Spatiotemporal Evolution, Spatial Heterogeneity and Driving Mechanisms of Agricultural Carbon Emissions in the Main Grain-Producing Areas of the Middle Reaches of the Yangtze River, China

Abstract

Achieving low-carbon agriculture in major grain-producing areas hinges on balancing emission reduction imperatives with the need to maintain food security. However, previous studies have largely examined agricultural carbon emissions (ACE) at administrative scales, potentially overlooking heterogeneous agricultural functions and emission pathways within urban agglomerations. This study integrates ACE accounting, spatial inequality and autocorrelation analyses, and Logarithmic Mean Divisia Index (LMDI) decomposition to investigate the spatiotemporal dynamics, spatial differentiation, and driving mechanisms of ACE in the Main Grain-Producing Areas of the Middle Reaches of the Yangtze River, China, using panel data from 31 prefecture-level cities during 2013–2023. The results show that ACE experienced expansion, decline, and subsequent stabilization, while agricultural carbon intensity decreased by 37.29%, indicating gradual decoupling between agricultural development and carbon emissions. Significant spatial heterogeneity and positive spatial dependence were observed, with intra-agglomeration disparities contributing more to overall inequality than inter-agglomeration differences. The three urban agglomerations also exhibited distinct emission characteristics associated with their agricultural functions and emission structures. LMDI decomposition showed that agricultural economic development was the main factor increasing ACE, whereas improvements in production efficiency and changes in rural population mitigated emissions. Overall, the results reveal that similar regional emission trends can arise from heterogeneous agricultural functions, emission structures, and development trajectories. This highlights the need for differentiated low-carbon transition pathways rather than uniform emission-reduction strategies and provides a basis for coordinating food security with low-carbon agricultural development.

1. Introduction

Agriculture plays a fundamental role in ensuring global food security and supporting sustainable socioeconomic development [1], but it is also a significant contributor to greenhouse gas (GHG) emissions [2,3,4]. According to the Intergovernmental Panel on Climate Change (IPCC) Sixth Assessment Report, the agriculture, forestry, and other land use (AFOLU) sector accounted for approximately 22% of global net anthropogenic greenhouse gas emissions in 2019 [5]. With population growth and increasing food demand, agricultural production has become increasingly dependent on energy-intensive inputs [6], making agricultural carbon emissions (ACE) a growing concern [7]. Achieving emission reductions while maintaining sufficient food production has therefore become a critical challenge for advancing the Sustainable Development Goals (SDGs) [8]. Unlike industrial emissions, ACE are closely tied to biological processes, farming practices, and regional environmental and socioeconomic conditions. Consequently, mitigation pathways vary substantially across regions due to differences in cropping systems, technological capabilities, and socioeconomic development levels [9,10]. This context dependence implies that effective strategies require a detailed understanding of regional emission characteristics, rather than uniform approaches.

China exemplifies this challenge. As one of the world’s largest agricultural producers [11], China has incorporated agricultural emission reduction into its broader carbon peaking and carbon neutrality strategy [12]. However, agricultural decarbonization cannot simply reduce production scale, as maintaining grain self-sufficiency remains a national priority. Under this constraint, China must pursue a transition toward more efficient and low-carbon production systems. The main grain-producing areas (MGPA) in the middle reaches of the Yangtze River (MRYR) provide a representative context for examining these issues. As a major grain-producing region, it simultaneously undertakes critical grain supply responsibilities and faces increasing pressure for low-carbon transition, providing an appropriate setting for investigating spatial heterogeneity and differentiated emission characteristics.

However, three gaps remain in the existing literature. First, although ACE have been examined at various administrative scales, the urban agglomeration—as a functional economic space that integrates metropolitan cores with surrounding agricultural hinterlands—has received limited attention. This matters because administrative boundaries may obscure the differentiated agricultural production characteristics and regional development linkages that shape emission patterns across functionally integrated regions. Second, spatial pattern analysis and driving mechanism decomposition have largely been conducted separately, leaving the relationship between regional emission disparities and their underlying driving mechanisms insufficiently explored. Third, limited attention has been paid to how heterogeneous agricultural production characteristics and emission structures within urban agglomeration systems shape differentiated emission dynamics, thereby requiring differentiated mitigation approaches. These gaps are particularly salient in major grain-producing regions, where emission reduction must be coordinated with food security objectives.

To address these limitations, this study develops an integrated analytical framework combining ACE accounting, the Theil index, spatial autocorrelation analysis, and LMDI decomposition to examine the spatiotemporal evolution, spatial heterogeneity, and driving mechanisms of ACE in the MGPA of the MRYR during 2013–2023. Three research questions are addressed: (1) How have ACE, agricultural carbon intensity, and carbon-source composition evolved over time? (2) What spatial disparities and agglomeration characteristics exist across the study region and its urban agglomerations? (3) What factors drive ACE changes, and how do their effects vary across regions?

This study contributes to the literature in three aspects. First, it extends the analytical unit of ACE research from administrative divisions to urban agglomeration systems, revealing that ACE disparities reflect not only geographical differences but also differentiated agricultural production characteristics and emission structures within integrated regional systems. Second, by integrating spatial pattern analysis and driving mechanism decomposition, this study clarifies how spatial disparities and emission changes are shaped by different regional development pathways, providing insights into the mechanisms underlying ACE evolution across urban agglomerations. Third, this study reveals how heterogeneous agricultural production characteristics and emission structures within urban agglomeration systems shape differentiated emission dynamics and transition pathways, highlighting the need for region-specific mitigation strategies.

The remaining sections proceed as follows. Section 2 reviews the literature on ACE measurement, spatial heterogeneity, and driving mechanisms. Section 3 documents the study area, data sources, and analytical methods. Section 4 presents the empirical results, which are further interpreted and discussed in Section 5. Section 6 summarizes the main findings, acknowledges the study limitations, and identifies directions for future research.

