Detecting Differences in the Impact of Construction Land Types on Carbon Emissions: A Case Study of Southwest China

Abstract

The area with the highest concentration of carbon emission activities is construction land. However, few studies have been conducted that investigated the different effects of various types of construction land on carbon emissions and the extent of their impact. To address this shortcoming, this study constructed a multi-indicator evaluation system with 393 counties in Southwest China and integrated ordinary least squares and spatial regression models to deeply analyze the different impacts of construction land types on carbon emissions. The results revealed that (1) in Southwest China, carbon emissions were generally distributed in clusters, with significant spatial variability and dependence; (2) the distribution of urban land scale, rural settlement land scale, and other construction land scale all showed obvious spatial clustering differences; (3) all three types of construction land’s effect on carbon emissions was positive, and the direction of impact was in line with theoretical expectations; and (4) the other construction land scale had the highest effect on carbon emissions, followed by rural settlement land scale, while the urban land scale was slightly lower. The findings help to further explain the different impacts of construction land types on carbon emissions and provide theoretical references for the government to formulate more refined emissions reduction policies.

1. Introduction

Climate change is a worldwide political, economic, and social issue, as well as a scientific one [1]. The continuous exploitation of energy and natural resources by humanity has led to a significant increase in emissions of pollutants, such as nitrogen oxides, ozone, sulfur dioxide, carbon monoxide, benzene, particulate matter, and carbon dioxide, which aggravate air pollution in urban environments while causing the global greenhouse effect and deteriorating air quality [2,3]. The Sixth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC) emphasizes that global climate change mitigation and adaptation actions are urgent and that controlling the global warming caused by human activities requires the limiting of cumulative CO₂ emissions to a specific level, that is, at least net zero. As the globe’s top energy user and carbon dioxide emitter, China’s total carbon emissions in 2020 were 10,243.4 Mt, accounting for 31.73% of the global total. To cope with global warming and reduce the CO₂ concentration in the atmosphere, China proposed to decrease CO₂ emissions as a percentage of gross domestic product (GDP) by more than 65% in 2030 compared with 2005 and boost non-fossil fuel utilization to around 25%, with the clear aim of achieving zero carbon emissions by 2060 [4,5]. From the perspective of reducing emissions, investigating the extent of the role played by factors influencing carbon emissions is key to dissecting the formation mechanism of the latter.

Carbon emissions reduction is both a climate issue and a development issue, involving politics, the economy, energy, the environment, and other aspects. Scholars have looked at the elements that influence CO₂ emissions, the connection between CO₂ emissions and economic growth, and the explanations for the CO₂ emissions disparities between different regions and industries from different perspectives. Most researchers consider such influencing factors to come from the following aspects: (1) Energy structure and efficiency. Different proportions of energy consumption produce different carbon emissions, lowering energy intensity is the major strategy to minimize carbon emissions, and adjusting energy structure helps to cut CO₂ emissions dramatically [6]. The diversification of the energy consumption structure facilitates the country’s transition from a predominantly high-carbon fuel to a predominantly low-carbon one [7], and carbon emissions are suppressed when energy efficiency is improved [8]. (2) Economic factors. Human economic activities are the primary contributors to the growth of CO₂ emissions on the planet, and the connection between economic growth and CO₂ emissions is bidirectional, regardless of the curve pattern they show [9]. At the same time, the leading industries and industrial structures that drive economic development have different impacts on carbon emissions [10]. (3) Population size. The size and expansion of the population are universally acknowledged as major drivers of anthropogenic CO₂ emissions, other greenhouse gas emissions, and environmental deterioration outcomes [11]. Jorgenson et al. explored the connection between CO₂ emissions and population in 85 countries and concluded that population had a substantially higher positive impact on total CO₂ emissions at the national level and was much larger than that of other human drivers [12]. (4) Technological progress. The connection between technological advancement and CO₂ emissions is complicated [13]. Technological advances save energy by improving energy efficiency or energy intensity, which, in turn, can effectively reduce carbon emissions [14]. However, every great technological advancement in modern societies has not only improved energy and environmental efficiency but has also boosted economic development and corresponding energy consumption [15]. Except for the aforementioned factors, other scholars have claimed that factors such as urbanization [16], transportation [17], international trade [18], and household energy consumption [19] also affect carbon emissions. In general, the existing literature has mainly explored the relationship between carbon emissions and the economy, population, technology, trade, and so on, and has concentrated less on the impact of construction land, which carries the greatest intensity of carbon emissions from human activities.

