Figure 1.
Electric power generation for the nine countries with test sites in this study from 2010 to 2018. Data from the International Energy Agency (IEA) [
6].
Figure 1.
Electric power generation for the nine countries with test sites in this study from 2010 to 2018. Data from the International Energy Agency (IEA) [
6].
Figure 2.
Jeweled carpet. Color composite of the monthly Visible Infrared Imaging Radiometer Suite (VIIRS) day/night band (DNB) cloud-free average radiance images of a portion of the Gangetic Plain, India. The color hues arise from the differing level of radiances from the three months of data. January 2015 as blue, June 2015 as green, and November 2015 as red.
Figure 2.
Jeweled carpet. Color composite of the monthly Visible Infrared Imaging Radiometer Suite (VIIRS) day/night band (DNB) cloud-free average radiance images of a portion of the Gangetic Plain, India. The color hues arise from the differing level of radiances from the three months of data. January 2015 as blue, June 2015 as green, and November 2015 as red.
Figure 3.
Grid points for the DNB temporal profiles for Nairobi, Kenya. It is an east–west north–south grid with center points 15 arc seconds apart. The Nairobit grid has 100 points east–west and 70 points north–south.
Figure 3.
Grid points for the DNB temporal profiles for Nairobi, Kenya. It is an east–west north–south grid with center points 15 arc seconds apart. The Nairobit grid has 100 points east–west and 70 points north–south.
Figure 4.
The effects of filtering to remove cloudy and moonlit data from a VIIRS DNB profile from Shanghai, China.
Figure 4.
The effects of filtering to remove cloudy and moonlit data from a VIIRS DNB profile from Shanghai, China.
Figure 5.
The source used for the dark offset in the DNB radiance calculation changed from an earth view to a space view on 12 January 2017, resulting in a slight upward shift in the radiance. This can be seen visually in the DNB temporal profiles that lack surface lighting or have extremely dim surface lighting. We apply a radiance adjustment of +0.125 nW to the pixels acquired prior the 12 January 2017.
Figure 5.
The source used for the dark offset in the DNB radiance calculation changed from an earth view to a space view on 12 January 2017, resulting in a slight upward shift in the radiance. This can be seen visually in the DNB temporal profiles that lack surface lighting or have extremely dim surface lighting. We apply a radiance adjustment of +0.125 nW to the pixels acquired prior the 12 January 2017.
Figure 6.
This DNB temporal profile was divided into three segments based on the sharp radiance changes to either side of the prolonged outage in Segment 2.
Figure 6.
This DNB temporal profile was divided into three segments based on the sharp radiance changes to either side of the prolonged outage in Segment 2.
Figure 7.
DNB radiance versus the satellite zenith (satz) angle profiles fall into four primary types: flat, convex, concave, and peak at nadir. For any location on the ground, there are a limited set of possible satz positions due to orbit precession. The red lines mark the median positions for each satz column. The orange horizontal line marks the inter quartile range (IQR).
Figure 7.
DNB radiance versus the satellite zenith (satz) angle profiles fall into four primary types: flat, convex, concave, and peak at nadir. For any location on the ground, there are a limited set of possible satz positions due to orbit precession. The red lines mark the median positions for each satz column. The orange horizontal line marks the inter quartile range (IQR).
Figure 8.
Nadir normalized DNB radiance versus the satellite zenith (satz) angle for the temporal profiles shown in
Figure 7. The medians (red lines) are now aligned within the satz column. The IQRs (green horizontal lines) declined, except in the case of peak at nadir.
Figure 8.
Nadir normalized DNB radiance versus the satellite zenith (satz) angle for the temporal profiles shown in
Figure 7. The medians (red lines) are now aligned within the satz column. The IQRs (green horizontal lines) declined, except in the case of peak at nadir.
Figure 9.
Scattergram of DNB radiances versus lunar illuminance (LI) for a grid cell in Harare, Zimbabwe. DNB radiances are corrected to remove the reflected moonlight component to the signal. The presence of reflected moonlight is evident from the blank wedge zone on the low-radiance end of the distribution. The wedge widens as the LI increases. The wedge dissipates when the reflected LI component is subtracted.
