Spatiotemporal Differentiation of Carbon Emission Efficiency and Influencing Factors in the Five Major Maize Producing Areas of China

Abstract

Understanding the carbon emission efficiency (CEE) of maize production and its determinants is critical to supporting China’s dual-carbon goals and advancing sustainable agriculture. This study employs a super-efficiency slack-based measure model (SBM) to evaluate the CEE of five major maize-producing regions in China from 2001 to 2022. Kernel density estimation and the Dagum Gini coefficient are used to analyze spatiotemporal disparities, while a geographically and temporally weighted regression (GTWR) model explores the underlying drivers. Results indicate that the national average maize CEE was 0.86, exhibiting a “W-shaped” fluctuation with turning points in 2009 and 2016. From 2001 to 2015, the Southwestern Mountainous Region led with an average efficiency of 0.76. Post-2015, the Northern Spring Maize Region emerged as the most efficient area, reaching 0.90. Efficiency levels have generally become more concentrated across regions, though the Southern Hilly and Northwest Irrigated Regions showed higher volatility. Inter-regional differences were the primary source of overall CEE disparity, with an average annual contribution of 46.66%, largely driven by the efficiency gap between the Northwest Irrigated Region and other areas. Spatial heterogeneity was evident in the impact of key factors. Agricultural mechanization, cropping structure, and environmental regulation exhibited region-specific effects. Rural economic development and agricultural fiscal support were positively associated with CEE, while urbanization had a negative correlation. These findings provide a theoretical foundation and policy reference for region-specific emission reduction strategies and the green transition of maize production in China.

1. Introduction

In recent years, global greenhouse gas (GHG) emissions have continued to rise, exacerbating impacts and threats to ecosystems and triggering severe climate and environmental challenges [1]. To address this crisis, the international community has established a series of regulatory frameworks, including the United Nations Framework Convention on Climate Change (UNFCCC) and the Kyoto Protocol, aiming to achieve global carbon emission reductions through coordinated policy measures [2]. Agriculture has been long overshadowed by energy-related issues in the policy and scientific debate surrounding climate change [3]. However, within the global GHG emission structure, agricultural activities contribute approximately 30%, equivalent to 15 billion tons of CO₂ annually [4]. Notably, agricultural systems possess significant potential for emission reduction and carbon sequestration, theoretically capable of offsetting up to 80% of agricultural emissions [5]. As a major developing country with both substantial agricultural production and carbon emissions, China’s agricultural sector plays a crucial role in ensuring food security. However, this comes with considerable GHG emissions, accounting for about 17% of the nation’s total emissions and showing an annual growth rate of 5% [6]. Maize, as China’s primary grain crop, serves as a fundamental resource for food production, industrial raw materials, and livestock feed [7]. With rising living standards and the expansion of animal husbandry, the demand for maize has steadily increased. In 2018, China’s maize output reached 257 million tons, 1.96 times that of wheat and 1.21 times that of rice, highlighting its critical role in food security.

However, the growing reliance on machinery, chemical fertilizers, and pesticides in maize production, coupled with regional disparities in natural conditions, technological levels, and management practices, has led to significant spatial variations in carbon emission efficiency. Maize-related GHG emissions mainly involve three aspects: (1) agricultural material inputs, including N₂O released from chemical fertilizer application, GHGs from pesticides and plastic film degradation, and energy consumption during irrigation; (2) land use changes, which include N₂O caused by soil carbon pool disturbance and CO₂ produced by tillage activities.; (3) energy consumption, mainly CO₂ released from agricultural machinery and diesel use. In response to these challenges, the low-carbon transformation of maize production in China has progressed through three major stages. Before 2015, it was dominated by extensive management, relying heavily on chemical fertilizers, pesticides and traditional agricultural machinery, with a lack of systematic low-carbon policy support. Since 2015, initiatives such as the zero-growth action plan for fertilizers and pesticides have promoted low-carbon technologies including soil testing and energy-efficient machinery. Following the proposal of the “Dual Carbon” strategy in 2020, low-carbon transformation accelerated, with region-specific policies and wider adoption of practices like water-saving irrigation and straw incorporation aimed at achieving coordinated development in yield, efficiency, and emissions reduction. Therefore, it is imperative to examine the carbon emission efficiency of maize production from a regional heterogeneity perspective, exploring strategies to reduce emissions and enhance carbon sequestration while ensuring stable yield increases and economic benefits.