As an important factor of production, construction land fundamentally supports the process of economic operations; however, its high input and high consumption-oriented crude utilization is a major driver of carbon emissions. Some scholars consider construction land as the main carbon source [20] and have focused their research on the relationship between total construction land and CO₂ emissions, arguing that CO₂ emissions and emissions intensity show an increasing trend with an increase in construction land area [21]. At the same time, agricultural land and forests with high carbon sink capacity are encroached upon by the scale of construction land expansion, which further aggravates atmospheric CO₂ concentration [22]. Some studies have also investigated the impact of specific construction land types on CO₂ emissions, including residential land [23], industrial land [24], commercial land [25], land for roads and transportation facilities [26], land for public facilities, and land for logistics and storage [27]. However, research is still lacking on the different impacts of various types of construction land on CO₂ emissions, as well as the effects’ extents. However, this aspect is important. Because the functions of various types of construction land are diverse, the carbon emissions increases they cause are bound to be different. A more accurate assessment of the degree of impact of various types of construction land has theoretical significance and practical value for the formulation of precise carbon emissions reduction policies focusing on specific land types. Therefore, it is necessary to explore in depth the variations in CO₂ emissions generated by different types of construction land in built-up areas to discern the construction land type that has the greatest influence on CO₂ emissions, which will, in turn, benefit the government’s formulation of more refined emissions reduction policies.

The fragile ecological environment, low land-use efficiency, unreasonable industrial structure, and rough development model of “high input, high pollution, and low output” in Southwest China have resulted in a dramatic upsurge in regional CO₂ emissions. Driven by rapid urbanization, the land use pattern is changing drastically, the ecological environment is suffering serious damage, and the pressure to achieve future energy savings and emissions reduction is high. Therefore, this topic is suitable as a research case.

County units are the basic units of China’s economic and social development [28], and conducting research on carbon emissions from county units can help to finely assess regional carbon emissions reduction pressure and promote regional economic low-carbon development. Therefore, this study adopted the ecologically fragile southwest region as a research topic and used traditional regression and spatial regression analysis methods to deeply analyze the different impacts of construction land types on carbon emissions. This paper proposes corresponding countermeasures as a theoretical reference for the government to formulate more refined emissions reduction policies.

2. Materials and Methods

2.1. Study Area

The southwest region, located within 21°08′ N to 33°41′ N and 97°21′ E to 110°11′ E, is an important barrier area in the upstream area of the Yangtze River. It is divided into four provinces and regions—Sichuan, Yunnan, Guizhou, Tibet, and Chongqing—and has a complex topography, sensitive ecological environment, and weak economic foundation. Driven by new urbanization in China, the construction land area is expanding, the intensity of human activities is increasing, and carbon emissions are also increasing. In 2015, the construction land area was 11,261.01 km², including 2643.71 km² of urban land area, 4368.92 km² of rural settlement land area, and 3248.39 km² of other types of construction land area, such as enterprises and mines, big industrial districts, and traffic roads. Altogether, carbon emissions amounted to 862.74 million tons. Considering the small amount of construction land area and CO₂ emissions in Tibet, it is not easy to evaluate the area of specific construction land categories. Therefore, referring to the existing literature [29,30,31], the southwest region in this study included only the 393 counties within Sichuan, Yunnan, Guizhou, and Chongqing (Figure 1).

2.2. Research Design

2.2.1. Research Ideas

The goal of this research was to quantitatively identify the varied impacts of construction land types on carbon emissions, as well as the direction and extent of such impacts. First, 393 counties in Southwest China were studied to examine the distribution features of carbon emissions and three types of construction land and to test whether there was a spatial dependence of carbon emissions among the counties. Second, given that various construction land types carry different intensities of human activity, a carbon emissions impact model was constructed based on three aspects: the urban land scale, the rural residential land scale, and the other construction land scale (including industrial and mining, transportation, and roads). In the model, carbon emissions were the dependent variable; the three indicators of urban land scale, rural residential land scale, and other construction land scale were the explanatory variables; and the three indicators of GDP, secondary industry output proportion in GDP, and resident population were the control variables. Of note, in the selection of control variables, we considered the role of fixed assets investment in human production and life; however, there was obvious covariance in this indicator in the indicator covariance test, and after repeated verification, the control variable of fixed assets was excluded to meet the covariance test criteria. Third, the Lagrange multiplier (LM) test was applied to the ordinary least squares (OLS) model to identify whether the model selection was reasonable. Fourth, the best model among the OLS and spatial regression models was selected to further analyze the relationship between various construction land types and carbon emissions, including in terms of significance, the direction of influence, and intensity of influence. Finally, the research results were analyzed (Figure 2).

