Incorporating Topological and Age Uncertainty into Event-Based Biogeography of Sand Spiders Supports Paleo-Islands in Galapagos and Ancient Connections among Neotropical Dry Forests

Round 1

Reviewer 1 Report

In the attachment.

Comments for author File: Comments.docx

Author Response

We thank the reviewer for the thoughtful comments.

I would suggest that the authors further develop their approach, maybe by a more detailed investigation of the taxon sampling (incomplete sampling) uncertainty, which is highly problematic for biogeographic reconstructions of some groups of organisms.

R: This is a valuable comment, and the issue of missing taxa is important, but in our view it would merit a paper devoted to this topic alone. On the other hand, our approach allows to incorporate missing taxa, because it takes uncertainty into account; researchers could add those missing taxa to their trees randomly or using prior information, and them perform biogeographic estimates on those trees. 

The only thing I miss might be the absence of the discussion on any potential downsides of their approach. However, it isn't easy to really see one.

R: Potential downsides are those relative to the DEC model (e.g. underestimation of extinction), but these are widely known and mentioned in the text. We have now added some text to discuss alternatives (e.g. joint estimation of phylogeny and biogeographic history as suggested by reviewer 3).

Reviewer 2 Report

Review of diversity-1323602 titled “Incorporating topological and age uncertainty into event-based biogeography of sand spiders supports paleo-islands in Galapagos and ancient connections among Neotropical dry forests”.

 

The manuscript presents an effort to account for some of the important biases in reconstructions of biogeographic histories, the uncertainty in clade ages and topology. The authors address these biases by estimating ancestral ranges using a number of trees from the posterior distribution of a previously published Bayesian analysis and by using stochastic maps,

 

I have to point out that I am not an expert in methods in biogeography. Thus, I can only judge the manuscript from a more superficial point of view of a general reader.

 

The entire manuscript reads very nicely and takes the reader, even if unfamiliar with some specifics, through all necessary details. I have no meaningful comments that could improve the manuscript, but have enjoyed reading it and perceive the work as an important contribution to biogeography. Also, I urge the editors to get an opinion from at least another reviewer versed in biogeographical methodology.

Author Response

We thank the reviewer for the thoughtful comments!

Reviewer 3 Report

In the present ms Magalhaes et al. study the biogeographic history of a genus of spiders (Sicarius) that is found in the neotropics. They develop an analytical approach that allows to incorporate phylogenetic uncertainty in the biogeographic reconstructions and using this approach they find support for the role of islands that are nowadays submerged for the colonization of Galapagos by these spiders. Their results also suggest that in the past Neotropical dry forests have been interconnected. Large part of the ms centers on the problems of phylogenetic uncertainty and their effects on biogeographic reconstructions. This is certainly an interesting and relevant topic and has been recently widely discussed (along with other potential sources of uncertainty such as – uncertainty in age estimates, fossils placemats, fossil distributions etc.).

 

However, I cannot recommend publication of the ms in its current shape as there is a major flow that needs to be addressed. Given that focus of the ms it is very surprising to see that some recent literature on the topic has not been cited and discussed. In fact, at least two recent papers that have developed methods to handle various sources of uncertainty in biogeographical analyses come to mind (Yan et al., 2021 and Landis at al., 2020). Both the Yan et al and the Landis at al approaches go even further and consider, in addition to phylogenetic uncertainty, uncertainty in fossil placements, age estimates etc. The Yan et al approach is an empirical Bayesian method while Landis at al approach is implanted in RevBayes and is a fully Bayesian framework. Therefore, given the heavy focus on methodology in the ms one would expect to see some comparison between the method proposed here and these two alternatives and some justification of why the method that the authors propose is a better choice for their analyses. At least a priory I cannot see a good argument for not using the Landis at all approach given that the present dataset is rather small. See Yan et al for comparison of the performance of both methods and suggestions on when to use either one. Thus, if the authors want to keep the focus as it is they must rework the ms referring to the relevant recent literature and they must eventually provide arguments for the use of their methods (as well as compare its performance with other alternatives). Alternatively, much of the methodological discussion can be removed shifting the focus to the focal taxon and its biogeographical history reconstructed using the method of Landis et al for example.

