An Exponential Algorithm for Bottom Reflectance Retrieval in Clear Optically Shallow Waters from Multispectral Imagery without Ground Data

Round 1

Reviewer 1 Report

General remarks:

There are visible some shortcomings in the structure of the paper, for example, the results section is very short whereas some paragraphs from the Discussion directly relates to some of your results. I suggest moving some relevant parts of the discussion to the results section.

Since the benthic habitat mapping term can be understood in different ways, I recommend defining your approach. Consider the definitions mentioned in:

Brown, C.J.; Smith, S.J.; Lawton, P.; Anderson, J.T. Benthic habitat mapping: A review of progress towards improved understanding of the spatial ecology of the seafloor using acoustic techniques. Estuarine, Coastal and Shelf Science 2011, 92, 502-520, doi:10.1016/j.ecss.2011.02.007

Lecours, V.; Devillers, R.; Schneider, D.C.; Lucieer, V.L.; Brown, C.J.; Edinger, E.N. Spatial scale and geographic context in benthic habitat mapping: review and future directions. Marine Ecology Progress Series 2015, 535, 259-284, doi:10.3354/meps11378

The Discussion also needs some improvements in relation to a comparison with the other research in the scientific discipline. For example:

Misiuk, B.; Brown, C.J.; Robert, K.; Lacharité, M. Harmonizing Multi-Source Sonar Backscatter Datasets for Seabed Mapping Using Bulk Shift Approaches. Remote Sensing 2020, 12, doi:10.3390/rs12040601

Janowski, L.; Madricardo, F.; Fogarin, S.; Kruss, A.; Molinaroli, E.; Kubowicz-Grajewska, A.; Tegowski, J. Spatial and Temporal Changes of Tidal Inlet Using Object-Based Image Analysis of Multibeam Echosounder Measurements: A Case from the Lagoon of Venice, Italy. Remote Sensing 2020, 12, 2117, doi:10.3390/rs12132117.

Some specific comments:

line 47: vague (so on).

line 105: it seems that a sentence should not begin with "and"

Figure 3b: Provide label units of the measured values

line 357: after all, a better explanation would be helpful. Despite the small coverage by other classes, to ensure transparency of the research, it would be useful to take all classes of benthic habitats in this area. 

lines 366-368: Vague. What do you mean by dark and light signals? 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

The study is interesting and authors providing a new modification to Stumpf et al. and Lee et al. algorithm to improve bottom reflectance. However, there are some concerns and suggestions that should be considered before publication. 

• The manuscript needs to be checked by a native English speaker. There are some confusing passages that I had to go back and forth to understand. It makes it very difficult for the reader to follow the text.
• The authors say, their approach is novel. However, this is not very clear from the texts. Merging bathymetric data and multi-spectral images are in general common, although, the algorithms may differ. I suggest the authors to high light and clarify this part. 
• It is not clear, how this model stands against the usual approach and algorithms. The authors say this is an improvement over Stumpf et al., and Lee et al., however, there is no clear comparisons. 
• In the Introduction, the authors need to add more literature that is more related to their study rather than using general RS literature. 
• Why the authors used two image sets to test ? It is not clear what benefit does it have. As a reader I am more interested to see how much this algorithm improves the accuracy compared to other methods rather than other satellite images.
• The accuracy matrices are not mentioned. How much would be the accuracy of the habitat type detection using this algorithm compared not using this algorithm ?
• The Allen Atlas has been used for comparisons, however, the atlas itself is not entirely based on in situ data and habitat detections have different accuracy. So, there is already some noise and inaccuracy entering the authors algorithm. How do authors go about this inherited inaccuracy in the data since you had no actual field observations ? please discuss this in the discussion section.  

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript presents a novel method for inverting bottom reflectance and bathymetry by integrating log-transformed ratio (blue/green) and a semi-analytical model. This work is interesting but there several major issues that require significant improvements before a possible publication. The paper should be restructured to separate methods, results, and datasets. Everything is mixed up in the current version. Moreover, the methods need better clarification. For instance, it not clear how some parameters (e.g., g) are calculated and also why the bottom reflectance retrievals are limited to 550 nm only? On the other hand, the proposed method should be compared with the state of the art methods. There is no comparison with other methods and also no in-situ data is provided for the bottom reflectance. Here are more detailed comments:

 

The title is not informative enough. This is because the novelty of work is not clear as bottom reflectance retrieval is an established topic. Here, the authors should revise the title to reflect the contribution of their work. Moreover, the developed methodology is limited to oceanic islands and reefs only?

 

Line 19: I suggest replacing “log rotation” with “log-transformed”.

 

Line 45: more appropriate definition of optically-shallow waters is needed (e.g., significant bottom-reflected radiance). This statement is not precise: “the clear oceanic water depth is less than 20 m, which the maximum depth the blue-green band can penetrate; the bottom is covered with coral reefs, seagrass, sand, and so on.”

 

Lines 48-49: this is not precise to say “Visible wavelengths have the characteristics of high atmospheric transmittance and a small attenuation coefficient for water”. In general, water has high attenuation given the strong absorption by pure water. This sentence should be revised to clarify that the attenuation is weaker in short wavelengths compared to the longer wavelengths.

