Analyzing Integrated Carbon Emissions from Regional Transport and Land Use in the Context of National Spatial Planning

Abstract

Against the backdrop of intensified governance of territorial spatial planning, investigating carbon emissions from the perspective of territorial spatial planning for transport-land use integration holds significant academic and practical value. Taking Cangnan County as the case study, this research first dissects the reciprocal feedback mechanism between regional transport and land use at the territorial spatial planning level, while exploring transport-influencing factors. Subsequently, it constructs an integrated reciprocal feedback system for regional transport and land use by integrating accessibility drivers, cost matrices, and neighborhood weights through land use simulation–prediction models and the four-stage transport model. Finally, based on critical land use factors, diverse development scenarios under this integrated system are formulated; carbon emissions from transport and land use under each scenario are quantified; and their interrelationships are analyzed across multiple dimensions to explore the nexus of carbon emissions in transport–land use integration. Results indicate the following: (1) Integrated feedback enhances model accuracy (Kappa: 0.795→0.893; overall accuracy: 0.893→0.915), facilitating more precise land use simulation. (2) The county’s core construction area demonstrates the highest carbon emissions across all scenarios, meriting prioritized attention. (3) As deduced from the analysis of territorial spatial land use patterns, the significantly higher transport carbon emissions under the ecological protection priority scenario, compared to other scenarios, originate from over-concentrated construction land and imbalanced planning of carbon source land. These findings offer insights for regional planning; policy recommendations for Cangnan County include expanding carbon sink land, scientifically planning carbon source land, optimizing transport structures, and promoting new energy vehicles to advance carbon emission reduction and sustainable development.

1. Introduction

Over the past decade, the contradiction between economic development and ecological preservation in our country has become increasingly prominent, and it has emerged as an issue that requires urgent attention. Global carbon emissions continue to rise due to the continuous increase in energy consumption. In this regard, countries regard improving energy efficiency, reducing carbon emissions, and constructing low-carbon cities as key objectives in the pursuit of sustainable growth [1,2]. The Chinese government also released the Guidelines on Fully and Faithfully Implementing the New Development Philosophy to Achieve Carbon Peak and Carbon Neutrality in 2021, setting forth the strategic goals of achieving carbon peaking before 2030 and carbon neutrality before 2060. As the basis for economic development, the transportation industry has become one of the top three carbon-emitting industries in China, and it is expected that by 2030, the carbon dioxide emissions will be about four times the level of 2000 [3,4,5]. In this context, the transportation industry has become a key component in reducing carbon emissions [6,7], and about half of total energy consumption happens through road transportation. Clearly, in the context of territorial spatial planning, investigating carbon emissions arising from the integration of regional transportation and land use is of substantial significance for optimizing land use structure and improving carbon emission reduction efficiency.

At present, scholars both domestically and internationally have undertaken extensive academic inquiries to conduct in-depth analyses of carbon emissions associated with transportation systems and land use patterns [8,9]. This research includes the following.

➀ Exploring the relationship and integration model of land use and transportation from different levels of macro, medium and micro [10,11]. This research emphasizes the connection between land use and transportation It mainly adopts an integrated approach combining model simulation, remote sensing, and GIS to analyze the relationship between urban spatiotemporal growth and transportation changes. Furthermore, it proposes policy suggestions that adapt to the urban spatial layout and support the coupled development of different transportation modes based on the relationship between urban land use spatial structure and the spatial layout of urban transportation systems. Fesearch on the integration model of transportation and land use mainly includes five aspects: the intrinsic connection [12], the optimal allocation scheme [13,14], the comprehensive benefits [15], and interaction between macro and micro levels [16]. First, spatial interaction models based on the Lowry theory [17] explore the mutual influences among spatial elements to reveal the intrinsic connections between transportation and land use; second, models based on mathematical programming use mathematical optimization methods [2,18] to obtain optimal allocation schemes for transportation and land use; third, models based on the input–output approach evaluate the comprehensive benefits of transportation and land use by analyzing the economic connections between inputs and outputs [19]; fourth, models based on socioeconomic theories [20,21] explore the interactive relationship between transportation and land use and its impact on urban development from a socioeconomic perspective; fifth, models based on micro-simulation [11] study the micro-level interactions between transportation and land use by simulating individual behaviors.

➁ The aforementioned integrated models are often confined to intra-urban contexts, with relatively insufficient consideration of land use and transportation relationships at the regional scale. In the current era where territorial spatial planning has been repeatedly emphasized, the connotations of the integration of transportation and land use [22] have been significantly expanded. Integrated models should not only focus on the relationships between transportation and land use within cities but also emphasize the close connections between regional natural land use resources (such as cultivated land, forest land, and grassland) [23] and both urban–rural transportation and the overall regional transportation system. At the level of integrating territorial spatial land use and transportation, relevant models currently mainly focus on the simulation and prediction of land use. Simulation and prediction models consist of two categories: quantitative models [24] and spatial models. Quantitative simulation models mainly focus on changes in various land use types at the quantitative level and are unable to predict changes in their spatial locations. Currently, the main quantitative simulation models include grey prediction models [25], system dynamics models [26], artificial neural network models [27], and regression analysis models, among others. However, they can only provide simulations at the quantitative level and fail to reflect the dynamic evolution of spatial layouts.

