1. Introduction
The urbanization process has spurred the consumption of energy and resources, which leads to climate change and environmental degradation, and has a significant impact on the natural environment and human production lifestyles [
1]. Existing studies have shown that cities are the main sources of anthropogenic CO
2 emissions, with the amount accounting for more than 75% of global CO
2 emissions [
2]. Urban expansion exacerbates vulnerability to climate change, while the advanced technologies and wealth that accompany urban development can be used to improve the capability for climate change mitigation and adaptation. Therefore, cities should play a key role in tackling climate change [
3].
To quantify CO2 emissions scientifically and accurately in urban areas is an important topic in the research of urban climate change. Recently, the needs for high-resolution carbon emissions data have risen due to both scientific and policy-related reasons.
In terms of the scientific needs, high-resolution carbon emissions data is the prior data for the inversion of observation data. At present, scientists can obtain remote sensing data of carbon concentration through satellites, but they still need to use the atmospheric inversion method to translate the observations from the satellite to the distribution of carbon fluxes that best fit the actual distribution. It is the distribution of carbon fluxes that can help scientists better understand the mechanisms of the global and regional carbon cycle and the carbon exchange among the atmosphere, the sea, and the land. The existing observation data and the atmospheric transport model are the two basic elements for the inversion method to simulate the spatial distribution of carbon concentration in a given area. However, due to the limited number of observation points, other prior information is needed to reduce the uncertainty of the inversion results. The high-resolution carbon emissions data based on regional carbon emissions inventory can be used as the prior data.
As for policymakers, there are two needs. Firstly, the high-resolution carbon emissions data combined with satellite observations can provide policymakers with a scientific and independent way to verify the effects of carbon mitigation policies. Even though this need has been concentrated at the global or national scale in the past, as cities become more important in the field of climate change, the need for high-resolution carbon emissions data at the urban scale has also increased. Secondly, high-resolution carbon emissions data can provide policymakers with more information than the traditional carbon emissions inventories, such as spatial information. Spatial attributes are one of the basic elements of the urban planning system. Wang et al. [
4] demonstrated that urban spatial planning has a great impact on its carbon footprint, and that more work needs to be done to integrate the spatial planning with carbon mitigation. High-resolution carbon emission data can directly build the link between carbon emissions and urban spatial forms, which will raise the awareness of policymakers to integrate carbon reduction policies with urban planning.
Quantitative methods for CO
2 emissions include the mass balance method, the ground-based monitoring method, and the satellite observations method [
5,
6]. The mass balance method is the most commonly used estimation method, and it has been adopted by many studies to estimate the CO
2 emissions of different cities. Based on urban energy statistics, those studies compiled an energy-related CO
2 emissions inventory and then further analyzed the relationship between CO
2 emissions and other social–economic factors, such as the speed of economic development and the changes in industrial structure [
7,
8,
9]. These studies can help policymakers understand the amounts of CO
2 emitted in cities and find the driving forces of the emissions. However, due to the lack of accuracy, the results still contain large uncertainty [
10]. What is more, due to the lack of information about the spatial attributes, it cannot demonstrate the spatial differences in CO
2 emissions within the urban domain.
There are difficulties surrounding using the ground-based monitoring method to monitor the overall carbon emissions of a whole region due to the limited number of observation sites [
11]. Generally, the data observed by the ground-based monitor is used as a priori data for carbon flux inversion methods to aid in the adjusting of estimation.
In recent years, with the number of CO
2-monitoring satellites increased, there have been research achievements regarding satellite monitoring data. Satellite observations provide a more independent, broad, and efficient means of monitoring atmospheric CO
2 emissions [
12], and many studies used atmospheric measurements to produce temporal and spatial CO
2 data at city scales [
13,
14]. Scholars from the United States, France, Japan, and other countries have compared the data from CO
2 inventories with the satellite inversion data for those countries based on their CO
2 satellite data. The results conclude that CO
2 inventory data is consistent with satellite observations [
15,
16]. Nevertheless, studies using the satellite monitoring method are still subject to many uncertainties, including the mismatch between the resolution of the transport model, the spatial variability of the actual carbon fluxes and concentrations, and the uncertainties of parameters input into the inversion model [
17].
