Dynamic Inversion of Inland Aquaculture Water Quality Based on UAVs-WSN Spectral Analysis

Round 1

Reviewer 1 Report

The manuscript entitled "Dynamic inversion of aquaculture water quality based on UAVs-WSN spectral analysis" is well written and deals with remote sensing of water and data processing. The remote sensing of water is much more challenging that research on vegetation or soil due to the low response of that target as well as special requirements for the sun-sensor configuration what brings several data processing issues and noise removals. Thus, water can be mixed with either soil or vegetation components or even both what makes this target a real challenging task.

Authors also provide good insights for another emerging task that is CNN and therefore I judge the research topic being suitable for the Remote Sensing journal.

However, authors could provide a methodological flowchart so that the sections (in the Methods section) could be better followed and replicated (for those interested in the subject). I mean this because there is a mix of methods, sensors, and data processing steps that could be better explained. Another important aspect is the fact the there is only one study area with a size area of one hectare. Is this the usual size of the fish farms? Was this study restricted to this area only? It may not be representative.

Another major concern is the fact that there are not enough sample points to allow good Geostatistic and modeling (when we thought that some of the points should be dedicated to validation and tests. Additionally, if two missions were performed (two dates!), why change detection was not performed? And results presented for each date? These aspects should be justified accordingly in the revised version of the manuscript.

Concerns:

L14-15: define aquaculture and importance;

L16-17: complex process? Avoid such terms;

L19: almost a real time processing?

L22-23: at a regional scale? I would say this work is local but potential would be regional;

L31: RMSE;

L32: in fish farms at a regional scale? Small water reservoirs;

L38: explain why water quality is important (and in which temporal scale this information would be required and how much remote sensing could contribute with that compared to ground level sensors; this aspect could be explored within precision agriculture context;

L59: explain again UAV and WSN acronyms;

L60: please provide references for that affirmation;

L68-70: please explain briefly the importance of DO and TUB;

L70: present here both DO and TUB abbreviations;

L72: please explain DNS-DNNs;

L76: add here possible drawback, although the study area is small…;

L78: change complicated to difficult due to low reflectance response;

L200: is the usual size for a fish farm; why choosing only one area? does it have applicability?

L202: please provide scientific name and fish density; also describe better the area, how deep, how stable it is;

L203: high requirements in which sense?

L263: explain better the 924 sets;

L271: Pearson?

L306: how many pixels? Size?

L306-307: how many bands? how many variables?

L327: please explain better the input data;

L346-347: seems to be methodology (same for L379-380);

Figure7: change title of the axis, example: Predicted Value (ppm) rather than Predicted Value / ppm (keep it for other figures as well);

Figures7,8: which date acquisition? The lat-long axis does not make much sense, please change it to meters (m);

L441-447: new information shall be presented in the discussion section first; otherwise, change conclusions to conclusion and future research perspectives;

In general, the paper is very interesting. Congratulations.

Thank you for the opportunity to evaluate this research.

Author Response

Please find attached our detailed response letter.

Author Response File: Author Response.docx

Reviewer 2 Report

The manuscript is in general interesting. However, the writing and organization of the paper is relatively poor which makes it hard to follow. Moreover, more experiments are required to better understand the effectiveness and robustness of the proposed method. I suggest to apply the method on multitemporal imagery rather than a single image. I am providing some more detailed comments as follows:

1-     Lines 22-23: the sentence is not complete. Please revise.

2-     The way it is written, the advantages of satellite remote sensing such as large spatial extent, etc. are totally neglected.

3-     Line 36: it should be ‘global food production...”. The writing of the paper requires significant improvement.

4-     Line 53: what do you mean by “distribution effect”? are you referring to the spatial distribution?

5-     Lines 59-62: please provide some references here.

6-     Line 75: is the aim of this water quality monitoring is limited just to agricultural purposes?

7-     Point #3 in the introduction: the results shouldn’t be provided in the introduction. The organization of the paper needs improvement.