2. Literature Review

2.1. Research on ACE Measurement and Carbon-Source Composition

Accurate ACE measurement serves as the foundation for identifying emission sources, assessing reduction potential, and formulating low-carbon agricultural policies. ACE accounting methods differ mainly in analytical scope and data requirements. Life-cycle assessment (LCA) [13,14,15] and input–output analysis [16] capture broader indirect emissions but require detailed process-level or inter-sectoral data. By contrast, the IPCC emission-factor accounting [17,18,19,20,21] provides a relatively transparent and operationally consistent framework for quantifying emissions from specific agricultural activities, making it particularly suitable for prefecture-level panel analysis across multiple agricultural regions where consistent estimation of emission levels and carbon-source composition is essential for spatial comparison.

In China, existing studies have generally estimated ACE from four major sources: agricultural inputs (fertilizers, pesticides, diesel, plastic film, irrigation) [22], methane emissions from rice cultivation, livestock breeding (enteric fermentation and manure management) [23], and crop straw burning [24]. Because the scale and composition of these agricultural activities vary across regions, their contributions to ACE may also differ geographically. However, most existing studies have focused on aggregate ACE, with limited attention to how carbon-source composition varies across different regions.

2.2. Research on Spatial Heterogeneity and the Urban-Agglomeration Perspective

ACE exhibit pronounced spatial heterogeneity, and characterizing their spatiotemporal evolution patterns is essential for understanding regional disparities and identifying priority areas for mitigation. This heterogeneity reflects differences in farming systems [25,26], resource endowments [27], production intensity [28,29], and the spatial organization of agricultural activities [30]. Existing studies in China have examined the spatial distribution and evolution of ACE at national [11,22,31,32,33], provincial [24,34,35,36], and county [37] scales. Studies using inequality measures such as the Theil index [38,39] and Gini coefficient [22,40] have generally identified persistent regional disparities in ACE, with intra-regional differences constituting an important component of overall spatial inequality [41]. Kernel density analysis has further revealed the dynamic evolution and polarization of regional emission distributions [32], while spatial autocorrelation analyses have consistently identified significant spatial clustering of ACE [42]. These findings suggest that spatial disparities in ACE are associated with differences in agricultural specialization, resource endowments, production intensity, and urbanization levels.

However, these studies have predominantly characterized spatial heterogeneity within conventional administrative units, leaving the internal differentiation of functionally connected urban-agglomeration systems less explored. Administrative boundaries may obscure cross-boundary interactions, while agricultural production decisions in peripheral areas are often influenced by market demands, technological diffusion, and policy incentives originating from nearby urban centers [43]. As comprehensive spatial entities that integrate urban cores with surrounding agricultural zones, urban agglomerations provide a complementary spatial framework for examining the spatial heterogeneity and differentiated characteristics of ACE. Previous research has also highlighted the importance of the urban-agglomeration perspective for examining regional environmental and emission-related issues [44]. However, the specific cross-city flows of agricultural products, labor, and technology are not directly quantified in the present study and are therefore not treated as directly tested mechanisms. Rather, we use the urban-agglomeration perspective as a complementary spatial scale for examining the spatial heterogeneity and differentiated emission characteristics of ACE.

Despite this analytical potential, existing studies have rarely integrated the urban agglomeration perspective with systematic analyses of ACE evolution, spatial heterogeneity, and driving mechanisms, particularly in MGPA where food security and carbon mitigation objectives must be simultaneously considered.

2.3. Research on Driving Mechanisms of ACE

ACE are shaped by the combined effects of economic development, structural transformation, technological progress, and demographic change. Broadly, the influencing factors can be classified into four categories. Economic development factors, typically measured by agricultural output value, are important drivers of ACE growth, as production expansion increases demand for fertilizers, machinery, and irrigation [15,20]. Structural factors refer to changes in cropping patterns and the share of planting versus livestock sectors, which affect emission profiles due to different emission intensities across activities [36,45]. Technological efficiency factors encompass improvements in production efficiency and green technology adoption [23], and are considered among the most important mitigating forces. Demographic factors, particularly rural population changes [46], may influence ACE by affecting agricultural labor availability and production scale.

Various analytical approaches have been employed to investigate the roles of driving factors in ACE changes. Impacts by Regression on Population, Affluence, and Technology (STIRPAT) models [36,47,48] and spatial econometric models [49,50] are widely used to identify statistical relationships and spatial spillover effects, but they are primarily designed to test significance and direction rather than quantify the contribution of individual factors to observed emission changes. By contrast, the Logarithmic Mean Divisia Index (LMDI) method provides a complete additive decomposition without residual terms and allows flexible factor specification [23,51,52]. Because this study focuses on quantifying the contribution of different driving factors to ACE changes over time, LMDI is particularly suitable and was adopted for the analysis.

Taken together, existing studies provide a solid basis for ACE accounting and driver analysis, but less is known about how spatial heterogeneity and emission drivers jointly vary within major grain-producing urban agglomerations. This gap motivates the integrated framework adopted in this study.

3. Materials and Methods

3.1. Analytical Framework

Figure 1 presents the analytical framework of this study. Following a sequential logic, the framework integrates four analytical components: ACE accounting to quantify emission magnitude and source composition; Theil index decomposition to assess overall inequality and its intra- and inter-regional components; spatial autocorrelation analysis to identify clustering patterns; and LMDI decomposition to attribute emission changes to driving factors. By integrating spatial analysis with driving factor decomposition, the framework moves from identifying the “what” and “where” to explaining the “why”, providing an integrated basis for regionally differentiated policy recommendations.

