Satellite-Derived Bathymetry for Selected Shallow Maltese Coastal Zones
Round 1
Reviewer 1 Report
The manuscript is well written and the research topic is of interest to several readers from a wide variety of backgrounds. The idea of using satellite images to get bathymetric data is not novel, but it still has a lot of challenges to overcome.
I have a few concerns that should be responded to after recommending the work for publication:
- The introduction is too basic and the gap in knowledge regarding the acquisition of bathymetric data from satellite images is not clearly stated. No previous works are cited nor previous methodologies or results are criticized.
- Some interesting information is given about the study sites but it is not used later in the research, the authors may want to revise the pertinence of those paragraphs. Or even better, the authors may cross the hydrodynamic information they provide, with the morphological variations they recorded.
- Section 3 should be maned Methods. Also, in this section, a more comprehensive explanation of the algorithms is needed.
- An explanation is needed regarding the fact that all of the algorithms fail in shallow waters. In the introduction, the authors seem to develop this research to improve and facilitate the recording of bathymetric data in shallow waters, but there precisely is where the results obtained are worst.
- Also, in the introduction, it seems as if the SDB is presented as an alternative to in-situ measuring, but then in the conclusions, they are presented as complementary. Please change the wording to be coherent.
Author Response
We thank the reviewer for the comments. We followed the suggestions/comments and we feel that the manuscript is much stronger now. Reply to comments in the attached file
Author Response File: 
Author Response.docx
Reviewer 2 Report
The authors used optical satellite imagery from Sentinel-2 and in situ bathymetry data from a water surface drone, combined with a LiDAR survey, to conduct a bathymetric study of two typical shallow water areas in Malta. Machine learning methods such as random forest, K-star, multi-layer perceptron and linear regression were applied to the comparison of bathymetric model construction and the output accuracy tests were all satisfactory. However, a number of queries exist.
1. I believe that the references to previous studies in the Introduction section are too brief and need to be completed.
2. The LiDAR data mentioned in the paper were measured in 2018 and satellite data were imaged between 2020-2021, whereas the UAV data were real-time in-situ measurements. However, the paper also mentions that dynamic sedimentation along the coast can 'periodically' change the local seafloor topography. Therefore, I think it is worth discussing whether the temporal differences among the different data are sufficiently 'periodic' to cause some evolution of the seafloor topography.
3. Of the four machine learning methods chosen by the authors, the modelling accuracy results suggest K-star > random forest > multi-layer perceptron. However, in our general perception, as a machine learning method, the multilayer perceptron should be the latest and more accurate potential, followed by the random forest and then the K-star. I think it would have been more convincing, or something could have been identified from it, if the parameters and the tuning process of each method could have been listed. In addition, why there is an inversion of the measured accuracy and the generally perceived arithmetic in the experiments is also a question that would be interesting to be discussed.
Author Response
We thank the reviewer for the comments. We followed the suggestions/comments and we feel that the manuscript is much stronger now. Reply to comments in the attached file.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
The author has made detailed and corresponding changes to my comments, making the article fuller and more scientific in content. I agree to proceed with the acceptance.