Spatial simulation models, building on quantitative predictions and incorporating spatial simulation, can more accurately simulate spatial layout structures. The main ones include Cellular Automata (CA) models [28], CLUE-S models, and GEOMOD models, etc. [29]. Researchers both domestically and internationally have further improved these models and proposed coupled models, such as CA-Markov models [30,31], FLUS models [32,33], and Plus models [7]. These are also commonly used models at the current stage.

Studies on the factors influencing carbon emissions from land use and transportation have been carried out previously. In the context of land use carbon emissions, the primary influencing factors encompass natural factors such as topography, climatic conditions, and vegetation coverage [34,35], alongside socioeconomic factors including economy, population, and lifestyle [36]. With respect to carbon emissions from transportation, the research elements are primarily categorized into four dimensions: urban scale, urban spatial structure [37,38], land use intensity [5], and transportation accessibility. In terms of research methodologies, they are mainly divided into two aspects: one involves analyzing the impact of urban land use on transportation carbon emissions through model-based analysis [39,40], and the other entails identifying the optimal scheme for low-carbon urban development by comparing transportation carbon emissions under different scenarios via simulation and prediction [7,41].

In summary, current research on the integration of traffic and land use is largely carried out an urban scale, focusing on analyzing the relationships between the spatial distribution of internal urban land use types, their degree of mixing, and transportation accessibility. Such research can directly guide the internal planning and development of cities but lacks a grasp of overall regional trends. Scholars have considered the impact of land use on transportation carbon emissions, yet they have overlooked the feedback effect of transportation on land use: changes in land use structure induce changes in transportation demand, which in turn drive urban expansion and land development. Accordingly, this research takes Cangnan County as a case study area, elucidates the carbon emission linkages within the integration process of regional transportation and land use from the perspective of territorial spatial planning, uncovers their inherent driving mechanisms, and seeks to address the following research questions. (1) What is the evolutionary mechanism of regional land use from the perspective of territorial spatial planning? (2) What constitutes the mutual feedback mechanism between regional transportation and land use within the framework of territorial spatial planning? (3) What approaches are there to account for carbon emissions linked to regional transportation and land use? (4) What approaches can be employed to quantify carbon emissions from regional transportation and land use across different scenarios? The present research aims to supply a scientific grounding for regional carbon emission reduction, optimization of transportation structures, and formulation of land management decisions.

2. Overview of the Study Area and Data Sources

2.1. Overview of the Study Area

This research selects Cangnan County, Zhejiang Province as the study area. Situated at the southernmost coastal point of Zhejiang Province, Cangnan County is bordered by the East China Sea to the east and southeast, shares boundaries with Fuding City (Fujian Province) to the southwest, adjoins Taishun County to the west, and neighbors Pingyang and Wencheng Counties to the north. By year-end 2022, the county’s registered permanent population stood at 955,000 residents, comprising 480,200 urban inhabitants and 474,800 rural residents. Its gross regional product reached RMB 42.756 billion. According to the current status of land use in 2023, 4061.27 hectares of wetlands, 17,658.11 hectares of cultivated land, 4817.27 hectares of plantation land, 64,593.61 hectares of forest land, 703.01 hectares of grassland, 186.93 hectares of commercial service land, 1047.52 hectares of industrial and mining land, 5418.9 hectares of residential land, 669.82 hectares of public management and public service land, 652.38 hectares of special land, 3543.38 hectares of water and water conservancy facilities, and 1738.44 hectares of other land. The urban core of Cangnan lies 81 km from Wenzhou’s city center and 432 km from Hangzhou (provincial capital). National Highway 104 traverses the urban area north−south. Administratively, the county comprises 16 towns, 2 ethnic townships, 54 communities, 354 administrative villages, and 19 urban residents’ committees. The specific study area is delineated in Figure 1.

2.2. Data Sources and Processing

The dataset employed in this paper mainly include Cangnan County land use data, socioeconomic data, remote sensing geographic information data, digital elevation data, etc., and the data sources and pre-processing methods are shown in

Table 1. Types and sources of data.

Data Type	Data Sources
Administrative divisions vector data	National Geographic Information Resources Catalog Service System
Land use data	Data Center for Resource and Environmental Sciences, Chinese Academy of Sciences
DEM data, slope data	Geo-spatial Data Cloud (GDC)
Socioeconomic data, population density data, energy data	Wenzhou Statistical Yearbook, Cangnan County Statistical Yearbook
Geographic vector data (including road network data, settlement distribution data, railroad station location data)	National Geographic Information Resources Catalog Service System

.

This study employs land use datasets from Cangnan County for five time periods (2000, 2005, 2010, 2015, 2020) to analyze quantitative changes in and simulate spatial patterns of land use. The base data, sourced from the Resource and Environment Science Data Center of the Chinese Academy of Sciences, is in raster format with a resolution of 30 m × 30 m. These datasets were cropped to extract Cangnan County’s land use distribution and categorized into six types: cropland, woodland, grassland, water bodies, built-up land, and other land.

3. Research Methodology

From the perspective of national land use planning, the core of regional traffic and land use integration is the realization of a mutual feedback mechanism. In the process of training model accuracy, the mutual feedback between traffic and land use is taken into consideration. Building upon traditional land use simulation optimization, this paper further explores the impact of transportation on land use, constructs an integrated system between the two, and then re-simulates and structurally optimizes land use. The mutual feedback mechanism of the integrated system described in this paper is shown in Figure 2.

