Investigating the Impacts of Built-Up Land Allocation on Carbon Emissions in 88 Cities of the Yangtze River Economic Belt Based on Panel Regressions

Abstract

The supply of built-up land determines the depths of human activities, leading to the differences in scale and intensity of carbon emissions. However, the relationship between the composition of built-up land and carbon emissions has not been fully investigated. In response, this study collects the panel data of 88 cities along the Yangtze River Economic Belt, China, and uses the fixed effect model and system GMM model, to explore the impacts of specific subtypes of built-up land on carbon emissions averaged by economic output and urban land. The findings show that industrial land and commercial land are the main contributors to increase carbon emissions; the increased proportions of land subtypes related to supporting facilities and infrastructures show significant restraining effects; carbon emission was a dynamic process with time-lagged effects. As a result, reallocating the structure of urban built-up land can directly and indirectly adjust the intensity of carbon emissions. Policy recommendations focus on the balanced supplies of production and ecological land.

1. Introduction

In recent decades, global warming, the most challenging climate change, has caused multiple natural disasters such as biodiversity loss, extreme weather, and sea level rise. This environmental crisis has aroused widespread concerns of scholars and decision makers due to its destructive and unpredictable nature [1,2]. Carbon emissions produced by increased human activities are widely regarded as the main source causing global warming [3]. Land resources is a basic carrier of the prosperity of economy and society, undoubtfully influencing the scale and intensity of carbon emissions [4,5]. It is estimated that more than one-quarter of global carbon emissions was influenced by the land-use changes [6]. Generally, the watershed-based economic belt is a sensitive area where land-use conflicts frequently occur between economic activities and environmental conservation, making it a hot focus of research on carbon emission and carbon sink [7,8]. In 2018, the Intergovernmental Panel on Climate Change (IPCC) claimed the urgency of a 1.5 °C temperature control target [9]. At the general debate of the 75th Session of the United Nations General Assembly in 2020, China announced its ambitions to realize the carbon peaking and carbon neutrality (“double carbon”) goals in the mid-21st century, which stressed the importance of adjusting the allocation of natural resources to control and reduce the carbon emissions [10].

Facing the goal of “double carbon”, carbon emission has become an important measurement index for environmental governance and land use management in China. Researchers propose multiple methods to measure carbon emissions [11,12,13,14,15,16]. These methods differ due to multiple perspectives. Zheng et al. (2018) calculated the carbon emission by estimating the production of industry and service, which means the emission is production-based [11]. Mi et al. (2016) measured carbon emission based on the consumption of goods and services as well as their carbon footprints [12]. David et al. (2011) proposed a method that measures carbon emissions using the consumption of fossil fuels [13]. Lenzen et al. (2021) considered carbon emissions from multiple scopes of human activities within a city, and made a thematic dataset covering global cities [14]. Some studies also calculated the carbon emission account based on both fossil fuel consumption and economic activities by covering the industrial chains [15,16]. The progress of measurement methods makes it available and convenient for carbon-emission-related studies.

Carbon emission occurs but differs with respect to the production on or with land resources. Scholars have studied the relationship between land resources and carbon emissions. Han et al. (2022) and Ma et al. (2021) indicated that the misallocation of land resources impacted the land market, hindered the upgrade of industrial structure, and further exacerbated urban carbon emissions [17,18]. Among all the land-use types, urban built-up land is the greatest contributor to carbon emissions due to its constructive purposes. Bae and Ryu (2020) pointed out that over 80% of the world’s greenhouse gases were emitted from urban lands [19]. Yu et al. (2022) indicated that the agglomeration of cities has temporally heterogeneous impacts on carbon intensity [20]. Liu et al. (2019) reported that energy-related carbon emissions can be restricted by limiting built-up land expansion [21]. In detail, Zhang et al. (2021) found that higher utilization efficiency of industrial land can benefit to reduce regional carbon emissions [22]. Zhou et al. (2022) revealed that the misallocation of industrial land and commercial land produced a negative impact on carbon emission efficiency by limiting the development of high-end industries [23]. Moreover, studies suggested that optimizing land-use structures helped to reduce carbon emissions and increase carbon sink [24,25]. In sum, the existing literature explored the impacts of land resources on carbon emissions by solely focusing on one kind or a whole of land use. More details about these effects were hidden. The individual effect of each built-up land subtype on carbon emission was rarely explored.

The existing literature has revealed factors that impact carbon emissions from multiple aspects. The demographic factor was a basic representative of human activities, Liu et al. (2021) reported that the increase in urban population size led to an increase in total CO₂ emissions and a decrease in per capita CO₂ emission [26]. Studies also showed that the investment in science and technology contributed to industrial upgrading, thus leading to the improvement of overall production efficiency [27,28]; however, the investment in green technology benefited to reduce the carbon emission of unit product production [29]. For other economic factors, Shahzad et al. (2017) reported that the increase in trade openness and financial development can promote carbon emission [30]. Muhammad et al. (2019) indicated that the trade situation can reflect the ecological niche of leading industry in the production and sales chain of a city, which had different impacts on carbon emissions [31]. For the industrial structure, Dong et al. (2020) reported that different industries played different roles in carbon emissions [32].

