Environmental Research 221 (2023) 115257

Contents lists available at ScienceDirect

Environmental Research
journal homepage: www.elsevier.com/locate/envres

Spatially explicit carbon emissions by remote sensing and social sensing

Feng Gao a, b, Jie Wu a, b, *, Jinghao Xiao a, b, Xiaohui Li a, b, Shunyi Liao a, b, Wangyang Chen a, b
a
Guangzhou Urban Planning & Design Survey Research Institute, Guangzhou, 510060, China
b
Guangdong Enterprise Key Laboratory for Urban Sensing, Monitoring and Early Warning, Guangzhou, 510060, China

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

Keywords: Scientific simulation of carbon emissions is an important prerequisite for achieving low-carbon green develop­
Carbon emissions ment and carbon peak and carbon neutralization. This study proposed a carbon emissions spatialization method
Remote sensing based on nighttime light (NTL) remote sensing and municipal electricity social sensing. First, the economics-
Social sensing
energy comprehensive index (EECI) was proposed by integrating the NTL and municipal electricity consump­
Electricity consumption
Nighttime light
tion (EC) data. Second, the carbon emissions were spatialized at a fine scale based on NTL, EC, and EECI,
respectively. Finally, the geographical detector model was applied to quantify the influencing factors on carbon
emissions from the perspectives of individuals and interactions. Results show that combining remote sensing and
social sensing data helps depict carbon emissions accurately. The factor analysis found that GDP and population
were the basis of carbon emissions, while the secondary industry and urbanization rate were the direct factors.
This study is expected to provide constructive suggestions and methods for emission reduction, carbon peak, and
carbon neutrality in high-density cities in China.

1. Introduction challenges (Guan et al., 2012, 2014; B. Zheng et al., 2018). Therefore, to
achieve carbon peak and carbon neutralization, China needs to formu­
The world has consumed a large amount of fossil energy since the late feasible and fair emission reduction policies.
Industrial Revolution in the process of industrialization, resulting in In recent years, carbon emissions have gained increasing attention
high greenhouse gas emissions, mainly carbon dioxide (Petit et al., 1999; from many disciplines, including geography, economics, and environ­
Raupach et al., 2007). A series of ecological and environmental prob­ mental sciences. Among them, carbon emission spatialization has
lems such as climate warming decreased biodiversity, and extreme become one of the hot spots of related research, which is of great sig­
weather has become major challenges to the survival and sustainable nificance for energy conservation and emission reduction, low-carbon
development of human society (Cai et al., 2014; McMichael et al., 2006). development, and spatial planning at regional and urban micro scales.
In 2006, China surpassed the United States as the world’s largest emitter Most existing studies spatialized carbon emissions from two perspec­
of carbon emissions. From 2006 to 2016, China’s total carbon emissions tives: bottom-up and top-down.
accounted for 27.3% of the world’s emissions, with an average annual On the one hand, carbon emission spatialization from a bottom-up
growth rate of about 4.22%. Under the dual pressure of maintaining a perspective means improving the accuracy of carbon emission spatiali­
balance between economic development and carbon emission reduction, zation based on a large number of fine individual carbon emission
China is still gradually implementing a series of measures, including source data, which usually requires individual level statistical data and
carbon peaking and carbon neutrality (Feng et al. n.d.; Liu et al., 2015; field investigation. For example, Chen et al. (2021) calculated the spatial
Liu et al., 2015). In November 2015, the Chinese government pledged to distribution of industrial carbon emissions in Guangdong Province based
the international community to peak CO2 emissions around 2030. China on energy consumption and the Cement production process. Gurney
announced in September 2020 that it would strive to become carbon et al. (2020) calculated the spatial results of carbon emission of USA
neutral by 2060. In this context, due to the obvious regional urbaniza­ with a spatial resolution of 1 km by using the carbon emission data of
tion gap and great differences in economic and social development and industrial, commercial, residential and consumer industries. Chuai and
energy consumption structure in China, different regions of China are Feng (2019) obtained spatialization results of carbon emissions with a
faced with differentiated emission reduction requirements and resolution of 300 m based on detailed energy consumption data from

* Corresponding author. Guangzhou Urban Planning & Design Survey Research Institute, Guangzhou, 510060, China.
E-mail address: [email protected] (J. Wu).

https://doi.org/10.1016/j.envres.2023.115257
Received 28 September 2022; Received in revised form 5 December 2022; Accepted 9 January 2023
Available online 13 January 2023
0013-9351/© 2023 Elsevier Inc. All rights reserved.
F. Gao et al. Environmental Research 221 (2023) 115257

