Revision of Ryan-Keogh et al. “A new global oceanic multi-model net primary productivity data product”

General comment

The paper submitted by Ryan-Keogh et al. focus on satellite NPP estimates based on the existing algorithms. To this aim, the authors have used the longest satellite time-series available from the ESA OC CCI project. The paper is well written and easy to be read and it is a useful work for the ocean colour community.

Overall, I have few specific comments to be included in the revised version of the manuscript. For instance, one point to be addressed is about the differences between the NPP products based on ESA OC CCI dataset and the NPP estimates for MODIS and SeaWiFS. I suggest to extend this paragraph also including some lines about the difference ocean color bio-optical algorithms used in the NPP computations. For instance, the use of GSM instead of QAA that is used in OC-CCI to retrieve the backscatter; this is valid also for the chlorophyll concentration estimates.

On my opinion, the paper needs only minor corrections before the publication.

Specific comments:

Lines 26-27: I’d suggest to rephrase the sentence and include a more generic concept about the spatial and temporal variability of phytoplankton dynamics instead of only the seasonal bloom as a key biological process which play an important role in the carbon.

Lines 30-36: For me this sentence is not clear, please could you rephrase. There are also some repetions that need to be removed.

Lines 40-41: Maybe you can listed areas where there is an increase and areas where there is a decrase of NPP.

Lines 42: DIC is the first time cited. Please include the acronyms. In addition, I suggest to include just a sentence to introduce the role and importance of DIC and its connection with the biological carbon pump.

Lines 43-47: This paragraph can be reduced in terms of lines. An option should be to refer directly to the trophic chain or food web.

Lines 55-57: In which sense the NPP is a better proxy environmental changes? It is because the phytoplankton chlorophyll is also influenced by physiological adaptations to light, nutrient and temperature? Maybe you can add a short sentence and a reference to better clarify this point.

Line 60: Nutrient availability instead of macro- and micro-nutrient specification.

Lines 66-68: You should include a reference at the end of the sentence. For instance Johnson and Bif (2021; Nature Geoscience).

Lines 87-89: You should include also that this programme is funded by th European Space Agency (ESA). Maybe some lines about the programme. The main goal of the OC-CCI is to have a long-term consistency merged ocean color dataset. OC-CCI data products are the result of the merging of SeaWiFS, MERIS, MODIS, and VIIRS observations in which the inter-sensor biases are removed. version 6.0 includes an updated processing, which mostly accounts for the ageing of the MODIS sensor and the new sensors (Sentinel-3/OLCI).

Lines 97-99: Maybe, you should say like: NPP from OC-CCI is not available yet since it is not a standard product of the programme. Kulk et al., (2020) have used the dataset to develop and validate an algorithm that however is not listed in the suite of variables of the OC-CCI.

Lines 105-106: Probably you should smooth the sentence a little bit as: however, it is difficult to obtain all the variables needed for the detection of NPP such as the mixed layer depth, the nitracline, ecc

Lines 122-123: The spectral slope is not a product listed in the OC-CCI project, so you should say: we have computed the spectral slope of bbp (xx) following.

Line 129: Maybe a reference also for the criteria of 0.125 kg/m3

Line 145: Did you use monthly climatology data or monthly data for the entire time-series?

Lines 180-185: You should start as: “As additional term of comparison, monthly NPP data of.

Lines 397: I suggest also to include the algorithms at the base of the ocean color quantities such as chlorophyll, backscatter ecc. For instance the chlorophyll in ESA OC-CCI dataset is based on different algorithms as well as the backscattering coefficient is based on the QAA algorithm. The Behrenfeld NPP CbPM is based on the GSM that sometimes overetimates backscattering values at 443. For this reason, you should extend the last paragraph of the results and discussion taking into account this additional source of differences (algorithms).

Review of “A new global oceanic multi-model net primary productivity data product”, by Ryan-Keogh et al.

This manuscript introduces a new data product that consists of an ensemble mean and associated variances derived from a suite of diverse global satellite net primary production (NPP) models. The data product is generated by application of the various NPP models to an established, merged multi-mission ocean color record that spans the full modern satellite record (1998-2022).  The manuscript describes the approach and some of the basic spatio-temporal patterns observed in the product.  It is well-written and the graphics are good quality.  I recommend this manuscript for publication without major changes.  It does not introduce any new science per se, but provides a product that will hopefully be of use to the broader science community.  I offer the following points out of general interest and perhaps to better-clarify certain points.

First, the authors acknowledge the need for this product to provide an alternative to other data products already in existence (e.g., those hosted by Copernicus Marine Services or the Oregon State University Ocean Productivity website).  However, in doing so, I feel that a massive disclaimer is needed stating that advantages of this ‘ensemble approach’ may be fully mitigated by combination of estimates of varying quality (as assessed by exercises such as the Primary Productivity Algorithm Round Robin series, Campbell et al. (2002); Carr et al., (2006); Friedrichs et al. (2009); Saba et al. (2011); Lee et al. (2015)).  While exercises such as these are not definitive, the community has dedicated tremendous effort to trying to establish the fidelity of these models.  This has been challenging and is limited partly by satellite – in situ matchups, but I feel that we were at least converging on a narrative that CAFE>CbPM>VGPM.  Personally, I would gladly use the CAFE model applied the OC-CCI dataset as the preferred data product (but I realize not everyone may feel the same :).  Perhaps, a bit of a philosophical point, but I think it deserves some discussion. It could easily be placed in the paragraph starting on Line 96 to balance the justification for the current product.

I’m concerned about the global annual integrated NPP values.  They are really high, much moreso than we have reported in the original publications and related work.  I wonder if the integrals are unduly influenced by a low number of ‘spurious’ values?  They might be easily traceable to something in the input fields (e.g., coastal Chl-a retrievals > 50 mg/m3, or spurious bbp retreivals in the case of the CbPM).  It would take more than a pixel or two, but implementing these ‘traps’ on the input, as well as the resultant NPP, can be important.  Also, in the case of the Behrenfeld-CbPM, the formulation inappropriately uses Kd490 to estimate the euphotic depths and mixed layer growth irradiances, both of which can be significantly overestimated in this way.  In Westberry et al. (2008), we point out that the global annual NPP is reduced by nearly 2x (from 67 to 35 Pg C) by simply replacing the Kd490 terms with simple Chl-dependent parameterizations (e.g., Morel-type relationships).

Figure 2a, I think you should truncate the CV map at 0.6 or so I order to see more spatial structure?  Just a thought ...

Lines 262-265, The convergence or divergence of model NPP over time is interesting.  Is it related to the inter-sensor merging of the OC-CCI record?  Is it related to the fact that some models utilize products that others do not (e.g., MLD)?  Resolving this is beyond the scope of this paper, but to me, this is related to the point above regarding the advantages and disadvantages of the ensemble mean versus a single model that we believe performs best. 

Line 238, typo should read ‘Westberry-CBPM’