There is no simple one-to-one correspondence between regional transportation, land use, transportation carbon emissions, and land use carbon emissions. Land use influences transportation, leading to transportation carbon emissions, while transportation, in turn, affects land use, thereby impacting land use carbon emissions. These four factors interact and feed back into one another. The specific feedback mechanisms are illustrated in Figure 3.

This study employs the GeoSOS−FLUS model for land use prediction, which is developed based on the FLUS model [42]; it is designed to simulate land use dynamics and future scenarios under the combined influence of anthropogenic activities and natural factors. The FLUS methodology operates through two core stages: First, an artificial neural network (ANN) algorithm integrates single-temporal land use data with multidimensional driving factors to generate spatially explicit suitability probability distributions for each land use category within the study area. Second, an Adaptive Inertia Competition Mechanism (AICM) [43], utilizing a roulette wheel selection approach, is employed to resolve the uncertainties and complexities arising from the interactions between natural processes and human activities during land use type conversions. This dual-mechanism framework ensures high simulation accuracy, enabling the model’s outputs to closely approximate the actual spatial configuration of land use patterns.

3.1. Artificial Neural Network-Based Probability Calculation of Site Distribution Suitability

Neural network algorithms are divided into input layers, hidden layers, and output layers. The input layer consists of the driving factors that affect land use changes, while the output layer represents the various types of land use [32,44]. The formula is as follows:

$$sp(p,k,t)={\sum }_j{w}_{j,k}\ast sigmoid\left({net}_j(p,t)\right)={\sum }_j{w}_{j,k}\ast \frac{1}{1+e^{-\left({net}_j(p,t)\right)}}$$

(1)

where $sp(p,k,t)$ is the the occurrence probability of type of land use k at the training time t in the raster cell p; $w_{j,k}$ is the adaptive weight between the hidden layer and the output layer; $sigmoid\left(ne{t}_j(p,t)\right)$ is the activation function of the hidden layer to the output layer; and $e^{-}\left(ne{t}_j(p,t)\right)$ is the signal received by the grid cell in the hidden layer t during training.

3.2. Computation of Cellular Automata Based on the Adaptive Inertia Mechanism

Through the calculation of land use distribution probabilities via artificial neural networks, the correlations among suitability probabilities of different land use types can be distinctly manifested. Based on this, in combination with neighborhood effects, adaptive inertia coefficients, conversion costs, and constraints, the overall probability of land use change is computed. Finally, the final land use type for each grid is determined through a roulette selection mechanism [32].

Neighborhood Effect: The field effect reflects the expansion capacity of land use types during the process of land use change, as described by the following formula:

$${\Omega }_{p,k}^t={\displaystyle \frac{\sum N\ast Ncon(c_p^{t-1}=k)}{N\ast N-1}}\ast W_k$$

(2)

where ${\Omega }_{p,k}^t$ is the influence of the neighborhood on the class of land k in the cell of the grid p during the t-th iteration; $con(c_p^{t-1}=k)$ is the effect of the neighborhood of $t-1$ after the second-generation selection N*N for the type of land use k and the total number of grids occupied; and $W_K$ is the weight of the neighborhood expansion of the type of land use k.

Adaptive inertia coefficient: The essence of the adaptive inertial competition mechanism within the FLUS model lies in adaptive inertia. Specifically, the inertia coefficient for each land type is determined by the gap between the existing land quantity and the land demand, and it is adaptively adjusted during the iteration process, thus prompting the quantity of each land use type to evolve toward the preset target [32]. The adaptive inertia coefficient (${Inertial}_k^t$) of type k of land use at time t is given by the following formula:

$${Inertial}_k^t=\left\{\begin{array}{cc}{Inertial}_k^{t-1}\hfill & if|D_k^{t-1}|\le |D_k^{t-2}|\hfill \\ {Inertial}_k^{t-1}\times {\displaystyle \frac{D_k^{t-2}}{D_k^{t-1}}}\hfill & if0>D_k^{t-2}>D_k^{t-1}\hfill \\ {Inertial}_k^{t-1}\times {\displaystyle \frac{D_k^{t-1}}{D_k^{t-2}}}\hfill & if{D}_k^{t-1}>D_k^{t-2}>0\hfill \end{array}\right.$$

(3)

where $D_k^{t-1}$ denotes the discrepancy in grid cell counts between macro-level demand and supply quantity for land use type k at the $t-1$-th iteration.

Conversion cost: this denotes the difficulty of mutual conversion between land use types and takes the form of a binary matrix, where 0 indicates the impossibility of converting one land use type to another, and 1 indicates the possibility of such a conversion.

Roulette selection mechanism: The overall probability of each land use type is determined by a combination of suitability probabilities, neighborhood effects, adaptive inertia coefficients, and conversion costs. The combined probability ${TP}_{p,k}^t$ that grid cell P converts from its original land use type to land use type k at iteration t is given by the following formula:

$${TP}_{p,k}^t=P_{p,k}\ast {\mathsf{\Omega }}_{p,k}^t\ast {Inertiala}_k^t\ast (1-{sc}_{c⤏k})$$

(4)

where $P_{p,k}$ is the suitability probability; ${\mathsf{\Omega }}_{p,k}$ is the neighborhood effect; ${Inertiala}_k^t$ is the adaptive inertia coefficient; and ${sc}_{c⤏k}$ is the original land use type c converted to land use type k at the required conversion cost.

