Reconstructed 3-D Ocean Temperature Derived from Remotely Sensed Sea Surface Measurements for Mixed Layer Depth Analysis
Round 1
Reviewer 1 Report
Just editorial comments:
Line 127, "IPRCIPRC-Argo" should be "IPRC-Argo".
Line 164, ")" should be removed.
Author Response
Manuscript ID:remotesensing-656415
Re: “Reconstructed 3-D Ocean Temperature Derived from Remotely Sensed Sea Surface Measurements for Mixed Layer Depth Analysis”
December 10, 2019
Dear Reviewer,
Enclosed is a corresponding letter to each reviewer, a manuscript with tracking changes and a revised manuscript, “Reconstructed 3-D Ocean Temperature Derived from Remotely Sensed Sea Surface Measurements for Mixed Layer Depth Analysis” by Jeong et al., which we are pleased to resubmit for publication in Remote Sensing.
During this revision, we fully explained reviewers’ concerns and comments. We provided more validations in details for our MLR model results, as Table 1 in the revised MS.
We thank you in advance for your consideration of this manuscript for Remote Sensing publication.
Sincerely,
Young-Heon Jo, Ph.D.
Associate Professor
Department of Oceanography
Pusan National University
Busan, 609-735, Republic of Korea
joyoung@pusan.ac.kr
Author Response File: 
Author Response.docx
Reviewer 2 Report
The manuscript in this version improves a lot compared with the previous version. The new version added many detailed discussions including subsurface temperature anomaly variability and method performance around western boundary current region (e.g., Kuroshio, Gulf Stream, Brazil-Malvinas confluence) and MLD trend variability and related driving process. Their method and data are reliable, and the results are interesting. The study of MLD has lots of significance in biogeochemical and ocean heat studies. I would like to give the manuscript a minor revision.
Here are detailed major and minor comments
Major comments:
In lines 83-89, the authors said the method is simple and convenient to update the data sets. How is the accuracy comparison of this method with other complex methods (model-based, data merged via optimal interpolation)? I saw the authors list RMES and the correlation coefficient of the current MLR method in Table 1. How is the RMES and correlation and coefficient of other methods? I would like to see more analysis of MLD variability and related explanations. For example, in the pre-hiatus period, the reduction of MLD in the north Atlantic (close to Green Land) correlates well with SSTA. However, in the post-hiatus period, the drop of MLD in the Southern Ocean (west of South American) and SSTA are not consistent. The authors may need to discuss the different responses between them. In the discussion of MLD spatial and temporal variability, I suggest the authors increase a link of MLD variability with other heat and carbon variability observed. This can increase the paper’s influence by attracting more readersMinor comments:
The authors mentioned the global MLDA dropped in the pre-hiatus is probably driven by MLDA variability in the North Atlantic. How is the MLDA drop in the post-hiatus? Line 48, remove “try to”, it is limiting to understand … Line 86, have depended on à depends on
Author Response
Manuscript ID:remotesensing-656415
Re: “Reconstructed 3-D Ocean Temperature Derived from Remotely Sensed Sea Surface Measurements for Mixed Layer Depth Analysis”
December 10, 2019
Dear Reviewer,
Enclosed is a corresponding letter to each reviewer, a manuscript with tracking changes and a revised manuscript, “Reconstructed 3-D Ocean Temperature Derived from Remotely Sensed Sea Surface Measurements for Mixed Layer Depth Analysis” by Jeong et al., which we are pleased to resubmit for publication in Remote Sensing.
During this revision, we fully explained reviewers’ concerns and comments. We provided more validations in details for our MLR model results, as Table 1 in the revised MS.
We thank you in advance for your consideration of this manuscript for Remote Sensing publication.
Sincerely,
Young-Heon Jo, Ph.D.
Associate Professor
Department of Oceanography
Pusan National University
Busan, 609-735, Republic of Korea
joyoung@pusan.ac.kr
Author Response File: Author Response.docx
Reviewer 3 Report
This study proposed a MLR method to reconstruct subsurface temperature from remote sensing data for analyzing the mixed layer depth variation. The authors show a clear understanding on the scientific issue and well-presents those ideas in the manuscript. Significant yet interesting results are achieved. However, several issues should be well concerned and addressed before it can be accepted for publication in Remote Sensing.
