An Improved Correction Method of Nighttime Light Data Based on EVI and WorldPop Data
Round 1
Reviewer 1 Report
The manuscript had been meticulously revised and is ready to be published as is.
Author Response
Thank you for your modification in the previous manuscript. This article is greatly improved on the basis of your suggestions. Thank you again for your patience！
Reviewer 2 Report
The manuscript has been improved. From my point of view, the paper is almost suitable for publication in Remote Sensing.
Author Response
Thank you for your modification in the previous manuscript. This article is greatly improved on the basis of your suggestions. Thank you again for your patience！
Reviewer 3 Report
In this paper the authors proposed an invariant region method based on ENVI and WorldPop data to achieve inter-calibration and saturation correction of DMSP/OLS data using regression data on long time series of night light images. The data were collected from the year 2001 to 2018 for 30 provinces in China. The GDP and EPC were estimated using correction methods. Methodology used in the paper is adequate with sound conclusions. I am interested why is the predicted GDP in Table 1 negative under linear model for the year 2001? Also authors use MARE and MAPE metrics for displaying the precision of the models. The range of precision for each measure could be better decribed, for example the interpretation of MAPE below 10% is excellent predicted accuracy, etc. After making minor changes in the manuscript I suggest publication of this interesting and relevant paper.
Author Response
Points 1: why is the predicted GDP in Table 1 negative under linear model for the year 2001?
Response 1: Thank you for your suggestions. Table 1 in the manuscript compares the accuracy of the estimated statistical gross domestic product (GDP) on time series at the national scale through the three regression models (linear, quadratic polynomial and power function model), so as to obtain the optimal fitting equation to compare the accuracy of the nighttime light data corrected in this paper and corrected by reference correction method to predict GDP, so as to illustrate that the proposed correction method based on EVI and WorldPop data can better correct the night light data.
The growth index of GDP with TNL under the power function model is 2.94. Therefore, it can be seen that the growth rate of GDP is far greater than that of TNL in time series. Therefore, it cannot get better fitting results under the linear regression equation, and the results obtained in Table 1 of the manuscript also illustrate this result. When fitting with linear regression equation, the linear regression parameters (a and b) are determined according to the overall data. Therefore, the a value obtained is the average slope, which is far greater than the slope of GDP-TNL in 2001-2002, and the b value obtained will greatly reduce the predicted GDP value in 2001 and 2002. Therefore, under this prediction equation, the predicted GDP in 2001 is negative.
Author Response File: 
Author Response.docx
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
Major comments:
- This study might set the bar too low to present its novelty. The objectives of this study should be better explained. What is the difference with the previous similar studies?
- The manuscript adequately describes the equation used. However, the authors do not explain why in the approach. For example, average DN value of the pixel is based on the correction formulas from formula 10 and 11.
- The study generates consistent nighttime light data set from 2001 to 2018. The established regression model was used to fit the NPP/VIIRS data from 2014 to 2018 to the DMSP/OLS data scale. Could the authors transform the DMSP/OLS data to NPP/VIIRS data?
- We can’t see the spatial details. Fig 2 and 3 are too small.
Minor comments:
- The x-axis in Fig 7 ranges are different with a, c and b, d.
Author Response
Response to Reviewer 1 Comments
We appreciate valuable suggestions and comments from the anonymous reviewer. We have revised our manuscript according to your helpful suggestions.
Major comments:
Points 1: This study might set the bar too low to present its novelty. The objectives of this study should be better explained. What is the difference with the previous similar studies?
Response 1: Thank you for your suggestions. Aiming at the problems in the current research, this paper proposes a correction method of nighttime light data based on EVI data, so as to obtain the nighttime light dataset from 2001 to 2018 to represent economic data reliably and accurately. By comparing the correction method mentioned in this manuscript with those correction method mentioned in the following literature [1, 2], we illustrate the superiority of the performance of the correction method mentioned in this manuscript. We have modified lines 156-174 to make the purpose of the study clearer.
Points 2: The manuscript adequately describes the equation used. However, the authors do not explain why in the approach. For example, average DN value of the pixel is based on the correction formulas from formula 10 and 11.
Response 2: We have strengthened the explanation of the formula used in the lines 356-361, 367-370, 373-378 and 381-384.
Points 3: The study generates consistent nighttime light data set from 2001 to 2018. The established regression model was used to fit the NPP/VIIRS data from 2014 to 2018 to the DMSP/OLS data scale. Could the authors transform the DMSP/OLS data to NPP/VIIRS data?
Response 3: Thank you for your suggestions. Since DMSP/OLS data is no longer updated and NPP/VIIRS data is still being updated, in theory, DMSP/OLS data should be converted to NPP/VIIRS data. However, due to the short time span of NPP/VIIRS data in this study (only five years), thus considering the accuracy and validity of the fitting results, the NPP/VIIRS data from 2014 to 2018 should be converted to DMSP/OLS data. At present, most scholars [1-3] adopt the method of transforming NPP/VIIRS data into DMSP/OLS data.
Points 4: We can’t see the spatial details. Fig 2 and 3 are too small.
Response 4: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Minor comments:
Points 1: The x-axis in Fig 7 ranges are different with a, c and b, d.
Response 1: The X & Y axis tick label have be improved by using scientific notation.
Response to Reviewer 2 Comments
The authors have discussed about the Nighttime lights correction with EVI data and validated with GDP. However, this paper can be improved in many ways.
Points 1: Writing and grammar can be improved. For example, there are many inconsistency and grammar issues like [Line 310]" Obtaining historical stable images in China, it is necessary to consider the image data collected by sensors in different years and different sensors in the same year. "; [Line 404] "It assumes that..." - are you referring to a specific method?; [Line 466] "sharp increase trend" -"sharp increase" or "increasing trend".
Response 1: Thank you for your suggestion. We have modified the corresponding sentences in the line 309 and 458. The assumption in the line 394 is a precondition of the hypothesis, which is considered to be within the two intervals of 2012-2015 and 2016-2018, there will not be a large number of pixels changing from dim to bright in the two stages.
Points 2: The articles need to be better organized - e.g., you can use another level of heading (instead of number in circles) to show 4.2.1. Inter-calibration steps.
Response 2: We have readjusted the structure of the article, such as merging the second part and the first part of the article into the first part, explaining the methods adopted by researchers in correcting the nighttime light data, and summarizing some common shortcomings of the current correction methods. We have modified the corresponding sentences to remove the circles in the line 312-332.
Points 3: The authors have mentioned the "reference correction method" [Line 475] at the end of the methodology part, can you please explicitly define what exactly is the reference method. It is very hard to follow after you have introduced various data processing steps.
Response 3: The reference correction method mentioned in this paper is the method used in the following literature [1, 2] for DMSP/OLS and NPP/VIIRS data correction. This part has been added in the line 446-453.
Points 4: As for the data validation, why compare NTL with GDP since NTL is not a direct measurement for GDP. GDP can be affected by economic structures at various subnational levels. Basically the graphs and tables you have presented is only helping readers to explore the relationship between lights and total GDP. However, for regions that mainly reply on agriculture production, how do you use nighttime lights to capture this.
Response 4: Thank you for your suggestion. We are very much in favor of your opinion. Agriculture is an important part of social and economic activities, especially in a traditional agricultural country like China. However, most of the agricultural production is located in the darker area of the nighttime light image, so the nighttime light data cannot represent the agricultural part of economic activities [4]. It is necessary to remove the contribution of agricultural sector to economic activities when using nighttime data to predict GDP. The GDP data in the following paragraphs leave out the data of the agriculture sector. We have added the description part in the line 212-217. And the content related to GDP has been recalculated.
Points 5: Why EVI index is introduced instead of other remote sensing data? If an additional data source is added (like population distribution from GHSL https://ghsl.jrc.ec.europa.eu/), will this improve the model performance.
Response 5: Thank you for your suggestion. In this manuscript, EVI data is introduced to realize the saturation correction of DMSP/OLS data based on considering the negative correlation between vegetation and human activities (the intensity of human non-agricultural activities in the central area of the city is high, and the vegetation coverage is generally less; in the rural areas, the intensity of human non-agricultural activities is low, and the vegetation coverage is more) [5, 6]. And EVI data is a continuous dataset in time series, which can correct DMSP/OLS data in different years. The GHSL, a global information baseline describing the spatial evolution of human settlements in the past 40 years, was developed by using a symbolic machine learning model trained by the collected high-resolution samples, multitemporal Landsat imagery in the epochs 1975, 1990, 2000 and 2015 [7]. Although we can interpolate the data between the years to match the period of the nightlight data, it is difficult for us to obtain GHSL data on continuous time series as accurately as EVI data. However, we agree with your opinion very much. At the same time, your view also gives us great inspiration. Applying GHSL data to DMSP/OLS data correction can effectively improve the performance of the model, and can better distinguish and identify built-up areas and non-built areas Therefore, the next stage of work will focus on how to combine DMSP/OLS data with GHSL data to better study how cities expand.
Response to Reviewer 3 Comments
This work aims to accomplish two most critical thresholds in performing long time series analysis using Nighttime Light (NTL), which are
Correction of saturation in DMSP-OLS annual composite
Inter-annual and intra-annual correction of DMSP-OLS annual composite
Bridging DMSP-OLS and VIIRS-DNB nighttime light imagery
The authors then verified the resulted NTL series by comparing GDP and EPC of selected provinces in China, and confirmed to have very high agreement.
The manuscript is very well laid out with some room for improvement as listed below.
Points 1: Line 80: EVI should be described more in Introduction section.
Response 1: Thank you for your valuable suggestions. We have added some description on the EVI data in the line 104-108.
Points 2: Line 161: DMSP/OLS Data -> DMSP/OLS data.
Response 2: Thank you for your patience in reading this manuscript. We have read through the full manuscript, and have changed all the contents related to DMSP/OLS Data to DMSP/OLS data.
Points 3: L194, L204, L215: The group which makes these products (EOG, Earth Observation Group) is no longer at NOAA/NCEI and now affiliated to the Paynes Institute for Public Policy, Colorado School of Mines.
Response 3: We have changed the corresponding institutions and websites to the current ones.
Points 4: Figure 1: NPP/VIIRS Stable Data for China should not need inter-annual correction?
Response 4: Thank you for your suggestions. Although the NPP/VIIRS data is collected by the same sensor, it can be seen from Figure 4-a that the NPP/VIIRS data before uncorrected are not all increasing. The TNL data from 2015 to 2016 and from 2017 to 2018 show a slight downward trend. Therefore, we believe that in order to make the results more appropriate to China's actual urbanization process, the nighttime light data more accurately predicting GDP and EPC, the NPP/VIIRS data should be carried out with inter-annual correction.
Points 5: Figure 2 & 3: These two figures does not show the difference clearly. Suggest enlarge some smaller regions to show the effect of correction. NPP/VIIRS should also be included in the caption of Figure 3.
Response 5: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Points 6: Line 314 to 324: The english writing in this paragraph is unclear. Please improve.
Response 6: We have modified the corresponding sentences in the line 312-320.
Points 7: L372: The reference [48] only exams data up to year 2008, it is not very convincing to extrapolate the conclusion all the way to 2018. Authors should make clear about the assumption they are making here.
Response 7: Thank you for your suggestion. The author has noticed this problem and updated the latest references [8, 9]. It can be seen from reference [9] that China's urbanization rate showed an upward trend from 2004 to 2017. From reference [8], we can see that China's urbanization increase rate is 1.36% from 2000 to 2010 and 1.1% from 2010 to 2018. Therefore, according to these two references, the author can assume that there is no reverse urbanization process in China from 2001 to 2018.
Points 8: L402-404: Is the research period up to year 2018 or 2019?
Response 8: Thank you for your suggestions. The research period is 2002-2018, which has been revised in the line 392-393.
Points 9: L403-410: The english writing in this paragraph is unclear, please improve.
Response 9: Thank you for your suggestions. We have modified the corresponding sentences in the line 393-399.
Points 10: L515-516, Figure 6: What is the reference correction used here? Please make it clear.
Response 10: The reference correction method mentioned in this paper is the method used in the following literature [1, 2] for DMSP/OLS and NPP/VIIRS data correction. This part has been added in line 446-453.
Points 11: Figure 7: The X & Y axis tick label should be improved by using scientific notation or difference unit.
Response 11: Thank you for your suggestions. The X & Y axis tick label have be improved by using scientific notation.
Points 12: Table 1 & 2: What is the data source of GDP and EPC?