8-     Section 2: it should be “related works”. I strongly suggest to improve the writing of the paper.

9-     More relevant and recent studies on the uses of spectral data to monitor in-water constituents should be added to the paper.

10-  Lines 151-152: This statement is not fully correct. There has been a lot of progress in monitoring inland water. Of course, no method is perfect but the advantages of satellite remote sensing should be carefully discussed. In general, the use of different source of spectral data is a tradeoff between different factors (e.g., spatial coverage, temporal resolution, etc.).

11-  Lines 171-174: all the methods should be properly cited. I would suggest to study some of these works:

A comprehensive review on water quality parameters estimation using remote sensing techniques

Novel Spectra-Derived Features for Empirical Retrieval of Water Quality Parameters: Demonstrations for OLI, MSI, and OLCI Sensors

Mapping water quality parameters with sentinel-3 ocean and land colour instrument imagery in the Baltic Sea

First experiences in mapping lake water quality parameters with sentinel-2 MSI imagery

12-  Table 1 and the text before: it sounds like the advertisement of the sensor! The actual configuration of the sensor used in this study should be mentioned (spectral, spatial resolution, etc.) and the unnecessary material should be removed.

13-  What about the radiometric calibration of the spectral sensor onboard the UAV? This needs clarification whether you have used calibration panels or not.

14-  Line 236: I suggest to show samples of the collected spectra and their associated values of water quality parameters.

15-  Table 2. How did you end up with these specific band combinations? There are methods to automatically identify the optimal band combination of the spectral features such as optimal band ratio analysis (OBRA). See these papers:

Evaluating the potential for remote bathymetric mapping of a turbid, sand‐bed river: 1. Field spectroscopy and radiative transfer modeling

Multiple Optimal Depth Predictors Analysis (MODPA) for river bathymetry: Findings from spectroradiometry, simulations, and satellite imagery

Spectrally based remote sensing of river bathymetry

16-  Fig. 8: the date of the image should be clarified. Did you apply the method on other dates? The robustness of the method should be investigated by applying the model on the multitemporal images.

Author Response

Please find attached our detailed response letter. Thanks.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The revised version of the manuscript contains several improvements compared to the first version. The authors provided additional references, figures, and sentences in order to allow reproducible research. The answer letter also covers the majority of my concerns. Therefore, I do recommend the publication of this manuscript.

When still possible, please prepare some additional improvements (not really necessary, but it would look better if you perform it). I would suggest increasing the font size of the text a bit in the flowchart (Figure1). Please check how it looks by changing the color of the internal text from white to black. 

In Figure 2, please enhance the caption description. I would suggest, location of the study area within national (a), province (b) and local (c) context. Indicate also which image (sensor) and RGB composition was used. Please take this care of such kind of description in all figures and tables.

In fact, such changes are not really necessary, but it is just for better looking. This is why I am selecting the option "Accept in present form". 

Thank you for the opportunity to evaluate this paper. Congratulations.

Author Response

Please find attached our response letter. Thanks.

Author Response File: Author Response.docx

Reviewer 2 Report

The paper is improved with respect to the initial submission. However, I have still some comments that should be clarified before any possible publication:

 

The calibration panel is not clear on the image provided by authors. Is it a white panel? Which band is shown on the provided image?

 

I am not still convinced that why authors don't apply the method on the other dates. The results on a single image can be subjective and probably not reflecting the pros and cons of the method. specially, it is important to understand where the method fails? 

 

The authors added some citations but not enough/properly. For instance, they claimed in the response letter that they use OBRA for selection of the wavelengths, however, the method is not mentioned and cited in the paper. Moreover, the recent studies on spectral analysis of water quality parameters should be cited thoroughly. Some of the appropriate references were provided in the previous comments which are recommended to be fully considered.

Author Response

Please find attached our response letter. Thanks.

Author Response File: Author Response.docx