Population Genetics to Population Genomics: Revisiting Multispecies Connectivity of the Hawaiian Archipelago^†

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript “Population Genetics to Population Genomics: Revisiting Multispecies Connectivity of the Hawaiian Archipelago” aims to 1. Find congruent barriers between taxonomically diverse species and 2. Identify if, by increasing the number of markers and depth of genomic information, the pre-identified barriers are shifted. The authors did a good job in taking the opportunity of an ideal natural sampling design and from the data-rich study area, selecting species with different traits, using the most suitable datasets and methods. Although the manuscript is not completely innovative, since other multispecies studies were already performed in the region, it adds a layer on population structure patterns, and can contribute to decision on genetic markers selection for future studies. I only have some minor suggestions of improvements throughout the manuscript. Please see my specific comments below.

Abstract

L35 - add that the pairwise comparison was for genetic differences

Introduction

L74 - challenge

L76 - marine - aquatic? Freshwater fish also disperse during larval stage

Figure 1 - brown or orange? Maybe by removing the stroke of the rectangles, it will be easier to visualize the figure. It is not clear why some islands are in blue and some are in green.

Methods

Some information from other parts of the manuscript should be moved to the methods:

1. a) which species were pooled per location and not per island (it is in L308)
2. b) the different morphotypes for M. capitata (L314)
3. c) that some sampling locations had a difficult access (L356)
4. d) that the correlation between Fst and larval model was tested (L467)

Table 1 - I imagine that HI means Hawaiian Islands, but in the legend it is mentioned Hawaiian Archipelago, what makes it a bit confusing.

Table 2 - The manuscript has a large number of tables. I suggest removing this one (or move to supplementary material), and include the complete names together with figure 1, in the legend.

Table 3 - Including a figure with the sampling location for each species in the supplementary material would be very helpful.

Table 5 - Can be moved to supplementary material

L278 - p≤0.05 - p>0.05?

 

Results

L308 - The pools per location should be explained in the methods

L314 - The comparisons between morphotypes should be in the methods

L325 - I suggest keeping the discussion about minimum pool size for the discussion

L328 - I suggest keeping the discussion about the species dispersal capacity for the discussion

L337 - Again I suggest leaving the discussion about pool sizes to the discussion

L355 - Move the explanation about sampling design to methods

 

Discussion

L379 - the first paragraph is a bit too repetitive with the introduction. I suggest focusing on the importance of finding congruent patterns of connectivity and why this is relevant.

L423 - greatest dispersal potential - as larvae or in general, as adults also?

L423 - the species analyzed by Iacchei had a larger or smaller dispersal capacity than the ones analyzed in the present study? It is not clear.

L436 - “low sample size” which can over or underestimate the divergence?

L467 - “we find that FST correlates significantly 467 with passive larval model predictions” - where is this on the methods?

L491 - “model other” such as?

Conclusion

L511 “each island should be considered a distinct management unit” - consistent to previous studies?

Author Response

Reviewer 1:

 

The manuscript “Population Genetics to Population Genomics: Revisiting Multispecies Connectivity of the Hawaiian Archipelago” aims to 1. Find congruent barriers between taxonomically diverse species and 2. Identify if, by increasing the number of markers and depth of genomic information, the pre-identified barriers are shifted. The authors did a good job in taking the opportunity of an ideal natural sampling design and from the data-rich study area, selecting species with different traits, using the most suitable datasets and methods. Although the manuscript is not completely innovative, since other multispecies studies were already performed in the region, it adds a layer on population structure patterns, and can contribute to decision on genetic markers selection for future studies. I only have some minor suggestions of improvements throughout the manuscript. Please see my specific comments below.

Abstract

L35 - add that the pairwise comparison was for genetic differences

-Thank you for this correction, it has been fixed in the updated text

Introduction

L74 - challenge

-Thank you for this correction, it has been fixed in the updated text

L76 - marine - aquatic? Freshwater fish also disperse during larval stage

-We agree that aquatic larval dispersal is more inclusive in this broad statement and we have updated the text to reflect this.

Figure 1 - brown or orange? Maybe by removing the stroke of the rectangles, it will be easier to visualize the figure. It is not clear why some islands are in blue and some are in green.

-Thank you for these suggestions to improve clarity on this figure. We have updated the legend as red-orange for clarification. We have also added an additional sentence to the legend to clarify color scaling. The Northwestern Hawaiian islands are mostly submerged atolls, represented in blues, while the land above sea-level is scaled in green shades. The thin line surrounding them represents the 1000m bathymetric isoline surrounding the islands and atolls.

