Elsevier

Remote Sensing of Environment

Volume 190, 1 March 2017, Pages 366-382
Remote Sensing of Environment

A global analysis of factors controlling VIIRS nighttime light levels from densely populated areas

https://doi.org/10.1016/j.rse.2017.01.006Get rights and content

Highlights

  • We analysed VIIRS night lights of all urban areas globally.

  • Cities' brightness was explained by variables at both the country and city levels.

  • Seasonal changes in vegetation and snow cover affect cities' brightness.

Abstract

Remote sensing of nighttime lights has been shown as a good surrogate for estimating population and economic activity at national and sub-national scales, using DMSP satellites. However, few studies have examined the factors explaining differences in nighttime brightness of cities at a global scale. In this study, we derived quantitative estimates of nighttime lights with the new VIIRS sensor onboard the Suomi NPP satellite in January 2014 and in July 2014, with two variables: mean brightness and percent lit area. We performed a global analysis of all densely populated areas (n = 4153, mostly corresponding to metropolitan areas), which we defined using high spatial resolution Landscan population data. National GDP per capita was better in explaining nighttime brightness levels (0.60 < Rs < 0.70) than GDP density at a spatial resolution of 0.25° (0.25 < Rs < 0.43), or than a city-level measure of GDP per capita (in proportion to each city's fraction of the national population; 0.49 < Rs < 0.62). We found that in addition to GDP per capita, the nighttime brightness of densely populated areas was positively correlated with MODIS derived percent urban area (0.46 < Rs < 0.60), the density of the road network (0.51 < Rs < 0.67), and with latitude (0.31 < Rs < 0.42) at p < 0.001. NDVI values (representing vegetation cover) were found to be negatively correlated with cities' brightness in winter time (− 0.48 < Rs <  0.22), whereas snow cover (enhancing artificial light reflectance) was found to be positively correlated with cities' brightness in winter time (0.17 < Rs < 0.35). Overall, the generalized linear model we built was able to explain > 45% of the variability in cities' nighttime brightness, when both physical and socio-economic variables were included. Within the generalized linear model, the percent of national GDP derived from income (rents) from natural gas and oil, was also found as one of the statistically significant variables. Our findings show that cities' nighttime brightness can change with the seasons as a function of vegetation and snow cover, two variables affecting surface albedo. Explaining cities' nighttime brightness is therefore affected not only by country level factors (such as GDP), but also by the built environment and by climatic factors.

Introduction

Artificial nighttime lights present one of humanity's unique footprints that can be seen from space (Croft, 1978). Resulting light pollution has been shown to negatively impact the community of astronomers and our ability to observe the night sky (Cinzano et al., 2001). However, the negative effects that light pollution has on ecological systems and on our health, through changes in circadian exposure to light and changes in the wavelengths we are exposed to, might have more important and far-reaching consequences (Longcore and Rich, 2004, Falchi et al., 2011, Gaston et al., 2013). Light pollution and artificial lighting has been shown to vary greatly in space and in time, as a function of population and economic activity. However, most studies examining the factors explaining global spatial variability in lit areas were conducted at national and provincial levels using the DMSP/OLS sensor (e.g., Elvidge et al., 1997, Chen and Nordhaus, 2011, Wu et al., 2013, Keola et al., 2015). While offering the only globally available time series of nighttime lights imagery from 1992 onwards (Bennie et al., 2014a), DMSP imagery has various drawbacks as it is not calibrated, its spatial resolution is coarse, it contains overglow beyond urban boundaries and it is saturated in urban areas (Small et al., 2005, Doll, 2008). Temporal changes in cities' lights and the spatial characteristics of cities' nighttime brightness have been examined in several countries using DMSP data (e.g., Lo, 2002, Ma et al., 2012, Zhang and Seto, 2013). Most of the studies which used DMSP data for urban studies have used annual datasets, whereas daily and monthly datasets were used to identify more dynamic and time varying features, such as forest fires, wars and fishing vessels (Huang et al., 2014). New studies using DMSP datasets for quantifying urban patterns are continuously being published (e.g., Ma et al., 2015, Weidmann and Schutte, 2016), however, annual products of DMSP night lights data are no longer being produced, the last one available being that of 2013.