Given the need for scientific, independent, continuous, and reliable data on CO
2 emissions and the critical importance of data verification to the study of climate change, integrating the three CO
2 quantification methods has become a trend in current academic research. Researchers used the atmospheric inversion method to translate the observations from the satellite to the optimized distribution of carbon fluxes concentration that best fit the actual distribution. The atmospheric inversion method uses a statistical method, which relies on the uncertainty in the prior estimate of emissions and other observation errors, to adjust and produce an optimized posterior estimate. Meanwhile, the mass balanced method and the ground-based observation method combined with the advanced technology of the Geographic Information System (GIS) are used to produce high-resolution carbon emission data. The combined method has been developed to determine the spatial and temporal distribution of CO
2 emissions at the global and regional level [
18].
At the city level, high-temporal resolution CO
2 emissions inventories have been established as well. For instance, the US Vulcan project established an inventory of CO
2 emissions of the different sectors of US cities for each hour in 2002 [
19,
20]. Based on the Vulcan project, the Hestia project further improved the spatial and temporal resolution of the previous inventory data and attempted to develop a data system for CO
2 emissions on the scale of urban buildings, streets, and manufacturing facilities. A high-resolution carbon emissions data system has been constructed for four cities; these are Indianapolis, Indiana [
21], Los Angeles, California [
22], Salt Lake City, Utah [
6], and Baltimore, Maryland. Based on the high-resolution energy-related CO
2 emissions data, the Indianapolis Flux Experiment (INFLUX) project was conducted by integrating the data of the high-resolution CO
2 emissions inventory, tower monitoring, and aircraft monitoring to explore ways to reduce uncertainties and establish the Measurement, Reporting, and Verifying (MRV) system. The results indicate that high-resolution energy-related CO
2 inventory data based on the bottom–up method, combined with the atmospheric inversion method, can be used to more accurately predict regional CO
2 emissions trends [
22,
23].
Compared with developed countries, due to the lack of data, studies of high-resolution CO
2 emissions spatial gridded data systems in China are relatively rare. In 2016, China launched a carbon satellite, named the TanSat, and relevant research has already been carried out [
24]. Wang et al. [
25] established a 10 km
10 km CO
2 emissions map to reveal the spatial distribution of CO
2 emissions in China based on the point-source data of CO
2 emissions from power plants using the data derived from the First China Pollution Source Census (FCPSC). This research mainly introduced the methodology of mapping the high-resolution CO
2 emissions map and deduced the characteristics of CO
2 emissions in China. Cai et al. [
26] created a China high-resolution emission gridded database from 288 Chinese cities and analyzed different characteristics of city CO
2 emissions by their locations and development states. This research revealed different driving factors and the degree of their impact on city CO
2 emissions in China’s different cities in various area regions. The study suggested that cities’ low carbon development strategies should adapt to local conditions. However, this study mainly analyzed the intercity differences of CO
2 emissions, and rarely analyzed the differences of CO
2 within the city. The internal differences of CO
2 emissions reveal the heterogeneous nature of the local situation, such as the urbanization state and land-use activities. The reasons behind these differences provide direction for urban policymakers to improve the process of transitioning to a low-carbon future.
This study takes Shanghai as a case study for several reasons: First, Shanghai is one of the world’s biggest cities, with an area of 6340 square kilometers and a population of more than 24 million in 2017. These characteristics make Shanghai a great representation of urban carbon emissions. Second, Shanghai’s statistic data are well organized and can fulfill the needs of establishing high-resolution CO2 emissions data. Third, Shanghai has been accelerating the speed of energy transition and low carbon development since 2010. In 2012, Shanghai was selected as the pilot city of low carbon development; then, in 2013, it pledged to cap its coal consumption by 2017. The effects of this transition still need to be verified.
Based on the previous study on Shanghai’s 1 km × 1 km high-resolution energy consumption and carbon emissions gridded data system in 2010 [
27], this study extended the data of energy consumption and its related CO
2 emissions from 2010 to 2015. Through the time-series gridded data, this study aims to answer the following questions: (1) What are the temporal and spatial characteristics of Shanghai’s CO
2 emissions? (2) What factors impact the energy-related CO
2 emissions in Shanghai? (3) Is there any effect of the carbon mitigation policy implemented by the Shanghai government since 2010? (4) What can be done for the policymakers to mitigate carbon emissions in the future?