For the above reasons, this paper aims to decompose the built-up land type into subtypes, focus on how the subtypes of built-up land impact carbon emissions, and distinguish the role how each land subtype plays in a same influential system. In this way, this paper provides more evidences on the importance of allocating the structures of urban land use, and recommends efficient land policies under the context of carbon emission restriction and reduction. By using the panel data of 88 cities covering 11 provinces that belong to the area of the Yangtze River Economic Belt (YREB) from 2012 to 2019, and taking industrial land, residential land, commercial land, green space and square land, public management and public service land, municipal utilities land, and transportation land as the cases for empirical study, this paper adopts penal regression model and System GMM model to investigate the impacts of specific built-up land subtypes on carbon emissions. This study contributes to the existing literature in three ways. First, it decomposes the built-up land into subtypes to explore more details of the effects. Second, it confirms the roles of each land subtype rather than a single subtype in impacting carbon emission. Third, it explores the time-lagged effects of both carbon emission and the subtypes of built-up land on the premise of the models’ robustness.

2. Materials and Methods

2.1. Research Area

In China, the Yangtze River Economic Belt (YREB) plays an important role in both the national economy and ecology. It includes 11 provinces and municipalities, including Shanghai, Jiangsu, Zhejiang, Anhui, Jiangxi, Hubei, Hunan, Chongqing, Sichuan, Yunnan, and Guizhou, covering an area of 2.05 million square kilometers, accounting for 21.4% of the total land area. In 2018, the total population of the YREB was 599 million, accounting for 42.9 percent of the national total. The population in the upper, middle, and lower reaches of the YREB was 199, 175, and 225 million, accounting for 33.2%, 29.2% and 37.6% of the total number of YREB, respectively. As for the economic output, the GDP in the upper, middle, and lower reaches of the YREB was 9.37, 9.78, and 21.15 trillion RMB, accounting for 23.3%, 24.3%, and 52% of the total in YREB. The total GDP in YREB is about CNY40.3 trillion, accounting for 44.1% of the country’s total.

Given the substantial economic activities and population agglomerations in YREB, it is of great importance to investigate the environmental effects of land-use change or allocation locally. Limited by the data, we selected 88 cities instead of the full coverage of YREB as the research area (Figure 1).

2.2. Methods

2.2.1. STIRPAT Model

The IPAT (Impacts, demographic, Affluence, Technology) model, originally proposed by Ehrlich and Holdren (1971) [33], was used to analyze the relationship between socioeconomic development and environmental problems. Among the improved and expanded patterns of the IPAT model, the ImPAT model and STIRPAT model were most widely used [34]. The basic formula of the STIRPAT model is as follows:

$$I=a\times P^b\times A^c\times T^d\times e$$

(1)

where P, A, and T represent the demographic, wealth, and technical factors; I denotes the environmental stress factor; a represents the model coefficient; b, c, and d represent the elasticity coefficient of demographic, wealth, and technical factor; and e denotes the error term.

The STIRPAT model is a stochastic representation of the IPAT equation that verifies the nonlinear hypotheses such as the EKC curves and decomposes and recombines the variables related to the research subject. Wang et al. (2016) combined the STIRPAT model and panel quantile to measure the impacts of socioeconomic data on carbon emissions [35]. Based on it, this paper extended the STIRPAT model by combining it with the panel regression model to process the multi-year data.

2.2.2. Modeling the Relationship between Built-Up Land Area and Averaged Carbon Emission

We first modeled the relationship between built-up land area and the carbon emission from the aspect of carbon emission intensity (CEI) and carbon emission per built-up land (CEB). Following the concept of the STIRPAT model, we took the derivative on both sides of the equation, and the transformed and standardized equation is as follows:

$$C{E}_{it}={\alpha }_0+\beta B{L}_{it}+{\gamma }_{n}C{V}_{nit}+{\epsilon }_{it}+V_t$$

(2)

In which,

$$C{V}_{nit}={\gamma }_{1}TE{C}_{it}+{\gamma }_{2}PU{L}_{it}+{\gamma }_{3}GP{C}_{it}+{\gamma }_{4}P{T}_{it}+{\gamma }_{5}I2GD{P}_{it}+{\gamma }_{6}I3GDP$$

(2a)

where t denotes the year, i denotes the city; CE denotes the carbon emission factors, which are measured by CEI and CEB, respectively; BL denotes the built-up land area; CV_n indicates the set of n control variables, which herein contains the factors of science and technology (TEC), population density (PUL), economic output (GPC), trade level (PT), and industrial structure (I2GDP, I3GDP); V is the time dummy variable; α₀ is a constant term, β denotes the coefficient of core explanatory variables, γ_n is the coefficient of the nth control variable, ε is the error term.

