Dynamic Water Quality Changes in the Main Stream of the Yangtze River from Multi-Source Remote Sensing Data
Round 1
Reviewer 1 Report
Most of my comments and edits are in the attached document. However, I will add here that there seems to be scientific strength and rigor in the remote sensing data analysis.
However, further explanation on the amount of remote sensing data that was used in the analysis will provide some additional context to the methodology. For example, 1) what percentage of the data was usable for this study or 2) this methodology consisted of selecting X% of the RS data as the training series in the calibration/verification phase of the work…….. etc.
In general, the paper is technically sound, and the mathematical calculations/models are appropriate for this study. However, clarifications around the other data that was collected (flow, temperature, and water level) is need.
Lastly, the discussion section needs to be revised to include several citations (currently, there are no citations in the discussion) and data from the literature needs to be incorporated in the discussion to strengthen this section.
Comments for author File: 
Comments.docx
Author Response
Please refer to the attachment
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This manuscript develops TN and TP prediction models based on Landsat 8 and Sentinel 2A images for the Yangtze River (China). Authors computed a wide set of bands combinations (feature types) for predicting TN and TP based on field measurements. Inversion models were also developed and applied for a multitemporal assessment of both water quality parameters.
The objective of the manuscript is interesting. I have several uncertainties regarding the water quality parameters field measurements, the process of training/validation and multitemporal prediction, and the range of water quality parameters respect to the accuracy of the models.
Line 117. "In research" or "In this research". An overall review of the writing style could be useful.
Line 119. What do you mean with “water conservancy resources”?
Line 127. Include a table with the exact number of images and their acquisition dates?
Line 139. What WTW water quality analyser? Further details are needed because no clear information of the analytical procedure is provided. Is you are using a field multiparametric measurement device, it should be carefully calibrated in order to guarantee that measured values are comparable to those obtained with standard laboratory based analytical methods. I always recommend a verification of field measurements with laboratory-based water quality analyses.
Lines 145-147. Specify further details of sampling and analytical procedures.
Section 2.3.2. Include additional statistics such as normalized RMSE in all cases and Residual Predictive Deviation for the predictions.
Section3. Include further details of measured water quality parameters.
Line 212 et seq. According to this information, your TN range was 3.5 mg/L and TP range was 0.5 mg/L. I have recommended the computation of the normalized RMSE because your models reach good Pearson correlation coefficient values, but RMSE is not satisfactory taking into consideration the range of your measurement. This is particularly concerning for TP.
Line 213. Provide references that justify your selection of a proportion 10:1.
Line 216. Specify what bands were employed?
Lin 239, Did you assess the generalization capability of your inversion models? I would like to know their performance with a second order polynomial function.
Table 1. Include nRMSE
Table 2. Include nRMSE and RPD.
Figure 5. The scattering of points in most of the graphs suggest significant uncertainties in the estimates. The value of R2 does not accurately reflect this dispersion.
Figure 6. Include the same plot with a lower order polynomial function in order to enrich the discussion of your results.
Section 4. In this section you apply your inversion models to a large number of images/dates. Some questions:
• What sensors/images were used?
• How to verify the accuracy of these new predictions?
• What is the impact of high nRMSE values on these new predictions?
Author Response
Please refer to the attachment
Author Response File: Author Response.pdf
Reviewer 3 Report
Dear Authors,
This article examines an interesting research topic for Remote Sensing application on sustainable water resources management, with the aim of analyzing the opportunity offered by Remote Sensing Data, to retrieve water quality particularly linked with Total nitrogen (TN) and total phosphorus (TP) as important indicators for water quality, in the main stream of the Yangtze River .
An interesting methodological approach is developed based on inversion model using single satellite data (Landsat 8 and Sentinel 2A) or joint based on multi source (both L8 and S2A data) integrated with measured data of the TN and TP parameters. The influence and correlation of hydrological and meteorological factors as well as human activities impacts on the water quality of the main stream of the Yangtze River is also analyzed.
The improved accuracy (in terms of R2 and RMSE) obtained with the joint inversion model confirm the utility of using multi-source satellite data, which can provide higher temporal frequency and improved radiometric/spectral information due to the configuration of different satellite platform.
The obtained results shows a promising way to obtain temporal/spatial Information on water quality as compared with traditional, ground-based measurements, which often considered laborious, time consuming and expensive.
Regarding the text:
In figure 2, line 236 the caption report Landsat 8 but according to the text should be Sentinel -2A, the figure number should be 3, being figure 2 for Landsat.
Author Response
Please refer to the attachment
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
NA