3.2. Study Area

The MGPA in the MRYR consists of 31 prefecture-level cities across Hubei, Hunan, and Jiangxi Provinces (Figure 2). Benefiting from a subtropical monsoon climate, fertile plains, and abundant water resources, the region has developed into one of China‘s most important grain production bases. It encompasses the Jianghan Plain, Dongting Lake Plain, and Poyang Lake Plain and is one of China’s major double-cropping rice production regions. The three plains differ in agricultural production conditions and socioeconomic development. The Jianghan Plain, where the Wuhan Metropolitan Area (WHMA) is located, has experienced relatively intensive urban development while maintaining substantial agricultural production. Double-cropping rice production is particularly important in the Dongting Lake Plain, which is associated with the Changsha–Zhuzhou–Xiangtan (CZT) Urban Agglomeration. The Poyang Lake Plain, where the Poyang Lake Urban Agglomeration (PYLUA) is located, combines extensive rice cultivation with a more diversified agricultural structure. Such differences are likely to be reflected in both the level and composition of ACE. The region thus contains three major urban agglomerations—WHMA, CZT, and PYLUA. Their differences in agricultural production and socioeconomic development provide a useful basis for comparing the spatial heterogeneity of ACE across urban agglomerations.

3.3. Data Sources

Data used in this study were collected from provincial and municipal Statistical Yearbooks, Rural Statistical Yearbooks, and socioeconomic development bulletins of Hubei, Hunan, and Jiangxi Provinces, covering the period 2013–2023. The dataset includes agricultural input consumption (fertilizer, pesticide, diesel, plastic film, and irrigation), crop production (rice, wheat, and maize yields and sown areas), livestock inventories (cattle, pigs, sheep, and poultry), agricultural output value, and rural population. Agricultural output values were converted into constant 2013 prices using provincial agricultural price indices to eliminate the influence of inflation. Detailed data sources are summarized in

Table 1. Data categories, variables, and sources.

Category	Variables	Source
Agricultural inputs	Fertilizer, pesticide, diesel, agricultural plastic film, irrigation area	Provincial and municipal Statistical Yearbook; Hubei, Hunan, and Jiangxi Rural Statistical Yearbook; Statistical Bulletin on National Economic and Social Development in Hubei, Hunan, and Jiangxi Province (2014–2024)
Rice Fields	Rice sown area
Crop production	Rice, wheat, and maize yields
Livestock	Cattle, pigs, sheep, and poultry inventories
Socioeconomic	Gross output value of agriculture, gross output value of farming, forestry, animal husbandry and fishery, rural employed population	Provincial Statistical Yearbook (2014–2024)

.

A small number of missing observations were identified for wheat and maize yields, which are used as activity data for estimating straw burning emissions (SBE). Specifically, 9 wheat yield observations and 8 maize yield observations were missing, accounting for 2.64% and 2.35% of the respective datasets. These missing values were dispersed across several cities and years rather than concentrated in specific regions or periods. Linear interpolation [53] was applied to impute these missing values using each city’s own time-series data. Given the short-term stability of cropping patterns and production capacity, this approach preserves temporal continuity while minimizing potential bias in emission estimation. Detailed information on missing values is provided in Appendix A (

Table A1. Detailed missing observations for crop yield data.

Variable	City	Missing Years	No. of Missing
Wheat yield	Shangrao	2017, 2018, 2019	3
Wheat yield	Fuzhou	2016, 2017, 2018	3
Wheat yield	Zhuzhou	2013, 2014, 2015	3
Maize yield	Yingtan	2013, 2014, 2015	3
Maize yield	Yichun	2017	1
Maize yield	Fuzhou	2017	1
Maize yield	Yiyang	2022	1
Maize yield	Hengyang	2016	1
Maize yield	Changde	2019	1

).

3.4. Agricultural Carbon Emission Accounting

ACE were estimated based on the carbon emission coefficient approach, which has been widely applied in studies of agricultural GHG emissions in China. Following the guidelines of the IPCC, methane (CH₄) and nitrous oxide (N₂O) emissions were converted into carbon dioxide equivalent (CO₂-eq) emissions to ensure comparability among different GHGs.

Considering the characteristics of agricultural production in the MGPA of the MRYR and drawing upon previous studies, ACE were calculated from four major sources: (1) agricultural material inputs, including fertilizers, pesticides, agricultural diesel, plastic films, and irrigation; (2) methane emissions from rice cultivation; (3) carbon emissions generated by crop straw burning; and (4) livestock breeding emissions (LBE) resulting from enteric fermentation and manure management.

Agricultural material emissions (AME) were estimated using the consumption of agricultural production inputs and their corresponding emission coefficients. Paddy field emissions (PFE) mainly consisted of CH₄ emissions generated during rice cultivation. To account for regional differences in rice production conditions, the emission coefficients proposed by Meng et al. [24] were adopted. SBE were estimated based on the production of rice, wheat, and maize and the corresponding emission factors. LBE included emissions from enteric fermentation and manure management of major livestock categories, including cattle, pigs, sheep, and poultry. The emission coefficients used in this study are presented in

Table 2. Carbon emission coefficient of major agricultural materials.

Source	Carbon Emission Factor	Reference
Fertilizer	0.8956 kg C·kg−1	[11]
Pesticides	4.9341 kg C·kg−1
Diesel	0.5927 kg C·kg−1	[54]
Agricultural Plastic Film	5.1800 kg C·kg−1
Irrigation	266.48 kg C·hm−2	[23]

Note: Different emission factors are expressed according to their original activity units.

,

Table 3. Carbon emission coefficient of straw burning.

Source	Carbon Emission Factor (kg C·kg−1)	Reference
Corn	0.93	[24]
Wheat	0.92
Rice	0.93

and

Table 4. Carbon emission coefficient of livestock breeding (unit: kg C·head⁻¹·year⁻¹).

Source	Enteric Fermentation	Fecal Emission	Reference
Cattle	51.00	2.87	[55]
Pig	1.00	4.53
Sheep	5.00	0.49
Poultry	-	0.04

.

