Flood Hazard Risk Mapping Using a Pseudo Supervised Random Forest

Round 1

Reviewer 1 Report

Dear Editor,

Thank you for giving me the opportunity to review this manuscript titled "Flood Hazard Risk Mapping Using a Pseudo Supervised Random Forest, by: Esfandiari et al.

 

There are no issues of conflicting interest, and I have no personal or professional Affiliation with the authors. 

 

This is an interesting manuscript which uses remotely sensed data to develop a model produced a flood extent prediction model. The methodological approach and the investigation are well established and through a thorough discussion the authors come to useful conclusions. The manuscript presents original data, and I think that the scientific content is of interest to the readers of the Remote Sensing Journal and should be published in this journal with minor revisions.

In general, the manuscript seems well done and well presented. The English is fine, but some sentences require a rephrasing. The tables and figures are nicely prepared, and the text is well organized. The approach of this manuscript is well supported and the authors have taken into account a great amount of the relevant prior works.

 

 

Specific comments:

Title: Flood Hazard Risk Mapping Using a Pseudo Supervised Random Forest

The title is appropriate for the study area

Abstract.

Abstract should be concise and includes only your findings.

1. Introduction

The manuscript seems well written and well presented

2. Materials and Methods
3. 1 Study area

The maps are not readable, particularly figure 1. Please add the localities names in all the maps/images e.g., Fredericton and Ottawa cities in Fig. 1a and Baskatong Reservoir, Quebec, Chaudière Bridge. Please, reorganize Fig.1 and add figure labels on the top left side of the figure e.g., "a, b, and c".  Please, add latitudes and longitudes on Fig. 1a and label the names of the study areas. Enhance the scale bar in Fig. 1b.

2.2 Datasets

- "The Landsat images are pan-sharpened by integrating the panchromatic band (15 m) and multispectral bands (30 m) using UNB (University of New Brunswick) pan-sharpening method available in PCI Geomatica software".

Could you please use "fusion" instead of "integrating" and please specify the utilized multispectral bands (30 m).

- Please add the abbreviation "UNB" after "University of New Brunswick".

- Please refer to "Table 1" in the second paragraph of "Datasets"

- In page 6, please add the equation 1 after "For accuracy assessment, the ground truth points were selected using a Normalized Water Difference Index (NDWI)"

- In page 6, Equation (1) demonstrates how NDWI is calculated using green (Green) and near infrared (NIR) bands. Please, specify what number of the bands and specify the sensor either OLI or Sentinel-2.

2.3. Conditioning Factor

- Please, add the importance of "Aspect". Why you did not extract stream-networks instead of Aspect?

Discussion

The discussion has been presented clearly, but this part must be concise and supported with previous literatures. You can benefits from the recent papers of flood hazard risk mapping in discussion and introduction sections.

 

Conclusions

Conclusions should be concise and the grammar should be checked as some errors occur in the entire manuscript.

How can extend the results in other regions with similar/different climates? At the end of the manuscript, explain the implications and future works considering the outputs of current study.

 

Author Response

Dear Reviewer,

 

Thank you for giving us the opportunity to submit a revised draft of the manuscript “Flood Hazard Risk Mapping Using Pseudo Supervised Random Forest”. We appreciate the time and effort dedicated to providing feedback on our manuscript and are grateful for your comments. Your suggestions are applied in the manuscript and the changes are tracked within the manuscript. 

Please find the attached document as our response to your comments.

 

Sincerely,

Morteza Esfandiari

 

Author Response File: Author Response.docx

Reviewer 2 Report

• Add a short explanation about the UNB pan-sharpening algorithm in section 2.2. Datasets.
• The location of formula (1) is out of order thus needs to be moved from page 6 to page 7 or 8 to avoid confusion.
• In Figure 3 caption: Correct the following typo: “dos” to “dots”
• Page 8: Correct the following typo: “0.275 meters” to “0.275 meter”
• Many important references discussing flood modeling using remote sensing and GIS were left out, and previous work was not reviewed. A couple key references are:

• Ghoneim, E., Foody, G. 2013. Assessing flash flood hazard in an arid mountainous region. Arabian Journal of Geosciences 6 (4): 1191-1202.

Author Response

Dear Reviewer,

 

Thank revising our article titled “Flood Hazard Risk Mapping Using Pseudo Supervised Random Forest”. We truly appreciate your time and effort that you dedicated to provide valuable comments on our manuscript. Your comments are applied in the manuscript and highlighted within the file.

Please find the attached document as our response to your comments.

 

Sincerely,

Morteza Esfandiari

Author Response File: Author Response.docx

Reviewer 3 Report

This paper presents a very interesting framework using pseudo supervised Random Forest (RF) for fluvial flood risk mapping. The RF-model uses predictions of HAND model along with altitude, slope, aspect distance from river and land use maps as training input. The input of the flood maps of the HAND model were considered as pseudo training data sets. Furthermore, a rand sample consensus method were used to detect outliers in order to improve the accuracy of the RF-model.

To test the effectives of the model, case studies in Canada have been implemented to compare prediction results of RF-Model with Hand-model results in front of historic flood events. The figures and tables look good. The research process is well outlined and the framework is well explained.

However, several points must be clarified and several questions must be answered before this draft can be approved for publication. The main concerns are:

• In the Discussion there is no comparison to other studies using RF for fluvial flood extent prediction. Try to interpret and discuss yours results in perspective of previous studies and of the working hypotheses. The findings and their implications should be discussed in the broadest context possible.
• The manuscript does not contain an explicit objective statement within the abstract, starting with: “The objective of this…. “. Such a statement, however, is extremely useful for readers in several ways: they use it for instance to judge if the manuscript is worth reading, and to focus on the main issues while reading.
• Please describe the approach of the HAND model in more detail.
• The HAND model is a simplified conceptual model and therefore not so precise as a physical-based hydrodynamic model. Are there any data / flood maps accessible in order to compare your approach with an hydrodynamic model?

A set of technical issues and comments for the paper are provided here:

• Please use line numbers! The review process otherwise proves to be very laborious. (please check the appendix)
• Line 76: …, Hydrodynamic models: orthographic error.
• Line 127: …Machin Learning: orthographic error.
• Line 284: Figure 4. Cont.: Description of Figure 4 is missing.
• Lines 311-312: What’s the difference between flood mapping and flood hazard mapping?
• Line 346: … 146meters: Please correct to 146 meters.

Author Response

Dear Reviewer,

Thank you for reviewing our paper titled “Flood Hazard Risk Mapping Using Pseudo Supervised Random Forest”. We appreciate your time and effort that you dedicated to provide helpful comments on our manuscript and are thankful for your suggestions. Your comments are applied in the manuscript and the changes are highlighted within the manuscript.

Please find the attached document as our response to your comments.

 

Sincerely,

Morteza Esfandiari

Author Response File: Author Response.docx