Generating the Nighttime Light of the Human Settlements by

Furthermore, the cumulative digital numbers (CDNs) and number of light area pixels (NLAP) of the generated stable light and NOAA/NGDC stable light wer...
0 downloads 0 Views 5MB Size
Download PDF
Article pubs.acs.org/est

Generating the Nighttime Light of the Human Settlements by Identifying Periodic Components from DMSP/OLS Satellite Imagery Husi Letu,*,†,∥ Masanao Hara,‡ Gegen Tana,§ Yuhai Bao,∥ and Fumihiko Nishio§,∥ †

Research and Information Cener, Tokai University, 2-28-4 Tomigawa, Shibuya-ku, Tokyo 151-0063, Japan VTI Research Institute, VisionTech, Inc., 2-1-16 Umezono, Tsukuba City, Ibaraki 305-0045, Japan § Center for Environmental Remote Sensing (CEReS), Chiba University, 1-33 Yayoi-cho, Inage-ku, Chiba 263-8522, Japan ∥ Remote Sensing and GIS Key Laboratory, Inner Mongolia Normal University, 81 Zhaowuda street, Hohhot 010022, China ‡

S Supporting Information *

ABSTRACT: Nighttime lights of the human settlements (hereafter, “stable lights”) are seen as a valuable proxy of social economic activity and greenhouse gas emissions at the subnational level. In this study, we propose an improved method to generate the stable lights from Defense Meteorological Satellite Program/Operational Linescan System (DMSP/ OLS) daily nighttime light data for 1999. The study area includes Japan, China, India, and other 10 countries in East Asia. A noise reduction filter (NRF) was employed to generate a stable light from DMSP/OLS timeseries daily nighttime light data. It was found that noise from amplitude of the 1-year periodic component is included in the stable light. To remove the amplitude of the 1-year periodic component noise included in the stable light, the NRF method was improved to extract the periodic component. Then, new stable light was generated by removing the amplitude of the 1-year periodic component using the improved NRF method. The resulting stable light was evaluated by comparing it with the conventional nighttime stable light provided by the National Oceanic and Atmosphere Administration/National Geophysical Data Center (NOAA/NGDC). It is indicated that DNs of the NOAA stable light image are lower than those of the new stable light image. This might be attributable to the influence of attenuation effects from thin warm water clouds. However, due to overglow effect of the thin cloud, light area in new stable light is larger than NOAA stable light. Furthermore, the cumulative digital numbers (CDNs) and number of light area pixels (NLAP) of the generated stable light and NOAA/NGDC stable light were applied to estimate socioeconomic variables of population, electric power consumption, gross domestic product, and CO2 emissions from fossil fuel consumption. It is shown that the correlations of the population and CO2FF with new stable light data are higher than those in NOAA stable light data; correlations of the EPC and GDP with NOAA stable light data are higher those in the new stable light data.

■. INTRODUCTION

version 4 of stable light product for 1992−2009 (http://ngdc. noaa.gov/eog/download.html).26 In the stable light generation methods, DMSP thermal infrared data and visual determination techniques were used to eliminate clouds by utilizing their low relative temperature. However, when relying on a visual determination technique, criteria for distinguishing cloud from cloud-free might vary among observers, making it difficult to establish quantitative standards for cloud determination. Furthermore, in high-latitude regions of the Northern Hemisphere, wintertime ground temperatures are often lower than cloud temperatures, making it difficult to identify clouds on the basis of thermal infrared data.27 Additionally, detecting thin

The visible and near-infrared channel (VIS) of the Defense Meteorological Satellite Program/Operational Linescan System (DMSP/OLS) satellite sensors is capable of observing clouds affected by moonlight, lights from human settlements, gas flares, fishing boats, and ephemeral events such as fires, auroras, and lightning-affected clouds.1 Among the nighttime light in DMSP/OLS satellite data, light from human settlements remains stable throughout the year, and this “stable light” is important for estimating economic activity,2−5 electric power consumption,1,6,7 fossil fuel carbon dioxide emissions,1,7−11 urbanization and urban expansion,12−15 population,16−20 and light pollution at the national and regional levels.21−24 Elvidge et al.1,25 developed a stable light by averaging the digital numbers (DNs) of Operational Linescan System visible light (OLS-VIS) data across cloud-free observations. In 2010, the National Oceanic and Atmosphere Administration/National Geophysical Data Center (NOAA/NGDC) released © 2015 American Chemical Society