Methods

Some information from other parts of the manuscript should be moved to the methods:

-Agreed, we have done as requested.

1. a) which species were pooled per location and not per island (it is in L308)

-Expanded upon in methods and notated in Table 3

1. b) the different morphotypes for M. capitata (L314)

-Expanded upon in methods text.

1. c) that some sampling locations had a difficult access (L356)

-Moved to methods.

1. d) that the correlation between Fst and larval model was tested (L467)

-We did not actually perform this analysis, so we have revised this and cited the dissertation work of Conklin (2024) from which the correlation between FST  and the larval modelling was drawn. 

Table 1 - I imagine that HI means Hawaiian Islands, but in the legend it is mentioned Hawaiian Archipelago, what makes it a bit confusing.

-HI is the US legal abbreviation for Hawaiʻi, which we now spell out to avoid future confusion.

Table 2 - The manuscript has a large number of tables. I suggest removing this one (or move to supplementary material), and include the complete names together with figure 1, in the legend.

-Agreed, we have moved this table to Supplementary as requested, and added the complete site names to the figure legend as requested.

Table 3 - Including a figure with the sampling location for each species in the supplementary material would be very helpful.

-Agreed, we have added a map with the sampling locations for each species to the supplementary materials (Figs S2a-g) to illustrate the sampling information provided in Table 3.

Table 5 - Can be moved to supplementary material

-Agreed and it has been moved to Table S1

L278 - p≤0.05 - p>0.05?

 -Corrected, thank you.

Results

L308 - The pools per location should be explained in the methods

-We apologize that this was unclear, we have revised the text for clarity regarding the pooling of samples by sampling location and details of intra-island pools in the methods.

L314 - The comparisons between morphotypes should be in the methods

-Agreed and corrected

L325 - I suggest keeping the discussion about minimum pool size for the discussion

- Agreed, we have revised the manuscript to move this to the discussion.

L328 - I suggest keeping the discussion about the species dispersal capacity for the discussion

- Agreed, we have revised the manuscript to move this to the discussion.

L337 - Again I suggest leaving the discussion about pool sizes to the discussion 

- Agreed, we have revised the manuscript to move this to the discussion.

L355 - Move the explanation about sampling design to methods

- Agreed, this has been moved

Discussion

L379 - the first paragraph is a bit too repetitive with the introduction. I suggest focusing on the importance of finding congruent patterns of connectivity and why this is relevant.

- Agreed, we have revised the manuscript to narrow the focus and avoid repetition with the introduction as suggested.

L423 - greatest dispersal potential - as larvae or in general, as adults also?

- We apologize for the lack of clarity. Spiny lobsters have larvae specialized for open ocean dispersal (phyllosoma) with a very long pelagic larval duration (~7-11 months PLD). While adult lobsters can certainly migrate, the water depth between many of the Main Hawaiian islands exceeds 1000m with some channels reaching up to 4500m depth. This depth limits benthic dispersal among the Hawaiian Islands, so most dispersal in this system is via pelagic larvae, and we have clarified this point in the revised text.

L423 - the species analyzed by Iacchei had a larger or smaller dispersal capacity than the ones analyzed in the present study? It is not clear.

- We apologize for the lack of clarity. The species in Iacchei et al. are the same as the ones analyzed in our study (Panulirus marginatus and P. pennicilatus), but we meant to communicate that spiny lobsters have substantially longer PLD than the other species included here (Table 1).  We have revised the text to ensure our meaning is now clear.

L436 - “low sample size” which can over or underestimate the divergence?

- We apologize for the lack of clarity. Small pool sizes tend to affect significance of FST  calls (Anderson et al. 2014; Hivert et al. 2018; Kurland et al. 2019), leading to many loci not meeting filtering thresholds. Since statistical power is proportional to the number of loci in the analysis (Ryman et al. 2006; Morin et al. 2009), and the number of individuals in a pool is incorporated directly into the FST  calculations of both PoPoolation2 and {poolfstat}, it is unsurprising to us that mean FST  was not significantly different from zero. We have revised the text to explain that this is not an over or underestimation issue, but most likely a statistical power issue from fewer acceptable loci.

L467 - “we find that FST correlates significantly with passive larval model predictions” - where is this on the methods?