Response 12: Gross domestic product (GDP) and electric power consumption (EPC) from 2002 to 2019 year are provided by the China statistical yearbook which published by China National Bureau of statistics (http://www.stats.gov.cn/tjsj/ndsj/).
Points 13: Table 5 & 6: What is the unit of these numbers?
Response 13: It shows the mean absolute relative error between the real GDP and the GDP predicted by the models based on the GDP and corrected nighttime light data at provincial scale under the two correction methods in Table 5. At the same time, it shows the mean absolute relative error between the real EPC and the EPC predicted by the models based on the EPC and corrected nighttime light data at provincial scale under the two correction methods in Table 6. Therefore, these numbers are the mean absolute relative errors without units.
Points 14: Table 7: Why is the numbers shown in this table having fixed numbers after decimal point of .33 and .67?
Response 14: The numbers in table 7 are obtained through the classification of results in tables 5 and 6. The mean absolute relative errors of 30 provinces are divided into three categories [10]: 0-25% is high accuracy, 25%-50% is medium accuracy, and greater than 50% is inaccuracy. The reason of the numbers shown in Table 7 having fixed numbers after the decimal point of. 33 and. 67 is that the research object of this paper is 30 provinces in China, and the fixed decimal point (.00, .33, and .67) is obtained by dividing the total of the provinces after statistics by 30 provinces.
Response to Reviewer 4 Comments
The manuscript entitled "Nighttime light data correction based on EVI data by invariant region method" aimed to address the overglow and saturation effects of the DMSP/OLS data by using the EVI product.
I believe the manuscript is suitable for publication in Remote Sensing, after the authors have addressed the comments below.
Points 1: Page 1, line 19. You mentioned that (… on national and provincial scales, respectively). Where? You have to mention the country.
Response 1: Thank you for your suggestions. We have defined the research area of this paper to 30 provinces of China in the line 16. The national scale mentioned here is to take China as a whole research area, so as to analyze the fluctuation of TNL over time and the ability to predict GDP and EPC. The provincial scale refers to taking each province in China as an independent research area to analyze the ability of predicting GDP and EPC in each province.
Points 2: Page 2, line 46-50. There are other nighttime sensors. Therefore, when you mentioned only the DMSP/OLS and NPP/VIIRS, you have to say something like “the most widely used” or “the popular nighttime sensors”.
Response 2: Thank you for your suggestions. We have modified the corresponding sentences in the line 49.
Points 3: Page 2, line 77. Change the sentence to be (… is more serious than that between the difference products of the DMSP/OLS).
Response 3: Thank you for your suggestions. We have modified the corresponding sentences in the line 145-147.
Points 4: Page 4, line 185. A site location map is more important than Figures 2 and 3.
Response 4: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Points 5: Page 8, line 321. You mentioned that (… the stable nighttime light image mask of 2018 and the nighttime light image ….). I believe you mean light image mask of 2008 not 2018. Check.
Response 5: Thank you for your patience in reading this manuscript, the year of the stable nighttime light image mask is actually 2008. We have modified the corresponding sentences in the line 318.
Points 6: Page 9, Equations (4) and (5). These equations are similar. Thus, just keep one of them.
Response 6: Thank you for your valuable advice. Formula (5) is a general formula for the normalization of DN value, while formula (6) is a special case where the value of is taken as 0. The retention of formula (6) can simplify the writing of subsequent formulas. Therefore, experts are requested to agree to retain formula (5) and formula (6).
Points 7: Page 10, line 362. Put a reference after “Intra-annual composition”.
Response 7: Thank you for your suggestions. The corresponding reference has been added in the line 355.
Points 8: Page 10, line 374. Put a reference after “inter-annual series correction”.
Response 8: Thank you for your suggestions. The corresponding reference has been added in the line 366.
Points 9: Page 10, Equation (10). Would there be a difference if you used the inter-annual series correction that proposed by Liu et al. (2012) [Ref. 48].
Response 9: Using the inter-annual correction method proposed by Liu et al. it can ensure that the DN value of the pixel in the next year's image is not less than that of the previous year's image. This adjustment order can avoid weakening the DN value of the pixel in the early annual image, but will cause the DN value of the pixel in the later year's image to be excessively enlarged. Therefore, the correction in the opposite direction should be carried out on the basis of the inter-annual correction method proposed by Liu et al. to ensure that the DN value of pixel in the image of the previous year is not greater than that of the pixel in the image of the next year. Then taking the average value after two inter-annual correction as the final DN value, we can obtain a more realistic nighttime light image.
Points 10: Page 11, lines 402-403. Are you sure it is 2019 not 2018.
Response 10: The research period is from 2002 to 2018. We have modified the corresponding sentences in the line 392-393.
Points 11: Page 11, lines 405-410. Why did not you choose the 2018 annual composite data as a binary-value mask instead of 2016?
Response 11: At present, the monthly composites published by the Paynes Institute for Public Policy, Colorado School of Mines are all monthly products, has not been filtered to screen out lights from aurora, fires, boats, and other temporal lights. Therefore, the nighttime light images we synthesized for 2018 also include these short-term light data. The annual composite data published only contain the 2015 annual composite data and 2016 annual composite data which has been filtered to screen out short-term lights. Hence, we choose the 2016 annual composite data as a binary-value mask instead of 2018.
Points 12: Page 13, lines 467-468. Re-write the sentence.
Response 12: Thank you for your suggestions. We have modified the corresponding sentences in the line 459-461.
Points 13: Page 13, lines 474-475. Re-write the sentence.
Response 13: Thank you for your suggestions. We have modified the corresponding sentences in the line 466-467.
Points 14: Page 13, line 475. What do you mean by the reference correction method? You have to define it.
Response 14: The reference correction method mentioned in this paper is the method used in the following literature [1, 2] for DMSP/OLS and NPP/VIIRS data correction. This part has been added in line 446-453.
Points 15: Page 16, lines 540-542. Re-write the sentence.
Response 15: Thank you for your suggestions. We have modified the corresponding sentences in the line 535-538.
Points 16: Page 16, lines 548-554. From which tables did you get the results of the MARE before and after 2013? If there is no table, you have to create one that summarises the results.
Response 16: Thank you for your suggestions. We added lines 21-22 in Table 1 and lines 21-22 lines in Table 2 to obtain the corresponding conclusion data in the paper.
Points 17: Page, Figure 7. Change the x-axis unit to be in million and ranges from 0 to 30.
Response 17: Thank you for your suggestions. The X & Y axis tick label have be improved by using scientific notation.
Points 18: Page 17, lines 572-578. Same as Comment No. (16).
Response 18: Thank you for your suggestions. We added lines 21-22 in Table 1 and lines 21-22 lines in Table 2 to obtain the corresponding conclusion data in the paper.
Points 19: Page 17, lines 579-580. You have to support your discussion with the results. Thus, the MARE results (12.3 and 7.22) must be reported here.
Response 19: In the lines 561-562, we have added the MARE of the adjusted nighttime lighting data to forecast the GDP and EPC, so as to better support our discussion.
Points 20: Page 19, line 603. Check the results (0.83 and 0.845).
Response 20: The 0.859 in the line 674 is the R2 value of the model between the TNL of the 2015 NPP/VIIRS annual composite data published by the Paynes Institute for Public Policy, Colorado School of Mines and GDP. The 0.827 in the line 674 is the R2 value of the model between the TNL of the 2015 NPP/VIIRS annual composite data published by the Paynes Institute for Public Policy, Colorado School of Mines and EPC. These two R2 values are used to show that there are still short-term lights existing in the annual NPP/VIIRS images which only the background noise removed.
The reason of the numbers (0.83 and 0.845) changing: Agriculture is an important part of social and economic activities, especially in a traditional agricultural country like China. However, most of the agricultural production is located in the darker area of the nighttime light image, so the nighttime light data cannot represent the agricultural part of economic activities [4]. It is necessary to remove the contribution of agricultural sector to economic activities when using nighttime data to predict GDP. The GDP data in the following paragraphs leave out the data of the agriculture sector. We have added the description part in the line 212-217. And the content related to GDP has been recalculated.
Points 21: Page 19, line 614. Check the results (0.866).
Response 21: The 0.863 in the line 584 is the R2 values of the model between TNL of the F162010 radiometric calibration nighttime light image and GDP. The R2 values of the model between TNL of nighttime light image of 2010 under the correction method in this paper and GDP was 0.817, the R2 values of the model between TNL of nighttime light image of 2010 under the reference correction method and GDP was 0.773, so it can be found that the correction method in this paper plays a greater role in reducing the "saturation effect" of the image than the reference correction method. We have added the explanation in the lines 583-587.
The reason of the number (0.866) changing: Agriculture is an important part of social and economic activities, especially in a traditional agricultural country like China. However, most of the agricultural production is located in the darker area of the nighttime light image, so the nighttime light data cannot represent the agricultural part of economic activities [4]. It is necessary to remove the contribution of agricultural sector to economic activities when using nighttime data to predict GDP. The GDP data in the following paragraphs leave out the data of the agriculture sector. We have added the description part in the line 212-217. And the content related to GDP has been recalculated.
Points 22: Pages 19-20, lines 619-631. When you use a term “better than”, you have to write the result (value) in parentheses. For example:
By comparing the results of 2015, it can be seen that the R2 value (0.764) of the model between TNL after correction …..
Response 22: Thank you for your suggestion. We have reviewed the manuscript again and change the sentences in the lines 591-596.
Points 23: Page 24, line 724. The DMSP/OLS must be written as a one word, there must be no space between the slash and letters. Please check the whole manuscript.
Response 23: Thank you for your reminder. We have reviewed the manuscript and have completed the modification of the corresponding content.
Response to Reviewer 5 Comments
The paper deals with the implementation of a procedure for building a long time series of night light images from 2001 to 2018 by fitting the DMSP/OLS Data and NPP/VIIRS data. The topic is current and almost “popular”, as witness by the relative large numbers of papers published in the very last years.
Points 1: In this framework, my main challenge with this paper is to understand how much different it is from the other already published on the topic, as well as an assessment of the results achieved by comparison with output of very recent methods, for example with the results shown in Xuecao Li et al., 2020 (https://www.nature.com/articles/s41597-020-0510-y). Moreover, even if I’m not an English native speaker, I can say that due to the poor level of English used, the difficulties in understand the current value of the paper are further increased.
Response 1: Aiming at the problems in the current research, this paper proposes a correction method of nighttime light data based on EVI data, so as to obtain the nighttime light dataset from 2001 to 2018 to represent economic data reliably and accurately. By comparing the correction method mentioned in this manuscript with those correction method mentioned in the following literature [1, 2], the author shows the advantages of the correction method mentioned in this manuscript. We have modified lines 156-174 to make the purpose of the study clearer.
As for English language, this manuscript has been submitted to MDPI for English embellishment.
Points 2: The paper is not well structured, section 2 should be a brief inset of the Introduction section, where the author should better state which is the approach they’re going to use and why.
Response 2: Thank you for your suggestion. We have rearranged the structure of this manuscript, merging the second part and the first part into the first part. In this manuscript, we explain the methods used by researchers in nighttime light data correction, and summarize some common shortcomings of the current correction methods in the lines 156-164.
Points 3: Concerning the methodology, even if it seems quite coherent, there are several doubts. The main one is related with Section 4.2.1. Inter-calibration, where the authors states (Line 314-315) that they assign to each DN higher than 0 a value equal to 1. My doubt is, how many pixels show values below 0? If they are present, are they physically based?
Response 3: The DN values of the original DMSP/OLS nighttime light data from 2001 to 2013 range from 0 to 63, the saturated DN value is 63, and the DN values of the dark, non-light areas and unstable light areas are 0. Therefore, the regions where the DN values are greater than 0 are stable light areas. In this manuscript, pixels with DN value greater than 0 are assigned as 1, while pixels with DN value equal to 0 are still kept at 0, so as to distinguish the stable light area and non-light area (unstable light area) in nighttime light image. Because the nighttime light images are collected by the same sensor in the different years and different sensors in the same year, we believe that the stable bright pixels obtained by different sensors in the same year should be consistent. The bright pixels in the previous year must be the stable bright pixels in the later year. Finally, the stable nighttime light image mask of 2001-2013 is obtained by the method shown in the lines 312-320. Taking F182013 nighttime lighting image as an example, the number of pixels with DN value equal to 0 is 7682588, and the number of pixels with DN value greater than 0 is 1509110.
The regions where the DN values are equal to 0 are dark and non-light areas.
Points 4: Figure 2 and 3 are useless, as well as equations 1 and 2.
