Central American mountains inhibit eastern North Pacific seasonal tropical cyclone activity

The eastern North Pacific (ENP) has the highest density of tropical cyclones (TCs) on earth, and yet the controls on TCs, from individual events to seasonal totals, remain poorly understood. One effect that has not been fully considered is the unique geography of the Central American mountains. Although observational studies suggest these mountains can readily fuel individual TCs through dynamical processes, here we show that these mountains indeed play the opposite role on the seasonal timescale, hindering seasonal ENP TC activity by up to 35%. We found that these mountains significantly interrupt the abundant moisture transport from the Caribbean Sea to the ENP, limiting deep convection over the open ocean area where TCs preferentially occur. This study advances our fundamental understanding of ENP TC genesis mechanisms across the weather-to-climate timescales, and also highlights the importance of topography representation in improving the ENP regional climate simulations, as well as TC seasonal predictions and future projections.

This paper examines the impact of Central American mountains on tropical cyclone activity, using a suite of simulations with a regional weather model. The authors find that Central American mountains reduce tropical cyclone activity by blocking moisture transport from the Atlantic, which in turn reduces convective activity and mid-tropospheric relative humidity, a key driver of tropical cyclone genesis. Analogous results are found when closing the gaps in the Central American mountains. The results are shown to be robust via simulations at 9 km and 27 km resolutions, and supporting analysis of background climate fields.
Overall, I find this an impressively well-thought-out and interesting study. I think it contributes significantly to understanding of the controls on tropical cyclone activity in the eastern North Pacific, and importantly helps clarify the differences in the role of this topography at weather vs climate time scales, which has been a confusing point in the literature. Additionally, the paper is for the most part quite clearly written, with helpful figures. I think the general contents of this paper are very appropriate for Nature Communications.
However, I have a few larger critiques/questions which I would like to see addressed listed immediately below, and then line by line suggestions for how to improve the paper listed farther below.
1) The work is motivated by talking about how the ENP tropical cyclone genesis density is the highest on the planet. However, the conclusion is that mountains actually decrease TC activity in this region. If mountains are not driving the anomalously high TC genesis in this region, what do you think is and how can the community move forward in understanding this region? I suggest exploring this question a bit in the discussion. 2) I think it should be clarified earlier in the text that the models used are not atmosphereocean coupled, and that this is a limitation of this study. For someone unfamiliar with the particular version of WHARF that might not be immediately clear. I think this should be added in addition to the existing text in the Discussion/Conclusion on this point. 3) In [183-228] the authors argue that the Atlantic-Pacific integrated water vapor transport is driving the changes in TCs and precipitation. While I agree with this conclusion, and find Figure 4 compelling, I am a bit confused by Figure 3, and the related explanations. I would think the role of this Atlantic-Pacific water vapor transport could be more cleanly diagnosed using a moisture budget decomposition (ala Seager and Vecchi 2010 or Baldwin and Vecchi 2016 links copied below), to decompose the role of changes in humidity versus winds, at monthly mean vs transient time scales. Right now, the plot showing divergent wind does not seem very relevant to me, because it is u*q not u by itself that matters for changes in precipitation/ convection. To me, the plots showing changes in buoyancy and convection are helpful, but still not fully addressing the question of where the moisture is coming from, as far as I understand right now. I would appreciate it if the authors could help me understand their logic a bit more here, to square with how I would normally think about this problem, or try the alternative approach I suggest of a moisture budget decomposition and see if it is helpful.  Fig. 13 and 14) feel like they are the beginnings of an entirely different paper. I understand the connection to the current work, and find the results of this analysis compelling, but I'm wondering if it would be better for the scientific community if this was a separate paper. I also have mixed feelings about the current structure of introducing these new results in the final paragraph of the paper. If you definitely want to include them, could you introduce them in the results section, and then discuss them further in the Discussion section? 5) In [234-235] you assert that "the anomalous deep convection and heating also modulate the occurrence of the easterly waves". While you show the changes in easterly waves, I am not sure how you can conclude that this is due to the change in deep convection and heating. Can you please back this up with either reference to relevant literature or some further explanation? 6) [362-374] I find this description of how the gaps are closed rather confusing. First, what is meant by "At each meridional grid within the box"? Second, does a random factor of order 1 mean between 1 and 9? Finally, can you explain more clearly why you use this somewhat complicated method with some randomization, as it's not immediately intuitive to me? Finally, is the method (including the Gaussian filter of 5 zonal and 3 meridional gridcells) same for the 9 km simulations? 7) [517-521] The authors say they will make the code used in this study available upon request. I want to ensure that this is consistent with Nat Comms data policies-it would be best if this could be put on github or the like to be openly accessible.