The total ACE were calculated as follows:

$$C=\sum E_i=\sum S_i\times {\sigma }_i$$

(1)

where $C$ denotes total ACE, $E_i$ represents carbon emissions of the specific source $i$, $S_i$ represents the activity level of emission source $i$, ${\sigma }_i$ represents carbon emission coefficient of the specific source $i$.

To eliminate the influence of agricultural production scale and better reflect carbon efficiency in agricultural production, agricultural carbon intensity (ACI) was further calculated as:

$$ACI={\displaystyle \frac{C}{AVG}}$$

(2)

where $C$ denotes ACE, $AVG$ denotes gross agricultural production.

3.5. Theil Index

The Theil index [32,33] was applied to quantify spatial inequality in ACE and decompose it into intra- and inter-agglomeration components, thereby revealing their respective contributions to overall disparities.

The formulae are specified as follows:

$$T=T_{wr}+T_{br}$$

(3)

$$T_{wr}={\displaystyle {\sum }_{i=1}^M}\left({\displaystyle \frac{N_i}{N}}\ast {\displaystyle \frac{{\overline{X}}_i}{\overline{X}}}\ast T_i\right)$$

(4)

$$T_{br}={\displaystyle {\sum }_{i=1}^M}\left({\displaystyle \frac{N_i}{N}}\ast {\displaystyle \frac{{\overline{X}}_i}{\overline{X}}}\ast \ln {\displaystyle \frac{{\overline{X}}_i}{\overline{X}}}\right)$$

(5)

$$T_i={\displaystyle \frac{1}{N_i}}{\displaystyle {\sum }_{j=1}^{N_i}}{\displaystyle \frac{X_{ij}}{{\overline{X}}_i}}\ln {\displaystyle \frac{X_{ij}}{{\overline{X}}_i}}$$

(6)

where $T$ denotes the overall disparity in ACE, $T_{wr}$ and $T_{br}$ represent the intra-agglomeration and inter-agglomeration disparity components, respectively; $M$ is the number of urban agglomerations; $N$ is the total number of cities; $\overline{X}$ is the mean ACE across all cities; $N_i$ is the number of cities in the $i$-th urban agglomeration; $T_i$ is the Theil index of the $i$-th urban agglomeration; ${\overline{X}}_i$ is the mean ACE of the $i$-th urban agglomeration; and $X_{ij}$ is the ACE of $j$-th city within the $i$-th urban agglomeration.

3.6. Spatial Autocorrelation Analysis

Spatial autocorrelation analysis [39,56,57] is a method to explore the spatial agglomeration state and agglomeration mode by measuring the spatial correlation of attribute values. This study constructs an economic-distance spatial weight matrix based on gross agricultural production to characterize the spatial interdependencies among cities. Unlike geographic distance matrices, which primarily measure spatial associations based on geographic proximity, the economic-distance matrix further accounts for differences in regional economic development levels and reflects potential economic linkages between cities. Given that agricultural production practices, resource input intensity, and carbon emission characteristics are closely associated with regional economic conditions, the use of the economic-distance matrix can more effectively identify the spatial patterns of ACE within urban agglomerations.

The Global Moran’s I index [58] was adopted here to describe the overall spatial correlation of ACE in the MGPA in MRYR. Then, the Local Moran’s I index was used to reveal the heterogeneity of local spatial agglomeration types of each unit in the study area. The formula for calculating the Global Moran’s I index is as follows:

$$I={\displaystyle \frac{{\sum }_{i=1}^n{\sum }_{j=1}^n{w}_{ij}\left(y_i-\overline{y}\right)\left(y_j-\overline{y}\right)}{m_2{\sum }_{i=1}^n{\sum }_{j=1}^n{w}_{ij}}}$$

(7)

The formula for calculating the Local Moran’s I index is as follows:

$$I_i={\displaystyle \frac{y_i-\overline{y}}{m_2}}{\displaystyle {\sum }_{j=1,j\ne i}^n}{w}_{ij}\left(y_j-\overline{y}\right)$$

(8)

$$m_2={\displaystyle \frac{1}{n}}{\displaystyle {\sum }_{k=1}^n}{\left(y_k-\overline{y}\right)}^2$$

(9)

where $n$ is the number of cities, $y_i$ is ACE of the $i$-th city, $\overline{y}$ denotes the average value of ACE in each city, $w_{ij}$ represents the economic weight matrix based on the total agricultural output value of city $i$ and city $j$, $m_2$ represents the second central moment of ACE.

3.7. Driving Factor Decomposition Using the Extended Kaya Identity and LMDI Method

This research identifies the drivers of ACE and decomposes these emissions based on an extension of the Kaya equation and the LMDI decomposition framework. The Kaya equation describes the relationship between ACE and key drivers such as energy consumption, economic growth, and population, providing a systematic framework for analyzing carbon-related issues. The equation can be expressed as follows:

$$C={\displaystyle \frac{C}{E}}\times {\displaystyle \frac{E}{\mathit{GDP}}}\times {\displaystyle \frac{\mathit{GDP}}{P}}\times P$$

(10)

$$\mathit{CI}={\displaystyle \frac{C}{E}}$$

(11)

$$\mathit{EI}={\displaystyle \frac{E}{\mathit{GDP}}}$$

(12)

$$G={\displaystyle \frac{\mathit{GDP}}{P}}$$

(13)

where $C$ represents carbon emissions; $E$ represents total energy consumption; $\mathit{GDP}$ represents gross domestic product; $P$ represents total population size; $\mathit{CI}$ represents carbon intensity; $\mathit{EI}$ represents energy intensity; $G$ represents economic output per capita.