Existing research on agricultural carbon emissions has primarily focused on three key areas: quantification and efficiency measurement, identification of influencing factors, and the exploration of underlying mechanisms. Investigation spans multi-scale assessments—from national and regional levels down to provinces, cities, and even counties [8,9,10,11,12,13]—covering the agriculture sector overall, crop cultivation, and animal husbandry [14,15]. Detailed estimations have also been conducted for various crops, including rice and soybeans [16,17], which have helped establish a relatively clear picture of the spatial patterns and status of agricultural carbon emissions in China. In terms of efficiency measurement tools, recent studies applied the Data Envelopment Analysis (DEA) model to evaluate agricultural carbon emission efficiency [18,19,20], while others employed SBM models that account for undesirable outputs, enabling more accurate efficiency assessments and an analysis of spatiotemporal evolution. Regarding influencing mechanisms, commonly used methodologies include the Tobit model [21,22,23], the Logarithmic Mean Divisia Index (LMDI) [24,25], and spatial econometric models [26,27]. Researchers have constructed comprehensive indicator systems incorporating variables such as agricultural mechanization, industrial structure, digital economy, environmental regulation, urbanization level, and public agricultural support [28,29,30,31,32,33]. However, due to limitations in data quality and model precision, a consensus on the key driving forces has yet to be established.

In summary, significant progress has been made in the study of agricultural carbon emissions, providing valuable insights for understanding the development of low-carbon agriculture. Nevertheless, several important gaps remain. First, most existing studies focus on agriculture as a whole or on broad sub-sectors, with limited attention given to crop-specific carbon emissions, particularly for maize, despite its importance. Second, current efficiency evaluation frameworks tend to emphasize agriculture’s role as a carbon source while neglecting its potential as a carbon sink, leading to measurements that lack both accuracy and comprehensiveness. Third, while many studies examine the spatiotemporal evolution of emission efficiency, few have systematically analyzed the underlying sources of regional disparities, which hurts efforts to develop coordinated carbon reduction strategies among regions. Finally, most research on influencing mechanisms has treated the effects of various factors as uniform across space and time, without adequately considering spatial heterogeneity in their impacts. This limits the effectiveness of region-specific, differentiated mitigation policies.

This study presents a comprehensive evaluation of carbon emission efficiency in China’s maize production by integrating carbon sequestration into the accounting framework while systematically categorizing carbon sources. Employing an SBM model with undesirable outputs, we quantify carbon emission efficiency across China’s five major maize-producing regions. Kernel density estimation is applied to construct continuous distribution curves, revealing the temporal evolution of efficiency and intra-regional agglomeration patterns, thereby precisely showing the status of low-carbon development in each production zone. To dissect regional disparities, we utilize the Dagum Gini coefficient decomposition method, partitioning overall differences into three components: intra-regional, inter-regional, and hypervariable density contributions. This approach elucidates the structural mechanisms underlying spatial heterogeneity and identifies key shortcomings in low-carbon maize production, offering quantitative insights to overcome bottlenecks in regional collaborative development. Furthermore, a geographically and temporally weighted regression (GTWR) model is adopted to investigate the spatiotemporal heterogeneity of influencing mechanisms from three dimensions: industrial, social, and governmental. This analysis systematically identifies constraints and advantages in low-carbon maize production, culminating in targeted policy recommendations. Our findings provide a theoretical foundation for enhancing carbon emission efficiency, fostering inter-regional carbon reduction synergies and facilitating the transition toward sustainable maize production.