- We clarify that this is reporting prior work rather than original analyses in this manuscript. This correlation was performed in a previously published dissertation on passive larval dispersal that we now reference (Conklin 2024) more clearly to avoid future confusion.

L491 - “model other” such as?

- We have revised the text to clarify that we are referring here to Crandall et al.’s (2019) work testing a suite of seven population structure models including panmixia, island model, several different regional groups, and stepping-stone structure. For nearly 70% of the 41 species in that study, the stepping-stone model was the best fit for the observed population genetic data. We have clarified our message by revising this section of the manuscript in an attempt to avoid future confusion.

Conclusion

L511 “each island should be considered a distinct management unit” - consistent to previous studies?

- Somewhat - previous studies lacked the statistical power to resolve fine-scale structure at the level of individual islands, but this finding is entirely consistent with predictions from the coalescent modeling and larval dispersal simulation work. We have revised the text in both the Intro and Discussion to better communicate that our findings are different than but not at odds with the previous studies. We simply have greater statistical power to resolve even finer scale structure within the broader regions detected in previous genetic studies and our findings are consistent with both theoretical expectations and larval dispersal simulations now outlined in the introduction.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

I appreciate the premise and ambition of this manuscript, for it tackled two worthwhile research objectives. This manuscript was well-written with clear goals, extensively researched, and exploited an impressive dataset. It has key components that would make it a valuable contribution to the scientific literature.

 

There are also some limitations to this study. I think these can be overcome but would require reanalysis of the data. Overall, I found the data analysis underwhelming. I am an advocate for keeping analyses simple when it is appropriate to do so. However, inferences of gene flow and genetic structure are based solely on F_ST values. This provides a narrow window into patterns of differentiation. Why not perform a multivariate analysis, or a clustering analysis, or an AMOVA or estimate migration rates, or generate migration surfaces? One answer is that the dataset is inappropriate for some of these analyses since individuals were pooled to estimate site-specific allele frequencies. I am fine with PoolSeq, but I do think its limitations should be acknowledged. For example, from what I can tell, the dataset was never examined for whether closely related individuals were collected from the same sampling site, which could bias allele frequency estimates.

 

Also, I was hoping for more in terms of a direct comparison in the performance of the PoolSeq dataset to other marker types. Such is suggested in the last sentence of the Introduction. But the output derived from the different datasets are not directly compared, instead only the inferences of dispersal barriers. Why not regress the with F_ST estimates from microsatellite and mitochondrial markers with geographic distance, as was done with the PoolSeq data? What is the correlation between the pairwise F_ST values derived from the different marker types? For this assessment to truly be valuable, it should provide a side-by-side comparison of the output from each marker type.

 

The Introduction felt lengthy and could be more concise: it covers a lot of ground. Is the theme of the manuscript marine connectivity, comparing genetic markers, or both? It bounces between these two and could be condensed. This is then mirrored in the Discussion: use of subheadings based on the two main themes would help keep similar content together.

 

In summary, I do not think there are any fatal flaws in the study design and the subject matter would be of interest to the readership of this journal and the scientific community at large. The methods, data generation, and interpretation are mostly suitable for publication. However, I found the data analysis lacking and think readers will as well. I would recommend revising the manuscript to provide more insights into connectivity and comparisons between marker types.

 

Below are more specific comments.

 

1. Page 3, lines 123-124: Change to Conklin et al. to be consistent with other citations (e.g., line 116).
2. Page 5, lines 164-169: I really appreciate that the research questions are so explicit.
3. Page 6, line 201: Thank you for providing the indigenous names as a cross-reference.
4. Page 9, line 285: The sentence just ends.
5. Page 11, line 347: Check figure reference.
6. Page 11, Figure 2: A general comment I have is the lack of discussion on the biological significance of these results. Treating statistical difference in p-values from zero as evidence of robust biological patterns is problematic. Figure 2 exemplifies this: the correlations between F_ST and geographic distance are barely note-worthy. Some of the slopes, such as for Mfla, appear to be heavily influenced by negative F_ST values, which is below the theoretical limit of F_ST. Or I could be misinterpreting the figure. Line 349 says there was no IBD among islands or sampling sites, but then what are the points in Figure 2? Are they F_ST values between islands, which was significant (line 348), or between all sampling sites?
7. Page 12, Figure 3: This figure is too much. Simplify: what is the message you are trying to get across? The map does not need every piece of information.
8. Page 13, Figure 4: I do not like this heatmap. Interpreting the figure means finding an individual cell in the heatmap and finding the corresponding cell in the diagonal, which is difficult. Why not a bar graph for each site showing the proportion of pairwise comparisons that were significant? Avoid the red-green color scheme, it is not colorblind friendly.
9. Page 13, lines 401-403: This is a stretch as no actual cost comparison was provided.
10. Page 15, lines 466-474: I am confused, was this a novel analysis specific to this study? Why bury it in the Discussion? Describe in the Methods and present in the Results.
11. Page 16, lines 520-526: I am confused as to the threshold used to determine “significant” genetic breaks. Lines 276-301 state that significance was determine using exact test or t-tests. But then it says here that the detection threshold for significant population structure was 0.0016. Which was it? The remainder of the sentence implies that all the pairwise comparison below 0.0016 were not significant, presumably following the exact tests. So 0.0016 really is not the threshold of significance, it is just the highest F_ST observed in a nonsignificant test? Did any statistically significant tests have a F_ST less than 0.0016? Also, I will again bring up the biological significance of this value. I would argue whether less than 100 effective migrants per generation constitutes a “genetic barrier”.