Based on the findings of a previous study [23,50], this research extended the Kaya identity by incorporating agricultural characteristics to better reflect the mechanisms underlying changes in ACE. Considering that agricultural production is closely related to rural labor and agricultural activities, the population factor in the original Kaya identity was replaced by rural population size. Thus, agricultural emission intensity (PE), agricultural sector structure (AS), agricultural economic development level (EL), and rural population size (N_x) were used to analyze the mechanisms underlying changes in ACE.

$$C=PE\times AS\times EL\times N_x$$

(14)

where $PE={\displaystyle \frac{C}{G_n}}$, $G_n$ denotes the gross value of agriculture; $AS={\displaystyle \frac{G_n}{G_A}}$, $G_A$ denotes the gross value of agriculture, forestry, livestock and fisheries; $EL={\displaystyle \frac{G_A}{N_x}}$, $N_x$ denotes rural population of the region.

LMDI is a decomposition method to measure the influencing factors of ACE. This method can eliminate the residual term, accurately decomposing ACE and clarifying the contribution value of each factor [32,59,60,61]. The LMDI additive decomposition was used to quantify the influence of each factor on carbon emissions, specifically, as follows:

$$\Delta C=C_T-C_0=\Delta PE+\Delta AS+\Delta EL+\Delta N_x$$

(15)

$$\Delta F={\displaystyle \frac{C_T-C_0}{\mathit{ln}{C}_T-\mathit{ln}{C}_0}}\times \mathit{ln}{\displaystyle \frac{F_T}{F_0}}$$

(16)

where $\Delta C$ is the total effect; $T$ denotes the period (t = 1, 2, …, T), and $0$ denotes the base period. $\Delta F$ denotes the contribution of factor $F$ to the change in ACE. If the amount of carbon emission change caused by the factor is positive, the factor has a promoting effect on carbon emissions. If it is a negative value, the factor has an inhibitory effect.

PE represents changes in ACE per unit of agricultural economic output and is measured by ACI. When the PE effect is negative, it indicates a decline in ACE per unit of agricultural economic output, suggesting that improvements in resource-use efficiency and agricultural production efficiency have contributed to mitigating ACE growth.

AS represents the relative contribution of crop production within the broader agricultural sector, reflecting structural differences between crop production and other agricultural sectors. Changes in this structural composition can further influence ACE because different agricultural sectors differ in production processes, input requirements, and emission characteristics.

EL is measured by agricultural economic output per rural population, which reflects the level of agricultural economic development and output capacity per rural population. An increase in agricultural economic productivity may be associated with expanded agricultural production activities and increased production inputs, thereby contributing to ACE growth.

N_x, as a demographic factor, is used to reflect changes in rural population size. Changes in rural population size may affect agricultural production scale, labor supply, and demand for agricultural inputs, thereby influencing ACE.

4. Results

4.1. Spatiotemporal Evolution and Source Composition of ACE

As illustrated in Figure 3, total ACE in the MGPA of the MRYR exhibited a distinct three-stage evolution from 2013 to 2023. From 2013 to 2015, ACE increased from 60.62 million tons to 71.65 million tons. This increase was likely associated with the expansion of grain production and the accompanying growth in agricultural inputs, including fertilizers, diesel, and irrigation, as well as the expansion of rice cultivation area. From 2016 to 2020, ACE declined steadily to 63.88 million tons. This downward trend coincided with the implementation of China’s “Zero Growth in Fertilizer Use by 2020” action plan and agricultural supply-side structural reform, which may have contributed to more efficient use of agricultural inputs and reductions in input-related emissions. From 2021 to 2023, ACE stabilized, with annual growth rates remaining within ±1%. This stabilization may reflect a balance between continued improvements in agricultural production efficiency and the requirement to maintain stable grain production.

In contrast, ACI demonstrated a steady downward trend throughout the entire study period, declining from 1.34 tons per 10,000 CNY in 2013 to 0.84 tons per 10,000 CNY in 2023, representing a cumulative reduction of 37.29% (Figure 3). The continuous decline in ACI suggests that technological progress and efficiency improvements have outpaced the emission-increasing effects of agricultural economic expansion, indicating a gradual decoupling between agricultural economic growth and carbon emissions.

Regarding carbon-source composition (Figure 3), PFE consistently constituted the largest source of ACE, accounting for 37.39% to 47.16% of total emissions over the study period. AME ranked second, while LBE and SBE contributed smaller and declining shares. The contribution of PFE increased from 37.39% in 2013 to 46.71% in 2023, while the shares of AME and LBE declined correspondingly. This increasing share of planting-related emissions indicates that rice cultivation has become an increasingly dominant component of ACE in the study region, whereas the relative contribution of livestock-related emissions has gradually decreased.

4.2. Spatial Heterogeneity and Evolution Patterns of ACE

Figure 4 presents the spatial distribution of ACE and ACI across the 31 cities in 2013 and 2023. In 2023, the highest ACE value (Changde, 10.42 million tons) was 19.4 times the lowest (Ezhou, 0.54 million tons), indicating substantial inter-city disparity. ACI also exhibited notable heterogeneity, ranging from 0.84 tons per 10,000 CNY (Wuhan) to 8.32 tons per 10,000 CNY (Ji’an). Relative to 2013, ACE increased in 16 cities and decreased in 15, while ACI declined in 29 of the 31 cities, with increases observed only in Xiaogan and Hengyang (Figure 4), suggesting a widespread improvement in carbon efficiency across the study region.

Spatially, the distribution of ACE exhibited a pronounced “south-high, north-low” pattern by 2023, with high-emission zones concentrated in the southern portion of the study area and low-emission zones in the northern portion (Figure 5). At the urban agglomeration scale, three differentiated spatial patterns were observed.

The WHMA consistently exhibited a clear core–periphery spatial pattern, where Wuhan and its neighboring cities remained low-emission areas, while peripheral cities were predominantly classified as medium-high or high-emission areas.

The CZT Urban Agglomeration showed a contiguous concentration of high-emission areas, with high-value zones progressively extending from Yueyang toward Changde and Hengyang during the study period.