2.2.2. Selection of Indicators

The factors that affect CO₂ emissions are complex and varied, and the mechanisms of action among them are complicated. No consensus on them has been reached in the academic community. To ensure the representativeness and accessibility of the indicators, we selected economic development, industrial structure, and population size as control variables. According to the secondary land-use classification standard of the Chinese Academy of Sciences, construction land was divided into three categories: urban land, rural settlement land, and other construction land (

Table 1. Definitions of the variables.

Variable	Influencing Factors	Evaluation Indicators	Indicator Meaning	Expected Direction
Dependent variable	Carbon emissions	CO2 emissions	Indicates total CO2 emissions by county (Mt)
Explanatory variables	Construction land type	Urban land scale	Indicates land in large, medium, and small cities and built-up areas above county towns (km2)	+
		Rural settlement land scale	Indicates land for rural settlements independent of towns (km2)	+
		Other construction land scale	Indicates land for factories and mines, large industrial areas, oil fields, salt fields, quarries, etc., as well as transportation roads, airports, and special land (km2)	+
Control variables	Economic development level	GDP	Indicates the scale of regional economic development (billion CNY)	+
	Industry structure	Secondary industry output proportion in GDP	Reflects the rationality of industrial structure	+
	Population size	Resident population	Indicates regional population size (million people)	+

).

• Dependent variable

Carbon emissions are processes that release CO₂ into the atmosphere, including carbon emissions from non-anthropogenic sources, such as oceans, soils, rocks, and organisms, as well as carbon emissions from anthropogenic sources, such as energy consumption, industrial production, domestic waste disposal, and wastewater treatment. Currently, the increase in atmospheric CO₂ concentration is mainly caused by human activity [32]. Considering the different resource endowments and economic development bases between regions, carbon emissions will vary, and the impact on the environment will be different. Therefore, we adopted the most basic administrative unit in China [33], namely, the county, as the research object, and chose county carbon emissions as the control variable.

2.

Explanatory variables

The core of regional development is construction land, and construction land with different land types carries different human activities and building functions, while different human behaviors and building functions have different corresponding energy consumption demands due to different content, intensity, and density of use [34], as well as different impacts on carbon emissions.

Urban land refers to land in the built-up areas of large, medium, and small cities and county towns, as determined by the overall land-use planning. Urban land is the basic spatial carrier for human production and living, leisure and recreation, and economic and social development, and it is also the carrier of carbon emissions [35]. Yuan et al. used a remote coupling framework to analyze the connection between different land-use types and CO₂ emissions intensity in Jiangsu, China, and verified that urban land use and economic development affect carbon emissions transfer to some extent [36]. Chuai et al. used a linear programming model to optimize the land-use structure and found that limiting the urban land scale would play a key role in carbon emissions reduction [37]. For this reason, we selected the urban land scale as an explanatory variable to verify its impact on carbon emissions.

Rural settlement land is the main place where rural populations congregate, and their scale, morphology, and spatial distribution reflect the intensity of production and livelihoods of rural populations. Moon argued that rural areas play a vital role in sustainable development, and the multifunctionality of rural settlement land is seen as a major source of carbon emissions [38]. Minx et al. used mixed methods to estimate the carbon footprint of urban and other human settlements in the UK and concluded that the carbon footprint of rural settlement land was higher than that of urban areas [39]. Qiu et al. used the STIRPAT model and various panel regression methods to explore the effects of urbanization on energy consumption and CO₂ emissions from rural settlement land. The results showed a positive impact of the total energy intensity of rural settlement land on carbon emissions [40]. Considering the role of rural settlement land in carbon emissions, we selected it as an explanatory variable in the study.

“Other construction land” is a collective term that refers to land for factories, mines, large industrial zones, oil fields, salt fields, quarries, and so on, as well as transportation roads, airports, and special land. These land types are all important components of construction land, but they are small in scale and relatively scattered, and have similar spectral and structural characteristics, making it difficult to further divide and count them; therefore, they were uniformly classified as other construction land with reference to the Chinese Academy of Sciences secondary land use classification standard. Many studies evaluated the impact of factories, mines, industrial areas, or traffic roads on carbon emissions and concluded that most have a close relationship with carbon emissions [41,42]. Therefore, we also studied other construction land as an explanatory variable.