2.2.3. Modeling the Relationship between Individual Subtypes of Built-Up Land and Carbon Emissions

Similar to Equation (2), we changed the formula by replacing the land-use subtype variables with the total land area. The basic formula is as follows:

$$C{E}_{it}={\alpha }_0+{\beta }_{m}SB{L}_{mit}+{\gamma }_{n}C{V}_{nit}+{\epsilon }_{it}+V_t$$

(3)

In which,

$${\beta }_{m}SB{L}_{mit}={\beta }_{1}IB{L}_{it}+{\beta }_{2}CSB{L}_{it}+{\beta }_{3}RB{L}_{it}+{\beta }_{4}GSB{L}_{it}+{\beta }_{5}APB{L}_{it}+{\beta }_{6}RTB{L}_{it}+{\beta }_{7}MB{L}_{it}$$

(3a)

where t denotes the year, i denotes the city; CE denotes the carbon emission factors, which are measured by CEI and CEB, respectively; SBL_m means the set of m subtypes of built-up land, which contains at most industrial land (IBL), commercial land (CSBL), residential land (RBL), green space and square land (GSBL), public management and public service land (APBL), transportation land (RTBL), and municipal utilities land (MBL); CV_n indicates the set of n control variables; V is the time dummy variable; α₀ is a constant term, β_m denotes the coefficient of m core explanatory variables, γ_n is the coefficient of the nth control variable, ε is the error term.

Among the optional subtypes, industrial land and commercial land are the main carriers for production and consumption and relate most closely to carbon emission [22,23,36,37,38]. After a series of exploratory attempts, we built model to explore the impacts of these two subtypes by including two corresponding variables in the same equation (shown in Equation (3b)). Further, we focused on the rest subtypes of built-up land and clarified the impact of them by neglecting the variables of industrial land and commercial land, which is shown in Equation (3c).

$$C{E}_{it}={\alpha }_0+{\beta }_{1}IB{L}_{it}+{\beta }_{2}CSB{L}_{it}+{\beta }_{3}RB{L}_{it}+{\beta }_{4}GSB{L}_{it}+{\beta }_{5}APB{L}_{it}+{\beta }_{6}RTB{L}_{it}+{\beta }_{7}MB{L}_{it}+{\gamma }_{n}C{V}_{nit}+{\epsilon }_{it}+V_t$$

(3b)

$$C{E}_{it}={\alpha }_0+{\beta }_{1}IB{L}_{it}+{\beta }_{2}CSB{L}_{it}+{\gamma }_{n}C{V}_{nit}+{\epsilon }_{it}+V_t$$

(3c)

$$C{E}_{it}={\alpha }_0+{\beta }_{1}RB{L}_{it}+{\beta }_{2}GSB{L}_{it}+{\beta }_{3}APB{L}_{it}+{\beta }_{4}RTB{L}_{it}+{\beta }_{5}MB{L}_{it}+{\gamma }_{n}C{V}_{nit}+{\epsilon }_{it}+V_t$$

(3d)

where the interpretation of all letters and variables is consistent with Equations (3) and (3a).

2.2.4. Modeling the Relationship between Individual Subtype of Built-Up Land and Carbon Emission by Considering the Time-Lagged Effects

The utilization of both static and dynamic model estimations is necessary to ensure the robustness of the findings. To relieve the endogeneity of the modeling, we introduced the system-GMM model which generally includes the first-order or more lagged terms of response variables as the explanatory one [39]. The basic formula is as follows:

$$C{E}_{it}={\alpha }_0+{\theta }_{k}C{E}_{it-k}+{\beta }_{m}SB{L}_{mit}+{\gamma }_{n}C{V}_{nit}+{\epsilon }_{it}+V_t$$

(4)

where t denotes the year, i denotes the city, k is the lagged order; CE denotes the carbon emission factors, which are measured by CEI and CEB, respectively; SBL_m means the set of m subtypes of built-up land; CV_n indicates the set of n control variables; V is the time dummy variable; α₀ is a constant term, θ denotes the coefficient of the lagged term of the response variable, β_m denotes the coefficient of m core explanatory variables, γ_n is the coefficient of the nth control variable, and ε is the error term.

Commercial land and industrial land were chosen as the explanatory variable in Equation (4a). The rest land subtypes related to supporting facilities and infrastructures were used in Equation (4b). In these models, the lagged terms were added to explore the time-lagged effects and to confirm the robustness of the models we built previously [40].

$$C{E}_{it}={\alpha }_0+{\theta }_{k}C{E}_{it-k}+{\beta }_{1}IB{L}_{it}+{\beta }_{2}CSB{L}_{it}+{\gamma }_{n}C{V}_{nit}+{\epsilon }_{it}+V_t$$

(4a)

$$C{E}_{it}={\alpha }_0+{\theta }_{k}C{E}_{it-k}+{\beta }_{1}RB{L}_{it}+{\beta }_{2}GSB{L}_{it}+{\beta }_{3}APB{L}_{it}+{\beta }_{4}RTB{L}_{it}+{\beta }_{5}MB{L}_{it}+{\gamma }_{n}C{V}_{nit}+{\epsilon }_{it}+V_t$$

(4b)

where θ denotes the coefficient of the lagged term of the response variable, the interpretation of all the other letters and variables is consistent with Equations (3), (3a) and (4).