Response 4: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
We appreciate the reviewer’s suggestions earnestly and hope that the revision answers the issues raised by the reviewer.
Once again, thank you very much for your comments and suggestions.
References:
- Lv, Q.; Liu, H.; Wang, J.; et al. Multiscale analysis on spatiotemporal dynamics of energy consumption CO2 emissions in China: Utilizing the integrated of DMSP-OLS and NPP-VIIRS nighttime light datasets. Sci. Total Environ. 2020, 703, 134394.
- Li, X.; Gong, L. Correction and fitting of night light images of DMSP / OLS and VIIRS /DNB
. Bulletin of Surveying and Mapping. 2019, 138-146.
- Zhu, X.; Ma, M.; Yang, H.; et al. Modeling the Spatiotemporal Dynamics of Gross Domestic Product in China Using Extended Temporal Coverage Nighttime Light Data. Remote Sens.-Basel. 2017, 9, 626-644.
- Tilottama, G.; Powell, R.; Elvidge, C.; et al. Shedding Light on the Global Distribution of Economic Activity. The Open Geography Journal. 2010, 3, 147-160.
- Pozzi, F.; Small, C. Analysis of Urban Land Cover and Population Density in the United States. Photogramm. Eng. Rem. S. 2005, 6, 719-726.
- Weng, Q.H.; Lu, D.S.; Schubring, J. Estimation of land surface temperature-vegetation abundance relationship for urban heat island studies. Remote Sens. Environ. 2004, 89, 467-483.
- Florczyk, A.J.; Corbane, C.; Ehrlich, D.; Freire, S.; Kemper, T.; Maffenini, L.; Melchiorri, M.; Pesaresi, M.; Politis, P.; Schiavina, M.; et al. GHSL Data Package 2019; EUR 29788 EN; Publications Office of the European Union: Luxembourg, 2019; ISBN 978-92-76-08725-0. Available online: https://ghsl.jrc.ec.europa.eu/documents/GHSL_Data_Package_2019.pdf
- Hao, Y.; Zheng, S.; Zhao, M.; et al. Reexamining the relationships among urbanization, industrial structure, and environmental pollution in China—New evidence using the dynamic threshold panel model. Energy Reports. 2020, 6, 28-39.
- Li, T.; Wu, Z. Urbanization and Education Development in China. Research in Educational Development. 2019, 39, 1-10.
- Shi, K.; Yu, B.; Huang, Y.; et al. Evaluating the Ability of NPP-VIIRS Nighttime Light Data to Estimate the Gross Domestic Product and the Electric Power Consumption of China at Multiple Scales: A Comparison with DMSP-OLS Data. Remote Sens.-Basel. 2014, 6, 1705-1724.
Author Response File: Author Response.docx
Reviewer 2 Report
The authors have discussed about the Nighttime lights correction with EVI data and validated with GDP. However, this paper can be improved in many ways.
1. Writing and grammar can be improved. For example, there are many inconsistency and grammar issues like [Line 310]" Obtaining historical stable images in China, it is necessary to consider the image data collected
311 by sensors in different years and different sensors in the same year. "; [Line 404] "It assumes that..." - are you referring to a specific method?; [Line 466] "sharp increase trend" -"sharp increase" or "increasing trend".
2. The articles need to be better organized - e.g., you can use another level of heading (instead of number in circles) to show 4.2.1. Inter-calibration steps.
3. The authors have mentioned the "reference correction method" [Line 475] at the end of the methodology part, can you please explicitly define what exactly is the reference method. It is very hard to follow after you have introduced various data processing steps.
4. As for the data validation, why compare NTL with GDP since NTL is not a direct measurement for GDP. GDP can be affected by economic structures at various subnational levels. Basically the graphs and tables you have presented is only helping readers to explore the relationship between lights and total GDP. However, for regions that mainly reply on agriculture production, how do you use nighttime lights to capture this.
5. Why EVI index is introduced instead of other remote sensing data? If an additional data source is added (like population distribution from GHSL https://ghsl.jrc.ec.europa.eu/), will this improve the model performance.
Author Response
Response to Reviewer 2 Comments
We appreciate valuable suggestions and comments from the anonymous reviewer. We have revised our manuscript according to your helpful suggestions.
Points 1: Writing and grammar can be improved. For example, there are many inconsistency and grammar issues like [Line 310]" Obtaining historical stable images in China, it is necessary to consider the image data collected by sensors in different years and different sensors in the same year. "; [Line 404] "It assumes that..." - are you referring to a specific method?; [Line 466] "sharp increase trend" -"sharp increase" or "increasing trend".
Response 1: Thank you for your suggestion. We have modified the corresponding sentences in the line 309 and 458. The assumption in the line 394 is a precondition of the hypothesis, which is considered to be within the two intervals of 2012-2015 and 2016-2018, there will not be a large number of pixels changing from dim to bright in the two stages.
Points 2: The articles need to be better organized - e.g., you can use another level of heading (instead of number in circles) to show 4.2.1. Inter-calibration steps.
Response 2: We have readjusted the structure of the article, such as merging the second part and the first part of the article into the first part, explaining the methods adopted by researchers in correcting the nighttime light data, and summarizing some common shortcomings of the current correction methods. We have modified the corresponding sentences to remove the circles in the line 312-332.
Points 3: The authors have mentioned the "reference correction method" [Line 475] at the end of the methodology part, can you please explicitly define what exactly is the reference method. It is very hard to follow after you have introduced various data processing steps.
Response 3: The reference correction method mentioned in this paper is the method used in the following literature [1, 2] for DMSP/OLS and NPP/VIIRS data correction. This part has been added in the line 446-453.
Points 4: As for the data validation, why compare NTL with GDP since NTL is not a direct measurement for GDP. GDP can be affected by economic structures at various subnational levels. Basically the graphs and tables you have presented is only helping readers to explore the relationship between lights and total GDP. However, for regions that mainly reply on agriculture production, how do you use nighttime lights to capture this.
Response 4: Thank you for your suggestion. We are very much in favor of your opinion. Agriculture is an important part of social and economic activities, especially in a traditional agricultural country like China. However, most of the agricultural production is located in the darker area of the nighttime light image, so the nighttime light data cannot represent the agricultural part of economic activities [3]. It is necessary to remove the contribution of agricultural sector to economic activities when using nighttime data to predict GDP. The GDP data in the following paragraphs leave out the data of the agriculture sector. We have added the description part in the line 212-217. And the content related to GDP has been recalculated.
Points 5: Why EVI index is introduced instead of other remote sensing data? If an additional data source is added (like population distribution from GHSL https://ghsl.jrc.ec.europa.eu/), will this improve the model performance.
Response 5: Thank you for your suggestion. In this manuscript, EVI data is introduced to realize the saturation correction of DMSP/OLS data based on considering the negative correlation between vegetation and human activities (the intensity of human non-agricultural activities in the central area of the city is high, and the vegetation coverage is generally less; in the rural areas, the intensity of human non-agricultural activities is low, and the vegetation coverage is more) [4, 5]. And EVI data is a continuous dataset in time series, which can correct DMSP/OLS data in different years. The GHSL, a global information baseline describing the spatial evolution of human settlements in the past 40 years, was developed by using a symbolic machine learning model trained by the collected high-resolution samples, multitemporal Landsat imagery in the epochs 1975, 1990, 2000 and 2015 [6]. Although we can interpolate the data between the years to match the period of the nightlight data, it is difficult for us to obtain GHSL data on continuous time series as accurately as EVI data. However, we agree with your opinion very much. At the same time, your view also gives us great inspiration. Applying GHSL data to DMSP/OLS data correction can effectively improve the performance of the model, and can better distinguish and identify built-up areas and non-built areas Therefore, the next stage of work will focus on how to combine DMSP/OLS data with GHSL data to better study how cities expand.
Response to Reviewer 1 Comments
We appreciate valuable suggestions and comments from the anonymous reviewer. We have revised our manuscript according to your helpful suggestions.
Major comments:
Points 1: This study might set the bar too low to present its novelty. The objectives of this study should be better explained. What is the difference with the previous similar studies?
Response 1: Thank you for your suggestions. Aiming at the problems in the current research, this paper proposes a correction method of nighttime light data based on EVI data, so as to obtain the nighttime light dataset from 2001 to 2018 to represent economic data reliably and accurately. By comparing the correction method mentioned in this manuscript with those correction method mentioned in the following literature [1, 2], we illustrate the superiority of the performance of the correction method mentioned in this manuscript. We have modified lines 156-174 to make the purpose of the study clearer.
Points 2: The manuscript adequately describes the equation used. However, the authors do not explain why in the approach. For example, average DN value of the pixel is based on the correction formulas from formula 10 and 11.
Response 2: We have strengthened the explanation of the formula used in the lines 356-361, 367-370, 373-378 and 381-384.
Points 3: The study generates consistent nighttime light data set from 2001 to 2018. The established regression model was used to fit the NPP/VIIRS data from 2014 to 2018 to the DMSP/OLS data scale. Could the authors transform the DMSP/OLS data to NPP/VIIRS data?
Response 3: Thank you for your suggestions. Since DMSP/OLS data is no longer updated and NPP/VIIRS data is still being updated, in theory, DMSP/OLS data should be converted to NPP/VIIRS data. However, due to the short time span of NPP/VIIRS data in this study (only five years), thus considering the accuracy and validity of the fitting results, the NPP/VIIRS data from 2014 to 2018 should be converted to DMSP/OLS data. At present, most scholars [1, 2, 7] adopt the method of transforming NPP/VIIRS data into DMSP/OLS data.
Points 4: We can’t see the spatial details. Fig 2 and 3 are too small.
Response 4: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Minor comments:
Points 1: The x-axis in Fig 7 ranges are different with a, c and b, d.
Response 1: The X & Y axis tick label have be improved by using scientific notation.
Response to Reviewer 3 Comments
This work aims to accomplish two most critical thresholds in performing long time series analysis using Nighttime Light (NTL), which are
Correction of saturation in DMSP-OLS annual composite
Inter-annual and intra-annual correction of DMSP-OLS annual composite
Bridging DMSP-OLS and VIIRS-DNB nighttime light imagery
The authors then verified the resulted NTL series by comparing GDP and EPC of selected provinces in China, and confirmed to have very high agreement.
The manuscript is very well laid out with some room for improvement as listed below.
Points 1: Line 80: EVI should be described more in Introduction section.
Response 1: Thank you for your valuable suggestions. We have added some description on the EVI data in the line 104-108.
Points 2: Line 161: DMSP/OLS Data -> DMSP/OLS data.
Response 2: Thank you for your patience in reading this manuscript. We have read through the full manuscript, and have changed all the contents related to DMSP/OLS Data to DMSP/OLS data.
Points 3: L194, L204, L215: The group which makes these products (EOG, Earth Observation Group) is no longer at NOAA/NCEI and now affiliated to the Paynes Institute for Public Policy, Colorado School of Mines.
Response 3: We have changed the corresponding institutions and websites to the current ones.
Points 4: Figure 1: NPP/VIIRS Stable Data for China should not need inter-annual correction?
Response 4: Thank you for your suggestions. Although the NPP/VIIRS data is collected by the same sensor, it can be seen from Figure 4-a that the NPP/VIIRS data before uncorrected are not all increasing. The TNL data from 2015 to 2016 and from 2017 to 2018 show a slight downward trend. Therefore, we believe that in order to make the results more appropriate to China's actual urbanization process, the nighttime light data more accurately predicting GDP and EPC, the NPP/VIIRS data should be carried out with inter-annual correction.
Points 5: Figure 2 & 3: These two figures does not show the difference clearly. Suggest enlarge some smaller regions to show the effect of correction. NPP/VIIRS should also be included in the caption of Figure 3.
Response 5: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Points 6: Line 314 to 324: The english writing in this paragraph is unclear. Please improve.
Response 6: We have modified the corresponding sentences in the line 312-320.
Points 7: L372: The reference [48] only exams data up to year 2008, it is not very convincing to extrapolate the conclusion all the way to 2018. Authors should make clear about the assumption they are making here.
Response 7: Thank you for your suggestion. The author has noticed this problem and updated the latest references [8, 9]. It can be seen from reference [9] that China's urbanization rate showed an upward trend from 2004 to 2017. From reference [8], we can see that China's urbanization increase rate is 1.36% from 2000 to 2010 and 1.1% from 2010 to 2018. Therefore, according to these two references, the author can assume that there is no reverse urbanization process in China from 2001 to 2018.
Points 8: L402-404: Is the research period up to year 2018 or 2019?
