Spatiotemporal Analysis of Influencing Factors of Carbon Emission in Public Buildings in China

Abstract

The rapid development of public buildings has greatly increased the country’s energy consumption and carbon emissions. Excessive carbon emissions contribute to global warming. This paper aims to measure the carbon emissions in the operation of public buildings, and to identify the multiple influencing factors of carbon emissions in operational public buildings. First, the spatial and temporal variation characteristics of carbon emissions from public buildings in 30 provinces of China from 2008–2019 are analyzed. Second, a green building index is constructed, and the STIRPAT (Stochastic Impacts by Regression on Population, Affluence, and Technology) model is utilized to explore the relationship between each influencing factor and carbon emissions, using spatial and temporal geographically weighted regression analysis. The results show that the effects of population, urbanization rate, GDP per capita, green building index, and industrial structure on carbon emissions from public buildings all show spatial correlation and differences. There are east-west differences in the operational carbon emissions of public buildings in China’s provinces. Cluster evolution shows a spatially increasing trend from west to east. To some extent, policymakers can develop appropriate policies for different provinces through the findings.

1. Introduction

Climate change has become one of the most important globally recognized issues. Reducing CO₂ emissions helps mitigate climate change, and carbon emission and environmental protection issues have aroused the attention of various countries. Many countries have begun to measure carbon emissions and take action to reduce them [1,2]. Recently, the Chinese government announced that it strives to achieve carbon peaking by 2030, and carbon neutrality by 2060. More detailed plans have also been specified to reach this goal. By 2030, carbon dioxide emissions per unit of GDP will drop 65% compared to 2005, the proportion of non-fossil energy consumption will reach about 25%, and the total installed capacity of solar and wind power generation will reach more than 1.2 billion kilowatts. The Ministry of Ecology and Environment has proposed that during the 14th and 15th Five-Year Plan periods, China will carry out CO₂ emission peaking actions and specify the peaking targets and action plans for localities and industries. China’s total construction carbon emissions were 4.93 billion tons in 2018, accounting for 51% of the national carbon emissions. Carbon emissions from the production phase of building materials account for 28% of the total national carbon emissions, the construction phase accounts for 1%, and the building operation phase is 22% [3].

According to the data in 2018, the existing stock of public buildings in China’s urban and rural areas is 12.8 billion square meters, accounting for 21.3% of the total civil construction area. From the annual data, the energy consumption for the construction of public buildings accounted for 44% of the total building construction energy consumption in 2018, and the total energy consumption for the operation of public buildings excluding northern heating accounted for 33% of the total building operation energy consumption [4].

In 2015, the energy consumption of public buildings in China was 34.1 billion tons of standard coal equivalent, accounting for about 40% of the total energy consumption of civil buildings. However, the public building area accounted for only 18% of the total area of civil buildings. Furthermore, a study showed that the energy intensity of public buildings in China was four times that of residential buildings from 2000 to 2015. Public buildings are important places for human activities, and their construction, operation, renewal, and demolition processes all generate significant energy consumption. Therefore, the CO₂ emissions of public buildings in China have attracted more attention [5].

Green buildings play a very important role in reducing carbon emissions. Wu et al. (2017) [6] found that with commercial buildings in China, green buildings are lower in carbon emissions than non-green buildings, in the operational phase. The Chinese government proposed the Green Building Creation Initiative in 2020, which aims to reach 70% of the green building area in new urban buildings in that year by 2022. The effect of green building on carbon reduction in the building sector and its spatial evolution is the focus of this study.