3.3. Carbon Emission Measurement Methods

(1) The carbon emissions from land use are assessed using direct carbon emission measurement methods [45]. The formula is as follows.

$$E1=\sum e_i=\sum S_i\ast {\delta }_i$$

(5)

where $E1$ represents the total direct carbon emissions; $e_i$ represents the amount of carbon emissions (of absorption) generated by different land use types; $S_i$ represents the area of different land use types; and ${\delta }_i$ represents the carbon emission (absorption) coefficients of different land use types.

Based on the 2018 National Greenhouse Gas Inventory from the Third Biennial Update Report on Climate Change of the People’s Republic of China (released by the Ministry of Ecology and Environment in 2023), this study determined carbon emission coefficients for various land use types through a critical review of the existing literature [46] integrated with socioeconomic and physio-geographical conditions specific to Cangnan County. The derived coefficients are presented in

Table 2. Direct carbon emission factors for land use.

Land Use Type	Carbon Emission Factor
plowed land	42.2
woodland	−64.4
grasslands	−2.10
body of water	−21.8
unused land	−0.5

.

The quantification of carbon emissions from construction land employs an indirect methodology for measuring carbon emissions [47] with the following formula:

$$E2=\sum e_j=\sum m_j\ast c_j\ast {\sigma }_j$$

(6)

where $E2$ is the carbon emission from construction land; $e_j$ is the carbon emission of various energy sources in the study area; j is the type of energy; $m_j$ is the terminal consumption of different energy sources; $c_j$ is the coefficient of converting various energy sources into standard coal; and ${\sigma }_j$ is the carbon emission coefficient of each energy source. $c_j$ and ${\sigma }_j$ are taken as in

Table 3. Standard coal conversion factors and carbon emission factors for different energy sources.

Type of Energy	Standard Coal Conversion Factor	Carbon Emission Factor (t C/t)
raw coal	0.7143 (t standard coal/t)	0.7559
coke (processed coal used in blast furnace)	0.9714 (t standard coal/t)	0.8550
diesel	1.4714 (t standard coal/t)	0.5538
crude oil	1.4286 (t standard coal/t)	0.5857
gasoline	1.4714 (t standard coal/t)	0.5714
diesel fuel	1.4571 (t standard coal/t)	0.5921
fuel oil	1.4286 (t standard coal/t)	0.6185
liquefied petroleum gas	1.7143 (t standard coal/t)	0.5042
thermodynamic	3.4122 (t standard coal/multifarious coke)	0.2601
electrical power	0.1229 (kg/kwh)	0.2132

.

This article predicts land carbon emissions by simulating and optimizing future land use, taking into account the uncertainty of future urban energy consumption data. It calculates the total carbon emissions from construction land in Cangnan County based on energy consumption per unit of GDP [48]. Its calculation formula is

$$E3=E_w\ast {\displaystyle \frac{{GDP}_c\ast P_c}{{GDP}_w\ast P_w}}$$

(7)

where $E3$ is the carbon emission from construction land in Cangnan County, and $E_w$ is the carbon emission from construction land in Wenzhou City, which is measured according to the equation. ${GDP}_c$ and ${GDP}_w$ are the gross regional product of Cangnan County and Wenzhou City, respectively; and $P_c$ and $P_w$ are the unit GDP energy consumption of Cangnan County and Wenzhou City. The total carbon emissions from land use are the sum of carbon emissions and carbon absorption for each type of land use:

$$E=E1+E3$$

(8)

where E is the total carbon emissions from land use. $E1$ represents direct carbon emissions, and $E3$ represents indirect carbon emissions.

(2) Transportation Carbon Emission Measurement

At present, the measurement methods for road traffic carbon emissions both domestically and internationally mainly adopt the IPCC mobile emission source calculation method, also known as the emission factor method [48]. It is divided into two forms: “top-down” and “bottom-up”. Among them, the ”top-down” approach estimates transportation carbon emissions using energy consumption and energy conversion factors [49], with the following formula:

$$E={\sum }_i[S_i\ast {EF}_i]$$

(9)

In the formula: E denotes the total transportation carbon emissions in kg; i is the fuel type, containing gasoline, diesel, kerosene, natural gas, etc.; $S_i$ denotes the consumption of fuel i, in $TJ$; and ${EF}_i$ is the emission coefficient of the fuel i in kg/TJ.

The “bottom-up” approach uses data on different modes of transportation, vehicle types, distance traveled, etc., to calculate transportation carbon emissions [50]. The formula is as follows:

$$E={\sum }_{i,j}[D_{i,j}\ast S_{i,j}\ast P_i\ast G_i\ast {EF}_{i,j}]$$

(10)

where j denotes the type of transportation; $D_{i,j}$ denotes the type of transportation using i fuel for the j total miles traveled by the type of transportation, in km; $S_{i,j}$ is the total distance traveled by the mode of transportation using i fuel j mode of transportation, in L/km; $P_i$ is the fuel density of the fuel i, in kg/L; $G_i$ is the net calorific value of the fuel i, in TJ/kg; and ${EF}_{i,j}$ is the emission factor of the fuel i, in kg/TJ.

4. Analysis of Results

4.1. Driver Selection

Combined with the selection principle of driving factors and referring to the previous research results [51], this paper identifies eight driving factors for the simulation and prediction of land use, as shown in

Table 4. Selection of drivers of land use change.

