General Comment

The paper present Lagrangian Coherent Structures as part of a set of methods for identifying coherent eddies and tracking them. The aim is: compute the coherency of the recent past, synchronize the analysis with ocean color products, and provide a high temporally resolved atlas of coherent eddy boundaries. There are many researches on ‘eddy’ identification with automatic algorithms, machine learning, etc. Examples of papers on this issue are:

• ESSD, 14, 1087–1107, 2022: 1exp: a new global mesoscale eddy trajectory atlas derived from altimetry
• Assessment of global eddies from satellite data by a scale-selective eddy identification algorithm (SEIA). Clim Dyn (2023). https://doi.org/10.1007/s00382-023-06946-w
• Global Oceanic Eddy Identification: A Deep Learning Method From Argo Profiles and Altimetry Data DOI=10.3389/fmars.2021.646926   

Some published papers are also providing eddy mesoscale atlases at a global or regional level, others are on idealized eddy resolving model applied to North Pacific to identify rotationally coherent Lagrangian vortices (e.g.  https://doi.org/10.1029/2019JC015576).

In practice I consider the article not original and based on algorithms consolidated in the literature. However, the work, done by a young researcher, is interesting, since it enriches an important theme.

Specific comments

The paper is presenting a general literature revue in &1.2, but from my point of view is insufficient. It should include also a discussion on open questions and propose the solutions e.g. (few questions coming in my mind)

• SSH eddies overestimate coherent core and fail to reveal more than half of Lagrangian eddies: is the paper providing an answer?
• Looking at figures and movies, is there a significant difference between mesoscale eddies from random ocean pieces in material coherence?
• How effective is transport, including chaotic stirring and filamentation outside of eddy cores?

The definition of Lagrangian eddy must be made explicit. There are many possible choices, but in this paper is based on concept of coherent rotation of water parcels over the eddy lifetime (RCLV and LAVD are two alternatives, but here the second is adopted),

I really enjoyed this manuscript: I learned more from it than I generally do from manuscripts I review, it was well written and, I believe that the dataset they are introducing is of value, potentially significant value. Having said this, I do have a number of comments/suggestions, none of which relate to significant issues with the approach or results. These fall into two categories, simple editorial suggestions to improve clarity and, somewhat more significant comments, suggesting additional analyses or plots. I want to emphasize that these are suggestions for the authors to follow or not. Both sets of comments are indicated in the attached marked-up version of the manuscript, however, I will repeat the more significant ones here. (Oh yeah, I have also highlighted in yellow some of the definitions used. These, which the authors should ignore, were for my own benefit as I reviewed the manuscript.)

1) I’m a little confused about the definition of CD. My sense is that it should have dimensions of length: contour area / perimeter of the enclosing boundary but it is presented as unitless? I’m assuming that ‘contour area’ is the area enclosed by a closed contour?

2) I struggled a bit with what was meant by overlap of RCLV over SLA or vice versa. It might help to label a few examples of each in Fig. 1 and to say, when introducing the idea, that rarely is an eddy of one species not almost completely contained in one of the other species. I have more comments related to this in the marked-up version of the manuscript. 

3) It is argued that the area versus lifespan plot of RCLV eddies show a linear relationship while a similar plot for SLA eddies does not. First, the linearity of the relationship looks relatively weak but I’m willing to go along with it. I also agree that the SLA distribution is less linear. However, if one concentrates on the same area—lifespan range as that of the RCLV eddies, 0-3 km, 0-140 days the distribution looks like it may be close to that of RCLV, i.e., as linear, although it’s hard to tell from the plot as is.

4) Also related to Fig. 6 I think that it might be interesting to show a figure similar to this one but with the color of each bin being the fraction of eddies in the bin that are RCLV eddies. This would give a sense for the probability of overlap as a function of size and duration. I think that it would be informative but would have to see the plot to be sure. Just an idea.

5) I found the distributions of eddies shown in Appendix E, in support of the discussion of Fig. 8 to be really interesting. There is some discussion of this but I had to dig to sort things out. I’m guessing that a lot of readers will skip the appendices and, as a result, will miss this. It might be worth moving Figs. E1 and E2 into the body of the text and discussing in a little more detail. 

6) I found Fig. 9, showing the notion of trapped parcels, to be very helpful. However, I wondered while looking at the figure whether or not a parcel near the edge of the RCLV eddy would have remained in the eddy—the example shown is for a parcel near the center of the RCLV eddy. I guess that what I’m saying is that I would have been more convinced had a parcel been chosen near its edge.  I understand that parcels on the edge of these eddies may leak out of them due to turbulence but still, I would have used one closer to the edge. Actually, an interesting plot would show the portion of the initial RCLV eddy for which, say, less than 5% leaked out in one color (maybe red), the portion for which less than 5% remained in the corresponding SLA contour in a second color  (maybe green) and the region between the two in gray. I also realize that this is just one example but such a plot would still convey a lot of information.

7) Also related to Fig. 9 you discuss the August feature but not the rather startling increase at the end of September. A useful corollary plot would be one showing the area of the SLA eddy, the area of the RCLV eddy, the  mean Chl value in SLA and the mean Chl value in RCLV all  four lines as a function of time.

Again, I really enjoyed the manuscript.

Hope that this helps.

Here is a listing of all comments in the attached marked-up manuscript.

Notes in ‘essd-2023-425’

Notes in Document 

'essd-2023-425':

Comment:        The definition of ephemeral is ‘lasting for a very short time’. ‘Short’ is, of course, a subjective term but my sense is that it is much shorter than the length of time characteristic of many mesoscale eddies. I think that I would remove ephemeral and add ‘and temporal scales of weeks to months’ at the end of the sentence’. Or just remove it without adding a temporal scale at the end of the sentence since you mention the temporal scale in a few sentences.

