Journal of Cleaner Production 328 (2021) 129488

Contents lists available at ScienceDirect

Journal of Cleaner Production

journal homepage: www.elsevier.com/locate/jclepro

An improved approach for monitoring urban built-up areas by combining

NPP-VIIRS nighttime light, NDVI, NDWI, and NDBI
Yuanmao Zheng a, b, Lina Tang a, Haowei Wang a, *
a
Key Lab of Urban Environment and Health, Institute of Urban Environment, Chinese Academy of Sciences, Xiamen, 361021, China
b
University of Chinese Academy of Sciences, Beijing, 100049, China

A R T I C L E I N F O A B S T R A C T

Handling editor: Zhifu Mi Timely and accurate extraction of urban built-up areas is crucial to addressing environmental problems related to
fast changes in urban land cover, which is fundamental for optimizing land use patterns and supporting global
Keywords: sustainable development. Nighttime light (NTL) from the Suomi National Polar-orbiting Partnership Visible
NPP-VIIRS nighttime light Infrared Imaging Radiometer Suite (NPP-VIIRS) offer a new data source for extracting urban information.
Landsat multispectral image
However, this kind of data suffer from drawbacks of blooming effects. To address this problem, in this study, the
Urban built-up areas
Enhanced Nighttime Light Urban Index (ENUI) approach, which involves the combination of NPP-VIIRS NTL
Multi-temporal monitoring
Blooming effects with the Normalized Difference Vegetation Index (NDVI), Normalized Difference Water Index (NDWI), and
Normalized Difference Built-up Index (NDBI), is proposed and tested. This approach was used to rapidly monitor
the urban built-up areas in the Guangdong-Hong Kong-Macao Greater Bay Area in 2012, 2015, and 2018. The
average overall accuracy and Map-level Image Classification Efficacy (MICE) for the extraction results are
93.56% and 0.77, respectively, while those of the Local-Optimized Thresholding (LOT) are 86.48% and 0.54,
respectively; meanwhile, the average F-score values, user’s accuracy and producer’s accuracy for urban areas
using the proposed approach increased by 9.98%, 10.90% and 8.67%, respectively, compared with the LOT.
These findings suggest that this approach has a higher extraction accuracy than the LOT; this is primarily
ascribed to the integration of NTL data with the NDVI, NDWI, and NDBI, which increases the variability of
nighttime light in the urban core area and adequately alleviates the blooming effects of nighttime light brightness
in water bodies and vegetated areas. The proposed approach shows great potentials to accurately and effectively
monitor multi-temporal urban information and address environmental issues using NPP-VIIRS NTL data in global
urban agglomerations.

1. Introduction land (d’Amour et al., 2017; Liu et al., 2019), elevated ecological risks
(Tang et al., 2018), air and water pollution (McGrane, 2016), hydro­
The earth has become an urban planet (Wigginton et al., 2016). The logical and ecological environmental damage (Brunori et al., 2017),
core of urbanization is motivated by the search for a better life (Stokes urban heat-island effects (Meng et al., 2018), and increased energy use
and Seto, 2019). Although urban areas make up only a small fraction of (Liu et al., 2018). Synoptically, this rapid urbanization has led to the
the total global land area, they host more than 54% of its population, depletion of resources and multiple environmental problems and di­
furthermore, the urban population is expected to rise by over 2 billion by sasters, exacerbated the problems of water scarcity, air pollution, and
2050 (United Nations, 2014). The surging increase of the urban popu­ urban heat-island effects, and imposed strong pressure on global and
lation has been accompanied by a sharp increase in urban built-up areas regional sustainable development (Tang and Ma, 2018; Xian et al., 2019;
(Wigginton et al., 2016; Seto et al., 2011; Kuang et al., 2016). While He et al., 2017). Urban built-up areas expansion have far-reaching im­
population migration and urbanization is helping to lift hundreds of pacts on human society and living environments all over the world.
millions of people out of poverty, it also has serious environmental Therefore, timely and accurate monitoring of urban built-up areas is
consequences and presents immense societal challenges, including essential for research on land cover evolution and climate change and is
increasing greenhouse gas emissions (Ban et al., 2015), loss of arable also the basis for optimizing land use patterns and supporting

* Corresponding author.
E-mail addresses: yuanmaozheng@iue.ac.cn (Y. Zheng), lntang@iue.ac.cn (L. Tang), hwwang@iue.ac.cn (H. Wang).

https://doi.org/10.1016/j.jclepro.2021.129488
Received 19 August 2021; Received in revised form 8 October 2021; Accepted 23 October 2021
Available online 27 October 2021
0959-6526/© 2021 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/4.0/).
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

sustainable urban development, management, and planning (Schneider cities in China in 2012 by using a threshold method; Goldblatt et al.
et al., 2010; Ban et al., 2015; Feng et al., 2019). To avoid confusion, (2018) extracted the urban land cover in India, Mexico, and the United
herein, the term “urban built-up areas” refers to impervious surfaces, i. States; Sharma et al. (2016) determined the global urban built-up area in
e., man-made covering and constructions (Liu et al., 2018). 2014; Chen et al. (2017) identified 33 core urban areas in Shanghai and
In response to the increasing requirement for monitoring urban built- extracted a large amount of urban surface details, thus providing highly
up areas, since 2000, remote sensing data has been employed to produce accurate data sources for the remote sensing monitoring of urban land;
urban expansion maps (Schneider et al., 2010; Liu et al., 2018; Stokes while Li and Chen (2018) selected five urban agglomerations in China
and Seto, 2019). Since such data are suitable for performing contem­ from 2012 to 2017 and evaluated the urban spatiotemporal develop­
poraneous observations over a large area and obtaining up-to-date and ment mode using NPP-VIIRS NTL and the Genetic Algorithm-based
periodic data. The data sources of these maps mainly include medium- urban cluster automatic threshold (GA-UCAT). Additionally, Yu et al.
and high-resolution remote sensing images, such as Landsat series im­ (2018) took six major cities and 32 other cities in the United States and
ages (Tourea et al., 2018; Goldblatt et al., 2016). Landsat data have been demonstrated that preprocessing via logarithmic transformation could
employed to map urban growth in various regions, such as Bagan and remarkably increase the ability to use NPP-VIIRS NTL composite data for
Yamagata (2012) for Tokyo, Li et al. (2015) for Beijing, Tourea et al. extracting urban built-up areas. However, it remains challenging to
(2018) for the Greater Accra region in Ghana, and Goldblatt et al. (2016) determine urban built-up areas by NPP-VIIRS NTL data alone since these
for India. In particular, Michishita et al. (2012) extracted urban built-up data are subject to blooming effects. Such effects arise from incoherent
areas in land cover using a Landsat-5 Thematic Mapper image. Addi­ light radiating from the light source in all directions; for instance, the
tionally, based on Landsat image, Sexton et al. (2013) and Song et al. dispersion of light brightness to surrounding areas (Baugh et al., 2013)
(2016) extracted multi-temporal urban land characteristics, which can leads to blooming effects of nighttime light brightness, which causes the
more accurately determine the boundary between urban areas and the overestimation of urban built-up areas (He et al., 2014).
surrounding environment. The blooming effects of NPP-VIIRS NTL data—especially in water
However, these medium-and high-resolution data are cannot be bodies and vegetated regions of urban built-up areas—hinder the wide
easily used to map the rapid expansion of large urban agglomerations. use of NPP-VIIRS data for accurately extracting urban land information.
Approaches based on medium- and high-resolution remote sensing im­ Therefore, in this study, a new improved approach for extracting urban
ages are labor-intensive and are therefore unsuitable for the inversion of built-up areas, namely the Enhanced Nighttime Light Urban Index
changes for urban built-up areas in the large-scale or global urban ag­ (ENUI), is proposed and tested. Our intention is to diminish the
glomerations (Liu et al., 2012; Schneider et al., 2010), it is difficult to blooming effects by incorporating optical band derivatives, namely
reflect the multi-temporal monitoring in large-scale urban built-up areas vegetation indices, water indices and building indices. It is hoped that
in a short period, especially in rapidly urbanizing agglomerations such the proposed approach can provide a new method and reference for the
as the Guangdong-Hong Kong-Macao Greater Bay Area (GBA) (Ma et al., extraction and change monitoring of urban built-up areas in the large-
2019). Therefore, it is urgently needed to adopt new approaches to scale urban agglomerations around the world.
extract and quantify urban built-up areas to monitor urban extensions
and evolution in multiple periods. 2. Data
Recently, nighttime light (NTL) from the Suomi National Polar-
orbiting Partnership Visible Infrared Imaging Radiometer Suite (NPP- 2.1. Research area
VIIRS) have emerged as a valuable data to effectively determine urban
expansion patterns and monitor urban built-up areas (Bagan and The GBA was selected as the research area since it has experienced
Yamagata, 2015; Small and Elvidge, 2013; Ouyang et al., 2019). It is unprecedented urbanization over the past 40 years, accompanied by
particularly significant that NTL based method can be applied in areas rapid urban expansion and economic growth. The GBA urban agglom­
which are closely associated with urban development and can be used eration is a megalopolis located in a bay/delta area in Southern China
for purposes including the estimation of population distribution (Levin and has among the highest population densities and economic levels of
and Duke, 2012; Yang et al., 2013) and carbon emissions (Meng et al., any region on earth (Ma et al., 2019). The GBA comprises 11 cities
2014; Shi et al., 2018), urban land extraction (Goldblatt et al., 2018; Ma (Fig. 1) and had a resident population of around 71 million in 2018. This
et al., 2012), urbanization monitoring (Pandey et al., 2013; Zhou et al., area has long been the “vanguard” of China’s urban development.
2014), and the analysis of the evolution of urban agglomerations (Wei Although the GBA takes up less than 1% of China’s land area, it
et al., 2014; Zhang and Seto, 2013). Therefore, NTL data can provide accounted for 12.17% of the national GDP in 2017.
alternative or supplementary methods for drawing urban maps at the The GBA is located at low latitudes (21–25◦ N) and has a subtropical
regional and global scale (Elvidge et al., 2007; Lu et al., 2008). monsoon climate that is warm and humid all year round. Most of the
NPP-VIIRS NTL data can be used to highly efficiently detect low-level area is below 200 m a.s.l., with the Pearl River Delta plain in the center,
nighttime light with a high radiation resolution, wide coverage with hilly and mountainous areas in the north, east, and west, and sea in the
3000 km, etc., which make them more suitable for extracting urban land south. Therefore, the coastal and central regions are relatively flat.
areas (Cole et al., 2017; Yu et al., 2015). NPP-VIIRS NTL data have high
resolution and can be used to detect the nighttime light within the ra­ 2.2. Data sources
diation range from 3 × 10− 9 to 0.02 W cm− 2.sr− 1 across a geographical
area which spans latitudes between 75◦ S and 65◦ N with a width of 3000 (1) NPP-VIIRS NTL data. The NPP-VIIRS NTL remote sensing data
km and a return cycle of 12 h (Elvidge et al., 2013). Additionally, employed in this research were acquired from the National
NPP-VIIRS NTL data are of much better quality than the NTL data from Oceanic and Atmospheric Administration National Geophysical
the Defense Meteorological Satellite Program-Operational Linescan Data Center (NOAA NGDC; http://ngdc.noaa.gov/eog/downlo
System (DMSP-OLS) (Elvidge et al., 2013; Schueler et al., 2013); the ad.html). The datasets were the synthesized data from 2012,
spatial resolution is increased from 1 km to 500 m and the digital 2015, and 2018 and were re-sampled by bilinear fitting to a
number (DN) is changed from the relative low light intensity value with spatial resolution of 30 m.
6-bit quantization limit to the calibration radiance value with 14-bit (2) Normalized Difference Vegetation Index (NDVI), Normalized
quantization, which effectively solves the problem of saturation Difference Water Index (NDWI), and Normalized Difference
(Elvidge et al., 2013). Researchers have used NPP-VIIRS NTL to deter­ Built-up Index (NDBI). The corresponding NDVI, NDWI, and
mine the urban land area at the urban, national, and global scales. For NDBI were calculated using Landsat-8 OLI image spectra with a
instance, Shi et al. (2014) quantified the urban land area in 12 large spatial resolution of 30 m from 2012, 2015, and 2018. The data

