Suitability of NASA’s Black Marble Daily Nighttime Lights for Population Studies at Varying Spatial and Temporal Scales

Round 1

Reviewer 1 Report

Dear Author,

thanks for the work, it is very informative and honest.

I miss a discussion about why might not be such a good correlation between the population changes and some of the light intensity.

For example, this lack of correlation has also been found with other variables recently, like in the case of light pollution.

Kyba, C. C., Altıntaş, Y. Ö., Walker, C. E., & Newhouse, M. (2023). Citizen scientists report global rapid reductions in the visibility of stars from 2011 to 2022. Science, 379(6629), 265-268.

It can be explained by the lack of sensitivity to blue light and infrared sensitivity of VIIRS, plus its panchromatic nature. It is likely that in many of the cases studied, after the electric renovations or the earthquake the new lights where LEDs, which will produce a reduction in the VIIRS signal without any reduction in human eyes(Sánchez de Miguel et. al. 2019). In the future, SDGSAT-1, JL1, or ISS data can complement VIIRS data. Also, might be that late-night data is not so correlated with the variables considered. See,  Bustamante-Calabria et. al. 2021 and Roman et. al. 2018.

Sánchez de Miguel, A. , Kyba, C. C., Aubé, M., Zamorano, J., Cardiel, N., Tapia, C., ... & Gaston, K. J. (2019). Colour remote sensing of the impact of artificial light at night (I): The potential of the International Space Station and other DSLR-based platforms. Remote sensing of environment, 224, 92-103.

 Bustamante-Calabria, M., Sánchez de Miguel, A., Martín-Ruiz, S., Ortiz, J. L., Vílchez, J. M., Pelegrina, A., ... & Gaston, K. J. (2021). Effects of the COVID-19 lockdown on urban light emissions: ground and satellite comparison. Remote Sensing, 13(2), 258.

Román, M. O., Wang, Z., Sun, Q., Kalb, V., Miller, S. D., Molthan, A., ... & Masuoka, E. J. (2018). NASA's Black Marble nighttime lights product suite. Remote Sensing of Environment, 210, 113-143.

Another topic is the use of Sum of Lights. The use of that variable is conceptually incorrect. Regardless of how popular it is. It does not have a physical meaning.  It comes from the use of the DMSP, an uncalibrated satellite with a long list of calibration issues. VIIRS does not have these issues. When the light on a certain area is aggregated, the right way to present it is on W/sr, W (if a power angular distribution of radiance is assumed) or provide the average radiance of the area. 

You can read more about this on Wikipedia: https://en.wikipedia.org/wiki/Radiance

For an example I can give you a very popular article in this journal:

Sánchez de Miguel, A., Bennie, J., Rosenfeld, E., Dzurjak, S., & Gaston, K. J. (2021). First estimation of global trends in nocturnal power emissions reveals acceleration of light pollution. Remote Sensing, 13(16), 3311

 

Author Response

1

I miss a discussion about why might not be such a good correlation between the population changes and some of the light intensity. For example, this lack of correlation has also been found with other variables recently, like in the case of light pollution. Kyba, C. C., Altıntaş, Y. Ö., Walker, C. E., & Newhouse, M. (2023). Citizen scientists report global rapid reductions in the visibility of stars from 2011 to 2022. Science, 379(6629), 265-268. It can be explained by the lack of sensitivity to blue light and infrared sensitivity of VIIRS, plus its panchromatic nature. It is likely that in many of the cases studied, after the electric renovations or the earthquake the new lights where LEDs, which will produce a reduction in the VIIRS signal without any reduction in human eyes(Sánchez de Miguel et. al. 2019). In the future, SDGSAT-1, JL1, or ISS data can complement VIIRS data. Also, might be that late-night data is not so correlated with the variables considered. See,  Bustamante-Calabria et. al. 2021 and Roman et. al. 2018. Sánchez de Miguel, A. , Kyba, C. C., Aubé, M., Zamorano, J., Cardiel, N., Tapia, C., ... & Gaston, K. J. (2019). Colour remote sensing of the impact of artificial light at night (I): The potential of the International Space Station and other DSLR-based platforms. Remote sensing of environment, 224, 92-103. Bustamante-Calabria, M., Sánchez de Miguel, A., Martín-Ruiz, S., Ortiz, J. L., Vílchez, J. M., Pelegrina, A., ... & Gaston, K. J. (2021). Effects of the COVID-19 lockdown on urban light emissions: ground and satellite comparison. Remote Sensing, 13(2), 258. Román, M. O., Wang, Z., Sun, Q., Kalb, V., Miller, S. D., Molthan, A., ... & Masuoka, E. J. (2018). NASA's Black Marble nighttime lights product suite. Remote Sensing of Environment, 210, 113-143.