2.3. Variables

2.3.1. Response Variable

Carbon emission comes from not only the production links, but also the consumption links and other human activities. Therefore, this study applied the total carbon emissions of energy consumption, production of building materials and activities of various social sectors as the response variable. We measured it through two aspects, the first way is by dividing the total amount of carbon emission within a city by the total area of the built-up land in this city, and the second one is the carbon emission intensity which is the ratio of the emission of CO₂ to the GDP of the city.

2.3.2. Explanatory Variables

The explanatory variables cover the overall built-up land and its main subtypes. Limited by the data sources, the total area of built-up land and 7 subtypes of built-up land were selected as the explanatory variables. In detail, the built-up land subtypes included industrial land, commercial land, residential land, green space and square land, public management and public service land, municipal utilities land, and transportation land. The data value of every subtype we used is the proportion of the area of each built-up land subtype to the total area of built-up land.

The descriptions of urban built-up land and its subtypes are as follows.

• Urban built-up land. Land used for engineering, construction, and other facilities within the jurisdiction of the city.
• Industrial land. Land used for industrial and mining production.
• Commercial land. Land used for business, commercial affairs, and recreational facilities.
• Residential land. Land used for urban housing and its supporting community service facilities.
• Green space and square land. Land used for public space such as park, protective green space, and square within urban area, excluding the ancillary green space in other built-up lands.
• Public management and public service land. Land used for institutions and facilities such as government apparatus, scientific research, culture, education, sports, health, and social welfare.
• Municipal utilities land. Land used for water supply, drainage, power supply, gas supply, heat supply, communication, postal, radio and television, sanitation, fire protection, and other facilities for urban infrastructure.
• Transportation land. Land used for railway, highway, airport, port and wharf, urban rail transit, roads, stations, and other ancillary facilities.

2.3.3. Control Variables

We also selected five variables that directly influence carbon emissions to control for the expanded form of the STIRPAT model.

Science and technology. The influences of science and technology on the efficiency of carbon emissions cannot be ignored [41,42]. Therefore, the proportion of annual investment in science and technology to the public financial expenditure is used to measure the city’s investment in science and technology (TEC), the equation is as follows:

TEC = Science and technology expenditure/Public financial expenditure

Trade Level. The import and export of goods can be regarded as a kind of inward or outward flow of carbon account. In this study, the foreign trade coefficient was used as a measure of the trade situation of a city. We used it by measuring the proportion of both imports and exports in the gross domestic production of the city (IEG), the equation is as follows:

IEG = (Import trade volume + Export trade volume)/Gross domestic production

Industrial structure. The differences in industrial composition have different impacts on carbon emission [43]. The proportion of the output of secondary (I2GDP) and tertiary industry (I3GDP) to the total were, respectively, included as the variables that present the industrial structure, the equations are as follows:

I2GDP = Secondary industry output/Gross domestic production

I3GDP = Tertiary industry output/Gross domestic production

Population Density. Regarding the vital role of urban human activities in carbon emission, we chose the data of urban population and urban built-up land. We measured this variable by using the urban population divided by urban built-up land (PUL), the equation is as follows:

PUL = Permanent urban resident population/Area of urban built-up land

Economic output. Considering that carbon emission is a typical consequence of economic activities, both the intensity and efficiency of carbon emission have close linkages to the scale of economic output [44]. We thus measured it by averaging the gross domestic production by the population of the urban area (GPC), the equation is as follows:

GPC = Gross domestic production/ Permanent urban resident population

The descriptive statistics of all variables were shown in

Table 1. Descriptive statistics of the variables.

Type	Variable	Mean	Standard Deviation	Maximum	Minimum	Unit
Response variable	CEB	28.5607	22.6869	235.1588	3.0753	1 × 105 tons/km2
	CEI	1.6006	1.5200	10.6387	0.2154	1 × 105 tons/1 × 108 RMB
Explanatory variable	IBL	20.6646	7.6861	40.3865	2.3530	%
	CSBL	7.1340	3.9129	27.7605	0.1520	%
	RBL	30.6566	6.1676	58.0884	8.2831	%
	GSBL	11.2496	5.2730	37.6058	0.0702	%
	APBL	9.2060	3.2017	20.9677	1.8779	%
	RTBL	14.5491	4.5076	24.4000	0.5369	%
	MBL	3.8266	3.2829	19.9924	0.0301	%
	BL	157.3798	187.1139	1319.1600	21.2300	km2
Controlled variable	TEC	0.025	1.844	13.966	0.261	%
	IEG	24.9715	26.2040	190.2347	1.1013	%
	I2GDP	48.629	7.882	75.860	1.930	%
	I3GDP	41.452	8.921	77.490	20.660	%
	PUL	0.7019	0.1770	1.4416	0.3147	1 × 104 persons /km2
	GPC	6.0026	4.4187	26.6070	0.9256	1 × 108 RMB/1 × 104 persons

.

