Transition Pathways for Carbon Peaking in Mountainous Megacities: A Multi-Scenario Simulation in Chongqing, China
Lijun
Liang
1✉
Emaillianglijunuse@bistu.edu.cn
Mengze
Ma
1
Email19135729274@163.com
Jianglin
Feng
1
Emailfengjianglin1986@163.com
1
Beijing Information Science & Technology University
No.55 Taihang Road, Changping District
102206
Beijing, Beijing
China
Lijun Liang1*, Mengze Ma2, Jianglin Feng3
1 Beijing Information Science & Technology University, No.55 Taihang Road, Changping District, Beijing, Beijing 102206, China
2 Beijing Information Science & Technology University, No.55 Taihang Road, Changping District, Beijing, Beijing 102206, China
3 Beijing Information Science & Technology University, No.55 Taihang Road, Changping District, Beijing, Beijing 102206, China
*Corresponding author: Lijun Liang, E-mail(s): lianglijunuse@bistu.edu.cn
Other authors' e-mail(s): 19135729274@163.com; fengjianglin1986@163.com
Abstract
This research addresses several challenges, including the lack of comprehensive regional structural variables in carbon peak pathways for mountainous cities, limited policy flexibility in scenario design, and a deficiency of case studies focused on western China, by selecting Chongqing as a representative case. Utilizing the STIRPAT model, the study conducts a thorough assessment of critical regional factors such as industrial structure, energy intensity, and energy mix. To enhance the accuracy of parameter estimation, ridge regression is employed with data sourced from the China Emissions Accounting and Data System (CEADS). Seven multi-scenario combinations are developed to forecast carbon emission trajectories from 2023 to 2050. The results reveal that: (1) population size, industrial structure, and energy structure are the principal determinants of Chongqing’s carbon emissions, with population size exerting the most pronounced marginal impact; (2) under the baseline scenario, carbon emissions are projected to increase, reaching a peak of 208.009 million tons of CO₂ in 2037; conversely, the “Low Growth + Efficient Carbon Reduction” scenario, characterized by accelerated improvements in energy intensity and intensified industrial restructuring, could advance the peak to 2035 at 202.040 million tons, representing a reduction of approximately 2.9% relative to the baseline; and (3) the “Low Growth + Smart-Driven + High-Efficiency Decarbonization” pathway emerges as the most effective strategy for achieving carbon emission peaking and promoting clean production in Chongqing and comparable regions.
Keywords
Carbon emissions peaking
STIRPAT model
Mountainous cities
Multi-scenario analysis
Emission reduction pathways
1. Introduction
Driven by accelerated global industrialization, mountainous cities exemplify regional environmental change hotspots where terrain-mediated spatial patterns amplify carbon-climate feedbacks. Chongqing, as a transitional zone between plateau and plain, embodies such dynamics, Carbon dioxide (CO2) emissions are increasing, posing a growing threat to ecosystems, economies and societies. To address this global challenge, many countries have committed themselves to carbon peaking and carbon neutrality targets, striving to achieve net-zero GHG emissions by mid-century. However, China faces many unique challenges in achieving these goals (Wang et al., 2021; Wang et al., 2024).
While macro-level policies exert considerable influence on the temporal dynamics of regional carbon emissions, these emissions are also intrinsically linked to a complex interplay of localized factors, including population size, the level of economic development, prevailing industrial structure, available energy resources, and unique topographic characteristics. As the only municipality directly under the central government in western China, Chongqing holds a crucial strategic position in China’s coordinated regional development framework. Chongqing is situated at the intersection of the “Yangtze River Economic Belt” and the “Chengdu-Chongqing Twin-city Economic Circle,” and is among the most industrialized and urbanized cities in Southwest China.
While the "Economic Operation of Chongqing Municipality in 2024" report published by the Chongqing Municipal Bureau of Statistics indicates a GDP exceeding 3.2 trillion yuan, it also reveals that Chongqing continues to face substantial challenges related to energy consumption intensity and structural imbalances, particularly given the persistently high proportion of secondary industry within its economic composition. On the other hand, Chongqing’s characteristic “mountain-hill-valley” topography has constrained the city’s spatial development and caused notable regional disparities in carbon emissions. Coupled with a population exceeding 30×10⁶ and ongoing urbanization, this has created intense pressure on energy demand and carbon emission reduction efforts. In this context, Chongqing’s carbon emission issues are both representative and complex, reflecting common challenges faced by resource- and industrial-based cities in carbon reduction and transformation, while also highlighting the unique needs and practical difficulties of mountainous cities in Southwest China in identifying pathways to carbon peaking.
Given the above complexity, accurately identifying the peak carbon emission period in mountainous cities is crucial for formulating scientific and effective emission reduction strategies, optimizing the energy structure, and promoting sustainable socio-economic development in similar regions. However, regionally-focused analyses of carbon peaking pathways for mountainous urban systems under terrain constraints remain sparse, despite their critical role in balancing economic growth and ecological vulnerability.Moreover, from an energy policy perspective, there is a more pronounced scarcity of research on how energy structure transition and efficiency improvement policies, particularly under complex topographic constraints, systematically influence regional carbon peaking pathways.
Existing studies on carbon peak research mainly focus on the national level (Mahapatra et al., 2021; Wojewodzki et al., 2023; Chen et al., 2025) (e.g., comparative analysis of the pathways of countries along the Belt and Road (Zheng et al., 2025)). Meanwhile, different industries and sectors such as iron and steel (Song et al., 2023; Zhang et al., 2024), petrochemicals (Zhang et al., 2023), light industry (Wang et al., 2023), transportation (Zhang et al., 2019; Pang et al., 2022; Sun et al., 2023; Cai et al., 2024), construction (Luo et al., 2025; Qiao et al., 2025), and energy consumption (Liu et al., 2025) have also become research hotspots. In terms of content, existing research mainly covers four directions: first, international practices and experiences of carbon peaking (Han et al., 2022; Wang et al., 2024); second, carbon peaking prediction methods and their evolution (Zhang et al., 2023); third, identification of carbon emission driving mechanisms and factor analysis (Fang et al., 2025); and fourth, peaking pathways and scenario simulations for specific regions or industries (Wu et al., 2023). Regarding research methods (Sahabuddin et al., 2023; Luo et al., 2024; Zhang et al., 2024), the STIRPAT (Stochastic Impacts by Regression on Population, Affluence, and Technology) model, LMDI (Logarithmic Mean Divisia Index) decomposition model, Logistic growth model, LEAP model, gray prediction model, BP neural network model, EKC curve, and Kaya constant equation are mainly used. Among them, STIRPAT has become a classic method for carbon emission driver identification and peak prediction due to its ability to parametrically handle the nonlinear effects of demographic, economic, and technological variables, with advantages of high scalability and variable flexibility (Gao et al., 2022; Liu et al., 2024).
While existing studies cover national/industrial carbon pathways, regionally-focused analyses of terrain-constrained urban systems remain sparse, particularly for inland mountainous cities in Western China. This study addresses this gap by investigating Chongqing—a prototypical mountainous municipality where industrial agglomeration, topographic fragmentation, and carbon-intensive energy structures intersect.
Despite the abundance of research findings, there are several limitations in applied research within specific regions. For example, Although the traditional STIRPAT model is comprehensive, variables reflecting regional structural differences, such as industrial structure, energy intensity, and energy composition, are often insufficiently incorporated into actual modeling, limiting the model's explanatory power regarding the logic of regional carbon emission evolution. Most forecasting studies have not systematically designed scenarios that consider the regional policy context and economic and social development trends, resulting in limited guidance of forecasting results for practical decision-making. Most existing studies focus on the eastern, central, and coastal regions of China, with relatively little research on western and southwestern China, especially on cities characterized by mountainous topography and overlapping industrial bases, making it difficult to provide replicable and implementable identification of peak paths and policy recommendations for cities in the western part of the country.