Received: Revised: Accepted: Published: 10503

May 27, 2015 August 10, 2015 August 17, 2015 August 17, 2015 DOI: 10.1021/acs.est.5b02471 Environ. Sci. Technol. 2015, 49, 10503−10509

Article

Environmental Science & Technology

Table 1. Stable Light Data and Statistical Data of the Population, Electric Power Consumption, GDP and CO2 Emission by Fossil Fuels Consumption in 13 Countries of Asia country

CDNNEW*a ( × 104)

NLAPNEW*b ( × 103)

CDNNOAA*c ( × 104)

NLAPNOAA*d ( × 103)

Population ( × 106)

EPCTOT*e ( × 109 kWh)

GDP ( × 109 USD)

CO2FF*f (104 kt)

Bangladesh Philippines Vietnam China India Japan Pakistan South Korea Thailand Singapore Mongolia Sri Lanka Nepal

9.97 19.41 20.62 907.56 468.92 625.92 125.63 195.57 197.16 2.26 7.67 5.52 3.29

4.54 8.19 10.25 383.12 202.33 194.43 55.38 62.35 87.68 1.46 4.68 2.29 2.04

10.74 18.94 14.30 642.43 341.19 530.77 99.30 160.51 67.58 0.63 1.88 7.96 1.77

4.40 6.79 6.04 245.10 147.22 166.22 45.83 56.09 28.25 0.23 0.77 3.87 0.85

140.58 76.02 76.60 1252.74 1025.01 126.65 140.58 46.62 61.62 3.96 2.38 19.06 22.69

13.21 35.65 19.99 1179.80 397.71 989.90 48.22 227.44 84.61 28.46 2.53 4.92 1.29

45.69 83.00 28.68 1254.96 466.87 4432.60 62.97 445.40 122.63 85.96 1.06 15.66 5.03

2.52 6.92 4.77 336.23 114.44 119.80 10.04 39.98 18.28 5.01 0.76 0.85 0.32

*a

CDNNEW: Cumulative DNs of the new stable light. *bNLAPNEW: Number of light area pixels in the new stable light. *cCDNNOAA: Cumulative DNs of the NOAA stable light. *dNLAPNOAA: Number of light area pixels in the NOAA stable light. *eEPCTOT: Total electric power consumption. *f CO2FF: CO2 emission by fossil fuels consumption.

Figure 1. Flowchart for generating the stable light using improved NRF method.

the time-series daily nighttime light data. To reduce the underestimation of the DNs caused by thin clouds, compositing maximum DNs (CMD) method using daily OLS-VIS data over the span of 10 days (hereafter, 10-day CMD) was employed to develop the time-series nighttime light data set. Then, the stable light was developed from the time-series 10-day CMD data set by using the NRF method to remove the periodic components. Furthermore, the characteristics of the stable light (hereinafter, the new stable light) were investigated by comparing with other stable light obtained from the NOAA/ NGDC. Finally, population, electric power consumption (EPC), GDP, and CO2 emissions from fossil fuel consumption (CO2FF) were estimated from cumulative digital numbers (CDNs) and number of light area pixels (NLAP) of the new stable light and NOAA/NGDC stable light data, respectively.

warm clouds, such as fog, is difficult when using only OLS thermal band data and OLS-VIS data.25 Hara et al.28,29 developed a noise reduction filter (NRF) for separating periodic components, such as seasonal variation of the vegetation index and lunar illumination, from satellite timeseries data. Furthermore, stable light image was generated from time-series daily nighttime light data using the NRF by eliminating the cloud pixels as a random noise without using the thermal infrared data. The periodic components due to sunlight and moonlight were included in the generated stable light image.6 However, periodic components are a kind of noise in nighttime light imagery when generating the stable light image. Thus, a quantitative analysis of the periodic components in daily nighttime light data is important for eliminating these components from stable light imagery. The aim of the present study was to generate a stable light using all of the DMSP/OLS daily nighttime light data for 1999 by the NRF method. First, amplitude of periodic components were identified by analyzing the variation of the average DNs of

■. DATA AND METHODS Using Data. Daily DMSP-F14/OLS-VIS nighttime light data for 1999 was obtained from the Satellite Image Database 10504

DOI: 10.1021/acs.est.5b02471 Environ. Sci. Technol. 2015, 49, 10503−10509

Article

Environmental Science & Technology

Figure 2. (a) Average DNs of the time-series daily mosaic imagery which is processed by the method shown in Figure 2. (b) Variation of average DNs of the image in Figure 3a at low, middle and high latitude (The marked points by “○” shows the date of full moon) .