2.4. Data Sources

The analysis was carried out on a dataset of 88 cities at the prefecture level or above, which are located along the Yangtze River Economic Belt ranging from 2012 to 2019. The carbon emission data were obtained from an open-source database named Carbon Emission Accounts and Datasets (CEADS). The calculation of each city’s carbon emission covered the consumption of 17 fossil fuels by 47 socio-economic sectors of a city, using emission factors of industrial processes in each sector developed by the IPCC and NDRC (National Development and Reform Commission of China). For example, the data of the socio-economic sector was retrieved from the energy balance table in a city’s yearbook; for the cities with no recorded energy balance table, their sectors’ consumptions were estimated by considering GDP and population structure. Another example is that the emission from the production process of building materials is also considered [15,45]. Land supply data and demographic data were obtained from the Statistical Yearbooks of China Urban Construction. Other social and economic data were all obtained from the China Statistical Yearbook. Some data were obtained via secondary calculation. Some obvious discrete and incorrect data were revised according to the annual report of the government. All data were standardized to eliminate the influences of scale and unit.

3. Results

3.1. Impacts of Individual Land Subtype of Built-Up Land on Carbon Emission

The over-identification test and Hausman test could compare whether our data fit the random effect model (REM) or the fixed effect model (FEM) and determine which model should be selected. According to the over-identification test and Hausman test outcomes in

Table 2. Results of over-identification test and Hausman test.

	Model	1	2	3	4
CEI	over-identification test (chi2)	26.1300 ***	45.9050 ***	37.8900 ***	34.1990 ***
	Hausman test (chi2)	38.1400 ***	57.1700 ***	56.1800 ***	38.1000 ***
	Model	5	6	7	8
CEB	over-identification test (chi2)	56.8920 ***	52.2880 ***	48.6970 ***	48.3150 ***
	Hausman test (chi2)	28.7400 ***	58.0400 ***	53.6300 ***	40.8900 ***

Note: *** denotes statistical significance at the 1% levels.

, the two-factor fixed effect models were more suitable compared to the random effect one.

Table 3. Results of regression using the two-factor fixed effect model.

		Model 1	Model 2	Model 3	Model 4	Model 5	Model 6	Model 7	Model 8
Explanatory variable	BL	0.2401 **				0.0849
	IBL		0.1222	0.1577 *			0.1613 **	0.1698 **
	CSBL		0.0893 *	0.1184 ***			0.0931 **	0.1041 **
	RBL		−0.1905		−0.3033 *		−0.162		−0.3066 *
	GSBL		−0.0427		−0.1448 *		−0.028		−0.1590 **
	APBL		−0.0234		−0.0695 **		−0.0022		−0.0606 *
	MBL		−0.0381		−0.0913 **		0.011		−0.0587 *
	RTBL		−0.0134		−0.0693 **		−0.0007		−0.0734 **
Controlled variable	TEC	0.0874	0.1079	0.0847	0.106	0.0332	0.055	0.0273	0.0495
	IEG	−0.1893 **	−0.1348 **	−0.1274 *	−0.1321 **	−0.2107 ***	−0.1888 ***	−0.1885 ***	−0.1853 ***
	I2GDP	0.1387	0.1513 **	0.1719 *	0.1394 *	0.0412	0.0492	0.0715	0.0354
	I3GDP	0.1275	0.1338 *	0.1386	0.1280 *	0.0308	0.0282	0.0435	0.0207
	PUL	0.0853 *	0.0496	0.0482	0.0543	0.1617 ***	0.1457 ***	0.1437 ***	0.1484 ***
	GPC	0.1592	0.1717 *	0.1749	0.1669 *	0.2483 **	0.2340 **	0.2599 **	0.2342 **
	Constant	27.2930 ***	28.3812 ***	26.0222 ***	28.4755 ***	10.0002	8.4518	8.9198 ***	8.7258 ***
	N	704	704	704	704	704	704	704	704
	F	12.5200 ***	7.9000 ***	11.9400 ***	8.8100 ***	3.9600 ***	2.8400 ***	3.9700 ***	2.6100 **
	R2	0.243	0.293	0.27	0.2868	0.1453	0.2386	0.201	0.2224

Note: *, **, *** denote statistical significance at the 10%, 5%, 1% levels, respectively. The response variable in models 1–4 is CEI, and CEB in model 5–8.

shows the results of models 1–8 regressed by the two-factor fixed effect panel data model, which revealed the relationship between the subtypes of the built-up land and the carbon emission of the city. Specifically, models 1–4 correspond to Equations (2a), (3b)–(3d), so do model 5–8.

In Model 1, the coefficient of the built-up land area was 0.2401 at a 5% level significance, meaning that every 1 unit increase of the built-up land area led to a 0.2401 unit increase in CEI. It shows a significant positive effect of the supplies of the built-up land to a city’s production and its carbon emission. While in Model 5, the coefficient of the built-up land area was 0.0849 but not significant, indicating that the change in the total area of the built-up land has no statistic relation to either the promotion or the reduction in the CEB.

