charles-Antoine Guerin
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(4533 KB) - Metadata XML
This study analyzes surface current velocity data obtained by high-frequency (HFR) ocean radar in the Red River (RR) plume region within the Gulf of Tonkin, characterizing surface circulation and investigating the dispersion of passive tracers.
It describes the dispersion processes in this sea area in detail for several cases, which is an interesting study.
The HFR data and drifter data generally agree. However, there is some question as to whether the accuracy of the HFR current velocity data remains the same even during periods of large river discharge.
Although the paper makes no mention of it, large river discharge leads to lower salinity and reduced electrical conductivity of the seawater.
Since ocean radar waves propagate along the sea surface, low seawater electrical conductivity causes greater radiowave attenuation, reducing the signal-to-noise ratio (SNR) of the Doppler spectrum data.
I think this results in reduced current velocity accuracy.
The comparison with drifter data was conducted during November and December, when river outflow is small.
Therefore, the accuracy of the HFR velocity data is likely better than that during periods of large river discharge.
This study performs 2D interpolation of current velocity vectors. When the number of measurement data used for the interpolation differs, the accuracy of the interpolated data also varies.
Particularly when river discharge is high, there may be many missing HFR radial velocity measurements. Consequently, the resulting current field is primarily based on 2D interpolation results and is represented as a uniform flow, which may affect the analysis results.
It is necessary to discuss how the number of Doppler spectra sufficient to determine the radial velocity differs between periods of low river discharge and periods of high discharge, and how this difference affects the accuracy of the 2D interpolated current velocity vectors.
Other comments:
(1) Around Line 165:
The explanation was unclear.
Is this simply calculating the distance between the actual buoy and the virtual buoy every 30 minutes for up to 48 hours after buoy deployment?
The explanation here seems different.
t_i is time and N is the number of buoys, but are they adding these in Equation 1?
(2) Figure 6: Where is the location? What about the confidence interval? Wouldn't the rotary spectrum be better?