This paper includes the following four parts. The first part introduces the research methods used in this study, including the method of creating, adding, and updating gridded data from 2010 to 2015 and the mechanism of the geographic detector. In the second part, we elaborate on the three spatial characteristics of Shanghai CO2 emissions. We analyze the impact mechanism of Shanghai CO2 emissions from the perspectives of large point sources and area sources in the third part. In the last part, we conclude some findings and policy recommendations.
4. Impact Mechanisms of CO2 Emissions
Based on the characteristics of emission sources, we divided Shanghai’s ffCO2 emissions into two categories, which we named the large point sources and area sources. The large point sources indicate those large-scale energy infrastructures, such as coal-fired power plants, steel mills, refineries, and airports. In 2015, the total ffCO2 emissions estimated via the point source, the line sources, and the area sources were 201.7 Mt, and the large point sources located in 57 girds emitted 129.90 Mt of CO2, accounting for 64.4%. The area sources include those “low-quantity and wide-area diffused CO2 emissions sources” (e.g., emissions from natural gas boilers) and linear sources (e.g., vehicle fuel road emissions). In 2015, CO2 emissions from area sources accounted for 35.6% of total gridded emissions. The impact mechanisms of the two CO2 emission sources are different; we will discuss this in the following part.
4.1. Impact Mechanisms of CO2 Emissions from Large Point Sources (LPS)
Large point sources (LPS) for ffCO2 emissions in Shanghai are the leading factors shaping the temporal and spatial characteristics. Among them, the coal-related CO2 emissions of LPS are mainly generated from five power generation bases in Shanghai, namely, Waigaoqiao, Shidongkou, Wujing-Minhang, Caojing, and Lingang. Coal consumption in Shanghai reached a peak of 61.49 Mt in 2011, and coal consumption by power plants reached its maximum of 36.67 Mt at the same time. Since Shanghai’s GDP growth has continued to slow down and the imported power from other provinces has increased substantially, Shanghai’s local coal-fired power generation and coal consumption have decreased rapidly. Coal consumption in 2015 was more than 25% less than it was during the peak year of 2011. According to the 13th Five-Year Plan of Shanghai for energy development, the newly increased demand for electricity in Shanghai in the near future will be met by the power imported from other provinces or will be generated from natural gas, which indicates that the impact of key sources of coal-related CO2 emissions is locked in, and can be clearly predicted and controlled.
The oil-related CO
2 emissions of LPS are mainly from the three petrochemical parks and the two airports (see
Figure 1a). In 2015, the total consumption of oil from those chemical parks reached 9.09 Mt, accounting for 81.6% of the total oil consumption in the industrial sector. Given that the Jinshan Petrochemical and Shanghai Chemical Industry Park are located on the edge of Hangzhou Bay, and the coast of Hangzhou Bay is planned to be one of the national oil refining and petrochemical bases, Gaoqiao Petrochemical will be moved to the coastal areas of Hangzhou Bay as well. Besides, the two airports will remain the current location for a long time. Thus, it is safe to conclude that the large point sources of oil-related CO
2 emissions will be located on the north shore of Hangzhou Bay, Hongqiao Airport, and Pudong Airport in the long term. The geographical information on all of LPS is shown in
Figure 1a.
Besides, Shanghai launched its pilot carbon emission trading system (ETS) in November 2013. The ETS covered all the LPS by 2015. Hence, the government can take advantage of the ETS to mitigate CO2 emissions from LPS. First, the government can obtain information on energy consumption and CO2 emissions from these LPS through their annual CO2 emissions report, which will help policymakers track the trend of Shanghai CO2 emissions. Second, the government can encourage LPS to reduce CO2 emissions by adjusting the volume of free allocation. Therefore, the ETS will enhance the downward trend of CO2 from LPS.
Therefore, we conclude that CO2 emissions from LPS in Shanghai has been strictly controlled by the government, and its trend is controllable and predictable. The policies aiming to control coal consumption that have been implemented since 2012 have had positive effects on the reduction of CO2 emissions. This transition also demonstrates that the government can have a great influence on the spatial layout of urban’s CO2 emissions by implementing targeted policies to mitigate ffCO2 emissions from large point sources.