various land uses. However, the data needed for fine-scale carbon et al., 2012, 2014). Therefore, it was not enough to analyze only the
emission spatialization from a bottom-up perspective are usually diffi­ single impact of factors on carbon emissions.
cult to obtain due to data availability (Cai et al., 2020). In addition, data In this study, we comprehensively combined NTL remote sensing and
collation and summary are time-consuming, and this method is rela­ social sensing in the carbon emissions spatialization (Fig. 1). A new
tively lack of timeliness and universality, and it is difficult to be widely indicator, the economics-energy comprehensive index (EECI) was pro­
applied in other regions. posed by integrating the NTL data and municipal electricity consump­
On the other hand, the top-down method refers to the spatialization tion (EC) big data at fine scale, which was applied to spatially allocate
of carbon emissions based on easily accessible open data, such as carbon emissions based on industrial proportion of carbon emissions and
nighttime light (NTL) remote sensing data and the total energy con­ urban functional areas. Then, the results of carbon emissions spatiali­
sumption, to realize the transformation of carbon emissions from the zations based on NTL, EC, and EECI were compared with the open data,
scale of administrative boundaries to the scale of pixels. This method can respectively. Finally, the geographical detector model was applied to
clearly and objectively demonstrate the spatial distribution of carbon quantify both the individual and interactive influences of factors on
emissions. Open data, represented by NTL data, can be applied to the carbon emissions in Guangzhou. This study is expected to provide
spatial study of greenhouse gas emissions related to social population, constructive suggestions and methods for energy conservation and
energy consumption and energy consumption due to its ability to reflect emission reduction, carbon peak and carbon neutrality of high-density
the intensity of human activities (Drewniok and Rohr, 1997). For cities in China.
example, two of the most widely used data on the spatial distribution of
carbon emissions, ODIAC and EDGAR, were based on top-down ap­ 2. Data and methods
proaches. The ODIAC data were based on global NTL data and power
plant emissions data (Oda et al., 2018), while the EDGAR data were 2.1. Study area
based on global land use and human activity data (Muntean, 2014).
Based on DMSP-OLS night light data, Shi et al. analyzed and compared The research area of this paper, Guangzhou City, is located in the
the spatial-temporal variation of carbon emissions and its influencing south-central part of Guangdong Province and is the core of the
factors at provincial and prefectural levels in China, and found that the Guangdong Hong Kong Macao Bay area (Fig. 2). By 2021, the city has 11
secondary industry was the main influencing factor of carbon emissions districts with a total area of 7434.4 km2 and a resident population of
(Shi et al., 2019). more than 18 million (Gao et al., 2021a).
Although many existing studies have made great contributions to the As an important central city of the country, Guangzhou has a rela­
study of carbon emission spatialization, there are still some problems to tively complete industrial system, the service industry is booming, the
be solved. First of all, spatial resolution is a non-negligible issue, which momentum of economic development continues to increase, and the
is crucial for the fineness of the spatial results of carbon emissions and its energy demand is strong, which puts forward higher requirements for
subsequent application. At present, the spatial resolution of most studies energy production and supply and energy security. Due to such factors
of this type is relatively coarse, mainly based on national (Muntean, as the substitution of industrial electric energy, the increase in the
2014; Oda et al., 2018; Shan and Hu; n. d), provincial (Chu et al., 2018; proportion of electric buses and taxis, and the increase of the urban
Wei et al., 2017), and urban scales (Long et al., 2021; Wang, 2022). population, the power consumption in 2020 was 99.7 billion kWh, an
Among them, the previous studies at urban scale have reached a finer increase of 24.3% over 2015 and an average annual growth of 4.4%.
resolution of 1 km × 1 km, 300 m × 300 m, and 150 m × 150 m in
carbon emission spatialization. With the development of science and
technology, the resolution of NTL data has been greatly improved. Some 2.2. Data
studies have used high-resolution NTL data, the Luojia data in carbon
emission spatialization (Lu et al., 2022; Wang, 2022). However, due to The data used in this study include essential urban land use cate­
the sensor of NTL data itself, the spatial distribution of data has the gories (EULUC) dataset, Luojia 1-01 NTL data, total carbon emission,
problem of oversaturation (Zhang et al., 2022; Zheng et al., 2019), but proportion of carbon emission by industry, and municipal electricity
this has not been noticed and solved in the research of carbon emission consumption data. Except for electricity consumption data, other data
spatialization. Simply put, this problem is that the high-value areas of are open-source data, as shown in Table 1.
the carbon emission spatialization results obtained using high-resolution
NTL data were too concentrated and it was difficult to further distin­ 2.2.1. Carbon emissions
guish their internal spatial heterogeneity. For high carbon emission re­ Combining the total amount and the proportion of carbon emissions
gions, it is of great significance to distinguish their internal by industry, the amount of carbon emissions by industry were calculated
characteristics and differences in detail. Therefore, integrating as Table 2. The total carbon emission of Guangzhou in 2019 was
high-resolution NTL data with other social sensing data is necessary to 112.479 million t, among which the secondary industry and trans­
spatially explicit of carbon emissions at a further finer scale. portation accounted for the largest proportion, reaching 41.2% and 37%
Second, although NTL reflects the intensity of human activities, it respectively, followed by residential & consumption (15.4%), tertiary
was rare to further distinguish the difference in energy consumption of industry (5.8%) and primary industry (0.6%).
human activities in the process of carbon emissions spatialization. Most
researches on the spatialization of carbon emissions either started from 2.2.2. Luojia 1-01 NTL data
the intensity of human activities based on NTL data (Lu et al., 2022; Compared with other NTL data such as DMSP/OLS and VIIRS, the
Wang, 2022), or carried out bottom up energy consumption calculation Luojia 1-01 NTL data has greatly improved in spatial resolution reaching
by region or location (Chuai and Feng, 2019; Long et al., 2021). 130 m (Liu et al., 2020). First, using the administrative boundary of
Therefore, it is insufficient to spatialize carbon emissions only consid­ Guangzhou as a mask, the Luojia 1-01 NTL data within the scope of
ering human activity intensity or bottom up energy consumption Guangzhou were extracted. Then, the Landsat 8 with higher spatial
calculation, which is an urgent problem to be solved. Third, previous resolution was used as the reference image, and the road intersection
studies have conducted influencing factors analysis of carbon emissions and building inflection point were selected as the feature points for
by using various types of regressions or correlation analysis. However, geometric calibration, and the error was controlled within 0.5 pixels
the interactive influences between factors on carbon emissions was (Lin et al., 2022). Finally, radiation correction was performed based on
seldom explored. In a highly urbanized environment, the formation of the radiation correction formula of Luojia 1-01 satellite product as
carbon emissions is caused by multiple factors, not a single factor (Guan follows:

2
F. Gao et al. Environmental Research 221 (2023) 115257

Fig. 1. The workflow of the study.

L = DN 3/2 × 10− 10
(1) emissions of residents’ consumption, that is, all carbon emissions
generated by residents’ spending on household consumer goods and
Where L is the radiance value after absolute radiation correction; DN is service-oriented expenditure. Therefore, this paper divided residential,
the image gradation value. administrative, educational, medical, sport and cultural functional areas
into residential and consumption. Moreover, agriculture, forestry, ani­
2.2.3. Municipal electricity consumption data mal husbandry, fishery and other industries in the primary industry are
The municipal EC big data recorded all the EC data of Guangzhou, difficult to distinguish between carbon source and carbon sink (Lu et al.,
including personal and non-personal categories, which is the monthly 2022), and the primary industry accounts for a very low proportion of
EC data of the whole year in 2019. The original data is in text format, carbon emissions in Guangzhou (0.6%). Therefore, the primary industry
recording the specific address and the average power consumption in was not included in the research scope.
2019, with more than 10 million records. This kind of municipal service
data (such as water consumption and EC data) has begun to appear in 2.3. Methods
the frontier studies due to its high spatial-temporal resolution and wide
population coverage (Guan et al., 2021; Yao et al., 2022). 2.3.1. Economic-energy comprehensive index (EECI)
The processing of EC data is presented in Fig. 3. First, all addresses The NTL remote sensing data reflects the intensity of social and
were geocoded by using Baidu map API with Python script (https://lbs. economic activities in the city (Huang et al., 2021; Zheng et al., 2021).
baidu.com/index.php?title=webapi/guide/webservice-geocoding). However, it is insufficient to consider only NTL data in carbon emission
After manual inspection and deleting address confusion, errors, and simulation. In addition to the degree of economic development, the
addresses that were not local, the final geocoding hit rate was 96.5%, energy consumption generated by economic development also needs to
and the accuracy is high enough to meet the needs of subsequent be considered. To measure the economic development and energy
research. Finally, a set of grids was generated with the same spatial consumption simultaneously, we integrated the Luojia 1-01 NTL data
position and resolution as the Luojia 1-01 data, and the geocoded EC and municipal electricity consumption big data by using Entropy-weight
data were summarized into the grid. Each grid has the information of approach, which is an objective weighting method considering the dis­
Luojia 1-01 DN value and the total EC. tribution of values (Tu et al., 2020; Zou et al., 2006). The Shannon en­
tropy of one type of indicators (NTL and EC), Hk , is measured as Eq. (2),
2.2.4. Essential functional land use where xki is the indicator in a grid, i; n is the total count of grids.
The EULUC dataset have five major categories (Residential, Com­ ∑
n
mercial, Industrial, Transportation, Public management, and service) Hk = − xki ln(xki ) (2)
and thirteen sub categories (Residential, Business office, Commercial 1

service, Industrial, Road, Transportation stations, Airport facilities, The comprehensive index (EECI) is calculated as Eq. (3), where xk is
Administrative, Educational, Medical, Sport and cultural, Greenspace, one type of indicators.
Water) (Gong et al., 2020).
∑
According to the industrial proportions of carbon emissions in this EECI =
Hx
∑k k (3)
study (Table 2), the EULUC data was reclassified into five categories k={NTL,EC}
Hk
corresponding to five industries (Fig. 4). Based on the above data, the
greenspace and water body were classified as the primary industry; The 2.3.2. Spatialization of carbon emissions
industrial was classified as the secondary industry; Business office and The spatialization of carbon emissions is four-fold shown in Fig. 5.
commercial service were classified as the tertiary industry; Road, First, based on the open data released by the government, the total
transportation stations and airport facilities were classified as the carbon emissions of Guangzhou in 2019 and the proportion of industries
transportation industry; Residential, administrative, educational, med­ were obtained, and then the carbon emissions of each industry was
ical, sport and cultural were residential and consumption. calculated, which can be expressed by the following formula.
In addition, residential and consumption refers to the indirect carbon

3
F. Gao et al. Environmental Research 221 (2023) 115257

Fig. 2. Study area.