2. Materials and Methods

2.1. Overview of the Study Area

The five major maize producing areas in China each have unique production conditions and development characteristics (Figure 1). The northern spring sowing maize areas include Heilongjiang, Jilin, Liaoning, Inner Mongolia, and Shaanxi, which have a temperate monsoon climate with a frost-free period of 120–160 days and an annual precipitation of 400–600 mm. The plains are vast and the black soil is fertile, with high levels of mechanization and scale. However, low-temperature damage occurs from time to time and the problem of thinning of the black soil layer gradually becomes apparent. The summer maize planting area in the Huang-Huai-Hai Region covers provinces such as Hebei, Shandong, Henan, and Shanxi. Under a temperate semi-humid climate, the frost-free period is 180–220 days, with an annual precipitation of 500–800 mm. The terrain is mainly plain, and the agricultural technology level is high. The yield is among the highest, but there are issues such as overexploitation of groundwater, soil compaction, and high incidences of pests and diseases. The Southwest Mountainous Maize Region involves provinces and regions such as Sichuan, Yunnan, Guizhou, and Chongqing. The subtropical humid climate brings sufficient precipitation, but the proportion of mountainous hills is high, the vertical climate difference is large, the lighting conditions are poor, small-scale farming is dominant, the degree of mechanization is low, and soil erosion and pest control are prominent challenges. The Southern Hilly Maize Region includes provinces such as Anhui, Jiangsu, Hubei, and Guangxi, with a warm and humid subtropical monsoon climate and abundant precipitation. However, the hilly terrain leads to scattered farmland, poor water and fertilizer retention of red soil, small-scale maize cultivation, and significant impacts of labor outflow. The Northwest Irrigated Maize Region includes Xinjiang, Gansu Hexi Corridor, and Ningxia. Under the temperate continental climate, there is sufficient sunshine and large temperature differences between day and night. Although the annual precipitation is low, it relies on irrigation to maintain production. The level and scale and mechanization are high, and water scarcity, soil salinization, and ecological fragility are issues that need attention. In 2022, the total planting area and yield of maize in the five major production areas accounted for 97.82% and 97.94% of the national total, which can well represent the production situation of maize in China.

2.2. Research Method

2.2.1. Super-Efficiency SBM Model

The super-efficient SBM model is more effective and has stronger resolution than the Charne–Cooper–Rhodes model (CCR) and Banker–Charnes–Cooper model (BCC), and traditional SBM models have limitations in handling decision-making unit (DMU) ranking problems that involve undesirable outputs and require efficiency scores greater than 1. Therefore, this study employs the non-oriented super-efficiency SBM model under constant returns to scale to evaluate the carbon emission efficiency of maize production in different regions. The super-efficiency SBM model under constant returns to scale is formulated as follows:

$$\eta =\min {\displaystyle \frac{\frac{1}{m}{\displaystyle \sum _{i=1}^m\frac{{\overline{x}}_i}{x_{ik}}}}{\frac{1}{z_1+z_2}\left({\displaystyle \sum _{r=1}^{z_1}\frac{{\overline{y}}_r^g}{y_{rk}^g}+{\displaystyle \sum _{r=1}^{z_2}\frac{{\overline{y}}_r^b}{y_{rk}^b}}}\right)}}$$

(1)

$$s.t.\left\{\begin{array}{l}\overline{x}\ge {\displaystyle \sum _{j=1,j\ne k}^n{x}_j{\lambda }_j}\\ {\overline{y}}^g\le {\displaystyle \sum _{j=1,j\ne k}^n{y}_j^g{\lambda }_j}\\ {\overline{y}}^b\ge {\displaystyle \sum _{j=1,j\ne k}^n{y}_j^b{\lambda }_j}\\ \overline{x}\ge x_0,{\overline{y}}^g\le y_0^g,{\overline{y}}^b\ge y_0^b,\lambda \ge 0\end{array}\right.$$

(2)

where η denotes the carbon emission efficiency of maize production and η ≥ 1 indicates that the production unit has reached the optimal production frontier. Here, n represents the total number of DMUs, and k corresponds to the k-th DMU. Each DMU utilizes m types of inputs x, produces $z_1$ types of desirable outputs $y^g$, and generates $z_2$ types of undesirable outputs $y^b$. The slack variables for inputs, desirable outputs, and undesirable outputs are denoted as $\overline{x}$, ${\overline{y}}^g$, and ${\overline{y}}^b$, respectively. The weight coefficient λ reflects the contribution of each DMU to the efficiency benchmark.