These issues are particularly pronounced in typical cities with large populations, high energy consumption, and complex topography, such as the “mountainous-valley-industrial” cities represented by Chongqing, where the pressure of industrial transformation overlaps spatial development constraints, and the evolution of carbon emissions is influenced by the combined effects of population density, industrial layout, and energy structure, urgently necessitating the development of adaptive, explanatory, and scenario-based research. There is an urgent need to establish a regional carbon peak prediction system with strong adaptability, sufficient explanatory power, and scientific scenario simulation.
Based on the historical panel data of Chongqing, a typical mountainous city, this paper adopts an extended STIRPAT model: while the model expands upon traditional demographic and economic indicators such as population size, per capita GDP, and urbanization rate, by incorporating the proportion of secondary industry, energy consumption intensity per unit of GDP, and the proportion of coal consumption, it aims to provide a more nuanced analysis of industry-energy coupling. on the other hand, it employs ridge regression to suppress multicollinearity and ensure the statistical robustness of the elasticity coefficients of each variable.
Furthermore, combination scenarios of six variables are designed according to three growth rates—“low, moderate, high”—to systematically measure the evolution of Chongqing’s carbon emissions under multiple scenarios from 2023 to 2050. The optimal carbon peaking path of “low-speed growth + intelligent drive + high-efficiency carbon reduction” is identified to realize three major carbon peak objectives simultaneously: differentiated expansion of the variable system, improved robustness of parameter estimation, and regionalized scenario design and future outlook. The results aim to enhance the methodological framework for carbon emission forecasting in Chongqing and, through this typical regional case, reveal the intrinsic driving forces and optimal path to peak carbon in “mountain-valley-industrial” cities, providing replicable decision support and policy references for similar regions.
The remainder of this paper proceeds as follows. Section 2 details the research methods and models employed in the analysis. Section 3 presents the empirical findings. Section 4 discusses the implications of these results and proposes corresponding policy recommendations. Finally, Section 5 concludes the study and outlines potential avenues for future research.
2. Methodology and model design
To create a regional carbon emission peak prediction model, this study is based on the STIRPAT framework and incorporates essential variables to enhance its architecture. The model's validity is assessed through rigorous data processing, and the evolution of regional carbon emissions is thoroughly analyzed via multivariate scenario simulations. This approach seeks to confirm the model's suitability and precision in a specific region.
The overall research idea is organized into the following six steps: identification of research directions and objectives; data collection and processing; model construction and optimization; model validation and error analysis; multiscenario design and forecasting; and analysis of results and policy implications. As shown in Fig. 1 below.
Fig. 1
Framework diagram of the study design
2.1 Case selection and data sources
As the only municipality directly governed by the central government in China’s western region, Chongqing City combines unique characteristics as a mountainous city, an industrial hub, and a core node of the Yangtze River Economic Belt, making it a distinctive and representative case study in regional carbon peaking research. Its uniqueness is reflected in the following three aspects.
(1) Topographical and spatial heterogeneity. Chongqing's geographical location, situated in the transitional zone between the Qinghai-Tibet plateau and the Yangtze River delta plain, presents a unique set of environmental and geological characteristics that warrant specific attention in research and urban planning, forming a three-dimensional topographical landscape characterized by “two rivers and four mountains,” with over 70% of its area covered by mountainous and hilly terrain. This complex topography results in a cluster-based urban spatial development pattern, limiting energy transmission efficiency. Carbon emissions show significant variation across urban-rural and topographical gradients.
(2) Dual constraints of industrial and energy structures. As a western industrial hub, Chongqing’s secondary industry has long accounted for a higher proportion of GDP than the national average (40.1% in 2023). High-energy-consuming sectors such as steel, chemicals, and machinery manufacturing, which contribute 25% of GDP, are responsible for 55% of carbon emissions, underscoring the urgency of industrial transformation (as shown in Fig. 2). Although the share of coal in the energy consumption structure decreased from 57.7% in 2015 to 50.7% in 2023, it remains above the national average, indicating a significant high-carbon lock-in effect.
(3) Ongoing pressure from population growth and urbanization. By the end of 2023, Chongqing's permanent resident population surpassed 34×10⁶, with an urbanization rate above 70%. Population concentration and urban expansion have intensified energy demand. Meanwhile, as an ecological barrier zone in the upper reaches of the Yangtze River, the conflict between ecological protection requirements and industrial development has become increasingly prominent.
These characteristics make Chongqing an ideal case study for exploring the carbon peaking pathways of “mountain-valley-industrial” composite cities, and its experiences can offer differentiated policy insights for cities in similar regions across Southwest China and the country as a whole.
Fig. 2
Terrain-Clustered Industrial Emissions: GDP Share vs CO₂Contribution (2023)
The primary data sources for this study are as follows: ① Carbon emissions calculation data. The China Carbon Accounting Database (CEADs) (2001–2022) and the China Multi-Scale Emission Inventory Model (MEIC) were utilized, applying the United Nations Intergovernmental Panel on Climate Change (IPCC) emission factor method for calculations; socioeconomic data: The Chongqing Statistical Yearbook (2002–2024) and the Bulletin of the Chongqing Municipal Ecology and Environment Bureau, covering indicators such as population size, per capita GDP, and urbanization rate; ② Industry and energy data. Extracted from the “Annual Report on Energy Consumption in Key Industries” published by the Chongqing Municipal Development and Reform Commission, including secondary industry added value, energy intensity, and coal consumption share; ③ Spatial geographic data. Sourced from the National Basic Geographic Information Center, used for terrain constraint effect analysis.
This paper aims to construct a regional carbon emissions peak prediction model based on the STIRPAT model. By introducing key variables to optimize the model architecture, rigorously processing data to validate the model’s effectiveness, and conducting multi-scenario simulations, the study seeks to deeply analyze the logic behind regional carbon emissions evolution, thereby verifying the model’s adaptability and accuracy in specific regions. The research approach is as follows:
Initially, drawing upon established research concerning carbon emission impact indicators, pivotal factors affecting carbon emissions—namely, population size, the level of economic development, and the state of industrial technology—were translated into quantifiable indicators. This process yielded an optimized STIRPAT model incorporating six measurable indicators: population size, industrial structure, energy structure, urbanization rate, energy intensity, and per capita GDP. Subsequently, leveraging relevant data from Chongqing spanning the period of 2001 to 2018, a ridge regression algorithm was employed to determine the elasticity coefficients for each indicator within the model. The model's validity was then assessed through a comparison of predicted values against actual values observed from 2019 to 2023. Ultimately, utilizing scenario analysis, three distinct rates of change ("low," "medium," and "high") were assigned to each of the six indicators. Based on these varying development scenarios, seven unique scenario combinations were formulated to conduct multi-scenario projections of Chongqing's carbon emissions from 2023 to 2050, thereby exploring pathways toward achieving the stated carbon peak target.