Figure 9.
Scattergram of DNB radiances versus lunar illuminance (LI) for a grid cell in Harare, Zimbabwe. DNB radiances are corrected to remove the reflected moonlight component to the signal. The presence of reflected moonlight is evident from the blank wedge zone on the low-radiance end of the distribution. The wedge widens as the LI increases. The wedge dissipates when the reflected LI component is subtracted.
Figure 10.
DNB temporal profiles from Houston and Harare, each with a Landscan population count of near 2000. The center of the Houston grid cell is at 29.7375 N, 95.4292 W. The center of the Harare grid cell is at 17.8625 S, 31.0333 E.
Figure 10.
DNB temporal profiles from Houston and Harare, each with a Landscan population count of near 2000. The center of the Houston grid cell is at 29.7375 N, 95.4292 W. The center of the Harare grid cell is at 17.8625 S, 31.0333 E.
Figure 11.
DNB histograms for the Houston and Harare points from
Figure 10. The two distributions are nearly 100% disjointed.
Figure 11.
DNB histograms for the Houston and Harare points from
Figure 10. The two distributions are nearly 100% disjointed.
Figure 12.
DNB temporal profile of a location in Yemen showing prominent annual cycles in the brightness of the lights. January is always the brightest month and July the dimmest. The lower panel shows the corresponding autocorrelation chart with a peak near the one-year mark.
Figure 12.
DNB temporal profile of a location in Yemen showing prominent annual cycles in the brightness of the lights. January is always the brightest month and July the dimmest. The lower panel shows the corresponding autocorrelation chart with a peak near the one-year mark.
Figure 13.
The % annual lighting deficit (orange horizonal lines) is the percent loss in monthly average radiances over a year relative to the brightest month. Open circles indicate anomalously bright radiances excluded from the analysis, associated with the annual Diwali holidays.
Figure 13.
The % annual lighting deficit (orange horizonal lines) is the percent loss in monthly average radiances over a year relative to the brightest month. Open circles indicate anomalously bright radiances excluded from the analysis, associated with the annual Diwali holidays.
Figure 14.
Monthly outage rates (orange horizontal lines) following the aerial bombing of Sana’a, Yemen, in late March, 2015. Following the bombing, the outage rates jumped to 100% for 10 of the first 12 months. Small radiance spikes are associated with Ramadan in 2016 and 2017, plus a slight rise in radiance above the detection limit resulted in fluctuating monthly outage rates starting from June 2016. The last recorded outage event was 28 October 2017.
Figure 14.
Monthly outage rates (orange horizontal lines) following the aerial bombing of Sana’a, Yemen, in late March, 2015. Following the bombing, the outage rates jumped to 100% for 10 of the first 12 months. Small radiance spikes are associated with Ramadan in 2016 and 2017, plus a slight rise in radiance above the detection limit resulted in fluctuating monthly outage rates starting from June 2016. The last recorded outage event was 28 October 2017.
Figure 15.
Upward trend in the DNB profile from Gaya, India.
Figure 15.
Upward trend in the DNB profile from Gaya, India.
Figure 16.
This grid cell started out with nearly zero lift. Lift began to rise midway through 2017.
Figure 16.
This grid cell started out with nearly zero lift. Lift began to rise midway through 2017.
Figure 17.
Histogram and scattergram of the 2018 mean values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the mean versus population count from 2018.
Figure 17.
Histogram and scattergram of the 2018 mean values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the mean versus population count from 2018.
Figure 18.
Picket-fence chart of the mean radiances for each of the test sites by year.
Figure 18.
Picket-fence chart of the mean radiances for each of the test sites by year.
Figure 19.
Picket-fence chart of the radiance per person for each of the test sites by year.
Figure 19.
Picket-fence chart of the radiance per person for each of the test sites by year.
Figure 20.
Histogram and scattergram of the 2018 variance values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the variance versus population count from 2018.
Figure 20.
Histogram and scattergram of the 2018 variance values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the variance versus population count from 2018.
Figure 21.
Picket-fence chart of the mean variance for each of the test sites by year.
Figure 21.