The PYLUA displayed localized high-emission clusters, mainly concentrated in the Yichun–Ji’an subregion and Shangrao.

Over the study period, the spatial distribution of ACE demonstrated strong persistence, with major high- and low-emission areas remaining relatively stable. Meanwhile, transitional changes occurred in several regions, particularly the expansion of high-emission areas within the CZT Urban Agglomeration and the increasing spatial differentiation between urban cores and peripheral cities in the WHMA.

4.3. Carbon Source Typology

To further explore the underlying differences in emission structures beyond aggregate carbon magnitudes, cities were classified according to the dominant contribution of different carbon sources. Based on the relative shares of AME, PFE, SBE, and LBE in total ACE, five emission typologies were identified: AME-led, PFE-led, SBE-led, AME–PFE-led, and PFE–SBE-led. The classification follows a dominant-source criterion: a city is assigned to a single-source typology if any source contributes 40% or more of its total ACE; otherwise, it is assigned to a dual-source typology comprising the two largest contributing sources. Although LBE was included in the accounting framework, no city met the threshold for an LBE-dominated category, indicating that livestock-related emissions did not constitute the dominant emission structure in the study region. This typology framework moves beyond total emission magnitude to capture the structural composition of emissions, providing a complementary lens for understanding spatial heterogeneity.

The spatial distribution of these typologies across the three urban agglomerations is presented in

Table 5. Spatial distribution of agricultural carbon emission typologies across cities.

City Cluster	ACE Type	Cities
WHMA	AME-led City	Yichang
	PFE-led City	Huanggang, Xiaogan, Xianning, Xiantao, Qianjiang, Tianmen, Huangshi, Jingzhou
	SBE-led City	Xiangyang
	AME-PFE-led City	Wuhan, Ezhou
	PFE-SBE-led City	Jingmen
CZT Urban Agglomeration	PFE-SBE-led City	Changsha, Zhuzhou, Xiangtan, Yueyang, Yiyang, Changde, Hengyang, Loudi
PYLUA	PFE-led City	Jiujiang, Jingdezhen, Yingtan, Xinyu, Shangrao, Fuzhou, Ji’an
	PFE-SBE-led City	Nanchang, Yichun, Pingxiang

Note: Typology classification is based on the proportional contribution of each emission source to total ACE in each city. AME = agricultural material emissions; PFE = paddy field emissions; SBE = straw burning emissions. Detailed estimation procedures and emission coefficients are provided in Section 3.4.

. The classification results reveal substantial differences in emission structures among cities, suggesting that spatially differentiated mitigation strategies are required because ACE heterogeneity is reflected not only in emission intensity but also in emission composition. This structural dimension of heterogeneity has direct implications for mitigation prioritization, as different source types respond to distinct policy interventions.

The spatial distribution of these typologies exhibits pronounced differences both across and within the three urban agglomerations. The WHMA displays the most diversified structure, encompassing four of the five identified typologies. Within this urban agglomeration, Wuhan and Ezhou exhibit an AME–PFE-led pattern, whereas peripheral cities such as Jingzhou and Huanggang show a distinct PFE-led pattern, with PFE contributions approaching or exceeding 50% of total emissions. Two exceptional cases were also identified within the WHMA. Yichang is the only AME-led city in the entire study region, and Xiangyang is the only SBE-led city.

The CZT Urban Agglomeration presents a structurally homogeneous pattern, with all eight cities classified as PFE–SBE-led. This pattern corresponds to the prevalence of rice-based agricultural systems in the Dongting Lake Plain, where PFE and SBE jointly constitute the major emission sources across the agglomeration.

The PYLUA exhibits an intermediate structure, consisting of PFE-led and PFE–SBE-led types. Among these, Ji’an stands out with the highest PFE across the entire study region. The absence of AME-led or SBE-led types indicates that emission structures in the PYLUA were dominated by fewer carbon sources than those observed in the WHMA.

Overall, the typology results reveal that ACE heterogeneity across urban agglomerations is manifested not only in emission magnitude but also in carbon-source composition.

4.4. Regional Inequality and Spatial Autocorrelation Analysis

The Theil index results revealed that the overall disparity in ACE across the MGPA of the MRYR followed a trajectory of initial widening, narrowing, and eventual stabilization over the study period (Figure 6). The overall Theil index increased from 0.21 in 2013 to a peak of 0.23 in 2016, then declined to 0.21 by 2018 and fluctuated within a narrow range of 0.21–0.22 thereafter. Overall, a considerable degree of heterogeneity was maintained throughout the period, while spatial disparities narrowed after 2016.

Further decomposition of the overall inequality (Figure 7) reveals that intra-regional differences consistently formed the principal source of overall disparity, with their contribution rate substantially exceeding that of inter-regional differences. This indicates that intra-agglomeration differences contributed more substantially to overall ACE disparities than inter-agglomeration differences. At the urban agglomeration scale, the long-term Theil index values for the WHMA and PYLUA were consistently higher than those for the CZT Urban Agglomeration, demonstrating a greater degree of intra-regional imbalance in ACE within the former two clusters. Specifically, intra-regional disparities within the WHMA showed a continuously widening trend, while those within the CZT Urban Agglomeration gradually narrowed after peaking in 2016. Intra-regional disparities in the PYLUA remained relatively stable after 2014.

The Theil index results, together with Moran’s I statistics, demonstrate that ACE disparities exhibit significant spatial dependence rather than random distribution. The global Moran’s I results (

Table 6. The global Moran’s I index from 2013 to 2023.

