Monitoring urban carbon emissions from energy consumption over China with DMSP/OLS nighttime light observations

This paper constructs a model to accurately estimate the urban CO2 emissions in 2000, 2005, and 2013 in China, using the combined data of DMSP/OLS nighttime light data and the provincial energy statistical yearbook data. We calculate and analyze the growth of urban built-up areas and carbon emissions in different time periods both all over the country and the four economic zones in China. It was shown a good fitting relationship between urban growth and carbon emissions, with the R2 at 0.6188 in 2000, 0.7132 in 2005, and 0.7195 in 2013. The growth rate of developed land area was 13.4% from 2000 to 2005 and 15.9% from 2005 to 2013. During the same period, CO2 emissions had been increasing as well, at an average annual growth rate of 12.2% from 2000 to 2005 and 6.5% from 2005 to 2013. From a spatial point of view, carbon emissions are far greater in the eastern region of China than in western China. The carbon emissions are the highest in major metropolitan cities such as Beijing, Shanghai, and Guangzhou. Per capita carbon emissions are also higher in eastern China, which is consistent with the people’s higher living standards. In some cities with large energy and heavy industry concentrations, especially in the northeastern and western regions, the growth rate of carbon emissions has risen faster than in other cities.


Introduction
Global climate change has become a critical global issue with severe threats to the environment, economic development, and public health (Zhou et al. 2013). Because of its massive volume, carbon dioxide (CO 2 ), a greenhouse gas (GHG) produced by human activities, has received a lot of attention. Carbon emissions from energy consumption are caused by the burning of fossil fuels, mainly coal, oil, and natural gas, which are the main sources of GHG emissions generated by human activities. Many scholars have conducted in-depth research on carbon energy emission calculations, emission reduction strategies, and influencing factors (Zhang et al. 2011;Dong and Zhang 2010). Chen et al. 2007; Wang and Watson 2010). During the past decades, China has experienced rapid economic growth. It is of great implications to investigate the environmental effects of the dramatic changes in land use and energy consumption for policy making towards sustainable development. Therefore, it is necessary to develop reliable approaches to estimate and analyze carbon emissions in different economic regions and implement strategic plans for carbon emission reduction.
The administrative regimes of China are divided into three levels: provincial level, municipal level, and county level. The statistics on energy carbon emissions in China are usually collected at the national or provincial level by the National Bureau of Statistics and its affiliated agencies. Since detailed city-level data is comparatively rare, it is difficult to analyze the temporal trends and spatial patterns of carbon emissions at a micro-scale. Therefore, more advanced techniques are needed to assess carbon emissions in China. Due to the limited sources of energy data in China, the related research (Zhao et al. 2010;Wang and Zhu 2008;Su 2015) mainly used energy statistics to calculate the energy consumption of CO 2 emissions according to the IPCC standard. Up to now, the most used approaches for calculating carbon emissions in the world are the material balance algorithm, the life cycle method, emission coefficient method, model method, and the method of decision tree of real objects (IPCC 1990(IPCC , 2001(IPCC , 2006(IPCC , 2007Jarvis, et al. 1997;Lee 1998;Malhi et al. 1998;Wang and Gu 2006;Che 2010). Based on detailed fuel classification, Schimel (1995) estimated the amount of global anthropogenic CO 2 . Their results show that the consumption of fossil fuels and cement production in the 1980s accounted for 78% of the world's total anthropogenic CO 2 emissions. Park and Hong (2013) adopted the fuel CO 2 estimation method provided by IPCC and estimated the seasonal emission of Korean energy CO 2 using the energy consumption data provided by Korea's energy statistics system and conducted correlation analysis with economic and energy data. Iihan and Ali (2010) analyzed the long-term causality between Turkey's carbon emissions and economic growth, energy consumption, and employment, using CO 2 data provided by the World Development Institute with an auto-regressive co-integration method. Biogeochemical models are ususlly used to simulate the processes of the ecological carbon cycle to obtain the flux of GHG, such as the MS-MRT model used in the Kyoto Protocol and the SDA model used in the USA for carbon emissions.
On the account of the research background and knowledge gap, this paper constructed a model to estimate and calculate the municipal carbon emissions of China using nighttime lighting data and provincial-level carbon emissions data.

Study area and data sources
The study area covers mainland China, excluding Hong Kong, Macao, and Taiwan. The spatial distribution of CO 2 emissions in mainland China was analyzed respectively at the provincial and municipal levels.
Five main types of data sets were used in this paper: (1) Nighttime lighting data from the DMSP/OLS in 2000, 2005, and 2013. This data was obtained from the NGDC (National Geophysical Data Center), a unit of NOAA (National Oceanic and Atmospheric Administration) of the USA. The data provides a spatial resolution of 30 arc seconds including stable light from cities and towns with human activities. The instantaneous light of transient events such as fire and abandoned combustion has been eliminated. Because there are a few differences and some noises, it is also necessary to adjust these night light data for de-noising, cutting, relative radiation calibration, and geographic coordinate conversion. The data is used to get light brightness values and extract urban and urban built-up areas.  Figure 1 presents the overall research ideas and the technical route of this paper. In this study, we use night lighting data to extract urban built-up areas in China and verify the accuracy of the extraction with TM data, to provide a basis for counting the lighting brightness values in Chinese provinces and municipalities.