2
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

Fig. 1. The location of the Guangdong-Hong Kong-Macao Greater Bay Area (GBA).

were acquired from the International Scientific Data Service 2003). The calculation formula of the ENUI is as follows:
Platform, Computer Network Information Center, Chinese
ENUI = NTL × (1 − NDVI) × (1 − NDWIB ) × NDBIB (2)
Academy of Sciences (CAS).
(3) Land use/cover (LUC) type datasets. To confirm the extraction NIR − RED
accuracy, the built-up areas extracted using the proposed NDVI = (3)
NIR + RED
approach were verified using LUC datasets with a 30-m spatial
resolution for 2012, 2015, and 2018. The LUC datasets were G − NIR
NDWI = (4)
obtained from the Resource and Environment Data Cloud Plat­ G + NIR
form of the CAS (http://www.resdc.cn/).
(4) Administrative division map. The administrative boundary vec­ NDBI =
MIR − NIR
(5)
tor of the study area was obtained from the administrative MIR + NIR
boundary data (1:4,000,000) of prefecture-level cities and
where NIR is the near-infrared band, RED is the red band, G is the green
counties released by the Chinese National Basic Geographic In­
band, MIR is the mid-infrared band, and NTL is the NPP-VIIRS NTL data
formation Center.
for the study area.
The NDVI was calculated from Landsat-8 OLI data. The value of the
3. Methods
NDVI ranges between − 1 and 1. Its values are positive for vegeta­
tion—including crops, shrubs, grasses, and forests, and are close to zero
3.1. Extraction methods of built-up areas
or negative for non-vegetated areas—including rocks, sand, and con­
crete surfaces (Pettorelli et al., 2005; Jones and Vaughan, 2010). Thus,
3.1.1. Local-Optimized Thresholding (LOT)
larger NDVI values indicate more vigorous green vegetation cover
According to the method proposed by Cao et al. (2009) and Liu et al.
(Huang et al., 2020). Although the NDVI is affected by scattering caused
(2012), the local-optimized threshold was determined according to the
by atmospheric effects, NDVI values obtained with accurate data pro­
threshold that is identified when the built-up areas determined using the
cessing methods are valuable (Huang et al., 2020). Advanced satellite
NPP-VIIRS data alone was most similar to the reference data in the
systems have been thoroughly calibrated to obtain accurate values of the
spatial range. Then, according to research by Shao et al. (2021), the
NDVI (Huang et al., 2020). The expression (1− NDVI) indicates that the
Map-level Image Classification Efficacy (MICE) are useful for comparing
greater non-vegetative weights to the urban core area rather than to
classification methods that are tested with different images.
urban peripheral areas, resulting in increased variability of data values
[ ]
Maximize MICEi = f (Ti ), Ti ∈ VIIRSimin , VIIRSimax (1) in the urban core area (Zhang et al., 2013). Therefore, the (1− NDVI)
value of urban core areas is close to 1 while that of the non-urban areas
where Ti is the threshold of the NTL in city i; MICEi was obtained by rich in vegetation is close to 0. The combination of (1− NDVI) and NTL
calculating the extracted built-up areas and the reference data; the built- can increase the speed of identification of the change characteristics in
up-area extraction was based on the threshold Ti using VIIRS NTL data; the urban core area.
and VIIRSmin
i and VIIRSmax
i are the minimum and maximum values of The value of the NDWI ranges from − 1 to 1. Theoretically, NDWI
VIIRS NTL data in city i, respectively. Finally, the regions where the values above 0 represent water bodies and NDWI values below 0 indi­
NPP-VIIRS data exceed the optimum threshold were determined as the cate non-water-body areas (McFeeters, 1996). However, in reality,
built-up areas in each city. water bodies can also have NDWI values < 0 due to the presence of bare
sediment in some rivers, lakes, and seas. The present work indicates that
3.1.2. Enhanced Nighttime Light Urban Index (ENUI) the threshold value of NDWI for water bodies is − 0.1. Therefore, in this
Following the research of Zhang et al. (2013) and Li et al. (2016), we study, an NDWI value of − 0.1 was taken as the threshold to binarize the
proposed the ENUI to improve the extraction of built-up areas using NDWI extraction result: that is, NDWI values > (− 0.1) were taken as 1
NPP-VIIRS NTL data and Landsat-8 OLI image spectra. Among these and NDWI values < (− 0.1) were taken as 0. Then, the binarization value
data, Landsat-8 OLI image spectra were used to calculate the NDVI, of the NDWI value was 1 and was set as NDWIB. The expression
NDWI, and NDBI (Pettorelli et al., 2005; McFeeters, 1996; Zha et al., (1− NDWIB) removes the water bodies. Thus, the combination of