2.2 Selection of research methods
To scientifically predict the trajectory of regional carbon emissions and identify the key driving factors, this study employs the Stochastic Impacts by Regression on Population, Affluence, and Technology (STIRPAT) model as its core analytical framework. The STIRPAT model is derived from the classical IPAT (Impact = Population × Affluence × Technology) identity and aims to quantify the stochastic effects of population, affluence, and technology on environmental pressures(Wei et al., 2024). Unlike the deterministic relationship of IPAT, the core feature of the STIRPAT model is its parameterized form, which allows researchers to estimate the elasticity coefficients of each driving factor through regression analysis. This enables a more accurate description of the nonlinear relationships and relative contributions of socio-economic variables to the environment. The model offers advantages such as a flexible structure, strong variable scalability (allowing inclusion of more variables reflecting regional characteristics), and robust explanatory power for complex socio-economic systems. It has been widely used in analyzing the driving mechanisms of carbon emissions and in medium- and long-term forecasting research(Usman et al., 2022).
Using this model and based on the region’s historical data, the peak of regional carbon emissions is predicted, comprehensively considering the impact of population (P), economic level (A), and technology (T) on environmental pressure (I). While this approach offers a valuable perspective, further research is needed to fully elucidate the complex interplay between social and economic development and environmental sustainability.
The fundamental expression of the STIRPAT model is presented in Eq. (1):
1
While the basic STIRPAT model offers valuable explanatory power, its applicability can be limited when analyzing the complexities of specific regional contexts. To improve the model's fit with actual regional carbon emissions and enhance the precision of characterizing the driving factors, this paper refines the basic STIRPAT model through parametric improvements, incorporating moderator variables to optimize its structure. Prior research (Lian et al., 2025) has demonstrated that industrial structure optimization exerts a significant constraint on carbon emissions.
Consequently, based on the characteristics of carbon emissions predominantly driven by the secondary industry, this study constructs a coupled indicator of industrial structure (IS) and energy intensity (EI) to represent the technological evolution dimension. The energy intensity (EI) is directly linked to the effectiveness of regional energy efficiency standards and energy conservation policies; while the energy mix (ES) reflects the progress of local fossil fuel subsidy reforms and clean energy incentive policies.
Furthermore, the traditional economic development level indicator is disaggregated into two distinct explanatory variables: per capita GDP (A) and urbanization rate (U). The resulting expanded model is presented in Eq. (2):
2
The following is obtained by taking the logarithms of both sides of Eq. (2):
3
In this equation, 'a' represents a constant coefficient, and 'e' denotes a correction coefficient. The terms 'b,' 'c,' 'd,' 'f,' 'g,' and 'h' are elasticity coefficients, quantifying the impact of changes in the independent variables on 'I.' Specifically, they indicate that a one-unit change in P, A, U, IS, EI, and ES will result in a change in 'I' of 'b,' 'c,' 'd,' 'f,' 'g,' and 'h' units, respectively. Refer to Table 1 for a detailed definition of each variable.
|
Variate
|
explanation
|
unit (of measure)
|
|---|---|---|
|
Carbon dioxide emissions (I)
|
Carbon dioxide emissions
|
10⁶ t of carbon dioxide
|
|
Population size (P)
|
End-of-year total population
|
10⁶ people
|
|
GDP per capita (A)
|
GDP per capita
|
104 (RMB)
|
|
Urbanization rate (U)
|
Urban population percentage
|
%
|
|
Industrial structure (IS)
|
Secondary sector GDP share
|
%
|
|
Energy intensity (EI)
|
Energy consumption per unit of GDP (equivalent to standard coal)
|
104 t / 108 yuan (RMB)
|
|
Energy mix (ES)
|
Share of coal in total energy consumption
|
%
|
2.3 Model validation
2.3.1 Verification and optimization of model variables
This paper undertakes an empirical analysis of Chongqing Municipality spanning from 2001 to 2018. Initially, correlation testing is performed on the variables under consideration. Subsequently, a linear regression analysis is implemented, with a specific focus on identifying and mitigating potential multicollinearity to enhance the model's stability and reliability. Following this, the processed data are incorporated into the model to generate predicted carbon emissions for each year, which are then compared with historical data to validate the model's predictive accuracy. Finally, historical data from Chongqing are integrated and analyzed using Eq. (3) to diagnose potential multicollinearity.
Table 2 presents the results of the Pearson correlation analysis, which indicates a statistically significant linear relationship between all explanatory variables included in the model.
|
Variate
|
lnP
|
lnA
|
lnU
|
lnIS
|
lnEI
|
lnES
|
|---|---|---|---|---|---|---|
|
lnP
lnA
lnU
lnIS
lnEI
lnES
|
1
|
0.927**
|
0.896**
|
−0.425*
|
−0.946**
|
−0.768**
|
|
0.927**
|
1
|
0.994**
|
−0.527**
|
−0.997**
|
−0.821**
|
|
|
0.896**
|
0.994**
|
1
|
−0.599**
|
−0.987**
|
−0.830**
|
|
|
−0.425*
|
−0.527**
|
−0.599**
|
1
|
0.676**
|
0.701**
|
|
|
−0.946**
|
−0.997**
|
−0.987**
|
0.676**
|
1
|
0.830**
|
|
|
−0.768**
|
−0.821**
|
−0.830**
|
0.701**
|
0.830**
|
1
|
**. At the 0.01 level (two-tailed),the correlation is significant.
*. At the 0.05 level (two-tailed), the correlation is significant.
This study utilizes a least squares regression model and incorporates a multicollinearity diagnostic analysis to assess the relationships among the independent variables, as presented in Table 3. The variance inflation factor (VIF) results indicate that the majority of explanatory variables exhibit VIF values exceeding the conventional threshold of 10, suggesting the presence of significant multicollinearity within the model.
|
Variate
|
Unstandardized coefficient
|
T-Statistic value
|
Significance
|
Covariance statistics
|
|||
|
B
|
Standard error
|
Beta
|
Tolerances
|
VIF
|
|||
|
lna
lnP
lnA
lnU
lnIS
lnEI
lnES
|
93.821
|
17.597
|
5.332
|
< 0.001
|
|||
|
−9.113
|
1.908
|
−1.004
|
−4.775
|
< 0.001
|
0.005
|
21.089
|
|
|
1.662
|
0.514
|
3.296
|
3.233
|
0.009
|
0.001
|
495.528
|
|
|
−3.825
|
1.463
|
−1.719
|
−2.615
|
0.450
|
0.014
|
206.101
|
|
|
−0.884
|
1.017
|
−0.070
|
−0.869
|
0.750
|
0.113
|
3.126
|
|
|
−0.340
|
1.036
|
−0.337
|
−0.328
|
0.431
|
0.001
|
502.655
|
|
|
0.166
|
0.467
|
0.033
|
0.356
|
0.353
|
0.097
|
4.137
|
|
To make the fitting results more realistic and address the multicollinearity problem among the variables, ridge regression analysis was performed on all variables in the model. The ridge regression results are shown in Fig. 3, according to the ridge regression results, the penalty term of the ridge parameter (i.e., k value) was selected as 0.7.
