3.1. Study Area
China is a typical emerging-market country that has experienced unprecedented urbanization over the past few decades, thus marked by rapid economic growth and massive rural-to-urban migration. China’s urban landscape has undergone dramatic changes, with cities expanding both vertically and horizontally to accommodate the influx of people and economic activities [
23]. The rapid pace of urbanization in China has raised concerns about its environmental impact, particularly in terms of carbon emissions. Recognizing the challenges posed by rapid urbanization, the Chinese government has emphasized the need for sustainable urban development strategies. Intensive urban land use has emerged as a promising approach to promoting compact, efficient, and environmentally friendly urban growth [
24,
25,
26]. By optimizing land use patterns, reducing sprawl, and improving infrastructure efficiency, intensive urban development aims to mitigate environmental impacts while supporting economic growth and social well-being. The “Land Saving and Intense Usage Rules” published by the Ministry of Land and Resources outlined a fundamental “Five Regulations” policy of slower expansion, inventory optimization, flow efficiency, and quality enhancement [
27]. By increasing the effectiveness of industrial inputs, enabling the treatment of pollutants, or reducing commuting time, intensive urban land use might aid in the construction of low-carbon cities [
16,
25].
3.2. Data Description
Based on the literature review and data availability, this study adopted the input–density–output framework to describe the urban land use intensity. Capital intensity, labor intensity, and R&D investment intensity were selected as the land input levels. Population intensity was selected to define the land density. Economic output per unit land area was regarded as the land output intensity.
The data used in this study include the carbon emissions, urban land area, GDP, population, number of labors, research and development investment, gross investment in fixed assets, and energy consumption of the sample cities. The data sources are described in
Table 1.
Given the heterogeneity of development levels among China’s cities, we divided the cities into three groups based on their development level, which was measured by per capita GDP: this yielded a high development level, middle development level, and low development level. According to data availability, 153 cities were selected as sample cities.
The data descriptive statistics are shown in
Table 2. The wide range of carbon emissions, socioeconomic data, and land use intensity levels among cities, as indicated by the substantial standard deviations, suggests disparities in environmental performance and sustainability practices across urban areas. The data description underscores the importance of categorized exploration with consideration of city heterogeneity and targeted policy interventions tailored to the specific characteristics and challenges faced by individual cities.
Data of carbon emissions and intensive urban land use indicators are presented in spatial distribution in
Figure 2. It is shown that the emissions and land use intensity vary across regions greatly, which indicates a huge development gap among cities. Cities in western China, which are usually regarded as lagging development regions, have high carbon emissions, with high capital intensity and R&D investment intensity. On the contrary, cities in eastern coastal China, which have good economic conditions, present high urban land use intensity and low emissions.
3.3. Model Development
The STIRPAT model, an acronym for “Stochastic Impacts by Regression on Population, Affluence, and Technology”, has gained widespread popularity in assessing environmental impacts and identifying the factors that influence carbon emissions [
29,
30]. This multiplicative model is built upon the IPAT identity, which posits that environmental impact is the product of population, affluence, and technology factors. By utilizing the STIRPAT model, researchers in environmental studies can effectively examine the relationship between population, economic development, and environmental impact factors. One significant advantage of the STIRPAT model, when compared to other approaches such as the IPAT model, is its flexibility in incorporating representative drivers from various perspectives [
31,
32,
33,
34]. For instance, factors such as land use and urbanization can be integrated into the model based on the specific requirements of the study. This allows for a more comprehensive analysis of the relationships between population, economic factors, and environmental impact. As a result of its ability to encompass diverse drivers, the empirical findings derived from the STIRPAT model tend to be more reasonable and credible. Researchers can tailor the model to their specific research context, thus ensuring that relevant factors are included and contributing to a more accurate understanding of the complex dynamics between population, economic development, and environmental outcomes.
This study used the STIRPAT model to analyze the impact of intensive urban land use on carbon emissions in China’s cities. The specification of the STIRPAT model can be expressed as follows:
In the model, the constant term a serves as a scaling factor, while the exponents b, c, and d represent the respective powers of the variables P, A, and T, which are estimated. Additionally, the error term “e” accounts for unexplained variation. The subscript i signifies that these variables (I, P, A, T, and e) vary among different observational units.
An additive regression model in which all variables are in logarithmic form facilitates estimation and hypothesis testing:
In this equation, the coefficients b, c, and d represent the proportional changes in environmental impacts for each 1% change in population size, wealth level, and technology level.
With the representative drivers of intensive land use indicators added into the model in this research, the extended STIRPAT model with the ordinary least squares (OLS) method is denoted as follows:
where α
0 denotes a constant; e denotes the error term; CE, POP, PCGDP, EC, LI, KI, RI, PI, and OI denote the carbon emission, population, per capita GDP, energy use intensity, labor intensity, capital intensity, R&D investment intensity, population density, economic output intensity, respectively; and α
1, α
2, …, and α
8 represent the parameters to be estimated.
The OLS model is a global parameter estimation technique that can only represent the average contribution, because it assumes invariant coefficients [
35]. However, carbon emissions and their determinants have a tendency to be geographically autocorrelated [
36,
37]. The spatial nonstationarity of geographic elements is ignored by the OLS model, which can easily provide biased findings or ineffective estimations [
38]. The geographical weighted regression (GWR) model can describe the spatial characteristics of impacts, while spatial heterogeneity is taken into account by conducting local estimation [
39]. The GWR model can be expressed as follows:
where
is the carbon emissions, and
are the jth independant variables at location i.
and
are the estimated coefficients at location i;
are the coordinates of location i, and
is the random error at location i.