Dear authors,

I find the manuscript (ms) a very important contribution to the field of applied ecology, particularly to understand the mechanisms that shape the shelf-regulatory and resilient mechanisms of biological systems against natural and anthropogenic perturbations. My overall assessment is that this ms can be of potential interest for the readership of PRSB, while there are some important elements, at analytical and interpretation level, that should be revised, and/or providing well-funded rebuttals, since they are not totally convincing in the current version. However, having said that, I understand that the paper has the chance of being outstanding following revision.

First, though the study is conceptually and analytically sound by applying SEM to this complex set of relationships under the biomass-diversity-stability framework, there is an important empirical relationship/link missed in the SEM, whose inclusion must change the rest of direct and indirect relationships and likely their significance and strength effect. I refer to the relationship between synchrony and the species-averaged stability, which is often strongly correlated in most studies since they have shared variance across common mechanisms and structural biodiversity facets. The same reference that the authors use to support the inclusion of this metric (Thibaut and Connolly 2013) but also many others studies, shows this relationship, which must be included the SEM.

The second and even more important element is that part of the theoretical foundation of diversity-stability theories is missing in the SEM, and the whole ms and discussion. I refer to taxonomic biodiversity facets such as evenness (whose effect was also shown by Thibaut and Connolly 2013) and richness, as they both have an important influence on stability. I think that they both must be also included in the SEM too, and/or explain which is their role. If any of the taxonomic metric is correlated to the functional biodiversity metrics included (i.e. corr > 0.6 or <-0.6), the authors must show it and interpret the results accordintly. In addition, in the case that the stability metrics analysed are spatially correlated before the SEM (not the residuals), authors may consider to include as well a beta-diversity metric in the SEM.

OTHER ISSUES

• The independence of all communities must be convincingly reported as they are relatively close and the environmental drivers used are strongly spatially correlated, which can make that some of the relationships found are spatially driven. Also, in addition to the residuals of the SEM relationships, it would be useful to check the spatial auto-correlation of the stability metrics analysed before SEM.

• In terms of environmental drivers, I miss the inclusion of the structural complexity of habitats and seascape as a critical element shaping the stability of coastal communities. More complex and heterogeneous habitats are more prone to display asynchronic fluctuations of the species of coastal communities. Therefore, in addition to all the pelagic environmental drivers here used, I would suggest to include any seascape metric reflecting the degree of spatial heterogeneity of the habitat in addition to depth.

• Anthropogenic impacts. It is not clear what is the ‘human impact’ indicator measuring, and therefore it is difficult to link to the interpretation of the results. As far as I understand, the human impact indicator used is cumulative pressure indicator associated to human coastal development but not necessary totally associated to fishing impact, or at least, this is not convincingly shown. To me, it would make more sense to relate the higher asynchrony to the erosion of the spatial structure of populations and habitats that selectively remove certain traits associated to some habitats, and also erodes structural and functional connectivity that asynchronize populations. Note that the papers by Anderson or Hsieh you refer are associated to a different mechanism of fishing impact in large fish stocks and off-shore populations. This other mechanism is the demographic erosion by the size-selective effect of fishing, which makes the populations more dependent of young ages classes and then on population recruitment; this process is different to the one you refer in the ms on the trait-selective removal within a community reducing traits diversity and indirectly affecting synchrony, and also the species fluctuating degree. Zac Hsieh has other interesting paper in 2010 in Aquatic Science touching spatial elements that could be more useful.

• In the discussion, authors often refer just to climate change and anthropogenic impacts as the main drivers of stability. However, the authors should refer to three elements ‘natural environmental variability, climate change, and anthropogenic impacts’. Mean SST, CV SST and CV chlorophyll, are more associated to environmental geographic gradients and natural environmental heterogeneity (and predictability) rather than to climate related drivers. Authors may considerer this natural variability as important element in the discussion rather that rely so much on climate change.

• Some important explanations of mechanisms are to complex and convoluted in one sentence and need more space and elaboration (e.g. lines 244-247, or lines 250-252, among others).

• Figure 1 need to include the values of the strength of the effect in the arrows, and use a shape of arrows that is not triangular, that is very confusing (e.g. mean SST to CWM_TL or CV chl to trait diversity).

• Authors refer many times to the STAR protocol but I don’t see it explained in the methods.

• Some information on the spatial buffering area used to calculate environmental indicators for each community should be given.