Line by line comments: 1) [13] suggest to say "tropical cyclones (TCs) on earth" 2) [18] suggest to remove "," after "show" 3) [23] "mechanism" should be "mechanisms" 4) [32] suggest to remove "the" before "TC activity density" 5) [47] suggest to move citations 12,13 to end of the sentence 6) [54] should be "take into consideration the unique" 7) [55-56] suggest to write just "Sierra Madre in North America" as I've seen varying interpretations of exactly what "Sierra Madre Occidental" refers to 8) [56] suggest to say "~1km high" 9) [57] suggest to replace "marked" with "interrupted" or "split-up" 10) [63] suggest to remove "-" between "mountain" and "gaps" 11) [82] should be "model"  [368] suggest to change to "a size of 5 zonal and 3 meridional gridcells and a standard deviation…" 41) [396] "purposed" should be "proposed" 42) [400] suggest to remove "the" before "climate change" 43) [402] suggest to add "the" before "vertical and horizontal wind speeds" 44) [409] remove "that" before "determined by" 45) [416] shouldn't this be "determined from" not "defined by"? since you provide the actual definition on the next page 46) [425] should have "is" before "determined" 47) [433] based on the prior line, shouldn't there be a factor of 2 in front of (h10m-h*env)? 48) [440] "diabetic" should be "diabatic" 49) [448] suggest to write "; as such, we need" 50) [449] add "such" before TCs 51) [452] do you mean "that are adapted"? 52) [456] add "the" before "95%" 53) [492,498] "smoothened" should be "smoothed" 54) [501] suggest to change to "similar to most subseasonal-to-seasonal prediction models; as such, the results" 55) [Fig 1 boxplots] What lat-lon region is the averaging performed over? Please specify in caption. 56) [678] here and elsewhere you write "denoated". It should be "denoted" 57) [679] "experiment" should be "experiments" 58) [683,699] suggest to add "the" before "two-sided t-test" 59) [ (14) which is more than three times than the total figures included in the main manuscript text (4). Moreover, these supplemental figures are referred to in both the results and methodology of this manuscript and seem more integral to the manuscript that just being included as supplemental material. This reviewer would like to see substantial revision of the manuscript, incorporating some of the more integral supplemental figures provided into the primary manuscript. There are also a few extraneous topics included in the methodology that seem only tangential to the main them of the manuscript (for example: how easterly waves and precipitation are affected by differences in topography) that this reviewer would prefer to see incorporated into an additional manuscript.
Therefore I am recommending major revisions to this manuscript before I can consider this manuscript suitable for publication. My comments sorted by major and specific are below:

Major Comments:
1) There is a lot of extraneous material in the methodology section that does not seem applicable to the manuscript theme which focuses on how Central American mountains can impact TC activity in the ENP. While the first four sections of the methods (Highresolution climate model simulations, Experiment designs, Modified genesis potential index, Buoyancy diagnostics) are all relevant to the manuscript, the final two sections (Easterly wave tracking and CMIP6 Precipitation; Lines 439-504) are too far removed from the manuscript theme to add overall value to this manuscript and seem like unnecessary extra material.
While it is true that easterly waves and precipitation are both meteorological phenomena that can influence ENP TC activity, the detailed description of the methods and results presented in the supplemental figures go beyond the scope of what this manuscript was inferred to discuss in the introduction (line 73-77).
The recommendation of this reviewer would be to remove this extraneous material that is already supplemental as organized by the authors. Specifically lines 439-504 and supplementary figures 12-14. These results and figures are high quality but would be more suitable for a subsequent manuscript.
2) There are a number of instances where this reviewer would have preferred to structurally include the supplemental figures as regular figures in the text. For example, Supplemental figure 1 is a nice introductory figure that is the only figure discussed in the introduction. Why it is considered supplementary when it provides important background information provided in lines 52-67? There are a number of instances where it would have been preferable to include supplemental figures as regular figures in the manuscript and this will be occasionally described in the specific comments below.

Specific Comments:
L43: Instead of "tightly" use "closely" instead? L108-117: This reviewer very much appreciates that the 27-km simulation results are being compared to the 9-km simulations described in this portion of the results. But I would have preferred to see more details comparing the two simulations in the following section (lines 146-153) especially given the knowledge higher resolution would be more likely to capture important mesoscale circulations produced by the terrain gaps and have better resolution to capture TC circulation and wind field aspects not touched on in this manuscript. L180 and in other places: "state-of-art" is superfluous.