 

Lines 52-53: both empirical and analytical methods can be applied to either multispectral or hyperspectral data. Please revise the sentence.

 

Lines 56-60: the physics-based inversion is not limited to the hyperspectral data only. There are several studies showing the applicability of such inversions using multispectral data. For instance, there are works with the physics-based Water Color Simulator (WASI) model (some of them recently published in Remote Sensing) that should be addressed.

 

Lines 95-96: A revision is required here. Please avoid mixing the bottom reflectance retrieval method with the accuracy assessment (these two should be discussed separately). The method and assessment metrics should be presented as standalone subsections. Then, seems you have both simulated and real image dataset to examine the effectiveness of the proposed method. Restructuring the manuscript can help the reader to better follow the paper.

 

Fig.1: this figure is out of place because none of the parameters presented in the flow chart are not defined yet. This flowchart might be moved to the last part of the methods. I suggest making it more generic and standalone. At the moment, it is very hard to follow all those notations and steps.

 

Line 125: “The distribution of the bottom reflectance spectrum” should be clarified. What does it mean and how do you calculate it?

 

Section 2.2: the equations require more clarification. There are several unclear points. For instance, how do you calculate g? the bottom reflectance should be wavelength dependent. Thus, how the reflectance at different wavelengths are derived from the model based on the blue/green ratio?

 

Lines 152-154: “This means that the properties are the same on one image scene, so that the content of the simulation shown in the following is a result at 0.2 mg/m3 of Chl-a [28 - 31].” This is unclear. The Chl-a concentration is fixed to 0.2 mg/m3?

 

Lines 198-199: “which provides a full-color image with a resolution of 0.5 m and a four-band multispectral image with a resolution of 2 m.” what do you mean by “full-color”? WorldView-2 provides 8 spectral bands. This is in contrast with what is written in the paper (4 bands). Only 4 bands are used instead of 8?

 

Lines 207-216: the radiometric calibration and sun glint removal techniques are unclear. These pre-processing steps should be clearly described in the method sections.

 

Line 262: the inversion of the bottom reflectance is performed only at 550 nm? This was also unclear in the methods section that how the bottom reflectance is inverted across the spectrum.

 

Fig. 8: the figure is out of place. This should be part of the methodology. The manuscript requires a significant amount of work to reorganize different sections and provide a clear description of the methods.

 

The accuracy assessment is unclear. The depth estimates and retrieved bottom reflectance should be compared with in-situ data. Moreover, the bottom reflectance is retrieved only for 550 nm? This undermines the full potential of multispectral data in the classification of bottom properties.

 

Lines 318-319: “The bottom reflectance and water depth inversion results of the two data sets were generally the same”. This statement is not precise, the WV-2 vs Landsat-8 depths are correlated but they are not the same!

 Line 338: why all of the in-situ depths are not considered for accuracy assessment and only some random samples are considered?

 

Section 4.4: here a new dataset is introduced! Datasets, methods, and results are mixed up and it is hard to follow the paper. I strongly recommend a major reorganization.

 

The proposed method should be compared with other available methods in the literature:

 

Water Column Correction for Coral Reef Studies by Remote Sensing

Mapping Substrate Types and Compositions in Shallow Streams

Remote Sensing of Shallow Coastal Benthic Substrates: In situ Spectra and Mapping of Eelgrass (Zostera marina) in the Gulf Islands National Park Reserve of Canada

Bottom Reflectance in Ocean Color Satellite Remote Sensing for Coral Reef Environments

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Thank you very much for the replies. Just please make sure your responses have been reflected in the manuscript. The point is to have those information in the manuscript so the reader won't have similar questions. 

Author Response

We have checked and corrected the revised manuscript carefully according to the comments suggested by the Reviewers point by point. All responses to the Reviewers’ questions were marked and reflected using the "Track Changes" function in our manuscript which we will upload this time.

Once again, thank you very much for your constructive comments and suggestions!

Author Response File: Author Response.docx

Reviewer 3 Report

The authors made major changes and replied in detail to my previous comments. The manuscript has been improved significantly. There are still few points to be considered before a possible publication: 

 

line 192: I had pointed out this issue also during the previous revision that the physics-based inversions are not limited to the processing of hyperspectral data only. They can be applied to multispectral images as well to retrieve the in-water constituents and bathymetry (in optically-shallow waters). Here are some reference works published in Remote Sensing based upon physics-based models applied on multispectral images:

Water Constituents and Water Depth Retrieval from Sentinel-2A—A First Evaluation in an Oligotrophic Lake

Physics-based Bathymetry and Water Quality Retrieval Using PlanetScope Imagery: Impacts of 2020 COVID-19 Lockdown and 2019 Extreme Flood in the Venice Lagoon

 

What is the motivation to exclude 4 bands of WorldView imagery in your analyses? I expect to take advantage of all bands. especially, red-edge band and others can contribute to the 

 

The choice of dark-object subtraction method for atmospheric correction should be better justified. This is while there are several advanced and sophisticated methods for mitigating the atmospheric artifacts.

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx