A Method to Construct an Environmental Vulnerability Model Based on Multi-Source Data to Evaluate the Hazard of Short-Term Precipitation-Induced Flooding

Round 1

Reviewer 1 Report (Previous Reviewer 1)

The present study proposes a method for short-term floods vulnerability evaluation. First, it is based on the analysis of remote sensing data for floods detection and land use classification. Then, the evaluation of indicators of exposure, sensitivity and adaptive capacity allows the vulnerability assessment. The proposed vulnerability framework is applied to the Henan Province.

Compared with the first version, the manuscript has been slightly improved and some figures have been better commented.

However, the manuscript still presents several criticalities.

1) Some parts of the manuscript remain not clear and seem not linked to the rest of the text. Hereafter, some examples:

-Pag.2, line 21: “Physical vulnerability is the probability of loss of disaster bearer under different disaster intensity, and is a quantitative estimation of disaster resistance of disaster bearer at a micro scale [15]. The social vulnerability of flood disasters is mainly studied in this paper. Currently, flood vulnerability assessment can be roughly classified into four categories.” In this context, what is the function of the statement: “The social vulnerability of flood disasters is mainly studied in this paper”? it is not linked to the sentence before neither to the successive!

-Pag 7: The indicators: Exposure, Sensitivity are defined positive and the Adaptive capacity negative but it is unclear to what authors refer. Probably to the formula (3)?

2) Some part of the text are redundant and repeat things already explained. For example:

-Pag.6. line6: “The Exposure-Sensitivity-Adaptive capacity framework is used to build an environmental vulnerability model for short-term precipitation-induced flooding.” Pag. 7, section 2.3.1. “In this paper, the Exposure-Sensitivity-Adaptive capacity framework is used as the vulnerability assessment system of the short-term flood hazard.”

-section 3.2.1: authors repeat the meaning of exposure, sensitivity, Adaptive capacity, composite vulnerability but they have already explained these parameters before, in section 2.3, page 7.

3) Some definition are unclear: for example what is a “disaster feature”, or a “Flood submerging feature”? probably, I suppose that is the areal extension of the flood and the water depth. In a scientific work, the language should be more rigorous and precise, instead authors often use generic and unclear definitions all over the text.

-Pag.7, formula (3): The meaning of:“f is the vulnerability index of rainstorm flood…” is not clear. Generally, Vulnerability refers to an exposed community or to a particular area/region.

4) other observations:

-section 2.2: In my opinion, the explanation of the methodology is not clear. First of all, authors should insert a synthetic definition of the method. Then, they should try to synthesize the rest of the paragraph deleting redundant information.

-Section 3.1, line3: Auhors state: “The buildings, mountain shadows and other land use features can be mistakenly identified as flood disaster feature using the traditional feature extraction method. The accuracy of the U-Net++ network is better than the traditional method.”  But, what is the traditional method? ii is not clear. It should be explained.

In my opinion, the manuscript should be completely revised to improve comprehensibility and conciseness and an extensive editing of English language and style is required.

Author Response

1. Some parts of the manuscript remain not clear and seem not linked to the rest of the text. Hereafter, some examples:

-Pag.2, line 21: “Physical vulnerability is the probability of loss of disaster bearer under different disaster intensity, and is a quantitative estimation of disaster resistance of disaster bearer at a micro scale [15]. The social vulnerability of flood disasters is mainly studied in this paper. Currently, flood vulnerability assessment can be roughly classified into four categories.” In this context, what is the function of the statement: “The social vulnerability of flood disasters is mainly studied in this paper”? it is not linked to the sentence before neither to the successive!

-Pag 7: The indicators: Exposure, Sensitivity are defined positive and the Adaptive capacity negative but it is unclear to what authors refer. Probably to the formula (3)?

Response: Thank you very much for your suggestion. We have rearranged the paragraph in page 2, the specific modifications are as follows:” Physical vulnerability is the probability of loss of disaster bearer under different disaster intensity, and is a quantitative estimation of disaster resistance of disaster bearer at a micro scale [14]. Social vulnerability is related to societal, economic, cultural and other factors. Social vulnerability is the natural attribute during disasters [15]. Currently, flood vulnerability assessment can be roughly classified into four categories.” The indicators: Exposure, Sensitivity are defined positive and the Adaptive capacity negative，which refer to formula (3), the attribute of the indicator.

2. Some part of the text are redundant and repeat things already explained. For example:

-Pag.6. line6: “The Exposure-Sensitivity-Adaptive capacity framework is used to build an environmental vulnerability model for short-term precipitation-induced flooding.” Pag. 7, section 2.3.1. “In this paper, the Exposure-Sensitivity-Adaptive capacity framework is used as the vulnerability assessment system of the short-term flood hazard.”

-section 3.2.1: authors repeat the meaning of exposure, sensitivity, Adaptive capacity, composite vulnerability but they have already explained these parameters before, in section 2.3, page 7.

Response: We thank you for the helpful suggestion. Based on your comments, the recent papers have been revised in this paper. The specific modifications have been shown in red in the manuscript.

3. Some definition are unclear: for example what is a “disaster feature”, or a “Flood submerging feature”? probably, I suppose that is the areal extension of the flood and the water depth. In a scientific work, the language should be more rigorous and precise, instead authors often use generic and unclear definitions all over the text.

-Pag.7, formula (3): The meaning of:“f is the vulnerability index of rainstorm flood…” is not clear. Generally, Vulnerability refers to an exposed community or to a particular area/region.

Response: Thank you for your constructive comments on our manuscript. We have carefully considered the suggestion of Reviewers and make changes. The specific modifications have been shown in red in the manuscript.

4. other observations:

-section 2.2: In my opinion, the explanation of the methodology is not clear. First of all, authors should insert a synthetic definition of the method. Then, they should try to synthesize the rest of the paragraph deleting redundant information.

-Section 3.1, line3: Auhors state: “The buildings, mountain shadows and other land use features can be mistakenly identified as flood disaster feature using the traditional feature extraction method. The accuracy of the U-Net++ network is better than the traditional method.”  But, what is the traditional method? ii is not clear. It should be explained.

Response: We feel great thanks for your professional review work on our paper. Based on your comments, we have revised the description and organized section2.2. The specific modifications are as follows:

“In this paper, we propose an end-to-end architecture where flood areal extension extraction and land use classification is performed simultaneously. Figure 2 shows the significance of our method. U-NET++ is used as the backbone network for flood areal extension extraction and land use classification based on multi-source remote sensing images. U-Net++ alleviates the problems caused by the network depth through effective integration of U-Net with different depths. Integrating deep supervision, new skip connection layer structure with pruning scheme were used to obtain better training effect [33]. On this basis, multispectral remote sensing images and SAR images of the same period in the study area are taken as the inputs. Flood areal extension was extracted from SAR images, affected land use classification were extracted from multispectral images. The final extraction results are used to overlapped analysis. Then, the flood areal extension and surrounding land use influenced is used to environmental vulnerability evaluation.

The U-Net++ network includes convolution layer, down sampling layer, up sampling layer, skip connection layer and input / output layer. The shallow and deep features of different proportions of ground objects can be fused using multi-scale skip connection layer. Compare to other networks, more details on remote sensing images can be retained. It is part of the main reasons for selecting this network. X^{i, j} represents convolution nodes with different proportions, and the relationship between nodes is defined as follows:

Where H represents the convolution operation, D represents down-sampling operation,

Figure 2. Network structure of flood areal extension extraction and land use classification

U represents up-sampling operation, i is defined as the index number in the sequence of down-sampling operation along the encoder, and j is defined as the index number in the sequence of skip connection operation.

Deep supervision was introduced in the U-Net++ network to obtain the best depth in multiple models of the U-Net. Sigmoid activation function and 1×1 convolution was attached to X^0,1, X^0,2, X^0,3, and X^0,4 nodes of the outputs. Where C is equal to the number of cate-gories of the given data set, and N is the number of all pixels. On the basis of the above operation, a new loss function is defined, which is mainly composed of pixel level cross entropy loss and Dice coefficient loss. The loss function has the advantages of both smoothing the gradient and redistributing the imbalance class. This loss is defined as following:

where y_n,c ∈ Y represents the target label, p_n,c ∈ P represents the prediction probability of the current category, c represents the current category, and n represents the nth pixel. In addition, the training model can be pruned through deep supervision to reduce the parameters that have to be saved. It can be realized in the following two ways: 1) integration mode: collect the segmentation results of all model branches, and then take the average value; 2) pruning mode: select one from the multi-scale U-Net model as the output. Leveraging on the U-Net++ network, the flood areal extension and land use classification is extracted which is applied to the environmental vulnerability of short-term precipitation-induced flooding.”

Then, in the manuscript, we have added the traditional feature extraction, The specific modifications are as follows: “The buildings, mountain shadows and other land use features can be mistakenly identified as flood areal extension using the traditional feature extraction methods. Which refer to threshold method and ostu method.”

5. In my opinion, the manuscript should be completely revised to improve comprehensibility and conciseness and an extensive editing of English language and style is required.

Thank you for your decision and constructive comments on our manuscript. We have carefully considered the suggestion of Reviewers and make changes. We have tried our best to improve and mad some changes in the manuscript. We really hope that this revised letter has been substantially improved.

Author Response File: Author Response.pdf

Reviewer 2 Report (Previous Reviewer 2)

The authors have substantially revised and improved the re-submission manuscript than past submission. They have also addressed most of my previous comments in this submission, however, I still have some minor comments that need to be addressed or clarified before consideration for publication. Please see comments below.

1.      Authors should clearly define the terminologies flood disasters, flood inundation, flood hazards. As mentioned before, the detection of water using satellite images might be flooding, but may not be flood disasters if there is no harm.

2.      Page 3, is the permanent population 99 million estimated using 2021 worldpop data or based on census data? Please clarify it. If it is based on census data, suggest to add year of the census data.

3.      In Figure 3, Table 2 and also in the main texts, dense à density ?

4.      In Figure 3, what do you mean by medical institution amount and school institution amount? amount à number ?

5.      Page 7, α, β, δ It is the primary index weight. à α, β, and δ are the primary index weights. ?

6.      Fig. 4, number sub-levels in the figure are missing.

7.      Page 10, please clarify that what do you mean by “Erosion Analysis”. Is the term “Erosion” commonly used term in the land cover change analysis?

8.      Figure 6, suggest to make it clear that which figures show daily precipitation and which figure shows cumulative precipitation. Please also add date for daily precipitation.

9.      Conclusions: The major findings of this study are stills completely missing. More conclusions can be drawn from this study.

Author Response

1. Authors should clearly define the terminologies flood disasters, flood inundation, flood hazards. As mentioned before, the detection of water using satellite images might be flooding, but may not be flood disasters if there is no harm.

Response: We feel great thanks for your professional review work on our letter. In our manuscript, flood disaster, flood inundation, flood hazards refer to the water areal extension is larger than the inherent water extension based on remote sensing image extraction results. The increase of water area based on satellite observation refers to the change of water when remote sensing image data is obtained. However, due to the observation frequency and mobility of the satellite, we cannot guarantee that this is the maximum flood areal extension.

2. Page 3, is the permanent population 99 million estimated using 2021 worldpop data or based on census data? Please clarify it. If it is based on census data, suggest to add year of the census data.

Response: We sincerely thank the reviewer for careful reading. Based on your comments, we have revised the description, The specific modifications are as follows：” Henan Province (31°23′ ~ 36°22′N, 110°21′~ 116°39′E, Fig.1) is located in the central of China. The total area of the province is 167000 square kilometers, and the permanent population is about 99 million according to 2020 Henan Statistical Yearbook.”

3. In Figure 3, Table 2 and also in the main texts, dense à density ?

Response: We sincerely thank the reviewer for careful reading. A suggested by the reviewer, we have revised the dense to density in figure3 and table 2.

4. In Figure 3, what do you mean by medical institution amount and school institution amount? amount à number ?

Response: We sincerely thank the reviewer for careful reading. A suggested by the reviewer, we have revised the amount to number in figure3.

5. Page 7, α, β, δ It is the primary index weight. à α, β, and δ are the primary index weights. ?

Response: We thank you for the helpful suggestion. Based on your comments, the recent papers have been revised the description, the specific modifications are as follows: “Where f is the vulnerability index of exposed region, E is the exposure index, S is the sensitivity index, and P is the adaptive capacity index. α, β, δ are the primary index weights.”

6. Fig. 4, number sub-levels in the figure are missing.

Response: Thanks for your careful checks. Based on your comments, we have added the number sub-levels in figure 4.

7. Page 10, please clarify that what do you mean by “Erosion Analysis”. Is the term “Erosion” commonly used term in the land cover change analysis?

Response: Thank you very much for your suggestion. We have revised the description in the manuscript, the modified description has been marked in red .

8. Figure 6, suggest to make it clear that which figures show daily precipitation and which figure shows cumulative precipitation. Please also add date for daily precipitation.

Response: Thank you for your careful checks. We have supplemented this information in our text. the specific modifications are as follows:

9. Conclusions: The major findings of this study are stills completely missing. More conclusions can be drawn from this study.

Response: Thank you for your decision and constructive comments on our manuscript. We have carefully considered the suggestion of Reviewers and make changes. We have tried our best to improve and mad some changes in the manuscript. The specific modifications are as follows:” The present study proposes a novel framework of short-term flood vulnerability assessment through a deep learning-based U-Net++, multi-source remote sensing im-ages as well as exposure-sensitivity-adaptive capacity indicators. Which can complete the large-scale, rapid and automatic evaluation of urban short-term flood disaster. The exhaustive framework is demonstrated over Henan Province, the short-term flood oc-curred from July 17 to 23, 2021. Taking 18 cities in Henan Province as assessment units, the overall relative vulnerability distribution of Henan Province is obtained.

1. U-Net++ is used as the backbone network to realize the classification of land use and the extraction of flood areal extension, and analyze the flood areal extension and the results of the surrounding land use erosion. This short-term precipitation has increased the water area by 0.18%. The short-term precipitation has the greatest impact on the crops, weak impact on vegetation and bare, and very weak impact on the building. The model can quickly and pertinently analyze the land use are af-fected by short-term flood disasters, and provide data support for disaster recon-struction and recovery. The proposed feature extraction framework, owing to the consideration of land use characteristics and real-time extraction of short-term flood areal extension, promises its wide applicability, especially over resource and data-constrained short-term flood regions.
2. In this paper, the exposure-sensitivity- adaptive capacity framework is used as the vulnerability assessment system of short-term flood, and the EVAM is constructed based on multi-source data. The vulnerability level is divided into four levels: very high, high, medium and low vulnerability. The proportion of the four levels in the study aera from high to low is 22.22%, 22.22%, 38.89% and 16.67% respectively. The EVAM proposed has well transferability for vulnerability analysis to other types of disasters and can better estimate the relative vulnerability of other regions.”

If there is any conclusion that is not considered in our manuscript, we hope you can give more suggestions, and we will add it to the manuscript, or put it into the follow-up study.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report (Previous Reviewer 1)

Authors precisely responded to comments improving the organization of the manuscript and the clearness  of the sections that was pointed out.

Respect to the previous version, English language has been improved.

In my opinion, it misses a minor modification to section 2.3.

As the equation (3) is inserted after the statements:

1) Exposure indicators are positive…

2) Sensitivity indicators are positive…

3) Adaptive capacity indicators are negative…

It is unclear what you are speaking about! 

In my opinion, authors should adopt one of the following solutions:

a) Put the equation (3) before the explanation of parameters exposure, sensitivity, adaptive capacity…

b) Cite before the equation (3), for example, in such a way:

“Hereafter, the vulnerability indicators (exposure, sensitivity and adaptive capacity) are explained and their contribute (positive/negative) in the following eq.3 (paragraph 2.3.1) is specified…”

Author Response

We feel great thanks for your professional review work on our paper. Based on your comments, we have revised the description and organized section2.3. The specific modifications has been marked in the manuscript. We have carefully considered the suggestion of Reviewers and make changes. We have tried our best to improve and mad some changes in the manuscript. We really hope that this revised letter has been substantially improved.

Author Response File: Author Response.pdf

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

The article: “A method to construct an environmental vulnerability model based on multi-source data to evaluate the hazard of short-term precipitation-induced flooding” proposes a model based on multi-source remote sensing data to  evaluate the vulnerability of urban areas in Henan Province. In my opinion, results of the work (showed in figure 7) could be interesting but they are presented in a very unclear manner.

The text is not well organized in the various sections. The abstract is too long and confused; the introduction does not present clearly the “state of art” and the novelty of the study. Conclusions deal about methods already described in the proper section. Then, I found that some sentences are repeated (in the same way) in different parts of the manuscript.

Moreover, an important missing about methodology is that the validation phase  is not clear. If a model is presented, it needs to be validated comparing results of the model with observed data.

To conclude, I think that the manuscript needs a complete revision and it is far from a publication stage.

I strongly suggest to reread everything carefully and review the English language.

The manuscript should be reconsidered after major Revision.

Reviewer 2 Report

Major Comments:

-          This manuscript focused on surface feature classifications using satellite remote sensing images and also environmental vulnerability index development. The research topics are interesting, and can be useful for post disaster reconstructions. However, the manuscript needs substantial revisions and clarifications before consideration for publication. There are two independent parts in the manuscript (surface feature classification and environmental vulnerability analysis), but their linkages or relationships in the manuscript are completely missing or unclear.