Response 8: Thank you for your suggestions. The research period is 2002-2018, which has been revised in the line 392-393.
Points 9: L403-410: The english writing in this paragraph is unclear, please improve.
Response 9: Thank you for your suggestions. We have modified the corresponding sentences in the line 393-399.
Points 10: L515-516, Figure 6: What is the reference correction used here? Please make it clear.
Response 10: The reference correction method mentioned in this paper is the method used in the following literature [1, 2] for DMSP/OLS and NPP/VIIRS data correction. This part has been added in line 446-453.
Points 11: Figure 7: The X & Y axis tick label should be improved by using scientific notation or difference unit.
Response 11: Thank you for your suggestions. The X & Y axis tick label have be improved by using scientific notation.
Points 12: Table 1 & 2: What is the data source of GDP and EPC?
Response 12: Gross domestic product (GDP) and electric power consumption (EPC) from 2002 to 2019 year are provided by the China statistical yearbook which published by China National Bureau of statistics (http://www.stats.gov.cn/tjsj/ndsj/).
Points 13: Table 5 & 6: What is the unit of these numbers?
Response 13: It shows the mean absolute relative error between the real GDP and the GDP predicted by the models based on the GDP and corrected nighttime light data at provincial scale under the two correction methods in Table 5. At the same time, it shows the mean absolute relative error between the real EPC and the EPC predicted by the models based on the EPC and corrected nighttime light data at provincial scale under the two correction methods in Table 6. Therefore, these numbers are the mean absolute relative errors without units.
Points 14: Table 7: Why is the numbers shown in this table having fixed numbers after decimal point of .33 and .67?
Response 14: The numbers in table 7 are obtained through the classification of results in tables 5 and 6. The mean absolute relative errors of 30 provinces are divided into three categories [10]: 0-25% is high accuracy, 25%-50% is medium accuracy, and greater than 50% is inaccuracy. The reason of the numbers shown in Table 7 having fixed numbers after the decimal point of. 33 and. 67 is that the research object of this paper is 30 provinces in China, and the fixed decimal point (.00, .33, and .67) is obtained by dividing the total of the provinces after statistics by 30 provinces.
Response to Reviewer 4 Comments
The manuscript entitled "Nighttime light data correction based on EVI data by invariant region method" aimed to address the overglow and saturation effects of the DMSP/OLS data by using the EVI product.
I believe the manuscript is suitable for publication in Remote Sensing, after the authors have addressed the comments below.
Points 1: Page 1, line 19. You mentioned that (… on national and provincial scales, respectively). Where? You have to mention the country.
Response 1: Thank you for your suggestions. We have defined the research area of this paper to 30 provinces of China in the line 16. The national scale mentioned here is to take China as a whole research area, so as to analyze the fluctuation of TNL over time and the ability to predict GDP and EPC. The provincial scale refers to taking each province in China as an independent research area to analyze the ability of predicting GDP and EPC in each province.
Points 2: Page 2, line 46-50. There are other nighttime sensors. Therefore, when you mentioned only the DMSP/OLS and NPP/VIIRS, you have to say something like “the most widely used” or “the popular nighttime sensors”.
Response 2: Thank you for your suggestions. We have modified the corresponding sentences in the line 49.
Points 3: Page 2, line 77. Change the sentence to be (… is more serious than that between the difference products of the DMSP/OLS).
Response 3: Thank you for your suggestions. We have modified the corresponding sentences in the line 145-147.
Points 4: Page 4, line 185. A site location map is more important than Figures 2 and 3.
Response 4: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Points 5: Page 8, line 321. You mentioned that (… the stable nighttime light image mask of 2018 and the nighttime light image ….). I believe you mean light image mask of 2008 not 2018. Check.
Response 5: Thank you for your patience in reading this manuscript, the year of the stable nighttime light image mask is actually 2008. We have modified the corresponding sentences in the line 318.
Points 6: Page 9, Equations (4) and (5). These equations are similar. Thus, just keep one of them.
Response 6: Thank you for your valuable advice. Formula (5) is a general formula for the normalization of DN value, while formula (6) is a special case where the value of is taken as 0. The retention of formula (6) can simplify the writing of subsequent formulas. Therefore, experts are requested to agree to retain formula (5) and formula (6).
Points 7: Page 10, line 362. Put a reference after “Intra-annual composition”.
Response 7: Thank you for your suggestions. The corresponding reference has been added in the line 355.
Points 8: Page 10, line 374. Put a reference after “inter-annual series correction”.
Response 8: Thank you for your suggestions. The corresponding reference has been added in the line 366.
Points 9: Page 10, Equation (10). Would there be a difference if you used the inter-annual series correction that proposed by Liu et al. (2012) [Ref. 48].
Response 9: Using the inter-annual correction method proposed by Liu et al. it can ensure that the DN value of the pixel in the next year's image is not less than that of the previous year's image. This adjustment order can avoid weakening the DN value of the pixel in the early annual image, but will cause the DN value of the pixel in the later year's image to be excessively enlarged. Therefore, the correction in the opposite direction should be carried out on the basis of the inter-annual correction method proposed by Liu et al. to ensure that the DN value of pixel in the image of the previous year is not greater than that of the pixel in the image of the next year. Then taking the average value after two inter-annual correction as the final DN value, we can obtain a more realistic nighttime light image.
Points 10: Page 11, lines 402-403. Are you sure it is 2019 not 2018.
Response 10: The research period is from 2002 to 2018. We have modified the corresponding sentences in the line 392-393.
Points 11: Page 11, lines 405-410. Why did not you choose the 2018 annual composite data as a binary-value mask instead of 2016?
Response 11: At present, the monthly composites published by the Paynes Institute for Public Policy, Colorado School of Mines are all monthly products, has not been filtered to screen out lights from aurora, fires, boats, and other temporal lights. Therefore, the nighttime light images we synthesized for 2018 also include these short-term light data. The annual composite data published only contain the 2015 annual composite data and 2016 annual composite data which has been filtered to screen out short-term lights. Hence, we choose the 2016 annual composite data as a binary-value mask instead of 2018.
Points 12: Page 13, lines 467-468. Re-write the sentence.
Response 12: Thank you for your suggestions. We have modified the corresponding sentences in the line 459-461.
Points 13: Page 13, lines 474-475. Re-write the sentence.
Response 13: Thank you for your suggestions. We have modified the corresponding sentences in the line 466-467.
Points 14: Page 13, line 475. What do you mean by the reference correction method? You have to define it.
Response 14: The reference correction method mentioned in this paper is the method used in the following literature [1, 2] for DMSP/OLS and NPP/VIIRS data correction. This part has been added in line 446-453.
Points 15: Page 16, lines 540-542. Re-write the sentence.
Response 15: Thank you for your suggestions. We have modified the corresponding sentences in the line 535-538.
Points 16: Page 16, lines 548-554. From which tables did you get the results of the MARE before and after 2013? If there is no table, you have to create one that summarises the results.
Response 16: Thank you for your suggestions. We added lines 21-22 in Table 1 and lines 21-22 lines in Table 2 to obtain the corresponding conclusion data in the paper.
Points 17: Page, Figure 7. Change the x-axis unit to be in million and ranges from 0 to 30.
Response 17: Thank you for your suggestions. The X & Y axis tick label have be improved by using scientific notation.
Points 18: Page 17, lines 572-578. Same as Comment No. (16).
Response 18: Thank you for your suggestions. We added lines 21-22 in Table 1 and lines 21-22 lines in Table 2 to obtain the corresponding conclusion data in the paper.
Points 19: Page 17, lines 579-580. You have to support your discussion with the results. Thus, the MARE results (12.3 and 7.22) must be reported here.
Response 19: In the lines 561-562, we have added the MARE of the adjusted nighttime lighting data to forecast the GDP and EPC, so as to better support our discussion.
Points 20: Page 19, line 603. Check the results (0.83 and 0.845).
Response 20: The 0.859 in the line 674 is the R2 value of the model between the TNL of the 2015 NPP/VIIRS annual composite data published by the Paynes Institute for Public Policy, Colorado School of Mines and GDP. The 0.827 in the line 674 is the R2 value of the model between the TNL of the 2015 NPP/VIIRS annual composite data published by the Paynes Institute for Public Policy, Colorado School of Mines and EPC. These two R2 values are used to show that there are still short-term lights existing in the annual NPP/VIIRS images which only the background noise removed.
The reason of the numbers (0.83 and 0.845) changing: Agriculture is an important part of social and economic activities, especially in a traditional agricultural country like China. However, most of the agricultural production is located in the darker area of the nighttime light image, so the nighttime light data cannot represent the agricultural part of economic activities [3]. It is necessary to remove the contribution of agricultural sector to economic activities when using nighttime data to predict GDP. The GDP data in the following paragraphs leave out the data of the agriculture sector. We have added the description part in the line 212-217. And the content related to GDP has been recalculated.
Points 21: Page 19, line 614. Check the results (0.866).
Response 21: The 0.863 in the line 584 is the R2 values of the model between TNL of the F162010 radiometric calibration nighttime light image and GDP. The R2 values of the model between TNL of nighttime light image of 2010 under the correction method in this paper and GDP was 0.817, the R2 values of the model between TNL of nighttime light image of 2010 under the reference correction method and GDP was 0.773, so it can be found that the correction method in this paper plays a greater role in reducing the "saturation effect" of the image than the reference correction method. We have added the explanation in the lines 583-587.
The reason of the number (0.866) changing: Agriculture is an important part of social and economic activities, especially in a traditional agricultural country like China. However, most of the agricultural production is located in the darker area of the nighttime light image, so the nighttime light data cannot represent the agricultural part of economic activities [3]. It is necessary to remove the contribution of agricultural sector to economic activities when using nighttime data to predict GDP. The GDP data in the following paragraphs leave out the data of the agriculture sector. We have added the description part in the line 212-217. And the content related to GDP has been recalculated.
Points 22: Pages 19-20, lines 619-631. When you use a term “better than”, you have to write the result (value) in parentheses. For example:
By comparing the results of 2015, it can be seen that the R2 value (0.764) of the model between TNL after correction …..
Response 22: Thank you for your suggestion. We have reviewed the manuscript again and change the sentences in the lines 591-596.
Points 23: Page 24, line 724. The DMSP/OLS must be written as a one word, there must be no space between the slash and letters. Please check the whole manuscript.
Response 23: Thank you for your reminder. We have reviewed the manuscript and have completed the modification of the corresponding content.
Response to Reviewer 5 Comments
The paper deals with the implementation of a procedure for building a long time series of night light images from 2001 to 2018 by fitting the DMSP/OLS Data and NPP/VIIRS data. The topic is current and almost “popular”, as witness by the relative large numbers of papers published in the very last years.
Points 1: In this framework, my main challenge with this paper is to understand how much different it is from the other already published on the topic, as well as an assessment of the results achieved by comparison with output of very recent methods, for example with the results shown in Xuecao Li et al., 2020 (https://www.nature.com/articles/s41597-020-0510-y). Moreover, even if I’m not an English native speaker, I can say that due to the poor level of English used, the difficulties in understand the current value of the paper are further increased.
Response 1: Aiming at the problems in the current research, this paper proposes a correction method of nighttime light data based on EVI data, so as to obtain the nighttime light dataset from 2001 to 2018 to represent economic data reliably and accurately. By comparing the correction method mentioned in this manuscript with those correction method mentioned in the following literature [1, 2], the author shows the advantages of the correction method mentioned in this manuscript. We have modified lines 156-174 to make the purpose of the study clearer.
As for English language, this manuscript has been submitted to MDPI for English embellishment.
Points 2: The paper is not well structured, section 2 should be a brief inset of the Introduction section, where the author should better state which is the approach they’re going to use and why.
Response 2: Thank you for your suggestion. We have rearranged the structure of this manuscript, merging the second part and the first part into the first part. In this manuscript, we explain the methods used by researchers in nighttime light data correction, and summarize some common shortcomings of the current correction methods in the lines 156-164.
Points 3: Concerning the methodology, even if it seems quite coherent, there are several doubts. The main one is related with Section 4.2.1. Inter-calibration, where the authors states (Line 314-315) that they assign to each DN higher than 0 a value equal to 1. My doubt is, how many pixels show values below 0? If they are present, are they physically based?