Author Response

I appreciate the premise and ambition of this manuscript, for it tackled two worthwhile research objectives. This manuscript was well-written with clear goals, extensively researched, and exploited an impressive dataset. It has key components that would make it a valuable contribution to the scientific literature.

- We thank the referee for their kind words.

There are also some limitations to this study. I think these can be overcome but would require reanalysis of the data. Overall, I found the data analysis underwhelming. I am an advocate for keeping analyses simple when it is appropriate to do so. However, inferences of gene flow and genetic structure are based solely on FST values. This provides a narrow window into patterns of differentiation. Why not perform a multivariate analysis, or a clustering analysis, or an AMOVA or estimate migration rates, or generate migration surfaces? One answer is that the dataset is inappropriate for some of these analyses since individuals were pooled to estimate site-specific allele frequencies. I am fine with PoolSeq, but I do think its limitations should be acknowledged. For example, from what I can tell, the dataset was never examined for whether closely related individuals were collected from the same sampling site, which could bias allele frequency estimates.

-We appreciate the feedback and agree with the referee that the analyses are simple, but felt that they communicated our point. We have added to these analyses as suggested, but possibilities with PoolSeq data are limited to those based on allele frequency differences among populations so options like relatedness or migration are not available. We have now added both AMOVA and k-means clustering analyses as a new table, and included a series of PCA plots in the supplementary materials. 

Also, I was hoping for more in terms of a direct comparison in the performance of the PoolSeq dataset to other marker types. Such is suggested in the last sentence of the Introduction. But the output derived from the different datasets are not directly compared, instead only the inferences of dispersal barriers. Why not regress the with FST estimates from microsatellite and mitochondrial markers with geographic distance, as was done with the PoolSeq data? What is the correlation between the pairwise FST values derived from the different marker types? For this assessment to truly be valuable, it should provide a side-by-side comparison of the output from each marker type.

- We understand the point the referee is making, but respectfully disagree that our assessment is not valuable without such a side-by-side comparison of marker type. Both Weersing & Toonen 2009 and Selkoe & Toonen 2011 reported the analyses proposed by the referee and the latter showed log-fold differences between mtDNA and microsatellite estimates across more than 100 species. Multiple other studies have performed similar comparisons and confirmed little or no correlation among marker classes from the same individuals (e.g., Teske et al. 2018; Zimmerman et al. 2020; Englmaier et al. 2024). A poor correlation among FST  values from microsatellites, mitochondrial markers and SNPs is actually expected, because FST  scales inversely to the maximum within population heterozygosity of loci, which varies dramatically among these marker types (Hedrick 2005; Holsinger & Weir 2009; Meirmans & Hedrick 2011; Bird et al. 2011; Alcala & Rosenberg 2021). We believe that the time for such marker comparisons is past, which is why we did not undertake this proposed analysis. 

From our perspective, the direct side-by-side comparison of the FST  values among different markers is of far less importance than the conservation and management inferences derived from them, so we decided to focus on the latter in our manuscript. 

The Introduction felt lengthy and could be more concise: it covers a lot of ground. Is the theme of the manuscript marine connectivity, comparing genetic markers, or both? It bounces between these two and could be condensed. This is then mirrored in the Discussion: use of subheadings based on the two main themes would help keep similar content together.