2. Literature Review

2.1. Spatiotemporal Analysis of Carbon Emission

Spatiotemporal analysis is a method that considers both temporal data and spatial position, and is mainly used to solve how coherent entities change over time. Scholars have performed many researches with different methods to study the spatiotemporal analysis of carbon emission and its influencing factors in different fields. For example, Chen et al. (2021) analyzed the temporal and spatial characteristics of industrial carbon emissions in four regions of Guangdong province from 2005 to 2015, and concluded that industrial carbon emissions have a trend of eastward expansion [7]. The spatial dynamic analysis model (SDDM) was used to study the impact of different technological progress factors on carbon emissions [8]. Cui et al. (2021) explored spatiotemporal dynamic evolution of carbon emission intensity and per capita carbon emissions from planting industry in 31 provinces in China across 20 years. The spatial inequality is measured by Theil index and its contribution rate [9]. Wang et al. (2020) employed the standard deviation ellipse method and tapio decoupling method to reveal the spatiotemporal characteristics of the relationship between carbon emissions from transportation industry and economic growth [10]. Hu et al. (2020) studied the spatial and temporal evolution relationship between economic growth and carbon emissions in Belt and Road countries [11]. Han et al. (2021) revealed the spatiotemporal characteristics of carbon intensity of 20 industries by extending the spatial weight matrix and spatial dubin model [12]. Falahatkar et al. (2020) quantified the relationship between carbon dioxide emission and urban form in 15 Iranian cities, and believed that carbon dioxide emission level was positively correlated with urban area growth and urban complexity increase [13]. Some scholars have studied the carbon emission spatiotemporal effect in the construction industry. Bai et al. (2021) estimated the building inventory and carbon emissions embodied by buildings in 31 provinces of China from 1997 to 2016, and proposed a spatiotemporal decomposition model to identify driving forces [14]. To sum up, in the construction industry, there are few spatial analyses on carbon emissions during the operation of public buildings.

2.2. Influencing Factors of Buildings Carbon Emission

The extraction of raw materials, on-site construction activities, and building operations produced the majority of carbon emissions of the construction sector [4]. Hard coal and its derivatives were the largest carbon dioxide emitters in China’s construction industry [15]. In view of the significant impact of the construction industry, prior studies have been carried out to investigate influencing factors in order to develop mitigation strategies. For example, Lu et al. (2016) have analyzed the influencing factors of carbon emissions from construction activities in China, including energy intensity, energy structure, unit cost, level of construction automation, and machine efficiency [16]. Similarly, Zhang, Yan et al. (2019) stated that building scale, building structure type, and production efficiency of material are the three main driving factors [4]. Wu, shen et al. (2019) used the STIRPAT model and found that the impact of population size, per capita GDP, energy intensity, and industrial structure on carbon emissions were heterogeneous among regions [17]. Mostafavi, Tahsildoost et al. concluded that strengthening the design parameters of envelope structure, optimizing the layout, and utilizing natural ventilation are conducive to reducing energy consumption of high-rises [18]. Tan, Lai, Gu, Zeng, & Li constructed a carbon emission prediction model including population, urbanization rate, and urban building area [19]. Wang et al. explored the driving forces of energy-related CO₂ emissions in the construction industry by implementing the comprehensive decomposition method, and finally found that technological progress of industrial output was the leading factor that suppressed CO₂ emissions [20]. Huang et al. propounded increased energy efficiency design for new buildings and energy-saving retrofit for existing buildings to carbon emission [21].

2.3. Research Gap

Based on a critical review of relevant studies, as well as substantive surveys and interviews with Chinese building industry professionals, we conclude that the spatial and temporal effects of carbon emissions from public buildings still require further research. Because using the STIRPAT model to decompose influence factor is more comprehensive, it is still worthwhile to use this model to study the factors that drive the carbon emission of operational public buildings.

3. Data Source and Methodologies

3.1. Study Area

China has 34 provincial districts. Four provincial districts (Tibet, Macao, Hong Kong, and Taiwan) are excluded due to data unavailability, so this study selected a total of 30 provincial districts. The study divided 30 provincial districts into four regions (East region, Central region, West region, and Northeast region) according to the National Bureau of Statistics (

Table 1. Four regions and their provincial districts.

Regions	Provincial District
East region	Beijing, Tianjin, Hebei, Shanghai, Jiangsu, Zhejiang, Fujian, Shandong, Guangdong, Hainan
Central region	Shanxi, Anhui, Jiangxi, Henan, Hubei, Hunan
West region	Inner Mongolia, Guangxi, Chongqing, Sichuan, Guizhou, Yunnan, Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang
Northeast region	Liaoning, Jilin, Heilongjiang

).