In Model 2, the coefficient of commercial land was 0.0893 and the significance was at a 10% level, illustrating that 1 unit increase in the ratio of the commercial land could bring a 0.0893 unit increase in CEI, while the coefficient of industrial land was 0.1222 and failed to pass the significance test; the coefficient of other subtypes of built-up land was all negative with no significance. The same phenomenon also occurred when setting the CEB as the response variable. In Model 6, the coefficient of commercial land was 0.0931 with a significance of a 5% level, and the coefficient of industrial land was 0.1613 passing the significance test at a 5% level.

From the results shown in Model 2 and Model 6, industrial land and commercial land proportions show a significant impact on the CEI and CEB. These two land types are separately the main carrying space of the second and third industries, which are highly related to the production of a city’s land for industry and business driving economic growth and the consequent carbon emission. It is noteworthy that, compared with CEB, changing the ratio of industrial land shows no significant impact on the CEI. A possible explanation is that limiting or reducing the intensity of carbon emissions relies heavily on pushing technical advancement rather than regulating a certain type of land. For the rest of the subtypes of built-up land, their little impacts on carbon emission can be due to the overwhelming need for the two main land types for production and business, namely industry land and commercial land.

Model 3 and Model 7 analyzed the impacts of industrial land and commercial land. When putting the CEI as the explained variable, the coefficient of the industrial land was 0.1577 at a 10% significance, and the coefficient of the commercial land was 0.1184 passing the 1% significance. When setting CEB as the response variable, the coefficient of the industrial land was 0.1698 and its significance level is 5%, and that of commercial land was 0.1041 at a 5% level of significance. It indicates that increasing the proportion of industrial land and commercial land has significant positive impacts to CEI and CEB. The carbon emission generated by the increase in industrial land is more than that of commercial land, revealing the differences in energy and resource consumption between the secondary and tertiary industries that generate economic growth. A similar phenomenon could be confirmed by the coefficient of industry structure, the impact of the proportion of secondary industry to the CEI is significant and higher than that of the proportion of tertiary industry.

All selected variables of land subtypes in models 4 and 8 show their significantly negative impacts to both CEI and CEB. In Model 4, the coefficient of residential land was −0.3033 at 10% significance, and that of green space and square land was −0.1448, passing the 10%-level significance test. Public management and public service land also negatively impact carbon emission, whose value is −0.0695 and significant at a 5% level. The coefficient of municipal utilities land was −0.0913 at 5%-level significance. Compared to other land subtypes, transportation land has the least but significantly negative impacts (−0.0693) gas emissions. In Model 8, the impacting value of residential land was −0.3066, greater than that of CEI. The coefficient of green space and square land was −0.1590 and its significance level is 5%; public management and public service land had less negative but significant impacts, −0.0606, compared to −0.0695 in the CEI model. The coefficients of municipal utilities land and transportation land were −0.0587 and −0.0734, passing the 10% and 5% level significance tests, separately.

From the results of Table 3, we conclude that the performance of all subtypes of built-up land can be divided into two groups. In one group, increasing the supplies of the lands whose main functions are carrying the expansions of high-intensity economic activities not only indicates the promoting the consumption of certain amounts of energy and building material, but also the continuous input of other resources, such as electricity, cement, and crops, for production and business. The exploitations of these lands directly and indirectly lead to the growth of carbon emissions. In the other group, the main functions of the land subtypes are carrying the expansions of supporting facilities and infrastructures, the increase in the proportion of these land types contributes to limiting and even reducing the carbon emission. For example, playing as a carbon sink land, green space and square land can carry more planting activities that can be used to absorb and neutralize carbon dioxide. Moreover, these non-industrial lands can help increase the efficiency of total production locally to decrease the carbon emission emitted during production and consumption.

3.2. Impacts of Individual Land Subtype of Built-Up Land Considering Temporally Lagged Terms

Table 4. Results of the parameter estimation and test outcomes of Sys-GMM model.

		Model 9	Model 10	Model 11	Model 12
Explanatory variable	CEBt-1	0.8217 ***	0.6238 ***
	CEIt-1			0.7075 ***	0.7232 ***
	CEIt-2			0.1778 **	0.1783 **
	IBL	0.0236 *		0.0329 ***
	CSBL	0.0033		0.0116
	RBL		−0.0215		−0.0425 **
	GSBL		−0.0328 **		−0.0265 *
	APBL		−0.0187		−0.0205 **
	MBL		−0.0145		−0.0214 *
	RTBL		−0.0252 *		−0.0222
Controlled variable	TEC	0.0112	0.0203	0.0057	0.0071
	IEG	−0.0950 ***	−0.1597***	−0.0040	0.0018
	I2GDP	−0.0720	−0.0283	0.0367	0.0319
	I3GDP	−0.0306	−0.021	0.0782 ***	0.0770 ***
	PUL	0.0591 ***	0.0875 ***	0.0003	0.0009
	GPC	0.0609 ***	0.0922 ***	−0.0346 **	−0.0359 **
	Constant	−1.5582	−3.2141 **	4.2404 **	4.2306 *
	N	616	616	528	528
p-value of tests	AR(1)	0.0070	0.0130	0.0110	0.0100
	AR(2)	0.1260	0.1300	0.1780	0.1840
	Hansen’J	0.8340	0.7290	0.6460	0.5690