Figure 3. (a) Variation of the DNs of the sample points (S1: West Siberia Plain, S2: China’s Taklamakan Desert, S3: Indian Ocean). (b) Simulation of the amplitude of the 1-year component and the bias component in the sample points using the NRF.

System (SIDaB, http://rms1.agsearch.agropedia.affrc.go.jp/ sidab/index-ja.html) of the Japanese Ministry of Agriculture, Forestry and Fisheries, Agriculture’s Information Resources System (AGROPEDIA), which is used in this study for generating the new stable light. The SIDaB distribute the OLSVIS nighttime data with a length of 8-bit (0−255).6 The DMSP-F14/OLS stable light product for 1999 obtained from the NOAA/NGDC (hereinafter, NOAA stable light) was used for comparing with new stable light data. NOAA stable light data is produced by all the available cloud-free data for that particular calendar year in the NGDC’s digital archive. Statistic data of the population, EPC, GDP, and CO2FF as shown in Table 1 were obtained from the World Bank statistics (http:// data.worldbank.org/indicator). The study area was defined as the area in the region 5° N−55° N, 68° E−150° E including 13 Asian countries, which is determined by the limitation of the coverage area of the DMSP/OLS satellite data due to the antenna in the SIDaB. Method for Generating the Stable Light. Hara et al.29 used the NRF method to generate the bias component as a stable light from time-series OLS-VIS data set by 10-day CMD method. However, we found that noise from amplitude of the 1-year periodic component is included in the extracted stable light. To improve the accuracy of the stable light, we improved the NRF method to generate the stable light by eliminating the noise from amplitude of the 1-year periodic component. The definitions of the bias components and periodic components are given in section S1 in the Supporting Information (SI). Equation 1 (SI) illustrates the original NRF method, which is

useful for extracting the bias components and coefficient of periodic components; eq 2 (SI) illustrates the improved NRF method, which is useful for extracting the amplitude of periodic components such as the amplitude of 1-year periodic components by sunlight illumination and the amplitude of 1month periodic components by moonlight illumination. The detailed description about the algorithm of NRF method is given in section S1 in the SI (see Figure S1). Figure 1 shows a flowchart for generating the stable light using improved NRF method. First, a daily mosaic OLS-VIS data set was used for identifying periodic components by analyzing the averaged DNs as a supplemental process (see Figure S2 in the SI). Then, the 10-day CMD method was used to develop a time-series composition data set. The bias component and other periodic components were extracted by the NRF method from the 10-day CMD data set. After that, three samples, one each from low-, mid-, and high-latitude regions were selected for evaluating the effect of the NRF method. Furthermore, DNs of the time-series composition data in the three samples were simulated by the NRF method as a supplemental process. Finally, stable light was extracted by removing periodic components from the bias component based on the results of the supplemental process. The 10-day CMD is a composition method for extracting the maximum DNs of the daily OLS-VIS mosaic image during 10 days (see Figures S3, S4, and S5 in the SI). In this study, the 10-day CMD method is employed to develop the time-series composition data set for generating the stable light. The 10-day CMD method is efficient for extracting the ground surface 10505