3.2. Data Source

The data of green buildings from 2008 to 2016 and 2018 to 2019 are obtained from the Chinese Green Building Evaluation Label Network. Green building data in 2017 are obtained through compilation of public information of green building projects on the website of provincial Construction Department. The remaining indicators are from China Energy Statistical Yearbook and China Statistical Yearbook for 2009–2020. The map data comes from the National Geomatics Center of China.

3.3. Methodology

3.3.1. Hot Spot Analysis

$G_i^{*}$ statistic is used to analyze the spatial aggregation degree of carbon emissions during the operation of provincial public buildings, as shown in Equation (1).

$$G_i^{\ast }(d)={\displaystyle \sum _{i=1}^n{w}_{ij}}(d)x_i/{\displaystyle \sum _{i=1}^n{x}_i}$$

(1)

$x_i$ is the attribute value of the unit $i$, and $x_i$ is the carbon emission in $i$th province public buildings during its operation; $w_{ij}$ is the spatial weight matrix.

$G_i^{*}$ is standardized by Equation (2) and the result is $Z(G_i^{*})$. The larger the value $Z(G_i^{*})$ is, the higher the spatial clustering in the region, indicating that it belongs to the hot spot area; the smaller it is, the lower the spatial clustering in the region, indicating it belongs to the cold spot area.

$$Z(G_i{}^{\ast })=\frac{G_i^{*}-E(G_i^{\ast })}{\sqrt{Var(G_i^{*})}}$$

(2)

where $E(G_i^{*})$, $Var(G_i^{*})$ are the expectation and variance of $G_i^{*}$, respectively.

3.3.2. Geographically and Temporally Weighted Regression Model

The geographically and temporally weighted regression (GTWR) model is a deepening of the geographically weighted regression model, as shown in Equation (3). By using regional panel data for spatial regression, the temporal attributes are linked to the spatial attributes in the GTWR model, which better reflects the spatial and temporal change information of the study area, and makes the estimation results more effective.

$$y_i={\beta }_0(u_i,v_i,t_i)+{\displaystyle \sum _{k=1}^p{\beta }_k(u_i,v_i,t_i)}{x}_{ik}+{\epsilon }_i$$

(3)

where $y_i$ is the dependent variable of sample $i$, $x_{ik}$ is the $k$th independent variable at the sample point $i$, $u_i$, $v_i$ are the latitude and longitude coordinates of the center of gravity, respectively, $(u_i,v_i, t_i)$ are the spatial and temporal coordinates of the $i$th sample, ${\beta }_k(u_i,v_i, t_i)$ is the regression coefficient on the $k$th independent variable at the $i$th sample point, ${\beta }_0(u_i,v_i, t_i)$ is the space-time intercept of the $i$th sample point, and ${\epsilon }_i$ is the residual term.

3.3.3. Model Specification

Enrlich and Holdren first put forward the classic IPAT model in the early 1970s, which stipulates the influence of external factors on the environment. External factors include population size (P), affluency (A), and technology (T). The IPAT model was improved and transformed into the nonlinear random STIRPAT model, which is often used to analyze influencing factors of carbon emissions in different industries [22]. For example, Ma et al. surveyed the driving factors of carbon dioxide emission from public buildings in a country [23,24]; Yang and Jia explored the spatial effects of technology progress channels on CO₂ emissions for the agricultural, industrial, construction, transportation, and wholesale sectors [25]. The STIRPAT model is expressed as Equation (4):

$$I_i=a{P}_i^b{A}_i^c{T}_i^d{\epsilon }_i$$

(4)

where $i$ denotes the regional unit. $I_i$, $P_i$, $A_i$ and $T_i$represent the impacts on the environment owing to population, affluence (per capita GDP), and technology factors in region $i$, respectively. Constant a represents the scale of the model. Meanwhile, $b$, $c$, and $d$ are the estimated coefficients of population, affluence (per capita GDP), and technology, respectively. $\epsilon$ is the random error term. We take the logarithm of the STIRPAT model, obtaining the following Equation (5):

$$\ln I_i=\ln a+b\ln P_i+c\ln A_i+d\ln T_i+\ln \epsilon$$

(5)

The STIRPAT equation allows the addition of plenty of relevant variables, and the transformation of the model into an extended version, as long as the dimensionality of these variables is reasonable [26,27].