Response 3: The DN values of the original DMSP/OLS nighttime light data from 2001 to 2013 range from 0 to 63, the saturated DN value is 63, and the DN values of the dark, non-light areas and unstable light areas are 0. Therefore, the regions where the DN values are greater than 0 are stable light areas. In this manuscript, pixels with DN value greater than 0 are assigned as 1, while pixels with DN value equal to 0 are still kept at 0, so as to distinguish the stable light area and non-light area (unstable light area) in nighttime light image. Because the nighttime light images are collected by the same sensor in the different years and different sensors in the same year, we believe that the stable bright pixels obtained by different sensors in the same year should be consistent. The bright pixels in the previous year must be the stable bright pixels in the later year. Finally, the stable nighttime light image mask of 2001-2013 is obtained by the method shown in the lines 312-320. Taking F182013 nighttime lighting image as an example, the number of pixels with DN value equal to 0 is 7682588, and the number of pixels with DN value greater than 0 is 1509110.
The regions where the DN values are equal to 0 are dark and non-light areas.
Points 4: Figure 2 and 3 are useless, as well as equations 1 and 2.
Response 4: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
We appreciate the reviewer’s suggestions earnestly and hope that the revision answers the issues raised by the reviewer.
Once again, thank you very much for your comments and suggestions.
References:
- Lv, Q.; Liu, H.; Wang, J.; et al. Multiscale analysis on spatiotemporal dynamics of energy consumption CO2 emissions in China: Utilizing the integrated of DMSP-OLS and NPP-VIIRS nighttime light datasets. Sci. Total Environ. 2020, 703, 134394.
- Li, X.; Gong, L. Correction and fitting of night light images of DMSP / OLS and VIIRS /DNB
. Bulletin of Surveying and Mapping. 2019, 138-146.
- Tilottama, G.; Powell, R.; Elvidge, C.; et al. Shedding Light on the Global Distribution of Economic Activity. The Open Geography Journal. 2010, 3, 147-160.
- Pozzi, F.; Small, C. Analysis of Urban Land Cover and Population Density in the United States. Photogramm. Eng. Rem. S. 2005, 6, 719-726.
- Weng, Q.H.; Lu, D.S.; Schubring, J. Estimation of land surface temperature-vegetation abundance relationship for urban heat island studies. Remote Sens. Environ. 2004, 89, 467-483.
- Florczyk, A.J.; Corbane, C.; Ehrlich, D.; Freire, S.; Kemper, T.; Maffenini, L.; Melchiorri, M.; Pesaresi, M.; Politis, P.; Schiavina, M.; et al. GHSL Data Package 2019; EUR 29788 EN; Publications Office of the European Union: Luxembourg, 2019; ISBN 978-92-76-08725-0. Available online: https://ghsl.jrc.ec.europa.eu/documents/GHSL_Data_Package_2019.pdf
- Zhu, X.; Ma, M.; Yang, H.; et al. Modeling the Spatiotemporal Dynamics of Gross Domestic Product in China Using Extended Temporal Coverage Nighttime Light Data. Remote Sens.-Basel. 2017, 9, 626-644.
- Hao, Y.; Zheng, S.; Zhao, M.; et al. Reexamining the relationships among urbanization, industrial structure, and environmental pollution in China—New evidence using the dynamic threshold panel model. Energy Reports. 2020, 6, 28-39.
- Li, T.; Wu, Z. Urbanization and Education Development in China. Research in Educational Development. 2019, 39, 1-10.
- Shi, K.; Yu, B.; Huang, Y.; et al. Evaluating the Ability of NPP-VIIRS Nighttime Light Data to Estimate the Gross Domestic Product and the Electric Power Consumption of China at Multiple Scales: A Comparison with DMSP-OLS Data. Remote Sens.-Basel. 2014, 6, 1705-1724.
Author Response File: Author Response.docx
Reviewer 3 Report
This work aims to accomplish two most critical thresholds in performing long time series analysis using Nighttime Light (NTL), which are
- Correction of saturation in DMSP-OLS annual composite
- Inter-annual and intra-annual correction of DMSP-OLS annual composite
- Bridging DMSP-OLS and VIIRS-DNB nighttime light imagery
The authors then verified the resulted NTL series by comparing GDP and EPC of selected provinces in China, and confirmed to have very high agreement.
The manuscript is very well laid out with some room for improvement as listed below.
[1] Line 80: EVI should be described more in Introduction section.
[2] Line 161: DMSP/OLS Data -> DMSP/OLS data.
[3] L194, L204, L215: The group which makes these products (EOG, Earth Observation Group) is no longer at NOAA/NCEI and now affiliated to the Paynes Institute for Public Policy, Colorado School of Mines.
[4] Figure 1: NPP/VIIRS Stable Data for China should not need inter-annual correction?
[5] Figure 2 & 3: These two figures does not show the difference clearly. Suggest enlarge some smaller regions to show the effect of correction. NPP/VIIRS should also be included in the caption of Figure 3.
[6] Line 314 to 324: The english writing in this paragraph is unclear. Please improve.
[7] L372: The reference [48] only exams data up to year 2008, it is not very convincing to extrapolate the conclusion all the way to 2018. Authors should make clear about the assumption they are making here.
[8] L402-404: Is the research period up to year 2018 or 2019?
[9] L403-410: The english writing in this paragraph is unclear, please improve.
[10] L515-516, Figure 6: What is the reference correction used here? Please make it clear.
[11] Figure 7: The X & Y axis tick label should be improved by using scientific notation or difference unit.
[12] Table 1 & 2: What is the data source of GDP and EPC?
[13] Table 5 & 6: What is the unit of these numbers?
[14] Table 7: Why is the numbers shown in this table having fixed numbers after decimal point of .33 and .67?
Author Response
Response to Reviewer 3 Comments
We appreciate valuable suggestions and comments from the anonymous reviewer. We have revised our manuscript according to your helpful suggestions.
Points 1: Line 80: EVI should be described more in Introduction section.
Response 1: Thank you for your valuable suggestions. We have added some description on the EVI data in the line 104-108.
Points 2: Line 161: DMSP/OLS Data -> DMSP/OLS data.
Response 2: Thank you for your patience in reading this manuscript. We have read through the full manuscript, and have changed all the contents related to DMSP/OLS Data to DMSP/OLS data.
Points 3: L194, L204, L215: The group which makes these products (EOG, Earth Observation Group) is no longer at NOAA/NCEI and now affiliated to the Paynes Institute for Public Policy, Colorado School of Mines.
Response 3: We have changed the corresponding institutions and websites to the current ones.
Points 4: Figure 1: NPP/VIIRS Stable Data for China should not need inter-annual correction?
Response 4: Thank you for your suggestions. Although the NPP/VIIRS data is collected by the same sensor, it can be seen from Figure 4-a that the NPP/VIIRS data before uncorrected are not all increasing. The TNL data from 2015 to 2016 and from 2017 to 2018 show a slight downward trend. Therefore, we believe that in order to make the results more appropriate to China's actual urbanization process, the nighttime light data more accurately predicting GDP and EPC, the NPP/VIIRS data should be carried out with inter-annual correction.
Points 5: Figure 2 & 3: These two figures does not show the difference clearly. Suggest enlarge some smaller regions to show the effect of correction. NPP/VIIRS should also be included in the caption of Figure 3.
Response 5: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Points 6: Line 314 to 324: The english writing in this paragraph is unclear. Please improve.
Response 6: We have modified the corresponding sentences in the line 312-320.
Points 7: L372: The reference [48] only exams data up to year 2008, it is not very convincing to extrapolate the conclusion all the way to 2018. Authors should make clear about the assumption they are making here.
Response 7: Thank you for your suggestion. The author has noticed this problem and updated the latest references [1, 2]. It can be seen from reference [2] that China's urbanization rate showed an upward trend from 2004 to 2017. From reference [1], we can see that China's urbanization increase rate is 1.36% from 2000 to 2010 and 1.1% from 2010 to 2018. Therefore, according to these two references, the author can assume that there is no reverse urbanization process in China from 2001 to 2018.
Points 8: L402-404: Is the research period up to year 2018 or 2019?
Response 8: Thank you for your suggestions. The research period is 2002-2018, which has been revised in the line 392-393.
Points 9: L403-410: The english writing in this paragraph is unclear, please improve.
Response 9: Thank you for your suggestions. We have modified the corresponding sentences in the line 393-399.
Points 10: L515-516, Figure 6: What is the reference correction used here? Please make it clear.
Response 10: The reference correction method mentioned in this paper is the method used in the following literature [3, 4] for DMSP/OLS and NPP/VIIRS data correction. This part has been added in line 446-453.
Points 11: Figure 7: The X & Y axis tick label should be improved by using scientific notation or difference unit.
Response 11: Thank you for your suggestions. The X & Y axis tick label have be improved by using scientific notation.
Points 12: Table 1 & 2: What is the data source of GDP and EPC?
Response 12: Gross domestic product (GDP) and electric power consumption (EPC) from 2002 to 2019 year are provided by the China statistical yearbook which published by China National Bureau of statistics (http://www.stats.gov.cn/tjsj/ndsj/).
Points 13: Table 5 & 6: What is the unit of these numbers?
Response 13: It shows the mean absolute relative error between the real GDP and the GDP predicted by the models based on the GDP and corrected nighttime light data at provincial scale under the two correction methods in Table 5. At the same time, it shows the mean absolute relative error between the real EPC and the EPC predicted by the models based on the EPC and corrected nighttime light data at provincial scale under the two correction methods in Table 6. Therefore, these numbers are the mean absolute relative errors without units.
Points 14: Table 7: Why is the numbers shown in this table having fixed numbers after decimal point of .33 and .67?
Response 14: The numbers in table 7 are obtained through the classification of results in tables 5 and 6. The mean absolute relative errors of 30 provinces are divided into three categories [5]: 0-25% is high accuracy, 25%-50% is medium accuracy, and greater than 50% is inaccuracy. The reason of the numbers shown in Table 7 having fixed numbers after the decimal point of. 33 and. 67 is that the research object of this paper is 30 provinces in China, and the fixed decimal point (.00, .33, and .67) is obtained by dividing the total of the provinces after statistics by 30 provinces.
We appreciate the reviewer’s suggestions earnestly and hope that the revision answers the issues raised by the reviewer.
Once again, thank you very much for your comments and suggestions.
References:
- Hao, Y.; Zheng, S.; Zhao, M.; et al. Reexamining the relationships among urbanization, industrial structure, and environmental pollution in China—New evidence using the dynamic threshold panel model. Energy Reports. 2020, 6, 28-39.
- Li, T.; Wu, Z. Urbanization and Education Development in China. Research in Educational Development. 2019, 39, 1-10.
- Lv, Q.; Liu, H.; Wang, J.; et al. Multiscale analysis on spatiotemporal dynamics of energy consumption CO2 emissions in China: Utilizing the integrated of DMSP-OLS and NPP-VIIRS nighttime light datasets. Sci. Total Environ. 2020, 703, 134394.
- Li, X.; Gong, L. Correction and fitting of night light images of DMSP / OLS and VIIRS /DNB
. Bulletin of Surveying and Mapping. 2019, 138-146.
- Shi, K.; Yu, B.; Huang, Y.; et al. Evaluating the Ability of NPP-VIIRS Nighttime Light Data to Estimate the Gross Domestic Product and the Electric Power Consumption of China at Multiple Scales: A Comparison with DMSP-OLS Data. Remote Sens.-Basel. 2014, 6, 1705-1724.
Author Response File: Author Response.docx
Reviewer 4 Report
The manuscript entitled "Nighttime light data correction based on EVI data by invariant region method" aimed to address the overglow and saturation effects of the DMSP/OLS data by using the EVI product.
I believe the manuscript is suitable for publication in Remote Sensing, after the authors have addressed the comments below.
1) Page 1, line 19. You mentioned that (… on national and provincial scales, respectively). Where? You have to mention the country.
2) Page 2, line 46-50. There are other nighttime sensors. Therefore, when you mentioned only the DMSP/OLS and NPP/VIIRS, you have to say something like “the most widely used” or “the popular nighttime sensors”.
3) Page 2, line 77. Change the sentence to be (… is more serious than that between the difference products of the DMSP/OLS).
4) Page 4, line 185. A site location map is more important than Figures 2 and 3.
5) Page 8, line 321. You mentioned that (… the stable nighttime light image mask of 2018 and the nighttime light image ….). I believe you mean light image mask of 2008 not 2018. Check.
6) Page 9, Equations (4) and (5). These equations are similar. Thus, just keep one of them.
7) Page 10, line 362. Put a reference after “Intra-annual composition”.
8) Page 10, line 374. Put a reference after “inter-annual series correction”.
9) Page 10, Equation (10). Would there be a difference if you used the inter-annual series correction that proposed by Liu et al. (2012) [Ref. 48].