DOI: 10.1021/acs.est.5b02471 Environ. Sci. Technol. 2015, 49, 10503−10509

Article

Environmental Science & Technology

observation image will be affected by solar radiation and that satellite images will be increasingly bright at higher latitudes, an effect that can be seen in Figure 2a. DN variation shows significant dependence on latitude and indicates monthly periodicity. The dots in Figure 2b show that peak DNs are generally consistent with full moons and that there are 12 peaks over the course of the year. In some cases, DN peaks do not coincide precisely with the full moon, but instead occur just before or after. This is likely because areas of cloud cover are more luminous by reflected moonlight, so average DNs of the images with high cloud cover will be higher than those on image with low cloud cover. Shorter, more irregular examples of periodicity can also be seen. Such irregular variation is likely due to random elements, such as clouds. Because there is periodic variation from sunlight and moonlight illumination in the nighttime light imagery, stable light can be extracted by separating each periodic component from nighttime light data. Simulation of Periodic Components in Nighttime Light Data. To quantitatively investigate the relationship between nighttime light components and the periodic component, DNs of the 10-day CMD data set, at three sample pixels, were simulated by applying NRF. Three sample pixels were taken where stable light was not expected to be observed, one from each of the plains of western Siberia (S1: 54° 86′ N, 93° 49′ E), the Taklamakan Desert in China (S2: 38° 26′ N, 86° 98′ E), and the Indian Ocean (S3: 17° 47′ N, 91° 36′ E). Figure 3a shows the variation in time-series 10-day CMD data obtained by applying NRF at the sampled point. The DNs increase from January to June, then decrease from June to December. The amount of variation of the DNs increases with latitude, and the 1-year period is confirmed clearly from the data. To demonstrate the periodicity of time-series data from each sample point, the NRF method (see eq 2 in the SI) was used to approximate the amplitude and the bias component. Figure 3b shows the relation between the amplitudes of the 1-year periodic component and the bias component. Because there is no stable light at the three sampled points, the DNs should be low and constant. However, the results of analysis using the 10day CMD data set show that the DNs of the sampled points

information from time-series satellite data when the observed area is covered by clouds and haze. Cloud noise included in the 10-day CMD data set was removed by the NRF method. The NRF method was used to generate the bias component and amplitude of periodic components in the nighttime light data by removing random noise from the time-series CMD composition data set.

■. RESULTS AND DISCUSSION Identifying Periodic Components from Average DNs of the Daily Nighttime Light Data. To analyze periodic Table 2. Comparison of DNs at the Same Points of the Bias Component Image and the Amplitude of the 1-Year Periodic Component name of the components DNs of bias component DNs of one year period component difference of DNs

sample 1: West Siberia Plain

sample 2: Taklamakan Desert

sample 3: Indian Ocean

sample 4: Sapporo City

112

68

53

126

110

66

53

1

2

2

0

125

components in nighttime light data, the average DNs of the time-series zonal mean images (Figure 2a) were generated from a daily mosaic data set by the processing method shown in Figure S2 in the SI. Figure 2b shows the variations in DNs at low- (5°N), mid- (30°N), and high-latitude (55°N) regions in Figure 2a. The DNs vary throughout the year, with high DNs near the summer solstice in June and low DNs near the winter solstice in December. Additionally, DNs shows some dependence on latitude and indicates higher at high-latitude regions and lower at low-latitude regions, which is probably due to the twilight zone as a function of latitude. In mid-latitude regions, however, the influence of cloud cover in the daily mosaic data set results in lower average DNs. A reasonable cause for this is that DMSP satellites observe the study area during 18:00 to 24:00, Japan Standard Time (JST). At the summer solstice, the sun has not yet set at 18:00, so it can be inferred that the