In order to deeply explore the mechanism of carbon emission of public buildings, considering green buildings’ specific characteristics, and looking for supporting references from a great deal of relevant previous studies, this study developed an extended version of the STIRPAT model using several meaningful variables retrieved from the population, affluence, and technology levels, respectively. The extended STIRPAT model is expressed in Equation (6):

$$\begin{array}{l}\ln C_{it}=\ln a+{\beta }_0(u_i,v_i,t_i)+{\beta }_1(u_i,v_i,t_i)\ast \ln P_{it}+{\beta }_2(u_i,v_i,t_i)\ast \ln U_{it}+{\beta }_3(u_i,v_i,t_i)\ast \ln I{S}_{it}\\ +{\beta }_4(u_i,v_i,t_i)\ast \ln G_{it}+{\beta }_5(u_i,v_i,t_i)\ast \ln IG{B}_{it}+\ln {\epsilon }_i\end{array}$$

(6)

where $C_{it}$ refers to the carbon emission in the public building sector in province $i$ over time $t$. $(u_i,v_i, t_i)$ represents spatial coordinates of province $i$ (i = 1, 2, 3, ..., 30). ${\beta }_k$ ($k$ = 1, 2, 3, 4, 5) denotes the $k$th regression coefficient in the $i$th province. The meaning and units of the variables are shown in

Table 2. Declaration of the model variables.

Nomenclature	Variable	Unit	Type	Supporting References
C	Carbon emission of public buildings	tCO2	dependent variable
P	Population	Ten thousand people	explained variable	[28]
U	Urbanization level	%	explained variable	[24,29,30]
IS	Industrial structure	1	explained variable	[24,30,31]
G	GDP per capita	Yuan	explained variable	[32]
IGB	Index of green buildings		explained variable

.

3.3.4. Index Calculation

• Calculation methods of CO₂ emission

The operational energy consumption of public buildings includes heating, air conditioning, ventilation, lighting, elevators, cooking, domestic hot water, office electrical equipment, and comprehensive service equipment and facilities. Corresponding energy types include electricity, gas (natural gas, gas, and LPG), fuel oil (diesel), and coal combustion. This study uses a macro model for measuring carbon emissions from buildings based on energy balance sheets.

This paper mainly studies the operational stage of carbon emissions in public buildings. Because China’s energy statistics yearbook does not provide building energy consumption directly [33], we need to select the energy consumption as public buildings’ operational consumption. The specific accounting boundaries are shown in

Table 3. Specific accounting boundary of public building.

Chinese Region Terminal Energy Category in Statistics Yearbook	Chinese Region Terminal Energy Category in Statistics Yearbook
Wholesale, retail trade and hotel, restaurants, and others	Coal, electricity, heat, liquefied petroleum gas, natural gas

.

This study mainly measures carbon emissions during the use of public buildings. Carbon emission in Chinese public buildings is measured by the end-use consumption of energy in each region in the China Energy Statistics Yearbook. The industries involved in public buildings are Transport, Storage and Post, Wholesale and Retail Trades, Hotels and Catering Services, and Other.

Energy type measurement includes coal, electricity, natural gas, LPG, and thermal power. Oil is not counted because it is mostly used in public buildings for transportation involving cars, and is not counted as energy consumption inside buildings for the time being. To obtain more meaningful and comprehensive results, we included three types of energy sources, such as coal, natural gas, and liquefied petroleum gas. According to the calculation method provided by IPCC (Equation (7)), coal, electricity, and heat consumed in the operation of public buildings are taken as the sources of carbon emissions.

$$C={\displaystyle \sum _j^{30}({\displaystyle \sum _i^3{C}_{ij}+C_{ej}+C_{hj}})}={\displaystyle \sum _j^{30}({\displaystyle \sum _i^3{E}_{ij}\times O_{ij}\times LC{V}_{ij}\times C{F}_{ij}\times \frac{44}{12}+E_{ej}\times \delta e}}+E_{hj}\times {\delta }_h)$$