4.2. Impact Mechanisms of CO2 Emissions from Area Sources
Compared with CO2 emissions from LPS, there are more complicated factors that affect the spatial pattern of CO2 emissions from area sources. ffCO2 emissions from area sources have increased from 49.51 Mt to 71.84 Mt between 2010–2015, and its share in the total gridded ffCO2 increased from 23.3% to 35.6%. This trend shows the increasing impact of ffCO2 from area source on the total ffCO2 emissions in Shanghai. In this study, we applied the Geo Detector method and tools to identify and analyze the drivers that affect the spatial pattern of ffCO2 emissions from area sources.
According to the specific situation of Shanghai, this study selects five impact factors controlling CO
2 emissions from area sources, including population density, human activity intensity, land-use type, energy efficiency, and urban planning. Following the requirements of the Geo Detector model, we transformed each factor into a categorical variable. Continuous variables such as population density, different land types, and energy consumption intensity are classified into categorical variables according to the natural breaks classification method of GIS. Population activity intensity consists of six categories according to population density at the daytime and nighttime, and their ratio of daytime-to-nighttime at the district level. The population density is ranked by high, middle, and low levels, and the ratio of daytime-to-nighttime population is divided into two levels: namely, the high ratio of the daytime-to-nighttime population, and the low ratio of the daytime-to-nighttime population. The meaning and data sources of all the factors are shown in
Table 4, and the spatial distribution of those variables is shown in
Figure 6.
We used the factor detector from the Geo Detector software to analyze the determinant power of the 10 identified factors that are responsible for the spatial variability in the area sources’ CO
2 emissions in Shanghai. The spatial strata of CO
2 emissions from area sources are the dependent variable Y, and the spatial strata of identified factors are the independent variable X. The q-statistic method in the Geo Detector software is used to measure the degree to which variable X can explain the spatially stratified heterogeneity of Y. The statistical significance is tested at the level of
p ≤ 0.05. The q and
p values of each factor are shown in
Table 5.
The results show that the explanation of all the factors on CO2 emissions from area resources is statistically significant. Urban circular layers (X10) and population density (X1) are the most influential factors in 2010, with q values of 0.32 and 0.30, respectively. These are followed by the population activity intensity factor (X2), with a q value of 0.28. Each of these three factors can explain around 30% of Shanghai’s area sources’ CO2 emissions, and are the leading factors of surface-derived CO2. Furthermore, the values of q for traffic land (X5), constructed land development intensity (X3), and city-level types (X9) ranged from 0.19–0.23, indicating that these factors also have strong determinant power for area sources’ CO2 emissions. These results demonstrate that urban planning can influence the urban CO2 emissions from area sources.
Compared with 2010, the q values of all the factors are higher in 2015, except for industrial land (X7), which indicates that the determinant power of this factor is significantly enhanced. Among them, the top three factors are still circular layers (X10), population density (X1), and population activity intensity (X2), and this consistency demonstrates the locking effect of these leading factors. The increase in the q value is the largest for energy consumption per unit of output value (X8), increasing from 0.11 in 2010 to 0.28 in 2015, indicating that the policies aiming to adjust the structure of the economy have had an effect on the spatial distribution of CO2 emissions from area sources from 2010 to 2015. However, the explanatory power of industrial land is still weak. We think that this weak relationship is reasonable, as Shanghai’s industrial land planning intends to concentrate the dispersed industrial land to the industrial park, which will increase the trend of spatial clustering of industrial land. This trend is different from the dispersive character of ffCO2 emissions from area sources.
Besides those leading factors, we find that different types of land use have limited power to determine the spatial distribution of CO
2 emissions from area sources. In 2015, the determinant power of traffic land (X5) is the highest (q = 0.26) among the four types of construction land, followed by public building land (X6, q = 0.18). The determinant power of industrial land (X7) is the lowest (q < 0.1). This conclusion derived from our study is different from those of many previous studies, which suggests the changes in land-use types have a strong relationship with urban carbon emissions [
46]. We conclude the reasons that caused this difference are: First, the principle of statistical correlation is different. The Geo Detector mechanism we used focused on analyzing the spatial heterogeneity of two variables, and assumed that if the spatial distribution of the two variables tends to be consistent, there is a statistical correlation between them. In contrast, the regression analysis that the other studies used is based on the values of different variables. Spatial distribution can reveal the difference of a variable by way of spatial heterogeneity, while the value used in the regression model can only represent the total value. Therefore, even if the total value is consistent, when the internal difference is huge, the correlation results will be different. Second, the urban development is unbalanced. As the high-resolution CO
2 emissions products provide more specific spatial information about types of land use and quantitative CO
2 emissions, the urban unbalanced development characteristics can be detected. The results show that the factor of industrial land has greater power in the area of the Outer Ring (q = 0.14), while it has little determinant power (q = 0.03) in the Inner Ring and Middle Ring area.
