RFim: A Real-Time Inundation Extent Model for Large Floodplains Based on Remote Sensing Big Data and Water Level Observations
Round 1
Reviewer 1 Report
The major flaw of this work is the setup of the manuscript objective: the creation of a real-time flood inundation extent without using satellite imagery (but that relies on satellite imagery to achieve this) from just water levels. In general, the research question and explicit research objectives need to be more clearly stated upfront because as it stands, a reader needs to go back and forth through the manuscript to surmise what the paper aims to accomplish since this was not clearly stated upfront. Other major questions that need to be clarified:
if the water level data is available until 2016, why was only imagery for 2002-2011 used in the study?
what does RFim stand for? I couldn't find anywhere in the manuscript that explicitly explains the name of the simulation model
Lines 140-147 discuss the methodology in brief, but does not explain how the model is then applied in other locations and regions of the world with scarcer water level data?
Lines 182-188 I do not understand how the relating of inundation level to pixel values was done which makes the remainder of the paper very difficult to follow given this is probably the most important contribution of this paper yet it is very unclear how this was performed statistically.
Until the issue discussed in comment 4 gets resolved, the paper is not valid further on.
Author Response
Dear Reviewer,
Thank you for your insight comments and suggestions. We have modified the manuscript accordingly. We trust that all of your comments have been addressed accordingly in the revised manuscript. If you have further suggestions for changes, please let us know. The detailed corrections are listed below point by point:
Please see the attachment.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Brief summary:
The topic and research presented in the manuscript entitled RFim: a real-time flood inundation model based on remote sensing big data is of great interest and the questions being addressed are practically important for flood management and mitigation. The authors demonstrate a solid knowledge of the presented issue and data processing. The methods applied are appropriate. The presented approach for flood modeling, which is based on the combination of remote sensing images and daily water level observations, is innovative and brings new possibilities and prospects for more accurate large-scale flood inundation prediction. Although the presented model has certain limitations, its real-time applicability is promising for future, especially, in connection with the progress in high-resolution mappers and their deployment on satellites producing images with higher resolutions and shorter revisit cycles. Based on my review, I suggest a minor revision providing authors with some comments.
General comments:
Title of the paper – At this stage, is it appropriate to call the model “real-time” when real-time hourly/daily satellite images are not available yet and only historical images or past daily water levels were used? Is not it then near real-time or potentially real-time model? This issue needs to be resolved and clear information needs to be provided for the reader of this paper. Also in the text, the model is sometimes called “real-time” (e.g. lines 82-83) and sometimes “near real-time” (e.g. lines 357, 384-385) or having “great potential for real-time flood simulation” (line 390).
Abstract – To be more concrete, include the word “large-scale” or “large floodplain” in this sentence: e.g. “This approach has been applied in East Dongting Lake, the study area representing a large floodplain, for inundation simulation at a 30 m resolution.
Moreover, consider the suitability of using “large-scale” or “large floodplain” also in the title of the paper, in particular: “RFim: a real-time large-scale flood inundation model based on remote sensing big data” or “RFim: a real-time flood inundation model for large floodplains based on remote sensing big data” as well as consider including the information on the use of in situ water level observations, as another essential model input data except for remote sensing images, in the title of the paper: e.g. “RFim: a real-time flood inundation model for large floodplains based on remote sensing big data and water level observations.
1. Introduction – The main aim of the paper as well as its partial objectives are missing. Please, specify them in the end of this section.
5. Analysis / 6. Discussion – I suggest to join section 5. Analysis with section 6. Discussion into one section 5. Discussion. The reason is that in both sections you deal with the limitations of the presented approach which are usually included in the Discussion section.
6. Discussion – The authors should also discuss how the results/model differs from other similar studies or alternative opinions that can be found in recent literature. In this respect, there is no publication stated in the Discussion section.
Finally, the authors should also deal with the generality and reproducibility of the research outcomes. Is the proposed method reproducible for other flood-prone areas (e.g. large rivers) than water bodies? Furthermore, is it possible to simultaneously incorporate daily water level observations from more than one gauging station into the proposed model?
Specific comments
Lines 105-106 – It would be appropriate to specify that such resolutions (25-100 m) are used for large-scale flood inundation mapping, e.g. “Generally, large-scale flood inundation models run at …”
Figure 4 and lines 153-154 – Specify in the text what resampling method/technique was used here.
Line 250 – I think the word “was” is missing in this part of the sentence: “in every grid research area was divided into, …”
Author Response
Dear Reviewer,
Thank you for your insight comments and suggestions. We have modified the manuscript accordingly. We trust that all of your comments have been addressed accordingly in the revised manuscript. If you have further suggestions for changes, please let us know. The detailed corrections are listed below point by point:
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
Comments:
The main topic of this manuscript is a new model for real-time flood inundation prediction
using remote sensing big data. In general, the manuscript is good. The introduction provides
adequate references, the study area and the data are properly described. The methodology is a key aspect of this work. This reviewer suggests improving point (5) “Establishing the relationship
between the inundation extent and the water level”. The results are quite promising in this target
area. However, it could be positive to add some considerations about the application of this
method at a global scale. The conclusions can be improved. As such, this work could be published after addressing some considerations.