- Agreed.  As explained above, we feel the comparison of genetic markers is less useful, so we have undertaken a major revision to streamline the introduction and refocus it around the theme of marine connectivity as suggested.  

In summary, I do not think there are any fatal flaws in the study design and the subject matter would be of interest to the readership of this journal and the scientific community at large. The methods, data generation, and interpretation are mostly suitable for publication. However, I found the data analysis lacking and think readers will as well. I would recommend revising the manuscript to provide more insights into connectivity and comparisons between marker types.

- We thank the referee for their assessment and suggestions.  We have incorporated nearly all of them other than the direct comparisons among marker types, as explained above.

Below are more specific comments. 

1. Page 3, lines 123-124: Change to Conklin et al. to be consistent with other citations (e.g., line 116).

-Agreed and corrected in the text.

1. Page 5, lines 164-169: I really appreciate that the research questions are so explicit.

-Thank you for your positive feedback, we appreciate it!

1. Page 6, line 201: Thank you for providing the indigenous names as a cross-reference.

-Thank you for your positive feedback, we appreciate it!

1. Page 9, line 285: The sentence just ends.

-Corrected, thank you.

1. Page 11, line 347: Check figure reference.

-Corrected, thank you.

1. Page 11, Figure 2: A general comment I have is the lack of discussion on the biological significance of these results. Treating statistical difference in p-values from zero as evidence of robust biological patterns is problematic. Figure 2 exemplifies this: the correlations between FST and geographic distance are barely note-worthy. Some of the slopes, such as for Mfla, appear to be heavily influenced by negative FST values, which is below the theoretical limit of FST. Or I could be misinterpreting the figure. Line 349 says there was no IBD among islands or sampling sites, but then what are the points in Figure 2? Are they FST values between islands, which was significant (line 348), or between all sampling sites?

- Our apologies for the confusion and have tried to clarify our pairwise FST  approach in the text. We agree that the correlations between FST  and distance (IBD) are barely note-worthy but are entirely consistent with previous work. We have tried to clarify our reporting to explain that each point in Fig 2 (now Figure S4) represents the mean FST  between pools of individuals collected from a single island across all loci scored.  We have added a specific example to the figure legend to clarify this, and revised the figures to set all negative FST  estimates by the software to 0 before redrawing the plots to remove the influence pointed out by the referee. We have also moved this plot to the supplementary so that it is not distracting from the primary findings of the manuscript.

1. Page 12, Figure 3: This figure is too much. Simplify: what is the message you are trying to get across? The map does not need every piece of information.

- We respectfully disagree that the map does not need all this information, because the message we are trying to get across is the sum of connectivity work that has been done to date. This figure is the simplest summary of all available information from previous work (which is often requested from our local management agencies) and our new findings, so it needs to be included here. The other referee requested an additional simplified site map with sampling locations and sample sizes that has also been added to the supplementary materials.

1. Page 13, Figure 4: I do not like this heatmap. Interpreting the figure means finding an individual cell in the heatmap and finding the corresponding cell in the diagonal, which is difficult. Why not a bar graph for each site showing the proportion of pairwise comparisons that were significant? Avoid the red-green color scheme, it is not colorblind friendly.

- We agree that the heatmap is difficult to read and provides little benefit.  We have created the suggested proportional summary barchart and moved the offending heatmap to the supplementary materials.

1. Page 13, lines 401-403: This is a stretch as no actual cost comparison was provided.

- Any cost comparison would be outdated virtually as soon as published because sequencing and lab costs are so volatile. We now specify that this study used ~70 sequence libraries for pool-seq, which would have required ~3000 sequenced libraries if done by sequencing each individual rather than in pools. We believe that it is entirely defensible and not at all a stretch to assert that sequencing of 70 libraries will always be far cheaper than sequencing 3000.

1. Page 15, lines 466-474: I am confused, was this a novel analysis specific to this study? Why bury it in the Discussion? Describe in the Methods and present in the Results.

- We apologize for the confusion, this is not a novel analysis specific to this study. The comparison comes from Conklin (2024) and we have revised the text here and above to attribute the analysis with proper citation, and better explain the rationale for that comparison.