Table 1 Table 2
Data and sources. Proportion and quantity of carbon emissions by industry.
Data Time Source Industry Proportion of Carbon Emissions Carbon Emissions
/Million t
EULUC 2019 Tsinghua University, China (Gong et al.,
2020) Residential & Consumption 15.4% 17.322
(http://data.ess.tsinghua.edu.cn) Primary Industry 0.6% 0.675
Luojia 1-01 NTL 2019 Wuhan University, China (http://59.17 Secondary Industry 41.2% 46.341
5.109.173:8888/app/login.html) Tertiary Industry 5.8% 6.524
Total carbon emission 2019 Carbon Emission Accounts and Datasets Traffic 37.0% 41.617
(CEADs) (Shan, 2017; Shan et al., 2018,
2019) (https://www.ceads.net/)
Proportion of carbon 2019 Guangzhou Institute of Energy Conversion, of NTL-DN, EC and EECI respectively to obtain the carbon emissions
emission by industry Chinese Academy of Sciences (Lu et al., 2022) represented by the unit value of NTL-DN, EC and EECI in the same in­
Municipal electricity 2019 Local power supply department
consumption data
dustry, which is called carbon emission coefficient in this study
expressed as the following formula.
Carbon emissionsi
Carbon emissionsi = Total carbon emissions × Proportioni (4) NTLCarbon emission Coefficient i = (5)
Total NTLDNi
where Carbon emissionsi is the carbon emissions of the ith industry.
Carbon emissionsi
Second, the total values of NTL-DN, EC, and EECI of each industry ECCarbon emission Coefficient i = (6)
Total ECi
were counted based on the land use boundaries of each industry. Third,
the total carbon emissions of each industry are divided by the total value

4
F. Gao et al. Environmental Research 221 (2023) 115257

Fig. 3. Processing of EC data.

Carbon emissionsi [ ]/[ ]

EECICarbon emission Coefficient i = (7) ∑n ∑n ( ) ∑n ∑n ∑n
Total EECIi I = n i=1 (xi − x) j=1 Wij xj − x (x − x) 2
W
i ij
i=1 i=1 j=1

where Carbon emissionsi is the carbon emissions of the ith industry,

(8)
Total NTLDNi is the total value of NTL-DN within the land use boundary
of the ith industry, Total ECi is the total value of EC within the land use Where: n is the number of grid pixels at 130 m resolution; i and j are the
boundary of the ith industry, Total EECIi is the total value of EECI within carbon emissions of i and j space units, respectively; x is the average
the land use boundary of the ith industry. value of carbon emission of each grid pixel; Wij is a spatial weight ma­
It has been found that the intensity of NTL and energy consumption trix. The weight matrix based on adjacency relationship was selected to
are highly correlated with carbon emissions (Chen et al., 2021; Doll perform Z-test on Moran’s I value to reveal the spatial distribution
et al., 2000; Wu et al., 2022). Therefore, in the step 4, three results of characteristics of carbon emissions (Anselin, 2010).
carbon emissions were spatialized by three indicators, respectively. For The local Moran’s I index was applied to conduct local spatial
example, the NTL-DN value of each grid was multiplied by the autocorrelation analysis on the carbon emission in Guangzhou, and
NTL-Carbon emissions coefficient according to the industry of each grid identify the high-value and low-value clusters of carbon emissions to
to assign carbon emissions. Finally, the simulation results of carbon reflect the spatial dependence and heterogeneity of carbon emissions.
emissions by industry with a resolution of 130 m were obtained. The expression of local Moran’s I is:
/
2.3.3. Spatial autocorrelation analysis ∑n ∑n
Ii = Wij xj xi (9)
The global Moran’s I index was used as the global spatial autocor­ j=1 i=1

relation indicator to analyze the spatial correlation and difference of

urban carbon emissions distribution in Guangzhou. The expression is: Wherein, when Ii is greater than 0, there is a positive spatial correlation;
When Ii is less than 0, there is a negative spatial correlation (Anselin,
2010).

5
F. Gao et al. Environmental Research 221 (2023) 115257

Fig. 4. The reclassification of essential functional land use by industry of Guangzhou in 2019.

Fig. 5. The workflow of the spatially explicit carbon emissions.

2.3.4. Geographical detector model et al., 2021; Fan et al., 2021; Gao et al., 2021b). The factor detector
This study adopted the geographical detector model to quantify the model and interaction detector model were sub-models of geographical
influencing factors of carbon emissions distribution. It is a spatial sta­ detector model, which were applied in this study. The geographical
tistical model based on spatial variance analysis with the advantage of detector model was performed by using the open access software in R
analyzing individual and interactive influence at the same time (Wang named GD (https://CRAN.R-project.org/package=geodetector).
et al., 2010, 2016), which has been widely used in spatial studies (Deng The factor detector sub-model was applied to analyze the individual

6
F. Gao et al. Environmental Research 221 (2023) 115257

influences of factors on carbon emissions distribution based on the q Table 4

value, and its formula is: Summary statistics of variables.

