The current manuscript describes an analysis primarily focused on evaluating alternative approaches for assessing phytoplankton carbon biomass.  This is a challenging and important issue.  Unfortunately, none of the measurements conducted during this study were direct analytical measurements of phytoplankton carbon, so the authors are limited to comparisons between different indirect proxies.  This limitation is clearly articulated in the manuscript along with a brief discussion of underlying uncertainties with each proxy.  The manuscript is written in an almost conversational manner, which I found enjoyable to read.  That said, in future versions it would be beneficial if the authors found a colleague whose first language is English to provide a final edit to clean up some of the writing to make it clearer to readers.

I found the manuscript to be fundamentally flawed in ways that compromise the overall conclusions.  The central measurements of this study were total chlorophyll (tchl), beam attenuation (cp), particulate backscatter (bbp), POC, and cell counts.  Phytoplankton carbon (Cphyto) is assessed from tchl based on the study of Sathyendranath et al (2009) (hereafter S2009), from bbp following Graff et al. (2015) (hereafter G2015), and from cell counts following an approach similar to Martinez-Vincent et al 2013) (hereafter MV2013).  It should be noted that in both G2015 and MV2013 a significant relationship is reported between bbp and Cphyto, with the former representing the only assessment where direct measurements of Cphyto were used (as recognized in the current paper).  In G2015, it was noted that, while the MV2013 paper reported a linear relationship between bbp and cell volume, the slope was much too large compared to results based on analytical Cphyto data, highlighting a challenge with converting cell number and cell size data into Cphyto that is relevant to the current study.  Samples measured during the G2015 and MV2013 studies included phytoplankton populations that were very similar to populations sampled during the current study (e.g., oligotrophic waters dominated by Prochlorococcus and Synechococcus).  There is no mechanistic reason I am aware of that similar populations measured in different geographical places would be expected to exhibit vastly different relationships with bbp.  Thus, one must question why the current study failed to find a significant relationship between bbp and Cphyto? 

Before investigating this question further, it is important to evaluate the S2009 study.  In that study, near surface chlorophyll concentrations (i.e., >95% of samples were from <40m) were compared to POC values.  S2009 then assumes that at any given chlorophyll concentration, the lowest observed POC can be assumed to approximately represent Cphyto and that POC values above this minimum can be attributed to increasing concentrations of other non-phytoplankton particulate carbon forms (detritus, bacteria, viruses).  The fundamental flaw in this approach is that chlorophyll concentration is a reflection of 3 primary determinants; biomass, nutrient limitation of growth, and photoacclimation to light.  The fact that physiological factors can be responsible for greater than an order of magnitude variation in chlorophyll concentration invalidates the foundational assumption of the S2009 approach.  In addition, since the S2009 data set likely included few if any measurements from below the mixed layer, it is questionable whether the approach can be robustly applied to data in the current study from samples collected below the mixed layer. Given the aforementioned issues, one should be skeptical  about the Cphyto values presented in the current study that are based on the S2009 approach.

One of the primary messages in the current study is that bbp provides an unreliable estimate of Cphyto (at least for the current study region – but see comment above regarding mechanisms), and indeed this message is the main theme of Secton 4.2 and figure 8.  However, inspection of figure 2 reveals that the current data set is poorly suited for assessing a bbp:Cphyto relationship.  Specifically, the dynamic range in observed particle mass for the full cruise transect is driven by differences between the moderately productive waters sampled at stations 1-4 and oligotrophic sites at stations 5-20.  While Tchl, POC, cp and cell count data are available across this gradient, bbp is entirely absent in the more productive waters.  This lack of bbp data is a great disadvantage for assessing a bbp:Cphyto relationship, and it also influences the interpretation of histograms presented in the manuscript (Fig. 3, Fig 5).  Despite the smaller dynamic range in bbp, the authors nevertheless find is good relationships between bbp and cp, bbp and POC, and cp and POC.  Where things fall apart is when tchl is compared to the IOPs (fig. 3c,d). So lets give this more consideration.