Methods
The detailed analyzing process is as followed: Firstly, nighttime lighting data is processed, including image projection conversion, re-sampling, and cropping. The original images without radiometrically calibrating are mutually corrected, and the series of images are subjected to steps of year-fusing and inter-annual correction. In the mutually corrected step, we recalculate the new DN values of each raster in China annually by processing its DN value from the raw data (in the same year) with a group of parameters obtained from the regressions between the DN values extracted from F16 in 2006 and values from other series of light data in the "benchmark city". There are 33 groups of parameters since 34 series of raw data existed from 1992 to 2013. The benchmark city here is Hegang in Heilongjiang province, the reason for which has been elucidated by scholars in remote sensing fields (Cao et al. 2015). Secondly, the carbon emissions of each province are calculated using the statistical yearbook data. Thirdly, we construct an urban carbon emissions estimation model based on regression analysis of carbon emissions and lighting brightness in each province. Finally, the characteristics of urban carbon emissions, urban expansion, and provincial carbon emissions in spatial-temporal perspective are analyzed.

Urban area extraction
According to Wang et al. (2019), the urban area to be extracted can be divided into three parts: the demarcation zone, the urban area outside the demarcation zone, and the urban area within the demarcation zone. After obtaining the urban area and the area within the demarcation zone, we can then calculate the whole urban area of each city. The whole urban area of each city is composed of two parts: one is the urban area within the demarcation zone, and the other is the main urban area outside the demarcation zone. The whole urban area can be expressed as Eq (1):

Construction of the municipal energy carbon emission model
Similar to the related literature (Elvidge et al. 1997;Doll et al. 2000;Raupach et al. 2010;Meng et al. 2014;Su et al. 2014), a city-level CO 2 emission estimation model can be constructed if the nighttime light correlates to the total yearly carbon emissions of all provinces. The inversion model is shown in Eq. (2): where E i is the consumption of energy i, according to the standard coal unit, 10 4 t; K i is the coefficient of carbon emission of energy i, (10 4 t carbon)/(10 4 t standard coal); i is the type of energy; and the value of K i is calculated according to the default value of the IPCC Carbon Emissions Calculation Guidelines.

Accuracy verification results
We further accurately calibrated the extracted urban area in this paper. Given the relatively high spatial resolution of Landsat TM images (30 m × 30 m), scholars generally believe that the construction of map spots based on TM image data extraction is a reliable verification data source (Cao et al. 2009;He et al. 2006). Su et al. (2015) used TM image data to calibrate the extract urban areas based on DMSP/OLS nighttime light data, and their results showed a significant correlation between the total number of pixels of the urban area extracted by the two datasets.
In this paper, the accuracy of the DMSP/OLS nighttime lighting data was calibrated by using the TM data image map of Beijing in 2005. The verification results presented in Fig. 4 suggest a highly consistent relationship between the urban areas extracted respectively by light data and TM data. The difference between the numbers of pixels is 528, and the accuracy reaches 90.3%. Therefore, it is scientifically feasible to use the nighttime lighting data neighborhood analysis method to extract China's urban areas. the increase has been relatively rapid, with an average annual growth rate of 12.1%; from 2005 to 2013, the growth rate has slowed down, with an average annual growth rate of 6.5%.

Correlation analysis of energy carbon emission and nighttime light brightness
A correlation analysis of provincial carbon emissions and provincial nighttime light brightness was conducted to explore whether there is a correlation between nighttime light brightness and carbon emissions. The fitted results shown in Fig. 6 suggest a clear linear relationship between the nighttime lighting data and the CO 2 emission statistics, with R 2 moving between 0.62 and 0.72. These results are consistent with the conclusion that there is a linear relationship between night light data and CO 2 at the global level, national level, provincial level, and some urban levels.

Municipal carbon emission results
According to the above fitting analysis, the provincial nighttime light brightness linearly correlates with the CO 2 emissions. Through provincial and municipal administrative division maps of China, zone statistics were applied to the nighttime lighting data, so we can obtain statistics on the nighttime lighting data of more than 300 cities in China in 2000, 2005, and 2013. By using the municipal energy carbon emission model in From the perspective of regional heterogeneity, most of the CO 2 emissions were concentrated in the eastern regions of China. After 2000, China's economic development began to switch towards harmonious development in each region, and therefore, carbon emissions began to grow steadily in the eastern and western regions. The eastern region optimized its economic structure and devoted itself to improving the quality of development. At the same time, the economic growth rate and public policies in the central region were relatively stable, and its carbon emissions had stabilized. For the western region, carbon emissions were relatively low. Since the beginning of the western development policy, Fig. 6 Fitting relationship between the total value of light data at night and the CO 2 emission statistics regional carbon emissions began to rise slowly over the period. Though the western economy did not grown rapidly due to insufficient funds and imperfect infrastructure, carbon emissions in western China also increased in 2005 and 2013 compared to 2000. As for northeastern China, since the 1990s, due to the gradual decline of the old industrial bases, the gap between the northeastern region's economy and the developed eastern coastal regions continuously expanded. Since 2003, after the country officially began to make important strategic decisions for the rejuvenation of the industries in that region, the CO 2 emissions in northeastern China began to rise rapidly. Regional GDP was one of the factors influencing the total amount of carbon emissions in different regions.
To further analyze the temporal and spatial changes of CO 2 emissions in cities in, the annual average growth of each city from 2000 to 2013 was calculated and classified. There were five types of energy sources of city CO 2 emissions all over China. The criteria for its classification were shown in Table 2, including the slower growth type, the slow growth type, medium-speed growth type, the fast growth type, and the faster growth type. The results showed that 9 faster growth cities, and 17 fast growth cities, which were mainly concentrated in the eastern areas with comparatively well-developed economy (Fig. 8). In addition, 89 cities were classified as slow growth types, and 211 cities as slower growth types, which were mainly concentrated in the less developed regions such as the western region and the northeastern and central regions (Fig. 8). It can be concluded that the growth rate of total CO 2 emissions in different regions of China is closely related to its economic development speed and degree of development.