2.2.2. Kernel Density Estimation

Kernel density estimation (KDE) is a widely adopted nonparametric approach for investigating temporal dynamic characteristics, as it requires no predefined functional form and can effectively represent variable distributions through continuous curves [34,35]. Given these advantages, we employ KDE to characterize the temporal evolution of maize carbon emission efficiency across China’s five major production regions. The fundamental equation is expressed as

$$f(x)={\displaystyle \frac{1}{Gh}}{\displaystyle \sum _{i=1}^{n}B\left(\frac{x_i-\overline{x}}{L}\right)}$$

(3)

where $f\left(x\right)$ denotes the kernel density function, G represents the sample size, L is the bandwidth parameter, $B( )$ is the kernel function, and $x_i$ indicates the efficiency value of the i-th DMU.

2.2.3. Dagum Gini Coefficient Decomposition

Dagum (1997) [36] proposed a refined approach to decomposing the overall Gini coefficient, enabling the identification of inequality sources from multiple dimensions. Specifically, the total Gini index can be decomposed into three components: intra-regional inequality, inter-regional inequality, and the hypervariable density. This method not only fully accounts for the distributional characteristics of each subgroup but also incorporates overlapping and cross-distributions among different subsamples, allowing for a more accurate and nuanced reflection of inequality sources [36]. The decomposition formula is given as follows:

$$G={\displaystyle \sum _{j=1}^K{\displaystyle \sum _{h=1}^K{\displaystyle \sum _{i=1}^{n_j}{\displaystyle \sum _{r=1}^{n_k}\left|{\rho }_{ji}-{\rho }_{hr}\right|}}}}/2n^2\overline{y}$$

(4)

Specifically, ${\rho }_{ji}$ and ${\rho }_{hr}$ denote the maize carbon emission efficiency of the j-th province within region i and the h-th province within region r, respectively. Parameter n indicates the total number of provinces contained in the respective regions.

2.2.4. Geographically and Temporally Weighted Regression (GTWR) Model

Conventional panel regression approaches (e.g., OLS and Tobit models) can only estimate the average effects of explanatory variables across the entire sample, failing to capture spatiotemporal heterogeneity. To overcome this limitation of traditional global models in capturing localized spatiotemporal variation, the academic community has developed various methods for analyzing spatiotemporal differences, mainly including Bayesian spatiotemporally varying coefficients models and Markov transition models, etc. [37,38]. While each of these approaches has its own advantages, they often fall short in effectively capturing complex and dynamic spatiotemporal heterogeneity, particularly when both spatial and temporal non-stationarity are present. To address this issue, the GTWR model enables localized parameter estimation that varies continuously over space and time and allows for a more nuanced analysis of spatiotemporal processes. Therefore, GTWR provides a more flexible and robust framework for exploring spatiotemporal heterogeneity compared to traditional methods [39,40]. Accordingly, this study employs the GTWR model to examine the spatiotemporal heterogeneity of influencing mechanisms. The basic formulation is as follows:

$$y_j={\beta }_0\left(u_j,v_j,t_j\right)+{\displaystyle {\sum }_k{\beta }_k\left(u_j,v_j,t_j\right)x_{jk}+{\omega }_j}$$

(5)

where $y_j$ and $x_{jk}$ represent the dependent variable and explanatory variable, respectively (with variance inflation factors (VIF) ≤ 10 for all predictors); $u_{j, }{v}_{j, }{t}_{j, }$ denote the longitude, latitude, and temporal coordinates of observation j, defining its spatiotemporal location; ${\beta }_0$ is the intercept term; ${\beta }_k$ represents the regression coefficient for the k-th explanatory variable at location j; and $w_j$ is the residual term.

2.3. Indicator System

2.3.1. Indicator System for Measuring Maize Carbon Emission Efficiency

In this study, carbon emissions from corn production were measured in carbon dioxide equivalents (CO₂-eq), reflecting greenhouse gas emissions from major agricultural inputs and production operations. Drawing on existing research [10,17], an input–output index system for evaluating the carbon emission efficiency of corn production was constructed (see

Table 1. Index system of carbon emission efficiency of maize.