Point-to-Point Reply to Referee #1
First, we would like to thank the referee for the invaluable comments and help us improve the manuscript. We have carefully followed each of your comments listed in red and revised our paper accordingly. Our replies to your comments are as follows: Referee #1: This paper was previously submitted to Nature Geoscience and I served as a reviewer on that submission. The manuscript is largely unchanged from that submission, so most of my comments from my previous review still stand. Some of the minor edits did help clarify some points, though.
Overall, I feel that this paper is very well written. However, I still feel that the most important result of this study is that climate models used to simulate future TC activity have low resolution with too heavily smoothed topography. Thus, they would not accurately represent the TC activity in the ENP, which also has downstream impacts. This point should be emphasized earlier and not just in the final paragraph of the conclusion. Additional discussion of the heavily smoothed terrain in these climate models as part of the reason the authors chose to simulate the terrain in different ways for their study is needed.
Reply: Thank you for taking the time to serve as a referee of our work for the second time. Your constructive comments and positive feedbacks are highly appreciated. We agree with you that this paper may have important implications on topography representation used in climate models, as such; we add some discussions in the Introduction. Please see lines 72-82. However, as criticized by Referee #2 and #3 about manuscript structure, in the revision, we removed the topic of how smoothened topography could influence precipitation in the CMIP6 (Supplementary Figure 13 and 14 of the initial submission). We will expand the results presented in Supplementary   Figure 13 and 14 and make more comprehensive analyses in a separate future study.
Referee #1: Instead, the paper focuses on how the presence of the Central American mountains actually inhibits TC activity on seasonal to climate timescales by blocking moisture transport into the ENP. While this is a very interesting finding and is helpful for gaining a better understanding of the role the mountains play in ENP TC activity, even on synoptic timescales, I think that the implications for the climate models are more impactful. There is also no discussion on whether the climatological values of moisture in the ENP are sufficient without any additional moisture transport, to support cyclogenesis.
Reply: We agree this paper has significant implications for climate models' the ENP are sufficient without any additional moisture transport, to support cyclogenesis", as can be seen from Figure 1 in the revision, the strong moisture transport from Atlantic to the ENP is persistent on the seasonal timescale, thus it's difficult to consider a scenario that ENP climatological moisture is not influence by this moisture transport from Atlantic. Comparing our CTL with ERA5 reanalysis, ENP 700hPa relative humidity is also reasonably simulated, and the climatological value is comparable to those in the western North Pacific warm pool region. Note that, western North Pacific has the strongest TC activity over the globe that primarily attributed to the warm sea surface temperature and abundant moisture supply. All these results manifest that CTL simulated moisture in the ENP is faithful and the climatological values are sufficient for TC genesis.   We would like to thank the Referee for taking the time to review our manuscript and offering helpful suggestions to improve it. We are particularly grateful for your suggestions on the moisture budget decomposition, which led to a very nice addition to the manuscript. We appreciate your constructive comments and have carefully considered each of these comments listed in red and revised our paper accordingly.
Our replies to your comments are as follows: Referee #2: Overall, I find this an impressively well-thought-out and interesting study.   In CTL, it clearly shows that precipitation minus evaporation (P-E) is primarily controlled by the mean flow (monthly mean) moisture flux convergence, while the moisture flux convergence by transient eddy plays a minor but negative role, which indicates that high-frequency synoptic variability tends to transport the moisture from equator to mid-latitude in the ENP. Surface term largely induces positive moisture flux convergence, but its contribution to the total column integrated moisture is negligibly small.    Figure R2-3m,n). However, we note that the significant increase (decrease) of flow convergence only confines to the narrow area in the eastern portion of ENP close to the coastline ( Figure R2-3b,h,n), whereas the increase (decrease) of mositure flux (qu) convergence occurs at much broader area ( Figure R2-2a,d). It suggests that, in NMT (NGP), the increased (decreased) moisture flux from the Carribbean (please see Figure 1) enhances (reduces) the ENP moisture content ( Figure R2-3e,k,q), also the buoyancy, primarily due to stronger (weaker) moisture flow convergence, and thus increase (decrease) moisture convection there ( Figure   R2  Referee #2: [517-521] The authors say they will make the code used in this study available upon request. I want to ensure that this is consistent with Nat Comms data policies-it would be best if this could be put on github or the like to be openly accessible.
Reply: As the simulations produce more than 100TB of output files, we currently don't have appropriate online storage like GitHub to directly share the model outputs and make it public available. However, we uploaded the modified topography data  [140] should be "mountains as well" Reply: Done.
[142-143] Does the removal of these jets also explain the nmt decrease in TC activity in this region?