Recently, new studies have attempted using finer spatial resolution (≤ 1 m) nighttime imagery to examine the factors explaining spatial patterns of nighttime lights within cities (Kuechly et al., 2012, Hale et al., 2013, Levin et al., 2014, Katz and Levin, 2016). Astronaut photography taken from the International Space Station presents an additional source of information about spatial patterns of cities at nighttime (de Miguel et al., 2014, de Miguel, 2015). Levin and Duke (2012) have used ISS imagery showing that not all towns and cities are equally lit, and that economic, infrastructure and demographic factors can explain differences in brightness levels of localities in Israel and the West Bank. Kyba et al. (2014) have used VIIRS DNB data to study the relationship between population size and the sum of lights from cities and communities in the USA and Germany, finding differences in light emission between cities of these two countries, and several recent studies have used VIIRS data to examine the nighttime brightness of cities in China (Ma et al., 2014a, Ma et al., 2014b, Shi et al., 2014a, Shi et al., 2014b) and in the USA (Chen et al., 2015). In addition, Elvidge et al. (2016) have used VIIRS data to detect and measure radiant emissions from gas flares globally, forming one of the major industrial sources of light pollution, which can even be detected night-time images of Landsat 8 in the visible bands (Levin and Phinn, 2016).

Urban areas are of high importance as most of the world's population resides in cities, with 78% of global carbon emissions attributed to cities (Grimm et al., 2008). In this paper our aim was to use the new monthly global cloud-free mosaics from the VIIRS sensor onboard the Suomi-NPP (launched in 2011), to examine the factors explaining spatial variability in nighttime lights at the city level, comparing densely populated areas (mostly urban areas) globally. We hypothesized that urban form and urban density (and other factors including percent urban area, NDVI, snow cover etc.) will also affect brightness levels, and not just socio-economic factors such as national GDP and population size. In addition, we aimed to examine the difference between using lit areas (i.e., areas above a certain threshold of nighttime lights brightness, as usually done in studies using DMSP data) and using calibrated brightness levels in radiance values, on the resulting factors explaining inter-city variability in nighttime lights.

Section snippets

Methods

The Visible/Infrared Imager/Radiometer Suite (VIIRS) was launched in October 28, 2011, collecting high quality nighttime images at a spatial resolution of 750 m in the Day/Night Bands (DNB), between 500 and 900 nm (Miller et al., 2012, Miller et al., 2013). Recent studies have shown the improved quality of VIIRS nighttime lights images over those acquired by the DMSP/OLS sensor (Elvidge et al., 2013, Li et al., 2013, Miller et al., 2013, Shi et al., 2014a, Shi et al., 2014b). There are now

City level

Altogether, we identified 4153 populated areas globally, mostly corresponding to cities and metropolitan areas (Fig. 1; see Supplementary KML file for the polygons of all cities). Their median area was 29.3 km2 (with a maximum of 3927 km2, for Jakarta, Indonesia), their median population being 172,000 (with a maximum of 30.4 million people for Tokyo, Japan), the median population density being 5476 people/km2 (with a maximum of 39,605 people/km2 for Hong Kong), and the median brightness of these

Discussion

Overall, our global mapping identified 4154 densely populated areas, 13.9% more than the 3646 metropolitan urban areas identified by Angel et al. (2011) who used MODIS derived urban land cover and population data. Previous global studies which analyzed differences in nighttime light brightness at the country or state level often focused on four main variables: population size, urban area, GDP and electric power consumption (e.g., Elvidge et al., 1997, Elvidge et al., 1999, Small et al., 2005,

Conclusions

Nighttime light remote sensing is still in its infancy stage and is basically qualitative, compared with daytime optical remote sensing and microwave remote sensing. There is still a lack of understanding of the mechanisms behind nighttime light remote sensing, due to the lack of studies at the ground level and the relative lack of understanding of nighttime light transfer from lighting sources through the air to the sensor. To advance nighttime light remote sensing, there is an urgent need for

Acknowledgments

The work of Qingling Zhang was supported in part by the One Hundred Talents Program of the Chinese Academy of Science ([2015], ARP: Y674141) and the Fundamental Research Program of Shenzhen S&T Innovation Committee ([2016], JCYJ20160429191303198). We thank the anonymous reviewers and Editor whose suggestions contributed to the manuscript.