First Indicators	Second Indicators	Unit
Input Indicators	Fertilizer usage	104 t
	Usage of agricultural film	104 t
	Pesticide usage	104 t
	Seed usage	104 t
	Sowing area	104 hm2
	Number of working days	104 day
	Mechanical power	104 kW
	Diesel consumption	104 t
Output Indicators	Maize production, Carbon sink capacity	104 t
	Maize output value	104 yuan
	Carbon emissions	104 t

). The input includes four types of elements: agricultural material inputs, including fertilizer usage, agricultural film usage, pesticide usage, and seed usage. The intensity of the use of these production resources directly affects carbon emission efficiency, and their carbon footprint comes from manufacturing, transportation, and field application processes. Increasing usage will promote carbon emission growth, land input is characterized by the planting area of corn, and the disturbance effect of planting activities on soil carbon pool is related to carbon sequestration capacity. Reasonable utilization can maintain carbon sink stability. Labor input is measured by number of labor days in corn production, and the scale of labor operations and business models affect resource utilization efficiency. The level of skills and production methods affect emission redundancy through the degree of technology adoption. Energy input includes mechanical power and diesel consumption, both of which reflect the degree of dependence of agricultural production on fossil fuels. Energy utilization efficiency is influenced by the promotion of energy-saving technologies and structural optimization and is directly related to total carbon emissions.

Desirable outputs include maize output value, physical yield, and carbon sequestration, while the undesirable output is the total carbon emissions generated during the production process. The estimation of carbon emissions follows the emission coefficient method [9,41]. Emissions are calculated from three main components: agricultural material usage, land utilization, and energy consumption. For agricultural inputs, carbon emissions are attributed to fertilizer, pesticide, and plastic film use, as well as irrigation. Emission factors are adopted from the Oak Ridge National Laboratory (ORNL, Oak Ridge, TN, USA) and Deng R [42]. Land use-related emissions include N₂O emissions from soil carbon stock disturbances caused by maize cultivation, and CO₂ emissions from irrigation and tillage practices. The corresponding emission coefficients are based on IPCC guidelines. Energy-related emissions are mainly derived from diesel consumption and machinery operation during maize production, with emission coefficients referenced from Wang [43].

The carbon sources and emission factors involved in maize carbon emissions refer to

Table 2. Indicator system for carbon emission accounting for maize.

Dimension	Carbon Source	Emission Factor	Reference Source
Material input	Nitrogenous fertilizer	1.53 kg/kg	China Life Cycle Database (CLCD)
	Phosphate fertilizer	1.63 kg/kg
	Potash fertilizer	0.65 kg/kg
	Compound fertilizer	1.77 kg/kg
	Seed	1.84 kg/kg	Ecoinvent 2.2
	Agricultural film	5.180 kg/t	Institute of Agricultural Resources and Ecological Environment, Nanjing Agricultural University
	Pesticide	4.934 kg/kg	Oak Ridge National Laboratory (ORNL)
Energy input	Agricultural machinery	0.18 kg/kW	IPCC
	Diesel oil	0.5927 kg/kg
Land use	Soil damage	2.532 kg N2O/hm2
	Irrigation	266.48 kg/hm2
	Land plowing	312.6 kg/hm2	College of Agriculture and Biotechnology, China Agricultural University

. All GHG emissions—including CO₂, CH₄, and N₂O—are converted into CO₂-equivalent units to facilitate aggregation. The conversion coefficients used are CO₂:CH₄:N₂O = 1:25:298 (based on IPCC AR4), and the molecular weight conversion for carbon to CO₂ is 44/12. Accordingly, the total carbon emissions are estimated using the emission coefficient method as follows:

$$Ac={\displaystyle \sum A{c}_{\partial t}}={\displaystyle \sum C_t·{\lambda }_t}$$

(6)

where $Ac$ denotes total carbon emissions (kg CO₂-eq), $t$ is three types of carbon emission sources, $C_t$ represents the utilization quantity of the t-th carbon sources, and ${\lambda }_t$ is the corresponding emission factor for the carbon emission source.