Panama
Reply: In NMT, since we removed the mountains, the entire regional circulation (not just the gap wind jets) was changed (please see the Figure R2-3g,m). We think the similar mechanism to NGP also applies to this case, that is, the dynamical role of moisture transport dominates the TC activity changes along the ENP coastline.
[154] You say "mountains can fuel TC genesis". Is this backed up by the literature?
Or would it be more accurate to say "mountain gaps can fuel TC genesis"?
[157] suggest to replace "it provides" with "there is" Reply: Done.
[179] suggest to move second apostrophe around the two pseudo phrases before the parentheses.
Reply: The change has been made.
[188] suggest to replace "which is directly resulting" with "which directly results" Reply: Done.
[230] I think this would read clearer if you said "Removal of the mountains' blocking effect" Reply: Done.
[230] "diabetic" should be "diabatic" (here and anywhere else-I think I saw this one other place) Reply: We apologize for this typo. Two changes have been made.
[304] What region is the nest performed over? Can you specify that here for clarity?
Reply: The nested domain coverage is now shown in Supplementary Figure 1a, which roughly covers the area [178°W-65°W, 2°S-36°N]. We have clarified this in the revised manuscript.
[311] suggest to remove "We specially note that" Reply: Done.
[313] What is meant by "convexity" here? I'm not used to seeing this term describing topography.
Reply: Topographic convexity measures how convex or sharp the subgrid-scale topography is by statically relating the characteristics of the mountain waves to the subgrid-scale topography. It is an important parameter in the WRF topography-induced gravity wave drag parameterization.
[318] Can you briefly say how these relative vorticity thresholds are chosen?
[338] suggest to add "with observations" after "precipitation" Reply: Done.  [345-346] suggest to change to "This bias in South American orographic precipitation" Reply: Done.
Reply: As the model simulation consists of a total 174 TC seasons while IBTrACS observation only has 29 TC seasons, the maximum value of simulated ensemble mean seasonal TC genesis density is consequently smaller than that in the observation, as ensemble mean tends to average out extreme values. Here we want to emphasize that the spatial pattern, rather than the amplitude, of TC genesis density shows agreement between CTL and IBTrACS observation, especially in the area between 120°W and 100°W. In the Figure 3 of revised manuscript, we labeled contour plots level for better clarity.
[719] should be "Sensitivity" Reply: Thanks for pointing this.
[724] should be "quadrent" Reply: Thanks for pointing this, we corrected it to "quadrant".
[733] "minus" should be "subtracting" Reply: Done. Reply: As the seasonal TC activity in the ENP is not only controlled by the IVT following a linear relationship, the linear regression coefficient between the IVT and ENP TC days is statistically less significant than that shown in Figure 6a  By subtracting the NMT/NGP with the corresponding season CTL 6-member ensemble mean to calculate the anomalies, as shown in Figure 6a, some of the influence from these modes of climate variability can be reduced, making the results cleaner for the influence of IVT. In addition, we concatenated the NMT and NGP anomalies to double the sample size when we determined the statistical confidence level. Reply: Yes, thanks for pointing this.
[ Fig S10] is "changes" in initial bolded line a mistake?
Reply: We have added "TC activity" before "changes" for clarity. Now it is Supplementary Figure 8 in the revised manuscript.
[ Fig S11] should be "distribution" Reply: The change has been made.
[ Fig S12] should be "Easterly waves" Reply: Done. Note that we replace this figure with new Supplementary Figure 10 in the revised manuscript to better clarify the easterly waves changes in the ENP are primarily attributed to those easterly waves generated locally in the ENP, rather than transported from the Atlantic.
[ Fig S13] How do you construct the high res and low res topography fields in panels gand h? Do you average the topography across models?
Reply: Yes, they are averaged from the available outputs from the CMIP6. Note that Supplementary Figure 13 was removed in the revised manuscript based on two Referees' comments, as this part of analysis is less related to the current study.
[ Fig S13] should say "purposes, precipitation on various" Reply: Thanks for pointing this. Supplementary Figure 13 was removed in the revised manuscript.
[ Fig S13] should say "Refer to Methods" Reply: Thanks for pointing this. Supplementary Figure 13 was removed in the revised manuscript.
[ Fig S14] What does TCM stand for? Also this is also a June-Nov average too, right?Should maybe specify this.
Reply: "TCM" stands for "tropical channel model", which is the same model that we

Point-to-point Reply to Referee #3
We would like to thank the referee for taking the time to review our manuscript and offering invaluable comments and suggestions to improve the manuscript. We have carefully followed each of your comments listed in red and revised our paper accordingly. Our replies to your comments are as follows: Reply: Thanks for your thoughtful recommendation. Please refer to the reply above.