3
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

(1− NDWIB) and NTL can remove the blooming effects of NTL brightness area” derived from LUC datasets, which are published by the CAS, were
to rivers and other water bodies. used as the reference data to verify the extraction results of LOT and
The value of NDBI ranges between − 1 and 1. Research suggests that ENUI. Therefore, the extraction results in the present research have
positive values of NDBI represent urban land areas and negative values credibility.
of NDBI represent non-urban land areas (Zha et al., 2003). The NDBI In studies performed by Shao et al. (2019) and Shao et al. (2021),
calculation results were binarized: positive NDBI values were given a only one index was used to verify the accuracy of map classification and
value of 1 and negative NDBI values were given a value of 0. The extraction; in particular, if only the Overall Accuracy (OA) of this single
binarization value of NDBI was 1 and was set as NDBIB. Because the index is used for verification, the accuracy evaluation will contain errors
common phenomenon involving the mixing of different types of ground and will be unreliable. Therefore, in this study, accuracy and reliability
objects in cities, which tends to result in a slight overestimation in the were evaluated using the MICE coefficient, OA, Producer’s Accuracy
urban land area extracted by NDBIB (Li and Chen, 2018); however, some (PA), User’s Accuracy (UA), and F-score. The MICE coefficient is used to
blooming effects of NTL can be reduced. The reasons for reduced measure consistency or accuracy and is a common index for measuring
blooming in vegetated and bare-soil areas are as follows: since the strong classification accuracy. The MICE value ranges between 0 and 1, and
light source of NTL in urban core areas radiates in all directions, there higher values suggest a higher reliability of the classification result.
are blooming effects in the vegetated and bare-soil areas outside urban The MICE, OA, PA, UA, and F-score were calculated according to the
land areas. Therefore, despite the fact that NDBIB contains small errors, confusion matrix table (Shao et al., 2019, 2021). The error matrix (Shao
the blooming effects of NTL in vegetated areas (greenbelts, etc.) and et al., 2019, 2021) is shown in Table 1, and the calculation formulas of
bare-soil areas outside urban land areas can be reduced by multiplying PA and UA are shown in the same table.
NDBIB and NTL images to find the intersection areas. According to the error matrix in Table 1, the MICE was calculated as
In summary, through the introduction of the NDVI, NDWI, and NDBI, follows:
the blooming effects of NTL brightness in water bodies and vegetated ∑n njj ∑n nj 2
areas outside urban land areas were eliminated, thus increasing the MICE =
j=1 n − j=1 ( n )
∑n nj 2 (7)
extraction accuracy for urban built-up areas. 1− j=1 n )
(
Following the method of Li et al. (2016) and Shao et al. (2021), the
ENUI was calculated to extract built-up areas. Specifically, the LUC where njj represents the proportion of correctly classified samples in
datasets were used as a reference to determine the optimal threshold of class j in the diagonal of the error matrix (Table 1), nj is a simplified
the ENUI. Since the spatial resolution of LUC datasets (30 m) is much presentation of n+j, which is a commonly applied to represent a refer­
higher than that of NPP-VIIRS NTL data (500 m), it is acceptable to use ence total, and n is the total number of land cover types.
LUC datasets to produce reference data for evaluation (Henderson et al., Then, according to the Table 1, the OA was expressed as:
2003; Cao et al., 2009; He et al., 2006). The selection of the optimal ∑n /
OA = njj n (8)
threshold for ENUI can be expressed as follows: j=1
[ ]
( ) Finally, the F-score was calculated as follows:
Maximize MICEj = f Tj , Tj ∈ ENUIjmin , ENUIjmax (6)
Fscore = 2 × (PA × UA) / (PA + UA) (9)
where Tj is the threshold of ENUI in city j and MICEj is the MICE coef­
ficient obtained by calculating the built-up areas extracted using ENUI 4. Results
and the reference data. Specifically, the built-up areas were extracted
using ENUI based on threshold Tj, and then the extracted built-up areas 4.1. Advantages of the ENUI
and reference data were used to obtain the MICE. ENUIjmin and
ENUIjmax are the minimum and maximum values of the ENUI in city j, In this study, the GBA was taken as the research area to verify
respectively. Finally, the ENUI values that are higher than the optimal whether the approach involving the integration of NPP-VIIRS NTL with
threshold were extracted into the built-up areas of each city. the NDVI, NDWI, and NDBI is the optimal choice for the extraction of
urban built-up areas (Fig. 2). A comparison with reference data with a
high resolution of 30 m (Fig. 2i) shows that the extraction results using
3.2. Evaluation of extraction accuracy the NDVI and NDBI overestimate the built-up areas (Fig. 2d and e) and
the results extracted using only NPP-VIIRS NTL data also overestimate
The reliability and accuracy of the reference data are key to the the built-up areas (Fig. 2g). Meanwhile, the water bodies areas extracted
success of the proposed approach. The LUC datasets obtained from the using the NDWI (Fig. 2f) are more consistent with the distribution of
manual interpretation of Landsat images released by the CAS have been water bodies in the reference data. Finally, the ENUI shows an excellent
proven to reflect the real LUC status of China (Liu et al., 2012). In this performance (Fig. 2h). Based on the results for 2015 (Fig. 2), it is
research, following the studies of He et al. (2014), the data of “urban

Table 1
The error matrix and various accuracy metrics.
Map data Reference data

j=1 j=2 … j=J Map total User’s accuracy

i=1 n11 n12 n1J n1+ n11/n1+

i=2 n21 n22 n2J n2+ n22/n2+
…
i=J nJ1 nJ2 nJJ nJ+ nJJ/nJ+
Reference total n+1 n+2 n+J n
Producer’s accuracy n11/n+1 n22/n+2 nJJ/n+J

Note: njj is the fraction of correctly classified samples of class j in the diagonal of the error matrix, nij is
the fraction of samples in row i and column j, n is the total number of land cover types, ni+ is the total
fraction of class i extracted from map data, and n+j is the total fraction of class j obtained from the
reference data.

4
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

Fig. 2. The results of different approaches for extracting built-up areas in the GBA in 2015. (a) values of the Normalized Difference Vegetation Index (NDVI). (b)
values of the Normalized Difference Built-up Index (NDBI). (c) values of the Normalized Difference Water Index (NDWI). (d) built-up areas extracted via the NDVI. (e)
built-up areas extracted via the NDBI. (f) water bodies extracted via the NDWI. (g) built-up areas extracted via only the NPP-VIIRS NTL data. (h) built-up areas
extracted via the Enhanced Nighttime Light Urban Index (ENUI). (i) reference data.

Fig. 3. Horizontal transects of a Landsat-8

OLI false-color image of Shenzhen, China,
from 2015. In the image, purple represents
the urban area, green represents the vege­
tated area, and dark blue represents the
water area. The pixel values of the NDVI,
NDBI, and NDWI are represented by the
green, red, and blue lines, respectively. Note:
The data in the figure are normalized to the
same visual range and the horizontal axis
represents the distance value of transect
pixels for images. (For interpretation of the
references to color in this figure legend, the
reader is referred to the Web version of this
article.)

concluded that, rather than being used separately, the four types of data to further illustrate the superior performance of the ENUI. Firstly, based
(i.e., NPP-VIIRS data and the NDVI, NDWI, and NDBI) can be effectively on the analysis of the NDVI, NDBI, and NDWI of Shenzhen, China, in
integrated using the ENUI since the spatial distribution of built-up areas 2015, as shown in Fig. 3, the results are derived from the urban char­
obtained with this approach is more consistent with the reference data. acteristics of NDBI with higher DN pixel values and NDVI with lower DN
As shown in Fig. 2, the ENUI was found to perform better than the pixel values in the core built-up areas, and there are NDWI values with
approach using NPP-VIIRS data alone. The following two aspects aimed higher pixel values in the water bodies areas. These results show that the