1. Page 16, lines 520-526: I am confused as to the threshold used to determine “significant” genetic breaks. Lines 276-301 state that significance was determine using exact test or t-tests. But then it says here that the detection threshold for significant population structure was 0.0016. Which was it? The remainder of the sentence implies that all the pairwise comparison below 0.0016 were not significant, presumably following the exact tests. So 0.0016 really is not the threshold of significance, it is just the highest FST observed in a nonsignificant test? Did any statistically significant tests have a FST less than 0.0016? Also, I will again bring up the biological significance of this value. I would argue whether less than 100 effective migrants per generation constitutes a “genetic barrier”.

- We apologize for the confusion about significance versus “threshold” values. For every SNP locus, significance of the FST  value was tested using a Fisher’s exact test as described. This was used to filter out loci with FST  calls we could not be confident in. We were then left with an FST  for each of the thousands of SNP loci. In order to determine if the mean FST  between two sites (pools) was significantly different from zero, a t-test was performed. The significance of this test was used to determine if two sites were differentiated or not (i.e. pairwise FST between the two pools was significantly greater than zero). We did not set any thresholds in our analyses - that was simply the smallest FST  value (0.0016) observed among significantly differentiated sites. We agree that this FST  value does not constitute convincing evidence of a barrier to gene flow. However, as we point out above, these values can be put into context with the simulations performed previously. The point we were trying to make is not that these constitute a genetic barrier, but that there is limited gene flow among sites with significant genetic differentiation. Putting that into context, understanding the effective number of migrants is fewer than 100 per generation matters greatly to our local marine resource managers, because it is hard to count on such limited exchange among islands for replenishment of an overfished population. We appreciate the feedback and have reworded this section to be more explicit and hopefully avoid future confusion.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

I appreciate the effort that went into revising this manuscript and the responses to my original comments. Thank you. The manuscript is improved, with clearer focus and a more robust array of analyses to support the conclusions.

 

I will be upfront: I still have critiques of the manuscript and think it can be improved. However, this is sound science and a worthy of publication. Lots of work goes into revising a manuscript and addressing every tedious comment from reviewers, most of which are not overly consequential. I will provide my additional comments, which you are welcome to address as you see fit. However, I will not require that these comments be addressed for publication and instead recommend accepting the manuscript with minor revisions. I will pass these thoughts along to the editor so they can decide on the next steps.

 

Here are my comments.

 

1. Page 2: The Introduction still feels too long and looking for a cohesive direction. The first paragraph is too generic to set-up the study: begin with something more concrete. The first 2 sentences of the Abstract are great and clearly present the topic and its importance. We do not need a generic literature review of connectivity.
2. Lines 75-76: The audience for this journal is likely more knowledgeable about fish biology than genetics. I think the jargon and theoretical musings can be removed for readability. The particulars of Wright’s Fst may be over the heads of most readers.
3. Line 79: I would replace “score” with “genotype” for the same reason as my previous comment: non-geneticists may be confused what it means to “score” SNPs.
4. Page 3, line 93: When you say, “concordant with multispecies connectivity”, do you mean concordant with Toonen et al.?
5. Page 4, Figure 1: After staring at this, I realized I was being confused by the black 1000 m bathymetric isolines. I thought the black lines implied some level of grouping between sites, which made it confusing to interpret some of the genetics breaks. I would consider removing the isolines so the focus can be the sampling sites and the location of the genetic breaks.
6. Lines 131-132: This is satisfactory overview of RADseq, but I do not feel like an equivalent description is given to pool-seq (line 141). Assume the audience are not geneticists and provide a brief description of the approach and why you chose it.
7. Lines 135-140: The way the research questions were posed in the original version was superior. The question spanning lines 138-141 is particularly long and confusing.
8. Page 7, line 188: Space needed between Hawaiian and species.
9. Page 8, lines 240-241: Are in-text citations needed for this software? Other programs mentioned include their citation.
10. Page 9, lines 278-280: Why not group collection sites by island in the AMOVA to see the extent to which genetic variation is partitioned by island? That is a core aspect of this manuscript, yet the hierarchical levels used in the AMOVA are regional.
11. Line 288: I am glad to see this, but the figure in the supplemental (Figure S4) still has negative Fst values.
12. Page 10, lines 318-326: I appreciated this walk-through of the results and how it related to sample size.
13. Line 327: Can you actually provide a number? That way readers do not have to figure out how many pools there are. Also, can you provide the BIC scores in the supplemental?
14. Page 12, line 353: I still think this figure is too busy and confusing, but not a hill I will die on. It did dawn on my: I wonder if it would be possible to flip inferences of connectivity. Rather than presenting all comparisons between islands, why not just show the areas where you did find connectivity? Perhaps draw a line showing connectivity between islands/sites. We know from the text that most of the pairwise comparisons are significantly different (page 10), so they dominant the figure. Much more biologically interesting are the sites where there is connectivity. That would more appealing visually and reduce the busyness. I also think a second map with the genetic clusters identified by the K-means clustering would be insight as to whether the two approaches are finding the same patterns.
15. Page 12, lines 367-374: The Introduction mostly dealt with the concept of measuring marine connectivity using genetic markers, so for the Discussion to immediately jump into implications for conservation networks was a bit of a whiplash.
16. Page 13, lines 389-391: I still disagree that this was thorough investigation of the trade-off between marker types. Without any direct comparison of costs or statistical power, this feels like an exaggeration. I think the remainder of the paragraph is fair and reasonable given the nature of this study.
17. Line 397: “Consistent”.
18. Line 405-406: The second half of this sentence (after the “and”) reads like an incomplete thought.
19. Line 416: Assume non-geneticists are reading and wondering what is COI and what type of marker is it?
20. Page 14, line 455: Insert an “of” between “most” and “the”.
21. Line 461; Compared to “our genetic results”? Do you mean Conklin compared simulations to the results of this study? Does this mean these results have been published? I am assuming this is the 2024 dissertation (number 113) that is being cited, not the 2024 Conklin et al. manuscript in review (number 50).