We think the revised manuscript is now more balanced. L43: Instead of "tightly" use "closely" instead?
Reply: The change has been made.

L41-51: A few additional citations worth including in this portion of the introduction
is the dramatic influence of convectively coupled Kelvin waves which have also been shown to have a significant influence on tropical cyclone development in the ENP.
Reply: Thank you for introducing these great works. We have added those references. Reply: We agree that 9-km simulation is more capable to resolve TC internal dynamics than 27-km and it is very intriguing to disentangle how Central American mountains can influence the ENP TC structures. As noted by Supplementary Figure 9 (also shown below), the probability density functions (PDFs) of TC maximum 10-m wind and minimum sea level pressure do show some differences, while the differences are not homogeneous. For example, in weak-to-moderate TC intensity regime (i.e. > 23 m/s but < 40 m/s or < 1005 hPa but > 970 hPa), the fraction in NMT is apparently more than that in CTL and NGP, but this difference reverses when we focus on strong TC intensity regime (i.e. > 40 m/s or < 970 hPa). However, as these differences in TC intensity is statistically less significant than the seasonal mean TC activity metrics (i.e. TC numbers, TC days) differences, and the main scope of this manuscript is to reveal the opposite role of Central American mountains played in hindering the seasonal ENP TC activity. We do not have enough space to fit the discussion of the TC structural changes to this short paper in Nature Communications.
We    L397: Again state-of-art is superfluous and not necessary here.
Reply: The change has been made.
L439-504: This portion of the methods section seems unnecessarily to the rest of the manuscript. How the CTL simulation depicts easterly waves and precipitation patterns is likely an important discussion but seems beyond the scope of this study that already includes 18 figures (14 supplementary). Recommend cutting this portion of the manuscript as discussed in major comment #1.
Reply: As noted above, we much appreciate your suggestions on the shape of manuscript. The topic of how smoothened topography could influence precipitation in the CMIP6 is removed in the manuscript. But as arises from the comments by Referee #1 and #2, we still remain the easterly wave related analysis. We have reduced the total number of figures to 16 (6 in the main text and 10 in the supplementary) in the revised manuscript.
We would like to thank the referee for taking the time to review our manuscript and offering invaluable comments and suggestions to improve the manuscript. We have carefully followed each of your comments listed in red and revised our paper accordingly. Our replies to your comments are as follows: This is my first review of this manuscript, and I was also asked to step in to assess whether the comments of Reviewer 1 were addressed. This study argues that the mountains of Central America hinder east Pacific tropical cyclogenesis by suppressing the moisture flux into the east Pacific from the Atlantic. This result suggests that models need to properly represent the mountains of Central America in order to represent and predict cyclogenesis in the east Pacific.
This is an interesting and relevant study that deserved publication Nature Communications. I feel that the authors have done an adequate job in responding to previous reviewer comments, especially to Reviewer 1, and I just have some additional minor comments for the authors to consider in revising the paper. I recommend that this paper be published after minor revisions.
Reply: Thank you for your encouraging comments.
Minor Comments: 1) line 102. Stating that three sets of simulations will be conducted, but providing no information on the nature of these simulations until later, is a bit awkward. Possibly a brief introduction can be given here to the differences between the simulations.
Reply: We thank you for these constructive comments. We have followed your suggestion and added a brief introduction about these simulations, please see lines 102-105.
2) lines 113-117. A more faithful discussion of the errors in control precipitation relative to the observed dataset should be conducted. In particular, the control simulation produces too much precipitation to the west of the mountains, and the Reply: By assessing the changes in Genesis Potential Index, we find that the environmental favorability for TCs is increased in the NMT simulation, but decreased in the NGP simulation. To save space, we want to state these opposite-sign changes in one sentence, and thus we parenthesize "less" and "NGP" in the sentence. We agree that the inconsistency between the prescribed SSTs and the simulated wind patterns in the ENP near-coastal area is a limitation of this study. However, we hypothesize its influence on seasonal TC activity may not be significant, at least in terms of seasonal accumulated number of TCs and TC days, due to the following reasons: 1) After using the correct SST climatology, we find that the cold patch near the Costa Rica dome is quite weak in the boreal summer (see Figures below). Along   10°N, the SST difference is less than 0.5°C at the Costa Rica Dome (around 88°W, 10°N), and the SST remains above 27.5°C. This is consistent with Figure 8   Reply: Please see our reply to 5). Based on the June-November seasonal mean SST ( Fig. R4-3), we believe the cold SST in the Costa Rica Dome during June-November is too weak to have a significant influence on TC activities. However, a full understanding of this issue will require future coupled climate model simulations as we discussed in the revision (please see lines 311-316).