3.

Control variables

Economic development and energy consumption, as important transmission channels, are considered the culprits in the environmental degradation process [43]. Chuzhi et al. considered economic scale as the main driver with an incremental effect on carbon emissions [44]. Song et al. constructed an incremental carbon emissions factor decomposition model to investigate the impact of economic size on CO₂ emissions in the Yangtze River Delta region and discovered that economic size has the greatest impact on the incremental increase in CO₂ emissions [45]. Saboori et al. argued that there is no causal connection between CO₂ emissions and the economy in the short term. In contrast, in the long run, there is a unidirectional causal relationship between the economy and carbon emissions [46]. Regardless of whether scholars agree or disagree, the role of the relationship between economic scale and carbon emissions cannot be ignored. GDP is the value of all final goods and services produced in a country or region during a certain period and includes four different components: consumption, private investment, government spending, and net exports. It is often considered the best indicator of a country’s or region’s economic situation [47], and therefore, we chose GDP as a control variable in the study.

The secondary industry output proportion in GDP usually refers to the contribution of extractive industries; manufacturing industries; electricity, gas, and water production and supply industries; and construction industries in national economic development, which is mainly used to reflect the rationality of regional industrial structure. Zhang et al. considered secondary industries among the three major industries that have relatively high potential as the main source of carbon emissions [48]. Li et al. argued that an increase in the proportion of secondary industries would lead to an increase in CO₂ emissions intensity in the Yangtze River Delta prefecture-level cities [49]. Therefore, we chose secondary industry output proportion in GDP as a control variable to analyze the influence of the industrial structure of different proportions on carbon emissions.

Population size refers to the number of people in a country or region during a certain period [50]. Changes in population size lead to changes in productive life needs, such as food, water, housing, transportation, education, and work, and the energy consumed to meet these needs varies. The direct and indirect effects of population on energy consumption are widely recognized as a major driver of anthropogenic CO₂ emissions, other greenhouse gas emissions, and environmental degradation outcomes at the national level [12]. Anser et al. found that population size in less developed countries affects CO₂ emissions in a more significant way than in developed countries [51]. Yang et al. concluded that population size can suppress an increase in the cost value of industrial carbon emissions [52]. To reflect the authenticity of the county population data as much as possible, we used the resident population to indicate the population size as a control variable.

2.3. Data Sources

The construction land data for Southwest China were obtained from the 2015 land-use data of the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC). Based on the National Land Use/Cover Classification System for Remote Sensing Monitoring, construction land use is divided into three categories: urban land, rural settlement land, and other construction land. The 2015 carbon emissions data were obtained from the Carbon Emission Accounts and Datasets (CEADs). Economic and social data, such as GDP, secondary industry output proportion in GDP, and resident population, were mainly drawn from the 2016 China County Statistical Yearbook, with a few being obtained from the China City Statistical Yearbook, Municipal Statistical Yearbook, and Statistical Bulletin, among others.

2.4. Research Methodology

2.4.1. Spatial Autocorrelation

Due to the spatial dependence among counties, their carbon emissions tend to have spatial interactions. Therefore, the global spatial autocorrelation (Moran’s I) index was used to quantitatively evaluate the degree of spatial correlation of carbon emissions in counties and their spatial patterns. ArcGIS software (v.10.2; 2013), spatial statistical tools, and the GIS natural breakpoint method were used to hierarchically and visually classify the carbon emissions and construction land types of 393 counties in Southwest China. Moran’s I index is expressed as follows:

$$Moran’sI={\displaystyle \sum _{c=1}^n{\displaystyle \sum _{j=1}^n\left(x_c-\overline{x}\right)\left(x_j-\overline{x}\right)}}/S^2{\displaystyle \sum _{c=1}^n{\displaystyle \sum _{j=1}^n{W}_{cj}}}$$

(1)

$$S^2={\displaystyle \sum _{c=1}^n{(x_c-\overline{x})}^2}/n$$

(2)

where Moran’s I is the spatial autocorrelation index of county carbon emissions; x_c and x_j denote the carbon emissions of county c and county j, respectively; $\overline{x}$ denotes the average carbon emissions; W_cj denotes the spatial weight matrix of 393 counties; S² is the variance. When Moran’s I index is greater than 0, it means that there is a beneficial relationship between county carbon emissions and the surrounding counties; when the value tends to 1, it means that the correlation of county carbon emissions is more obvious, that is, the county units with higher (or lower) carbon emissions show a spatial clustering distribution. When the Moran’s I index is less than 0, it means that a negative relationship between county carbon emissions and the surrounding counties, and when it gets closer to −1, it implies that the spatial differences in county carbon emissions are greater; when the index is equal to 0, it implies that there is no spatial correlation, that is, the spatial pattern of county carbon emissions is random.