Picket-fence chart of the mean variance for each of the test sites by year.
Figure 22.
Histogram and scattergram of the 2018 skew values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the skew versus population count from 2018.
Figure 22.
Histogram and scattergram of the 2018 skew values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the skew versus population count from 2018.
Figure 23.
Picket-fence chart of the mean skew for each of the test sites by year.
Figure 23.
Picket-fence chart of the mean skew for each of the test sites by year.
Figure 24.
Skew versus mean for Pyongyang versus Harare from 2013. Pyongyang’s skew is anomalously high. This is caused by the high-end outliers present in the DNB profiles, where the data cloud is pressed against the sensor’s detection limit.
Figure 24.
Skew versus mean for Pyongyang versus Harare from 2013. Pyongyang’s skew is anomalously high. This is caused by the high-end outliers present in the DNB profiles, where the data cloud is pressed against the sensor’s detection limit.
Figure 25.
Histogram and scattergram of the 2018 kurtosis values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the kurtosis versus population count from 2018.
Figure 25.
Histogram and scattergram of the 2018 kurtosis values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the kurtosis versus population count from 2018.
Figure 26.
Picket-fence chart of the mean kurtosis for each of the test sites by year.
Figure 26.
Picket-fence chart of the mean kurtosis for each of the test sites by year.
Figure 27.
Kurtosis versus the mean radiance for Pyongyang and Harare. Pyongyang has an anomalously high kurtosis.
Figure 27.
Kurtosis versus the mean radiance for Pyongyang and Harare. Pyongyang has an anomalously high kurtosis.
Figure 28.
Histogram and scattergram of the 2018 Cumulative Distribution Function (CDF) slope values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the CDF slope versus population count from 2018.
Figure 28.
Histogram and scattergram of the 2018 Cumulative Distribution Function (CDF) slope values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the CDF slope versus population count from 2018.
Figure 29.
Picket-fence chart of the mean CDF slope for each of the test sites by year.
Figure 29.
Picket-fence chart of the mean CDF slope for each of the test sites by year.
Figure 30.
Histogram and scattergram of the 2018 lift index values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as the “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the lift index versus population count from 2018.
Figure 30.
Histogram and scattergram of the 2018 lift index values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as the “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the lift index versus population count from 2018.
Figure 31.
Picket-fence chart of the mean lift for each of the test sites by year.
Figure 31.
Picket-fence chart of the mean lift for each of the test sites by year.
Figure 32.
Histogram and scattergram of the 2018 ALD values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the ALD versus population count from 2018.
Figure 32.
Histogram and scattergram of the 2018 ALD values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the ALD versus population count from 2018.
Figure 33.
Picket-fence chart of the mean ALD % for each of the test sites by year.
Figure 33.
Picket-fence chart of the mean ALD % for each of the test sites by year.
Figure 34.
Histogram and scattergram of the 2018 primay cycle lag from the autocorrelation analysis. Values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) are blue and for the developed set are red (Houston, Seoul, Shanghai, Beijing). The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the lag versus population count from 2018.
Figure 34.
Histogram and scattergram of the 2018 primay cycle lag from the autocorrelation analysis. Values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) are blue and for the developed set are red (Houston, Seoul, Shanghai, Beijing). The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the lag versus population count from 2018.
Figure 35.
Picket-fence chart of the percent of grid cells with annual cycling for each of the test sites.
Figure 35.
Picket-fence chart of the percent of grid cells with annual cycling for each of the test sites.
Figure 36.
Histogram and scattergram of the 2018 slope of the Longterm trend (LTT) values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the LTT versus the population count from 2018.
Figure 36.
Histogram and scattergram of the 2018 slope of the Longterm trend (LTT) values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the LTT versus the population count from 2018.
Figure 37.
Picket-fence chart of the mean slope of the long-term trend for each of the test sites.
Figure 37.
Picket-fence chart of the mean slope of the long-term trend for each of the test sites.
Figure 38.
Histogram and scattergram of the 2018 dispersion values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the dispersion versus population count from 2018.
Figure 38.