(essd-2023-425, p.1)

Comment:        I’m a little confused about the definition of CD. My sense is that it should have dimensions of length: contour area / perimeter of the enclosing boundary but it is presented as unitless below? I’m assuming that ‘contour area’ is the area enclosed by a closed contour?

(essd-2023-425, p.7)

Comment:        Does the intensity of the contours represent age or something else?

(essd-2023-425, p.9)

Comment:        It took me a bit to sort out what you are saying here. I think that the thing that I missed at first was that in virtually all cases an RCLV contour completely or almost completely contains an SLA contour or vice-versa when they are referencing the same feature. I had imagined cases where one could contain a significant region that was outside the other; e.g., eddies of the same size but overlapping by, say, 50%. In retrospect, I see that this is very unlikely since they are working with the same SLA field although there are some cases where the overlap is not complete, such as the cyclone at 21 N, 229 E. It would have helped had you labeled several examples in Fig. 1, examples of RCLV overlapping SLA, SLA overlapping RCLV and non-overlapping examples. You might also comment on the rarity of cases where the identified eddies are clearly the same but have more than, say, 10% of the area that is non-overlapping.

(essd-2023-425, p.11)

Comment:        It looks to me like they decrease in size over the last 20% of their lifetimes versus the last 50%—the 70% box looks like the mirror image about 0 of the 30% box.

(essd-2023-425, p.13)

Comment:        The contours in Fig. 6 seem to suggest that for SLA eddies similar in size to RCLV eddies (<~ 40,000 km^2) the size/lifespan relationship is much tighter than for the full range of SLA eddies; much closer to that of RCLV eddies.

(essd-2023-425, p.13)

Comment:        Tough to see the e^4 label on the y-axis.

(essd-2023-425, p.14)

Comment:        Should this be ‘of anticyclonic SLA’

(essd-2023-425, p.15)

Comment:        What’s really striking is that the relative distributions of cyclonic eddies are very similar over the entire domain while the distribution of anticyclonic eddies is similar everywhere except in the region west of the islands (Figs. E1 and E2). Think that this is what you are trying to say but I had to dig to sort it out.



Hmmm… I just looked at the cyclonic distribution a little more carefully and a striking feature of that distribution is the near total lack of RCLV cyclonic eddies in the region 195-200 E, 18-20 N. I think that there’s more here that is worth discussing.

(essd-2023-425, p.15)  

Comment:        Does it mean that a subset of anticyclonic Lee Eddies are not coherent or that virtually none of them are?

(essd-2023-425, p.15)

Comment:        I wonder if the SLA distribution would be better defined if you constrained it to eddies <~40,000 km^2? I guess that this would be similar to what you find in Fig. E3.

(essd-2023-425, p.16)

Comment:        A more dramatic and convincing demonstration would have been two parcels very close to one another but with one leaving the region and the other remaining in the eddy. For example, the end point of a parcel that is near (but just inside of the 23 April RCLV boundary) the end point of the orange trajectory. The orange parcel that you picked is very close to the edge of the SLA boundary while the blue parcel is very near the center of the RCLV eddy.

(essd-2023-425, p.17)

Comment:        I think that I would use ‘of’ instead of ‘in’’ here?

(essd-2023-425, p.17)

Comment:        I assume that this region includes the region inside the boundary. Might it be a better measure of the background to exclude the region with the boundary, effectively an annulus although not perfectly circular?

(essd-2023-425, p.17)

Comment:        Yes, but wouldn’t this be true of almost any closed convex—not that RCLV boundaries are necessarily convex but they are like close to being so—contour on the interior of the SLA boundary since the Chl concentration decreases away from the middle of the eddy?

(essd-2023-425, p.17)

Comment:        Are there times when this RCLV eddy is not completely inside the SLA boundary? If it is not completely contained maybe rewording as ‘despite the fact that at most xx% of RCLV eddy lies outside of the SLA eddy’? I’m guessing that xx would be less than 5%, which would make your point here as well as providing an example of “overlap”. Actually, I would find a simple plot of fraction outside, the xx above, versus time (panel e in the above) to be interesting unless—or maybe even if—it is zero most of the time.

(essd-2023-425, p.17)

Comment:        mid to late July?

(essd-2023-425, p.17)

Comment:        How did you get values for 7 and 15 April when it appears that you have not defined the RCLV eddy? Maybe I’m not uderstanding how the red line was calculated?

(essd-2023-425, p.18)

Comment:        You discuss the August feature but don’t mention the rather startling anomaly peak on 8 October. I’m guessing that it relates to uncertainties in the measurements either of area or Chl concentration. A useful plot would be one showing the area of the SLA eddy and the RCLV eddy as a function of time along with the mean Chl values in the eddies.

(essd-2023-425, p.18)

Comment:        Given the large spread in the scatter plot of Fig. 6, arguing that it represents a linear relationship is a bit of a stretch. Furthermore, it appears to me that SLA eddies in a similar size/duration range have a similar ‘linear’ relationship but I agree that it is hard to tell for sure from the plots.



Thinking a bit more about this it might be interesting to show a figure similar to Fig. 6 but with the color of each bin being the fraction of eddies in the bin that are RCLV eddies. This would give a sense of for the probabillity of overlap as a function of size and duration. I think that it would be informative but would have to see the plot to be sure. Just an idea.

(essd-2023-425, p.19)

Comment:        You might want to expand acronyms the first time they appear in an Appendix, which is referenced prior to the definition of the acronym in the main text.

(essd-2023-425, p.22)

Comment:        Why do you show this in percentage here but present in the text as a factor. Personally, I prefer factor but, regardless, I would be consistent—less mental gymnastics for the reader.

(essd-2023-425, p.27)