Fig. 3
Ridge regression trace
To mitigate the potential detrimental effects of multicollinearity on prediction accuracy, this study employs ridge regression estimation to model the data. By incorporating a ridge parameter penalty term, we effectively implement partial least squares estimation. The resulting parameter estimates following model regularization are presented in Table 4.
|
Variatet
|
Coefficient (B)
|
Standard error (SE)
|
Standardized coefficient (Beta)
|
T-Statistic value (T)
|
Significance (Sig)
|
|---|---|---|---|---|---|
|
lnP
|
2.525 785 78
|
0.274 884 85
|
0.241 314 29
|
9.188 523 15
|
0.000 000 15
|
|
lnA
|
0.079 895 82
|
0.008 343 32
|
0.190 969 91
|
9.576 025 87
|
0.000 000 09
|
|
lnU
|
0.207 557 23
|
0.030 227 23
|
0.172 237 88
|
6.866 564 45
|
0.000 005 35
|
|
lnIS
|
0.615 728 27
|
0.316 361 20
|
0.098 873 81
|
1.946 282 55
|
0.040 600 99
|
|
lnEI
|
−0.125 949 69
|
0.014 585 86
|
−0.162 062 29
|
−8.635 053 31
|
0.000 000 33
|
|
lnES
|
−0.280 568 81
|
0.130 841 95
|
−0.079 895 87
|
−2.144 333 80
|
0.048 795 87
|
|
constant
|
−17.713 679 69
|
3.175 003 76
|
0.000 000 00
|
−5.579 105 11
|
0.000 052 66
|
| * R2 = 0.885 149 37 F value = 6.206 908 14 Sig F = 0.006 061 86 | |||||
| The ridge regression analysis, as presented in Table 4, demonstrates a strong model fit, with the coefficient of determination (R²) explaining approximately 88.5% of the variance in carbon emissions in Chongqing. The statistically significant F-statistic (p ≈ 0.006, α = 0.05) leads to the rejection of the null hypothesis, confirming the overall significance of the model and indicating that at least one independent variable exerts a significant influence on carbon emissions. This outcome substantiates the capacity of ridge regression to maintain high explanatory power and robustness, even after mitigating multicollinearity. Furthermore, the statistical significance of all variable coefficients (p < 0.05) reinforces the rationality of variable selection and the validity of parameter estimation, thereby providing a reliable mathematical basis for subsequent multi-scenario predictions. Substituting the fitted coefficients from Table 4 into Eq. (3) yields the ridge regression equation, presented as Eq. (4): | |||||
4
The preceding analysis culminates in the development of a predictive model for Chongqing's annual CO2 emissions, as represented by Eq. (5).
5
2.3.2 Comparison of predicted results
Table 4 demonstrates the significant impact of all variables within the model on Chongqing's carbon emissions, as evidenced by the data fitting results. However, the magnitude of influence varies across these variables, with the following descending order of importance: P (0.956) > IS (0.568) > ES (0.482) > U (0.349) > EI (0.155) > A (0.081). Subsequently, to validate the predictive capability of the model, carbon emission data for Chongqing from 2019 to 2023 were input into Eq. (5). Table 5 presents a comparative analysis of the actual and predicted carbon emission values observed over the five-year period under investigation.
|
Vintages
|
Measured value/× 10⁶ t CO2
|
Model predictions/× 10⁶ t of CO2
|
Mean absolute error(MAE)/%
|
|---|---|---|---|
|
2019
|
156.254681
|
162.0613593
|
3.7161627
|
|
2020
|
152.8564364
|
159.0917716
|
4.0792101
|
|
2021
|
165.278991
|
172.2951832
|
4.2450599
|
|
2022
|
171.673362
|
178.0483988
|
3.7134688
|
|
2023
|
169.957426
|
177.4210672
|
4.3914773
|
The discrepancy between measured CO₂ emissions and model predictions for Chongqing over the past five years has consistently remained low, with a mean absolute error (MAE) of only 4.03%. Consistent with established research practices in carbon emissions modeling (Li et al., 2023), an MAE below 5% is generally considered indicative of high predictive accuracy and model stability. Furthermore, the relatively small annual error fluctuations (3.71%–4.39%) fall within the typical error range (3%–8%) reported in existing literature for optimized STIRPAT models (Tian et al., 2025), suggesting that the fitted predictive model exhibits robust performance and good applicability.
2.3.3 Comparison of model optimization methods
To verify the superiority of ridge regression in handling multicollinearity between variables, this study simultaneously employs LASSO regression and elastic net regression for comparative analysis. Models were trained using the same historical data as the ridge regression method (2001 to 2018) and compared using the same predicted values (2019 to 2023). The comparison results are shown in Table 6.
|
Regression Model
|
RMSE(model training)
|
R²(model training)
|
RMSE(model prediction)
|
MAE(model prediction)
|
|---|---|---|---|---|
|
Ridge regression
|
0.12
|
0.88
|
4.31
|
4.03%
|
|
LASSO regression
|
0.18
|
0.85
|
5.89
|
5.42%
|
|
Elastic network regression
|
0.15
|
0.79
|
4.98
|
4.61%
|
Ridge regression had the highest goodness of fit in the training set (R² = 0.88), indicating it had the strongest explanatory power for historical data. In the test set, ridge regression showed the lowest prediction error (MAE = 4.03%), which was significantly better than LASSO (5.42%) and elastic net (4.61%). LASSO regression over-compresses variable coefficients, resulting in the loss of key driver information. Although elastic net accounts for variable selection and stability, its mixed parameters require repeated tuning, and its prediction accuracy remains lower than that of ridge regression.
The core rationale for choosing ridge regression is as follows. The primary goal of this study is to identify the elasticity coefficients of all driving factors (especially policy-sensitive variables such as industrial structure and energy structure) and to determine the driving factors of energy consumption. LASSO’s variable screening property loses key policy information; there is a high degree of collinearity among variables in the Chongqing case (VIF > 10), and the L2 regularization of ridge regression maximally retains coefficient stability, which is insufficient for predicting policy variables. The subjective interference of mixed parameters in elastic net is avoided; the k value of ridge regression is determined by the ridge trace plot, and its methodological transparency is higher than that of the parameter tuning process in elastic net.
In summary, ridge regression is the best method in this study due to its comprehensive advantages in coefficient stability, policy explanatory power, and prediction accuracy.
3. Results analysis under ultidimensional scenario combination
This paper aims to use carbon emission data from Chongqing and an improved STIRPAT model, combined with scenario analysis, to develop multiple scenario combinations. By setting different rates of change for key factors influencing carbon emissions, this approach will offer a more comprehensive prediction of future carbon emission trends, as well as the timing and peak levels of emissions under various development pathways for Chongqing.
While the STIRPAT model has been extensively applied to identify drivers of carbon emissions, existing studies predominantly focus on analyzing the impact of changes in single parameters. This approach, while capable of identifying individual drivers, often lacks a systematic simulation of the complex interplay between multi-dimensional factors, such as combinations of policy variables, developmental trajectories, and technological advancements. Furthermore, current research on carbon peak pathways commonly suffers from limitations including a single-dimensional approach to scenario construction, overly idealized parameter settings, and insufficient consideration of regional heterogeneity and policy response feedback mechanisms. To address these limitations, this study proposes a multi-dimensional rate combination scheme based on six core variables, constructing seven representative scenario paths. This framework aims to systematically describe the evolution trajectory and peaking trends of carbon emissions under different policy and development trade-offs. Subsequently, the study analyzes and compares the carbon emission characteristics, peak year, and peak difference across each typical scenario.
