Disaster-Caused Power Outage Detection at Night Using VIIRS DNB Images

Round 1

Reviewer 1 Report

This paper developed a comprehensive and important power outage detection model (PODM) to process multiple VIIRS day/night band (DNB) data and quantitatively characterize both the total power outage percent and spatial distribution of power outage caused by disaster. The estimated power outage percent is close to in-situ ground truth. The method developed in this paper can help enhance the future usage of night time observations for disaster-related power outage assessment and providing assistance to disaster relief planning.  It is suggested to address the comments listed below to improve both technical content and English language of the paper before it can be accepted.

1. To make lunar correction and atmospheric correction more comprehensible, please add plots of time series of regional mean lunar phase, lunar zenith angle and AOD.  

2. In Figure 2, what does MVCM mean? In Step 2: Cloud and edge affected image removal, it is not clear the meaning of removing Along-track edge images manually. Please explain what does along-track edge mean.

3. Figure 9 is a bit crowded. The colors used for ROI mean and lunar corrected curves are too close to tell. Different colors with clear contrast can be helpful.

4. In Table 2, what does the “Outages Sur.” stand for? It shows Outages Sur. =17.34% at Post 1 day. However, the text in Line 404 shows 18.97%. Should these two numbers match each other since the deviation is determined as 11.14% at Line 405? In Line 404, should 18.97%, be 17.34%? Need to give the source of the data (http links if it is available) when stating “according to the data provided by the power company”.

5. To make this paper more valuable, it is important to estimate the uncertainty in using the PODM method to estimate the power outage percent. Especially, in Table 2, the estimated regional outage percent should have an associated error bar. This includes the uncertainties embedded in the lunar and atmospheric corrections, applied cloud screening with confident cloudy pixels >10%, and estimation of the reference average regional radiance from multiple days of DNB data before the disaster.

6. Also in table 2, it is not clear about the intension of two “Prior 30 days” columns. There is very little discussion about different columns in this Table. Only texts from Line 401 to 406 are about this Table. It is suggested to expand discussion of this Table. Especially, need to explain how “Prior 30 days”, “Prior 15 days” are defined. Since valid DNB data are sparse in the time series as shown in Figure 9, it should make it clear how many days of valid DNB data are used in the estimation in this Table.

 Also need to explain the meaning of Time (UTC) row for prior 30 days. Is this UTC time a mean time calculated from multiple days? Need a space between “30” and “days” in the Table and make corrections to other places.

7. In Fig.10, are the Stacked image and averaged image, daily disaster images redundant? It seems that they all appeared twice in the flowchart? This flow chart can be simplified to avoid redundancy and to make it clearer.

8. The English language of this paper requires professional editing to make the paper more comprehensible.

For example, in Figure 8 caption, “shooting date” seems to mean “observation date”.

In Figure 10, “daily disaster image” can be renamed as “daily image after disaster” to make it more appropriate.

There are other places that need language improvements.

Author Response

Dear reviewer,

    I truly  appreciate for your help in reviewing of this manuscript.

    The response to your concern has been listed in the attached file.

    Please review.

                                                                                                           Yours

                                                                                                         Haodong

Author Response File: Author Response.docx

Reviewer 2 Report

This paper was written clearly in an accessible way. Based on VIIRS DNB nighttime images, the authors detected the power outage degree and spatial distribution by building the power outage detection model (PODM) and drawing the power outage spatial distribution map (POSDM). Although the topic of the paper is well-established, some problems exist in the manuscript about the main objective(s) along with the novelty and completeness level. I think the paper needs minor revisions. I have provided extensive remarks on the manuscript, which the authors should consult. My main concerns for the paper may be summarized as follows:

 

1.      In the introduction, you should point out the problems or limitations of the current studies instead of just listing previous research results. I couldn't understand the importance and main objective of the study along with the originality in the literature. To be specific, please emphasize the importance and novelty of the study compared with previous studies. In addition, subject integrity and transitions in paragraphs are quite weak.

2.      In power outage detection model diagram, I couldn’t understand why Section 3.3 was placed before the latter methods? I think Section 3.4 and Section 3.5 are both pre-processing steps, which should be placed before Section 3.3. Please reconsider and reorganize the power outage detection model diagram.

3.      In Section 3, I think it’s necessary to point out the time range of VIIRS DNB images (i.e., three months before and 15 days after the disaster) and explain the reasons why you choose images in this time period.

4.      In Section 3.8, I disagree that the difference map should be generated by the absolute value of radiance difference. If so, whether nighttime light increases or decreases significantly in the region after the disaster, the region would be considered as power outage area. This is unreasonable and may result in overestimation of the power outage area.

5.      As for accuracy assessment of the results, the study only calculated the bias of the detected power outage rate at region level, which is weak and not good for further analysis. There may be other potentially suitable disaster map datasets available which could be used as reference data. It is suggested to add the confusion matrices to quantitatively assess the accuracy of POSDM at pixel level.

6.      In Section 4.3, (1) the process of drawing the power outage spatial distribution map should be moved to Methods. (2) in Figure 10, the Difference Image was produced by subtracting the prior disaster average image map and the post-disaster image map, which is inconsistent with the method described in Section 3.8 “Calculate the radiance difference …… take the absolute value to obtain the difference map.”. Please check and keep it consistent throughout. (3) I find it difficult to understand the meaning of “Daily disaster image” and “Averaged image” in Figure 10. Please clarify and add some illustration in the caption.

7.      The formats of the citation and references need further modification. Please consult the journal’s reference style for the exact appearance of references.

8.      Check all formats, e.g. PODM and POSDM should appear as abbreviated form except for the first appearance.

9.      The presentation should be further improved by native speakers, especially for those grammatical errors, typographical errors, and bad structured sentences.

10. The anisotropy of VIIRS DNB should be discussed as it may impact the change detection accuracy, please refer to following papers:

Li, X., Ma, R., Zhang, Q., Li, D., Liu, S., He, T., Zhao, L., 2019. Anisotropic characteristic of artificial light at night – Systematic investigation with VIIRS DNB multi-temporal observations. Remote Sensing of Environment. 233, 111357

Li, X., Shang, X., Zhang, Q., Li, D., Chen, F., Jia, M., Wang, Y., 2022. Using radiant intensity to characterize the anisotropy of satellite-derived city light at night. Remote Sensing of Environment. 271, 112920

Tong, K.P., Kyba, C.C.M., Heygster, G., Kuechly, H.U., Notholt, J., Kollth, Z., 2020. Angular distribution of upwelling artificial light in Europe as observed by Suomi–NPP satellite. Journal of Quantitative Spectroscopy and Radiative Transfer, 107009

Kyba, C., Ruhtz, T., Lindemann, C., Fischer, J., Hölker, F., 2013. Two camera system for measurement of urban uplight angular distribution. In: International Radiation Symposium (IRC/IAMAS) Radiation Processes in the Atmosphere and Ocean. 

Author Response

Dear reviewer,

    I truly  appreciate for your help in reviewing of this manuscript.

   The manuscript has completed the second round of revision. 

    The renewed response to your concern has been listed in the attached file. Please review.

    Thank you very much.

                                                                                                           Yours

                                                                                                         Haodong

Author Response File: Author Response.docx

Reviewer 3 Report

Very interesting paper, very carefully prepared. My only point is to correct the formatting - it's common to have no spaces before the brackets

Author Response

Dear reviewer,

    I truly  appreciate for your help in reviewing of this manuscript.

    The  spaces before the brackets have been removed in manuscript.

    Please review.

                                                                                                           Yours

                                                                                                         Haodong