(7)

where $C$ denotes the total carbon emissions from public building operation in each province, $C_{ij}$ refers to carbon emission of the consumption of fossil energy $i$ in the $j$ province, and $C_{ej}$ and $C_{hj}$ represent the carbon emissions from the secondary energy consumption of electricity and heat in the $j$ province. $E_{ij}$ denotes the consumption of fossil energy $i$ in the $j$ province; $O_{ij}$ refers to the oxidation rate of the fossil energy $i$ in the province $j$; $LC{V}_{ij}$ represents the average low-level calorific value of fossil energy $i$ in province $j$; $C{F}_{ij}$ denotes the carbon emission factor of fossil energy $i$ in province $j$; factor $\frac{44}{12}$ refers to the ratio of CO₂ molecules to carbon atoms by weight; carbon emissions can be converted into CO₂ emissions by multiplying by this coefficient, $E_{ej}$ and $E_{hj}$ denote the electricity consumption and heat consumption in province $j$, respectively, ${\delta }_{ej}$ and ${\delta }_{hj}$ denote the carbon emission factor of electricity and heat consumption in province $j$.

The carbon emissions generated during the use of public buildings are estimated by referring to the low level calorific value, carbon emission factor, and carbon oxidation rate provided by IPCC.

Since the carbon emission factor of coal is not directly provided in IPCC, coal is considered as raw coal for calculation. The carbon emission factor of coal is 25.8 TC/TJ, the low calorific value is 20.908 GJ/T, and the carbon oxidation rate is 0.899. LNG is converted to natural gas volume for calculation, depending on its density as 0.42~0.46 g/cm³.

The average CO₂ emission factors (kg-CO₂/kWh) of the national regional power grids in 2011 and 2012, as queried by the NDRC and the Guidelines for Provincial Greenhouse Gas Inventories, are shown in

Table 4. Electricity carbon emission factors.

Regional Grid	Coverage of Provinces and Cities	2011	2012	Average Value
North China Regional Grid	Beijing, Tianjin, Heibei, Shanxi, Shandong, Western Inner Mongolia	0.8967	0.8843	0.8905
East China Regional Grid	Liaoning, Jilin, Heilongjiang, Eastern Inner Mongolia	0.8189	0.7769	0.7979
Northeast Regional Grid	Shanghai, Jiangsu, Zhejiang, Anhui, Fujian	0.7129	0.7035	0.7082
Central China Regional Grid	Henan, Hubei, Hunan, Jiangxi, Sichuan, Chongqing	0.5955	0.5257	0.5606
Northwest Regional Grid	Shaanxi, Gansu, Qinghai, Ningxia, Xinjiang	0.6860	0.6671	0.67655
Southern Regional Grid	Guangdong, Guangxi, Yunnan, Guizhou, Hainan	0.5748	0.5271	0.55095

.

Electricity carbon emission coefficients obtained based on public data query are generally measured by the government or relevant departments in a unified manner, which is easily accessible, and their data source is authoritative. However, the data are not published annually, which is not conducive to the measurement of time series of building carbon emission data. The average value of these two years was used in this study. Inner Mongolia power emission factor is taken as the average value of 0.8442 in the east and west.

Table 5. Coefficient Thermal CO₂ emission (tCO₂/MWh).

Provinces and Cities	Thermal CO2 Emission Coefficient	Provinces and Cities	Thermal CO2 Emission Coefficient	Provinces and Cities	Thermal CO2 Emission Coefficient	Provinces and Cities	Thermal CO2 Emission Coefficient
Anhui	116	Guizhou	292	Hunan	110	Sichuan	105
Beijing	88	Hainan	57	Jilin	132	Tianjin	108
Fujian	112	Hebei	122	Jiangsu	109	Xinjiang	109
Gansu	110	Henan	124	Jiangxi	134	Yunnan	149
Guangdong	93	Heilongjiang	155	Liaoning	130	Zhejiang	104
Guangxi	153	Hubei	122	Inner Mongolia	160	Chongqing	98
Ningxia	120	Qinghai	245	Shandong	114	Shanxi	116
Shaanxi	149

shows Coefficient Thermal CO₂ emission [34].