Added citations mentioned.

Added the following paragraph: 

“Furthermore, change in brightness over time may be attributed to limitations of the SNPP-VIIRS/DNB sensors or errors stemming from seasonal variations and environmental factors [80], [81]. The NPP satellite's DNB sensors, which supply data for the VIIRS Black Marble, have a limited range of EM wavelengths between 0.5 µm to 0.9 µm [82]. This could create bias favorable to incandescent light bulbs, which operate within the sensor’s range, over light-emitting diodes (LEDs), which tend to emit light below 0.5 µm. The latter can lead to increased skyglow and reduced upward radiance recorded by satellites  [83], [84].”

 

Another topic is the use of Sum of Lights. The use of that variable is conceptually incorrect. Regardless of how popular it is. It does not have a physical meaning.  It comes from the use of the DMSP, an uncalibrated satellite with a long list of calibration issues. VIIRS does not have these issues. When the light on a certain area is aggregated, the right way to present it is on W/sr, W (if a power angular distribution of radiance is assumed) or provide the average radiance of the area. You can read more about this on Wikipedia: https://en.wikipedia.org/wiki/Radiance For an example I can give you a very popular article in this journal: Sánchez de Miguel, A., Bennie, J., Rosenfeld, E., Dzurjak, S., & Gaston, K. J. (2021). First estimation of global trends in nocturnal power emissions reveals acceleration of light pollution. Remote Sensing, 13(16), 3311 https://www.mdpi.com/2072-4292/13/16/3311 

Added Appendix B with Graph:

A major concern of this study is its use of the Sum of Light (SoL) aggregation method because summing radiance values can include background noise that influences the results. This can reduce the signal-to-noise ratio (SNR), thereby reducing the quality of the data. The product team uses algorithms to clean the VIIRS DNB data to reduce the influence of background noise [85]; yet, several studies have used exclusionary methods such as setting bottom thresholds or excluding areas that are known to produce background noise—usually open spaces or non-built-up areas—would improve the SNR [86]–[89]; however, this removal process may inadvertently exclude dimly lit areas that may be of interest [90]. We performed a sensitivity analysis to explore whether our results are affected by aberrations that may be introduced by accumulation of noise through SoL. To better understand how these excluded areas influenced our results, we used the Global Human Settlement Layer (GHSL) GHS Settlement Characteristics, derived from Sentinel2 composite (2018) and other GHS R2023A data (GHS-BUILT-C R2023A) [91] downloaded from the Joint Research Centre (https://ghsl.jrc.ec.europa.eu/download.php?ds=builtC). We tested two GHS-BUILT-C layers: the Morphological Settlement Zone (MSZ) and the Functional classification (RES vs. NRES) (FUN) to distinguish the built-up and non-built-up areas. 

Using a python script, we resampled both GHS-BUILT-C layers separately to match the alignment, resolution and extent of the NTL rasters. Using the monthly NTL averages for Qatar, the SoL were calculated for each respective classified zone in the GHSL MSZ and FUN layers. For MSZ, the zones 11 through 25, which represent built-up spaces with buildings, were summed for each month from January 2012 to May 2018. Similarly for FUN, the zones 1 and 2, which represent built-up residential and non-residential zones, respectively, were summed for each month. Both results were transformed using Log10 and plotted along with the Log10 of the Total SoL values from our original analysis, which summed up all of the lights within Qatar without regard to classification (Figure 11). Masking non-built-up areas reduced the absolute magnitude of luminosity; however, the monthly variations remained unchanged. The relative change in luminosity was consistent whether non-built-up pixels were included in the analysis or not. 

 

Reviewer 2 Report

The manuscript entitled "Suitability of NASA’s Black Marble Nighttime Lights for Pop-2 ulation Studies at Varying Spatial and Temporal Scales" address important scientific problem of population estimation by using geospatial datasets. Overall, the manuscript is structured well and can be accepted.

The introduction section can be improved/restructured as the first paragraph portrays the results and conclusion. 

The four case studies utilizing varying spatiotemporal datasets for demographic research, disaster mitigation and adaptation planning, and infrastructure development make this work quite comprehensive. However, the usage of different NTL products does not match the title of the manuscript, as the manuscript title only mentions "NASA’s Black Marble Nighttime Lights". 

The "Materials" and "Methods and Results by Case Study" sections seem confusing. The "materials" section does not mention "NPP/VIIRS Moonlight-adjusted Nighttime Lights" nor in table 1 however they are mentioned in line 458. 

The study results are unsurprising as NTL is not the best tool for population count in rural areas, as proven by many studies. However, the detailed analysis and validations make this study credible. 