Figure 4. Stable light image in the study area. 10506

DOI: 10.1021/acs.est.5b02471 Environ. Sci. Technol. 2015, 49, 10503−10509

Article

Environmental Science & Technology

moonlight illumination was eliminated by the 10-day CMD method. Generating the Stable Light Using NRF Method. Based on the above results, new stable light images were generated as the difference between the bias component (see Figure S6 in the SI) and the amplitude of the 1-year periodic component (see Figure S7 in the SI). Figure 4 shows the new stable light image generated in this study. Comparing this image with the bias component shows that the components due to sunlight in the bias component have been eliminated, and the distribution of the nighttime light from human settlements is seen clearly. As can be seen from the magnification of the image around Hokkaido, stable light has been extracted and other periodic components have been eliminated. Figure 5 shows the 10-day CMD image, new stable light image, and NOAA stable light image in the surrounding areas of Pearl River Delta, Taiwan, Soul, New Delhi, and Tokyo. From the images it was confirmed that the DNs of the NOAA stable light image are lower than those of the new stable light image and those of the 10-day CMD image, especially in highDN regions. This might be attributable to the influence of attenuation effects from thin warm water clouds. Because thin cloud can obscure and diffuse light emissions from Earth’s surface, DNs of the nighttime stable light in human settlement areas might be underestimated when generating the NOAA stable light from cloud-free daily nighttime light dada with low gain setting and low moonlight illumination by averaging the DNs of these data. The 10-day CMD method used for generating new stable light data can avoid being affected by attenuation from thin warm water clouds through extracting the maximum DNs during the 10 days. Thus, it is considered that new stable light data have not influenced by attenuation effect from thin warm water clouds. At the same time, the thin cloud increases atmospheric scattering, which possibly contributes to overglow effects. The overglow effect increases the DNs in urban surrounding areas with low light intensity and leads to an increase in the urban light area.13 The 10-day CMD method used for generating new stable light can extract the light from overglow effect, increasing the light area in urban surrounding area. As seen from Figure 5, it is evident that light area in new stable light image has more overglow than NOAA stable light image. Applications: Estimation of the Population, Electric Power Consumption, GDP, and CO2 Emissions. Figure 6 shows a comparison of regression analyses between statistical data (population, EPC, GDP, and CO2 fossil-fuel emissions (CO2FF)) and, separately, new stable light data and NOAA stable light data. There are good linear relations between CDN and NLAP of the stable light data and the four parameters other than population and GDP. The CDN and NLAP were calculated by removing the background noise with a threshold less than 10 from stable light data. The coefficients of determination in the regression analysis of the NLAP with population and CO2FF are higher than those for CDN. Furthermore, it is seen that the correlations of the population and CO2FF with CDN and NLAP in the new stable light data are higher than those in NOAA stable light data; correlations of the EPC and GDP with CDN and NLAP in NOAA stable light data are higher those in the new stable light data. As shown in Table 1, the CDN and NLAP of the new stable light data are larger than those of the NOAA stable light data. This is attributed to the influence of the method for generating

Figure 5. Comparison of 10-day CMD images (December 1st to 10th, 1999), new stable light images and NOAA stable light images for 1999: (a) Pearl River Delta, (b) Taiwan, (c) Soul, (d) New Delhi, and (e) Tokyo. (New stable light image and 10-day CMD image were converted into 6-bit data for comparison with NOAA stable light.)

also vary with a 1-year period due to the effects of sunlight illumination. As a result, the bias component is higher than expected. To quantitatively evaluate the relation between the bias component and the 1-year periodic component, a sample point that includes the city of Sapporo in Hokkaido prefecture, Japan, (S4: 43° 4′ N, 140° 54′ E) was added to the three points sampled at low, middle, and high latitudes. As shown in Table 2, amplitude of 1-year periodic component shows increased effects of sunlight with increased latitude, but that effect can be eliminated by taking the difference between the bias component and the amplitude of the 1-year periodic component. The 1-month periodic component induced by 10507

DOI: 10.1021/acs.est.5b02471 Environ. Sci. Technol. 2015, 49, 10503−10509

Article

Environmental Science & Technology

Figure 6. Relationship between stable light data and statistical data of population, electric power consumption, GDP, and CO2 emissions by fossil fuels consumptions for 13 countries of Asia.

resulting in different coefficients in the regression analysis between two table lights and socioeconomical variables. (Summary and conclusions: see section S5 in the SI).

the NOAA stable light by averaging the cloud-free OLS-VIS data. The differences in approaches to creating new stable light data and NOAA stable light data are summarized as follows: (1) New stable light data were generated from daily nighttime light data during 1 year. The NRF method was used to eliminate clouds and other noise components caused by various gain settings, moonlight, and sunlight illumination. A maximum-values composition method was used for developing the nighttime light data set when generating the new stable light data. (2) NOAA stable light data were created from cloudfree daily nighttime light data from October 1st to April 30th of the next year with low gain setting and low moonlight illumination.1 A visual determination technique, OLS thermal band data, and ground surface temperature data were used to detect cloud distributions. A method of averaging the DNs was used for generating the NOAA stable light data. Due to the difference in approaches, it is likely that the characteristics of the two stable lights are quite different from each other,