Note: *, **, *** denote statistical significance at the 10%, 5%, 1% levels, respectively. The response variable in Model 9–10 is CEB, and CEI in model 11–12.

shows the results of Model 9-12 regressed by the system-GMM model, which revealed the relationship between the built-up land subtypes and the carbon emission by considering the time-lagged effects. Specifically, model 9/11 corresponded to Equation (4a), and model 10/12 corresponded to Equation (4b). Our dynamic panel data models passed the AR (1) and AR (2) tests. In all models, AR (1) was significantly less than 0.05 while AR (2) was not significant, indicating that the response variable has a temporal correlation with its first-order lagged term. Our models also passed the Hansen J test, proving that the lagged term of the response variable, using as the instrument variables, was selected in an appropriate way, and there is no over-identification problem [46].

As shown in Table 4, the first-order lagged terms of CEB in Model 9 and Model 10 were both positive, and significant at the level of 1%. The coefficients are 0.8217 and 0.6238. It illustrates that CEB was positively impacted by the changes in emission in the previous year. The carbon emission averaged by the input of built-up land has a cumulative effect due to the historic performance of it. A similar phenomenon could be found in Model 11 and Model 12, whose coefficients of the first-order and the second-order lagged terms of CEI were significantly positive. For the contemporaneous CEI, the impacts from the emission in the previous year were greater than that in an earlier time.

Drawing back to Model 9, it is found that, the coefficient of industrial land was 0.0236 and significant at a 10% level, while the coefficient of commercial land was positive but failed to pass the significance test. Industrial land still plays an important space in emitting the carbon emission among all subtypes of built-up land. The lagged impacts of previous emissions possibly mean that supplying industrial land can have an impact on carbon emission in the next year. In Model 10, the coefficients of green space and square land and transportation land are negatively passing the 10% significance test, similar to the results in the fixed effect panel data models. However, the coefficients of other subtypes are also negative but insignificant. The increase in green land contributes to reducing and absorbing greenhouse gases for a long period. The expansion of the transportation system can absolutely improve economic efficiency, and accelerate the development of low-carbon industries, restraining carbon emissions [47].

In Model 11, the coefficient of industrial land was 0.0329 and significant at 5% level, while that of commercial land is positive but insignificant. Obviously, the lands for industrial purposes carry expansive economic activities, challenging the intensity of carbon emission influenced by economic output. More interestingly, the historic (lasts for two years) CEIs exerted the enhancing effects to the present one, performing more significant cumulative effects of time series. However, the cumulative effects were slightly more in Model 12, which can be due to the existence of the offsetting effects of supplying specific land subtypes. The statistically significant coefficients of residential land, green space and square land, public management and public service land, and municipal utilities land were all negative, like them in the fixed effect models. The impact of residential land was obviously greater than that of other land types, and the possible reason is that supplying more residential lands can increase GDP through the market of real estate, with a relatively small increase in carbon dioxide as the cost.

Compared with the two-factor fixed effect panel data method, all explanatory variables in the system GMM model did not change the direction of their impacts, but the absolute value of their coefficients was smaller. It can be inferred that the time-lagged effects, also called the cumulative effects, to both CEB and CEI cannot be ignored when trying to adjust the structure of built-up land, and carbon emission reduction is a dynamic process. According to the performance of the variables that present the levels of trade and economic output, expanding overseas carbon accounts and increasing per capita output will be beneficial to reducing CEB and CEI, respectively.

4. Discussion

4.1. The Impact Mechanism of the Subtypes of Built-Up Land

The land resource is the basis for human activities. In general, carbon emission is a by-product of economic activities, which mostly occur in built-up land. From the aspect of urban land use, different land use function corresponds to its own unique way to transform energy and resources into the matters the city needed, producing various scales of emissions. That means reallocating the structure of built-up land can adjust the total amount of carbon emission. Ideally, the increase in built-up land can generate more carbon emissions compared to the other land types such as forest land and agricultural land. However, some land subtypes within urban areas, such as green space and square land, can even have the function of carbon sink. Aside from them, providing land for industry and business as much as possible can get the best and long-term economic returns; however, they have the worst environmental effects, including unpredictable carbon dioxide emissions.

Our study on the quantitative relationship between urban land supplies and carbon emissions concludes the generalized impact mechanism of built-up land subtypes among YREB cities. Following the previous studies, it is confirmed again that industrial land and commercial land are the largest contributors for carbon emissions from the perspective of urban land use [22,23,37,38]. The positive role of green space and square land in limiting carbon emission can be also easily understood due to the capacity of carbon fixation of green plant on it. On a larger scale, a city can be regarded as a giant and complicated system. Supplies of some land subtypes can provide a better living environment and a more efficient operating conditions for the residents. Increasing these types of land will improve the overall efficiency of the urban system, and reduce internal damages and external disturbances. That can explain the reasons why the increasing proportion of residential land, public management and public service land, municipal utilities land, and transportation land can restrain the growth of carbon emissions for a city.