∑
L Variables Definitions References
Nh σ2h
SSW Urban land rate (UR) The area rate of urban (Falahatkar and Rezaei,
q = 1 − L h=1 =1− (10) land 2020; Lu et al., 2022; Zhu
Nσ 2 SST
et al., 2022)
Proportion of Secondary industry (Li et al., 2021; M. Wang
Where q is the explanatory power of factors associated with the carbon secondary industry output value/Total et al., 2022; Zhu et al., 2022)
emissions distribution, h = 1, …, L are the stratification of y or factor x; (SI) output value
Nh and N represents the number of units in h and the whole study area, Total social Sum of regional retail (Lu et al., 2022; B. B. Yuan
consumption (SC) sales of goods and et al., 2015; Zhu et al., 2022)
respectively. σ2h and σ2 are the variance of units in h and the global
services
variance of y over the whole study area, respectively. SSW and SST GDP Gross regional product (M. Wang et al., 2022; Wei
represents the within the sum of squares and the total sum of squares, et al., 2017; S. Zheng et al.,
respectively. The q value ranges between 0 and 1, and the higher the q 2018)
value is, the stronger the influence of factor x on y (Wang et al., 2010). Population (POP) Permanent resident (M. Wang et al., 2022; Wei
population at yearend et al., 2017; S. Zheng et al.,
The interaction detector sub-model was applied to explore whether 2018)
the factor pair (x1 and x2 ) have an interactive influence on carbon
emissions, and how it compares with the individual influences. First, the
q value of factors x1 and x2 were calculated and marked as q (x1 ) and q industry area is the functional area with the highest average value of
(x2 ). Then, the interactive q value of factors x1 and x2 was calculated and NTL, EC, and EECI.
marked as q (x1 ∩x2 ) (Wang et al., 2016). This sub-model can categorize Fig. 6 shows the general distribution of NTL, EC, and EECI. The
the interactive relationship into five types by comparing the interactive spatial distribution of NTL data was relatively continuous, and it was
q value and the individual q values of each pair of factors (Table 3). distributed along the traffic line in the central area of the city, or in
patches (Fig. 6a). On the contrary, since the original EC data was co­
2.3.5. Variables ordinate point data with power consumption information, and then was
It has been reported that urban socioeconomic features play an counted through a grid with the same spatial resolution as NTL data, its
important role in practical urban planning and management for spatial distribution was relatively discrete (Fig. 6b). Therefore, this
reducing carbon emissions (Chuai and Feng, 2019; Wei et al., 2017; Zhu study further integrated NTL and EC data through entropy weight
et al., 2022). Based on the spatially explicit carbon emissions, this study method to obtain a new comprehensive index, EECI, whose spatial dis­
further analyzed the influencing factors of carbon emissions. The tribution is shown in Fig. 6c. It can be seen that EECI further integrated
dependent variable was the carbon emissions of each grid. To measure the spatial characteristics of NTL and EC data (Fig. 6c).
the carbon emissions-related socioeconomic features, potential factors
were selected from five dimensions based on data availability and
3.2. Spatialization of carbon emissions
related studies (Li et al., 2021; Lu et al., 2022; M. Wang et al., 2022; Y.
Wang et al., 2022; S. Zheng et al., 2018; Zhu et al., 2022). These factors
The results of the carbon emissions spatializations are displayed in
included urban land rate (UR), the proportion of secondary industry (SI),
Fig. 7. Since the results of carbon emission spatialization were converted
the total social consumption (SC), GDP, and population (POP) of each
from the original data of the three indicators (NTL, EC, and EECI), their
districts, which were obtained from statistical yearbook (http://tjj.gz.
spatial distribution characteristics were similar with indicators. The
gov.cn/) in 2019 (see Table 4).
spatial distribution of carbon emissions based on NTL was relatively
continuous, while the EC-based carbon emissions tended to be discrete.
3. Results and discussion Finally, the carbon emissions based on EECI integrated the spatial
characteristics of NTL-based and EC-based carbon emissions.
3.1. Spatial distribution of NTL, EC, and EECI Three types of functional areas were further selected to compare the
results of carbon emission spatialization. First, Fig. 7d shows Zhujiang
The NTL, EC, and EECI were selected as indicators to spatialize the New Town in Guangzhou, which is the largest CBD in South China. It has
carbon emissions, and their spatial distributions within different types of become a highly concentrated area of headquarters economy, finance,
urban functional areas are shown in Fig. 6. In terms of NTL, the average science and technology, commerce, and other high-end industries in
NTL DN value of the tertiary industry area was the highest (29,404), South China. Zhujiang New Town is the most densely populated place
followed by the secondary industry area (19537), traffic (18862), and with skyscrapers over 300 m in China (Lin et al., 2020), and its carbon
residential and consumption area (15,212). In terms of EC, the monthly emission scale is far higher than that of other areas in Guangzhou (Lu
EC of the tertiary industry area was the highest (69,862 kWh), followed et al., 2022). The high values of three types of carbon emissions were
by the secondary industry area (46,947 kWh), traffic (35,391 kWh), and mainly distributed on both sides of the green belt in the center of CBD.
residential and consumption area (25,734 kWh). In terms of EECI, the Among them, the high values of EC-based carbon emissions were
average of standardized EECI of tertiary industry area was the highest distributed in the southern part of CBD (Fig. 7f). Second, although the
(0.0025), followed by traffic (0.0015), the secondary industry area Guangzhou Baiyun International Airport is nearly 30 km away from the
(0.0014), and residential and consumption area (0.0011). From the city center, the surrounding economic development lags far behind the
statistical results of different urban functional areas, the tertiary city center, and it is the main source of traffic carbon emissions. In 2020,
the passenger throughput ranked first in the world, so its carbon emis­
Table 3 sion scale is large (Zhang, 2019). From Fig. 7i and j, there was significant
The five types of interactive relationship. difference between NTL-based and EC-based carbon emissions because
Interaction Description the whole airport was bright in NTL data, but the power supply stations
Enhanced, nonlinear q (x1 ∩x2 ) > q (x1 )+ q (x2 ) is concentrated in one part of the airport. Therefore, in this case where
Independent q (x1 ∩x2 ) = q (x1 )+ q (x2 ) the carbon emissions calculated from NTL and EC are very different, it is
Enhanced, bilinear q (x1 ∩x2 ) > Max(q (x1 ), q (x2 )) particularly necessary to integrate the two indicators before calculating
Weakened, unique Min(q (x1 ), q (x2 )) < q (x1 ∩x2 ) < Max(q (x1 ), q (x2 )) carbon emissions. By integrating the EC and NTL, the EECI-based carbon
Weakened, nonlinear q (x1 ∩x2 ) < Min(q (x1 ), q (x2 ))
emissions avoid the saturation of NTL-based carbon emissions

7
F. Gao et al. Environmental Research 221 (2023) 115257

Fig. 6. Spatial distribution of three indicators within different types of urban functional areas.

8
F. Gao et al. Environmental Research 221 (2023) 115257

Fig. 7. Spatialization of carbon emissions from different industry sectors in Guangzhou by using three different methods: NTL-based (a,e,i,m), EC-based (b,f,j,n),
EECI-based (c,g,k,o) at the scale of 130 m grid.

distribution and the over discrete distribution of EC-based carbon high, leading to the simulated carbon emissions were also high, so the
emissions (Fig. 7k). Third, Fig. 7l shows the industrial area with multiple difference in carbon emissions among the three sets of results was
types of industrial enterprises, including petrochemical, electronic relatively small. On the contrary, the carbon emission simulation results
product assembly, high-tech and so on. The main industrial sites were in the airport and the industrial zone were quite different. Specifically,
generally covered with high brightness at night and relatively discrete EC-based carbon emissions differed greatly from NTL-based and EECI-
high electricity consumption. based carbon emissions in airports and industrial areas, and the bivar­
We further compared the three carbon emission results through the iate Moran’s I between them was low, only 0.179 and 0.156 respec­
bivariate spatial autocorrelation and regression relationship (Fig. 8.). tively. Importantly, the difference between NTL-based carbon emissions
First, we found that the High-High clusters were mainly located in and EECI-based carbon emissions was the smallest, and the bivariate
downtown and the centers of peripheral and outer suburbs in both three Moran’s I between them reached 0.504. It can be seen from the bivariate
sets of bivariate spatial autocorrelations (Fig. 8a–c). Secondly, as CBD is spatial autocorrelation distribution that NTL-based carbon emissions
the core of the city with high night light intensity and high power and EECI-based carbon emissions better depicted the spatial concen­
consumption, the three indicators (NTL, EC, and EECI) of CBD were all tration characteristics of high carbon emissions in airports and industrial

9
F. Gao et al. Environmental Research 221 (2023) 115257

Fig. 8. Bivariate spatial autocorrelations and relationships between three results of carbon emissions spatialization based on NTL, EC, and EECI.

parks at a fine scale (Fig. 8b). interesting and challenging research topic and many advanced data and
In addition, the scatter plots of three groups of carbon emission re­ models have been applied, the accuracy verification is still a problem to
sults were compared. Similarly, the regression relationship between NTL be solved. The data sources of carbon emission statistics used in previous
based carbon emissions and EECI based carbon emissions is far stronger studies are inconsistent, which makes the total carbon emission in cases
than the other two groups of results (Fig. 8e). It is worth noting that, in incomparable and the carbon emission of space units difficult to
general, the spatial distribution characteristics of consistency among the compare. Therefore, most related studies performed accuracy verifica­
three groups of results were similar to carbon emissions, that is, the tion by comparing the spatial heterogeneity of carbon emission distri­
highest was in downtown, followed by peripheral and outer suburbs. bution, rather than direct numerical comparisons such as correlation
To sum up, by comparing the spatial distribution, bivariate spatial analysis (Chuai and Feng, 2019; Long et al., 2021; Lu et al., 2022; Wang,
autocorrelation and scatter plot of the three sets of results, the EECI- 2022).
based carbon emissions were relatively optimal. On the one hand, car­ The ODIAC is one of the few freely available and widely used data on
bon emissions spatialization based on NTL data show that high-value carbon emissions, which was generated based on DMSP-OLS NTL data
areas account for a larger proportion and have a certain saturation and global power plant emissions data with a spatial resolution of 1 km
phenomenon, which is a common problem of NTL remote sensing data (Oda et al., 2018). This study spatialized carbon emissions based on the
(Zhang et al., 2022; Zheng et al., 2019). On the other hand, the EC-based total amount of carbon emissions from CEADs (Shan et al., 2019) and the
carbon emissions spatialization tended to be distributed discretely due proportions of carbon emission by industry from the government report
to the limitation of discrete spatial distribution of EC coordinate point of "Guangzhou’s Carbon Emission Peak and the 14th Five-Year Plan
data. Results show that the carbon emissions spatialization by inte­ low-carbon Development Strategy Path Research” (Lu et al., 2022).
grating NTL and EC finely depicted the spatial distribution pattern of These two data provided the numerical information of carbon emissions,
carbon emissions by industry within the city, which overcame the with which this study developed a framework of carbon emissions
problem of both saturation and discrete distribution. spatializations considering the intensity of human activities and energy
consumption.
As mentioned above, it is should be noticed that the total amounts of
3.3. Verification
carbon emissions are different between ODIAC and CEADs, which may
make further correlation verification inappropriate. Specifically, the
Although the spatialization of carbon emissions has been an