3.1 Scenario combination
This paper explores Chongqing's future economic development, population growth trends, and policy planning implications by employing six variables. For each variable, three distinct change rates—"Low," "Medium," and "High"—were established. These rates were informed by historical carbon emission data for Chongqing, the average annual change rate of relevant indicators, and alignment with Chongqing's planning outlines. Specifically, the baseline values for each factor during the "14th Five-Year Plan" period were determined, and subsequently, change rates under different scenarios were designed based on these baseline values, as detailed in Table 7.
|
Rate of change
|
Year span
|
P
|
A
|
U
|
IS
|
EI
|
ES
|
|---|---|---|---|---|---|---|---|
|
Low
|
2021ཞ2025
|
0.15
|
8.00
|
3.00
|
−1.30
|
−5.50
|
−0.90
|
|
2026ཞ2030
|
0.15
|
7.50
|
2.50
|
−1.20
|
−5.40
|
−0.80
|
|
|
2031ཞ2035
|
0.10
|
7.00
|
2.00
|
−1.10
|
−5.30
|
−0.70
|
|
|
2036ཞ2040
|
0.05
|
6.50
|
1.50
|
−1.00
|
−5.20
|
−0.60
|
|
|
2041ཞ2045
|
0.05
|
6.00
|
1.00
|
−0.90
|
−5.10
|
−0.50
|
|
|
2046ཞ2050
|
0.05
|
5.50
|
0.50
|
−0.80
|
−5.00
|
−0.40
|
|
|
Medium
|
2021ཞ2025
|
0.20
|
9.00
|
3.50
|
−1.50
|
−6.00
|
−1.10
|
|
2026ཞ2030
|
0.20
|
8.50
|
3.00
|
−1.40
|
−5.90
|
−1.00
|
|
|
2031ཞ2035
|
0.15
|
8.00
|
2.50
|
−1.30
|
−5.80
|
−0.90
|
|
|
2036ཞ2040
|
0.10
|
7.50
|
2.00
|
−1.20
|
−5.70
|
−0.80
|
|
|
2041ཞ2045
|
0.10
|
7.00
|
1.50
|
−1.10
|
−5.60
|
−0.70
|
|
|
2046ཞ2050
|
0.05
|
6.50
|
1.00
|
−1.00
|
−5.50
|
−0.60
|
|
|
High
|
2021ཞ2025
|
0.25
|
10.00
|
4.00
|
−1.60
|
−6.50
|
−1.30
|
|
2026ཞ2030
|
0.25
|
9.50
|
3.50
|
−1.50
|
−6.40
|
−1.20
|
|
|
2031ཞ2035
|
0.20
|
9.00
|
3.00
|
−1.40
|
−6.30
|
−1.10
|
|
|
2036ཞ2040
|
0.15
|
8.50
|
2.50
|
−1.30
|
−6.20
|
−1.00
|
|
|
2041ཞ2045
|
0.15
|
8.00
|
2.00
|
−1.20
|
−6.10
|
−0.90
|
|
|
2046ཞ2050
|
0.10
|
7.50
|
1.50
|
−1.10
|
−6.00
|
−0.80
|
Considering the data presented in the preceding table, it is pertinent to acknowledge the demographic and economic context of Chongqing over the past decade. During this period, the city's permanent resident population exhibited an average annual growth rate of 0.21%. Concurrently, the average annual change rates for key economic indicators were as follows: per capita GDP increased by 9.08%, urbanization advanced by 3.66%, the secondary industry structure contracted by approximately 1.48%, energy intensity decreased by 6.14%, and the energy structure improved by 1.17%. These figures provide a broader perspective on the socioeconomic and energy-related trends within Chongqing during the period under consideration.
Based on these approximate values, a benchmark change rate (i.e., the medium-speed rate for the period from 2021 to 2025) is set. On the basis of the medium-speed rate, two additional rates—“low-speed” and “high-speed”—are established according to differences in planning and development to account for changes in influencing factors. When the change rates of all factors are at the “medium-speed” level, it means that regional development maintains its existing momentum. When the “population size,” “per capita GDP,” and “urbanization rate” are at the high-speed change rate, it indicates that the region prioritizes economic development and urbanization as its primary objectives. When “industrial structure,” “energy intensity,” and “energy structure” exhibit high-speed changes, it indicates that the region prioritizes low-carbon, green, and environmentally friendly development as its primary objectives. Conversely, when all factors exhibit low-speed changes, the opposite applies.
Based on the three rates of change for each influencing factor, this paper designs seven different combination scenarios. The aim is to predict Chongqing's CO2 emissions from 2023 to 2050, considering the region's historical context, and to comprehensively explore the optimal path to achieve the regional carbon peak target. Specific scenario combinations are shown in Table 8.
|
Scenario type
|
P
|
A
|
U
|
IS
|
EI
|
ES
|
|---|---|---|---|---|---|---|
|
1
|
medium
|
medium
|
medium
|
medium
|
medium
|
medium
|
|
2
|
medium
|
medium
|
medium
|
high
|
high
|
high
|
|
3
|
medium
|
low
|
high
|
high
|
medium
|
high
|
|
4
|
high
|
high
|
high
|
low
|
low
|
low
|
|
5
|
high
|
high
|
high
|
high
|
high
|
high
|
|
6
|
low
|
low
|
low
|
low
|
low
|
low
|
|
7
|
low
|
low
|
low
|
high
|
high
|
high
|
The scenario combination design in Table 7 reflects the differentiated change paths of key drivers under various development strategies. Scenario 1, for example, represents a baseline path that continues the current momentum of development, while Scenario 7 depicts a scenario that tightly controls population, economic, and urbanization growth. At the same time, it emphasizes vigorously promoting the adjustment of industrial structure, improving energy efficiency, and pursuing a deep carbon reduction path through clean energy transformation.
3.2 Prediction results
The trend of carbon emissions in Chongqing from 2023 to 2025 is shown in Fig. 4, while the predicted CO2 emissions under seven combined scenarios are presented in Table 9.
Fig. 4
Carbon emissions under different energy policy scenarios (2023–2050)
|
Scenario
|
1
|
2
|
3
|
4
|
5
|
6
|
7
|
|---|---|---|---|---|---|---|---|
|
Projected carbon peak years
|
2037
|
2036
|
2039
|
2043
|
2037
|
2038
|
2035
|
|
Peak/(×10⁶ t CO2)
|
208.009
|
210.086
|
224.041
|
228.524
|
219.150
|
205.223
|
202.040
|
As shown in Table 8, under Scenario 7 (“Low-Speed Growth + Efficient Carbon Reduction”), carbon emissions peak the earliest, with the peak year occurring in 2035 and the peak value reaching 202.040×10⁶ t CO₂. In contrast, under Scenario 4 (“High-Speed Development”), carbon emissions are projected to peak the latest and reach the highest peak value, with the peak year occurring in 2043 and a peak of 228.524×10⁶ t CO₂.
4. Discussion and policy recommendations
In the context of the "Dual Carbon" target, it is crucial for inland mountainous cities to scientifically identify the driving mechanisms of regional carbon emissions, accurately predict the peaking trajectory, and formulate operable emission reduction policies. This study takes Chongqing as a case, constructing a multi-scenario simulation framework by introducing the extended STIRPAT model and ridge regression method to predict and analyze the trend of carbon emissions under different development paths. On this basis, the paper discusses the model results and regional characteristics, summarizes the practical challenges faced, and proposes targeted policy recommendations for mountain cities, forming a closed-loop logical structure from "Cognition-Analysis-Suggestion".
4.1 Comprehensive discussion on peak carbon research in China's inland mountain cities
This study, employing an extended STIRPAT model and multi-scenario simulations, investigates the driving forces and peaking pathways of carbon emissions in Chongqing, a representative mountainous city. The findings not only offer immediate policy implications for Chongqing but also provide valuable theoretical insights and practical guidance for other inland mountainous cities in China striving to achieve carbon peaking. This section delves into a critical discussion of the following core issues, beginning with the centrality of regional structural variables in carbon emission reduction for mountainous cities.