We used the significance of the Z-test value to determine the degree of spatial agglomeration of carbon emissions in the county. Its formula is as follows:

$$Z_c=c-E(c)/\sqrt{Var(c)}$$

(3)

where Var(c) denotes the variance of county CO₂ emissions, E(c) is the mathematical expectation of county CO₂ emissions, and if the Z value is significant, it indicates that carbon emissions in Southwest China show a pattern of spatial clustering.

2.4.2. Ordinary Least Squares

The OLS method was used to explore the factors influencing CO₂ emissions in the Southwest region, with the aim of analyzing and verifying whether the selected indicators are reasonable [53] and whether they affect county-level CO₂ emissions in a significant way. The OLS model is expressed as follows:

$$y_c={\beta }_0+{\displaystyle {\sum }_{k=1}^n{\beta }_k}{x}_{ck}+e_c,\left[e_c~D\left(0,{\varphi }^{2}I\right)\right]$$

(4)

where k is the carbon emissions impact factor (k = 1, 2, …, 6); β₀ denotes the constant term of OLS; y_c is the carbon emission of the cth county; x_ck denotes the standardized value of the kth impact factor in the cth county; β_k is the regression coefficient corresponding to these six impact factors; and e is the error term of the OLS model, where e_c~D (0,φ²I) means that the error term obeys normal distribution and the variance is consistent, that is, the product of the error and covariance matrix is 0 [53]. To avoid differences in results due to different magnitudes, we standardized the carbon emissions (dependent variable) and the six independent variables logarithmically. To avoid multicollinearity in the independent variables constructed in the model, which affects the final results of the study, before constructing the OLS model, we performed a multicollinearity test with SPSS software (v.22.0, 2013) and calculated the variance inflation factor (VIF) for the six variables, which is calculated as follows:

$$VIF=\raisebox{1ex}{$1$}\!\left/ \!\raisebox{-1ex}{$1-R_k^2$}\right.$$

(5)

where $R_k^2$ denotes the coefficient of determination of the kth carbon emission influencing factor with other factors in the OLS regression model. The larger the VIF, the greater the possibility of multicollinearity between the models. Therefore, the selected VIF limit was 10; that is, when VIF is less than 10, it is considered to indicate that there is no multicollinearity among the six influencing factors of the constructed model, and when VIF is greater than or equal to 10, it is considered that there is a multicollinearity problem among the factors [54].

2.4.3. Spatial Regression Model

There may be spatial dependence between spatially adjacent county units, and OLS cannot reflect the spatial information of independent variables, such as construction land type. Therefore, it was also necessary to construct spatial regression models to analyze the spatial connection between construction land types and CO₂ emissions.

The spatial lag model (SLM) can reflect the impact of carbon emissions of one county unit on carbon emissions of other neighboring counties, that is, the spatial spillover effect [55]. SLM is expressed as follows:

$$y_c=\rho {\displaystyle \sum _{j=1}^n{W}_{cj}{y}_j}+\beta X_c+e_c,\left[e_c~D\left(0,{\varphi }^{2}I\right)\right]$$

(6)

where $\rho$ is the spatial autoregressive coefficient of SLM and W_cj denotes the spatial weight matrix.

In spatial regression analysis, the independent error terms may be spatially autocorrelated [56]. The spatial spillover effect of the independent error terms is taken into account using the spatial error model (SEM), which is expressed as follows:

$$y_c=\mathsf{\lambda }{\displaystyle \sum _{j=1}^n{W}_{cj}{\sigma }_j}+\beta X_c+e_c,\left[e_c~D\left(0,{\varphi }^{2}I\right)\right]$$

(7)

where $\sigma$ represents the spatial autocorrelation error term and λ is the spatial autocorrelation coefficient of the error term.

3. Results

3.1. Characteristics of Spatial Differences in Carbon Emissions

The descriptive statistics of county-level carbon emissions in Southwest China are shown in Figure 3. The carbon emissions of 393 counties in Southwest China were classified into five levels according to intensity: 0.0379–1.4953, 1.4954–3.2071, 3.2072–5.2024, 5.2025–8.6875, and 8.6876–17.3026 Mt. The higher the grade, the greater the carbon emissions. From the change in quantity, the carbon emissions of counties in Southwest China showed an obvious geographical unevenness, with a relatively small number of high-carbon-emission areas (only six counties) and a relatively large number of low carbon emission areas (200 counties).