10
F. Gao et al. Environmental Research 221 (2023) 115257

total carbon emissions of ODIAC was much larger than CEADs, and three suburbs with good ecological environment quality, mainly including
spatialization results of this study had smaller ranges of outliers (Fig. 9.). Conghua District, Zengcheng district and Nansha District.
The spatial heterogeneity of the research results was compared with The local spatial autocorrelation method was further applied to
ODIAC data in this study (Fig. 10). To be comparable, the results of the analyze the local characteristics of carbon emissions in Guangzhou. As
study were converted into the same spatial scale (1 km grid). First, on shown in Fig. 11, the high-high, high-low, low-high, and low-low ag­
the macro scale, the spatialization results of carbon emissions in this gregations were used to reflect the local spatial relationships. The high-
study was roughly consistent with ODIAC data, that is, the carbon value aggregations with statistical significance denoted as high-high
emission distribution had the feature of distance decay. Specifically, the were distributed in downtown and centers of peripheral and suburbs.
downtown area had the highest carbon emissions, followed by periph­ The low-value aggregations denoted as low-low were mainly distributed
eral and outer suburbs. Second, limited by spatial resolution of DMSP- in outer suburbs. The low-value regions surrounded by high-value re­
OLS NTL data, ODIAC data poorly depict the fine-scale spatial differ­ gions were denoted as low-high, which were mainly distributed in the
ences of intra-city carbon emissions. For example, three peaks were outer downtown center and peripheral. In general, the higher the carbon
identified in the EECI-based carbon emissions, including the airport, emissions, the stronger the local spatial autocorrelation significance.
CBD, and industrial area, which represent transportation, the tertiary
industry, and the secondary industry, respectively. However, these
3.5. Influencing factors of carbon emissions
typical areas of peak carbon emissions were not reflected in the ODIAC
data. Rather, large areas of the Yuexiu district and the west of Tianhe
This study further analyzed the socio-economic factors of urban
district were found to be peak carbon emissions in ODIAC data. It
carbon emissions by using the geographical detector model, aiming to
overestimated the carbon emissions in center area and underestimated
provide suggestions for energy conservation and emission reduction in
the carbon emissions in airport and industrial area since that ODIAC
Guangzhou. First, the total carbon emissions of each administrative re­
data was limited by its spatial resolution and the lack of consideration of
gion were counted based on the administrative boundary. Taking the
local energy consumption.
total carbon emissions of each administrative region of Guangzhou as
Therefore, the method of integrating high-resolution NTL remote
the dependent variable and the five economic indicators as the inde­
sensing data and municipal electricity consumption big data to simulate
pendent variables, the influencing factors of carbon emissions in
carbon emissions in this study is feasible, which can accurately simulate
Guangzhou were quantified (Fig. 12). The average q value reached 0.4 at
high carbon emission regions and depict the spatial heterogeneity of
the significance level of 1%, indicating that the fitting level and the
carbon emissions in small-scale urban regions.
reliability of the model were ideal.
The factor detector model found that five factors were positively
3.4. spatial autocorrelation of EECI-based carbon emissions associated with carbon emissions with different degrees of explanatory
powers. Among them, UR was the factor with the highest q value of
The Moran’s I value of EECI-based carbon emissions in this study was 0.654, followed by SI (0.439), SC (0.336), GDP (0.3), and POP (0.249).
0.33 at the significance level of 1%, which indicates that Guangzhou’s UR and SI were the key factors with their explanatory powers above the
carbon emissions were highly concentrated in space. The regions with mean level, which indicates that UR and SI explained 65.4% and 43.9%
high carbon emission intensity were close to other regions with high of the variance of carbon emissions in Guangzhou.
carbon emission intensity, and the regions with low carbon emission From the perspective of individual influence of each factor, the SI in
intensity were also close to each other. Guangzhou accounted for the highest proportion of carbon emissions in
Comparing the results of carbon emissions in Guangzhou with the 2019, exceeding 40%. Moreover, factor analysis results showed that the
data of land-use functional areas, it can be found that the high-value SI accounted for the second largest proportion of carbon emissions,
areas of carbon emissions were located in the CBD and business dis­ explaining 43.9% of the variance of the spatial distribution of carbon
trict with concentrated high-rise buildings in the center of Guangzhou, emissions in Guangzhou. It can be seen that the impact of the SI on the
the International Airport in the north, and residential areas with high carbon emissions of Guangzhou, a megacity, is crucial. SI indicates the
population density; Medium and high value areas were located in urban- proportion of the secondary industry in the region, which reflects a
rural fringe industrial areas with large energy consumption and high rational industrial structure (M. Wang et al., 2022). Its explanatory
carbon emission intensity. Heavy industries such as petrochemical in­ power was second only to UR, indicating the proportion of the secondary
dustry and automobile manufacturing were concentrated in Huangpu industry in Guangzhou was an important factor in carbon emissions. In
District in the East, Baiyun District in the north and Panyu District in the addition, although the proportion of residents and consumption in
south; The areas with low carbon emission intensity were located in the Guangzhou’s carbon emissions was very small (0.6%), the joint

Fig. 9. Box plots of carbon emissions in this study and ODIAC dataset in downtown, peripheral, and outer suburbs.

11
F. Gao et al. Environmental Research 221 (2023) 115257

Fig. 10. Comparisons between carbon emissions in this study and ODIAC dataset.

explanatory power of SC∩POP was also high, reaching 0.805. It in­ factors leading to a large number of carbon emissions.
dicates that the residential areas in the city center, which are densely
populated and have a high level of social consumption, also have a great 3.6. Summary
impact on urban carbon emissions. The interaction detector model
found that the interactive influence of every pair of factors was higher As for the spatialization of carbon emissions, previous works seldom
than their individual influence in the form of nonlinear enhance and considered both the human activities intensity and energy consumption.
bilinear enhance (Fig. 12b). Importantly, the joint explanatory power of This study was the first attempt to finely spatialize carbon emissions
GDP∩UR was the highest (0.935), followed by GDP∩SI (0.932) and based on remote sensing and social sensing by taking into account the
POP∩SI (0.901). Results show that although the individual explanatory human activities intensity and individual level electricity consumption.
power of GDP and POP to carbon emissions was low (ranking the last The results show that the integration of NTL and EC can help to more
two), the joint explanatory power of these two factors was high when accurately depict the spatial heterogeneity of urban carbon emissions
interacting with SI and UR. than previous studies.
From the perspective of interactive influences between factors, the As for the influencing factors analysis, most existing studies focused
interactive detector model results show that the economic development on the individual influence of each factors on carbon emissions, how­
and urban population were the basis of the increase in urban carbon ever, little attention has been paid to the interactive influences between
emissions, rather than the direct factors. The top three pair of factors factors. A multi-level quantitative analysis of factors on carbon emis­
(GDP∩UR, GDP∩SI, and POP∩SI) indicate that based on the continuous sions was conducted by applying the geographical detector model,
development of the regional economy and the gradual increase of the including the influence of single factor on carbon emissions and the
population, the gradual expansion of urban construction land and the interactive influence of pairs of factors on carbon emissions. The model
vigorous development of the secondary industry were the key direct results divide the influencing factors into two levels: basic background

12
F. Gao et al. Environmental Research 221 (2023) 115257

Fig. 11. Local Moran’s I of carbon emissions based on EECI (a) and significance distribution (b).

