Comment on wcd-2021-29

The authors examine the causes for the “significant wave height” (swh) in the western South Atlantic (wSA) region on intra-seasonal timescales. They use EOF analysis and cyclone tracking and find that the different phases of EOF1,2 of u10 are responsible for larger/smaller swh in wSA, as well as that the u10 variability is related to the intraseasonal variability within the Pacific South-American (PSA) pattern, suggesting a remote impact on the wSA atmosphere-ocean variability. Their results are important for further understanding of the causes of swh, which has large societal impacts (especially on intraseasonal timescales), thus I recommend this manuscript for the publication in the Weather and Climate Dynamics, subject to revisions suggested below.


Major comments
There are a few parts that I found a bit confusing and would like to see them clarified before the manuscript is published.
l. 33-38: you seem to talk about SAM and NAO as though they are primarily interannual indices, but they are really oscillations that we consider relevant on weekly/submonthly timescale. They do of course exhibit interannual, decadal, multidecadal variability. I just think that the wording is perhaps confusing, or maybe I am not understating the point here. ENSO of course is interannual and it can impact SAM/NAO on those timescales, but SAM/NAO are primarily subseasonal. I think this should be corrected throughout the manuscript. It is also confusing when you say that there is no link to SAM here, then discuss later in the results that there is a link?
l. [164][165][166][167][168][169][170]Fig. 5: phase A in timeseries is when PC is strongly positive (Fig. 5a), hence I would expect the EOF pattern for phase A to be a positive-monopole (i.e. Fig. 5b would then have red colours and 5c blue colours). Also, scale on colourbar seems wrong or maybe scale for PC timeseries is wrong? Is it really in metres? Also, I think it might be better to use u10 in Fig. 5 instead of swh, since you mostly look at composites for u10. Perhaps put current Fig. 5 in Appendix together with all other composites for swh EOFs.
For the sake of consistency, I think you should composite all quantities based on the same variable (i.e. u10, swh, cyclogenesis etc. composited over EOFs of u10; or alternatively over EOFs of swh) There seem to be differences between composites over EOFs of u10 and EOFs of swh (see below).
Section 3.2: in several places you mention that composites from u10 EOFs are similar/consistent with composites over swh EOFs. I see many differences between the two.  Fig. 6a); and swh composites show weaker anomalies. panels (b) show a meridional shift between swh and u10 composites (tracks in u10 composite are shifted polewards compared with tracks in swh composite). By how much it is hard to tell. Again, composites over swh show weaker anomalies. panels (c,d) are somewhat consistent (though it is hard to tell), but anomalies are weaker for swh composites. Perhaps the issues is that swh lags behind u10 -e.g. if you do lag-correlations between PCs of u10 and swh you may find a lead-lag relationship between the two. So instead of correlating the two at lag 0 like in Table 1, correlate them for several positive and negative lags, to establish a clearer relationship. If you then lag data accordingly you might then get the "same" results for swh and u10 composites -or just plot general lagcomposites. OR the swh and u10 peak in different locations. Another thing I can think of is that EOF1 and EOF2 may not be entirely independent at longer lags (at lag 0 they are by definition uncorrelated) and may represent propagating mode (i.e. if you did a POP analysis [and I am not suggesting you do it] you might find EOF1,2 of u10 to represent the same POP's real and imaginary components). Indeed, PSA (and also SAM) modes are like that and if EOFs1,2 of u10 are related to PSA modes then this can also be a part of the story (i.e. both modes impacting swh at different lags). Also, I think that u10/swh A & B composites should be shown over the same regions as cyclone tracks and genesis -that way a link between these quantities can be clearer; i.e. use Fig. 6 type plots also in Fig. 5b,c, & Figs. 7,8. Note that track densities following wind anomalies are likely consistent with positive baroclinic feedback (such as that presented in Robinson 2000). l. 182: you mention SAM: so do you ultimately find any links to SAM or not? l. 201-208: Similar to the above comments: swh and u10 seem to be out of phaseperhaps plotting both of them on the same plot (one in contours and one in shading) could help you (or me) whether they are out of phase and by how much. Again, there is likely a lead-lag relationship or they are simply peaking in different locations.
l. 216-217: I can also see SW-NE orientation south-west of SBB, which makes me wonder if this is what brings high shww to SBB? Fig. 9: I think I can see cyclone-anticyclone (trough-ridge) pairs in all panels, but the exact position, orientation and magnitude differ. For example, Fig. 9a,d have the pair oriented along the S. America coast (i.e. SW-NE), but in Fig. 9b,c the orientation is perpendicular to the coast (i.e. NW-SE).
Perhaps you could think about a future study where you could do a regime perspective (e.g. using K-means) to really classify different regimes that cause this swh. [just a suggestion for future work] Overall, I think that some lead-lag relationships are missing, and that once those are established everything will make sense.