5
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

Fig. 4. Latitudinal transects of NPP-VIIRS and ENUI for Guangzhou City, China, in 2015. In the image, the pixel values of the NPP-VIIRS data and ENUI data are
represented by the blue and red lines, respectively. (a) Landsat-8 OLI false-color image. (b) NPP-VIIRS data with a 500-m spatial resolution. (c) ENUI data with a 30-
m spatial resolution. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

combination of NPP-VIIRS, NDVI, NDWI, and NDBI data can express the values obtained using the LOT were 86.22% and 0.52, respectively
urban characteristics and appearance from multiple aspects since these (Fig. 5 and Table 2). Over the three study years, the average OA obtained
data types are complementary. Additionally, the combination of NPP- using the ENUI is 7.08% higher than that obtained using the LOT, while
VIIRS, NDVI, NDWI, and NDBI data removes small noise from clouds the average MICE is 0.23 higher.
or other features. Therefore, the ENUI can better reflect the overall As shown in Fig. 6, the UA and PA values are all higher for the
characteristics of land use in urban built-up areas. extraction using ENUI than the extraction using LOT for both urban and
Furthermore, the ENUI values of the urban areas in Guangzhou City non-urban areas for all of the 11 cities. Furthermore, for almost all of the
in 2015, and the value of the normalized NPP-VIIRS data, were calcu­ cities, the increment of urban UA was larger than that of urban PA, while
lated. A latitudinal cross-section of the results is shown in Fig. 4. As the increment of non-urban PA was larger than that of non-urban UA.
shown in the figure, the ENUI values increase toward the urban core For instance, in the extraction results for Guangzhou City in 2012, the
areas and decrease toward the non-urban core areas. Meanwhile, the increments in urban UA and urban PA were 13.07% and 8.26%,
latitudinal transects demonstrate a high variability of ENUI values in the respectively (i.e., the increment in urban UA was larger); meanwhile,
urban core areas. More importantly, the ENUI can remove and alleviate the increments in non-urban PA and non-urban UA were 3.82% and
the blooming effects of NPP-VIIRS data, especially water bodies and 1.59%, respectively (i.e., the increment in non-urban PA was larger).
vegetated areas in the urban core areas. Furthermore, latitudinal tran­ Additionally, the average F-score for urban areas in the study area for
sects also show that NPP-VIIRS data maintain high values for non-urban extraction using the ENUI increased by 9.98% compared with extraction
land cover, while ENUI values for these areas are very low, that is, close using the LOT (Table 2).
to or equal to 0. These results show that the extraction accuracy of the proposed ENUI
In summary, the ENUI combining NPP-VIIRS NTL, NDVI, NDWI, and is highly superior to that of the LOT, demonstrating that, using the
NDBI can extract urban areas more accurately and better exhibit the former, the confusion between non-urban areas (e.g., water and vege­
differences between urban and suburban areas. Additionally, the pro­ tation) and urban areas is significantly reduced and the accuracy of
posed approach can effectively alleviate the blooming effects of NPP- classification and extraction is remarkably improved.
VIIRS NTL data, specifically in water bodies and vegetated areas in In summary, as shown in Table 2, the OA, MICE, PA, UA and F-score
urban built-up areas, and increased variability of nighttime light in for the extraction performed using the ENUI were higher than those for
urban core areas. Therefore, the ENUI shows great potentials for the extraction using the LOT in the research area. Moreover, compared
extensive application in urban studies due to its simplicity and accuracy. with the LOT, the urban UA increment obtained by the ENUI extraction
was greater than the urban PA increment, whereas the non-urban PA
4.2. Comparison of extraction precision increment obtained by the ENUI extraction was greater than the non-
urban UA increment. Furthermore, the accuracy of built-up-area
4.2.1. Statistical comparison of five precision indicators extraction was higher using the ENUI than using the LOT in all stud­
A comparison of the OA and MICE for the 11 cities in the research ied cities.
area in 2012, 2015, and 2018 is shown in Fig. 5. All the cities show
higher OA and MICE values in built-up areas extracted via ENUI than 4.2.2. Geographical comparison of representative cities
those extracted via LOT for all years, indicating that the ENUI has a The research area contains 11 cities with varying levels of socio-
higher precision. For instance, in 2012, the mean OA and mean MICE economic and urban development. According to the economic devel­
obtained with the ENUI were 92.92% and 0.75, respectively, whereas opment level, there are three major city types in the research area: the

6
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

Fig. 5. A comparison of the OA and MICE values for urban-built-up-area extraction using the LOT and the ENUI in (a) 2012, (b) 2015, and (c) 2018. “LOT OA” and
“ENUI OA” represent the OA values obtained for the extraction using LOT and ENUI, respectively; “LOT MICE” and “ENUI MICE” represent the MICE values obtained
for the extraction using LOT and ENUI, respectively; and “OA Add” and “MICE Add” represent the increment in OA and MICE, respectively, obtained with the ENUI
extraction compared with the LOT extraction.

Table 2
Comparison of the extraction accuracy of built-up areas in the research area for different years using the LOT and ENUI.
Category Extraction accuracy using LOT (%) Extraction accuracy using ENUI (%)

2012 2015 2018 Mean 2012 2015 2018 Mean

Urban PA 72.59 74.16 75.64 74.13 81.54 83.26 83.60 82.80

UA 62.05 63.40 64.30 63.25 72.98 74.69 74.77 74.15
F-score 66.91 68.36 69.51 68.26 77.02 78.74 78.94 78.24
Non-urban PA 87.30 87.98 87.26 87.52 92.81 93.38 93.04 93.08
UA 92.59 92.80 93.12 92.84 96.04 96.47 96.04 96.18
F-score 89.87 90.33 90.09 90.10 94.40 94.90 94.52 94.60
OA 86.22 86.51 86.7 86.48 92.92 93.79 93.98 93.56
MICE 52.40 54.28 55.93 54.20 74.68 78.17 78.37 77.07

Note: PA: Producer’s Accuracy; UA: User’s Accuracy; OA: Overall Accuracy.

first type (high economic level) includes Guangzhou, Shenzhen, and areas and were therefore more coherent with the verification data.
Hong Kong; the second type (medium economic level) comprises Foshan
and Dongguan; and the third type (low economic level) includes the 4.3. Assessment of the expansion of built-up areas in 2012–2018
other six of the 11 cities in the GBA. In this study, to verify the extraction
accuracy in different types, the representative cities of Guangzhou, The results showed that, between 2012 and 2018, the built-up areas
Foshan, and Zhuhai were selected for comparison. Using Landsat-8 OLI of the GBA spread from the core areas of the urban agglomerations in all
data as the verification data, the LOT and ENUI were both used to extract directions (Fig. 8) following a circular radial expansion pattern. The
the built-up areas in these three representative cities and the results built-up areas in the GBA are primarily distributed in the central areas of
were compared. A comparison of the built-up-area extraction for 2015 urban agglomerations and exhibit notable multi-center distribution
using these two approaches is shown in Fig. 7. The results show that the characteristics, that is, there are four areas with concentrated built-up
built-up areas of these representative cities with different types extrac­ areas, namely in Guangzhou, Shenzhen, Dongguan, and Foshan.
ted using the ENUI all did not include any water bodies or vegetated To further evaluate the expansion of built-up areas in the GBA

7
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

Fig. 6. A comparison between (a) Urban PA, (b) Urban UA, (c) Non-urban PA, and (d) Non-urban UA of the two extraction approachs used in this study for 2012,
2015, and 2018. Note: GZ, Guangzhou; SZ, Shenzhen; ZH, Zhuhai; FS, Foshan; JM, Jiangmen; ZQ, Zhaoqing; HZ, Huizhou; DG, Dongguan; ZS, Zhongshan; HK, Hong
Kong; MC, Macao. A1, 2012 LOT; A2, 2012 ENUI; B1, 2015 LOT; B2, 2015 ENUI; C1, 2018 LOT; C2, 2018 ENUI.

between 2012 and 2018, the degree of expansion were quantitatively Additionally, based on NPP-VIIRS data, He et al. (2014) used a method
analyzed. It was found that the built-up area in the GBA increased by involving the integration of NTL, the NDVI, and Land Surface Temper­
626.16 km2 during the study period; the coverage of built-up areas in the ature Support Vector Machine Classification (INNL-SVM) to extract
GBA changed from 11.43% in 2012 to 12.55% in 2018. These findings urban land for eight megacities in China in 2012, and found that the
are consistent with the prosperous and vibrant economic development mean MICE value was 0.70. The extraction accuracy obtained in the
and the increase of the agglomeration scale of the GBA over the study present research using the ENUI is higher than those obtained by Li and
period. Chen (2018) and He et al. (2014). This difference in performance can be
attributed to the fact that, in the results of these two previous studies,
5. Discussion blooming effects of NPP-VIIRS NTL data may exist, especially the
blooming effects of NTL brightness in water bodies and vegetated areas.
5.1. Comparison of extraction accuracy between the ENUI and other Our research results show that the ENUI can obtain the spatial distri­
methods bution of built-up areas more accurately than previous approaches,
which indicates that it has enormous potentials for the extraction and
5.1.1. The accuracy of the ENUI was superior to other methods inversion of urban built-up areas.
Through an accuracy assessment, this study verified the effectiveness
of the ENUI by comparing it with the reference data—namely LUC 5.1.2. Precise extraction and analysis
datasets obtained from the manual interpretation of Landsat image­ The ENUI has an advantage over existing methods in terms of the
s—which had been published by the CAS. In the average extraction re­ reduction of the blooming effects of NTL brightness in water bodies and
sults for built-up areas for 2012, 2015, and 2018, the OA was 93.56% vegetated areas in urban areas. Because of the blooming effects of NPP-
and the MICE was 0.77 for the ENUI extraction, while the OA was VIIRS NTL data (Baugh et al., 2013), using the LOT to obtain built-up
86.48% and the MICE was 0.54 for the LOT extraction. That is, the ENUI areas based on NTL data alone results in a notable overestimation of
was shown to be superior to the LOT in terms of extraction accuracy. the extent of built-up areas. However, the error caused by these
In other studies based on NPP-VIIRS data, Li and Chen (2018) used blooming effects in urban core areas can be remarkably reduced by using
the GA-UCAT method to extract the urban areas of Beijing, Shanghai, the ENUI. By conducting case studies for the cities of Guangzhou and
Guangzhou, and Changsha in China, and found that the OA was between Huizhou, the decrease of such error effects using ENUI was demon­
0.854 and 0.913 and the MICE was between 0.699 and 0.722. strated, as shown in Fig. 9.

