Evaluation of Multi-Incidence Angle Polarimetric Gaofen-3 SAR Wave Mode Data for Significant Wave Height Retrieval
Round 1
Reviewer 1 Report
It is an interesting work to evaluate the performance of polarimetric Gaofen-3 SAR wave mode data with multi-incidence angle for significant wave height retrieval. Polarimetric SAR has been demonstrated to be very effective for SAR SWH estimation. However, how to fully utilize the SAR polarimetry information to achieve the optimal SWH estimation? There is still controversy on this issue. Also, the effect of incidence angle needs to be further analyzed, and the impact of radiometric calibration on SWH inversion needs to be deeply discussed. This work is helpful in accurately estimating SWH from Gaofen-3 wave mode SAR. But some issues needs to be clarified before its publication.
1. Gaofen-3 wave mode SAR images are selected for open ocean wave observations, but Fig. 1 shows some images are over the lands. What about these data? Please check the map of the selected Gaofen-3 wave mode data.
2. For Fig. 6(b), the temporal and spatial scales of the ERA5 dataset used for the validation are not clear. Please give some details of the data source.
3. Why are the incidence angles divided in the way as shown in Table I. Please explain the division criteria.
4. Page 12, Line 442: “Subsection” should be consistent in the paper. Please change to "Section".
5. Page 18-19, Line 588-590: “The radiometric calibration improves the performance of Gaofen-3 SWH inversion although not so significantly as for wind inversion.” please provide evidence to support this conclusion.
Author Response
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Comments
1. There should be a proper reason for the selection of the SR features used in the proposed technique, was there some kind of sensitivity analysis performed, no details on the selection criteria of the said features in line 144- 149 is provided.
2. What specific values of calibration constant were used? Were different values of this constant explored for the experiments?
3. Clearly mention that the Gaussian fit function is denoted by C or something like this.
4. One of the main components "GPR is a machine learning model" is not described completely/adequately. following should be considered:
1. How the model weights were selected
2. What features were inputted and what transformations were used
3. Which loss and optimisation functions were used
4. Epochs and bach size if any
5. Was the model pre-trained, if not how many training iterations were used?
6. No description of the kernel functions used is provided
7. Finally training, validation and testing ROC/ accuracy curves were not provided.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
The authors report on the course of analyzes aimed at optimization of the algorithms processing the satellite signal produced by the Synthetic Aperture Radar (SAR) technique into the value of the significant wave height (SWH). I consider the manuscript an important contribution to oceanography in the field of describing the dynamics of the sea surface. The use of satellite methods makes it possible to analyze the state of the sea surface in large areas and quickly update descriptions of wave fields. I would like to note that the SWH is an important parameter, but personally I would be interested in the statistical slope distribution of the wave slopes, which is necessary in the analysis of processes in the sea using optical methods.
From the formal point of view, the manuscript is well organized. However, I have the impression that the interest among oceanographers in the material presented in the manuscript is not broad due to the relatively hermetic nature of the terms used.
Check on line 210 if "m" should be replaced with "meters" and if on line 470 there should be a space in the place of the dash.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