10) Page 11, lines 402-403. Are you sure it is 2019 not 2018.
11) Page 11, lines 405-410. Why did not you choose the 2018 annual composite data as a binary-value mask instead of 2016?
12) Page 13, lines 467-468. Re-write the sentence.
13) Page 13, lines 474-475. Re-write the sentence.
14) Page 13, line 475. What do you mean by the reference correction method? You have to define it.
15) Page 16, lines 540-542. Re-write the sentence.
16) Page 16, lines 548-554. From which tables did you get the results of the MARE before and after 2013? If there is no table, you have to create one that summarises the results.
17) Page, Figure 7. Change the x-axis unit to be in million and ranges from 0 to 30.
18) Page 17, lines 572-578. Same as Comment No. (16).
19) Page 17, lines 579-580. You have to support your discussion with the results. Thus, the MARE results (12.3 and 7.22) must be reported here.
20) Page 19, line 603. Check the results (0.83 and 0.845).
21) Page 19, line 614. Check the results (0.866).
22) Pages 19-20, lines 619-631. When you use a term “better than”, you have to write the result (value) in parentheses. For example:
By comparing the results of 2015, it can be seen that the R2 value (0.764) of the model between TNL after correction …..
23) Page 24, line 724. The DMSP/OLS must be written as a one word, there must be no space between the slash and letters. Please check the whole manuscript.
Author Response
Response to Reviewer 4 Comments
We appreciate valuable suggestions and comments from the anonymous reviewer. We have revised our manuscript according to your helpful suggestions.
Points 1: Page 1, line 19. You mentioned that (… on national and provincial scales, respectively). Where? You have to mention the country.
Response 1: Thank you for your suggestions. We have defined the research area of this paper to 30 provinces of China in the line 16. The national scale mentioned here is to take China as a whole research area, so as to analyze the fluctuation of TNL over time and the ability to predict GDP and EPC. The provincial scale refers to taking each province in China as an independent research area to analyze the ability of predicting GDP and EPC in each province.
Points 2: Page 2, line 46-50. There are other nighttime sensors. Therefore, when you mentioned only the DMSP/OLS and NPP/VIIRS, you have to say something like “the most widely used” or “the popular nighttime sensors”.
Response 2: Thank you for your suggestions. We have modified the corresponding sentences in the line 49.
Points 3: Page 2, line 77. Change the sentence to be (… is more serious than that between the difference products of the DMSP/OLS).
Response 3: Thank you for your suggestions. We have modified the corresponding sentences in the line 145-147.
Points 4: Page 4, line 185. A site location map is more important than Figures 2 and 3.
Response 4: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Points 5: Page 8, line 321. You mentioned that (… the stable nighttime light image mask of 2018 and the nighttime light image ….). I believe you mean light image mask of 2008 not 2018. Check.
Response 5: Thank you for your patience in reading this manuscript, the year of the stable nighttime light image mask is actually 2008. We have modified the corresponding sentences in the line 318.
Points 6: Page 9, Equations (4) and (5). These equations are similar. Thus, just keep one of them.
Response 6: Thank you for your valuable advice. Formula (5) is a general formula for the normalization of DN value, while formula (6) is a special case where the value of is taken as 0. The retention of formula (6) can simplify the writing of subsequent formulas. Therefore, experts are requested to agree to retain formula (5) and formula (6).
Points 7: Page 10, line 362. Put a reference after “Intra-annual composition”.
Response 7: Thank you for your suggestions. The corresponding reference has been added in the line 355.
Points 8: Page 10, line 374. Put a reference after “inter-annual series correction”.
Response 8: Thank you for your suggestions. The corresponding reference has been added in the line 366.
Points 9: Page 10, Equation (10). Would there be a difference if you used the inter-annual series correction that proposed by Liu et al. (2012) [Ref. 48].
Response 9: Using the inter-annual correction method proposed by Liu et al. it can ensure that the DN value of the pixel in the next year's image is not less than that of the previous year's image. This adjustment order can avoid weakening the DN value of the pixel in the early annual image, but will cause the DN value of the pixel in the later year's image to be excessively enlarged. Therefore, the correction in the opposite direction should be carried out on the basis of the inter-annual correction method proposed by Liu et al. to ensure that the DN value of pixel in the image of the previous year is not greater than that of the pixel in the image of the next year. Then taking the average value after two inter-annual correction as the final DN value, we can obtain a more realistic nighttime light image.
Points 10: Page 11, lines 402-403. Are you sure it is 2019 not 2018.
Response 10: The research period is from 2002 to 2018. We have modified the corresponding sentences in the line 392-393.
Points 11: Page 11, lines 405-410. Why did not you choose the 2018 annual composite data as a binary-value mask instead of 2016?
Response 11: At present, the monthly composites published by the Paynes Institute for Public Policy, Colorado School of Mines are all monthly products, has not been filtered to screen out lights from aurora, fires, boats, and other temporal lights. Therefore, the nighttime light images we synthesized for 2018 also include these short-term light data. The annual composite data published only contain the 2015 annual composite data and 2016 annual composite data which has been filtered to screen out short-term lights. Hence, we choose the 2016 annual composite data as a binary-value mask instead of 2018.
Points 12: Page 13, lines 467-468. Re-write the sentence.
Response 12: Thank you for your suggestions. We have modified the corresponding sentences in the line 459-461.
Points 13: Page 13, lines 474-475. Re-write the sentence.
Response 13: Thank you for your suggestions. We have modified the corresponding sentences in the line 466-467.
Points 14: Page 13, line 475. What do you mean by the reference correction method? You have to define it.
Response 14: The reference correction method mentioned in this paper is the method used in the following literature [1, 2] for DMSP/OLS and NPP/VIIRS data correction. This part has been added in line 446-453.
Points 15: Page 16, lines 540-542. Re-write the sentence.
Response 15: Thank you for your suggestions. We have modified the corresponding sentences in the line 535-538.
Points 16: Page 16, lines 548-554. From which tables did you get the results of the MARE before and after 2013? If there is no table, you have to create one that summarises the results.
Response 16: Thank you for your suggestions. We added lines 21-22 in Table 1 and lines 21-22 lines in Table 2 to obtain the corresponding conclusion data in the paper.
Points 17: Page, Figure 7. Change the x-axis unit to be in million and ranges from 0 to 30.
Response 17: Thank you for your suggestions. The X & Y axis tick label have be improved by using scientific notation.
Points 18: Page 17, lines 572-578. Same as Comment No. (16).
Response 18: Thank you for your suggestions. We added lines 21-22 in Table 1 and lines 21-22 lines in Table 2 to obtain the corresponding conclusion data in the paper.
Points 19: Page 17, lines 579-580. You have to support your discussion with the results. Thus, the MARE results (12.3 and 7.22) must be reported here.
Response 19: In the lines 561-562, we have added the MARE of the adjusted nighttime lighting data to forecast the GDP and EPC, so as to better support our discussion.
Points 20: Page 19, line 603. Check the results (0.83 and 0.845).
Response 20: The 0.859 in the line 674 is the R2 value of the model between the TNL of the 2015 NPP/VIIRS annual composite data published by the Paynes Institute for Public Policy, Colorado School of Mines and GDP. The 0.827 in the line 674 is the R2 value of the model between the TNL of the 2015 NPP/VIIRS annual composite data published by the Paynes Institute for Public Policy, Colorado School of Mines and EPC. These two R2 values are used to show that there are still short-term lights existing in the annual NPP/VIIRS images which only the background noise removed.
The reason of the numbers (0.83 and 0.845) changing: Agriculture is an important part of social and economic activities, especially in a traditional agricultural country like China. However, most of the agricultural production is located in the darker area of the nighttime light image, so the nighttime light data cannot represent the agricultural part of economic activities [3]. It is necessary to remove the contribution of agricultural sector to economic activities when using nighttime data to predict GDP. The GDP data in the following paragraphs leave out the data of the agriculture sector. We have added the description part in the line 212-217. And the content related to GDP has been recalculated.
Points 21: Page 19, line 614. Check the results (0.866).
Response 21: The 0.863 in the line 584 is the R2 values of the model between TNL of the F162010 radiometric calibration nighttime light image and GDP. The R2 values of the model between TNL of nighttime light image of 2010 under the correction method in this paper and GDP was 0.817, the R2 values of the model between TNL of nighttime light image of 2010 under the reference correction method and GDP was 0.773, so it can be found that the correction method in this paper plays a greater role in reducing the "saturation effect" of the image than the reference correction method. We have added the explanation in the lines 583-587.
The reason of the number (0.866) changing: Agriculture is an important part of social and economic activities, especially in a traditional agricultural country like China. However, most of the agricultural production is located in the darker area of the nighttime light image, so the nighttime light data cannot represent the agricultural part of economic activities [3]. It is necessary to remove the contribution of agricultural sector to economic activities when using nighttime data to predict GDP. The GDP data in the following paragraphs leave out the data of the agriculture sector. We have added the description part in the line 212-217. And the content related to GDP has been recalculated.
Points 22: Pages 19-20, lines 619-631. When you use a term “better than”, you have to write the result (value) in parentheses. For example:
By comparing the results of 2015, it can be seen that the R2 value (0.764) of the model between TNL after correction …..
Response 22: Thank you for your suggestion. We have reviewed the manuscript again and change the sentences in the lines 591-596.
Points 23: Page 24, line 724. The DMSP/OLS must be written as a one word, there must be no space between the slash and letters. Please check the whole manuscript.
Response 23: Thank you for your reminder. We have reviewed the manuscript and have completed the modification of the corresponding content.
We appreciate the reviewer’s suggestions earnestly and hope that the revision answers the issues raised by the reviewer.
Once again, thank you very much for your comments and suggestions.
References:
- Lv, Q.; Liu, H.; Wang, J.; et al. Multiscale analysis on spatiotemporal dynamics of energy consumption CO2 emissions in China: Utilizing the integrated of DMSP-OLS and NPP-VIIRS nighttime light datasets. Sci. Total Environ. 2020, 703, 134394.
- Li, X.; Gong, L. Correction and fitting of night light images of DMSP / OLS and VIIRS /DNB. Bulletin of Surveying and Mapping. 2019, 138-146.
- Tilottama, G.; Powell, R.; Elvidge, C.; et al. Shedding Light on the Global Distribution of Economic Activity. The Open Geography Journal. 2010, 3, 147-160.
Author Response File: Author Response.docx
Reviewer 5 Report
The paper deals with the implementation of a procedure for building a long time series of night light images from 2001 to 2018 by fitting the DMSP/OLS Data and NPP/VIIRS data. The topic is current and almost “popular”, as witness by the relative large numbers of papers published in the very last years.
In this framework, my main challenge with this paper is to understand how much different it is from the other already published on the topic, as well as an assessment of the results achieved by comparison with output of very recent methods, for example with the results shown in Xuecao Li et al., 2020 (https://www.nature.com/articles/s41597-020-0510-y). Moreover, even if I’m not an English native speaker, I can say that due to the poor level of English used, the difficulties in understand the current value of the paper are further increased.
The paper is not well structured, section 2 should be a brief inset of the Introduction section, where the author should better state which is the approach they’re going to use and why.
Concerning the methodology, even if it seems quite coherent, there are several doubts. The main one is related with Section 4.2.1. Inter-calibration, where the authors states (Line 314-315) that they assign to each DN higher than 0 a value equal to 1. My doubt is, how many pixels show values below 0? If they are present, are they physically based?
Figure 2 and 3 are useless, as well as equations 1 and 2.
Author Response
Response to Reviewer 5 Comments
We appreciate valuable suggestions and comments from the anonymous reviewer. We have revised our manuscript according to your helpful suggestions.
Points 1: In this framework, my main challenge with this paper is to understand how much different it is from the other already published on the topic, as well as an assessment of the results achieved by comparison with output of very recent methods, for example with the results shown in Xuecao Li et al., 2020 (https://www.nature.com/articles/s41597-020-0510-y). Moreover, even if I’m not an English native speaker, I can say that due to the poor level of English used, the difficulties in understand the current value of the paper are further increased.
Response 1: Aiming at the problems in the current research, this paper proposes a correction method of nighttime light data based on EVI data, so as to obtain the nighttime light dataset from 2001 to 2018 to represent economic data reliably and accurately. By comparing the correction method mentioned in this manuscript with those correction method mentioned in the following literature [1, 2], the author shows the advantages of the correction method mentioned in this manuscript. We have modified lines 156-174 to make the purpose of the study clearer.