4.1.1 The central role of regional structure in carbon emission reduction for
mountainous cities
The results suggest that, compared to macro-level population and industrial planning, direct energy policy instruments (e.g., carbon pricing targeting energy intensity, coal consumption control targeting energy structure) may exhibit higher policy elasticity and more significant emission reduction effects in mountainous cities. This finding underscores the critical constraints imposed by regional structural factors, such as industrial layout and energy composition, on achieving carbon peaking in mountainous cities. While the conventional STIRPAT model often prioritizes population size, affluence, and technological advancement, this study expands upon this framework by integrating regional characteristic variables. Specifically, we incorporate industrial structure, energy intensity, and energy structure to investigate the influence of these localized factors on economic growth patterns. This enhancement significantly improves the model's explanatory power regarding the evolutionary logic of carbon emissions in Chongqing. Consequently, this validates the necessity of thoroughly characterizing the transformation of local industrial structure, the improvement of energy efficiency, and the process of energy decarbonization when studying regional carbon peaking.
For Chongqing and similar inland mountainous cities, regulating population growth (particularly optimizing the spatial distribution of population), accelerating the transformation of the industrial structure towards low-carbon and high-value-added sectors (e.g., reducing the proportion of energy-intensive industries), and deeply optimizing the energy structure (e.g., continuously reducing the proportion of coal consumption) represent the most direct and effective policy levers for achieving carbon peaking. Chongqing's 2022 data reveal that energy-intensive industries, accounting for 25% of GDP, contributed 55% of carbon emissions, and the proportion of coal consumption remains higher than the national average. This further corroborates the substantial emission reduction potential attainable through optimizing industrial and energy structures.
4.1.2 Multi-scenario simulations as a basis for differentiated policy formulation
This study employed seven scenarios to systematically simulate the evolution and peaking potential of carbon emissions in Chongqing under varying development pathways. The findings indicate that Chongqing's carbon emissions could peak as early as 2035 (at 202.040×106 t) under the "Low growth + high efficiency carbon reduction" scenario (Scenario 7). This represents a two-year advancement compared to the 2037 peak (208.009×106 t) projected in Scenario 1, with a corresponding peak reduction of approximately 2.9%. Conversely, the "High growth" scenario (Scenario 4) anticipates the highest peak (228.524×106 t) and a delayed peak year of 2043. This substantial divergence underscores the decision-making value of multi-scenario analysis for regional carbon peak planning. By quantifying the impact of various policy combinations – including economic growth rates, industrial restructuring efforts, energy efficiency improvements, and the pace of energy decarbonization – on peak timing and magnitude, it provides policymakers with a clear "menu of policy options."
Notably, the "Slow growth + Intelligent Drive + efficient carbon reduction" pathway emerges as the optimal strategy for achieving Chongqing's carbon peaking target. This provides a scientific foundation for balancing economic development with emission reduction goals in Chongqing and similar regions. The scenario analysis results emphasize that for "mountain-valley-industry" complex cities like Chongqing, controlling population growth and urbanization rates, while simultaneously accelerating industrial structure optimization and reducing energy intensity, constitutes the key pathway to achieving regional carbon peaking.
4.1.3 Unique Challenges and Opportunities Facing Carbon Peaking in Mountainous Cities of Western China
This study, focusing on Chongqing, addresses a gap in the existing literature on carbon peaking in western China, particularly concerning southwestern mountain cities. As a representative "mountain-valley-industry" complex city, Chongqing faces distinct challenges in reaching peak carbon: (1) Its predominantly mountainous and hilly terrain (over 70%) fosters a dispersed urban spatial development pattern, which impedes efficient energy transmission; (2) As a major industrial center in western China, Chongqing exhibits a high proportion of secondary industries, resulting in high energy consumption and carbon emission intensity. This is further compounded by the prevalence of energy-intensive industries (e.g., iron and steel, chemical manufacturing, machinery manufacturing) and a coal-dominated energy structure, creating a significant high-carbon lock-in effect; (3) Continuous population growth and urbanization (with over 34 million permanent residents and a urbanization rate exceeding 70%) exert persistent pressure on energy demand.
Furthermore, its role as an ecological barrier in the upper reaches of the Yangtze River intensifies the conflict between ecological preservation and industrial development. These challenges place significantly higher demands on Chongqing's carbon emission reduction efforts compared to those faced by cities on plains. However, this study also reveals opportunities: mountain cities can overcome topographical constraints and achieve earlier and lower carbon emissions peaks by precisely identifying core drivers and implementing tailored policy combinations (e.g., Scenario 7) to enhance urban planning. Chongqing's experience offers valuable insights for the broader southwest region, characterized by complex terrain, heavy industrial structures, and carbon-intensive energy systems, as well as for other mountain resource-based cities.
4.2 Policy recommendations for similar cities to achieve regional carbon peak
Drawing upon the detailed analysis of Chongqing's experience, this study aims to offer insights applicable to inland cities sharing similar characteristics, namely mountainous terrain, a substantial industrial base, and a phase of rapid development. To support the scientific and efficient attainment of regional carbon peaking targets, the following hierarchical and operational policy recommendations are proposed.
4.2.1 Energy Structure Transition and Infrastructure Planning
Prioritize energy structure transition through fiscal incentives and market-based tools, such as carbon pricing and subsidies for clean energy, to overcome terrain-induced energy transmission costs. While population distribution optimization remains relevant, energy policy instruments (e.g., coal-to-clean transition subsidies) should be the primary lever to reduce carbon lock-in effects in valley industrial clusters (Zhou et al., 2022).
Formulating and rigorously enforcing a roadmap for the transformation and eventual phasing-out of energy-intensive, high-emission industries through carbon pricing mechanisms and fossil fuel subsidy reforms is essential. Concurrently, vigorous development of strategic emerging industries, high-end manufacturing, and modern services is needed to substantially decrease the proportion of high-carbon industries within the secondary economic sector. Furthermore, an evaluation system for industrial low-carbon development should be established, integrating carbon emission intensity into industrial access and performance assessment standards. As the Chongqing case demonstrates, accelerating the reduction of the secondary sector's economic share is paramount for achieving an early carbon emissions peak.
Achieving coal consumption reduction targets exceeding national averages necessitates a multi-faceted approach. This requires a substantial increase in the proportion of clean energy sources, including natural gas, renewable energy, and imported hydropower, within the primary energy consumption mix. Furthermore, bolstering local clean energy infrastructure, such as distributed photovoltaic systems, energy storage solutions, and smart grids, is crucial to mitigating energy transmission limitations imposed by mountainous terrain. Prioritizing the establishment of ventilation corridors along river valleys (e.g., the Jialing River Basin) to dissipate accumulated heat, in conjunction with the deployment of distributed solar farms on south-facing slopes to minimize transmission losses (Zheng et al., 2025), represents a potentially effective strategy.
4.2.2 Market-Based Energy Policy Instruments and Dynamic Governance
To ensure a responsive and adaptive approach to carbon mitigation, a closed-loop system integrating "Monitoring-Forecasting-Early Warning-Response" is crucial(Gao et al., 2022). Drawing upon the multi-scenario simulation framework developed in this study, a normalized urban carbon emission monitoring, forecasting, and early warning system should be established. Scenario parameters should be updated and re-simulated periodically (e.g., every five years or following major policy adjustments) to assess the divergence between the current policy trajectory and the desired optimal peaking path (e.g., Scenario 7 in this study).
A dynamic policy toolbox should integrate real-time carbon market signals, where carbon price fluctuations directly inform subsidy adjustments for renewable energy. For instance, when carbon prices exceed a threshold (e.g., 100 RMB/t CO₂), subsidies for solar PV can be automatically increased to accelerate coal replacement. For instance, if forecasts indicate a potential delay in achieving peak emissions targets, policy interventions focused on adjusting industrial structure and reducing energy intensity should be prioritized.