In terms of the spatial distribution, CO₂ emissions in general showed a cluster-like aggregation distribution, and the regions with high total carbon emissions were mainly clustered in the suburban areas of megacities and typical industrial cities. Yubei District in Chongqing Municipality, Shuangliu District in Sichuan Province, Huaxi District, Wudang District, Xinyi City, and Panxian County in Guizhou Province were the centers of high CO₂ emissions, while the carbon emissions of other counties decreased with greater distance from high carbon emission areas, and the spatial distribution of CO₂ emissions was distinctly different (Figure 3).

Interactions often existed between different counties within the Southwest region; therefore, we adopted Moran’s I index in this study to further determine the spatial autocorrelation characteristics of carbon emissions in 393 counties. The fixed distance threshold method (FD) was used as the basis for the construction of the spatial weight matrix. The threshold distance was set as 145,819 m. The Moran’s I index was 0.3187, the p-value was 0.0000, and the Z-statistic was 21.5035, which indicates that the county-level carbon emissions in Southwest China showed significant spatial correlation characteristics. In other words, the carbon emissions of certain counties were influenced by the surrounding neighboring counties. The spatial interaction of carbon emissions between counties needs to be considered in the subsequent regression model selection of influencing factors.

3.2. Characteristics of Spatial Differences in Construction Land Types

The urban land scale, rural settlement land scale, and other construction land scale in 393 counties in Southwest China were divided into five class intervals (Figure 4). The higher the value, the larger the scale of various types of construction land. Overall, the distribution of construction land showed three centers of concentration, mostly in the urban core regions of provinces and cities (the Chengdu–Chongqing economic circle, the city group in central Yunnan, and the central Guizhou economic zone). Counties with urban land areas above 32.4166 km² were primarily concentrated in the central urban region of Chongqing Municipality; the Chengdu–Deyang–Mianyang–Leshan City Belt; the center of Kunming, the capital of Yunnan Province; and the central urban area of Guiyang, the capital of Guizhou Province. Counties with rural settlement land areas above 40.9846 km² were mainly distributed in Chengdu, Meishan, and Deyang in Sichuan Province and Kunming, Qujing, Yuxi, and Zhaotong in Yunnan Province. The number of counties with other construction land areas above 18.8446 km² was relatively high and was mainly distributed around Chongqing municipal district, Sichuan’s Chengdu-ring metropolitan area, and the city group in central Guizhou. In general, the spatial distribution of urban land scale, rural settlement land scale, and other construction land scale shows obvious spatial clustering difference characteristics.

3.3. Differential Characteristics of the Impact of Different Types of Construction Land on Carbon Emissions

Table 2. Results of collinearity test for factors influencing carbon emissions in Southwest China.

Variable	Tolerance	VIF
Urban land scale	0.3492	2.8639
Rural settlement land scale	0.6814	1.4676
Other construction land	0.4740	2.1098
GDP	0.1148	8.7122
Secondary industry output proportion in GDP	0.6948	1.4392
Residential population	0.2033	4.9196

illustrates the results of the multicollinearity test of the factors that influenced carbon emissions in Southwest China. There was no significant cointegration among the six variables, and they could be included in the regression model for the analysis.

On this basis, three models, namely, OLS, SLM, and SEM, were used to analyze the differential characteristics of the impact of construction land types on CO₂ emissions and to select the optimal model (

Table 3. Comparison of OLS, SLM, and SEM models.

Mode	Adjusted R2	AIC	Log-Likelihood	LM test	Robust LM Test
OLS	0.7066	698.6060	−342.3030	−	−
SLM	0.7356	662.5860	−323.2930	0.0000	0.0433
SEM	0.8092	552.8380	−269.4190	0.0000	0.0000

). The adjusted R² and log-likelihood of the spatial regression models (SLM, SEM) were higher than those of the OLS model, and the Akaike information criterion (AIC) value was lower, indicating that the spatial regression model was more suitable for this study. To determine the best spatial regression model in the two SLM and SEM spatial regression models, we also needed to analyze the LM test values. Both the LM-lag and LM-error statistics test results were noteworthy. On the one hand, this again verified the view that the spatial regression model was suitable for this study; on the other, it showed that it was still necessary to conduct robust LM-lag and robust LM-error statistics tests. The robust LM test showed that SEM was significant at the p < 0.01 level, which showed that the SEM passed the LM test. In addition, judging from the adjusted R², log-likelihood, and AIC criteria, it was likewise confirmed that SEM was the best model and more suitable for expplaining the variability of the impact of construction land type on CO₂ emissions.