■

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.est.5b02471. NRF method, process for identifying periodic components from daily nighttime data, the 10-day CMD method, the bias component and the amplitude of the yearly component, summary and conclusions. (PDF)

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel.: +81-90-2933-1386. Fax: +81-3-3481-0610. 10508

DOI: 10.1021/acs.est.5b02471 Environ. Sci. Technol. 2015, 49, 10503−10509

Article

Environmental Science & Technology Notes

(16) Briggs, D. J.; Gulliver, J.; Fecht, D.; Vienneau, D. M. Dasymetric modelling of small-area population distribution using land cover and light emissions data. Remote Sens. Environ. 2007, 108, 451−466. (17) Doll, C. N. H.; Pachauric, H. Estimating rural populations without access to electricity in developing countries through nighttime light satellite imagery. Energy Policy 2010, 38, 5661−5670. (18) Amaral, S.; Camara, C.; Monteiro, A. M. V.; Quintanilha, J. A.; Elvidge, C. D. Estimating population and energy consumption in Brazilian Amazonia using DMSP nighttime satellite data. Comput. Environ. Urban Syst., Computers 2005, 29, 179−195. (19) Sutton, P.; Roberts, D.; Elvidge, C. D.; Baugh, D. Census from Heaven: an estimate of the global human population using night-time satellite imagery. Int. J. Remote Sens. 2001, 22, 3061−3076. (20) Sutton, P. C.; Taylor, M. J.; Elvidge, C. D. Using DMSP OLS imagery to characterize urban populations in developed and developing countries. Remote Sensing and Digital Image Processing. 2010, 10, 329−348. (21) Chalkias, C.; Petrakis, M.; Psiloglou, B.; Lianou, M. Modelling of light pollution in suburban areas using remotely sensed imagery and GIS. J. Environ. Manage. 2006, 79, 57−63. (22) Kuechly, H. U.; Kyba, C.; Ruhtz, T.; Lindemann, C.; Wolter, C.; Fischer, J.; Hölker, F. Aerial survey and spatial analysis of sources of light pollution in Berlin, Germany. Remote Sens. Environ. 2012, 126, 39−50. (23) Levin, N.; Johansen, K.; Hacker, J. M.; Phinn, S. A new source for high spatial resolution night time imagesThe EROS-B commercial satellite. Remote Sens. Environ. 2014, 149, 1−12. (24) Chalkias, C.; Petrakis, M.; Psiloglou, B.; Lianou, M. Modelling of light pollution in suburban areas using remotely sensed imagery and GIS. J. Environ. Manage. 2006, 79, 57−63. (25) Elvidge, C. D.; Baugh, K. E.; Kihn, E. A.; Kroehl, H. W.; Davis, E. R. Mapping city lights with nighttime data from the DMSP Operational Linescan System. Photogramm. Eng. Remote Sens. 1997, 63, 727−734. (26) Baugh, K.; Elvidge, C. D; Ghosh, T.; Ziskin, D. Development of a 2009 Stable Lights Product using DMSP-OLS data. Proceedings of the Asia-Pacific Advanced Network 2010, 114−130. (27) Kanamitsu, M.; Ebisuzaki, W.; Woollen, J.; Yang, S. K.; Hnilo, J. J.; Fiorino, M.; Potter, G. L. NCEP−DOE AMIP-II Reanalysis (R-2). Bulletin of the American Meteorological Society. 2002, 83, 1631−1643. (28) Hara, M.; Okada, S.; Yagi, H.; Moriyama, T.; Shigehar, K.; Sugimori, Y. Developing and evaluation of the noise reduction filter for the time-series satellite imagery. Japan Society of Photogrammetry and Remote Sensing. 2003, 42, 48−59 (in Japanese). (29) Hara, M.; Okada, S.; Yagi, H.; Moriyama, T.; Shigehara, K.; Sugimori, Y. Progress for stable artificial lights distribution extraction accuracy and estimation of electric power consumption by means of DMSP/OLS nighttime imagery. Int. J. Remote Sens. Earth Sci. 2010, 1, 31−42.