References (76)

  • N. Levin et al.

    Illuminating the capabilities of Landsat 8 for mapping night lights

    Remote Sens. Environ.

    (2016)
  • N. Levin et al.

    A new source for high spatial resolution night time images - the EROS-B commercial satellite

    Remote Sens. Environ.

    (2014)
  • T. Ma et al.

    Quantitative estimation of urbanization dynamics using time series of DMSP/OLS nighttime light data: A comparative case study from China's cities

    Remote Sens. Environ.

    (2012)
  • T. Ma et al.

    Night-time light derived estimation of spatio-temporal characteristics of urbanization dynamics using DMSP/OLS satellite data

    Remote Sens. Environ.

    (2015)
  • M. Mehrara

    Energy consumption and economic growth: The case of oil exporting countries

    Energy Policy

    (2007)
  • D. Reiche

    Energy Policies of Gulf Cooperation Council (GCC) countries—possibilities and limitations of ecological modernization in rentier states

    Energy Policy

    (2010)
  • C. Small et al.

    Spatial analysis of global urban extent from DMSP-OLS night lights

    Remote Sens. Environ.

    (2005)
  • J. Verbesselt et al.

    Detecting trend and seasonal changes in satellite image time series

    Remote Sens. Environ.

    (2010)
  • J. Wu et al.

    Exploring factors affecting the relationship between light consumption and GDP based on DMSP/OLS nighttime satellite imagery

    Remote Sens. Environ.

    (2013)
  • K. Baugh et al.

    Nighttime Lights Compositing Using the VIIRS Day-Night Band: Preliminary Results

    (2013)
  • J. Bennie et al.

    Contrasting trends in light pollution across Europe based on satellite observed night time lights

    Sci. Rep.

    (2014)
  • J. Bennie et al.

    Mapping artificial lightscapes for ecological studies

    Methods Ecol. Evol.

    (2014)
  • B. Bhaduri et al.

    LandScan: Locating people is what matters

    Geoinformatics

    (2002)
  • N. Bharti et al.

    Explaining seasonal fluctuations of measles in Niger using nighttime lights imagery

    Science

    (2011)
  • F. Bickenbach et al.

    Night Lights and Regional GDP (No. 1888)

    (2013)
  • A. Bitan et al.

    The distribution of snow in Israel

    GeoJournal

    (1978)
  • K.W. Chan et al.

    Urbanization in China in the 1990s: new definition, different series, and revised trends

    China Rev.

    (2003)
  • X. Chen et al.

    Using luminosity data as a proxy for economic statistics

    Proc. Natl. Acad. Sci.

    (2011)
  • Z. Chen et al.

    Estimating house vacancy rate in metropolitan areas using NPP-VIIRS nighttime light composite data

    IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens.

    (2015)
  • P. Cinzano et al.

    The first world atlas of the artificial night sky brightness

    Mon. Not. R. Astron. Soc.

    (2001)
  • T. Croft

    Nighttime images of the earth from space

    Sci. Am.

    (1978)
  • K. Didan

    MOD13C2 MODIS/Terra Vegetation Indices Monthly L3 Global 0.05Deg CMG V006. NASA EOSDIS Land Processes DAAC

    (2015)
  • C.N. Doll

    CIESIN Thematic Guide to Night-Time Light Remote Sensing and Its Applications

    (2008)
  • C. Elvidge et al.

    Relation between satellite observed visible-near infrared emissions, population, economic activity and electric power consumption

    Int. J. Remote Sens.

    (1997)
  • C.D. Elvidge et al.

    Spectral identification of lighting type and character

    Sensors

    (2010)
  • C.D. Elvidge et al.

    Why VIIRS Data Are Superior to DMSP for Mapping Nighttime Lights

    (2013)
  • C.D. Elvidge et al.

    Methods for global survey of natural gas flaring from visible infrared imaging radiometer suite data

    Energies

    (2016)
  • R.L. Forstall et al.

    Which are the largest? Why lists of major urban areas vary so greatly

    Tijdschr. Econ. Soc. Geogr.

    (2009)
  • Cited by (152)

    • Tree removal regulation mitigates light pollution in suburban communities

      2024, Journal of Environmental Planning and Management
    View all citing articles on Scopus
    View full text