8
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

Fig. 7. A comparison between the extracted urban area based on Landsat-8 OLI data, the LOT, and the ENUI for representative cities in 2015. “Accuracy by LOT” and
“Accuracy by ENUI” represent the accuracy comparisons between the results extracted by LOT and ENUI, respectively, compared to the verification data.

Fig. 8. The expansion of built-up areas in the GBA from 2012 to 2018 based on extraction using the ENUI.

Firstly, the NDVI was introduced into the ENUI; thus, the NDVI was and E shows the effects of reducing the light intensity in vegetated re­
combined with NPP-VIIRS NTL. Since the vegetation coverage of urban gions in the centers of Guangzhou and Huizhou. Additionally, the
areas is generally lower than that of non-urban areas (Cao et al., 2009) combination of NDVI and NTL data can increase the variability of light
and the NDVI is negatively correlated with the built-up areas, by per­ intensity values in the urban core areas and allow urban surface texture
forming this combination, the blooming phenomenon of NTL brightness features to be visualized more clearly.
in vegetated regions can also be alleviated, thus mitigating the influence Secondly, since water bodies can be accurately extracted using the
of nighttlime light brightness on the extraction. For example, Fig. 9 A, C, NDWI, the use of the NDWI in the ENUI can effectively remove the

9
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

Fig. 9. A comparison of the extraction results for Guangzhou and Huizhou, China, using different methods. Note
A: water bodies and vegetated area. B: water bodies. C: vegetated area. D: bare soil. E: vegetated area. F: water bodies.

blooming effects in water bodies areas caused by nighttime light shown in Fig. 9 D. Thus, the built-up areas can be better distinguished
brightness from urban centers, as shown in Fig. 9 B and F. from non-built-up areas through the combination of the NDBI and NTL,
Finally, since the NDBI is higher in urban areas than in non-urban thereby increasing the extraction accuracy for built-up areas.
areas, most built-up areas can be extracted using the NDBI. The use of Therefore, in the present study, the ENUI (which involves the com­
the NDBI in the ENUI can alleviate the blooming effects of nighttime bination of NTL data and the NDVI, NDWI, and NDBI) was adopted to
light in vegetated areas, as shown in Fig. 9 C and E. It is also found that increase the extraction accuracy for built-up areas. In previous research,
the combination of the NDBI and NTL can alleviate the blooming effects the variability of nighttime light in the urban core area was increased by
of nighttime light in bare-soil areas located next to built-up areas, as combining NTL data with the NDVI (Zhang et al., 2013), indicating the

10
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

the ground–sensor geometrical path and the obstruction of surface light

caused by the periodic presence of canopy leads to the reduction of NTL
intensity throughout the city (Levin et al., 2017).

5.3.2. Future prospects

Timely and reliable information about urban areas is urgently
needed to implement urban sustainable development and management
(Ban et al., 2015). In this study, nighttime light data and daytime land
remote sensing monitoring were applied to accurately extract built-up
areas and help clarify the important characteristics of urbanization. In
future research, higher-resolution NTL data should be obtained to
Fig. 10. The coefficients of determination (R2) between the extracted built-up perform quantitative research on the spatiotemporal evolution of
areas and the NTL threshold. Note: GZ, Guangzhou; SZ, Shenzhen; ZH, Zhuhai; highways, streets, and the increasing construction of buildings in
FS, Foshan; JM, Jiangmen; ZQ, Zhaoqing; HZ, Huizhou; DG, Dongguan; ZS, informal residential areas with insufficient infrastructure, which have
Zhongshan; HK, Hong Kong; MC, Macao. “Average ENUI” and “Average LOT” greatly significant research value and benefits (Stokes and Seto, 2019).
are the averages of the ENUI and LOT results for the three study years, Second, improved calibration methods for NTL data should be
respectively. designed in the future to obtain consistent and long-term time series of
NTL data using various platforms and sensors. Due to the differences in
feasibility of combining NTL data and vegetation data. Our findings acquisition time, onboard calibration, spatial resolution, etc., this re­
indicate that the use of multi-temporal NPP-VIIRS NTL and Landsat data mains a huge challenge (Levin et al., 2020; Zheng et al., 2020). For
have enormous potentials for the large-scale monitoring of urban instance, according to Hu et al. (2018) and Ryan et al. (2019), ground
built-up areas. In future studies, integrating population data, surface stabilization and radiometric calibration light sources may provide an
temperature data, or topographic and geomorphic data could be used to effective method for mutual calibrations between DMSP-OLS and
construct a more accurate urban monitoring approach (Zhao et al., NPP-VIIRS NTL sensors. Therefore, the fusion of DMSP-OLS and
2016). NPP-VIIRS or Luojia 1-01 NTL data will be performed to obtain a longer
time series of high-resolution data in future, which are essential for
5.2. Relationships between light threshold and built-up areas solving the environmental problems arising from long-term rapid ur­
banization. Besides, in the near future, apart from observing natural
Due to the differences in the scale, geographical location, and eco­ factors such as urbanization information, researchers should consider
nomic level of the cities in the study area, the optimal thresholds of the combination of socioeconomic perspective, the process and risks of
nighttime light differ between cities—i.e., the correlations between the urbanization, and other relevant aspects in order to promote sustainable
extracted built-up areas and the light threshold are different in different urban development (Zhao et al., 2020).
cities. In this study, the R2 value was used to quantify the correlations
between the extracted built-up areas and the optimal nighttime light 6. Conclusion
thresholds. The R2 values obtained for the 11 studied cities using the
LOT and ENUI, respectively, are given in Fig. 10. In this work, in order to improve the extraction accuracy of urban
As shown in Fig. 10, the average R2 values are larger for the ENUI information, we proposed an improved approach to rapidly and accu­
extraction than for the LOT extraction in all 11 cities. Meanwhile, at the rately monitor the urban built-up areas in the GBA in 2012, 2015, and
urban-agglomeration scale, the average R2 was 0.416 for the ENUI and 2018. Additionally, multi-temporal data were applied to monitor urban
0.269 for the LOT; that is, there was a higher correlation between the built-up areas, which reduced the random error of single-temporal data,
NTL threshold and built-up areas using the ENUI than the LOT, which and these data were used to verify the accuracy of the proposed
further shows that the ENUI is superior for extracting built-up areas. approach for multiple periods and verify the stability of the approach.
From the results, the following conclusions are drawn:
5.3. Limitations
(1) The ENUI effectively reduces the blooming effects of nighttime
5.3.1. Limitations of data light brightness in water bodies and vegetated areas and increases
The most notable uncertainty associated with the use of NTL data is the variability of nighttime light data values in the urban core
that, compared with daytime remote sensing, the brightness of night­ area. Thus, the ENUI can extract built-up areas more accurately.
time scenes is significantly lower and the dynamic range is extremely (2) The extraction accuracy of built-up areas obtained using NPP-
large. Recently, Li et al. (2019) discovered that the satellite observation VIIRS data was significantly improved using the ENUI
angle in NTL images is a major factor in the radiance fluctuation. compared to a previous method, namely the LOT. The average OA
Therefore, the height of urban buildings should be considered to facil­ and MICE of the extraction achieved using the ENUI were 93.56%
itate the understanding of the relationship between the zenith angle of and 0.77, respectively, whereas the average OA and MICE using
satellite observation and NTL emission. the LOT were 86.48% and 0.54, respectively.
A second source of uncertainty is the influence of moonlight, clouds/ (3) The average F-score, UA, and PA of the urban-area extraction
aerosols, and seasonal variations in vegetation, which are primary using the ENUI increased by 9.98%, 10.90%, and 8.67%,
contributors to the errors in NTL data. The uncertainty of NTL mainly respectively, compared with those for the LOT extraction. This
includes environmental factors such as moonlight, clouds/aerosols, and further suggests that the ENUI achieved a significantly higher
surface reflectance (which hinder observation signals) as well as surface extraction accuracy for built-up areas than the LOT.
properties such as the seasonal variation of vegetation or snow (Román (4) Between 2012 and 2018, the built-up area in the GBA increased
et al., 2018; Zheng et al., 2019). Such factors significantly affect the by 626.16 km2 and expanded in a circular radial pattern. The
estimation of seasonal and long-term trends. Specifically, NTL radiation coverage of built-up areas in the GBA changed from 11.43% in
is affected by clouds/aerosols, which cause it to be affected (e.g., blur­ 2012 to 12.55% in 2018.
red) (Elvidge et al., 2017). Additionally, seasonal changes, such as those
caused by vegetation artifacts, may complicate the acquisition of NTL The proposed approach was used to accurately monitor urban built-
data from observation satellites as the canopy-level foliage occurs along up areas in the research area between 2012 and 2018. The presented