In the SEM model, the regression coefficients and significance of the factors that influenced carbon emissions in Southwest China are shown (

Table 4. SEM model results.

Variable	Coefficient	Standard Error	Z-Value	p
Constant	−3.4583	0.3848	−8.9869	0.0000
Urban land	0.0642	0.0318	2.0184	0.0436
Rural settlement	0.0859	0.0254	3.3761	0.0007
Other construction land	0.1140	0.0262	4.3442	0.0000
GDP	0.8159	0.0712	11.4602	0.0000
Secondary industry output proportion in GDP	0.1378	0.0811	1.6984	0.0894
Residential population	−0.2512	0.0874	−2.8741	0.0041

). Among the six variables, all five variables had significant effects on carbon emissions, except for the secondary industry output proportion in GDP. Among them, the urban land scale was significant at p < 0.05, while the other four variables were significant at p < 0.01. This also confirmed that the evaluation index system constructed in this study was reasonable.

For every 1% increase in urban land area, CO₂ emissions increased by 0.0642%. The growth of urban land area had a positive impact on CO₂ emissions, and the carbon emissions effect gradually increased over a long period. Urban land consumes a large amount of material and energy and is the most important carbon source. If the significant expansion of the urban land scale is limited, the marginal carbon effect of the expansion of the urban land scale will be weakened.

For every 1% increase in the rural settlement land area, CO₂ emissions increased by 0.0859%. This value presents a simultaneous change between rural settlement land and carbon emissions. Rural settlement land is an important part of construction land and a relatively unique land-use pattern. The primary reason for the rise in CO₂ emission intensity in rural areas is the increase in energy consumption in agricultural production and rural households’ domestic energy use. Therefore, the two affect each other.

For every 1% increase in the other construction land area, CO₂ emissions increased by 0.1140%. This type of construction land, such as industrial and mining, oilfield, transportation, and special land, which is independent of towns, was the most important carrier of industrial production and energy consumption. At present, the other construction land in Southwest China is still in a relatively crude stage of utilization and expansion, and the industrial structure carried is still basically energy-intensive, with a relatively insufficient degree of land-use intensification, which has an extremely strong driving effect on carbon emission intensity.

In terms of control variables, GDP had the highest coefficient and had a very significant positive impact on CO₂ emissions, as each 1% increase in GDP increased carbon emissions by 0.8159%. By contrast, the resident population had a very considerable negative impact on CO₂ emissions, as each 1% increase in the resident population decreased carbon emissions by 0.2512%. The non-significant secondary industry output proportion in GDP indicated that the industrialization level of Southwest China was lagging, and secondary industry was not the dominant industry. Therefore, it played a relatively small role in carbon emissions and was not the primary cause of the rise in carbon emissions in Southwest China.

4. Discussion

The spatial variability characteristics of carbon emissions and construction land type indicate that there is a close connection between them. To eliminate the errors from the different models, this study found the best explanatory model by comparing three models, namely, OLS, SLM, and SEM. The results demonstrated that the SEM model had the largest adjusted R² and log-likelihood and the smallest AIC value, which is suitable for explaining the relationship between different types of construction land scale and carbon emissions. Based on the traditional test, we further introduced the Lagrange multiplier test (LM) to confirm the SEM as the best model again.

In the SEM model, the construction land type has a very significant influence on CO₂ emissions, and the direction of that impact was consistent with expectations. The intensity of the effect of different types of construction land on carbon emissions varied, and the other construction land scale had the highest effect on carbon emissions, followed by the rural settlement land scale, while the urban land scale had a slightly lower effect. This demonstrated that construction land types such as industrial and mining, oil field, transportation, and special land had large carbon emissions, which was also consistent with the existing relevant studies and theoretical perceptions. Furthermore, along with the carbon emissions caused by fossil fuel consumption in agricultural production, the carbon emissions from rural residential land also include rural household energy use. Rural settlement land in Southwest China is characterized by a large number, small scale, and wide distribution, and rural carbon emissions are rarely regulated. Therefore, the influence of rural settlement land on carbon emissions is becoming increasingly significant. The carbon emission intensity in urban areas is lower than in rural areas [39]. It also confirms the view that carbon emissions due to agricultural production and rural household energy consumption are much greater than those of urban construction [57]. While urban land is the embodiment of industrialization and urbanization, the complex and diverse topography of Southwest China restricts the urbanization process in the region, which, to a certain extent, limits the growth of energy consumption, and thus, has a limited impact on carbon emissions. Therefore, Southwest China should fully consider the different types of construction land and building functions to carry the intensity of human activities and develop differentiated carbon emissions reduction measures.

