Comparison of Different Artificial Intelligence Techniques to Predict Floods in Jhelum River, Pakistan

Round 1

Reviewer 1 Report

This paper resembles a conference paper that has a long way to go before it is journal-worthy. There are some issues that must be resolved before resubmitting. The application of some AI method to a case and publication of the results as a paper cannot be considered a significant academic contribution. In addition, there are numerous comparable and superior methods available in the literature review that can be utilised in this instance. Moreover, I am unable to identify a literature review, which I find extremely odd given the abundance of studies in this research. Moreover, the presentation of the methodology is completely unacceptable. The proposed ANN is a simple ANN without any performance-enhancing modifications. While there are numerous methods in the literature, it is unclear why these methods have been utilised. In addition, the results were not supported by a sufficient discussion or an adequate statistical test.

My decision is "reject".

Author Response

The authors appreciated the reviewer’s comment. Yet, we believe that our work has publish-worthy merits that would benefit other peers working within the field of data-based modelling with applications in water engineering and management. In addition, upon addressing multiple comments from the other two reviewers, the work has been further improved with more highlighted novelty and significance. The detailed response to the reviewer comments is added in PDF file. 

Author Response File: Author Response.pdf

Reviewer 2 Report

The manuscript number water-1856293 is dedicated to evolution of different regression methods in predicting flood in Punjab watershed in Pakistan. The title and topic is interesting because of frequented flood events in the west Asia during the past decade. However, the manuscript suffer from different short coms.

In addition the theme of the current issue is “This special issue intends to provide relevant results on how human activities and climate change impact fluvial basins, freshwater wetlands,  and river’s functioning, focusing on hydrological, hydro-geomorphological, hydrochemistry, and hydro-ecological research”. Even the manuscript are silence about human or climate effects on the flood characteristics. I will list some major and minor points in the follow,

Major Comments:

·         What is the major contribution of the research? The nobility of the paper must be mentioned in the last paragraph of the introduction clearly. I did not find any considerable contribution in the manuscript. It seems as a case study

·         The used methods do not have any relevant sources and literature. It is impossible to use method without relevant cases in the literature except innovative approaches.

·         What is the pool of input variables? I did not find any list of input variables. Table 1 is not clear without positioning and map of the mentioned stations.

·         What is the advantages of feature selection method mentioned in section 2.4?

Has the Gamma test for feature selection been proposed by the authors or barrowed from other references? Is yes, where is its related literature review?

·         Why the proposed feature selection with gamma test method is in the optimal situation? I recommend check the feature selection results with another filter-based method such as Mutual Information method.

·         There are no clues to prove the optimality of developed regression methods. The authors must present some evidence to show the calibration of model structure (including number of layer, nodes, and nearest neighbor). May be it’s the reason of superiority of LLM method versus AI approaches.

·         How did you check the parsimony issue of the developed regression models?

Minor Comments:

·         In the manuscript, the abbreviations must be carefully spell out in the first use. Then, abbreviations are allowed to use in the manuscript.

·         Exclude the signs of training and testing samples in the scatter plots.

·         Conclusion and discussion are very weak. Different questions as mentioned some of them in the major comments must be consider in the these two parts.

Author Response

The authors are appreciative of the reviewer for providing a very concise and precise summary of our work. The detailed response to the reviewer comments is added in PDF File.

Author Response File: Author Response.pdf

Reviewer 3 Report

The manuscript used 5 models of machine learning to construct flood susceptibility maps in the Jhelum River, Punjab, Pakistan. In my opinion the manuscript addresses an interesting topic in the study of flood susceptibility. However, in its current format, I consider that the manuscript has important shortcomings that should be resolved, which are described below:

The contribution of this paper is unclear. Authors should clearly describe how their work contributes to developing the knowledge in its field.

- Abstract

The abstract looks like a shopping list, focusing too much on what has been done and too little on justifying the need for this research and its contribution to the advancement of the field.

- Introduction           

+ It is difficult to see the justification for the need of this research. The literature review is poor. The paper needs to clearly state what are the problems with the existing works (these types of approaches) and what problem(s) this particularly paper was going to address. Without this clearly problem statement readers would have difficulty to see the merit of this paper. The author only lists some references, I did not find the problem with the existing methods. The author should show us deep analysis about the gap among existing methods.

+ The main goal of the paper should be better explained. It is not clearly stated. The goal should clearly and concisely explain the main scientific contributions of this work.

- Study Area

+ I did not see any description of the occurred floods (flood characteristics) in the study area. Why do you choose this area for study? This section is a problem statement, and you should bring a logical reason for starting this research.

- Results

+ It is noted that critical analysis of results has not been reported in the paper. The result section is so poor in writing and obtaining results. Also, the written results do not have homogeneous due to no logical relationship between applied methods’ results.

The discussions do not present the contribution of the findings to the theoretical advancement of the field, and the methodological or conceptual limitations of the study and the future research directions, including the ways to overcome the existing limitations.

Author Response

The authors are appreciative of the reviewer for providing a very concise and precise summary of our work. The detailed response to reviewer is added in PDF File.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

.

Author Response

The authors are thankful for suggestions of reviewer 1. The English language,  style and spell check are done.  

Reviewer 2 Report

In the section dedicated to methodology, DLLR procedure and formulations didn't mentioned. Please described the method in brief.

Topology of ANNs is a multi dimensional point including number of hidden layer, number of nodes, number of input and output features. How do the authors prove the optimality of the presented ANNs (without showing the evidence, the comparison is not fair)? How many nodes in first and second hidden layers have been used?

The presented reasons to show the nobility of the manuscript is not satisfied me.

Author Response

The authors acknowledge and substantially concur with the reviewer’s suggestions. The response to reviewer 2 is added in pdf file.

Author Response File: Author Response.pdf

Reviewer 3 Report

Accept in present form

Author Response

The authors are very thankful for reviewer suggestions and acceptance for the manuscript.