11
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

maps of built-up areas in the GBA are available from upon request from Chen, Z., Yu, B., Song, W., Liu, H., Wu, Q., Shi, K., Wu, J., 2017. A new approach for
detecting urban centers and their spatial structure with nighttime light remote
authors. Furthermore, it was shown that the proposed approach requires
sensing. IEEE Trans. Geosci. Rem. Sens. 55 (11), 6305–6319. https://doi.org/
less time and resources to extract large-scale urban information using 10.1109/TGRS.2017.2725917.
NPP-VIIRS NTL data. This newly proposed approach was successfully Cole, T.A., Wanik, D.W., Molthan, A.L., Román, M.O., Griffin, R.E., 2017. Synergistic use
used to monitor urban dynamics in the GBA at a spatial resolution of 30 of nighttime satellite data, electric utility infrastructure, and ambient population to
improve power outage detections in urban areas. Rem. Sens. 9 (3), 286. https://doi.
m, and the results can be employed to supplement socioeconomic sta­ org/10.3390/rs9030286.
tistical data, which currently lack spatial information. d’Amour, C.B., Reitsma, F., Baiocchi, G., Barthel, S., Güneralp, B., Erb, K.H., Haberl, H.,
In summary, the ENUI is a promising prospect in urban and envi­ Creutzig, F., Seto, K.C., 2017. Future urban land expansion and implications for
global croplands. Proc. Natl. Acad. Sci. Unit. States Am. 114 (34), 8939–8944.
ronmental research since it accurately monitor the urban information in https://doi.org/10.1073/pnas.1606036114.
large-scale urban agglomerations and adequately reduce the blooming Elvidge, C.D., Baugh, K., Zhizhin, M., Hsu, F.C., Ghosh, T., 2017. VIIRS night-time lights.
effects of NPP-VIIRS NTL. The proposed approach can be adapted to Int. J. Rem. Sens. 38 (21), 5860–5879. https://doi.org/10.1080/
01431161.2017.1342050.
further address the DMSP-OLS NTL data from 1992 to 2012. After cor­ Elvidge, C.D., Baugh, K.E., Zhizhin, M., Hsu, F.C., 2013. Why VIIRS data are superior to
recting multiple-source NTL data, the proposed approach can obtain DMSP for mapping nighttime lights. Proc. Asia-Pacific Adv. Netw. 35 (62), 62–69.
more longer time series of urbanization information, which are essential https://doi.org/10.7125/APAN.35.7.
Elvidge, C.D., Tuttle, B.T., Sutton, P.C., Baugh, K.E., Howard, A.T., Milesi, C., Bhaduri, B.
for addressing the environmental problems arising from long-term and L., Nemani, R., 2007. Global distribution and density of constructed impervious
rapid urbanization and strengthening global or regional sustainable surfaces. Sensors 7 (9), 1962–1979. https://doi.org/10.3390/s7091962.
urban planning and management. Feng, Y., Wang, X., Du, W., Liu, J., Li, Y., 2019. Spatiotemporal characteristics and
driving forces of urban sprawl in China during 2003-2017. J. Clean. Prod. 241,
118061. https://doi.org/10.1016/j.jclepro.2019.118061.
CRediT authorship contribution statement Goldblatt, R., Stuhlmacher, M.F., Tellman, B., Clinton, N., Hanson, G., Georgescu, M.,
Wang, C., Serrano-Candela, F., Khandelwal, A.K., Cheng, W., Balling Jr., R.C., 2018.
Using Landsat and nighttime lights for supervised pixel-based image classification of
Yuanmao Zheng: Conceptualization, Methodology, Software, Data urban land cover. Rem. Sens. Environ. 205, 253–275. https://doi.org/10.1016/j.
curation, Writing – original draft, Visualization, Validation. Lina Tang: rse.2017.11.026.
Data curation, Writing – original draft, Visualization, Validation. Hao­ Goldblatt, R., You, W., Hanson, G., Khandelwal, K.A., 2016. Detecting the boundaries of
urban areas in India: a dataset for pixel-based image classification in Google earth
wei Wang: Investigation, Writing – review & editing, All authors have
engine. Rem. Sens. 8 (8), 634. https://doi.org/10.3390/rs8080634.
read and approved this manuscript. He, C., Li, J., Zhang, X., Liu, Z., Zhang, D., 2017. Will rapid urban expansion in the
drylands of northern China continue: a scenario analysis based on the Land Use
Scenario Dynamics-urban model and the Shared Socioeconomic Pathways. J. Clean.
Declaration of competing interest Prod. 165, 57–69. https://doi.org/10.1016/j.jclepro.2017.07.018.
He, C., Liu, Z., Tian, J., Ma, Q., 2014. Urban expansion dynamics and natural habitat loss
in China: a multiscale landscape perspective. Global Change Biol. 20 (9),
The authors declare that they have no known competing financial 2886–2902. https://doi.org/10.1111/gcb.12553.
interests or personal relationships that could have appeared to influence He, C., Shi, P., Li, J., Chen, J., Pan, Y., Li, J., Zhuo, L., Ichinose, T., 2006. Restoring
the work reported in this paper. urbanization process in China in the 1990s by using non-radiance-calibrated DMSP/
OLS nighttime light imagery and statistical data. Chin. Sci. Bull. 51 (13), 1614–1620.
https://doi.org/10.1007/s11434-006-2006-3.
Acknowledgments Henderson, M., Yeh, E.T., Gong, P., Elvidge, C., Baugh, K., 2003. Validation of urban
boundaries derived from global night-time satellite imagery. Int. J. Rem. Sens. 24
(3), 595–609. https://doi.org/10.1080/01431160304982.
The authors sincerely thank Prof. Jingzhu Zhao at the Institute of Hu, S., Ma, S., Yan, W., Lu, W., Zhao, X., 2018. Feasibility of a specialized ground light
Urban Environment, Chinese Academy of Sciences, for the generous source for night-time low-light calibration. Int. J. Rem. Sens. 39 (8), 2543–2559.
comments and inspiration, which greatly improved the manuscript. The https://doi.org/10.1080/01431161.2018.1430915.
Huang, S., Tang, L., Hupy, J., Shao, G., 2020. A commentary review on the use of
authors would also like to thank the anonymous reviewers for their normalized difference vegetation index (NDVI) in the era of popular remote sensing.
constructive comments and suggestions. This research was supported by J. For. Res. 31 (5) https://doi.org/10.1007/s11676-020-01155-1.
the “Strategic Priority Research Program (A)” of the Chinese Academy of Kuang, W., Liu, J., Dong, J., Chi, W., Zhang, C., 2016. The rapid and massive urban and
industrial land expansions in China between 1990 and 2010: a CLUD-based analysis
Sciences (Grant No. XDA23030103 and XDA23030105) and the Na­
of their trajectories, patterns, and drivers. Landsc. Urban Plann. 145, 21–33. https://
tional Natural Science Foundation of China (Grant No. 41471137). doi.org/10.1016/j.landurbplan.2015.10.001.
Levin, N., 2017. The impact of seasonal changes on observed nighttime brightness from
2014 to 2015 monthly VIIRS DNB composites. Rem. Sens. Environ. 193, 150–164.
Appendix A. Supplementary data https://doi.org/10.1016/j.rse.2017.03.003.
Levin, N., Duke, Y., 2012. High spatial resolution night-time light images for
Supplementary data to this article can be found online at https://doi. demographic and socioeconomic studies. Rem. Sens. Environ. 119, 1–10. https://
doi.org/10.1016/j.rse.2011.12.005.
org/10.1016/j.jclepro.2021.129488.
Levin, N., Kyba, C.C., Zhang, Q., De Miguel, A.S., Roman, M.O., Li, X., Portnov, B.A.,
Molthan, A.L., Jechow, A., Miller, S.D., Wang, Z., Shrestha, R.M., Elvidge, C.D.,
References 2020. Remote sensing of night lights: a review and an outlook for the future. Rem.
Sens. Environ. 237, 111443. https://doi.org/10.1016/j.rse.2019.11144.
Li, K., Chen, Y., 2018. A Genetic Algorithm-based urban cluster automatic threshold
Bagan, H., Yamagata, Y., 2012. Landsat analysis of urban growth: how Tokyo became the
method by combining VIIRS DNB, NDVI, and NDBI to monitor urbanization. Rem.
world’s largest megacity during the last 40 years. Rem. Sens. Environ. 127, 210–222.
Sens. 10 (2), 277. https://doi.org/10.3390/rs10020277.
https://doi.org/10.1016/j.rse.2012.09.011.
Li, Q., Lu, L., Weng, Q., Xie, Y., Guo, H., 2016. Monitoring urban dynamics in the
Bagan, H., Yamagata, Y., 2015. Analysis of urban growth and estimating population
southeast USA using time-series DMSP/OLS nightlight imagery. Rem. Sens. 8 (7),
density using satellite images of nighttime lights and land-use and population data.
578. https://doi.org/10.3390/rs8070578.
GIScience Remote Sens. 52 (6), 765–780. https://doi.org/10.1080/
Li, X., Gong, P., Liang, L., 2015. A 30-year (1984–2013) record of annual urban dynamics
15481603.2015.1072400.
of Beijing City derived from Landsat data. Rem. Sens. Environ. 166, 78–90. https://
Ban, Y., Jacob, A., Gamba, P., 2015. Spaceborne SAR data for global urban mapping at
doi.org/10.1016/j.rse.2015.06.007.
30 m resolution using a robust urban extractor. ISPRS J. Photogrammetry Remote
Li, X., Ma, R., Zhang, Q., Li, D., Liu, S., He, T., Zhao, L., 2019. Anisotropic characteristic
Sens. 103, 28–37. https://doi.org/10.1016/j.isprsjprs.2014.08.004.
of artificial light at night–Systematic investigation with VIIRS DNB multi-temporal
Baugh, K., Hsu, F.C., Elvidge, C.D., Zhizhin, M., 2013. Nighttime lights compositing using
observations. Rem. Sens. Environ. 233, 111357. https://doi.org/10.1016/j.
the VIIRS day-night band: preliminary results. Proc. Asia-Pacific Adv. Netw. 35,
rse.2019.111357.
70–86. https://doi.org/10.7125/APAN.35.8.
Liu, F., Zhang, Z., Zhao, X., Wang, X., Zuo, L., Wen, Q., Yi, L., Xu, J., Hu, S., Liu, B., 2019.
Brunori, E., Salvati, L., Mancinelli, R., Smiraglia, D., Biasi, R., 2017. Multi-temporal land
Chinese cropland losses due to urban expansion in the past four decades. Sci. Total
use and cover changing analysis: the environmental impact in Mediterranean area.
Environ. 650, 847–857. https://doi.org/10.1016/j.scitotenv.2018.09.091.
Int. J. Sustain. Dev. World Ecol. 24 (3), 276–288. https://doi.org/10.1080/
Liu, X., Hu, G., Chen, Y., Li, X., Xu, X., Li, S., Pei, F., Wang, S., 2018. High-resolution
13504509.2016.1205156.
multi-temporal mapping of global urban land using Landsat images based on the
Cao, X., Chen, J., Imura, H., Higashi, O., 2009. A SVM-based method to extract urban
Google Earth Engine Platform. Rem. Sens. Environ. 209, 227–239. https://doi.org/
areas from DMSP-OLS and SPOT VGT data. Rem. Sens. Environ. 113 (10),
10.1016/j.rse.2018.02.055.
2205–2209. https://doi.org/10.1016/j.rse.2009.06.001.

