Mapping of the Spatial Scope and Water Quality of Surface Water Based on the Google Earth Engine Cloud Platform and Landsat Time Series

Round 1

Reviewer 1 Report

The authors aim to assess spatial extent and water quality parameters of surface water bodies using Landsat satellite remote sensing data. They execute their analysis on the Google Earth Engine Platform. They build a case study on a Emergency Geomatics Service (EGS), and focus on Secchi disk, chlorophyll, and suspended solids.

Already the naming of the EGS led me astray at first. The EGS is the Candian agency that developed a mapping methodology (citation r17), which likely was adopted here. This needs to be made more clear, what the authors developed and what the outcomes of this study are.

Overall, the study seems to be an applied case study, with no novel information for remote sensing practitioners apart from applying documented methods to the specific case study area, it might be more suited to a water resources or generl sustainability related journal. The main outcome is showing that water resources in the case study province seem to be improving, but the processes pertaining to that are also not elaborated on.

The manuscript is generally well written, but needs to some reorganisation, especially separating results and discussion material and more clear methods section with a more robust error measurement section (for example, against which data was the error measured?). It could be the JRC data, but as the JRC data is also used in the surface water classification training data, this means the modelling and scoring process is already completely biased.

I am not sure, why the authors state in the abstract: "applied to water quality parameter inversion" ... this seems to continue with some other not fully clear concepts, such as: "Surface water extraction ... maps", as I assume the authors meant maps created to show the extent of water surface, not where water is extracted? Overall, I would recommend to be more precise and choose a term such as water bodies, surface water bodies, or surface water extent.

Still in the abstract: "The results show that the surface water in the study area from 1990 to 2022 could be accurately extracted." I can only reiterate, this needs to be re-articulated, there is no water extraction, but only the extent or the surface area of the water bodies.

I believe the authors refer with "the inversion results of the water quality parameters" to a regression or similar statistical mapping.

- Review claim that surface water area has shrunk ([2-4]): The citations 2-4 refer to reservoirs (not natural lakes or rivers), citation 3 and 4 refer to flood extreme events. Please rethink how to reframe your statements and with which citations to underpin those claims.

- p.2, ll.53: I roughly guess what the authors mean with "dependence problem is shielded" e.g. "that hardware is not a problem anymore", please add reference to underpin claim

- p.2, ll.55: "Chinese and foreign scholars" not sure why this needed to be emphasized, just "scholars" or "in the literature"

- p.4, ll.157-158: "the Emergency Geomatics Service (EGS) operational surface water mapping algorithm [17] was used to extract the surface water coverage area, and the verified model was utilized to retrieve the Chl-a, SS, and SDD" , firstly, citing singular studies as "verified models" without showing broader scientific consensus is already shaky. In addition, the study cited for suspended solids (r21) is only focusing on lakes. The generalisation to all water bodies is not very robust.

- p.5, l.176: .. we shrank ... in the “shp” file ... is a rather unprofessional reference to the processing related to the vector dataset containing the water bodies extents. How about removed, clipped?

- p.6, l. 189: JRC Global Surface Water Mapping Layers v1.4 and other datasets are NOT "obtained" from GEE! It is important to point to and cite the original sources of these datasets. It should then be articulated clearly that the datasets are also (made) accessible on the GEE platform.

- p.7, ll.234: water quality parameter inversion models review source references, the look like already "calibrated" regression formulas (which is literally named on the lines ll:466..)

- p.8, ll.257: throughout the manuscript the authors refer to "EGS operational surface water mapping algorithm", but it is not clear if their analysis describes the algorithm (which is cited r17/r18), or if these are two different/complementary water surface mapping algorithms?

- p8, section 4.1: The paragraph is not clear. How can the authors refer to a map lot (Fig.4) and claim that this shows how accurate their algorithm is? "It can be seen from Figure 4 that the surface water information for the rivers, lakes, and reservoirs in Zhejiang Province was accurately extracted using the EGS algorithm. The spatial location was more accurate, and the editing scope was perspicuous." This phrasing is not helpful and should be removed. The section goes on refering to figure 5 (also only maps) to state "The integrity and continuity of the surface water extraction were good, and the main river system 267 characteristics in Zhejiang Province were well preserved. The rivers and lakes were extracted more completely and accurately." A metrics table (table 3) is then shown later on page 9.