Fig. 12. Results of geographical detector model: factor detector (a) and interaction detector (b).

factors and direct influencing factors, which is the advantage of this expected to help formulate refined and differentiated carbon emission
study over previous studies. reduction plans, and provide suggestions for achieving carbon peak and
carbon neutrality.
4. Conclusions and policy implications
4.2. Policy implications
4.1. Conclusions
Results of this study have implications for guiding regional low-
This study proposed a method to simulate the spatial distribution of carbon planning. It is essential to propose refined and differentiated
carbon emissions by coupling remote sensing and social sensing, which low-carbon strategies due to the variation of the explanatory power of
can reflect the spatial heterogeneity of carbon emissions more accu­ each socioeconomic feature to carbon emissions. Based on the findings,
rately. This study found that it is feasible to apply large-scale NTL the following recommendations are made: first, as a pillar industry of the
remote sensing and high-precision electricity consumption data to spa­ national economy of Guangzhou, the second industry plays a decisive
tialize carbon emissions. This method avoids the disadvantages of both role in regional development (Chen et al., 2021). However, due to its
in spatial characteristics including the saturation problem of NTL data characteristics of high energy consumption and high pollution, the
and the discretization problem of electricity consumption point data. second industry generated a large number of carbon emissions (Shi
This method is comprehensive and effective because it considered the et al., 2019). The high explanatory power of the second industry hints
activity intensity of the population, the level of economic development, that it has been the major source of carbon emissions, and that
and the differences in energy consumption within the city. This study is Guangzhou should continue to carry out economic development and

13
F. Gao et al. Environmental Research 221 (2023) 115257

reform, further optimize the industrial structure, and attach importance should also be investigated comprehensively for sustainable develop­
to the significant role of industrial upgrading in low-carbon emission ment of cities and environment.
reduction.
Second, UR refers to the proportion of urban construction land area Credit author statement
in the administrative region of the study. The high urbanization rate of a
region was often accompanied by the loss of natural land, the expansion Feng Gao and Jie Wu designed the research; Feng Gao, Xiaohui Li, Jie
of impervious surfaces, and the large-scale infrastructure construction Wu performed experiments and computational analysis; Feng Gao built
on impervious surfaces, which brings a lot of carbon emissions (Ke et al., the database; Feng Gao drafted the paper; Feng Gao and Jie Wu
2022; Tang et al., 2021; Zhu et al., 2022). Given that the explanatory contributed to the Writing – original draft; Jinghao Xiao, Shunyi Liao
power of UR was far greater than other factors, this study suggests that and Wangyang Chen contributed to the Investigation and Validation.
Guangzhou should further strictly control urban land expansion and
protect existing natural land and agricultural land.
Third, the results of interaction model show that on the basis of the Declaration of competing interest
continuous growth of regional economy and population, the further
development of urban construction land and secondary industry is the The authors declare that they have no known competing financial
key direct factor leading to a large amount of carbon emissions. interests or personal relationships that could have appeared to influence
Therefore, the Guangzhou government should vigorously support the the work reported in this paper.
development of green industries and promote the transformation and
upgrading of Guangzhou’s industrial structure. Guangzhou is in a crit­ Data availability
ical period of development and has just entered the stage of improve­
ment. The sustainable development of the city is facing more challenges. The authors do not have permission to share data.
Guangzhou should always adhere to the transformation of its industrial
structure into a knowledge-intensive, high value-added and low-carbon Acknowledgments
industry, actively guide the low-carbon development of the tertiary in­
dustry, improve the overall quality and efficiency of the industrial This work was supported by the Guangdong Enterprise Key Labo­
structure through technological progress, and promote the ration­ ratory for Urban Sensing, Monitoring and Early Warning (No.
alization and advanced development of the industrial structure, so as to 2020B121202019), The Science and Technology Foundation of
achieve the goal of reducing carbon emissions and mitigating the Guangzhou Urban Planning & Design Survey Research Institute
greenhouse effect. (RDI2210202162), National Natural Science Foundation of China
In addition, the level of social consumption is the key to promoting (Grant No. 42271467).
economic development, but it mainly depends on the production scale of
the manufacturing industry. Although social consumption activities may References
not directly produce carbon emissions, they will indirectly increase
carbon emissions by stimulating manufacturing activities around the Anselin, L., 2010. Local indicators of spatial association-LISA. Geogr. Anal. 27, 93–115.
city center (R. R. Yuan et al., 2015; Zhu et al., 2022). Therefore, resi­ https://doi.org/10.1111/j.1538-4632.1995.tb00338.x.
Bertinelli, L., Black, D., 2004. Urbanization and growth. J. Urban Econ. 56, 80–96.
dents need to improve their awareness and education about energy https://doi.org/10.1016/j.jue.2004.03.003.
conservation and emission reduction (Bertinelli and Black, 2004). Cai, M., Shi, Y., Ren, C., 2020. Developing a high-resolution emission inventory tool for
Moreover, it is also important to protect the ecological environment and low-carbon city management using hybrid method – a pilot test in high-density Hong
Kong. Energy Build. 226, 110376 https://doi.org/10.1016/j.enbuild.2020.110376.
improve the carbon sink capacity of the ecosystem (Cui et al., 2020).
Cai, W., Borlace, S., Lengaigne, M., van Rensch, P., Collins, M., Vecchi, G.,
Timmermann, A., Santoso, A., McPhaden, M.J., Wu, L., England, M.H., Wang, G.,
4.3. Prospect Guilyardi, E., Jin, F.-F., 2014. Increasing frequency of extreme El Niño events due to
greenhouse warming. Nat. Clim. Change 4, 111–116. https://doi.org/10.1038/
nclimate2100.
The main contribution of this study was to propose a method to Chen, L., Xu, L., Cai, Y., Yang, Z., 2021. Spatiotemporal patterns of industrial carbon
spatialize carbon emissions by coupling remote sensing and social emissions at the city level. Resour. Conserv. Recycl. 169, 105499 https://doi.org/
10.1016/j.resconrec.2021.105499.
sensing, and the results of this method were better than existing studies
Chu, Y., Xie, L., Yuan, Z., 2018. Composition and spatiotemporal distribution of the agro-
and public datasets. This study shows that the combination of remote ecosystem carbon footprint: a case study in Hebei Province, north China. J. Clean.
sensing earth observation data and high-precision social perception data Prod. 190, 838–846. https://doi.org/10.1016/j.jclepro.2018.04.111.
is helpful to study urban energy consumption and carbon emissions from Chuai, X., Feng, J., 2019. High resolution carbon emissions simulation and spatial
heterogeneity analysis based on big data in Nanjing City, China. Sci. Total Environ.
both macro and micro perspectives. Under the international background 686, 828–837. https://doi.org/10.1016/j.scitotenv.2019.05.138.
of carbon peak and carbon neutrality, efficient methods are urgently Cui, X., Li, S., Gao, F., 2020. Examining spatial carbon metabolism: features, future
needed. This study is expected to provide practical methods for the simulation, and land-based mitigation. Ecol. Model. 438, 109325 https://doi.org/
10.1016/j.ecolmodel.2020.109325.
government and related work. Deng, X., Liu, Y., Gao, F., Liao, S., Zhou, F., Cai, G., 2021. Spatial distribution and
Although this study has its advantages, there are still shortcomings to mechanism of urban occupation mixture in Guangzhou: an optimized GeoDetector-
be solved in the future. First, this study tends to analyze the spatializa­ based index to compare individual and interactive effects. IJGI 10, 659. https://doi.
org/10.3390/ijgi10100659.
tion of carbon emissions of the city, so it is inevitable to lack some de­ Doll, C.N.H., Muller, J.-P., Elvidge, C.D., 2000. Night-time imagery as a tool for global
tails. For example, the carbon emissions of sub industries in each mapping of socioeconomic parameters and. Greenhouse Gas Emissions 29, 7.
industry can be explored in the future, and more detailed results and Drewniok, C., Rohr, K., 1997. Exterior orientation an automatic approach based on
fitting analytic landmark models. Rem. Sens. 14.
enlightenment will be obtained. Secondly, this study aimed to spatially Falahatkar, S., Rezaei, F., 2020. Towards low carbon cities: spatio-temporal dynamics of
explicit carbon emissions by industry at fine scale only in 2019 because urban form and carbon dioxide emissions. Remote Sens. Appl.: Society and
the latest open access Luojia NTL data and non-open access electricity Environment 18, 100317. https://doi.org/10.1016/j.rsase.2020.100317.
Fan, Z., Duan, J., Lu, Y., Zou, W., Lan, W., 2021. A geographical detector study on factors
consumption data available is for 2019. Therefore, future work should
influencing urban park use in Nanjing, China. Urban For. Urban Green. 59, 126996
explore the long-term spatiotemporal characteristics of carbon emis­ https://doi.org/10.1016/j.ufug.2021.126996.
sions by obtaining long time series data. Thirdly, more data acquisition Gao, F., Huang, G., Li, S., Huang, Z., Chai, L., 2021a. Integrating the eigendecomposition
and detailed investigations should be considered to further evaluate the approach and k-means clustering for inferring building functions with location-based
social media data. IJGI 10, 834. https://doi.org/10.3390/ijgi10120834.
accuracy verification of the results in the future work. In addition, the Gao, F., Li, S., Tan, Z., Wu, Z., Zhang, X., Huang, G., Huang, Z., 2021b. Understanding the
relationship between population distribution and carbon emissions modifiable areal unit problem in dockless bike sharing usage and exploring the