Year	I	Z	p-Value
2013	0.474	3.580	0.001
2014	0.490	3.624	0.001
2015	0.522	4.335	0.001
2016	0.473	4.052	0.001
2017	0.466	3.749	0.001
2018	0.501	4.211	0.001
2019	0.449	3.311	0.001
2020	0.477	4.355	0.001
2021	0.475	3.901	0.001
2022	0.458	3.699	0.001
2023	0.397	3.428	0.001

) indicate that ACE exhibited significant positive spatial autocorrelation throughout the study period, with all values statistically significant at the 1% level. This confirms the existence of significant positive spatial autocorrelation in ACE across the study area. In terms of temporal variation, Moran’s I showed an overall fluctuating downward trend, falling from 0.474 in 2013 to 0.397 in 2023—a cumulative decline of 16.2%—although several short-term fluctuations occurred during the study period. Overall, this indicates that the intensity of clustering weakened over time, despite the persistence of strong positive spatial autocorrelation.

Local spatial autocorrelation results (Figure 8) further reveal distinct clustering patterns across the study area. The high–high (H–H) clusters were primarily concentrated in the PYLUA, particularly in the Yichun–Ji’an subregion and Shangrao, and gradually expanded toward the CZT Urban Agglomeration over time. The Yichun–Ji’an subregion consistently maintained the most stable H–H clustering status throughout the study period. Correspondingly, the low–low (L–L) clusters were mainly concentrated in the central Hubei region and gradually expanded eastward into Jiangxi. In contrast, the low–high (L–H) and high–low (H–L) clusters appeared only sporadically in individual cities and isolated years, without forming sustained spatial associations. Taken together, both high-emission and low-emission clusters exhibited considerable temporal persistence during the study period.

4.5. Driving Forces of ACE Changes

Based on the LMDI decomposition framework, the contributions of four driving factors—PE, AS, EL, and N_x—to ACE changes in the MGPA of the MRYR were quantified for the period 2013–2023.

At the regional level, the cumulative net increase in ACE over the study period was 5.27 million tons. EL was the dominant positive contributor, increasing ACE by 63.38 million tons, while AS generated a relatively small increase of 1.17 million tons. In contrast, PE and N_x exerted substantial mitigation effects, reducing emissions by 30.86 million tons and 28.42 million tons, respectively. Among these factors, PE represented the largest mitigating effect. The positive contribution of EL was largely offset by the combined mitigation effects of PE and N_x.

At the urban agglomeration level, the direction and magnitude of these effects varied considerably (Figure 9). In the WHMA, EL produced the largest positive effect (11.44 million tons), whereas PE represented the strongest mitigation effect (−9.66 million tons). In the CZT Urban Agglomeration, EL (26.85 million tons) and AS (0.91 million tons) contributed to emission increases, while PE (−9.81 million tons) and N_x (−13.66 million tons) generated mitigation effects, with N_x showing the larger contribution. In the PYLUA, EL (25.08 million tons) and AS (0.75 million tons) increased emissions, whereas the combined mitigation effects of PE (−11.38 million tons) and N_x (−14.48 million tons) outweighed these positive effects, resulting in a net reduction of 0.03 million tons. The PYLUA was the only urban agglomeration showing a net decline in ACE during the study period.

5. Discussion

5.1. Agricultural Functional Differentiation and Spatial Heterogeneity of ACE

The spatial heterogeneity of ACE in the MRYR reflects not merely differences in emission magnitude among cities, but differentiated agricultural functions embedded within urban agglomeration systems. This perspective helps explain why cities with similar climatic conditions and agricultural resources may exhibit divergent emission trajectories.

The three urban agglomerations demonstrate distinct functional patterns. The WHMA exhibits a core–periphery structure, where the metropolitan core has gradually shifted toward higher-value non-agricultural activities while peripheral areas continue to undertake intensive agricultural production. The CZT Urban Agglomeration shows persistent high emissions associated with its role as a rice-production base, resulting in relatively homogeneous emission structures dominated by paddy-field and straw-related emissions. The PYLUA presents localized high-emission clusters, reflecting the coexistence of intensive rice cultivation and diversified agricultural activities.

The carbon-source typology further confirms that ACE heterogeneity is closely related to regional agricultural functions. Unlike simple differences in emission levels, variations in emission composition reveal how different urban agglomerations organize agricultural production. Moreover, the dominance of intra-agglomeration disparities in the Theil decomposition suggests that emission inequality is primarily generated within integrated regional systems rather than between separated administrative regions, highlighting the necessity of an urban agglomeration perspective. Core cities and surrounding agricultural areas perform different roles within the same urban agglomeration, creating internal differentiation in emission intensity and production structure.

Therefore, the urban-agglomeration perspective provides an analytical advantage by examining ACE patterns in relation to the functional differentiation between metropolitan cores and agricultural hinterlands. ACE patterns are thus better understood as outcomes of regional functional organization rather than isolated geographical differences. This finding is consistent with previous studies showing that intra-regional differences constitute an important source of spatial inequality in ACE [41]. However, our results further show that such disparities remain pronounced even within urban agglomerations, suggesting that internal agricultural functional differentiation may contribute to ACE heterogeneity.

5.2. Decoupling Pathways and Regional Differences in ACE Reduction

The LMDI results demonstrate that ACE changes were driven by different combinations of demographic, efficiency, structural, and economic effects across urban agglomerations. These differences suggest that agricultural decoupling is not determined by a universal mechanism, but is shaped by region-specific development pathways involving urbanization, agricultural specialization, and functional transformation.

The WHMA represents an efficiency-oriented pathway. As the most urbanized agglomeration, it has experienced substantial agricultural labor transition, creating conditions for mechanization, improved input management, and efficiency enhancement. Accordingly, technological and managerial improvements played a more important role in emission reduction. This pathway indicates that metropolitan areas with stronger economic and technological capacities may achieve decoupling through productivity-oriented transformation [62].