Typology	Driving Factor	Hidden Meaning
socioeconomic	population density	Population density at the center of each grid cell
	GDP	GDP at the center of each raster cell
physical geography	altitude (e.g., above street level)	Elevation value of the center point of each raster cell
	elevation	Slope of the center point of each raster cell
accessibility	Road network density	Fishing net density for each grid cell
	Distance to town center	Euclidean distance from each grid to town center
	Distance from road	Euclidean distance from each grid to the highway
	Distance to railroad station	Euclidean distance from each grid to the railroad station

.

4.2. Characterization of the Spatial and Temporal Evolution of Land Use

4.2.1. Quantitative Land Use Characterization

Drawing on remote sensing images of Cangnan County spanning from 2000 to 2020, the spatiotemporal evolutionary characteristics of land use in Cangnan County during this period are listed in

Table 5. Cangnan County in 2000–2020: combined area and percentage (square kilometers) of five-phase land use.

Particular Year	2000		2005		2010		2015		2020
topography	area	percentage	area	percentage	area	percentage	area	percentage	area	percentage
cultivated land	285.46	27.88%	267.56	26.12%	266.12	25.99%	261.61	25.56%	259.28	25.29%
forest land	670.57	65.50%	674.46	65.85%	665.73	65.02%	666.2	65.08%	655.8	63.96%
grassland	40.71	3.98%	38.77	3.79%	39.77	3.88%	39.58	3.87%	40.21	3.92%
water area	6.34	0.62%	6.21	0.61%	6.34	0.62%	6.03	0.59%	6.34	0.62%
construction land	20.25	1.98%	36.78	3.59%	45.51	4.44%	49.78	4.86%	63.16	6.16%

.

From Table 5, it can be observed that between 2000 and 2020, there was a significant change in land use in Cangnan County. Arable land as a whole exhibited a downward trend, falling from 27.88% in 2000 to 25.29% in 2020. The areas of forest land, grassland, and water area remained relatively stable, while construction land underwent notable changes, increasing from 1.98% in 2000 to 6.16% in 2020. Urban land expanded from 7.74 square kilometers to 20.47 square kilometers between 2000 and 2005. From 2015 to 2020, other types of construction land grew from 12.13 square kilometers to 25.84 square kilometers, mainly including transportation facilities, energy land, and special-purpose land. Through analysis, it can be seen that the construction land in Cangnan County has rapidly expanded, occupying a large amount of forest and arable land, most notably between 2000 and 2005.

4.2.2. Land Use Transfer Matrix Analysis

The land use transfer matrix is a two-dimensional matrix that functions to depict the relationships of land use changes within the same region over different time periods [52]. The conversion relationships and characteristics between land use types in Cangnan County during various time periods from 2000 to 2020, derived using ArcGIS 10.7 software, are shown in Figure 4.

As indicated in Figure 4, from 2000 to 2005, construction land expanded significantly, with the vast majority originating from cultivated land. Around 2000, China’s urbanization and industrialization progressed rapidly, resulting in swift economic growth, with the expansion of urban infrastructure primarily focused on areas surrounding city centers. From 2005 to 2010, the pace of expansion of construction land slowed down, mainly occurring at the edges of urban areas. Concurrently, the implementation of farmland protection policies has resulted in a reduced rate of loss of arable land. From 2010 to 2015, land for construction continued to decrease, reflecting the effectiveness of ecological protection and environmental protection policies. From 2015 to 2020, there was an inflow of land for arable land, forested land, and land for construction, and wetland protection and fallow farmland and forest and grassland projects continued to be promoted, with further development of the land being driven by urbanization and industrialization.

4.3. Transportation and Land Use Integration Analysis

4.3.1. Mutual Feed Mechanisms for the Integration of Transportation and Land Use

Viewed through the lens of territorial spatial planning, the essence of integrating regional transportation and land use lies in the materialization of the mutual feedback mechanism. In this study, the mutual feedback mechanism of the integrated transportation and land use system takes the GeoSOS-FLUS model as the basic framework, incorporates transportation accessibility as a driving factor, and realizes feedback through cost matrices and neighborhood weights.

Concretely, the operation involves taking initial land use data and multi-dimensional driving factors (including transportation accessibility) as inputs and calculating the spatial suitability probability of land use types by means of an artificial neural network. By integrating adaptive inertia coefficients, neighborhood effects, and a policy-oriented conversion cost matrix, it iteratively generates new land use layouts using a cellular automaton. This layout is fed back to the transportation module in real time, which updates regional accessibility indicators that are then reintroduced into the driving factor set, thereby forming a dynamic closed-loop of “land use change → transportation network response → accessibility update → re-optimization of land use layout”.

4.3.2. Transportation Accessibility Feedback Analysis

The interplay between territorial spatial land use and transportation structure is not a simple corresponding relationship. Land use acts on transportation, and transportation reacts on land use; the two form a dynamic system characterized by mutual feedback and coordinated development. Neighborhood weights and transition probability matrices before and after integration are constructed, and in conjunction with the adjusted transition probability matrix of land use types, the Markov model is employed to predict the quantity of land use in 2020. The accuracy of the simulation is verified via the Kappa coefficient, as shown in

Table 6. Kappa accuracy and overall accuracy values before and after integrated feedback.