Given that Shanghai’s ffCO
2 emissions show a significant circular layer structure, we continue to explore the q values of the factors that affect the spatial pattern of area sources’ CO
2 emissions from Shanghai’s different circular layers. The circular layers consist of three parts, namely the area within the Inner Ring (Inner Ring), the area between the Inner Ring and the Outer Ring (Middle Ring), and the area outside the Outer Ring (Outer Ring) in 2015. The results are shown in
Table 6.
The factor that is statistically significant (p ≤ 0.05) and has a larger q value (q ≥ 0.1) is the leading factor in determining the spatial pattern of CO2 emissions within a given circular area. The results show that the leading factors of each layer are very different, especially in the areas outside the Outer Ring compared to the other two areas. The energy consumption intensity (X8), population activity intensity (X2), and population density (X1) are the leading factors for CO2 emissions within the areas of the Inner Ring and the areas between the Inner Ring and the Outer Ring. In contrast, the constructed land development intensity (X3), traffic land (X5), and industrial land (X7) are the leading factors in CO2 emissions for areas outside the Outer Ring.
The determinant power of the same factor may have different effects on the area sources’ CO2 emissions in different circular layers. The q values of population density, population activity intensity, and energy consumption intensity declined from the inside to outside layer by layer, indicating that these factors have stronger determinant power on the spatial pattern of CO2 emissions in the central city than in external areas. The increasing q values of industrial land, public building land, and traffic land from inside to outside shows that the type of land use has a stronger determinant power in the outside area than it does in the inside area.
This variation demonstrates that the determinant power of key factors will change within the urban areas, and that the leading factors may change between a local area and a regional area. We think that most of the variation can be explained by the unbalanced development of the city. The urban circular layers’ structure is the external form of the city’s unbalanced development. In the inner area, the development of urbanization is maintained at a high level. The CO2 emissions in this area are mainly driven by factors related to the population, including population density and human activity. Besides, the factor of energy consumption intensity is also the determinant factor, indicating that improving energy efficiency is an important measure to reduce urban CO2 emissions, even in the most developed areas of the city. From the perspective of policy, first, the government should implement policies aiming to decrease the intensity of CO2 emissions of human activity. For instance, these policies can include increasing building energy efficiency standards and thus decreasing CO2 emissions from the residential and commercial sectors; encouraging the development of public transportation to decrease the use of private cars; and encouraging the mix of the residential and commercial areas to decrease commuting needs. Second, the removal of industries with high energy consumption and high labor intensity in the Inner Ring area will also have a great impact on carbon reduction. In the Outer Ring area, the dominant factors are different. Factors such as urban development intensity, traffic land, and industrial land, which have strong positive relationships with urban growth, have stronger power to determine CO2 emissions in this area. Therefore, urban planning in this area should focus on industrial structure adjustment and low-carbon manufacturing. Policies aiming to restrict the development of high energy consumption projects and increase the use of renewable energy should be encouraged. In conclusion, the concept of low carbon development should be integrated into the urban planning of the area outside the outer ring in different ways. The high resolution of carbon emissions data can be a useful tool to help policymakers locate the areas where low carbon development measures should be implemented.
5. Conclusions and Outlook
This study uses Shanghai as a case study, showing that high-resolution energy consumption and spatially gridded CO2 emissions data are useful in the management of urban CO2 emissions.
Firstly, a high-resolution spatially gridded urban dataset can directly display the spatial pattern of urban CO2 emission and facilitate our understanding of the imbalance of urban development. Based on this dataset, we conclude that high spatial agglomeration, CO2 emissions centralized along the river and coastline, and the three circular layers structure are the three notable temporal–spatial characteristics of Shanghai ffCO2 emissions.
Secondly, a high-resolution spatially gridded urban dataset can help policymakers identify the leading impact factor of urban CO2 emissions in a given area within the city. Thus, a more targeted policy that adapts to local conditions can be made to promote their low-carbon development. We divide sources of CO2 emissions in Shanghai into two categories: namely, the large point sources and the area sources. The impact mechanisms of the two different categories are discussed and concluded as following.