Author Response

I appreciate the effort that went into revising this manuscript and the responses to my original comments. Thank you. The manuscript is improved, with clearer focus and a more robust array of analyses to support the conclusions.

-Thank you very much for your feedback and we agree that your constructive criticism and suggestions have led to a better manuscript. 

I will be upfront: I still have critiques of the manuscript and think it can be improved. However, this is sound science and a worthy of publication. Lots of work goes into revising a manuscript and addressing every tedious comment from reviewers, most of which are not overly consequential. I will provide my additional comments, which you are welcome to address as you see fit. However, I will not require that these comments be addressed for publication and instead recommend accepting the manuscript with minor revisions. I will pass these thoughts along to the editor so they can decide on the next steps.

-Thank you for your feedback and support of this manuscript. We appreciate and admire your honest view of the peer review process and respect your decision to move forward with what you recommend us improve, without the requirement of incorporating everything. We have incorporated your comments and suggestions into this revision and appreciate your kind words. 

Here are my comments.

Page 2: The Introduction still feels too long and looking for a cohesive direction. The first paragraph is too generic to set-up the study: begin with something more concrete. The first 2 sentences of the Abstract are great and clearly present the topic and its importance. We do not need a generic literature review of connectivity.

-We have revised this section to better communicate our point. The referee suggests throughout that we need to better introduce the concepts for a general audience, and the importance of studying connectivity is exactly the set up for this study.  

1. Lines 75-76: The audience for this journal is likely more knowledgeable about fish biology than genetics. I think the jargon and theoretical musings can be removed for readability. The particulars of Wright’s Fst may be over the heads of most readers.

-Agreed, we have revised this section to improve readability and moved the relevant bit to the discussion instead.

1. Line 79: I would replace “score” with “genotype” for the same reason as my previous comment: non-geneticists may be confused what it means to “score” SNPs.

-We agree and have changed the text and the abstract to “genotyped” rather than “scored.”

1. Page 3, line 93: When you say, “concordant with multispecies connectivity”, do you mean concordant with Toonen et al.?

-Yes and we have clarified this in the text.

1. Page 4, Figure 1: After staring at this, I realized I was being confused by the black 1000 m bathymetric isolines. I thought the black lines implied some level of grouping between sites, which made it confusing to interpret some of the genetics breaks. I would consider removing the isolines so the focus can be the sampling sites and the location of the genetic breaks.

-We believe that the bathymetric isoline is helpful to orient readers who may not be familiar with the HI archipelago. Without them it may be easy to assume every island is isolated by a deep channel, which is not the case for clusters of islands such as the Maui Nui island complex. Thus we have instead clarified the legend to ensure readers are not confused and highlighted the importance of understanding the isolines rather than removing them. 

1. Lines 131-132: This is satisfactory overview of RADseq, but I do not feel like an equivalent description is given to pool-seq (line 141). Assume the audience are not geneticists and provide a brief description of the approach and why you chose it.

-We have now added a similar overview here when introducing pool-seq. 

1. Lines 135-140: The way the research questions were posed in the original version was superior. The question spanning lines 138-141 is particularly long and confusing.