 

 

 

Author Response

The manuscript entitled "Suitability of NASA’s Black Marble Nighttime Lights for Pop-2 pulation Studies at Varying Spatial and Temporal Scales" address important scientific problem of population estimation by using geospatial datasets. Overall, the manuscript is structured well and can be accepted. The introduction section can be improved/restructured as the first paragraph portrays the results and conclusion. 

Restructured the introduction for more clarity

 

The four case studies utilizing varying spatiotemporal datasets for demographic research, disaster mitigation and adaptation planning, and infrastructure development make this work quite comprehensive. However, the usage of different NTL products does not match the title of the manuscript, as the manuscript title only mentions "NASA’s Black Marble Nighttime Lights". We only used NASA VIIRS that we aggregated at different scales, this was made more clear in the revision of the paper.

 

The "Materials" and "Methods and Results by Case Study" sections seem confusing. The "materials" section does not mention "NPP/VIIRS Moonlight-adjusted Nighttime Lights" nor in table 1 however they are mentioned in line 458. Updated intro to include this name more accurately.

 

The study results are unsurprising as NTL is not the best tool for population count in rural areas, as proven by many studies. However, the detailed analysis and validations make this study credible. 

Added appendix B to include research on the question of Sum of Light method and masking built-up areas:

Appendix B
A major concern of this study is its use of the Sum of Light (SoL) aggregation method because summing radiance values can include background noise that influences the results. This can reduce the signal-to-noise ratio (SNR), thereby reducing the quality of the data. The product team uses algorithms to clean the VIIRS DNB data to reduce the influence of background noise [85]; yet, several studies have used exclusionary methods such as setting bottom thresholds or excluding areas that are known to produce background noise—usually open spaces or non-built-up areas—would improve the SNR [86]–[89]; however, this removal process may inadvertently exclude dimly lit areas that may be of interest [90]. We performed a sensitivity analysis to explore whether our results are affected by aberrations that may be introduced by accumulation of noise through SoL. To better understand how these excluded areas influenced our results, we used the Global Human Settlement Layer (GHSL) GHS Settlement Characteristics, derived from Sentinel2 composite (2018) and other GHS R2023A data (GHS-BUILT-C R2023A) [91] downloaded from the Joint Research Centre (https://ghsl.jrc.ec.europa.eu/download.php?ds=builtC). We tested two GHS-BUILT-C layers: the Morphological Settlement Zone (MSZ) and the Functional classification (RES vs. NRES) (FUN) to distinguish the built-up and non-built-up areas. 
Using a python script, we resampled both GHS-BUILT-C layers separately to match the alignment, resolution and extent of the NTL rasters. Using the monthly NTL averages for Qatar, the SoL were calculated for each respective classified zone in the GHSL MSZ and FUN layers. For MSZ, the zones 11 through 25, which represent built-up spaces with buildings, were summed for each month from January 2012 to May 2018. Similarly for FUN, the zones 1 and 2, which represent built-up residential and non-residential zones, respectively, were summed for each month. Both results were transformed using Log10 and plotted along with the Log10 of the Total SoL values from our original analysis, which summed up all of the lights within Qatar without regard to classification (Figure 11). Masking non-built-up areas reduced the absolute magnitude of luminosity; however, the monthly variations remained unchanged. The relative change in luminosity was consistent whether non-built-up pixels were included in the analysis or not. 

 

Reviewer 3 Report

This paper investigates the potential link between changes in NTL value and human dynamics, particularly population counts. The analysis results have good reference value for related research. I feel that the method can be improved to make the results more sound. I suggest a “Major Revision”. Please see my comments below.

1. Based on my own experience, the brightness values of NTL on different dates may have an overall deviation, which means that all values in one image may be higher or lower than those in the same place on another image. If SoL is used over a large area of non-urban pixels, the background NTL noise will accumulate and may lead to overall deviation. I suggest removing non-urban areas or areas with very low NTL value to reduce the impact of noise.

2. Some abbreviations do not have full names when they first appear. For example, Line 51 UN, Line 81 VIIRS.

3. Conclusion needs to be added to make the structure of the article complete.

Author Response

This paper investigates the potential link between changes in NTL value and human dynamics, particularly population counts. The analysis results have good reference value for related research. I feel that the method can be improved to make the results more sound. I suggest a “Major Revision”. Please see my comments below. 

 

1. Based on my own experience, the brightness values of NTL on different dates may have an overall deviation, which means that all values in one image may be higher or lower than those in the same place on another image. If SoL is used over a large area of non-urban pixels, the background NTL noise will accumulate and may lead to overall deviation. I suggest removing non-urban areas or areas with very low NTL value to reduce the impact of noise. 