In figure 3c we see that there is a curvilinear relationship between bbp and tchl when samples shallower than 150 m are included. Why is this?  In figure 2 we see a clear subsurface chlorophyll maximum in the oligotrophic region (where the bbp data are available).  This chlorophyll maximum is primary a consequence of photoacclimation (i.e., not biomass variability).  Thus, when bbp and tchl are compared, we get the curvilinear relationship shown in figure 3c, which can be interpreted as a strong indicator that chlorophyll is not a reliable index of Cphyto (i.e., in general, the photoacclimation contribution will increase in parallel with increasing chlorophyll concentration along the x axis).  If we look at the surface only data in figure 3c (solid symbols), we see little relationship between tchl and bbp.  The most straight forward explanation for this is that the dynamic range in Cphyto is very limited across stations 5-20, but there still is some variability in tchl as a result of changing phytoplankton division rates and mixed layer light levels.  If we now look at the data in figure 3d, we see essentially the same thing as in 3c, with one minor difference.  For the cp data in figure 3d, we now have optical measurements for the mesotrophic stations 1-4, which I believe show up as the cluster of cp values > 0.10. Aside from this data cluster, we again see the curvilinear relationship between cp and tchl for samples shallower than 150 m (open symbols) that is due to photoacclimation impacts in the tchl data, and the constrained range of variability in the surface only samples (which can be interpreted exactly as above for the bbp : tchl data).  While I don’t have the data to evaluate this, my guess is that the station 1-4 data cluster in figure 3d falls apart from the other open symbol data because these mesotrophic stations actually had higher phytoplankton biomass and that their photoacclimation state was different than the ‘below the mixed layer’ data from stations 5-20.  Since the authors report a strong relationship between cp and bbp, it would seem safe to assume that had bbp data been collected at stations 1-4 they too would have shown a good relationship Cphyto.

Given the above, my conclusion from the observations of the current study is that bbp, cp, and POC  are all correlated with each other and correlated with Cphyto, while assessments based on tchl are inaccurate.  This leave the question of how to interpret the cell count data.  I don’t know the answer to this question, but I encourage the authors to rethink their overall interpretations.

Other issues:

(1) line 421: this conclusion about the PPC data standing out from the PSC data indicating a significant influence of non-algal particles is neither logical nor supported by observations.

(2) line 501: this idea that bbp is largely influenced by submicrometer particles has its foundations in Mie theory as applied to homogeneous spheres and is antiquated.  It is well recognized now that particles significantly larger than a micron play an important role in bbp variability and that Mie predictions for homogeneous spheres are inadequate for characterizing backscattering properties of natural phytoplankton populations.

(3) lines 556-569:  these suggestions regarding a role for NAP are simply speculation.

(4) lines 589-599:  comments here about the role of non algal particles also seem weak.

(5) paragraph beginning on line 619:  I did not find the discussion in this paragraph regarding phytoplankton physiology mechanistically sound, nor the statements regarding the relative contribution of phytoplankton to POC.

(5) Section 4.4: The conclusions stated in this section regarding the utility of bbp for assessing Cphyto are incorrect (as explained above).  The proposed role of NAP variability suggested in the two Bellacicco papers is based on a flawed assessment where bbp associated with NAP is assessed through relationships between bbp and chlorophyll even in low chlorophyll waters where chlorophyll variability is predominantly physiological. 

Review of Antoine et al. ‘Potential of optical and ecological proxies to quantify phytoplankton carbon in oligotrophic waters.

This article explores the use of various approaches to estimate phytoplankton carbon across a range of waters in the Indian Ocean, with a focus on chlorophyll a, optical backscatter and flow cytometry as predictor variables. Such Cphyto estimates are of great value in marine biogeochemical studies, with important applications to satellite-based measurements. The results presented in this paper show a range of relationships between Cphyto and the various input variables, with significant correlations observed for the full data set, and weaker relationships for surface (<25 m depth) data.

Overall, I think that this a valuable and interesting study, with important implications for the field.  One significant limitation, however, is the lack of a ‘true’ (i.e. gold-standard) measurement of Cphyto.  Without this validation, it is not possible to say which method provides the best approximation for Cphyto, as the authors themselves acknowledge. Nonetheless, I think the paper is still useful, as we can (with some modifications to the current text) get a sense of how the different proxies produce different Cphyto estimates.  Other things I note, is the need for a bit more discussion on the role of light-acclimation in driving some of the observed variability, more discussion of the size biases in the cytometer data, and a more robust application of the pigment data to discuss the role of phytoplankton taxonomy.

Specific comments are listed below.

Specific comments:

Line 13: include physiological status as a source of variability

Line 16: ‘both are yet each in’??  check grammar here.

Line 36: ‘are’ missing before ‘accordingly’.