The author didn’t well summarize the recent progress on the subsurface temperature estimation from remote sensing. Accurately, in recent years, several new methods based on AI / machine learning (such as support vector machine (2015), random forest (2018), XGBoost (2019) and clustering-neural network (2019), etc.) and geospatial modeling (GWR (2018)) have been proposed for well estimating/retrieving subsurface temperature, and already achieved significant progress. These recent studies are very relevant to this research, they should be paid attention to and also well summarized or reviewed in the manuscript.
The author set up multiple linear regression model for reconstructing subsurface temperature by considering the temporal variation feature only, but why not also consider the spatial variation feature which is very common in the ocean, and it would be better to set up the spatial-nonstationary-based model like GWR (2018) as a comparison experiment.
The accuracy evaluation for the MLR model is not enough in this manuscript. The quantitative evaluation of the model performance should be conducted by using R2, RMSE and NRMSE criterions, and the accuracy variation with the depth and time should be further investigated and discussed, so as to make the results more convincing.
Minor issues:
Line 83 - 87: these arguments are not true. On one hand, the concept of estimating subsurface temperature from satellite-based data was also applied by Guinehut et al. This is not the way the current study differs from Guinehut et al paper. The current study uses additional inputs (wind speed), which, in my opinion, is the highlight of this manuscript.
On the other hand, 'most previous methods …on model-based data' is also not true. Examples are Lu et al. (2019) and Su et al. (2019), etc.
The error or accuracy of the surface remote sensing data/products should be discussed before they were used for modeling.
Line 189: a bi-linear method was used to interpolate the coefficients. Did the authors test other interpolation such as cubic or spline? How sensitive to the methods? It would be also interesting to show an example of these coefficients for the reader to see their horizontal continuity and spatial distribution.
The SST derived from gridded Argo data (in Line 257 for instance) is actually not the concept of the remote-sensed skin-layer temperature, but is subsurface temperature which is very close to the surface (e.g., 2.5 m for the Argo product). This point should be noted somewhere in the text.
Line 370: Although I personally agree the argument of the MLD calculation here, it should be noted that the consistency in figure 6 is mostly due to the seasonal component which is relatively easy to be estimated.
Table 1: unit of RMSE should be added.
Author Response
Manuscript ID:remotesensing-656415
Re: “Reconstructed 3-D Ocean Temperature Derived from Remotely Sensed Sea Surface Measurements for Mixed Layer Depth Analysis”
December 10, 2019
Dear Reviewer,
Enclosed is a corresponding letter to each reviewer, a manuscript with tracking changes and a revised manuscript, “Reconstructed 3-D Ocean Temperature Derived from Remotely Sensed Sea Surface Measurements for Mixed Layer Depth Analysis” by Jeong et al., which we are pleased to resubmit for publication in Remote Sensing.
During this revision, we fully explained reviewers’ concerns and comments. We provided more validations in details for our MLR model results, as Table 1 in the revised MS.
We thank you in advance for your consideration of this manuscript for Remote Sensing publication.
Sincerely,
Young-Heon Jo, Ph.D.
Associate Professor
Department of Oceanography
Pusan National University
Busan, 609-735, Republic of Korea
joyoung@pusan.ac.kr
Author Response File: Author Response.docx
Round 2
Reviewer 3 Report
The manuscript can be accepted for publication in present form.This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
This is an important and well written study for reconstructing a new mixed-layer depth (MLD) data product with a high spatial and temporal resolutions and diagnosing the decadal variability of global MLD. The data product has a good correlation with other MLD datasets which indicate a high quality. They found different trends of global MLD in three recent periods (pre-hiatus, hiatus, post-hiatus) and attribute them to the SST anomaly variability. The finding is new and interesting and has lots of significance on biogeochemical and ocean heat studies. I would like to give the manuscript a minor revision.
