A Spatiotemporally Constrained Interpolation Method for Missing Pixel Values in the Suomi-NPP VIIRS Monthly Composite Images: Taking Shanghai as an Example

Round 1

Reviewer 1 Report

My only comment is that the quality of the figs 2,6,7, and 8 should be improved. If Editor is OK with them then manuscript can be published as it is.

Author Response

Thank you very much for your valuable comments. With the guidance of your comments, we have thought more deeply about these issues, Hope our revisions can improve our paper.

Author Response File: Author Response.pdf

Reviewer 2 Report

This research topic seems exciting and appropriate for publication in Remote sensing. The manuscript, titled " A Spatiotemporally Constrained Interpolation Method for Missing Pixel Values in the Suomi-NPP VIIRS Monthly Composite Images: Taking Shanghai as an Example" (remote sensing 2296434), reported a spatiotemporal constrained interpolation method that considers temporal continuity, the relative stability of the urban spatial structure, and the relative smoothness of the night light of the same objects in time. However, there are some issues to be solved in the manuscript.

 

#1: The ratio between the main image and the area map in Figure 1 is not appropriate, and I recommend the authors to redraw it.

#2: Figure 2 is not complete enough and it is recommended to add a nine-dotted line.

#3: I propose the authors to be more specific, explanatory and simplified in order to be easily understandable by readers.

#4: I suggest the authors to verify the formatting of all formulas throughout the manuscript.

#5: The reasons for selecting nine time series interpolation methods should be added.

#6: The spatiotemporally constrained method proposed in this article has improved the accuracy of interpolating night light images in Shanghai. The discussion on the applicability for other cities should be added.

Author Response

Thank you very much for your significant comments and suggestions! With the guidance of the comments, we have thought more deeply about these issues, condensed the main questions, and improved the organization and presentation of the article. We have benefited from each of your comments, and we sincerely admire your scientific ability. Hope our revisions can improve our paper.

 

Author Response File: Author Response.pdf

Reviewer 3 Report

This manuscript presents a comparison of and added methods to improving the VIIRS Day/Night Band nighttime light monthly data with stray light degradation near summer solstice for Shanghai China. The importance of a continuous, quality data set is explained, 9 interpolation methods are outlined, and the new statistical based method for removing suspect data is presented. Six years of data are processed to evaluate the various configurations with three metrics of NP, TDN, and ADN used to evaluate performance.

My comments and questions are both major and minor in nature. I am familiar with satellite data and meteorological data assimilation but not as knowledgable about day/night band usage. Given my background, I had these major questions:

1) Section 4.3 and Figure 10 compare the STCI3 & STCI5 to a time interpolation data set. Which one? 9 were reviewed and presented in Fig 8 & 9. Is the purpose of this paper to also conclude which interpolation method was the optimal choice for this dataset when combined with STCI3 or 5? Is there a recommendation by the authors for 3 versus 5 or will both be investigated in the future with the future research discussed at the end of section 5?

2) The presentation of the 9 interpolation methods is thorough. The 3 metrics of performance are not explicitly defined.  Number of abnormal pixels is understood from the text but what makes the value an outlier for the data which did not use the STCI3/5 method? The same cutoff values presented in Table 1? I am not familiar with TDN, is it the sum of the pixel values in the scene? Is the difference from the original June value defined as Original - New? ADN (I assume) is a pixel difference of abs[Original - New] but could also be explicitly presented in the text.

3) If TDN is the summed pixel value of the scene, why is a better performance determined by the difference in TDN approaching zero (line 360-361)? Looking at Figure 2, I think the June months have higher values which can be due to both surface temperature and stray light. If stray light adds bias and not just noise to the image, would a nonzero TDN difference equal to the stray light bias be the desired outcome? If so, this would also impact the desired ADN histogram peak e.g. perhaps ADN = 1 is better than 0?

Minor comments and questions:

Line 122 - Severely lacking and Figure 2. Does this refer to the number of valid pixels in the scene, the quality of the data or both? A description of why Fig 2 shows "lacking" data would be helpful. Or is Fig 2 to show the close northern cutoff of useable data and illustrate why Shanghai data in June is more affected?  

Table 2 - Like my previous question about Section 4.3, one line for each SCTI method is confusing when each was applied to 9 interpolation methods. By design, the SCTI method should have zero NP pixels? That could be stated in the caption and text.

Figures 9 - could be altered to use a fixed y-axis range of -55 to 25 to allow ease in comparing the different years.  Also, a definition of the difference such as Original - New in the caption would be helpful.

Lines 384, 385, 394 Confirming that the range is 0-5 not 1-5 for the discussion.

Finally, I want to add that this paper is almost ready for publication. I chose "must be improved" and had major comments because I believe the text needs to be clarified before the final draft. The science is good and the method is very useful!

Author Response

Thank you very much for your significant comments and suggestions! With the guidance of the comments, we have thought more deeply about these issues, condensed the main questions, and improved the organization and presentation of the article. We have benefited from each of your comments, and we sincerely admire your scientific ability. Hope our revisions can improve our paper.

Author Response File: Author Response.pdf