As addressing this issue will require a major rewriting of the ms and additional analyses I am not providing specific comments on the current version here.

I would also suggest that if the authors decide to rework the ms they pay closer attention the English. I found several places where it could be further improved.

One other think that the authors may want to consider is some additional analyses to test the sensitivity of the results the set of settings used in the biogeographical analyses. Currently very limited of alternative settings has been evaluated.

 

References:

Yujing Yan, Charles C Davis, Dimitar Dimitrov, Zhiheng Wang, Carsten Rahbek, Michael Krabbe Borregaard, Phytogeographic History of the Tea Family Inferred Through High-Resolution Phylogeny and Fossils, Systematic Biology, 2021;, syab042, https://doi.org/10.1093/sysbio/syab042

Michael J Landis, Deren A R Eaton, Wendy L Clement, Brian Park, Elizabeth L Spriggs, Patrick W Sweeney, Erika J Edwards, Michael J Donoghue, Joint Phylogenetic Estimation of Geographic Movements and Biome Shifts during the Global Diversification of Viburnum, Systematic Biology, Volume 70, Issue 1, January 2021, Pages 67–85, https://doi.org/10.1093/sysbio/syaa027

 

Author Response

In the present ms Magalhaes et al. study the biogeographic history of a genus of spiders (Sicarius) that is found in the neotropics. They develop an analytical approach that allows to incorporate phylogenetic uncertainty in the biogeographic reconstructions and using this approach they find support for the role of islands that are nowadays submerged for the colonization of Galapagos by these spiders. Their results also suggest that in the past Neotropical dry forests have been interconnected. Large part of the ms centers on the problems of phylogenetic uncertainty and their effects on biogeographic reconstructions. This is certainly an interesting and relevant topic and has been recently widely discussed (along with other potential sources of uncertainty such as – uncertainty in age estimates, fossils placemats, fossil distributions etc.).

R: Thank you for your positive feedback.

However, I cannot recommend publication of the ms in its current shape as there is a major flow that needs to be addressed. Given that focus of the ms it is very surprising to see that some recent literature on the topic has not been cited and discussed. In fact, at least two recent papers that have developed methods to handle various sources of uncertainty in biogeographical analyses come to mind (Yan et al., 2021 and Landis at al., 2020). Both the Yan et al and the Landis at al approaches go even further and consider, in addition to phylogenetic uncertainty, uncertainty in fossil placements, age estimates etc. The Yan et al approach is an empirical Bayesian method while Landis at al approach is implanted in RevBayes and is a fully Bayesian framework.

R: Thank you for pointing these recent papers, which we missed during our initial literature survey. Both have been included in the introduction and discussion as examples of previous efforts in this direction.

Therefore, given the heavy focus on methodology in the ms one would expect to see some comparison between the method proposed here and these two alternatives and some justification of why the method that the authors propose is a better choice for their analyses.

R: We believe there is a little misunderstanding regarding the aims of our paper. We do not aim at proposing a novel method, we simply advocate that uncertainty should be incorporated in biogeographic analyses by using both multiple trees and stochastic maps – and we had even acknowledged that other papers had recently done this before with positive outcomes (e.g. Matos-Maraví et al 2021, and now Yan et al. in press). Our approach is not alternative to the ones presented in the papers mentioned by the reviewer; it is, in fact, essentially identical to the one used by Yan et al. in press, although we went further, by testing model fit for every one of the multiple trees and drew some conclusions from that. We do focus on methodology, but only to the extent in which it may affect our biological conclusions regarding the biogeographic history of sand spiders. We do provide a novel script, for the sake of facilitating the replication of analyses in subsequent studies. We believe that providing such script will be useful for others, as well as increase the citation of this study.