14
F. Gao et al. Environmental Research 221 (2023) 115257

interactive effects of built environment factors. Int. J. Geogr. Inf. Sci. 35, 1905–1925. Shan, Y., 2017. Methodology and applications of city level CO2 emission accounts in
https://doi.org/10.1080/13658816.2020.1863410. China. J. Clean. Prod. 11.
Gong, P., Chen, B., Li, Xuecao, Liu, H., Wang, J., Bai, Y., Chen, J., Chen, X., Fang, L., Shan, Y., Guan, D., Hubacek, K., Zheng, B., Davis, S.J., Jia, L., Liu, J., Liu, Z., Fromer, N.,
Feng, S., Feng, Y., Gong, Y., Gu, H., Huang, H., Huang, X., Jiao, H., Kang, Y., Lei, G., Mi, Z., Meng, J., Deng, X., Li, Y., Lin, J., Schroeder, H., Weisz, H., Schellnhuber, H.J.,
Li, A., Li, Xiaoting, Li, Xun, Li, Y., Li, Zhilin, Li, Zhongde, Liu, Chong, Liu, Chunxia, 2018. City-level climate change mitigation in China. Sci. Adv. 4, eaaq0390 https://
Liu, M., Liu, S., Mao, W., Miao, C., Ni, H., Pan, Q., Qi, S., Ren, Z., Shan, Z., Shen, S., doi.org/10.1126/sciadv.aaq0390.
Shi, M., Song, Y., Su, M., Ping Suen, H., Sun, B., Sun, F., Sun, J., Sun, L., Sun, W., Shan, Y., Hu, K., n.d. China CO2 Emission Accounts. Scientific Data 10...
Tian, T., Tong, X., Tseng, Y., Tu, Y., Wang, H., Wang, L., Wang, X., Wang, Z., Wu, T., Shan, Y., Liu, J., Liu, Z., Shao, S., Guan, D., 2019. An emissions-socioeconomic inventory
Xie, Y., Yang, Jian, Yang, Jun, Yuan, M., Yue, W., Zeng, H., Zhang, K., Zhang, N., of Chinese cities. Sci. Data 6, 190027. https://doi.org/10.1038/sdata.2019.27.
Zhang, T., Zhang, Y., Zhao, F., Zheng, Y., Zhou, Q., Clinton, N., Zhu, Z., Xu, B., 2020. Shi, K., Yu, B., Zhou, Y., Chen, Y., Yang, C., Chen, Z., Wu, J., 2019. China: a comparative
Mapping essential urban land use categories in China (EULUC-China): preliminary analysis between the provincial and prefectural levels. Appl. Energy 12.
results for 2018. Sci. Bull. 65, 182–187. https://doi.org/10.1016/j. Tang, X., Woodcock, C.E., Olofsson, P., Hutyra, L.R., 2021. Spatiotemporal assessment of
scib.2019.12.007. land use/land cover change and associated carbon emissions and uptake in the
Guan, D., Klasen, S., Hubacek, K., Feng, K., Liu, Z., He, K., Geng, Y., Zhang, Q., 2014. Mekong River Basin. Rem. Sens. Environ. 256, 112336 https://doi.org/10.1016/j.
Determinants of stagnating carbon intensity in China. Nat. Clim. Change 4, rse.2021.112336.
1017–1023. https://doi.org/10.1038/nclimate2388. Tu, W., Zhu, T., Xia, J., Zhou, Y., Lai, Y., Jiang, J., Li, Q., 2020. Portraying the spatial
Guan, D., Liu, Zhu, Geng, Yong, Lindner, Sören, Klaus, Hubacek, 2012. The gigatonne dynamics of urban vibrancy using multisource urban big data. Comput. Environ.
gap in China’s carbon dioxide inventories. Nat. Clim. Change 2, 672–675. https:// Urban Syst. 80, 101428 https://doi.org/10.1016/j.compenvurbsys.2019.101428.
doi.org/10.1038/nclimate1560. Wang, J., Li, X., Christakos, G., Liao, Y., Zhang, T., Gu, X., Zheng, X., 2010. Geographical
Guan, Q., Cheng, S., Pan, Y., Yao, Y., Zeng, W., 2021. Sensing mixed urban land-use detectors-based health risk assessment and its application in the neural tube defects
patterns using municipal water consumption time series. Ann. Assoc. Am. Geogr. study of the heshun region, China. Int. J. Geogr. Inf. Sci. 24, 107–127. https://doi.
111, 68–86. https://doi.org/10.1080/24694452.2020.1769463. org/10.1080/13658810802443457.
Gurney, K.R., Liang, J., Patarasuk, R., Song, Y., Huang, J., Roest, G., 2020. The vulcan Wang, J.-F., Zhang, T.-L., Fu, B.-J., 2016. A measure of spatial stratified heterogeneity.
version 3.0 high-resolution fossil fuel CO 2 emissions for the United States. JGR Ecol. Indicat. 67, 250–256. https://doi.org/10.1016/j.ecolind.2016.02.052.
Atmospheres 125. https://doi.org/10.1029/2020JD032974. Wang, M., Wang, Y., Wu, Y., Yue, X., Wang, Mengjiao, Hu, P., 2022. Identifying the
Huang, Z., Li, S., Gao, F., Wang, F., Lin, J., Tan, Z., 2021. Evaluating the performance of spatial heterogeneity in the effects of the construction land scale on carbon
LBSM data to estimate the gross domestic product of China at multiple scales: a emissions: case study of the Yangtze River Economic Belt, China. Environ. Res. 212,
comparison with NPP-VIIRS nighttime light data. J. Clean. Prod. 328, 129558 113397 https://doi.org/10.1016/j.envres.2022.113397.
https://doi.org/10.1016/j.jclepro.2021.129558. Wang, X., 2022. Carbon emission accounting and spatial distribution of industrial
Ke, Y., Xia, L., Huang, Y., Li, S., Zhang, Y., Liang, S., Yang, Z., 2022. The carbon entities in Beijing—combining nighttime light data and urban functional areas. Ecol.
emissions related to the land-use changes from 2000 to 2015 in Shenzhen, China: Inf. 14.
implication for exploring low-carbon development in megacities. J. Environ. Manag. Wei, W., Wang, X., Zhu, H., Li, J., Zhou, S., Zou, Z., Li, J.S., 2017. Carbon emissions of
319, 115660 https://doi.org/10.1016/j.jenvman.2022.115660. urban power grid in Jing-Jin-Ji region: characteristics and influential factors.
Li, W., Dong, F., Ji, Z., 2021. Research on coordination level and influencing factors J. Clean. Prod. 168, 428–440. https://doi.org/10.1016/j.jclepro.2017.09.015.
spatial heterogeneity of China’s urban CO2 emissions. Sustain. Cities Soc. 75, Wu, D., Geng, Y., Zhang, Y., Wei, W., 2022. Features and drivers of China’s urban-rural
103323 https://doi.org/10.1016/j.scs.2021.103323. household electricity consumption: evidence from residential survey. J. Clean. Prod.