The CZT represents a specialization-constrained pathway. Its persistent rice-based production structure and homogeneous emission composition—with all eight cities classified as PFE–SBE-led—suggest a degree of production lock-in, in which limited diversification constrains structural adjustment opportunities [63]. Under this condition, emission reduction relies more heavily on the rural population effect than on productivity improvement, indicating the need for technological upgrading and production-system transformation. It should be noted, however, that the relatively high SBE contribution in the CZT should be interpreted with caution, as SBE is estimated from crop yields using literature-based emission factors rather than from direct observations of burning practices. Therefore, the estimates should not be interpreted as evidence of widespread open burning. Instead, they indicate the potential importance of straw-related emissions associated with intensive rice production in the Dongting Lake Plain, suggesting that improved straw management remains a potential area for agricultural emission mitigation in this region.

The PYLUA exhibits a transitional pathway characterized by the coexistence of efficiency improvement and demographic adjustment. Its resource advantages have supported intensive agricultural production, but simultaneous progress in management improvement and labor transition created favorable conditions for emission reduction. However, this pathway may become less effective as demographic effects weaken with further urbanization [64].

Together, these findings reveal distinct emission-reduction pathways across the three urban agglomerations. The dominant positive effect of EL and the mitigating effect of PE are broadly consistent with previous LMDI-based studies [65]. However, our results further show that the relative importance of these effects varies markedly among urban agglomerations, reflecting differences in agricultural functions, emission structures, and development conditions. Thus, similar aggregate emission trends may arise from different underlying adjustment processes.

5.3. Toward Pathway-Oriented Low-Carbon Agricultural Transition

Existing ACE research has mainly focused on emission accounting, intensity reduction, and influencing factors at provincial or municipal administrative scales [24,34,35,36,66,67]. This study extends this literature by shifting attention from emission differences among administrative units toward differentiated transition pathways within urban agglomeration systems. The key implication is that ACE heterogeneity reflects not only differences in emission magnitude but also the differentiated functional roles of urban agglomerations within regional agricultural systems. Therefore, low-carbon agricultural transition is pathway-dependent rather than uniform: regions with similar development levels may exhibit distinct emission trajectories due to differences in agricultural functions, urbanization processes, and technological conditions.

This functional-differentiation perspective suggests three major mitigation pathways. First, efficiency-oriented regions, represented by the WHMA, should prioritize technology diffusion and efficiency improvement [68,69]. Rather than relying solely on emission reduction within metropolitan cores, these regions should enhance the transmission of advanced agricultural technologies and management practices to surrounding areas. Second, specialization-oriented regions, such as the CZT Urban Agglomeration characterized by concentrated rice production, should focus on reducing structural emission constraints through methane mitigation technologies, water-saving irrigation [70], and straw resource utilization [26,71]. Third, transitional regions, represented by the PYLUA, should accelerate efficiency-oriented transformation while demographic and structural factors still provide emission reduction potential.

These differentiated pathways demonstrate that effective ACE mitigation requires consideration of both current emission levels and the developmental processes underlying emission formation. Spatially targeted strategies, rather than uniform reduction policies, are therefore essential for coordinating food security objectives with low-carbon agricultural transformation in major grain-producing regions.

6. Conclusions

6.1. Main Findings

Based on ACE accounting, Theil index decomposition, spatial autocorrelation analysis, and LMDI factorization for 31 prefecture-level cities in the MGPA of the MRYR from 2013 to 2023, three main findings emerge.

First, ACE in the study region followed a phased trajectory of growth, decline, and stabilization over the study period. Total emissions increased from 2013 to 2015, then gradually declined before stabilizing after 2020. ACI declined steadily throughout the period, indicating improved carbon efficiency during agricultural development. In terms of carbon-source composition, PFE became an increasingly dominant component, while the relative contribution of livestock-related emissions gradually decreased.

Second, ACE exhibited significant spatial heterogeneity with a pronounced south-high, north-low distribution pattern. The three urban agglomerations displayed distinct emission configurations: the WHMA exhibited a persistent core–periphery structure, the CZT Urban Agglomeration showed a contiguous high-emission pattern, and the PYLUA displayed localized high-value concentration. Global Moran’s I confirmed significant positive spatial autocorrelation throughout the period. Theil index decomposition further revealed that intra-agglomeration disparities, rather than inter-agglomeration differences, were the primary source of overall spatial inequality.

Third, the contribution of driving factors—PE, AS, EL, and N_x—varied substantially across the three urban agglomerations, indicating that emission dynamics are shaped by region-specific agricultural characteristics and development contexts rather than by a uniform mechanism. These differences highlight the importance of considering regional development pathways in designing agricultural carbon mitigation strategies.

6.2. Limitations and Future Research

Several limitations of this study should be acknowledged. First, the ACE accounting relied on emission coefficients drawn primarily from IPCC default values and previous Chinese studies. While these coefficients are widely used and allow for cross-regional comparability, they may introduce uncertainty when applied to region-specific conditions. Future research could refine coefficient estimation using locally measured data.

Second, this study focused on ACE and did not consider the carbon sequestration capacity of farmland ecosystems. Soils, rice paddies, and agroforestry systems can act as significant carbon sinks, and their exclusion may limit a more comprehensive understanding of agricultural climate impacts. Future studies could integrate both emission sources and carbon sinks to provide a fuller accounting of net carbon balance.

Third, the analysis was conducted at the prefecture-level city scale, which does not capture farm-level heterogeneity or household production decisions. Additionally, up-stream emissions associated with the production and supply of agricultural inputs were beyond the system boundary of this study. Incorporating micro-level survey data and expanding system boundaries in future work would help reveal the behavioral foundations of agricultural emission patterns and provide a more complete assessment of agricultural carbon footprints.

Fourth, the analysis is based on prefecture-level panel data for 2013–2023 and therefore mainly captures contemporary patterns and drivers of ACE. Although the observed differences may also reflect longer-term historical trajectories, the available data do not allow their historical effects to be separately identified or quantitatively evaluated. Future research could incorporate longer historical records to further examine the long-term formation of regional differences in ACE.