Analysis of per capita carbon emissions
From the above analysis of the total amount of carbon emissions, due to differences in the status, policies, and industrial development structures of economic development during different time periods, the total amount of CO 2 emissions in the four major economic zones, the eastern, central, western, and northeast regions of China, present different characteristics. The eastern region, the most densely populated and best economically developed region, has released the most CO 2 emissions.
Further examination of the four major economic regions' per capita carbon emissions reveals that between 2000 and 2013, per capita carbon emissions increased. Due to better economic development and higher living standard in the eastern region, which has a greater total amount of energy consumption, the per capita carbon emissions are slightly higher. Because the northeastern region is dominated by heavy industries, the per capita carbon emission level is relatively higher due to its low energy efficiency and huge amount of consumption. Due to the much smaller population and higher energy consumption, the per capita carbon emissions in some central regions are even higher than in the eastern region. Per capita carbon emissions from 2000 to 2013 are shown in Fig. 9. The per capita carbon emission growth rate was 11.8% from 2000 to 2005, and 5.2% from 2005 to 2013. The calculation is shown in Eq. (4).

Discussions
By using DMSP/OLS night light remote sensing data to estimate city-level energy carbon emissions, the research and methodology in this paper not only overcome the shortage of energy carbon emission statistics in prefecture-level cities but also unify the methods of carbon emission assessment at provincial and municipal levels, which can help in making more reasonable carbon emission reduction policies. There are a few studies on energy carbon emissions using DMSP/ OLS night lighting data, most of which are at the global or national level, and very few is at the provincial or municipal level. The methods for estimating carbon emissions for municipal energy are still in their infancy. An important research extension would be to validate the estimated results with actual carbon emissions data on urban energy consumption. Equation (2) is established based on the high (4) Per capita carbon emissions = Total carbon emissions / Total population correlation between lighting brightness values and carbon emissions. Equation (2) has an implicit assumption that the ratio of CO 2 emission to light intensity is spatially homogeneous within a province. Using the ratio, this study realizes the estimation of municipal carbon emissions from the provincial units. The limitation of this model is the assumption that all cities within a province follow the proportion, but the cities in the province of carbon ratio may be a spatial heterogeneity. That is to say, in different cities of carbon proportion, light brightness may be different. If there is a municipal level of carbon emissions statistics, we can do further research.

Conclusions
In this paper, we constructed an estimation model for city carbon emissions to overcome the difficulty of collecting statistical data at the municipal level in China. The main conclusions can be summarized as follows: (  (3) By constructing a city-level carbon emission inversion model, city-level CO 2 emissions were obtained. In general, during 2000 to 2013, the CO 2 emissions were on the rise in China. It grew from 4.30 billion tons in 2000 to 12.64 billion tons in 2013, with an average annual growth rate of 8.6%. From 2000 to 2005, the growth rate was rapid with an average annual growth rate of 12.2%. From 2005 to 2013, the growth rate slowed down and became less volatile to 6.5%. The growth rate of carbon emissions varied in different regions of China. Faster growth types and fast growth types of cities were mainly concentrated in the more economically developed regions in eastern China. In conclusion, the regional distribution of carbon emissions is characterized by "high concentration in the east and low concentration in the west." In addition, carbon emissions per capita were also increasing year by year. The eastern region was far higher than the western region. (4) The annual average growth rate of cities from 2000 to 2013 can be divided into 5 types. The results show that there are 9 faster-growing cities and 17 fastgrowing cities, mainly concentrated in some eastern developed areas. In addition, 89 slow-growing cities and 211 slower-growing cities are found mainly in the less developed regions such as the western region, the northeastern and central regions. The growth rate of total CO 2 emissions in different regions of China is closely related to the economic development speed and degree of development of those regions.
The results have important implications for the emission reduction policy in China. The focus of emission reduction should be on the improvement of energy efficiency and capacity utilization in the heavy industrial cities, such as those in the western and northeastern provinces. For cities in the eastern and central regions that are dominated by light industry, the focus of their emission reduction should be on the adjustment of industry structure.