Construction of the Long-Term Global Surface Water Extent Dataset Based on Water-NDVI Spatio-Temporal Parameter Set
Round 1
Reviewer 1 Report
Attached please find the comment.
Comments for author File: 
Comments.pdf
Author Response
Please see the attachment.
Reviewer 2 Report
The Authors present a crucial, critical problem of examining water surface changes by applying and testing their own developed, elaborated model. All tests and analyses are properly designed and discussed. The overall content is clear and understandable. From this point of view, the paper could be published after minor language proofreading (some blind spaces, missing commas, etc.). Nevertheless, I would appreciate some little explanation. As one can read, the Authors validate their model based on selected areas from different continents. The test objects (lakes) are situated in a precise location. I wonder whether the shape of particular borderlines and hence the objects' layouts play any role in the examination process? Is the method developed universal (and how much)? Can we estimate the overall universality? Why did the Authors choose those particular reservoirs? Such questions should be anyhow answered either in the introductory part or in the discussion. Except that, I cannot see any shortcomings and congratulate the Authors their exciting work!
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
The authors presented a long-term surface water extent dataset, namely GSWED, using an NDVI index. The dataset is from 2000 to 2018 with 8-day and 250-m temporal and spatial resolutions, respectively. While this dataset is of interests of many stakeholders, I felt that the current manuscript needs to undergo a major revision before can be considered for publication.
These are my main concerns:
- The authors need to do a more exhaustive literature review. While they presented the advantage of GSWED over existing dataset such as GIEMS (D15, D3), GSWE, etc in some features, however, when studying changes in climate these datasets have greater advantages over a longer period. Furthermore, they have not mentioned the current operation MODIS-based flood map from NASA https://floodmap.modaps.eosdis.nasa.gov/. It has near-real time daily images at 250-m resolution. Another improved product from it is the cloud-free MODIS flood map from Tran et al. 2019 (https://journals.ametsoc.org/jhm/article/20/11/2203/344130). How the authors will compare GSWED with these datasets?
- The temporal interpolation part need improving. A flow chart would be helpful in this part. Since the MOD09Q1 is an 8-day accumulative product, cloud should have been mitigated significantly in the product. However, cloud obscuring normally happens when there is storms or hurricanes which cause floods. If the authors do the temporal interpolation, there is likely a chance of missing flood extent as pointed out in Tran et al. 2019. Is there any other interpolation methods that help preserve the temporal scale?
- Figure 8, 9, as I understand, water pixel is classified as binary (i.e. yes or no) not as Low to High right?
- It is good that the authors have taken shadows, snow, bare soil into post-classification. But in global scale, there are still much more information needs to be taken into account to remove over/under-estimation such as topography, land use. It looks like the authors have validated GSWED over the Amazon forest, I'm interested in seeing the validation results over the Amazon.
- The authors should publish the dataset in some accessible repositories in order to be reviewed as well.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Line 232-235, change “I” to “we”.
Reviewer 3 Report
The authors have adequately addressed most of my comments. The content of the manuscript is much clearer now.
