Inversion and Driving Force Analysis of Nutrient Concentrations in the Ecosystem of the Shenzhen-Hong Kong Bay Area

Round 1

Reviewer 1 Report

      The author Inverse the nutrient concentrations in the ecosystem of the Shenzhen-Hong Kong Bay Area by using four machine learning methods. Following comments should be concerned.

(1) The main doubt is that the author uses only 22 station data for model construction and verification, which means that only 5 points are used for verification (20%), which makes the results very uncertain.

(2) Line 140-142, "conducts water quality monitoring at 76 open water monitoring stations every month"? 76 stations or 22 stations? please confirm.

(3) In section 2.3.1, please use a table to show which indicators are selected to build the model？only the Band of Landsat imagery?

(4) In the Discussion section, please make an in-depth discussion on the advantages of model comparison and the basis for index selection. Furthermore, please analyze the reasons for the annual and interannual changes of nutrients. For example, TN and TP significantly decreased after 2009.

 

 

 

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

This study use remote sensing technology to analyze the driving forces of changes in water quality (total nitrogen (TN) and total phosphorus (TP)) from multiple aspects based on inversion results in the Shenzhen-Hong Kong bay area. This study evaluated the modeling effects of four machine learning methods. The results show that the machine learning algorithms are suitable for the inversion of non-optical activity parameters of coastal water bodies, and also shows the potential of remote sensing in large-scale, long-term water quality monitoring and comprehensive analysis of the driving forces.

In general, this manuscript presents interesting results and I think it would be suitable for publishing. However, I suggest improving the following points:

-    Improve the evaluation of model performance. To evaluate the models you have just used two statistics r and RMSE (section 2.3.6). Add some more statistics to evaluate the performance of the models (e.g. R2, the mean absolute difference (MAE) or the mean absolute percent difference (MAPE)) and justify why these statistics are used.

-    Section of conclusions should be rewritten. The conclusions made are a simple summary, do not add value to the work, do not highlight important results, limitations and future research and broader impacts. The Conclusions section should include reasoning as to why the methods and results are recommended.

Specific comments:

1.      In the result values (e.g. line 20 and 21) use two or three decimal places. Four is excessive. Modify this throughout the manuscript.

2.      In the keywords add some more such as: Inversion or Driving Force Analysis.

3.      Improve the quality of figure 1. Make it larger by occupying the entire available width. Add reference to the north.

4.      Line 156 and 159: add reference for CHIRPS precipitation data and for ERA5 wind data.

5.      In Table 1, under BPNN, what layers does "Number of layers" refer to? Does it refer to hidden layers? If so, please indicate in the table. What do you mean by "size of the first, second and third layer"? do you mean the number of neurons? clarify this in the table.

6.      Provide a brief analysis of the results shown in Figure 2.

7.      Improve the quality of figure 3. It is difficult to read the legends.

8.      Improve the quality of figures 4 and 5. They look a bit pixelated.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The author has made substantial reversion according to the comments, therefore, I recommend accept the paper in current form.