At least a priory I cannot see a good argument for not using the Landis at all approach given that the present dataset is rather small. See Yan et al for comparison of the performance of both methods and suggestions on when to use either one.

R: Yan et al. (in press) concluded that “both methods produced congruent results under the M1 time-stratified dispersal scenario, but differed slightly under the M0 dispersal scenario”. So, both methods yielded identical or nearly identical outcomes in this case, and thus it seems we have little to gain from applying Landis’ approach to our own dataset.

Thus, if the authors want to keep the focus as it is they must rework the ms referring to the relevant recent literature and they must eventually provide arguments for the use of their methods (as well as compare its performance with other alternatives). Alternatively, much of the methodological discussion can be removed shifting the focus to the focal taxon and its biogeographical history reconstructed using the method of Landis et al for example.

R: The reviewer suggests that we apply Landis’s method from scratch, either to compare it with our approach, or to replace it altogether. The first option seems to us like a different study designed specifically to compare both approaches, which is not our aim here; furthermore, this has already been done by Yan et al. (in press) who showed they give similar results. Thus, we believe that such a study will be short in originality. The second option implies a priori that Landis’ approach is superior, which needs to be justified, especially in the light of the conclusions of Yan et al. (in press) reproduced above, indicating it does not produce very different results. Furthermore, Landis’ approach requires re-estimation of the phylogenetic tree using specific software (RevBayes). Our approach, in contrast, can be used with trees estimated with any software that produces dated phylogenetic trees (BEAST, IQtree, ape’s chronos, treePL, etc.) and is thus more flexible and may be useful for researchers in a situation similar to ours (where a sample of dated trees was already available) or when the dataset is too large for Landis’ approach.

As addressing this issue will require a major rewriting of the ms and additional analyses I am not providing specific comments on the current version here.

I would also suggest that if the authors decide to rework the ms they pay closer attention the English. I found several places where it could be further improved.

R: We have revised the manuscript once again for mistakes. We will be happy to revise it once again provided the areas of improvement are pointed to us.

One other think that the authors may want to consider is some additional analyses to test the sensitivity of the results the set of settings used in the biogeographical analyses. Currently very limited of alternative settings has been evaluated.

R: We believe we reached the limit of reasonable models and parameters to test. We have compared not only multiple time-stratification scenarios, but also different biogeographic models (DIVA-like, DEC, BAYAREA-like and their +J variants). We justified other parameters (e.g. maximum range size) using explicit biological justifications and see no reason to explore other values for these parameters since the ones we used are biologically reasonable. We used a simple dispersal matrix that allows for equal dispersal probabilities among areas, independent of distances; however, it is not straightforward to propose alternative dispersal matrices, since the biomes might have changed their extensions through time and, in some cases, distances do not reflect dispersal probabilities (e.g. Atacama and Monte are in very close geographical proximity but separated by a formidable barrier to dispersal, the Andes). We are unaware of other “alternative settings” that could be explored to yield different results.  We thus believe the sensitivity of the results is well-explored.

References:

Yujing Yan, Charles C Davis, Dimitar Dimitrov, Zhiheng Wang, Carsten Rahbek, Michael Krabbe Borregaard, Phytogeographic History of the Tea Family Inferred Through High-Resolution Phylogeny and Fossils, Systematic Biology, in press;, syab042, https://doi.org/10.1093/sysbio/syab042

Michael J Landis, Deren A R Eaton, Wendy L Clement, Brian Park, Elizabeth L Spriggs, Patrick W Sweeney, Erika J Edwards, Michael J Donoghue, Joint Phylogenetic Estimation of Geographic Movements and Biome Shifts during the Global Diversification of Viburnum, Systematic Biology, Volume 70, Issue 1, January 2021, Pages 67–85, https://doi.org/10.1093/sysbio/syaa027