Furthermore, rigorous target decomposition and assessment are essential. Regional carbon peaking targets (time and peak value) should be scientifically allocated to districts and counties, key industries, and major enterprises, accompanied by the establishment of robust assessment and accountability mechanisms. The achievement of these peak targets should be integrated into the performance appraisal system for local governments and leading officials(Ding et al., 2023).
4.2.3 Terrain-Adaptive Energy Spatial Planning and Ecosystem Synergy
Strategic utilization of ventilation corridors, particularly those aligned with river valleys such as the Jialing River, is crucial for dissipating accumulated heat and concurrently guiding low-carbon development initiatives by optimizing grid connectivity for distributed solar/wind farms in south-facing slopes. Spatial planning should be strategically leveraged to steer such initiatives. Specifically, urban land spatial planning and industrial layout must meticulously account for topographical features and ecological sensitivities, directing industrial activities and population concentrations towards areas optimally suited for low-carbon practices. This integrated approach seeks to establish an intensive, efficient, and low-carbon spatial development pattern that harmonizes with the region's mountainous environment, safeguards ecological spaces, and enhances overall carbon sequestration capacity.
Developing clean energy resources tailored to mountain characteristics is also critical. Exploiting the abundant solar, wind, hydro (small hydropower), geothermal, and biomass energy resources prevalent in mountain regions, distributed renewable energy projects should be developed according to local conditions. This will increase local clean energy supply capacity and reduce dependence on fossil fuels and long-distance transmission, minimizing associated losses.
Improvements in mountain transportation and building energy efficiency are also necessary(Chen et al., 2025). In line with the unique transportation and building energy consumption patterns of mountain cities, vigorous development of public transport (particularly rail transport), promotion of new energy vehicles, and optimization of logistics systems are vital to reduce transport-related carbon emissions. Strict adherence to green building standards and the adoption of energy-saving building technologies and materials specifically designed for mountain climates should also be prioritized.
4.2.4 Digital Innovation for Energy-Carbon Synergistic Governance
Blockchain-based carbon data platforms should be embedded in real-time energy dispatch systems, using topographic and thermal data to prioritize clean energy allocation in valley industrial zones. An advanced monitoring system should be developed that integrates a blockchain-based carbon data storage platform with IoT monitoring devices installed in major emission-producing companies (Guo et al., 2025). This setup will allow for carbon emissions to be collected in milliseconds and analyzed using AI, reaching a high accuracy of 92% in detecting abnormal emission activities.
Intelligent trading within the carbon market can be facilitated by developing an AI-driven carbon quota forecasting system. By integrating enterprise production plans and industry trends, this system can generate optimal trading strategies. A pilot carbon futures price forecasting model should also be implemented to mitigate the risk of market volatility.
A regional collaborative platform is essential for fostering cooperation (Li et al., 2024). A cloud platform of AI collaborative emission reduction in Chengdu-Chongqing Twin Cities should be built to enable cross-regional energy scheduling and industrial complementarity. For example, when solar power generation is abundant in Chengdu, the platform can automatically direct data centers in Chongqing to increase their computing load; conversely, when Chongqing has surplus hydropower, it can transmit clean power to chemical companies in Sichuan.
Public participation innovation can be enhanced through the launch of AI-powered personal carbon accounts. These accounts, accessible through a mobile app, will display the carbon footprint of individuals' travel and consumption activities in real-time and generate personalized emissions reduction plans. By integrating with merchant systems for carbon credit redemption, this initiative is projected to reach 12 million citizens.
5. Conclusion and outlook
This study demonstrates that energy policy customization—prioritizing energy intensity reduction and coal-to-clean transition—is key to achieving carbon peaking in mountainous cities by 2035 under Scenario 7.
Specifically, it focuses on projecting carbon emissions and delineating carbon peaking pathways for inland mountainous cities, employing Chongqing as a case study to construct a scenario simulation and strategic analysis framework. Leveraging an extended STIRPAT model, this research incorporates regional structural variables and enhances model stability through ridge regression. Subsequently, a multi-scenario approach is developed to forecast future carbon emission trends. Following the identification of an optimal peaking pathway, policy recommendations with practical applicability, tailored to the regional characteristic system, are proposed. The paper concludes by summarizing the key findings and outlining promising avenues for future research.
(1)
(1) This research identifies energy structure (ES) and energy intensity (EI) as the most policy-responsive factors for achieving carbon peaking in mountainous megacities, as evidenced by their high standardized coefficients and the substantial emission reductions observed in Scenario 7. Although population size exhibits the greatest marginal impact, energy policies focused on reducing coal consumption and enhancing efficiency provide more direct and immediate mechanisms for governance intervention.
(2)
Scenario 7 illustrates that an assertive energy policy trajectory—emphasizing rapid transitions in energy structure and reductions in energy intensity—can effectively decouple carbon emissions from economic growth, attaining the earliest carbon peak in 2035 at the lowest emission level of 202.04 million tons. This outcome challenges the traditional notion of a trade-off between economic development and decarbonization, underscoring the pivotal role of proactive energy governance.
(3)
The feasibility of a carbon peaking pathway fundamentally depends on the transformation of the energy system. This transformation primarily involves the enforcement of energy efficiency standards and the provision of subsidies for clean energy infrastructure to lower energy intensity, complemented by industrial policies aimed at optimizing structural composition. Such energy-focused interventions produce more substantial and reliable emission reductions compared to indirect demographic management strategies.
While this study establishes a framework for carbon emission prediction and policy pathway analysis in inland mountainous cities, it possesses limitations regarding the dynamic simulation of response mechanisms. Specifically, the framework does not fully account for the temporal lags and interdependencies between policy interventions, technological advancements, and systemic feedback loops. Furthermore, the suitability and marginal effects of artificial intelligence (AI), big data, and related technologies in regional governance remain largely informed by pilot studies, necessitating more longitudinal empirical tracking and cross-sectoral assessments.
References
1.
Wang, Y. et al. Carbon peak and carbon neutrality in China: Goals, implementation path and prospects. China Geol. 4 (4), 720–746 (2021).
2.
Wang, S., Li, J. & Lu, Q. Optimization of carbon peaking achieving paths in China's transportation sector under digital feature clustering. Energy 313, 133887 (2024).
3.
Wojewodzki, M., Wei, Y., Cheong, T. S. & Shi, X. Urbanisation, agriculture and convergence of carbon emissions nexus: Global distribution dynamics analysis. J. Clean. Prod. 385, 135697 (2023).
4.
Mahapatra, B. & Irfan, M. Asymmetric impacts of energy efficiency on carbon emissions: A comparative analysis between developed and developing economies. Energy 227, 120485 (2021).
5.
Chen, R. et al. Integrating local climate zones and spatial modeling for carbon emission forecasting in the Guangdong-Hong Kong-Macao Greater Bay Area toward 2060. Urban Clim. 61, 102453 (2025).
6.
Zheng, Y. et al. Spatiotemporal assessment of land use carbon emission in the Guangdong-Hong Kong-Macao Greater Bay Area: A hybrid static-dynamic model integrated with PLUS. Urban Clim. 63, 102574 (2025).
7.
Zhang, P. L., Du, Q. J., Zhang, K. X. & Tian, W. T. Carbon Emission for China's Iron and Steel Industry: Peak Scenarios and Neutralization Pathways. Environ. Sci. 45 (11), 6336–6343 (2024).
8.
Song, X. C. et al. Life cycle carbon emission accounting and emission reduction potential assessment of steel industry. Environ. Sci. 44 (12), 6630–6642 (2023).