We added Appendix B to test the SoL method. We address this issue by averaging over several days and focusing on the light’s change over time for the same areas rather than comparing absolute values of different areas of different sizes. Furthermore, we did not want to assume the areas that people would be moving into or the areas that would be developed as seen in the Oaxaca Migration and the Qatar cases:

Appendix B

A major concern of this study is its use of the Sum of Light (SoL) aggregation method because summing radiance values can include background noise that influences the results. This can reduce the signal-to-noise ratio (SNR), thereby reducing the quality of the data. The product team uses algorithms to clean the VIIRS DNB data to reduce the influence of background noise [85]; yet, several studies have used exclusionary methods such as setting bottom thresholds or excluding areas that are known to produce background noise—usually open spaces or non-built-up areas—would improve the SNR [86]–[89]; however, this removal process may inadvertently exclude dimly lit areas that may be of interest [90]. We performed a sensitivity analysis to explore whether our results are affected by aberrations that may be introduced by accumulation of noise through SoL. To better understand how these excluded areas influenced our results, we used the Global Human Settlement Layer (GHSL) GHS Settlement Characteristics, derived from Sentinel2 composite (2018) and other GHS R2023A data (GHS-BUILT-C R2023A) [91] downloaded from the Joint Research Centre (https://ghsl.jrc.ec.europa.eu/download.php?ds=builtC). We tested two GHS-BUILT-C layers: the Morphological Settlement Zone (MSZ) and the Functional classification (RES vs. NRES) (FUN) to distinguish the built-up and non-built-up areas. 

Using a python script, we resampled both GHS-BUILT-C layers separately to match the alignment, resolution and extent of the NTL rasters. Using the monthly NTL averages for Qatar, the SoL were calculated for each respective classified zone in the GHSL MSZ and FUN layers. For MSZ, the zones 11 through 25, which represent built-up spaces with buildings, were summed for each month from January 2012 to May 2018. Similarly for FUN, the zones 1 and 2, which represent built-up residential and non-residential zones, respectively, were summed for each month. Both results were transformed using Log10 and plotted along with the Log10 of the Total SoL values from our original analysis, which summed up all of the lights within Qatar without regard to classification (Figure 11). Masking non-built-up areas reduced the absolute magnitude of luminosity; however, the monthly variations remained unchanged. The relative change in luminosity was consistent whether non-built-up pixels were included in the analysis or not. 

2. Some abbreviations do not have full names when they first appear. For example, Line 51 UN, Line 81 VIIRS. Corrected and reviewed all abbreviations.

 

3. Conclusion needs to be added to make the structure of the article complete.

 

Added more to the conclusion: 

Furthermore, change in brightness over time may be attributed to limitations of the SNPP-VIIRS/DNB sensors or errors stemming from seasonal variations and environmental factors [80], [81]. The NPP satellite's DNB sensors, which supply data for the VIIRS Black Marble, have a limited range of EM wavelengths between 0.5 µm to 0.9 µm [82]. This could create bias favorable to incandescent light bulbs, which operate within the sensor’s range, over light-emitting diodes (LEDs), which tend to emit light below 0.5 µm. The latter can lead to increased skyglow and reduced upward radiance recorded by satellites  [83], [84]. 

In the Za’atari Refugee Camp, NTL was able to capture the creation of the camp and provided information on the stability of nighttime electricity, which is linked to the wellbeing of camp residents of all ages, but it did not correlate to the changes in Pop in that camp. This could be because refugee camp infrastructure is often constructed up-front. Further research is needed to clarify whether NTL can be a proxy for Pop in refugee camps.

We sought to determine the practicality and limitations of NASA’s Black Marble for population studies at different spatial and temporal scales. We surmise our findings as such: NASA’s Black Marble is an earth-observing tool with an exceptional ability at detecting a variety of anthropogenic activity even at low levels of light in part due to its ability to capture data with near-global coverage at a daily frequency with a high radiometric range. We demonstrated Black Marble’s can be applied to a wide range of applications in disaster mitigation, SDG monitoring, settlement detection, and economic development. The NTL product captured seasonal changes in light as well as larger trends of infrastructure and settlement development; however, NTL was limited at correlating to changes in populations in certain cases. In the Qatar case study, the monthly population estimates correlated strongly to SoL NTL values but a pixel analysis revealed that the dense urban areas are not increasing in luminosity, rather the increases in NTL were attributed to economic development and diversification in areas industrial areas where population density is low. The signal of NTL is not a direct signature of population, but the qualities of the data provide invaluable context on the light conditions on the ground. Despite the challenges that remote sensing data has, both physical, and in calibration and validation, NTL demonstrates that under the proper context, it can provide detailed information that could help to answer a multitude of research questions on the conditions of people.

Round 2

Reviewer 3 Report

This manuscript has been revised and I appreciate the responses to my provious comments. I think it can be published in present form.