Statistical Assessment of Sea-Surface Salinity from SMAP: Arabian Sea, Bay of Bengal and a Promising Red Sea Application

Round 1

Reviewer 1 Report

General Comments:

General Decision: excellent work and results. A minor revision Language: fine – very minor corrections needed; line 48: advance = advancement, line 96: In the time mean; in the mean time?! Line 583: tho = to. Length: paper is very lengthy, and, in fact can be shortened. Scientific Method: excellent New addition to the field of oceanography remote sensing: yes

Details Comments:

Abstract: informative.

Introduction: Technical info about SMAP can be removed for the sake of shortness; e.g. lines 19–28, 31–38 and 54–60. Readers better directed to the SMAP official website instead. Some salinity data types and sources better moved; for the sake of shortness, to section 2 (data and methods). Lines; 41–42, 144–145 and 177, about “Practical Salinity Scale”/“dimensionless”: this is unnecessary information and unneeded repetition. All oceanographers nowadays are using practical salinity unit; PSU. Figure 1, b &c, the SSS scale does not show any variability in the Red Sea, where all the salinity values in the Red Sea above 36.5, so its better to change the scale till 40.5 or separate Red Sea in other Figure. Data and Methods: Again; lines 163–179, 180–187 and 188–197 can be removed for the sake of shortness, then readers advised to consult the JPL SMAP Sea Surface Salinity (SSS) CAP V4.2 and V4.0 Dataset Release. Data and methods: lines 139-207 can be summarized, since all the details of the data can be referred to the technical papers mentioned in the reference list. The collocation procedures: Lines 300-352, it contains rich information and details which could be confusing in comparing different procedures. It would be easier for the readers to follow the procedures if the authors introduce a flowchart of different steps. Results: Adding P-values for each correlation: i.e. correl/P-vaule, in tables 3, 4, 6, 7, 8 and 10 is crucial. In line 583: “The Red Sea is as salty as the Mediterranean Sea” this is not true. In line 589-590: “in the Red Sea, found correlations to be even higher (0.88/0.85)” this is relatively high value for a correlation. I doubt its significance! Simply because the Red Sea is very narrow. What is the P-value for this correlation? From figure 6 both data sets at L2 and L3 show high difference with Argo in the northern Red Sea with values close to 0.9 and above, this area is active convection region during winter and such differences could change the dynamic of the water column. I suggest mentioning these high differences in text. In lines 601-602: “Seasonally, there is evidence that SMAP captures the intrusion of the fresher, colder, and nutrient-rich Gulf of Aden Surface Water (GASW) in winter” this is good and clear agreement for the the Red Sea comes obviously out of the study. In line 605: “the eastern Red Sea slightly fresher than the western” this is another good outcome for the Red Sea. Figures 17 and 18 show the distributions of SSS from the two data sets compared with that of ATLAs (Figure 16). I suggest showing the differences between those data sets and atlas data. Discussion: informative. Conclusion: informative.

Best regards and happy holidays ..

Author Response

Please check the PDF version

Reviewer 2 Report

The present study assessed SMAP sea surface salinity retrievals in the North Indian Ocean, particularly in the Arabian Sea, the Bay of Bengal, and the Red Sea. This work is very intensive with huge analysis. The work provides very detail information and analysis with 19 figures and 10 tables. I strongly recommend to accept this paper without any revision.

Author Response

Please check the PDF version.