Section 4 is "Results", and only the results should be reported, but not discussed. Here they are discussed and already evaluated without any scoring or objective metrics. The chapter 4 should be slimmed down to only report results without interpretation!

- p.9, ll. 284-287: the section "In order to test the accuracy of the extraction results, the kappa coefficient and the overall accuracy (OA) were calculated by setting verification points to verify the quantitative accuracy of the classification results. The Kappa coefficients for each year were  greater than 0.90, and the OA values were greater than 0.95 (Table 3), indicating that the water information extraction results are credible." However, the metrics "Overall Accuracy (OA)" and "Kappa" are not defined. Especially, OA is not clear if it is matching surface area, matching surface boundary geometry, and also for Kappa no reference data is elaborated against what the metrics are tested. This needs to be much more clearly described in the methodology section, when only very shortly a coefficient of variation (CV) is mentioned, but it is not clear how the test data is selected and how meaningful these scores are.

- Results section, Sankey diagram of Secchi disk depth: I would recommend to invert the color scale, as high Secchi disk depth implied higher visibility and thus clearer water. Wouldn't that be better to have this also in nice blue instead of red?

- the Discussion section 5 starts with another table of statistics, whioch obviously belongs into the results section! Please rearrange.

- Figure 13:  "the GEE JRC database" I like to reiterate to a former comment, it is not a GEE dataset, it a JRC dataset that is acessible via GEE, that should be clearly documented!

- section 5.2: It is good to have section called "Uncertainty of research results", however this really is only a very superficial discussion of a few general points and doesn't contain any hints and quantitative uncertainty measures. This should be rename maybe "Limitations of the study"! Matter of fact, uncertainty quantification could have been much more robustly incorporated into the methods section, where the CV is the only half-heartedly described measure of error for this study.

- section 6 "Conclusions": Again the paragraph starts with a unnecessary emphasis on the GEE platform. In this study, nothing is special or in detail documented about using GEE, except that it was obviously easier to process many years of satellite data. There is one short reference to Random Forest on ll.199, and then at ll.440 without any additional explanation, the "validated inversion models" for water quality are not overly involved regression models. Furthermore, the study cited for suspended solids is only focusing on lakes. The generalisation to all water bodies is not very robust.

None of the used datasets are properly cited in the references! This must be improved.

Suddenly, in table 5 Sentinel 2 data is listed. This is unclear how this is now related to the study.

 

 

 

 

Author Response

Thank you for your valuable comments, the response letter is shown as attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

My main concerns are the complete lack of data on the water regime of the rivers, lakes and reservoirs under study. There is no information on the variability of water levels and discharges at the different seasons. It is unclear how the space images used correspond in time to the phases of the hydrological regime - floods and low flows. Without these data, it is impossible to assess the accuracy of inter-annual comparisons of water quality characteristics.

Minor comments:

Line 117 The purpose of the paper must be corrected

Line 144 such as the occurrence frequency and absolute change rate of the water body

 The meaning of the phrase is not clear. Please, identify what characteristic of water body you mean

Figure 2

RF method is mentioned, but not described even brifly

The figures 4, 7, 9, 11 are not informative. Please, find the other way to show the accuracy of the method used. Ay be it is better to use small areas on large scale, as on the Figure 5

Line 284 In order to test the accuracy of the extraction results, the kappa coefficient and the overall accuracy (OA)

Please determine kappa and OA (formula, citation)

Table 4 Water surface area changes seasonally. Please, indicate the seasons of the surveys (images) used

Line 303 but there were some differences in individual cities.

Why the information for the cities (not districts) is provided??

Line 315 The CV values of Hangzhou and Wenzhou were low (0.05), and the CV values of Lishui and Quzhou were higher (0.28 and 0.23, respectively).

Please, explain the cause of such difference in CV

Figures 8B, 10B, 12B are excessive

Author Response

Thank you for your valuable comments, the response letter is shown as attachment.

Author Response File: Author Response.pdf