Histogram and scattergram of the 2018 dispersion values for the developing set (Harare, Nairobi, Dhaka, and Pyongyang) versus the developed set (Houston, Seoul, Shanghai, Beijing). The developed set is red and developing set is blue. The two sets are referred to as “lower” and “upper”, respectively. The sample set excludes grid cells deemed to be permanent background, where no lighting is detected. The scattergram is the dispersion versus population count from 2018.
Figure 39.
Picket-fence chart of the mean dispersion for each of the test sites.
Figure 39.
Picket-fence chart of the mean dispersion for each of the test sites.
Figure 40.
Variance versus mean for grid cells in Nairobi and Dhaka. While the range of means is quite similar between the two cities, Nairobi has a higher variance. This is an indication that the voltage varies over time in Nairobi as compared to Dhaka.
Figure 40.
Variance versus mean for grid cells in Nairobi and Dhaka. While the range of means is quite similar between the two cities, Nairobi has a higher variance. This is an indication that the voltage varies over time in Nairobi as compared to Dhaka.
Figure 41.
Percent outage rates for the study sites by year.
Figure 41.
Percent outage rates for the study sites by year.
Figure 42.
Percent of population in the dark by year.
Figure 42.
Percent of population in the dark by year.
Figure 43.
Observed and satz-normalized radiances from a grid cell in Nairobi, Kenya.
Figure 43.
Observed and satz-normalized radiances from a grid cell in Nairobi, Kenya.
Figure 44.
Temporal profiles from a grid cell in Nairobi showing the variance reducing in a stepwise fashion following the satellite zenith, normalization, and lunar illuminance correction. In total, the variance drops 65%.
Figure 44.
Temporal profiles from a grid cell in Nairobi showing the variance reducing in a stepwise fashion following the satellite zenith, normalization, and lunar illuminance correction. In total, the variance drops 65%.
Table 1.
DNB temporal profiles extracted for this study.
Table 1.
DNB temporal profiles extracted for this study.
Location | Country | Population Covered | Grid Size | Status |
---|
Rohingya Camps | Bangladesh | 1,070,436 | 36 × 75 | Developing |
Pyongyang | North Korea | 2,353,190 | 75 × 40 | Developing |
Harare | Zimbabwe | 1,631,405 | 50 × 50 | Developing |
Nairobi | Kenya | 5,129,731 | 100 × 70 | Developing |
Dhaka | Bangladesh | 15,015,586 | 73 × 97 | Developing |
Sana’a | Yemen | 1,615,498 | 45 × 73 | Developing |
Mari’b | Yemen | 91,155 | 41 × 41 | Developing |
Charpakri | India | 16,862 | 10 × 10 | Developing |
Gaya | India | 727,277 | 31 × 31 | Developing |
Delhi | India | 7,022,476 | 25 × 250 | Developing |
Beijing | China | 12,451,668 | 100 × 100 | Developed |
Shanghai | China | 11,264,168 | 60 × 60 | Developed |
Wuhan | China | 7,048,797 | 120 × 120 | Developed |
Seoul | South Korea | 13,011,839 | 105 × 69 | Developed |
Houston | USA | 5,783,193 | 200 × 200 | Developed |
Santa Rosa, CA | USA | 15,175 | 145 × 145 | Developed |
Table 2.
Index classification accuracies for a developed (Houston, Seoul, Shanghai, Beijing) and developing set (Harare, Nairobi, Dhaka, Pyongyang).
Table 2.
Index classification accuracies for a developed (Houston, Seoul, Shanghai, Beijing) and developing set (Harare, Nairobi, Dhaka, Pyongyang).
Variable | Year | Accuracy Solo | Accuracy with Population |
---|
Mean | 2018 | 74% | 85% |
Variance | 2018 | 68% | 84% |
Skew | 2018 | 62% | 77% |
Kurtosis | 2018 | 62% | 75% |
Lift | 2018 | 70% | 86% |
CDF slope | 2018 | 53% | 71% |
ALD % | 2018 | 56% | 71% |
Outage % | 2018 | 57% | 66% |
Dispersion | 2108 | 58% | 75% |
Primary ACF cycle | All | 69% | 73% |
Long term trend | All | 59% | 72% |