Author Response File: Author Response.pdf

Reviewer 3 Report

Review of the draft “Statistical assessment of sea surface salinity from SMAP: Arabian Sea, Bay of Bengal and a promising Red Sea Application” By Viviane Menezes The draft reports on the performance of 3 different SMAP reprocessing products on the L2 and L3 level in inferring sea surface salinity (SSS) in the northern Indian Ocean, including the Red Sea. The ground information of salinity originates from the in situ data of time series stations (moorings), upper most Argo float measurements and ship based thermosalinograph measurements. Main results are the mean differences (bias), the root mean square differences (rmsq), linear correlation coefficients ( r )and skew of the different data products. In general, the draft is well organized, the English is easy to read and the figures are attractive. The analysis is well described and diligently carried out. Doubtless, the presentation of the satellite derived salinity data evaluation in the Red Sea is a new component and an exciting result of the study. However, the validation study of the northern Indian Ocean sea surface salinity data is quite lengthy – this could be summarized more and could give more interesting results. I give some proposals how to condense the draft in the following. In the beginning, in the abstract and in the introduction, the author reports about the lacks of satellite data in retrieving salinity, and the reader is informed that the retrieval algorithms advance from one reprocessing level to the next with improvements in low performance regions (line 46/47). However, the improvements are not lined out in the results (chapter 3) and the validation results are limited to the regional bias, rmsd, r and skew. I miss a more elaborated study about the performance in retrieving SSS in dependence of the distance to the coast, of seasons (rainy or non-rainy), of the SST and of the presence RFI. Furthermore, I would be interested in the improvements from the latest SMAP data versions in these aspects as announced in the introduction. The first 54 of 64 references are used in the introduction. This reflects the style of the draft: The items to which the authors wants to refer could be specified better (e.g. in lines 31, 38, 40, 70) it cannot be expected from the reader to read several references in order to understand to which subject the author is referring to. Alternatively, the author could summarize what the other study was finding rather than make a general statement. In the discussion section, I would expect more references to validation studies with similar procedures or results also of other satellite missions, and here unfortunately, the references are limited. In line 64, the author states that the focus is on the Northwestern Indian Ocean, I think it should be the northern Indian Ocean. An open question in the SSS validation studies is the collocation procedure. The applied procedure is well described in the present draft, however, I miss a comparison to the procedures in the literature given, and a critical review about the distance and time span, which is chosen solely according to the resolution of the available data but not according to the variability of SSS. Are distance and time span resolving the spatial and temporal variability of SSS, what are the issues concerning this choice? In the chapter 2 the author gives information which I consider being not relevant if are not being used to infer results, e.g. the temporal availability of the data or the median distances/time between the center of the grid cell and the position/measurement time of the Argo buoy. This repeats in section 3.2, where the regional and temporal variability of the data availability is mentioned, however, not used for interpretation of the results. Moreover, I don’t see the relevance of separating the validation into the TSG and Argo ground measurements, nor in the figures, neither in the tables. Concerning the figures, I do not favor scatter plots, I prefer pdfs which are more constructive and an integrative way to present the relevant information. However, it may be a matter of taste. With pdfs, the author could summarize the Figure 4 into two subplots instead of six. The products could be compared intuitively from the superposed lines rather than guessing how broad the scattering clouds may be. Moreover, for the dots of each measurement plotted in Figure 6 the reader is not able to evaluate which is the dot plotted last superposing the dot plotted before. Also here, a pdf would be much more instructive. In this Figure 6, the TSG differences of the sections in the eastern IO at 5 N and further north in the Bay of Bengal are evidently biased. This can be referred in the text like the author did, however, I do not see the relevance of separating the TSG and Argo data in the validation of the satellite derived SSS. The figure does not convince me, a pdf of the single differences could show the same info more clearly, otherwise the figure could be omitted. The figure 7 and 8 are more instructive. By the way, where are the differences between figures 7, 8 and 12? I guess these figures could be merged. The separation of the figures into different regions does not make any sense to me, just turn the draft quite lengthy. For the RAMA time series stations, the SSS comparison is validated with the deeper measurements (at 10 m, Figure 9). There is no estimate nor a critical sentence about the existence of a potential salinity stratification and its influence on the statistics. A lack of interpretation is also evident for Figure 10, it would be interesting why the correlation and rmsd show a minimum in the second half of the year in the Arabian Sea and in June in the Bay of Bengal. Figure 14 has the observation number in the horizontal axis, this should be a physical unit, or distance or time. In my opinion, these TSG data could be interpreted much better. The author could derive autocorrelation scales, here in space, for Fig. 9 in time, as Bingham did or Delcroix or Martins (see BAMS Boutin et al for being inspired) Figures 16 to 18 are too much, they have the character of a technical report. In my opinion the author could think of a better and condensed presentation, or in terms of annual amplitude, or selecting 4 months only, or presenting a kind of hovmoeller plot along the diagonal axes of the Red Sea…. However, they present new findings of the performance in satellite derived salinity and show really promising results. In the discussion (chapter 4) the author suddenly refers to a signal to noise ratio. This has nor been calculated neither shown before. A simple statement is not acceptable. A comparison with earlier validation studies are interesting if carried out in the same region, so if the author wants to convince me of a better performance of the newer data product version I would like to see the comparison for the same regions. The performance of the satellite derived salinity is certainly acceptable in the mean time, however, this should not be proven by showing some individual maps of superposed currents from another data set, not introduced before. This gimmick should be omitted. Moreover it is not justified why the L3 CAP data is used – as I understood from the study before, the L3 RSS data show better results…Why in the end of the discussion, the author only refers to the L3 CAP results, when Table 9 is stating, that the RSS 70 product shows better performance? Concerning the study of Fournier et al [18] , why the in situ data used as ground truth are different? Overall, I suggest to condense the lengthy material, enlarge the horizon to other validation studies and cut the eddy glimpse in the Gulf od Aden, i.e. major revision.

Author Response

Please check the PDF version.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments on ‘Statistical Assessment of Sea-Surface Salinity from SMAP: Arabian Sea, Bay of Bengal and a Promising Red Sea Application’ by Menezes. This work assessed the NASA’s SWAP sea surface salinity products in the North Indian Ocean, e.g., the Bay of Bengal, the Arabian Sea, and the extremely salty Red Sea. The author evaluated six SMAP products against in-situ measurements collected by a variety of instruments, and it is concluded that the SMAP reproduced SSS well not only in the Arabian Sea/the Bay of Bengal, but in the Red Sea. In general, the paper is well organized and written. The author collected state-of-art SSS datasets and compare against various in-situ measurements. The results are convincible. The evaluation of the SSS products in the coastal regions will be interested to the readers. In my opinion, it could be accepted for publication.

Author Response

Please check the PDF version.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

In my opinion, the paper could be ameliorated as I already stated in my first review. There, I have posed several questions and asked for a more profound interpretation of the differences found between in situ and satellite derived salinity data. Moreover, I have proposed to widen the study in respect to the methods applied in other studies dealing with the comparison of in situ and satellite data, and to deepen the study in terms of interpretation.

Questions remain open or are not considered at least as critical viewpoints

concerning the coverage of the scales of salinity variability by the in situ data and therefor worse performance of the satellite salinity data in regions of high vertical gradients and strong spatial and temporal variability. concerning the land contamination correction within the products which is also based on the in situ data availability concerning the RFI which should be available in the data used

Moreover, I find the style of the draft is more similar to a technical report. However, the data and the regions are innovative and a lot of work is done in this study, so the editor may decide if the quality is reaching the journal's need.

Best regards.