12
Y. Zheng et al. Journal of Cleaner Production 328 (2021) 129488

Liu, Z., He, C., Zhang, Q., Huang, Q., Yang, Y., 2012. Extracting the dynamics of urban Shi, K., Huang, C., Yu, B., Yin, B., Huang, Y., Wu, J., 2014. Evaluation of NPP-VIIRS
expansion in China using DMSP-OLS nighttime light data from 1992 to 2008. Landsc. night-time light composite data for extracting built-up urban areas. Rem. Sens. Lett.
Urban Plann. 106 (1), 62–72. https://doi.org/10.1016/j.landurbplan.2012.02.013. 5 (4), 358–366. https://doi.org/10.1080/2150704X.2014.905728.
Lu, D., Tian, H., Zhou, G., Ge, H., 2008. Regional mapping of human settlements in Small, C., Elvidge, C.D., 2013. Night on earth: mapping decadal changes of
southeastern China with multisensor remotely sensed data. Rem. Sens. Environ. 112 anthropogenic night light in Asia. Int. J. Appl. Earth Obs. Geoinf. 22, 40–52. https://
(9), 3668–3679. https://doi.org/10.1016/j.rse.2008.05.009. doi.org/10.1016/j.jag.2012.02.009.
Ma, P., Wang, W., Zhang, B., Wang, J., Shi, G., Huang, G., Chen, F., Jiang, L., Lin, H., Song, X., Sexton, J., Huang, C., Channan, S., Townshend, J., 2016. Characterizing the
2019. Remotely sensing large-and small-scale ground subsidence: a case study of the magnitude, timing and duration of urban growth from time series of Landsat-based
Guangdong-Hong Kong-Macao Greater Bay Area of China. Rem. Sens. Environ. 232, estimates of impervious cover. Rem. Sens. Environ. 175, 1–13. https://doi.org/
111282. https://doi.org/10.1016/j.rse.2019.111282. 10.1016/j.rse.2015.12.027.
Ma, T., Zhou, C., Pei, T., Haynie, S., Fan, J., 2012. Quantitative estimation of Stokes, E.C., Seto, K.C., 2019. Characterizing urban infrastructural transitions for the
urbanization dynamics using time series of DMSP/OLS nighttime light data: a Sustainable Development Goals using multi-temporal land, population, and
comparative case study from China’s cities. Rem. Sens. Environ. 124, 99–107. nighttime light data. Rem. Sens. Environ. 234, 111430. https://doi.org/10.1016/j.
https://doi.org/10.1016/j.rse.2012.04.018. rse.2019.111430.
McFeeters, S.K., 1996. The use of the Normalized Difference Water Index (NDWI) in the Tang, L., Ma, W., 2018. Assessment and management of urbanization-induced ecological
delineation of open water features. Int. J. Rem. Sens. 17 (7), 1425–1432. https://doi. risks. Int. J. Sustain. Dev. World Ecol. 25 (5), 383–386. https://doi.org/10.1080/
org/10.1080/01431169608948714. 13504509.2018.1446193.
McGrane, S.J., 2016. Impacts of urbanisation on hydrological and water quality Tang, L., Wang, L., Li, Q., Zhao, J., 2018. A framework designation for the assessment of
dynamics, and urban water management: a review. Hydrol. Sci. J. 61 (13), urban ecological risks. Int. J. Sustain. Dev. World Ecol. 25 (5), 387–395. https://doi.
2295–2311. https://doi.org/10.1080/02626667.2015.1128084. org/10.1080/13504509.2018.1434570.
Meng, L., Graus, W., Worrell, E., Huang, B., 2014. Estimating CO2 (carbon dioxide) Tourea, S.I., Stowa, D.A., Shiha, H., Weeksa, J., Lopez-Carr, D., 2018. Land cover and
emissions at urban scales by DMSP/OLS (Defense Meteorological Satellite Program’ s land use change analysis using multi-spatial resolution data and object-based image
Operational Linescan System) nighttime light imagery: methodological challenges analysis. Rem. Sens. Environ. 210, 259–268. https://doi.org/10.1016/j.
and a case study for China. Energy 71, 468–478. https://doi.org/10.1016/j. rse.2018.03.023.
energy.2014.04.103. United Nations, 2014. World Urbanization Prospects: the 2014 Revision, Highlights.
Meng, Q., Zhang, L., Sun, Z., Meng, F., Wang, L., Sun, Y., 2018. Characterizing spatial Department of Economic and Social Affairs, New York, NY, USA. https://esa.un.
and temporal trends of surface urban heat island effect in an urban main built-up org/unpd/wup/CD-ROM/. (Accessed 20 October 2019). Accessed.
area: a 12-year case study in Beijing, China. Rem. Sens. Environ. 204, 826–837. Wei, Y., Liu, H., Song, W., Yu, B., Xiu, C., 2014. Normalization of time series DMSP-OLS
https://doi.org/10.1016/j.rse.2017.09.019. nighttime light images for urban growth analysis with pseudo invariant features.
Michishita, R., Jiang, Z., Xu, B., 2012. Monitoring two decades of urbanization in the Landsc. Urban Plann. 128, 1–13. https://doi.org/10.1016/j.
poyang lake area, China through spectral unmixing. Rem. Sens. Environ. 117, 3–18. landurbplan.2014.04.015.
https://doi.org/10.1016/j.rse.2011.06.021. Wigginton, N.S., Fahrenkamp-Uppenbrink, J., Wible, B., Malakoff, D., 2016. Cities are
Ouyang, Z., Lin, M., Chen, J., Fan, P., Qian, S.S., Park, H., 2019. Improving estimates of the future. Science 352, 904–905. https://doi.org/10.1126/science.352.6288.904.
built-up area from night time light across globally distributed cities through hier- Xian, S., Li, L., Qi, Z., 2019. Toward a sustainable urban expansion: a case study of
archical modeling. Sci. Total Environ. 647, 1266–1280. https://doi.org/10.1016/j. Zhuhai, China. J. Clean. Prod. 230, 276–285. https://doi.org/10.1016/j.
scitotenv.2018.08.015. jclepro.2019.05.087.
Pandey, B., Joshi, P.K., Seto, K.C., 2013. Monitoring urbanization dynamics in India Yang, X., Yue, W., Gao, D., 2013. Spatial improvement of human population distribution