2.

Index of green building

Green buildings in China are classified as one-star, two-star, and three-star. Three stars are the highest level of green building. The index of green building is calculated through Equation (8).

$$IGB=D_1\times 1+D_2\times 2+D_3\times 3$$

(8)

where $D_1$, $D_2$, $D_3$ denote the number of one-star, two-star and three-star public green buildings in China, respectively.

3.

Industrial Structure

We use the percentage of added value of the tertiary industry to GDP to describe industrial structure [35].

4. Empirical Results

4.1. Spatial Distribution of Carbon Emission in Different Areas

The regional energy balance of the China Energy Statistics Yearbook for 2009–2020 was used to estimate the carbon emissions from the operation of public buildings in each province of the country using end-use energy consumption. The specific measurement results, in accordance with the previously stated zoning, are shown in Figure 1.

In general, the eastern region has more carbon emissions than the central, western, and northeastern regions. The top three provinces generating carbon emissions from public buildings from 2008–2018 were Guangdong, Jiangsu, and Beijing. In 2019, Shandong surpassed Beijing among the top three.

Among the eastern regions, Guangdong Province has the most carbon emissions from public buildings and Hainan Province has the least. All provinces show an increasing trend year by year. Jiangsu Province and Hebei Province have a faster growth rate. In the central region, Henan Province is has the highest carbon emissions, except for 2017, and Jiangxi Province has the lowest carbon emissions. Other regions are steadily increasing, however, not as much as the vast majority of the eastern region’s emissions.

Among the western regions, Sichuan Province consistently has the highest carbon emissions from public buildings, and Qinghai Province steadily has the lowest. Chongqing fluctuated more, with carbon emissions decreasing in 2015 and 2017. The growth rate is larger in Inner Mongolia and Xinjiang. Carbon emissions in Inner Mongolia province were always lower than Shaanxi province between 2008 and 2017, and exceeded Shaanxi province after 2018. Xinjiang Province surpassed Chongqing, Yunnan Province, and Guangxi Province in 2018.

In Northeast China, Liaoning Province has the highest carbon emissions from public buildings. Carbon emissions from public buildings in Heilongjiang Province were higher than those in Jilin Province, except in 2012 and 2013.

4.2. Hot Spot Analysis of Carbon Emissions from Public Building Operations

Hot spot analysis can reflect the spatial aggregation effect of carbon emissions from public buildings. The variation of the aggregation can be seen by plotting the hot spot and cold spot areas in different years. The natural interruption point grading method was used to classify the values of each year, calculated by Equations (1) and (2) into high, subhigh, sublow, and low value cluster areas in order of largest to smallest. The study area of this research is in the years 2008–2019, and the clustering results of 2010, 2013, 2016, and 2019 are drawn equally spaced for analysis, as shown in Figure 2.

From Figure 2, it can be seen that evolution of provincial carbon emission clustering in China shows a spatially increasing trend from west to east. Overall, the high-value clustering areas are mainly concentrated in Zhejiang, Jiangsu, Anhui, and Shanghai. The low-value clustering areas are mainly concentrated in Qinghai and Sichuan. In 2016, the number of provinces with high-value clustering areas increased, then decreased in 2019. Over time, the high value cluster areas first expanded and then contracted.

4.3. Spatial Effects of the Influencing Factors of Public Buildings Carbon Emissions

Carbon emissions from public building operations are the dependent variable; population, urbanization rate, industrial structure, GDP per capita, and green index are the explanatory variables in the STIRPAT model; the time range is 2008–2019; and the X and Y coordinates are the geographic coordinates of each province. With these factors, the runs can be entered into the spatio-temporal geographically weighted model to obtain the influence size of the five explanatory variables. In order to unify the comparison of influence size, the influence size values are arranged in descending order and divided by equal spacing, and the positive and negative influence are divided into six levels. Positive influence includes weak (WPI), medium (MPI), and strong positive influence (SPI), which indicates that the influence factor positively contributes to the carbon emission of public building operation; the negative influence includes weak (WNI), medium (MNI), and strong negative influence (SNI), which indicates that the influence factor negatively inhibits the carbon emission of public building operation [36]. Due to the number of years, this study selected 2010, 2013, 2016, and 2019 for spatial presentation and analysis. The GTWR model was run with an adjusted R² of 0.956 and an AICc of 199.253, indicating a better model effect.