Here are detailed major and minor comments
Major comments:
Since one of the main objectives of this study is to construct a new MLD product, I suggest the authors provide a web link to let the readers download the product. This can benefit the MLD community. In the analysis part, more scientific analysis is needed. For example, the significance of deepen and shallower MLD in different periods. How does this influence biogeochemistry and global ocean heat? May be another interesting way is to compute the total heat in the mixing layer globally and analyze its variability regionally and temporally. Line 306-320. The explanation for bias in the west boundary current extension region is the low resolution from argo product. Did you see similar bias when you use surface argo data to interpolate the subsurface temperature. If you did not see the similar bias, the reason is attributed to the resolution. If you do see the similar bias, the reason maybe not the resolution.Minor comments:
Lines 21-22, add this “Based on this relationship, high spatial resolution and extended ….” Lines 38-39, add this “contributes to the heat transmission from surface to the ocean interior”. Lines 35-42, add some biogeochemical significance of MLD except for large influence of heat. Line 45-47, the logics of these two sentences is little bit strange. Why “the floats do not stay in a particular area” induce “numerous argo floats used in global scale study”? Please rewrite. Line 190, “contributed to by?” rewrite it. Line 322, “was applied it to ?” rewrite it. Line 331, “seems to be because the ” rewrite it. Line 332, “after that time” what is the time? 1998 and 2015? Line 396, “with a slight trend decrease” change it to be “with a slight decreasing trend”Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
This paper is about reconstructs three-dimensional (3D) ocean thermal structures using only satellite sea surface measurements for a higher spatial and longer temporal resolution than that of Argo and diagnoses the decadal variation of global MLD variability. However, I am not convinced due to the following reasons: (1) most previous studies using the similar method are over regional ocean, such as South China Sea[11], Arabian sea[12] and Southwestern Pacific[14], and (2) more surface factors do also affect the vertical temperature profile including ocean wind and net heat flux.
Major concerns:
As shown in Figure 1, in the mid-latitude of North Pacific ocean, the SLA is positive but STA is negative. The reason would be the ocean dynamic process induced by wind, such as, oceanic Rossiby waves and Ekman pumping (e.g. Chelton and Schlax, 1996, Science; Qiu Bo, 2002, Journal of Physical Oceanography). Therefore, it should be careful to directly relate the SLA to the ocean temperature but neglecting other ocean dynamic processes. Beside of the above dynamic processes, the SLA induced by the temperature profile can be described by the ‘seawater state equation’. This equation is for density in terms of Salinity, Temperature and Pressure. Then, the density is related to the volume and further related to the sea surface height. Therefore, the salinity and pressure would also be included. Validation should be added. Compare the STA from satellites to the STA from Argo in the Argo grids. As the method here, the Argo data is taken as “ground truth”. So the results from satellites should be verified with the Argo data. Then, provide the RMSE and correlationship between the STAs from Argo and calculated from satellites.
Minor concerns:
The usage of abbreviations should be consistent throughout the entire article. No explanations of SSH and SST can be found in the main article. Line 61 “In particular, Guinehut et al. [16] used statistical relationships between sea subsurface temperature and sea surface fields using in-situ observations, reanalysis data, and altimetry data.”The word “use” appears twice in this sentence, which could be avoided by substituting one with another word.
The nouns in Line 62 “Their studies” and Line 65 “In their study” are inconsistent Line 74 “They used” is ambiguous, would prefer “Guinehut et al. [16] used” Line 154 “we estimated the subsurface temperature using MLR based on only the sea surface data”“based on only” should be “solely based on” or “based solely on”
Line 189 “The life cycles of both types of ENSO (El Niño and La Niña) are dominantly contributed to by thermocline feedback representing the vertical advection process of STAs [36–39]”“to by” should be “by”
Line 371 “To evaluate the quantitative accuracy of the variability in the calculated MLD”I would prefer “To quantitatively evaluate the accuracy of the variability in the calculated MLD”
Line 505 “Therefore, the MLDA trends also show insignificant variability compared to other periods”Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Major comment:
In figure 2, STA does not correlate with SLA in high latitude regions. One can (mis)understand that the authors’ model is not available in high latitude regions. It would be better that the authors can add another figure which claims their model can be used there as well.
Minor comments:
Line 71: “applied” should be “applies”.
Line 345: “8º” should be “8ºN”.
Line 427: ”(b) and” should be “and (b)”.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
I do not think the authors really improve their work according to my previous comments, but only change some statement and/or wording. As the authors agreed the other dynamic processes impact the SLA but they still not exclude the effects in their formula, I cannot be convinced by this revision. So I have to suggest to reject this paper. I hope they can include the other factors in their formula mentioned in my previous comments.
Reviewer 3 Report
Well modified. I agree to publish this paper.