A
9.Zhang, X. C. et al. Volatile organic compound emission characteristics and influences assessment of a petrochemical industrial park in the pearl river delta region. Environ. Sci. 43 (4), 1766–1776 (2022).
10.
Wang, C. et al. Embodied carbon emissions generated by international trade of China's light industry sector based on global supply chains perspective. Energy Strategy Reviews. 47, 101095 (2023).
A
11.Zhang, L., Long, R., Chen, H. & Geng, J. A review of China's road traffic carbon emissions. J. Clean. Prod. 207, 569–581 (2019).
A
12.Pang, K. et al. Forecasting of emission co-reduction of greenhouse gases and pollutants for the road transport sector in Lanzhou based on the LEAP model. Environ. Sci. 43 (7), 3386–3395 (2022).
A
13.Sun, Y., Yang, Y., Liu, S. & Li, Q. Research on transportation carbon emission peak prediction and judgment system in China. Sustainability 15 (20), 14880 (2023).
14.
Cai, J., Deng, Z. & Li, L. High-resolution mapping of transport CO2 emission in Beijing–Tianjin–Hebei region: Spatial-temporal characteristics and decoupling effects. Int. J. Sustainable Transp. 18 (4), 301–314 (2024).
A
15.Luo, C. et al. Carbon peak and carbon neutrality scenarios for key industries in Jiangsu Province based on LEAP model. Environ. Sci. 46 (1), 1–9 (2025).
A
16.Qiao, Y., Zhang, J., Huo, T., Du, Q. & Cai, W. Shaping China's carbon peak roadmaps: A dynamic model for provincial residential buildings. Sustainable Prod. Consum. 55, 389–406 (2025).
A
17.Liu, Y., Luo, W., Tang, X. & Wang, X. J. Integrating STIRPAT-Adaboost and data-driven for influencing factors analysis and scenario prediction of carbon emission during building materialization stage. Urban Clim. 62, 102555 (2025).
A
18.Han, W. U., Keyan, T. E. N. G. & Jun, L. C. Comparative study on carbon dioxide peaking in some countries and regions and its enlightenment to China. J. Environ. Eng. Technol. 12 (6), 2032–2038 (2022).
A
19.Wang, D. et al. A Review of Research Progress in Carbon Monitoring and Carbon Metering Methods: Comparison at Home and Abroad. Processes, 12(12). (2024).
A
20.Zhang, B. L., Bai, Z. H., Zhang, N., Lü, L. H. & Yang, P. J. Evaluation Method and Application for Urban Carbon Peaking & Neutrality Performance. Huan Jing ke Xue = Huanjing Kexue. 44 (7), 3649–3659 (2023).
A
21.Yuan, D., Zhang, L., Fan, Y. & Yang, R. Investigating spatio-temporal variations and contributing factors of land use-related carbon emissions in the Beijing-Tianjin-Hebei Region, China. Sci. Rep. 14 (1), 18976 (2024).
A
22.Fang, Y. & Zhao, L. Exploring the decoupling effect and driving mechanism of carbon emissions at macroscale: An empirical study from Wuhan metropolitan area. Urban Clim. 61, 102434 (2025).
A
23.Wu, H., Yang, Y. & Li, W. Analysis of spatiotemporal evolution characteristics and peak forecast of provincial carbon emissions under the dual carbon goal: Considering nine provinces in the Yellow River basin of China as an example. Atmospheric Pollution Res. 14 (8), 101828 (2023).
A
24.Sahabuddin, M. & Khan, I. Analysis of demand, generation, and emission for long-term sustainable power system planning using LEAP: The case of Bangladesh. J. Renew. Sustain. Energy, 15(3). (2023).
A
25.Luo, C., Gao, Y., Jiang, Y., Zhao, C. & Ge, H. Predictive modeling of carbon emissions in Jiangsu Province's construction industry: An MEA-BP approach. J. Building Eng. 86, 108903 (2024).
A
26.Zhang, J. Research on carbon emission decoupling factors based on STIRPAT model and LMDI decomposition. Environ. Sci. 45 (4), 1888–1897 (2024).
A
27.Gao, L., Zhao, Z. Y., Li, C. & Wang, C. Factors facilitating the development of low-carbon cities: evidence from China’s pilot cities. Heliyon, 8(11). (2022).
A
28.Liu, H. D., Qiu, W. & Chen, S. Accounting and driving factors analysis of energy carbon emissions in Heilongjiang Province. China Environ. Sci. 44 (7), 4117–4126 (2024).
A
29.Usman, A., Ozturk, I., Naqvi, S. M. M. A., Ullah, S. & Javed, M. I. Revealing the nexus between nuclear energy and ecological footprint in STIRPAT model of advanced economies: Fresh evidence from novel CS-ARDL model. Prog. Nucl. Energy. 148, 104220 (2022).
A
30.Wei, M., Cai, Z., Song, Y., Xu, J. & Lu, M. Spatiotemporal evolutionary characteristics and driving forces of carbon emissions in three Chinese urban agglomerations. Sustainable Cities Soc. 104, 105320 (2024).
A
31.Lian, Y. Q., Su, D. H. & Shi, S. X. Carbon Peak Prediction in Fujian Province Based on Combined STIRPAT and CNN-LSTM Models. Environ. Sci. 46 (1), 10–18 (2025).
A
32.Li, X. & Zhang, X. A comparative study of statistical and machine learning models on carbon dioxide emissions prediction of China. Environ. Sci. Pollut. Res. 30 (55), 117485–117502 (2023).
A
33.Tian, Y., Ren, X., Li, K. & Li, X. Carbon Dioxide Emission Forecast: A Review of Existing Models and Future Challenges. Sustainability 17 (4), 1471 (2025).
A
34.Zhou, J., Zhang, Z., Xu, X. & Chang, D. Does the transformation of resource-dependent cities promote the realization of the carbon-peaking goal? An analysis based on typical resource-dependent city clusters in China. J. Clean. Prod. 365, 132731 (2022).
A
35.Zheng, Z. & Shuang, Q. Scenario analysis under climate extreme of carbon peaking and neutrality in China: A hybrid interpretable machine learning model prediction. J. Clean. Prod. 495, 145086 (2025).
A
36.Gao, D., Li, G. & Yu, J. Does digitization improve green total factor energy efficiency? Evidence from Chinese 213 cities. Energy 247, 123395 (2022).
A
37.Ding, T., Li, J., Shi, X., Li, X. & Chen, Y. Is artificial intelligence associated with carbon emissions reduction? Case of China. Resour. Policy. 85, 103892 (2023).
A
38.Chen, W., Zhang, J., Huang, C. & Fu, S. How landscape characteristics impact land surface temperature in the context of urban spatial heterogeneity: A case study from Beijing, China Vol. 60, 102366 (Urban Climate, 2025).
A
39.Guo, X. & Xu, J. Carbon reduction in the AI era: How does urban digital intelligence transformation facilitate low-carbon development? Urban Clim. 61, 102449 (2025).
A
40.Li, Z. et al. The impact of digital economy on industrial carbon emission efficiency at the city level in China: Gravity movement trajectories and driving mechanisms Vol. 33, 103511 (Environmental Technology & Innovation, 2024).
Acknowledgements
The authors would like to thank the anonymous reviewers and editors for their valuable time and constructive comments. This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.
A
Author Contribution
These authors contributed equally to this work.
A
Data Availability
All data are in the manuscript files.
A
Funding
Declaration
This document has not received any funding or other forms of financial support.
Competing Interests Statement
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
No potential conflict of interest was reported by the authors.
All authors disclosed no relevant relationships