|This is my 2nd review of the manuscript “The Mediterranean Ocean Colour Level-3 operational multi-sensor processing” by Volpe et al. The paper is mostly a description of the near-real-time/delayed-time processing of ocean color data in the framework of CMEMS, followed by results mostly based on a validation analysis for Rrs, Chla and Kd. A comparison with CCI data is also performed.|
The paper appears well suited to the CMEMS special issue and the authors have generally followed my comments on the original version. So I don’t have major comments, except one. I would like the authors to further improve on the clarity of the NRT/DT/REP characteristics. The abstract concludes saying that the study “demonstrates that the NRT processing chain compares sufficiently well with the historical in situ datasets to be confidently used also for reprocessing the full data series”. This should not be interpreted literally, particularly by users of the NRT data; as I understood the text, no validation results are shown for (strictly) NRT data so that the results of the manuscript are not valid for this type of data. This should be clearly stated in appropriate places in the text and the conclusion of the abstract modified accordingly. Again if I understand well, the authors have recently applied their DT processing chain to the full time series for this paper; considering that the NASA missions are then in a R2018 status and the MERIS data associated with the latest ESA reprocessing (post 2012), this DT time series is in practice a REP time series at least up to 2018 when the calibration of MODIS and VIIRS might start diverging from the R2018 parameters. Again from a user point of view, the quality of the DT data might become degraded for new data. This should also be simply stated.
The authors should also consider the minor comments that follow to further improve the text.
l.16: “kd”: kd and Kd are used randomly in the manuscript. Please stick to one symbol.
l.2: remove “on”
l.11 : “The next…”
l.11-12: “radiance”, “irradiance” (2)
l.5: “processed with”
l.7: “the only changes being the input data…”
l.9: “(NASA processing version R2018.0) currently rely on”
l.10: “L2 data”
l.11: “which uses”
l.15: “As detailed later”
l.21: “VIIRS data are derived …”
l.22: “MODIS-AQUA data”
l.25: “In this work,”
l.26: “this chain, Chl and Kd490”: I interpret that ‘REP’ refers to Chl and Kd490 derived from CCI Rrs?
l.31: “chain is used … are obtained” (for tense consistency).
l.32: “MERIS data are from”
l.24: “removal flag of l2gen”
l.6: “nearest-neighbour approach”: ? if this is the case, and considering that the missing pixel is surrounded by valid values, which one is chosen?
l.10: this is in effect produces a sensor-specific daily L3 product, right? Is simple averaging used?
l.26: “estimate Rrs”: is a MODIS 510 nm band computed? Is there a special treatment as this band is kind of between 488 and 531?
l.29: “from which it is then possible to derive…”
l.10: “The magnitude of the differences” ? An order of magnitude difference would suggest a factor 10.
l.10: considering that its results are well described in this paragraph, I am wondering if having Fig. S1 in supplementary material is appropriate. It could also be included in the main text.
l.12: “443 nm”
l.17: “responsible for”
l.4: “reduce the spatial gradients”: but what is described here only relates to time.
l2: “on the use of a SeaWiFS daily climatology field”: this would help understanding the method description.
l.7: “sharp gradients”: at that stage between one-sensor daily map and its associated climatology?
l.21: I’d suggest to reword as: “the satellite Rrs benefiting from the bias correction are closer to the in situ measurements at all bands”.
l.4 : “For Case I …”: by D’Alimonte et al. ? I don’t think they used the MedOC data set. This sentence introduces some confusion in this paragraph (and might removed).
l.33: “whose functional …”: does this bring some additional information to the sentence? It might just be better to say “ratio between 490 and 555 nm”.
l.6: these are ‘grid points’ and no longer ‘pixels’.
l.8: “10 am”: Fig. S2 says 12am.
l.24: “NRT/DT”: this should be a place to clarify that the validation analysis is conducted with the processing chain in the DT mode (and not NRT).
l.9: “(RPD of 76%)”
l.11: “Table S2-S7”: from these tables, it seems that the band-shifted data are also compared with respect to in-situ data; this should be mentioned in the table caption. Similarly, if there are validation statistics for all bands of each mission, it is likely that in-situ data have also been band-shifted when necessary. This is not mentioned anywhere I think. Units should be mentioned.
l.13: Table 2: units (% or geophysical) should be indicated.
l.31: “larger”: in relative terms. APD is largest at 670 nm, but lowest in RMSD.
l.3: I’d say: “differences between these two products being smaller than 5%”, to avoid confusion and clarify that this is a comparison with respect to in-situ data.
l.4: “R2014 and R2018”: yes but note that there are other notable differences between Multi and CCI, such as the use of Polymer.
l.6: “is seen at 670 nm” (to avoid repetition).
l.11-12: isn’t it the same message as lines 1-2?
l.17: MERIS is also processed with POLYMER, no?
l.26: “merged..” what? Product?
l.9: “global set … a larger modal…”
l.18: “Figure 6 and Table S7”: why should stats for kd be similar with Rrs results?
l.19: it is also surprising that the Med algorithm is actually worse (in terms of bias).
l.20: “one such reason”
l.3: “CCI, 2016b” ?
l.6: “merging approach”
Fig.1: “in situ stations”
Fig. 3: what do the curves within the color bars mean?
Fig. 8: “red and orange”: isn’t it black and blue?
Fig.9: “NRT”: this should be NRT/DT.
Fig.10: “space-time distribution”: panel c?
Table 5: “the three matchup data set”: three? Units should be given when appropriate (same for all Tables S).
Table S9: “whose location is shown in Figure 10c”?