The authors have addressed most of my previous concerns. I have a few remaining suggestions, but they are mostly quite minor.
Specific comments:
-Abstract: Without conducting tests on real-world A. islandica proxies, I’m not sure that the final sentence of the abstract is entirely justified based on the results presented in this manuscript. Since this study is conducted entirely in the virtual reality of a pseudoproxy experiment (with no testing of real-world A. islandica), I don’t think it is justified to say that “The results show that the marine network of Arctica islandica can be used to skillfully reconstruct the spatial patterns of SSTs at the eastern NA basin.” This statement assumes that real-world A. islandica proxies behave like the pseudoproxies. Based only on the results presented in this study, this seems like a big assumption to make, particularly since the construction of the pseudoproxies assumes both stationarity and linearity, whereas the real-world proxies may not have a stationary or linear relationship to SST. Further, if the real proxies behave like the noise-contaminated pseudoproxies used here, then the skill of SST reconstructions may actually be quite limited (as suggested by relatively low RE values in Figs. 12S and 13S and quite high variance losses, even in locations relatively close to the proxy locations). The results of this study show that the choice of surrogate reality makes a big difference and that both CCA and PCR could be appropriate choices for reconstructing SST in this region, but I think the authors should be a little more circumspect (both in the abstract and in the conclusion) about what exactly this study tells us about the skill of a real-world reconstruction of SST with A. islandica.
-Line 83: “CFR” is misspelled “CRF.”
-Line 187: should read “Reduction _of_ Error”
-Line 236: “…non-climatic noise, _it_ is usually…”
-Are equations 2 and 3 correct? I could be wrong, but I think to reconstruct PC_{i,t}, you would sum across all j, not across all i. So I think Eqns. 2 and 3 should have SIGMA_{j=1-n} (where n is the number of proxies) rather than SIGMA_{i=1-10}. Again, I could be wrong, but the authors should double-check these two equations and confirm that they are indeed correct.
-I would highly recommend a couple minor modifications to the figures that will make them more easily readable and interpretable. First, the figure labels are so small that they are nearly impossible to read. The font should be larger. Second, the color scales on all of the SDR figures should be rescaled. Theoretically, SDR should never be greater than 1, so it isn’t necessary to have a color scale that goes up to 2. That just makes it harder to distinguish the spatial variation of SDR. Also, since SDR ranges from 0-1, a sequential color scheme would probably be most appropriate, with light shades representing low SDR and dark shades representing high SDR (as opposed to the currently-used diverging color scheme, which implies a distinct split in the data that are being mapped).
-RE figures: it’s unclear from the caption and colorbar what the black grid cells represent. Are these regions with RE<0? The spatial patterns of the black grid cells also look a little weird. There are some subplots with entire regions blacked-out, and some where the black grid cells instead form lines or unfilled polygons. I’m also not sure that the colorbar should stop at a minimum of zero, since RE can be much lower than zero. It looks like there are probably regions of the NA (particularly far from the proxy locations) where RE is in fact less zero, but it is hard to tell based on the current scaling of the colorbar. I think it should be shown much more clearly where RE is less than zero.
-Figure 1: Correlation coefficients are “columns 1 and 3,” not “columns 1 and 4,” right?
-Figures 1 and 2: it should be clearer in the captions that these are for noise-free pseudoproxies.
-Figure 7S: Is this caption supposed to say “As in Fig. 6S” rather than “As in Fig. 5S”? |