Line 63: reference after ‘change significantly’

Line 66.  I find the transition to this new paragraph a bit abrupt

Last line of introduction – do you relate the results to environmental variables also?

Line 93: replace ‘going’ with ‘sent’

Lines 94/95.  Given the significant latitudinal gradient sampled, did these sampling times represent a consistent part of the diel cycle (e.g. xx hours after sunrise or sunset).  If not, is it necessary to consider this?

Line 106: was the filtration for pigments conducted under low light?

Line 111: I think the ‘P’ in HPLC stands for ‘performance’

Line 169: I think it’s important to provide more information on the size cutoffs of the instrument (lower and upper).  What part of the size spectrum is being missed? This is mentioned very briefly on line 240, but it would be good to see it here, and with an upper cutoff also.

Line 200: I would have thought that the value of gamma differed significantly between the different phytoplankton assemblages, based on their size spectra.  What was the relative error in the mean gamma value, averaged for all samples?

Line 208: What was the vertical resolution / sampling frequency of vertical bbp measurements?

Line 213: I don’t quite understand this method, based on POC chl regressions.  It seems to me that the derived relationship would produce an average C:Chl ration, including a lot of detrital matter.  It makes more sense after looking at figure 5, which could be cited here.

First paragraph p. 10.  At this point, I was wondering about the various error terms.  These are addressed below, but it might be good to at least mention this here.

Line 245:  What is the theoretical basis underlying the relationship between the slope of the size distribution and absorption at 676?  Does it relate to pigment packaging?  Some more information would be helpful for non-specialists.

I gather that there were no size-fractionated chlorophyll data?  Those would have been really helpful to validate some of these results.

Line 304:  It’s true that conditions were more oligotrophic, but it’s worth mentioning the significant sub-surface chl maximum, which had chl values higher than observed to the south.

Fig. 2.  The red starts are not labelled in the legend (only in the main text). I would add dots to show the actual sampling points uses for the interpolations. For the bottom panel, rather than repeating chl, why not plot, for example, the PPC/PSP ratio, or POC:Tchl, which is mentioned in the text.

Last paragraph on p. 14.  This is very descriptive material, which I think could be removed, as it’s apparent from the figures.

Fig. 3 could go in a supplement, I think.

Line 370: Cphyto is listed here as having a non-linear relationship, but that variable is not shown in the plots.

Fig. 4 bottom left panel: the white points look rather non-linearly distributed.

Table 2: I don’t understand the last sentence in the table header.  Maybe change ‘panel’ to ‘figure’ along the column headers.

Line 403 and elsewhere.  Overall, it seems that the PPC/PSC ratios provide relatively little explanatory power.  Maybe this could be mentioned somewhere and the data not explicitly included, unless they provide additional useful information. In Fig. 7e, for example, all the points fall together.

Line 456: It’s worth noting that the POC – Tchl data were distributed over a much smaller range.  If you took a similarly narrow range of other variables, relationships might also not be statistically significant.

Line 466: worth emphasizing here the significant size bias of the cytometry data

Fig. 6: One on hand, I can understand why the authors used log transformation for visual purposes, but the fact that this is not applied consistently across the panels makes direct comparison rather difficult.  I would use all linear or log scaling.

Line 485.  I can see how the black points sit above the line, but I’m not sure if these data ‘stand out’ considering the magnitude of the error bars.

Line 501: replace ‘when’ at the end, by ‘while’

Line 505: replace ‘on’ with ‘in’

Line 509: I can’t really see the green line.

Fig. 7: I think it would be useful to plot the various Cphyto estimates against each other. 

Line 520: the PPC-dominated points are distributed over a very narrow range, which could explain the lack of a clear relationship. Are they ‘separated’ beyond the error bars of measurements?

Line 573: maybe ‘refractive’ instead of ‘refringent’ (I had to look that up).

Line 617. In addition to PSC/PPC ratios, there are other approaches to pigment-based taxonomy, including Chemtax, for example.  Would different results have been achieved with another method?  At the least, I think it would be useful to provide more information on the taxonomic composition of PPC-dominated waters. This comes up again on line 646.

Line 624: insert reference after ‘ratios’

Line 626: insert reference under ‘conditions’

Line 643: is the higher bbp due to larger surface area / volume ratio?

Line 690: There is no doubt that the cytometer provides a significantly size-biased view of the community.  I wonder how robust the ffc factor is across the different assemblages.

END OF REVIEW