Lin, J., Luo, S., Huang, Y., 2022. Poverty estimation at the county level by combining 365, 132837 https://doi.org/10.1016/j.jclepro.2022.132837.
LuoJia1-01 nighttime light data and points of interest. Geocarto Int. 37, 3590–3606. Yao, Y., Yan, X., Luo, P., Liang, Y., Ren, S., Hu, Y., Han, J., Guan, Q., 2022. Classifying
https://doi.org/10.1080/10106049.2020.1870166. land-use patterns by integrating time-series electricity data and high-spatial
Lin, J., Wan, H., Cui, Y., 2020. Analyzing the spatial factors related to the distributions of resolution remote sensing imagery. Int. J. Appl. Earth Obs. Geoinf. 106, 102664
building heights in urban areas: a comparative case study in Guangzhou and https://doi.org/10.1016/j.jag.2021.102664.
Shenzhen. Sustain. Cities Soc. 52, 101854 https://doi.org/10.1016/j. Yuan, B., Ren, S., Chen, X., 2015. The effects of urbanization, consumption ratio and
scs.2019.101854. consumption structure on residential indirect CO2 emissions in China: a regional
Liu, L., Zhou, H., Lan, M., Wang, Z., 2020. Linking Luojia 1-01 nightlight imagery to comparative analysis. Appl. Energy 140, 94–106. https://doi.org/10.1016/j.
urban crime. Appl. Geogr. 125, 102267 https://doi.org/10.1016/j. apenergy.2014.11.047.
apgeog.2020.102267. Yuan, R., Zhao, T., Xu, X., Kang, J., 2015. Regional characteristics of impact factors for
Liu, Z., Guan, D., Wei, W., Davis, S.J., Ciais, P., Bai, J., Peng, S., Zhang, Q., Hubacek, K., energy-related CO2 emissions in China, 1997–2010: evidence from tests for
Marland, G., Andres, R.J., Crawford-Brown, D., Lin, J., Zhao, H., Hong, C., Boden, T. threshold effects based on the STIRPAT model. Environ. Model. Assess. 20, 129–144.
A., Feng, K., Peters, G.P., Xi, F., Liu, J., Li, Y., Zhao, Y., Zeng, N., He, K., 2015. https://doi.org/10.1007/s10666-014-9424-4.
Reduced carbon emission estimates from fossil fuel combustion and cement Zhang, Q., Zheng, Z., Wu, Z., Cao, Z., Luo, R., 2022. Using multi-source geospatial
production in China. Nature 524, 335–338. https://doi.org/10.1038/nature14677. information to reduce the saturation problem of DMSP/OLS nighttime light data.
Liu, Zhu, Guan, Dabo, Moore, Scott, Lee, Henry, Su, Jun, Zhang, Qiang, 2015. Steps to Rem. Sens. 14, 3264. https://doi.org/10.3390/rs14143264.
China’s carbon peak. Nature 522, 279–281. https://doi.org/10.1038/522279a. Zhang, W., 2019. Effects of urbanization on airport CO2 emissions_ A geographically
Long, Z., Zhang, Z., Liang, S., Chen, X., Ding, B., Wang, B., Chen, Y., Sun, Y., Li, S., weighted approach using nighttime light data in China 12.
Yang, T., 2021. Spatially explicit carbon emissions at the county scale. Resour. Zheng, B., Zhang, Q., Davis, S.J., Ciais, P., Hong, C., Li, M., Liu, F., Tong, D., Li, H.,
Conserv. Recycl. 173, 105706 https://doi.org/10.1016/j.resconrec.2021.105706. He, K., 2018. Infrastructure shapes differences in the carbon intensities of Chinese
Lu, Y., Liang, Yingran, Lu, Siyan, Xiao, Yue, He, Xiaoyu, Lin, Jinyao, 2022. Spatialization cities. Environ. Sci. Technol. 52, 6032–6041. https://doi.org/10.1021/acs.
of carbon emissions in Guangzhou city by combining luojia1-01 nighttime light and est.7b05654.
urban functional zoning data. Journal of Geo-information Science 24, 1176–1188. Zheng, Z., Chen, Y., Wu, Z., Ye, X., Guo, G., Qian, Q., 2019. The desaturation method of
https://doi.org/10.12082/dqxxkx.2022.210610. DMSP/OLS nighttime light data based on vector data: taking the rapidly urbanized
McMichael, A.J., Woodruff, R.E., Hales, S., 2006. Climate change and human health: China as an example. Int. J. Geogr. Inf. Sci. 33, 431–453. https://doi.org/10.1080/
present and future risks. Lancet 367, 859–869. https://doi.org/10.1016/S0140-6736 13658816.2018.1538516.
(06)68079-3. Zheng, Z., Wu, Z., Chen, Y., Guo, G., Cao, Z., Yang, Z., Marinello, F., 2021. Africa’s
Muntean, M., 2014. Trend analysis from 1970 to 2008 and model evaluation of protected areas are brightening at night: a long-term light pollution monitor based
EDGARv4 global gridded anthropogenic mercury emissions. Sci. Total Environ. 14. on nighttime light imagery. Global Environ. Change 69, 102318. https://doi.org/
Oda, T., Maksyutov, S., Andres, R.J., 2018. The Open-source Data Inventory for 10.1016/j.gloenvcha.2021.102318.
Anthropogenic CO2, version 2016 (ODIAC2016): a global monthly fossil fuel CO2 Zhu, E., Qi, Q., Chen, L., Wu, X., 2022. The spatial-temporal patterns and multiple
gridded emissions data product for tracer transport simulations and surface flux driving mechanisms of carbon emissions in the process of urbanization: a case study
inversions 21. in Zhejiang, China. J. Clean. Prod. 358, 131954 https://doi.org/10.1016/j.
Petit, J.R., Jouzel, J., Raynaud, D., Barkov, N.I., Delaygue, G., Delmotte, M., jclepro.2022.131954.
Kotlyakov, V.M., Legrand, M., Lipenkov, V.Y., Lorius, C., Saltzman, E., 1999. Climate Zou, Z., Yun, Y., Sun, J., 2006. Entropy method for determination of weight of evaluating
and atmospheric history of the past 420,000 years from the Vostok ice core, indicators in fuzzy synthetic evaluation for water quality assessment. J. Environ. Sci.
Antarctica. Nature 399, 429–436. https://doi.org/10.1038/20859. 18, 1020–1023. https://doi.org/10.1016/S1001-0742(06)60032-6.
Raupach, M.R., Marland, G., Ciais, P., Field, C.B., 2007. Global and regional drivers of
accelerating CO2 emissions. Proc. Natl. Acad. Sci. USA 104, 10288–10293. https://
doi.org/10.1073/pnas.0700609104.

15