Modeling Land Use in 2020	Kappa Accuracy	Overall Accuracy
No integration feedback	0.795193	0.893272
Integration feedback is available	0.893001	0.915631

.

The results indicate that the Kappa accuracy and overall accuracy before integration are relatively low, while the parameter settings after integration are more reasonable, leading to higher land use simulation accuracy and higher reliability of the results. This enables further simulation and optimization of the future spatial distribution of land use.

4.4. Carbon Emission Analysis of Different Scenarios Based on FLUS Simulation Results

4.4.1. Scenario Configurations

After constructing an integrated system of transportation and land use, the Kappa coefficient reached 0.893, and the overall accuracy value reached 0.915, indicating a high level of model simulation accuracy. Consequently, under the constraints of quantity control and optimization criteria, the 2020 land use data served as input to simulate land use allocation patterns for 2035. Accounting for Cangnan’s current development trajectory and incorporating regional territorial spatial planning frameworks—including cultivated land preservation policies and permanent prime farmland protection—multiple scenarios were parameterized through adjustments to land transition probability matrices, neighborhood interaction factors, and spatial conversion cost matrices. The scenario is set as follows:

(1)

Natural Development Scenario

Under this scenario, land use change follows the trajectory observed during 2005–2020, evolving autonomously according to Cangnan County’s intrinsic drivers and constraints without external interventions. Using a 15-year projection horizon, land demand for 2035 was estimated via Markov chain modeling within the integrated transportation-land use framework. All parameters—including transition probability matrices, neighborhood factors, and land conversion cost matrices—remained consistent with those validated in the 2020 land use simulation.

(2)

Priority Scenario for Cultivated Land Protection

In this scenario, the expansion of cultivated land area is realized by restricting its transfer to other land categories based on natural development. In these settings, the probability of arable land being converted has been reduced, with a 40% decrease in transfers to forest land, grassland, and water bodies; a 60% decrease in transfers to developed land; and a 100% decrease in transfers of arable land to unused land. Concurrently, on the basis of natural development, the neighborhood factor for cultivated land is adjusted to increase by 0.2, while that for unused land is adjusted to decrease by 0.1. In the cost matrix, cultivated land is restricted from converting to unused land.

(3)

Priority Scenario for Ecological Protection

In this scenario, it is proposed to enhance the growth rates of forest and grassland based on natural development while slowing the expansion rates of arable land and urban areas. Specifically, first, we adjust the probabilities of land use type conversion, increasing the conversion of arable land to forest land by 20%, while reducing the conversion of forest and grassland to arable land and construction land by 50%; second, the neighborhood parameters for non-utilized land and urban construction land are decreased by 0.1, while the neighborhood parameters for forest land, grassland, and water bodies are increased by 0.1; third, the priority ranking for land conversion in the cost matrix is clearly established as forest land > grassland > water bodies > other land types.

(4)

Priority Scenario for Economic Development

In this scenario, for rapid urban land expansion, arable land, forests, grasslands, and unused land will be reduced. Meanwhile, the probability of other land types converting to urban land rises by 50%. Additionally, based on natural development, the neighborhood factor for construction land will be increased by 0.05, while the neighborhood factors for other land types will be decreased by 0.05. Furthermore, the cost matrix establishes that only other land types can transition to construction land, with the transition hierarchy being construction land > other land types.

4.4.2. Analysis of Land Use Carbon Emissions

After calculating the carbon emission values from land use for various types of land under different scenarios, the carbon emissions from carbon source land, as well as the carbon emissions and carbon sequestration amounts from carbon sink land, are shown in Figure 5.

Analysis of carbon emissions related to land use indicates significant differences in various development scenarios. The economic development priority scenario exhibits the highest carbon emissions at 1.86 million tons, primarily driven by urban construction. In contrast, the ecological protection priority scenario shows the lowest carbon emissions, at only 1.61 million tons. In terms of carbon sink capacity, the ecological protection scenario boasts the highest capacity at 42,500 tons, surpassing the natural development scenario’s 41,400 tons. A critical observation is that in the farmland protection priority and economic development priority scenarios, the expansion of farmland and construction land encroaches upon forests and grasslands. Given that forest land serves as a major carbon sink with significant carbon sequestration potential, effective strategies to mitigate land use carbon emissions must prioritize controlling urban land expansion and enhancing forest land coverage.

4.4.3. Analysis of Carbon Emissions from Transportation

From the perspective of territorial spatial planning, there exist significant distinctions between traffic analysis zones (TAZs)’ delineation at the county level and urban level. When partitioning TAZs in urban areas, more granular consideration must be given to transportation patterns across different functional zones, population density gradients, and land use intensity variations within the urban fabric. Conversely, county-level zoning (encompassing multiple towns and rural areas) necessitates comprehensive evaluation of hierarchical land use structures across urban–rural continuums, multimodal transportation networks linking regional development nodes, and spatially heterogeneous population distribution patterns shaped by geographic and socioeconomic factors. Given Cangnan County’s unique characteristics, the jurisdictional territory has been strategically divided into 10 major traffic zones as illustrated in Figure 6.

