Rainfall Similarity Search Based on Deep Learning by Using Precipitation Images
Round 1
Reviewer 1 Report
I have read this paper carefully. I am impressed by the presentation which is clear and logic.
The quality of the paper is high. There are only very minor mistakes as indicated below.
L180, Change “is” to “be” before “one of blocks…”.
L394, add “that” before “it only considers…”.
L399, add a period at the end of this paper.
Fig. 7, the fonts are all too small.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
The paper titled “Rainfall similarity search based on deep learning by using precipitation images” has proposed a rainfall similarity research method based on deep learning by using precipitation images. Although there is novelty in the paper but still the paper has to be revised as per the suggestions given below:
1. In the abstract there is no discussion about the results, so please write the details about the achieved result
2. In the introduction part, there should be some discussion about the existing problems and it should be little more detailed.
3. Figures are not properly formatted, it should be formatted properly
4. Implementation details is missing in the paper
5. Future work is also not mentioned
6. There are so many grammatical mistakes, proof reading is required.
7. Few numbers of recent papers are cited
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
This paper proposes a rainfall similarity research method based on deep learning by using precipitation images. An ensemble weighting method of Normalized Discounted Cumulative Gain-Improved Particle Swarm Optimization (NDCG-IPSO) is applied to the daily precipitation images on the Jialing Basin. The manuscript is interesting. My specific comments are as under:
- At the end of the abstract, please give the numerical results of the study in no more than two lines.
- Avoid lump-sum references in the introduction section.
- Please provide a section-wise breakup at the end of section 1.
- Highlight the novelty of current work in the introduction section, point-by-point.
- The author must provide the details of the cited works. What they have done and what were the shortcomings of their study? How this study will address those shortcomings?
- The literature should be updated and expanded. For example, Study of the hydrological time series similarity search based on Daubechies wavelet transform, functional data approach for short-term electricity demand forecasting, etc.
- What is the value of p used in Minkowski distance?
- Is the data freely available? If yes, the authors should provide a link for the retrieval of data.
- Please add the study limitations and future recommendations to the conclusion section.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 4 Report
This paper introduced to research on rainfall similarity using machine learning, however, there are some issues that are listed below.
1. Line 26. It is recommended to replace “identifying& classifying them” with “identifying and classifying them”.
2. Line 88. Does the color histogram was calculated locally or globally in a certain image? It should be clarified here very well.
3. Line 132. In equation 3, does any bias was involved in the process? It looks like, there is no bias involved, if so how is the classifier avoid the outliers?
4. In Figure 4. The last right most rectangular should be addressed. “)”.
5. Line 202. It is recommended to show the relationship between the distance calculation and detection of upstream along with downstream in this paragraph.
6. Line 206 and Line 215. Both are equation 8!!
7. Line 216. It is better to normalize the distances before fusing them.
8. Line 222. The idea of reference [18] does not clarify here very well.
9. Line 257. “decrease the iterations w” to “decrease the iterations k”.
10. In equation 13, what is the operation “.” Means? Dot multiplication?
11. In 265. “take the same value in [0, 4]” to “take the same value between 0 and 4”. Also, the learning ratio might be usually a fraction value.
12. Line 285. “Experiment and Result Analysis”, on the next page!
13. Line 298. The author does not mention the total images for training and validation.
14. Line 300. Too few images were used in the test.
15. In Table 2. It should state “accuracy” words under “NDCG@5” and “NDCG@10”.
16. In Table 2. It might clarify what scheme was used to measure the accuracy, for instance, mean square error.
17. In Table 3. Do the state-of-art methods use the same protocol for data division? Same dataset? Same number of image samples? It might be addressed here.
The author used deep learning, however, there is no network structure stated in this paper!
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 5 Report
Title: Rainfall similarity search based on deep learning by using 2 precipitation images
Manuscript-ID: applsci-2221368
1. The paper proposes a rainfall similarity research method based on deep learning by using precipitation images. The topic is very interesting however some points should be addressed throughout a paper.
2. In abstract, main contributions must be added.
3. At the end of abstract, show the finding of similarity of the proposed method.
4. There is no enough investigation on the feature extraction through regional, distribution, and center precipitations. Try to test your model with different features for each type to show the validity of your model.
5. Train your model with different datasets size.
6. How did you initialize optimization parameters (i. e., w, c1, c2, gamma1,2,3)?
7. What iteration number met your optimal solution with PSO algorithm?
8. In Table3, please put reference number beside each method for example BORDA[34] and put proposed NDCG-PSO.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
As the authors addressed my concerns, I recommend the paper for publication in its present form.
Reviewer 5 Report
No comment