Based on the regression results of the GTWR model, the spatial and temporal variability of the five influencing factors of carbon emissions of public buildings is analyzed one by one.

• Spatial and temporal variation in the effect of population on carbon emissions

Population is a positive influence on the carbon emissions of public buildings in each province. The maximum value of the population regression coefficient is 8.135 and the minimum value is 0.082, which can be divided into WPI, MPI, and SPI according to the influence level. From the results, the provinces with the highest impact are Shanghai, Zhejiang, Jiangsu, Shandong, Anhui, Tianjin, Beijing, and Fujian. Mainly with the growth of population, the service industry activities in public buildings are frequent, thus increasing the carbon emissions from public buildings. As shown in Figure 3, the impact of population on carbon emissions from public buildings is increasing year by year. Spatially, the SPI of population size is gradually spreading from the northeast and eastern coastal regions to the central and western regions. By 2016, population size has reached a SPI within 30 provinces in the study area.

2.

Spatial and temporal variation in the effect of urbanization on carbon emissions

During the study period, the maximum value is 30.386 and the minimum value is −8.649. According to the influence level, it can be divided into SNI, MNI, WNI, WPI, and SPI. Overall, the urbanization rate has a predominantly negative impact on carbon emissions from public buildings. In 2010, WNI dominates, occupying 18 provinces, followed by WPI, occupying 11 provinces. In 2013, 2016, and 2019, WNI dominates, occupying 15 provinces, followed by WPI, occupying 14 provinces. It indicates that the urbanization rate has a small impact on the carbon emissions of public buildings. The scale of carbon emissions from public buildings does not increase with the increase in urbanization level; instead, it decreases with the optimization of energy consumption structure and the improvement of energy utilization efficiency. During the study period, the number of provinces with negative impact levels shows a decreasing trend, and spatially, the area of positive effect gradually expands from the northeast and some southeastern provinces to the whole eastern region, and then gradually shifts to the central region and Xinjiang province (Figure 4). Public buildings are mostly located in urban areas and less in rural areas. The increase in urbanization rate will promote the development of tertiary industry, which will further promote the generation and operation of public buildings. When the urbanization rate reaches a certain level, it will curb the carbon emissions of running public buildings because the stability of urbanization will make people start to raise the awareness of energy saving, not only limited to the use of more focus on energy efficiency.

3.

Spatial and temporal variation in the effect of industrial structure on carbon emissions

The regression coefficient of the industrial structure has a maximum value of 0.315 and a minimum value of −0.426 in the study period. Industrial structure refers to the ratio of the value added of the tertiary sector to the total value added of production. It is classified as SPI, WPI, WNI, and MNI according to the influence level. As shown in Figure 5, it is generally a positive impact. 2010, 2013, 2016, and 2019 are dominated by weak positive impact, occupying 16, 21, 18, and 15 provinces, respectively. The positive effect of industrial structure is shifted from the east to the center. In 2010, the positive effect of industrial structure is in the eastern coastal region, northeastern region, and Xinjiang province, and that positive effect is partially shifted to the central region in 2019.

4.

Spatial and temporal variation in the effect of GDP per capita on carbon emissions

The maximum value of the regression coefficient for GDP per capita is 1.177 and the minimum value is −1.184. In 2010, the positive effect is dominant, with a total of 16 provinces in WPI and MPI. In 2013, the negative effect is dominant, with a total of 22 provinces in SNI, MNI, and WNI. In 2016, the positive effect is more pronounced, with 24 provinces in WPI. In 2019, the positive effect is pronounced, with a total of 22 provinces in WPI and MPI. As shown in Figure 6, the influence of GDP per capita shows a trend from positive to negative and then positive again. By 2019, the only regions with a negative effect are Shanghai, Jiangsu, Zhejiang, Anhui, Fujian, Jiangxi, Hubei, and Hunan. It indicates that population affluence is promoting the increase in carbon emissions from public buildings.