Based on the residents’ travel survey data of Cangnan County, we established trip generation and attraction prediction models by integrating household counts and employment opportunities across TAZs. The generation rate method was employed in these models, which first derives scenario-based projections of household numbers and employment opportunities in each major traffic zone through urban–rural spatial planning population forecasts. Subsequently, this method computes daily trip generation and attraction volumes for all traffic zones by applying calibrated generation rates to socioeconomic parameters. This approach particularly considers the county’s urban–rural integration characteristics, where urban zones demonstrate higher trip generation intensities correlated with concentrated employment centers, and rural zones require differentiated attraction rate adjustments based on agricultural production cycles and commuting patterns. For each macro-region under the four scenarios, their transportation carbon emissions were separately measured, and a comparative analysis chart was formulated based on the measurement results.

Due to the independent computational frameworks of trip production and attraction prediction models, discrepancies inevitably arise between total trip production and attraction volumes. Consequently, prior to implementing trip distribution modeling, it is imperative to reconcile trip production and attraction values across all TAZs within the study area to achieve systemic equilibrium. This study adopts a production-constrained balancing methodology, wherein trip attraction values are iteratively adjusted relative to trip production baselines until a PA equilibrium is attained. This calibration yields revised zonal trip production and attraction matrices, ensuring spatial consistency in demand allocation. The equilibrium allocation process aims to stabilize internal traffic flow patterns, thereby enhancing the model’s fidelity to real-world mobility dynamics. As illustrated in Figure 7, zonal trip production and attraction intensities exhibit pronounced spatial heterogeneity. High-demand clusters are predominantly localized within the northern urban development core, encompassing the Fengchi macro-zone, Xianju macro-zone, and Yanting macro-zone. These areas correlate with concentrated residential–commercial land use configurations and robust multimodal infrastructure connectivity, reflecting their role as primary trip generation hubs in the county’s integrated transportation network.

Based on the predicted traffic generation and attraction, a traffic distribution prediction model was established to forecast the traffic volume distribution between traffic zones within the region. The gravity model was employed to predict the traffic distribution under the natural development scenario of Cangnan County. Desire lines for each traffic zone were plotted based on traffic flow data from each road segment or traffic node, as shown in Figure 8. It can be observed from the figure that there are relatively more trips between the Qiaodun area and the Zaoxi area, Yishan area, Fengchi area, and Xianju area. In particular, the traffic demand in the Fengchi area is relatively high, which may lead to congestion during peak hours.

Based on the transportation carbon emissions of different regions under four scenarios, a comparative chart of transportation carbon emissions for various transportation districts in each scenario is presented in Figure 9.

From Figure 9, it is evident that the trends in traffic carbon emissions across the four scenarios are similar, with the Fengchi region exhibiting the highest emissions, followed by the Zaoxi and Xianju regions. These areas are located in the northern part of Cangnan County, which is densely populated and considered a key area for carbon reduction. The Yishan and Fengyang regions have comparatively lower traffic carbon emissions, primarily due to their terrain and sparse population, with residents relying more on public transportation.

4.4.4. Comprehensive Analysis of Carbon Emissions from Transportation and Land Use Integration

After completing the analysis of carbon emissions from land use and transportation within the integrated transportation and land use system, we further synthesized and examined the relationship between the two to assess the overall carbon emission impact in four scenarios. The resulting findings are presented in

Table 7. Sum of carbon emissions from transportation and net carbon emissions from land use for different scenarios in Cangnan County in 2035 (ten thousand tons).

	Natural Development	Prioritization of Arable Land Protection	Prioritize Ecological Protection	Prioritize Economic Development
Transportation carbon emissions	32.13	30.66	34.11	35.03
Net carbon emissions from land use	174.22	161.25	156.75	181.44
Total carbon emissions	206.35	191.90	190.86	216.46

.

With respect to total carbon emissions, the economic development priority scenario has the highest. The key reason for this is the massive expansion of construction land driving up land use emissions; conversion of forests and other carbon sink lands to construction land weakens regional carbon sequestration and storage capacity, while construction activities themselves generate substantial emissions, jointly making land use emissions here the highest. Given transportation emissions account for a relatively small share of total emissions, this scenario remains top in total emissions even if its transportation emissions are not the largest. Next is the natural development scenario, with total emissions reaching 2,063,500 tons, about 5% higher than the cultivated land protection and ecological protection priority scenarios. Land use changes here follow existing trends and historical data, reflecting spontaneous evolution of land resources. Total emissions of the cultivated land protection and ecological protection priority scenarios differ slightly, but show significant divergence in transportation and land use emissions. Specific values are shown in Figure 10.

As seen in the figure, the trends in transportation and land use carbon emissions under the four 2035 scenarios of Cangnan County are generally aligned. This stems from the deep coupling of land use changes and transportation systems; construction land distribution shapes traffic demand and travel patterns, while transportation layout and efficiency drive construction land expansion, jointly impacting both emission types. However, under the scenario prioritizing ecological protection, though land use carbon emissions are lowest, transportation emissions are abnormally high. As the earlier analysis of land use spatial layout and transportation network cost matrices shows, this is mainly because land use changes driven by the adaptive inertial competition mechanism significantly reduce conversion of original forests and grasslands to construction land. Specifically, construction land is more concentrated in the northern core areas (originally densely populated), raising road segment saturation, worsening regional congestion, and triggering abnormal shifts in traffic demand and travel patterns.

In different development scenarios, transportation carbon emissions in the economic development priority scenario reach 350,300 tons, which is approximately 30,000 tons higher than that in the natural development scenario. This indicates that economic development leads to an increase in traffic demand, resulting in a significant rise in carbon emissions. Even in the ecological protection priority scenario, transportation carbon emissions still amount to 341,100 tons, indicating that concentrated land use and insufficient optimization of the transportation network still result in relatively high emissions. In summary, land use exerts a notable influence on transportation carbon emissions.