Specific comments:
• This reviewer would suggest stating more clearly “why” and “how” the authors claim
that the method is “real-time” and it is based on “big-data”.
• Figure 2 could be improved so as to show the flowchart is a more “friendly” and “nice”
way to the potential readers.
• Could RFim method be used in different target areas? What type of conditions are
required?
• The correlation between Fig. 11 and Fig 12. is clear. Could the authors further explain
this behaviour? Could the authors explain how to apply this method to different
topographic conditions i.e. surface slopes? It is required to better explain the relationship
between the water level and flood inundation extend.
• Overall, this reviewer suggests reviewing the paper, trying to highlight the most important
concepts to help the readers in the interpretation of the results.
• Line 99: Check error
• Adding colour to Fig. 10?
• Improve Fig. 14
Author Response
Dear Reviewer,
Thank you for your insight comments and suggestions. We have modified the manuscript accordingly. We trust that all of your comments have been addressed accordingly in the revised manuscript. If you have further suggestions for changes, please let us know. The detailed corrections are listed below point by point:
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 4 Report
The authors develop a method for fast computation of flood-inundation extent in a large floodplain, which relies on establishing a relationship between water level data of a gauging station and flood extent retrieved from Landsat imagery. It’s a simple model with several drawbacks (only flood extent prediction, no extrapolation capabilities, need for gauging stations, etc.), which limit its general applicability. Nonetheless, it can still be useful for similar case studies as the one described in the paper. I think that the topic of the article is within scope of Remote Sensing. The manuscript is reasonably well-written. I have a few questions, suggestions and minor comments, intended to help improving the clarity of some parts of the paper, and that I would like to see addressed before I recommend publication. I provide more details below.
Title: The authors do highlight the limitations of their method in the paper, but the title chosen suggests that this is a general flood inundation modelling approach, that could be used in any riverine context and capable of predicting flood dynamics. I suggest adding to the title something like “large floodplains” to clarify the type of domain, and changing “flood inundation model” to “flood extent”.
Abstract, line 12: DEM is not an approach in itself, I think this should read “DEM-Based Methods".
Introduction, lines 50-52: “2D models require many parametric inputs, some of which may be unavailable”. In most 2D models the only parameter is the bottom roughness coefficient. I think that I understand what the authors mean (the more complex model set up, which requires geometry, boundary conditions data, etc.), but this is not clear from the sentence. They also indicate that 3D models have the same drawbacks, but it could be noted that they are more significant (data needs, computational time, etc. significantly increase with respect to a 2D model).
Introduction, lines 53-58. The authors only mention here the “planar method” (or “bathtub method”), but there are more sophisticated DEM-based methods, which consider the drainage connectivity, that could be mentioned here.
Line 98: There is a typo “.”
Lines 150-155: It would be useful to note the size (i.e., how many 30m grid cells cover the study area?)
Figure 4 caption: there is no (a) and (b) indicated in the figure.
Lines 163-181: The authors have included an appendix regarding the water extraction accuracy using NDWI, I think there should be a reference to it here.
Lines 191-202: In some parts of the text I find confusing the use of the term “grid”. It would be more precise to refer to a “cell/element of a grid” or “grid cell” rather than the grid (the grid being the collection of all cells).
Line 209: I think that the criteria to define the abnormal records is not complete. In Figure 7, for example, if 22.64 was classified initially as wet, instead of 22.02, which would be the abnormal record: 22.64 or 25.13 m? This is not clear to me from the method’s description.
Lines 252-256: This explanation could be improved. When the authors refer to “composite images”, are they referring to combine Landsat data from multiple bands? When they refer to mosaicking the images I assume that this is simply based on the mapping coordinates to produce a larger image, correct? In Figure 10, if the composite images do not correspond to a single day, it’s not clear to me how the predictions shown in (b), (d), (f) and (h) are made (which water level values from the gauging station are used)?
Figure 14 and Tables 3 and 4: Please clarify what does “percentage” refer to. There are two different “percentages” in the main and secondary y axes in Figure 14.
Line 350: 1000 km2, where does this number come from? The run time of a flood simulation is directly related to extent of the spatial domain and the spatial resolution used, but also model complexity plays a critical role. For example, 2D hydrodynamic models can solve simplified forms of the 2-D shallow water equations, instead of the full form, allowing for larger domains to be simulated. A paradigmatic example of this could be Global Flood Models (GFMs), which are now a practical reality.
Line 396-397: Please rewrite (or expand) the last sentence of section 7, I find it’s not clear what you mean by “the limitation on the water level of predicting inundation can be removed and the impact of topographic changes will be avoided”.
Author Response
Dear Reviewer,
Thank you for your insight comments and suggestions. We have modified the manuscript accordingly. We trust that all of your comments have been addressed accordingly in the revised manuscript. If you have further suggestions for changes, please let us know. The detailed corrections are listed below point by point:
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors properly addressed my concerns regarding the manuscript and thus improved the paper.
