PE&RS December 2015 - page 935

Nighttime-Lights-Derived Fossil Fuel Carbon
Dioxide Emission Maps and Their Limitations
Naizhuo Zhao, Eric L. Samson, and Nathan A. Currit
Abstract
Brightness of nighttime lights has been used as an indica-
tor for spatial disaggregation of CO
2
emission based on an
assumed linear relationship between the digital number (DN)
values of nighttime light imagery and the amount of CO
2
emissions. However, reliability of the linear relationship of
these variables has not been thoroughly examined. In this
study we find that the actual overall correlations are expo-
nential rather than linear. More specific analyses showed that
the DN values of nighttime light imagery first behaves linearly
(from 3 to 50) and then exponentially (from 51 to 63), corre-
lating to the amount of CO
2
emissions. Regardless of the use
of a linear or piecewise function, some featured limitations
are evident as we developed the methodology. Among signifi-
cant limitations, CO
2
emissions were not visualized in unlit
areas and a lack of variation existed in regions with the same
DN values of nighttime light imagery. Lastly CO
2
emissions in
urban core areas were grossly under-estimated.
Introduction
Imagery from the Defense Meteorological Satellite Program’s
Operational Linescan System (
DMSP-OLS
) that was originally
designed to collect information about moonlit clouds (Doll,
2008) has been widely used to assess and map socioeconomic
activities (e.g., wealth production, electric power consump-
tion, and fossil fuel carbon dioxide emissions (hereafter
referred to as CO
2
emissions)) (Doll
et al.
, 2000; Doll
et al.
,
2006; Sutton
et al.
, 2007). Since the National Oceanic and
Atmospheric Administration’s (
NOAA
) National Geophysi-
cal Data Center (
NGDC
) released stable lights annual image
products in which ephemeral lights are removed (Baugh
et
al.
, 2010), quantitatively estimating socioeconomic data from
DMSP-OLS
has become more reliable and convenient. The
DM-
SP-OLS
nighttime light image data are a unique remote sensing
dataset to monitor global human activities at a relatively high
spatial resolution (Chen and Nordhaus, 2010).
Spatially disaggregating socioeconomic factors is an im-
portant application of nighttime light imagery. Since Doll
et
al.
(2000) first used nighttime light image data to disaggregate
gross domestic product (
GDP
) and CO
2
emissions, global and
continental
GDP
and CO
2
emissions maps with relatively fine
spatial resolution (from 1 km × 1 km to 1° × 1°) have been
efficiently produced based on linear relationships between
the socioeconomic factors and brightness of nighttime lights
(or logarithm of the socioeconomic factors and logarithm of
brightness of nighttime lights) at the national or the sub-na-
tional scales (Doll
et al.
, 2006; Sutton and Costanza, 2002). In
recent years demographic data (e.g., gross domestic production
(
GDP
), electric power consumption, and CO
2
emissions) re-
ported at the national or province/state level were distributed
to each pixel in proportion to the
DN
value of the pixel of the
nighttime light images and consequent spatial analyses of the
socioeconomic factors were accomplished at much finer scales
(Ghosh
et al.
, 2010a and 2010b; Oda and Maksyutov, 2011;
Zhao
et al.
, 2011; Zhao
et al.
, 2012). Such studies were per-
formed dependent on assumed linear relationships between
digital number (
DN
) values of the nighttime light images and
socioeconomic factors at the pixel level. However, accuracy
of the linear relationship at the pixel level was never tested.
Although linear relationships between
DN
values of nighttime
light images and the socioeconomic factors do exist at the
national and the regional levels (Doll
et al.
, 2000; Lo, 2002;
Sutton
et al.
, 2007; Zhao
et al.
, 2012), it cannot be stated that
such linear relationships still exist at the pixel level due to the
Modifiable Areal Unit Problem that a statistical result obtained
at a spatial scale may vary at a different spatial scale (Wrigley
et al.
, 1996). The main reason for lack of tests for veracity of
linear relationships is the nonexistence of other dependable
gridded data with sufficiently fine spatial resolution that can
be used as baseline reference data (Zhao
et al.
, 2012).
In this study we examined relationships between bright-
ness of nighttime lights and the amount of CO
2
emissions
in the contiguous US. Carbon dioxide concentration in the
atmosphere is widely believed to be one of the main reasons
for global climate change (Allen
et al
, 2009; Cox
et al.
, 2000;
Solomon
et al.
, 2009) so mapping CO
2
emissions can greatly
contribute to management of CO
2
emissions and prediction of
global climate change. More importantly, the Vulcan US fossil
fuel CO
2
emission inventory can be used as baseline reference
data in testing remote sensing conclusions. The Vulcan US
fossil fuel CO
2
emission inventory is established by eight indi-
vidual industry/use sectors. Data of the sectors reported at the
county level or collected at a spatial resolution of 1° × 1° are
downscaled to 10 km × 10 km/0.1° × 0.1° spatial resolution
based on social and geographic attributions using
GIS
tools
(Gurney
et al.
, 2009). So far, the Vulcan CO
2
emission inven-
tory may be the most reliable dataset showing carbon dioxide
emissions in the US considering data accuracy and spatial
resolution (Ghosh
et al.
, 2010a).
The main objectives of this study are to: (a) explore actual
functional relationship forms between brightness of nighttime
lights (
DN
values of stable light annual image products) and
the amount of CO
2
emissions, and (b) analyze errors/limita-
tions generated when the brightness of nighttime lights are
solely used as a proxy to map CO
2
emissions. To fulfill these
objectives we first produced a CO
2
emissions map for the con-
tiguous US using a linear relationship between
DN
values of
the nighttime light images and the amount of CO
2
emissions.
Naizhuo Zhao is with the Department of Geosciences, Texas
Tech University, 2500 Broadway, Lubbock, Texas 79409 (zhao.
).
Eric L. Samson is with the Mayan Esteem Project, 222 Main
Street, Suite 204, Farmington, Connecticut 06032.
Nathan A. Currit is with the Department of Geography, Texas
State University, 601 University Drive, San Marcos, Texas 78666.
Photogrammetric Engineering & Remote Sensing
Vol. 81, No. 12, December 2015, pp. 935–943.
0099-1112/15/935–943
© 2015 American Society for Photogrammetry
and Remote Sensing
doi: 10.14358/PERS.81.12.935
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
December 2015
935
879...,925,926,927,928,929,930,931,932,933,934 936,937,938,939,940,941,942,943,944,945,...946
Powered by FlippingBook