5. Discussion

This study developed an integrated model of regional transportation and land use from the perspective of territorial spatial planning (GeoSOS-FLUS coupled with a traffic feedback mechanism). Taking Cangnan County as a case study, it simulated land use changes in different development scenarios and their comprehensive impacts on transportation carbon emissions and land use carbon emissions.

In contrast to most studies on land use–transport models that are centered on intra-urban commuting or single factors [11,53], the regional-scale integrated model constructed in this study innovatively extends the mutual feedback mechanism to the territorial spatial dimension, incorporating natural elements such as forestland and cropland. Our results indicate that the model significantly enhances simulation accuracy, with a Kappa coefficient reaching 0.893, thus validating that the regional transportation network constitutes the core driving force shaping the overall spatial pattern of land use across the entire region [10]. This model offers a more robust tool for the refined assessment of carbon effects associated with territorial spatial planning policies, thereby remedying the deficiencies of existing models with regard to a regional comprehensive perspective.

Unlike similar scenario simulation studies that primarily focus on land use carbon emissions [18,27], this study synchronously quantifies the comprehensive impacts of different territorial spatial policy orientations—including economic priority, ecological priority, and cropland priority—on the dual carbon emissions from transportation and land. The results not only confirm that economic priority causes a surge in land carbon emissions by 1.81 million tons and ecological priority increases carbon sequestration by 42,500 tons but also reveal a key discrepancy: although pure ecological constraints can achieve the lowest net land emissions of 1.57 million tons, they fail to automatically reduce transportation emissions as anticipated by the compact city theory. Instead, in this case, they even cause transportation emissions to rise to the highest level of 340,000 tons.

In response to the aforementioned phenomenon, this study, through analyses of spatial layout and transportation networks, proposes an explanation distinct from simplistic density-based theories: ecological constraints have led to the excessive concentration of newly added construction land in existing high-density urban districts, triggering severe traffic congestion that offsets the theoretical emission reduction benefits associated with increased density. This highlights a core contradiction at the regional scale, one that diverges from the focus of residential self-selection research: without coordinated spatial optimization and efficiency improvements in transportation systems, merely restricting urban sprawl may yield counterproductive outcomes. This finding furthers our understanding of the complexities inherent in “compact cities” and empirically verifies the existence of complex interactions involving synergies and trade-offs between transportation and land use carbon emissions, with the core coupling node residing in the spatial allocation of construction land.

Despite the results obtained above, this study still has certain limitations. The transportation carbon emission model, relying on macro-level flow forecasting, did not incorporate simulations of individual travel behavior, microscopic road network congestion dynamics, or their non-linear amplifying effects on carbon emissions. Future research could integrate microscopic traffic simulation models to more accurately capture congestion impacts. Although scenario designs were policy-oriented, the representation of transportation system responses within corresponding scenarios remained relatively coarse-grained. Future work should develop more refined combinatorial scenarios, such as “ecological protection coupled with transportation system optimization”. The spatiotemporal resolution of certain socioeconomic data (e.g., township-level GDP, precise energy consumption statistics) requires enhancement. Future studies could leverage multi-source big data integration to improve the characterization of driving forces and validation accuracy. Static average values were adopted for land use carbon sink coefficients, failing to account for dynamic changes induced by evolving factors such as forest age, tree species composition, and management practices. This simplification potentially undermines the temporal accuracy of carbon sequestration estimates.

Future research could focus on the following directions for further development: utilizing more sophisticated models that couple traffic congestion dynamics with carbon emissions; exploring collaborative emission reduction pathways through multidimensional policy combinations; incorporating multiple objectives such as climate resilience and biodiversity conservation into the low-carbon optimization framework for land use; and expanding applications for comparative studies across different regional types.

6. Conclusions

The integrated regional transportation-land use model developed in this study, framed within the national spatial planning perspective, significantly enhances the accuracy of land use simulation. This robustly validates regional transportation networks as the fundamental driver shaping the spatial pattern of land use across the entire territory, thereby providing an effective tool for precisely assessing the carbon implications of planning policies.

The rsults demonstrate that the economic development priority scenario yields the highest total carbon emissions stemming from the unconstrained expansion of construction land. Conversely, while the ecological protection priority scenario achieves the lowest net land use carbon emissions and the highest carbon sequestration, its transportation carbon emissions emerge as the highest among the four scenarios. This counterintuitive paradox stems from the fact that, under ecological constraints, new construction land is forced into excessive concentration within existing high-density urban areas. This concentration induces severe traffic congestion, effectively offsetting the theoretical emission reduction benefits associated with increased density.

The study reveals the complex relationship of synergies and trade-offs between transportation and land use carbon emissions, with the spatial allocation of construction land as the core coupling node. To achieve in-depth regional carbon emission reduction, it is urgent to break the traditional fragmented thinking in territorial spatial planning and construct a systematic framework for coordinated optimization of land and transportation. We must scientifically guide polycentric, cluster-based development to avoid excessive concentration; simultaneously improve public transport services and road network efficiency in hotspots; strictly protect natural carbon sink land; and implement differentiated strategies for high-emission areas so as to lead the low-carbonization of transportation through the low-carbonization of spatial structure.