As for English language, this manuscript has been submitted to MDPI for English embellishment.
Points 2: The paper is not well structured, section 2 should be a brief inset of the Introduction section, where the author should better state which is the approach they’re going to use and why.
Response 2: Thank you for your suggestion. We have rearranged the structure of this manuscript, merging the second part and the first part into the first part. In this manuscript, we explain the methods used by researchers in nighttime light data correction, and summarize some common shortcomings of the current correction methods in the lines 156-164.
Points 3: Concerning the methodology, even if it seems quite coherent, there are several doubts. The main one is related with Section 4.2.1. Inter-calibration, where the authors states (Line 314-315) that they assign to each DN higher than 0 a value equal to 1. My doubt is, how many pixels show values below 0? If they are present, are they physically based?
Response 3: The DN values of the original DMSP/OLS nighttime light data from 2001 to 2013 range from 0 to 63, the saturated DN value is 63, and the DN values of the dark, non-light areas and unstable light areas are 0. Therefore, the regions where the DN values are greater than 0 are stable light areas. In this manuscript, pixels with DN value greater than 0 are assigned as 1, while pixels with DN value equal to 0 are still kept at 0, so as to distinguish the stable light area and non-light area (unstable light area) in nighttime light image. Because the nighttime light images are collected by the same sensor in the different years and different sensors in the same year, we believe that the stable bright pixels obtained by different sensors in the same year should be consistent. The bright pixels in the previous year must be the stable bright pixels in the later year. Finally, the stable nighttime light image mask of 2001-2013 is obtained by the method shown in the lines 312-320. Taking F182013 nighttime lighting image as an example, the number of pixels with DN value equal to 0 is 7682588, and the number of pixels with DN value greater than 0 is 1509110.
The regions where the DN values are equal to 0 are dark and non-light areas.
Points 4: Figure 2 and 3 are useless, as well as equations 1 and 2.
Response 4: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Response to Reviewer 1 Comments
We appreciate valuable suggestions and comments from the anonymous reviewer. We have revised our manuscript according to your helpful suggestions.
Major comments:
Points 1: This study might set the bar too low to present its novelty. The objectives of this study should be better explained. What is the difference with the previous similar studies?
Response 1: Thank you for your suggestions. Aiming at the problems in the current research, this paper proposes a correction method of nighttime light data based on EVI data, so as to obtain the nighttime light dataset from 2001 to 2018 to represent economic data reliably and accurately. By comparing the correction method mentioned in this manuscript with those correction method mentioned in the following literature [1, 2], we illustrate the superiority of the performance of the correction method mentioned in this manuscript. We have modified lines 156-174 to make the purpose of the study clearer.
Points 2: The manuscript adequately describes the equation used. However, the authors do not explain why in the approach. For example, average DN value of the pixel is based on the correction formulas from formula 10 and 11.
Response 2: We have strengthened the explanation of the formula used in the lines 356-361, 367-370, 373-378 and 381-384.
Points 3: The study generates consistent nighttime light data set from 2001 to 2018. The established regression model was used to fit the NPP/VIIRS data from 2014 to 2018 to the DMSP/OLS data scale. Could the authors transform the DMSP/OLS data to NPP/VIIRS data?
Response 3: Thank you for your suggestions. Since DMSP/OLS data is no longer updated and NPP/VIIRS data is still being updated, in theory, DMSP/OLS data should be converted to NPP/VIIRS data. However, due to the short time span of NPP/VIIRS data in this study (only five years), thus considering the accuracy and validity of the fitting results, the NPP/VIIRS data from 2014 to 2018 should be converted to DMSP/OLS data. At present, most scholars [1-3] adopt the method of transforming NPP/VIIRS data into DMSP/OLS data.
Points 4: We can’t see the spatial details. Fig 2 and 3 are too small.
Response 4: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Minor comments:
Points 1: The x-axis in Fig 7 ranges are different with a, c and b, d.
Response 1: The X & Y axis tick label have be improved by using scientific notation.
Response to Reviewer 2 Comments
The authors have discussed about the Nighttime lights correction with EVI data and validated with GDP. However, this paper can be improved in many ways.
Points 1: Writing and grammar can be improved. For example, there are many inconsistency and grammar issues like [Line 310]" Obtaining historical stable images in China, it is necessary to consider the image data collected by sensors in different years and different sensors in the same year. "; [Line 404] "It assumes that..." - are you referring to a specific method?; [Line 466] "sharp increase trend" -"sharp increase" or "increasing trend".
Response 1: Thank you for your suggestion. We have modified the corresponding sentences in the line 309 and 458. The assumption in the line 394 is a precondition of the hypothesis, which is considered to be within the two intervals of 2012-2015 and 2016-2018, there will not be a large number of pixels changing from dim to bright in the two stages.
Points 2: The articles need to be better organized - e.g., you can use another level of heading (instead of number in circles) to show 4.2.1. Inter-calibration steps.
Response 2: We have readjusted the structure of the article, such as merging the second part and the first part of the article into the first part, explaining the methods adopted by researchers in correcting the nighttime light data, and summarizing some common shortcomings of the current correction methods. We have modified the corresponding sentences to remove the circles in the line 312-332.
Points 3: The authors have mentioned the "reference correction method" [Line 475] at the end of the methodology part, can you please explicitly define what exactly is the reference method. It is very hard to follow after you have introduced various data processing steps.
Response 3: The reference correction method mentioned in this paper is the method used in the following literature [1, 2] for DMSP/OLS and NPP/VIIRS data correction. This part has been added in the line 446-453.
Points 4: As for the data validation, why compare NTL with GDP since NTL is not a direct measurement for GDP. GDP can be affected by economic structures at various subnational levels. Basically the graphs and tables you have presented is only helping readers to explore the relationship between lights and total GDP. However, for regions that mainly reply on agriculture production, how do you use nighttime lights to capture this.
Response 4: Thank you for your suggestion. We are very much in favor of your opinion. Agriculture is an important part of social and economic activities, especially in a traditional agricultural country like China. However, most of the agricultural production is located in the darker area of the nighttime light image, so the nighttime light data cannot represent the agricultural part of economic activities [4]. It is necessary to remove the contribution of agricultural sector to economic activities when using nighttime data to predict GDP. The GDP data in the following paragraphs leave out the data of the agriculture sector. We have added the description part in the line 212-217. And the content related to GDP has been recalculated.
Points 5: Why EVI index is introduced instead of other remote sensing data? If an additional data source is added (like population distribution from GHSL https://ghsl.jrc.ec.europa.eu/), will this improve the model performance.
Response 5: Thank you for your suggestion. In this manuscript, EVI data is introduced to realize the saturation correction of DMSP/OLS data based on considering the negative correlation between vegetation and human activities (the intensity of human non-agricultural activities in the central area of the city is high, and the vegetation coverage is generally less; in the rural areas, the intensity of human non-agricultural activities is low, and the vegetation coverage is more) [5, 6]. And EVI data is a continuous dataset in time series, which can correct DMSP/OLS data in different years. The GHSL, a global information baseline describing the spatial evolution of human settlements in the past 40 years, was developed by using a symbolic machine learning model trained by the collected high-resolution samples, multitemporal Landsat imagery in the epochs 1975, 1990, 2000 and 2015 [7]. Although we can interpolate the data between the years to match the period of the nightlight data, it is difficult for us to obtain GHSL data on continuous time series as accurately as EVI data. However, we agree with your opinion very much. At the same time, your view also gives us great inspiration. Applying GHSL data to DMSP/OLS data correction can effectively improve the performance of the model, and can better distinguish and identify built-up areas and non-built areas Therefore, the next stage of work will focus on how to combine DMSP/OLS data with GHSL data to better study how cities expand.
Response to Reviewer 3 Comments
This work aims to accomplish two most critical thresholds in performing long time series analysis using Nighttime Light (NTL), which are
Correction of saturation in DMSP-OLS annual composite
Inter-annual and intra-annual correction of DMSP-OLS annual composite
Bridging DMSP-OLS and VIIRS-DNB nighttime light imagery
The authors then verified the resulted NTL series by comparing GDP and EPC of selected provinces in China, and confirmed to have very high agreement.
The manuscript is very well laid out with some room for improvement as listed below.
Points 1: Line 80: EVI should be described more in Introduction section.
Response 1: Thank you for your valuable suggestions. We have added some description on the EVI data in the line 104-108.
Points 2: Line 161: DMSP/OLS Data -> DMSP/OLS data.
Response 2: Thank you for your patience in reading this manuscript. We have read through the full manuscript, and have changed all the contents related to DMSP/OLS Data to DMSP/OLS data.
Points 3: L194, L204, L215: The group which makes these products (EOG, Earth Observation Group) is no longer at NOAA/NCEI and now affiliated to the Paynes Institute for Public Policy, Colorado School of Mines.
Response 3: We have changed the corresponding institutions and websites to the current ones.
Points 4: Figure 1: NPP/VIIRS Stable Data for China should not need inter-annual correction?
Response 4: Thank you for your suggestions. Although the NPP/VIIRS data is collected by the same sensor, it can be seen from Figure 4-a that the NPP/VIIRS data before uncorrected are not all increasing. The TNL data from 2015 to 2016 and from 2017 to 2018 show a slight downward trend. Therefore, we believe that in order to make the results more appropriate to China's actual urbanization process, the nighttime light data more accurately predicting GDP and EPC, the NPP/VIIRS data should be carried out with inter-annual correction.
Points 5: Figure 2 & 3: These two figures does not show the difference clearly. Suggest enlarge some smaller regions to show the effect of correction. NPP/VIIRS should also be included in the caption of Figure 3.
Response 5: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Points 6: Line 314 to 324: The english writing in this paragraph is unclear. Please improve.
Response 6: We have modified the corresponding sentences in the line 312-320.
Points 7: L372: The reference [48] only exams data up to year 2008, it is not very convincing to extrapolate the conclusion all the way to 2018. Authors should make clear about the assumption they are making here.
Response 7: Thank you for your suggestion. The author has noticed this problem and updated the latest references [8, 9]. It can be seen from reference [9] that China's urbanization rate showed an upward trend from 2004 to 2017. From reference [8], we can see that China's urbanization increase rate is 1.36% from 2000 to 2010 and 1.1% from 2010 to 2018. Therefore, according to these two references, the author can assume that there is no reverse urbanization process in China from 2001 to 2018.
Points 8: L402-404: Is the research period up to year 2018 or 2019?
Response 8: Thank you for your suggestions. The research period is 2002-2018, which has been revised in the line 392-393.
Points 9: L403-410: The english writing in this paragraph is unclear, please improve.
Response 9: Thank you for your suggestions. We have modified the corresponding sentences in the line 393-399.
Points 10: L515-516, Figure 6: What is the reference correction used here? Please make it clear.
Response 10: The reference correction method mentioned in this paper is the method used in the following literature [1, 2] for DMSP/OLS and NPP/VIIRS data correction. This part has been added in line 446-453.
Points 11: Figure 7: The X & Y axis tick label should be improved by using scientific notation or difference unit.
Response 11: Thank you for your suggestions. The X & Y axis tick label have be improved by using scientific notation.
Points 12: Table 1 & 2: What is the data source of GDP and EPC?
Response 12: Gross domestic product (GDP) and electric power consumption (EPC) from 2002 to 2019 year are provided by the China statistical yearbook which published by China National Bureau of statistics (http://www.stats.gov.cn/tjsj/ndsj/).
Points 13: Table 5 & 6: What is the unit of these numbers?
Response 13: It shows the mean absolute relative error between the real GDP and the GDP predicted by the models based on the GDP and corrected nighttime light data at provincial scale under the two correction methods in Table 5. At the same time, it shows the mean absolute relative error between the real EPC and the EPC predicted by the models based on the EPC and corrected nighttime light data at provincial scale under the two correction methods in Table 6. Therefore, these numbers are the mean absolute relative errors without units.
Points 14: Table 7: Why is the numbers shown in this table having fixed numbers after decimal point of .33 and .67?
Response 14: The numbers in table 7 are obtained through the classification of results in tables 5 and 6. The mean absolute relative errors of 30 provinces are divided into three categories [10]: 0-25% is high accuracy, 25%-50% is medium accuracy, and greater than 50% is inaccuracy. The reason of the numbers shown in Table 7 having fixed numbers after the decimal point of. 33 and. 67 is that the research object of this paper is 30 provinces in China, and the fixed decimal point (.00, .33, and .67) is obtained by dividing the total of the provinces after statistics by 30 provinces.