Although the correlation between construction land type and carbon emissions is intuitive, the negative correlation with the resident population was more puzzling. We speculate that this may have been a result of enhanced environmental policies and public environmental awareness, whereby people were aware of carbon emissions hazards and used more renewable and clean energy in their daily lives, resulting in a gradual reduction of the carbon emissions effect. The impact of knowledge, technology, and human capital on carbon emissions far exceeds the population size, and highly knowledgeable and skilled people play an important role in reducing carbon emission intensity [58]. Increasing the use and penetration of renewable energy and clean energy in people’s daily production and life carbon emissions.

This study responded to the issue of carbon emission differences within the region by looking at the differences in the impact of different types of construction land on CO₂ emissions and provides some empirical evidence for relevant research and policy formulation in this field. In view of this, corresponding policies can be formulated in the following aspects: first, by adjusting the land-use structure, strengthening land-use control and zoning control, and strictly controlling the occupation of land with high carbon sink capacity by construction land; second, by promoting new urbanization in Southwest China in an orderly manner and promoting the linkage of land use increase and decrease (moderate growth in urban development land and simultaneous reduction in rural settlement land); third, by planning cities and towns in such a way as to maintain a reasonable proportion of industrial land and other construction land in the region to suppress or mitigate the impact of carbon emissions.

It is worth noting that this study explored the impact of construction land types on CO₂ emissions and their intensity differences at the global level, and there may be spatial heterogeneity in the effect of construction land types on carbon emissions that was not considered in this study. The spatial heterogeneity of the influence of construction land types on carbon emissions can be further analyzed in the future using methods such as geographically weighted regression. The classification of construction land types in this study was mainly based on the Chinese Academy of Sciences secondary land-use classification standard, which can be subdivided into industrial and mining warehousing, commercial services, residential, and public services in the future based on the function of construction land. Meanwhile, the correlation between different land-use types and carbon emissions in Southwest China and the direction and intensity of the effect can be analyzed in depth using traditional regression and spatial regression analysis methods in the future.

5. Conclusions

This paper discusses the variability of the impact of construction land types on carbon emissions using a variety of econometric models, including traditional regression and spatial regression, in an ecologically fragile southwest region and draws the following main conclusions.

Carbon emissions in Southwest China were very spatially variable and spatially dependent. On the one hand, carbon emissions were generally distributed in clusters, and high CO₂ emission regions were mostly clustered in the suburbs of megacities and typical industrial cities; on the other hand, Moran’s I index showed that carbon emissions in Southwest China had significant spatial correlation and clustering. Similarly, the distribution of different types of construction land in Southwest China showed unevenness, and the urban land scale, rural settlement land scale, and other construction land scale all showed obvious spatial clustering differences.

The SEM model was utilized to detect the relationship between the construction land type and carbon emissions and its significance. The results showed that five factors had statistically significant effects on carbon emissions, which did not include the secondary industry output proportion in GDP. These five factors were the urban land scale, rural settlement land scale, other construction land scale, GDP, and resident population. In addition, the construction land type was significantly and positively correlated with carbon emissions, and the direction of influence was in line with theoretical expectations. The nature of various types of construction land differs, as does the intensity of its effect on carbon emissions. The degree of its effect was as follows: the other construction land scale had the highest effect on carbon emissions, followed by the rural settlement land scale, while the urban land scale had a slightly lower effect. Further, this study found that the spatial regression model, after comparing the OLS model and spatial regression models (SLM, SEM), considered the spatial interrelationship that existed between county units. Therefore, the processed data had a higher degree of fit and better reflected the spatial dependence of county carbon emissions.

In view of the important influence of construction land types on carbon emissions, government departments should not only consider traditional factors, such as regional economic development level, energy structure and efficiency, industrial structure, population size, and technological progress, when formulating carbon emissions reduction policies but also consider the differences in the impact of specific categories of construction land to effectively reduce atmospheric CO₂ concentrations and develop more refined carbon emissions reduction strategies.