-          In Eq.3, please clarify why the disaster response ability index, which is a capacity/resilience to reduce the vulnerability, is considered as sum function in the vulnerability index function. Good disaster response ability helps to reduce the exposer and sensitivity effects and reduces the vulnerability, and we usually subtract (divided by) this term in the vulnerability function.

-          I suggest to elaborate more on the AHP and weight calculation e.g., key factors considered to determine the scale, what factors were considered in expert scoring, etc.

-          Page 14, please clarify that how the threshold criteria for low, medium, high and very high vulnerably were determined.

-          Conclusion: most of the descriptions in the conclusion are repeated from methods, and the major findings of this study are completely missing.

Minor Comments:

-          In abstract: “….case to verify the availability of the evaluation method.”, I could not find any verification of method in the manuscript.

-          Page 2, I also suggest to add some description about physical vulnerability.

-          After you defined acronyms or abbreviations, suggest to use only the acronyms or abbreviations. e.g., the Environmental Vulnerability Analysis Model (EVAM), Analytic Hierarchy Process (AHP)

-          Page 3, please add year of the permanent population data.

-          Page 4, please add period of data acquisitions and year of the data. e.g., precipitation data acquired period, year for population density, GDP etc.

-          Table 1, are the information presentation in this table statistical results?

-          Page 10, most of the descriptions in the 1^st paragraph are repeated from the previous sections.

-          Page 10, please clarify that what analysis are shown in Fig. 4.

-          Page 10, please elaborate more on this description “data, we affirm that the identified area was kept at a high accuracy through direct visual interpretation.”

-          Page 10, “We can be seen that flood disasters have occurred in dense urban communities.”, detection of water can be flooding, but may not be flood disasters if there is no harm.

-          Figure 4, suggest to add dates for the map of affected area before and after the flood. In addition, suggest to use more clear color legend for land. It is currently hard to identify the land areas.

-          Page 11, “…bodies, and will cause a bad impact far exceeding the numbers can show in local areas.“ suggest to make it readable.

-          Page 12, “The visualization result of precipitation is shown in Fig.5.” is the figure number correct?

-          Page 13, please define the full form of CI and CR.

Reviewer 3 Report

The paper loos a good one to addess the short-term precipitation-induced flooding issues. However, there may be few areas the author want to make changes for improvements, as such:

1. In the abstract, you mentioned that the model could be divided into two parts. However, I think you already have developed this model and it is instead-------divided into two parts! 

2. In the introduction part, you have started the topic sentence with rainstorms flooding in China. However, I would ask the authors to incorporate similar events not only in China.....but also include the similar issues happened in different continents of the world. It will improve t he readability among the scholars to connect the diverse geographic settings and implications around the world

3. Page 2, last paragraph. Exposire is related to risk ----- this statement is already provide scholars. Thus the authors should cite this sentence.

4. In the method, authors have deploued U-NET++ backbone network to summarize data. However, there are several methods available to generate the outcomes. Why did the authors adopt this network? It would be nice to elaborate this in the first paragraph.

5. Figure 4, I will recommend building/built-up areas' color to be changed to red so that distinction can be made between vegetation and buildings. Also, the mapsmust be following cartographic basics - North arrow, scale, latitude and longitude, etc. 

6. Same applies to Figure 6.

7. Could you explain the novelty of your work and future scope in the conclusion part? Also, do you think that this model may be useful in the USA/Canada/ European Union if someone wants to adopt?

6. 

Reviewer 4 Report

Contemporary climate change impacts are of interest to researchers around the World. The increasingly frequent occurrence of extreme precipitation has been scientifically observed, and their consequences can be significant. This is precisely why such scientific works are gaining more and more importance.

Abstract should be improved. In its current form, methods are described almost throughout the entire abstract, while sufficient attention is not paid to the results.

The introduction could be shorter. The entire manuscript is long, and I can see the possibility for shortening, among other sections, in the introduction.

In section 2. Experimental data and methodology, you describe the construction of the Environmental Vulnerability Analysis Model (EVAM). It remains unclear whether the constructed model has been tested? I don't see that in your work.

The author's position remains unclear about land cover and disasters. It can be especially noticed in the section of Discussion. It would be good if the authors would express their opinion specifically for the area that represents their research subject.

Apart from these minor remarks, I have no comments to the research conducted and the results presented. I believe that the manuscript itself has significant scientific value.

Best regards.