4.2. Policy Recommendations

Our quantitative study also has some policy implications on the utilization and allocation of built-up land in YREB.

The land resource is limited, while the urban land resource is extremely scarcer. The pursuit of the additional value of scarce goods, such as lands, will naturally produce a crowding-out effect on the supply of environment-friendly lands, which drives urban land-use structures to better adapt to a high carbon emission model. This brings great challenges to environmentally sensitive areas, especially those watershed-based cities. In reality, the YREB has witnessed large-scale and rapid urbanization. How to balance development and environmental protection in most urbanizing areas of YREB? One pathway is to adjust the supplies of different types of land for production and ecology.

For a city, carbon emission is an inevitable consequence of the consumption of various resources, which provides the energy and products needed for basic or upgrading human activities. As a solution, the concept of sustainable development has been raised. Especially for the cities which are under the conditions of overdevelopment and overpopulated, environment-friendly land should be supplied selectively to provide more public service and reduce the emission pressure of the cities in YREB. A parallel strategy is to encourage the improvement of the utilization efficiency of built-up land, and only in this way can the demand for urbanized land be fundamentally reduced. More importantly, officials and civilians should reach a consensus that a city should not blindly pursue the economic benefits of land. In this context, the land supply strategy with trade-offs will help the urban system to provide more energy and products with less resources.

The structure of urban built-up land can partially determine the performance of the atmospheric environment, which means the land use policies can be improved to better fit our “double carbon” goals in YREB. The role of land resource allocation in carbon emission control should not be exaggerated. A more “suitable” reallocation of urban built-up land cannot work without contributors from other economic, social, and ecological areas. We should also attach importance to green technology such as carbon fixation and reduction technology, clean energy technology, and clean production technology, which benefit carbon dioxide reduction in immediate and decisive ways. From the perspective of the economic belt, regional carbon compensation policy based on land use change covering urban and rural areas along the YREB is also an important policy option.

4.3. Limitations and Future Research Directions

Our study also has several limitations. First, limited by the data sources of carbon emission calculation, some of the cities in the YREB were deliberately excluded. Second, variables that present the progress of the carbon sink and carbon emission reduction technologies were unavailable to be collected. Third, the interactions of subtypes of built-up land were not fully considered. The future research directions can be summed up in two: One is to deeply analyze the possible interactions between any two built-up land subtypes and their synergistic effects on carbon emissions; another is to clarify the impact mechanism of built-up land allocation by considering their spatial (cross-city) effects.

5. Conclusions

This study investigates the environmental outcomes of land resource supplies of 88 cities in the Yangtze River Economic Belt, China. Under the constraints of the STIRPAT model, this study focuses on the urban built-up land allocation and its derivative variables, introduces both static and dynamic models that consider fixed effects and time-lagged terms, to empirically explore the impacts of built-up land as well as its component to the carbon emission. We conclude the findings of this study as follows.

First, the subtypes of built-up land can be basically divided into two groups by considering their functions. The first group contains the land subtypes that are directly related to production and consumption. The land subtypes in the second group own main functions that carry the expansions of supporting facilities and infrastructures such as dwelling, transportation, education, health care, and others. The latter works as the catalyst that makes urban activities efficient and living conditions improved. These two groups of lands show opposite effects on carbon emissions in our experimental model.

Second, for the individual land types or subtypes, the total area of built-up land has a significantly positive impact on the CEI, confirming that converting any type of land into built-up purpose will lead to the growth of carbon emission, more or less. The land subtypes that contribute most to CEI and CEB are industrial land and commercial land, two basic carrying spaces for production and consumption. Industrial land exerts a larger impact on CEI and CEB compared to commercial land, which can be due to the greater emissions of the secondary industry compared to the tertiary industry.

Third, among the lands used for supporting facilities and infrastructures, only green space land and transportation land have significant restraining effects on the CEB in both static and dynamic models. Supplying the green space and square land can encourage planting activities that absorb carbon dioxide, while increasing the transportation land can enhance the connections between producer market and consumer market, improving the overall efficiency of economic system. As for the CEI, except for transportation land, all other land subtypes in the second group show significant and negative impacts on carbon emission averaged by economic output in both models.

Last but not the least, carbon emission is proved to be a timely autocorrelation action. According to time-lagged characteristics of carbon emission and its influential factors including land supplies, we realize that adjusting the carbon emission through reallocating the structure of urban built-up land is also a time-consuming behavior. It is necessary to make a spatial and quantitative urban land use planning in advance, which can deal with the long-term phenomenon by avoiding trial and error as much as possible. Under any circumstance of urban development, it should be noted that optimizing the built-up land allocation is the secondary way to reach the “double carbon” goals compared to technical advancement.