-Thank you for your feedback - we have returned to the original wording for clarity

1. Page 7, line 188: Space needed between Hawaiian and species.

-Thank you for catching this mistake. It has been corrected.

1. Page 8, lines 240-241: Are in-text citations needed for this software? Other programs mentioned include their citation.

-Thank you, we have added the appropriate citations for the software used

1. Page 9, lines 278-280: Why not group collection sites by island in the AMOVA to see the extent to which genetic variation is partitioned by island? That is a core aspect of this manuscript, yet the hierarchical levels used in the AMOVA are regional.

-The proposed analysis is simply not possible. AMOVA partitions genetic variance among vs within groups, which is impossible when n=1 for every group in the analyses. Further, the hierarchical levels in the AMOVA are not only regional - we already compare the biogeographic regions to k-means clustering in Table 6. The “diffN” analysis includes literally a single pair of sites in each case (n-1 clusters), which is the minimum amount of clustering possible for an AMOVA analysis. This diffN clustering in Table 6 is therefore as close as is possible to the island-by-island analysis requested, and in all but one case (P. lobata) has the greatest amount of support in the AMOVA. 

1. Line 288: I am glad to see this, but the figure in the supplemental (Figure S4) still has negative Fst values.

-Thank you for catching this - we generated a zero-corrected plot to compare with the one including negative FST values and then inserted the incorrect plot. Figure S4 now has the zero-corrected values.

1. Page 10, lines 318-326: I appreciated this walk-through of the results and how it related to sample size.

-Thank you very much!

1. Line 327: Can you actually provide a number? That way readers do not have to figure out how many pools there are. Also, can you provide the BIC scores in the supplemental?

-The number of clusters varied per species since all species did not include samples from every island. For instance, C. strigosus included samples from 12 islands with one pool per island. K-means clustering found the best fit of 10 clusters (#pools - 2). Table 6 includes this information in one convenient location

1. Page 12, line 353: I still think this figure is too busy and confusing, but not a hill I will die on. It did dawn on my: I wonder if it would be possible to flip inferences of connectivity. Rather than presenting all comparisons between islands, why not just show the areas where you did find connectivity? Perhaps draw a line showing connectivity between islands/sites. We know from the text that most of the pairwise comparisons are significantly different (page 10), so they dominant the figure. Much more biologically interesting are the sites where there is connectivity. That would more appealing visually and reduce the busyness. I also think a second map with the genetic clusters identified by the K-means clustering would be insight as to whether the two approaches are finding the same patterns.

-This is essentially what we have done by adding the k-means clustering, because the diffN analysis for every species converges on the optimal number of clusters being n-1 (a single pair of populations and all others distinct), which is the minimum grouping required to complete an AMOVA analysis. This pattern does not provide evidence of connectivity, it is a limitation of the analysis, and this information is presented in Table 6.

1. Page 12, lines 367-374: The Introduction mostly dealt with the concept of measuring marine connectivity using genetic markers, so for the Discussion to immediately jump into implications for conservation networks was a bit of a whiplash.

-Thank you for the feedback. We have now revised both the Introduction and Discussion again to improve readability and flow.

1. Page 13, lines 389-391: I still disagree that this was thorough investigation of the trade-off between marker types. Without any direct comparison of costs or statistical power, this feels like an exaggeration. I think the remainder of the paragraph is fair and reasonable given the nature of this study.

-We have toned down the language in the offending sentence. 

1. Line 397: “Consistent”.

-Corrected, thank you

1. Line 405-406: The second half of this sentence (after the “and”) reads like an incomplete thought.

-Revised as suggested.

1. Line 416: Assume non-geneticists are reading and wondering what is COI and what type of marker is it?

-We have clarified that we are referring to the “mitochondrial cytochrome c oxidase subunit I (COI) gene” here.

1. Page 14, line 455: Insert an “of” between “most” and “the”.

-Thank you for the correction!

1. Line 461; Compared to “our genetic results”? Do you mean Conklin compared simulations to the results of this study? Does this mean these results have been published? I am assuming this is the 2024 dissertation (number 113) that is being cited, not the 2024 Conklin et al. manuscript in review (number 50).

-While these results have not been published, Conklin compared our genetic dataset to her biophysical models generated for her dissertation as an opportunistic comparison of the same system. We have removed the paper in review, and added her dissertation citation to both sentences to avoid any uncertainty.