Response to Reviewer 4 Comments
The manuscript entitled "Nighttime light data correction based on EVI data by invariant region method" aimed to address the overglow and saturation effects of the DMSP/OLS data by using the EVI product.
I believe the manuscript is suitable for publication in Remote Sensing, after the authors have addressed the comments below.
Points 1: Page 1, line 19. You mentioned that (… on national and provincial scales, respectively). Where? You have to mention the country.
Response 1: Thank you for your suggestions. We have defined the research area of this paper to 30 provinces of China in the line 16. The national scale mentioned here is to take China as a whole research area, so as to analyze the fluctuation of TNL over time and the ability to predict GDP and EPC. The provincial scale refers to taking each province in China as an independent research area to analyze the ability of predicting GDP and EPC in each province.
Points 2: Page 2, line 46-50. There are other nighttime sensors. Therefore, when you mentioned only the DMSP/OLS and NPP/VIIRS, you have to say something like “the most widely used” or “the popular nighttime sensors”.
Response 2: Thank you for your suggestions. We have modified the corresponding sentences in the line 49.
Points 3: Page 2, line 77. Change the sentence to be (… is more serious than that between the difference products of the DMSP/OLS).
Response 3: Thank you for your suggestions. We have modified the corresponding sentences in the line 145-147.
Points 4: Page 4, line 185. A site location map is more important than Figures 2 and 3.
Response 4: Thank you for your suggestions. We have enlarged Figure 2 and 3, and selected two sampling areas (sampling area 1: Hohhot City, Heilongjiang Province, and sampling area 2: Yiwei City, Yunnan Province) with the red box to compare the results. By comparing the image of sampling area 1 without correction and which after correction, it can be found that the correction method can effectively remove the abnormal highlight points, and by comparing the image of sampling area 2 without correction and which after correction, it can be found that the correction method can remove the negative pixels in the dark area.
Points 5: Page 8, line 321. You mentioned that (… the stable nighttime light image mask of 2018 and the nighttime light image ….). I believe you mean light image mask of 2008 not 2018. Check.
Response 5: Thank you for your patience in reading this manuscript, the year of the stable nighttime light image mask is actually 2008. We have modified the corresponding sentences in the line 318.
Points 6: Page 9, Equations (4) and (5). These equations are similar. Thus, just keep one of them.
Response 6: Thank you for your valuable advice. Formula (5) is a general formula for the normalization of DN value, while formula (6) is a special case where the value of is taken as 0. The retention of formula (6) can simplify the writing of subsequent formulas. Therefore, experts are requested to agree to retain formula (5) and formula (6).
Points 7: Page 10, line 362. Put a reference after “Intra-annual composition”.
Response 7: Thank you for your suggestions. The corresponding reference has been added in the line 355.
Points 8: Page 10, line 374. Put a reference after “inter-annual series correction”.
Response 8: Thank you for your suggestions. The corresponding reference has been added in the line 366.
Points 9: Page 10, Equation (10). Would there be a difference if you used the inter-annual series correction that proposed by Liu et al. (2012) [Ref. 48].
Response 9: Using the inter-annual correction method proposed by Liu et al. it can ensure that the DN value of the pixel in the next year's image is not less than that of the previous year's image. This adjustment order can avoid weakening the DN value of the pixel in the early annual image, but will cause the DN value of the pixel in the later year's image to be excessively enlarged. Therefore, the correction in the opposite direction should be carried out on the basis of the inter-annual correction method proposed by Liu et al. to ensure that the DN value of pixel in the image of the previous year is not greater than that of the pixel in the image of the next year. Then taking the average value after two inter-annual correction as the final DN value, we can obtain a more realistic nighttime light image.
Points 10: Page 11, lines 402-403. Are you sure it is 2019 not 2018.
Response 10: The research period is from 2002 to 2018. We have modified the corresponding sentences in the line 392-393.
Points 11: Page 11, lines 405-410. Why did not you choose the 2018 annual composite data as a binary-value mask instead of 2016?
Response 11: At present, the monthly composites published by the Paynes Institute for Public Policy, Colorado School of Mines are all monthly products, has not been filtered to screen out lights from aurora, fires, boats, and other temporal lights. Therefore, the nighttime light images we synthesized for 2018 also include these short-term light data. The annual composite data published only contain the 2015 annual composite data and 2016 annual composite data which has been filtered to screen out short-term lights. Hence, we choose the 2016 annual composite data as a binary-value mask instead of 2018.
Points 12: Page 13, lines 467-468. Re-write the sentence.
Response 12: Thank you for your suggestions. We have modified the corresponding sentences in the line 459-461.
Points 13: Page 13, lines 474-475. Re-write the sentence.
Response 13: Thank you for your suggestions. We have modified the corresponding sentences in the line 466-467.
Points 14: Page 13, line 475. What do you mean by the reference correction method? You have to define it.
Response 14: The reference correction method mentioned in this paper is the method used in the following literature [1, 2] for DMSP/OLS and NPP/VIIRS data correction. This part has been added in line 446-453.
Points 15: Page 16, lines 540-542. Re-write the sentence.
Response 15: Thank you for your suggestions. We have modified the corresponding sentences in the line 535-538.
Points 16: Page 16, lines 548-554. From which tables did you get the results of the MARE before and after 2013? If there is no table, you have to create one that summarises the results.
Response 16: Thank you for your suggestions. We added lines 21-22 in Table 1 and lines 21-22 lines in Table 2 to obtain the corresponding conclusion data in the paper.
Points 17: Page, Figure 7. Change the x-axis unit to be in million and ranges from 0 to 30.
Response 17: Thank you for your suggestions. The X & Y axis tick label have be improved by using scientific notation.
Points 18: Page 17, lines 572-578. Same as Comment No. (16).
Response 18: Thank you for your suggestions. We added lines 21-22 in Table 1 and lines 21-22 lines in Table 2 to obtain the corresponding conclusion data in the paper.
Points 19: Page 17, lines 579-580. You have to support your discussion with the results. Thus, the MARE results (12.3 and 7.22) must be reported here.
Response 19: In the lines 561-562, we have added the MARE of the adjusted nighttime lighting data to forecast the GDP and EPC, so as to better support our discussion.
Points 20: Page 19, line 603. Check the results (0.83 and 0.845).
Response 20: The 0.859 in the line 674 is the R2 value of the model between the TNL of the 2015 NPP/VIIRS annual composite data published by the Paynes Institute for Public Policy, Colorado School of Mines and GDP. The 0.827 in the line 674 is the R2 value of the model between the TNL of the 2015 NPP/VIIRS annual composite data published by the Paynes Institute for Public Policy, Colorado School of Mines and EPC. These two R2 values are used to show that there are still short-term lights existing in the annual NPP/VIIRS images which only the background noise removed.
The reason of the numbers (0.83 and 0.845) changing: Agriculture is an important part of social and economic activities, especially in a traditional agricultural country like China. However, most of the agricultural production is located in the darker area of the nighttime light image, so the nighttime light data cannot represent the agricultural part of economic activities [4]. It is necessary to remove the contribution of agricultural sector to economic activities when using nighttime data to predict GDP. The GDP data in the following paragraphs leave out the data of the agriculture sector. We have added the description part in the line 212-217. And the content related to GDP has been recalculated.
Points 21: Page 19, line 614. Check the results (0.866).
Response 21: The 0.863 in the line 584 is the R2 values of the model between TNL of the F162010 radiometric calibration nighttime light image and GDP. The R2 values of the model between TNL of nighttime light image of 2010 under the correction method in this paper and GDP was 0.817, the R2 values of the model between TNL of nighttime light image of 2010 under the reference correction method and GDP was 0.773, so it can be found that the correction method in this paper plays a greater role in reducing the "saturation effect" of the image than the reference correction method. We have added the explanation in the lines 583-587.
The reason of the number (0.866) changing: Agriculture is an important part of social and economic activities, especially in a traditional agricultural country like China. However, most of the agricultural production is located in the darker area of the nighttime light image, so the nighttime light data cannot represent the agricultural part of economic activities [4]. It is necessary to remove the contribution of agricultural sector to economic activities when using nighttime data to predict GDP. The GDP data in the following paragraphs leave out the data of the agriculture sector. We have added the description part in the line 212-217. And the content related to GDP has been recalculated.
Points 22: Pages 19-20, lines 619-631. When you use a term “better than”, you have to write the result (value) in parentheses. For example:
By comparing the results of 2015, it can be seen that the R2 value (0.764) of the model between TNL after correction …..
Response 22: Thank you for your suggestion. We have reviewed the manuscript again and change the sentences in the lines 591-596.
Points 23: Page 24, line 724. The DMSP/OLS must be written as a one word, there must be no space between the slash and letters. Please check the whole manuscript.
Response 23: Thank you for your reminder. We have reviewed the manuscript and have completed the modification of the corresponding content.
We appreciate the reviewer’s suggestions earnestly and hope that the revision answers the issues raised by the reviewer.
Once again, thank you very much for your comments and suggestions.
References:
- Lv, Q.; Liu, H.; Wang, J.; et al. Multiscale analysis on spatiotemporal dynamics of energy consumption CO2 emissions in China: Utilizing the integrated of DMSP-OLS and NPP-VIIRS nighttime light datasets. Sci. Total Environ. 2020, 703, 134394.
- Li, X.; Gong, L. Correction and fitting of night light images of DMSP / OLS and VIIRS /DNB
. Bulletin of Surveying and Mapping. 2019, 138-146.
- Zhu, X.; Ma, M.; Yang, H.; et al. Modeling the Spatiotemporal Dynamics of Gross Domestic Product in China Using Extended Temporal Coverage Nighttime Light Data. Remote Sens.-Basel. 2017, 9, 626-644.
- Tilottama, G.; Powell, R.; Elvidge, C.; et al. Shedding Light on the Global Distribution of Economic Activity. The Open Geography Journal. 2010, 3, 147-160.
- Pozzi, F.; Small, C. Analysis of Urban Land Cover and Population Density in the United States. Photogramm. Eng. Rem. S. 2005, 6, 719-726.
- Weng, Q.H.; Lu, D.S.; Schubring, J. Estimation of land surface temperature-vegetation abundance relationship for urban heat island studies. Remote Sens. Environ. 2004, 89, 467-483.
- Florczyk, A.J.; Corbane, C.; Ehrlich, D.; Freire, S.; Kemper, T.; Maffenini, L.; Melchiorri, M.; Pesaresi, M.; Politis, P.; Schiavina, M.; et al. GHSL Data Package 2019; EUR 29788 EN; Publications Office of the European Union: Luxembourg, 2019; ISBN 978-92-76-08725-0. Available online: https://ghsl.jrc.ec.europa.eu/documents/GHSL_Data_Package_2019.pdf
- Hao, Y.; Zheng, S.; Zhao, M.; et al. Reexamining the relationships among urbanization, industrial structure, and environmental pollution in China—New evidence using the dynamic threshold panel model. Energy Reports. 2020, 6, 28-39.
- Li, T.; Wu, Z. Urbanization and Education Development in China. Research in Educational Development. 2019, 39, 1-10.
- Shi, K.; Yu, B.; Huang, Y.; et al. Evaluating the Ability of NPP-VIIRS Nighttime Light Data to Estimate the Gross Domestic Product and the Electric Power Consumption of China at Multiple Scales: A Comparison with DMSP-OLS Data. Remote Sens.-Basel. 2014, 6, 1705-1724.
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
The authors improved the manuscript and answered to the comments given. Nevertheless, this paper proposes a correction method of nighttime light data based on EVI data for VIIRS/DNB NTL data. The novelty is still not well explained. The Introduction section is correct but must be revised because the lack of connection among some sentences and paragraphs in some parts. There are many more advanced methods. It shows that this study might set the bar too low to present its novelty.
This manuscript might have some potential to provide more robust methodology discussion and scientific merits to benefit readers. I recommend the authors to reconsider this approach, revise the structure of this study and resubmit.
Reviewer 3 Report
The authors meticulously revised the manuscript and answered to the comments given. Nevertheless, the answer to my comment [4] which doubts the necessity to perform inter-annual correction for VIIRS/DNB NTL data does not make scientific sense.
Despite the VIIRS/DNB sensor being calibrated at each scan, with constant quality monitoring by ground staff, the authors decided to alter the values to match the trend of China's GDP and EPC. Such conduct without scientific sound reason fundamentally weaken the robustness of this study. Unless the authors has a scientifically sound reason to do so, I do not agree with such decision.
I recommend the authors to reconsider their approach, revise the structure of this study and resubmit.