Large point sources of CO2 emissions have a strong locking effect on the spatial pattern of CO2 emissions in Shanghai, and limited changes in the energy consumption of some of the grids would not change the current distribution. Based on the trend analysis of the large point sources, we can accurately predict the trends of total energy consumption and CO2 emissions in Shanghai over the long term. Energy consumption in Shanghai entered a steady growth stage in the last five years. In the future, the total energy consumption may enter a saturation period, but the structure of energy consumption will change dramatically. Currently, the Shanghai energy system is under transformation; the implementation of many actions such as the closure of small and medium-sized coal-fired boilers, the reduction in coal-fired power generation, the relocation of steel plants, and the construction of a modern service-oriented city will directly influence the total energy consumption and CO2 emissions. Even though it is hard to predict the changes of CO2 emissions from each grid in the short term, the trends of CO2 emissions for the longer term are predictable and controllable. For instance, we conclude that due to the continuous reduction of coal consumption in coal-fired power plants since 2011, the decline of CO2 emissions is recognized as a long-term trend. The proportion of the relevant grids’ emissions will decrease year by year, but the absolute amount is still large. Due to the relocation of Shanghai’s petrochemical industry and the enhancement of the spatial agglomeration of shipping centers and trade centers, oil consumption in Shanghai will continue to increase in the future. Therefore, the decline in oil-related CO2 emissions in 2013 is a short-term phenomenon. CO2 emissions from the grids that cover Hongqiao and Pudong Airports will account for larger proportions of the total long-term gridded CO2 emissions. Based on those trends, we expect Shanghai’s total CO2 emissions to reach their peak before the total energy consumption.
However, for policies that target the reduction of energy and coal consumption of large point sources so as to reduce the total urban ffCO2 emissions, their effects will gradually weaken. In order to promote Shanghai’s low carbon development, the local government should pay more attention to the ffCO2 emissions from area sources. The share of ffCO2 emissions from area sources in the total gridded ffCO2 emission in Shanghai has increased by 12.3% between 2010–2015, and this rising trend will continue.
Compared to the clear trend of CO2 emissions from the large point sources, the mechanisms of CO2 emissions from area sources are more complex. In this study, we applied the Geo Detector method to detect the drivers that have determined the configuration of energy-related carbon emissions in Shanghai. The results show that the urban circular layer structure is not only an important boundary of the layout of urban and industrial functions, but is also an important factor leading the spatial pattern of urban CO2. The CO2 emissions of Shanghai already had a clear structure of circular layers in 2010, and the circular layers had an even stronger determinant power of the spatial distribution of CO2 emissions in Shanghai by 2015. As the circular layers structure is a result of early urban planning, this trend demonstrates that urban planning has a determinant force on the CO2 emissions of area sources. Thus, the government should link the idea of low carbon development with its urban development strategy more closely. Urban planning can comprehensively influence the utilization and distribution of land, resources, buildings, transportation, and other elements in the process of urbanization. Mitigating and controlling CO2 emissions should also be included in urban planning. For instance, from the perspective of urban planning, the government can promote the construction of the public transportation system, encourage the mix of residential and commercial areas in urban planning, and encourage the use of the electric vehicle to reduce ffCO2 emissions from the transport sector. High-resolution CO2 emissions data can provide a useful tool to advance policymakers’ awareness of the combination of carbon mitigation and urban planning, help policymakers identify hot spots for CO2 emissions within the urban domain, and support policymakers to make a more targeted policy that is in line with the local social–economic context.
Finally, high-resolution CO2 emissions data can provide prior information for the inversion of atmospheric CO2 data observed by satellites, thus laying the foundation for the development of an integrated ground–airspace carbon monitoring network. This network will provide basic scientific data to facilitate policymaking and implementation to reduce CO2 emissions effectively, and hence improve measures aiming at climate change mitigation and adaptation. In order to complete this comparison, the following work still needs to be done to improve the accuracy and comprehensiveness of the data. Firstly, researchers need to allocate the data of CO2 emitted from waterway transportation to related grids and add the data of the carbon sinks for each grid cell. Secondly, researchers need to collect real-time data on residents’ natural gas consumption. We hope to use this study to maximize the accuracy of urban CO2 emission inventories and to promote the pertinence and the timeliness of the policy.