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors thank the Satellite Image Database System (SIDaB) of the Ministry of Agriculture, Forestry and Fisheries Agriculture Information Resources System (AGROPEDIA) of Japan and National Oceanic and Atmosphere Administration/ National Geophysical Data Center (NOAA/NGDC) for providing DMSP/OLS dailly nighttime light data and stable light product. This study was supported by National Natural Science Foundation of China (61261030).

■

REFERENCES

(1) Elvidge, C. D.; Imhoff, M. L.; Baugh, K. E.; Hobson, V. R.; Nelson, I.; Dietz, J. B. Nighttime light of the world: 1994−1995. ISPRS J. Photogramm. Remote Sens. 2001, 56, 81−99. (2) Elvidge, C. D.; Baugh, K. E.; Kihn, E. A.; Kroehl, H. W.; Davis, E. R.; Davis, C. Relation between satellite observed visible-near infrared emission, population, and energy consumption. Int. J. Remote Sens. 1997a, 18, 1373−1379. (3) Ghosh, T.; L Powell, R.; D Elvidge, C.; E Baugh, K.; Sutton, P. C.; Anderson, S. (2010). Shedding light on the global distribution of economic activity. Open Geography Journal 2010, 3 (3), 147−160. (4) Ghosh, T.; Powell, R. L.; Anderson, S.; Sutton, P. C.; Elvidge, C. D. Informal economy and remittance estimates of India using nighttime imagery. Ph.D. Thesis, Centre for Environment, Social and Economic Research Publications, 2010. (5) Zhao, N.; Currit, N.; Samson, E. Net primary production and gross domestic product in China derived from satellite imagery. Ecol. Econ. 2011, 70, 921−928. (6) Letu, H.; Hara, M.; Yagi, H.; Naoki, K.; Tana, G.; Nishio, F.; Okada, S. Estimating energy consumption from nighttime DMSP/ OLS imagery after correcting for saturation effects. Int. J. Remote Sens. 2010, 31, 4443−4458. (7) Letu, H.; Nakajima, T. Y.; Nishio, F. Regional-scale estimation of electric power and power plant CO2 emissions using defense meteorological satellite program operational linescan system nighttime satellite data. Environ. Sci. Technol. Lett. 2014, 1, 259−265. (8) Ghosh, T.; Elvidge, C. D.; Sutton, P. C.; Baugh, K. E.; Ziskin, D.; Tuttle, B. T. Creating a global grid of distributed fossil fuel CO2 emissions from nighttime satellite imagery. Energies 2010, 3, 1895− 1913. (9) Doll, C. N. H.; Muller, J. P.; Elvidge, C. D. Nighttime imagery as a tool for globe mapping of socioeconomic parameters and greenhouse gas emissions. Ambio 2000, 29, 157−162. (10) Raupach, M. R.; Rayner, P. J.; Paget, M. Regional variations in spatial structure of nightlights, population density and fossil-fuel CO2 emissions. Energy Policy 2010, 38 (9), 4756−4764. (11) Oda, T.; Maksyutov, S. A very high-resolution (1km × 1 km) global fossil fuel CO2 emission inventory derived using a point source database and satellite observations of nighttime lights. Atmos. Chem. Phys. 2011, 11, 543−556. (12) Zhang, Q.; Seto, K. C. Mapping urbanization dynamics at regional and global scales using multi-temporal dmsp/ols nighttime light data. Remote Sens. Environ. 2011, 115, 2320−2329. (13) Small, C.; Pozzi, F.; Elvidge, C. D. Spatial analysis of global urban extent from DMSP-OLS nighttime lights. Remote Sens. Environ. 2005, 96, 277−291. (14) Small, C.; Elvidge, C. D.; Balk, D.; Montgomery, M. Spatial scaling of stable night lights. Remote Sens. Environ. 2011, 115, 269− 280. (15) Huang, Qingxu; He, C.; Gao, B.; Yang, Y.; Liu, Z.; Zhao, Y.; Duo, Y. Detecting the 20 year city-size dynamics in China with a rank clock approach and DMSP/OLS nighttime data. Landscape and Urban Planning. 2015, 137, 138−148. 10509

DOI: 10.1021/acs.est.5b02471 Environ. Sci. Technol. 2015, 49, 10503−10509