5.

Spatial and temporal variation in the effect of the green building index on carbon emissions

The maximum value of the regression coefficient of the green building index is 0.832 and the minimum value is −0.098. As shown in Figure 7, in 2010, all 30 provinces have a positive effect. In 2013, the WNI dominated, occupying 16 provinces. In 2016, WPI occupied 17 provinces. In 2019, WPI occupies 21 provinces. In 2010, the green building index has a positive effect on carbon emissions from public buildings in all regions. However, by 2019, the inhibitory effect has 9 provinces. According to previous studies, green buildings contribute to carbon emission reduction than non-green buildings, but this study finds that the large-scale effect of carbon emission reduction from green buildings has not yet been developed.

5. Conclusions and Policy Suggestions

5.1. Conclusions

Using the carbon emission of public buildings in operation in China from 2008 to 2019, the GTWR model was used to detect the spatial distribution of the influence coefficients of population, urbanization, industrial structure, GDP per capita, and index of green buildings. We found significant spatial heterogeneity changes between the five factors. Most importantly, the factor of the index of green buildings.

The empirical results of the green building index and carbon emissions of public building operation help to determine the direction and intensity of green building development, and help to enrich the study of the impact of green buildings on the carbon emissions of public buildings from a regional perspective.

In terms of carbon emissions from public building operations, the top three public building carbon emissions from 2008–2018 were Guangdong, Jiangsu, and Beijing, and in 2019, Shandong surpassed Beijing among the top three. Overall, the total carbon emissions from public buildings in the eastern region, except Hainan, Fujian, and Tianjin, are greater than those in the central, western, and northeastern regions, and the growth rate is obvious.

The results of the hotspot analysis show that there are east-west differences in the operational carbon emissions of public buildings in Chinese provinces. The evolution of clustering shows a spatially increasing trend from west to east.

The STIRPAT model shows that population has a positive influence on public building carbon emissions in each province, and the positive influence of population gradually spreads from the northeast and the eastern coastal regions to the central and western regions; urbanization rate has a predominantly negative influence on public building carbon emissions; industrial structure has a positive influence; the influence of GDP per capita and green building index shows a trend of positive to negative and then positive; the large-scale effect of green building carbon emission reduction has not yet been formed.

5.2. Suggestions

According to the above conclusions, we can make the following recommendations.

First, total carbon emission control should be carried out at the regional level under the constraints of other indicators such as socio-economic development rate and industry economic growth. Focus on controlling carbon emission hotspot areas and emission reduction measures should be formulated for cold spot areas, according to development needs to avoid generating large amounts of carbon emissions due to rapid economic development. Cooperation between provinces can be strengthened to develop inter-provincial carbon emission trading policies to balance provincial carbon emissions.

Second, for provinces with large populations, public buildings should be retrofitted with energy efficiency. It would be advisable while developing the economy to adjust the energy consumption structure, change the economic development mode, adhere to the path of low carbon development, and slow down the rapid growth of carbon emissions caused by the rapid growth of the regional economic development level.

Third, the development of green buildings still needs to continue to improve, and the current growth of green buildings has not yet formed a scale effect on the carbon emission aspect of public buildings. Because of the high cost of green building construction, there is a need to support the construction and development of green buildings in economically disadvantaged areas.

However, this study is subject to several limitations. Firstly, this paper uses China as an example, and the analysis for green buildings can be extended to other emerging economies and developing countries. Secondly, temperature has a different impact on the heating and cooling of public buildings under different climate backgrounds. Further study can take a closer look at the micro-level of the impact of temperature on the energy consumption of public buildings. Finally, the empirical results of this paper are helpful for policy makers to develop differentiated emission reduction strategies for high and low carbon emission provincial administrative regions.