using dmsp/ols night time lights and spot-vgt data. Int. J. Appl. Earth Obs. Geoinf. based on multi-sensor remote-sensing data: an input for exposure assessment. Int. J.
23 (1), 49–61. https://doi.org/10.1016/j.jag.2012.11.005. Rem. Sens. 34 (15), 5569–5583. https://doi.org/10.1080/01431161.2013.792970.
Pettorelli, N., Vik, J.O., Mysterud, A., Gaillard, J.M., Tucker, C.J., Stenseth, N.C., 2005. Yu, B., Shi, K., Hu, Y., Huang, C., Chen, Z., Wu, J., 2015. Poverty evaluation using NPP-
Using the satellite-derived NDVI to assess ecological responses to environmental VIIRS nighttime light composite data at the county level in China. IEEE J. Select.
change. Trends Ecol. Evol. 20 (9), 503–510. https://doi.org/10.1016/j. Top. Appl. Earth Obs. Rem. Sens. 8 (3), 1217–1229. https://doi.org/10.1109/
tree.2005.05.011. JSTARS.2015.2399416.
Román, M.O., Wang, Z., Sun, Q., Kalb, V., Miller, S.D., Molthan, A., Schultz, L., et al., Yu, B., Tang, M., Wu, Q., Yang, C., Deng, S., Shi, K., Peng, C., Wu, J., Chen, Z., 2018.
2018. NASA’s Black Marble nighttime lights product suite. Rem. Sens. Environ. 210, Urban built-up area extraction from log-transformed NPP-VIIRS nighttime light
113–143. https://doi.org/10.1016/j.rse.2018.03.017. composite data. Geosci. Rem. Sens. Lett. IEEE 15 (8), 1279–1283. https://doi.org/
Ryan, R.E., Pagnutti, M., Burch, K., Leigh, L., Ruggles, T., Cao, C., Aaron, D., Blonski, S., 10.1109/LGRS.2018.2830797.
Helder, D., 2019. The terra vega active light source: a first step in a new approach to Zha, Y., Gao, J., Ni, S., 2003. Use of normalized difference built-up index in
perform nighttime absolute radiometric calibrations and early results calibrating the automatically mapping urban areas from TM imagery. Int. J. Rem. Sens. 24 (3),
VIIRS DNB. Rem. Sens. 11 (6), 710. https://doi.org/10.3390/rs11060710. 583–594. https://doi.org/10.1080/01431160304987.
Schneider, A., Friedl, M.A., Potere, D., 2010. Mapping global urban areas using MODIS Zhang, Q., Schaaf, C., Seto, K.C., 2013. The vegetation adjusted NTL urban index: a new
500-m data: new methods and datasets based on ‘urban ecoregions. Rem. Sens. approach to reduce saturation and increase variation in nighttime luminosity. Rem.
Environ. 114, 1733–1746. https://doi.org/10.1016/j.rse.2010.03.003. Sens. Environ. 129, 32–41. https://doi.org/10.1016/j.rse.2012.10.022.
Schueler, C.F., Lee, T.F., Miller, S.D., 2013. VIIRS constant spatial-resolution advantages. Zhang, Q., Seto, K.C., 2013. Can night-time light data identify typologies of
Int. J. Rem. Sens. 34 (16), 5761–5777. https://doi.org/10.1080/ urbanization?A global assessment of successes and failures. Rem. Sens. 5,
01431161.2013.796102. 3476–3494. https://doi.org/10.3390/rs5073476.
Seto, K.C., Fragkias, M., Güneralp, B., Reilly, M.K., 2011. A meta-analysis of global urban Zhao, J., Liu, X., Dong, R., Shao, G., 2016. Landsenses ecology and ecological planning
land expansion. PLoS One 6, e23777. https://doi.org/10.1371/journal. toward sustainable development. Int. J. Sustain. Dev. World Ecol. 23 (4), 293–297.
pone.0023777. https://doi.org/10.1080/13504509.2015.1119215.
Sexton, J.O., Song, X.-P., Huang, C., Channan, S., Baker, M.E., Townshend, J.R., 2013. Zhao, J., Yan, Y., Deng, H., Liu, G., Dai, L., Tang, L., Shi, L., Shao, G., 2020. Remarks
Urban growth of the Washington, DC–Baltimore, MD metropolitan region from 1984 about landsenses ecology and ecosystem services. Int. J. Sustain. Dev. World Ecol. 27
to 2010 by annual, Landsat-based estimates of impervious cover. Rem. Sens. (3), 196–201. https://doi.org/10.1080/13504509.2020.1718795.
Environ. 129, 42–53. https://doi.org/10.1016/j.rse.2012.10.025. Zheng, Y., He, Y., Wang, X., Gao, Y., 2020. Application review and prospect of nighttime
Shao, G., Tang, L., Liao, J., 2019. Overselling overall map accuracy misinforms about light remote sensing data. Remote Sens. Inf. 35 (3), 1–14. https://doi.org/10.3969/j.
research reliability. Landsc. Ecol. 34, 2487–2492. https://doi.org/10.1007/s10980- issn.1000-3177.2020.03.001.
019-00916-6. Zheng, Y., Shao, G., Tang, L., He, Y., Wang, X., Wang, Y., Wang, H., 2019. Rapid
Shao, G., Tang, L., Zhang, H., 2021. Introducing Image Classification Efficacies. IEEE assessment of a typhoon disaster based on NPP-VIIRS DNB daily data: the case of an
Access 9, 134809–134816. https://doi.org/10.1109/ACCESS.2021.3116526. urban agglomeration along western Taiwan straits, China. Rem. Sens. 11 (14), 1709.
Sharma, R.C., Tateishi, R., Hara, K., Gharechelou, S., Iizuka, K., 2016. Global mapping of https://doi.org/10.3390/rs11141709.
urban built-up areas of year 2014 by combining MODIS multispectral data with Zhou, Y., Smith, S.J., Elvidge, C.D., Zhao, K., Thomson, A., Imhoff, M., 2014. A cluster-
VIIRS nighttime light data. Int. J. Dig. Earth 9 (10), 1004–1020. https://doi.org/ based method to map urban area from DMSP/OLS nightlights. Rem. Sens. Environ.
10.1080/17538947.2016.1168879. 147, 173–185. https://doi.org/10.1016/j.rse.2014.03.004.
Shi, K., Chen, Y., Li, L., Huang, C., 2018. Spatiotemporal variations of urban CO2
emissions in China: a multiscale perspective. Appl. Energy 211, 218–229. https://
doi.org/10.1016/j.apenergy.2017.11.042.

13