Current maize carbon sequestration technologies have achieved phased breakthroughs, mainly including sustainable and regenerative technological practices such as straw biochar carbon sequestration technology, intelligent monitoring and carbon sequestration measurement technology, and conservation tillage with straw returning technology. As a result, the potential of farmland soil carbon sequestration and maize carbon sequestration has been significantly enhanced. Based on this, it is necessary to incorporate them into the expected outputs when measuring carbon emission efficiency to improve the accuracy of efficiency measurements. The crop carbon sink in expected outputs refers to atmospheric carbon fixation through photosynthesis during plant growth [44,45]. Following Li Kerang et al.’s methodology [46], we estimate maize carbon sequestration during its growth period by integrating the economic coefficient and carbon absorption rate. The specific formulation is

$$C_s={\displaystyle \sum C·D}={\displaystyle \sum \left(C·Y/H\right)}$$

(7)

where $C_s$ represents the carbon sequestration during the maize growth period (kg C/ha); $C$ indicates the carbon absorption rate, i.e., carbon absorbed per unit of organic matter synthesized by maize (kg C/kg biomass); $D$ is the total biological carbon content of maize (kg biomass/ha), calculated as economic yield divided by the economic coefficient; and H refers to the maize economic yield Y (kg/ha). Following empirical studies, the economic coefficient H, water content, and carbon absorption rate for maize are set at 0.4, 13%, and 0.471, respectively.

2.3.2. Indicator System for Influencing Factors

To systematically examine the drivers of maize carbon emission efficiency, this study, by comprehensively considering the actual situation of maize production, data availability, and existing research, constructs an evaluation framework encompassing three dimensions: industrial, social, and governmental factors (see

Table 3. Indicator system of factors affecting carbon emission efficiency of maize.

Type	Dimension	Variable Explanation	Expected Effect
Industry level	Fertilizer application level	Net amount of fertilizer applied	−
	Agricultural mechanization	Power of maize production machinery	−
	Crop planting structure	Proportion of maize planting area	+
Social level	Economic development level	Per capita disposable income	+
	Educational level	Years of education per capita	+
	Urbanization level	Proportion of urban population	−
Government level	Agricultural finance	Agricultural, forestry and water affairs expenditures	+
	Environmental regulation	Environmental protection expenditure	+

). At the industrial level, the impact of fertilizer use mainly comes from the carbon emissions generated during its production, transport, and field application, and excessive use can lead to unnecessary emissions. The level of agricultural mechanization is related to the intensity of fossil energy consumption, and the energy efficiency and usage rate of equipment directly affect the total carbon emissions [47]. The crop planting structure helps match resources more efficiently among crops, improving land and input use, which in turn influences emission levels. At the social level, rural economic development provides basic support for low-carbon technology investment and facility improvement, indirectly improving efficiency. The education level of farmers affects their understanding and ability to apply low-carbon methods, influencing how precisely agricultural activities are managed. Urbanization changes how land and resources are distributed between cities and rural areas, which indirectly affects how inputs are used in farming. At the governmental level, two indicators are used. Firstly, fiscal support for agriculture, measured by government expenditures on agriculture, forestry, and water affairs. Agricultural subsidies help lower the cost barrier for adopting low-carbon technologies. Secondly, environmental regulation intensity, measured by government spending on environmental protection. Environmental regulations encourage producers to take emission-reducing actions by setting rules that limit unnecessary pollution.

To reduce problems such as biased estimates, incorrect significance of coefficients, and unclear results that may be caused by multicollinearity, this study first tested the selected factors before running the geographically and temporally weighted regression (GTWR) model. The variance inflation factor (VIF) was used to check for multicollinearity. The results showed that the VIF values for all variables were below the standard threshold of 10, showing that multicollinearity was not a serious problem in the model.

2.4. Data Sources

The data on carbon sources, input–output variables, and influencing factors were obtained from official statistical yearbooks, including the National Compilation of Cost-Benefit Data for Agricultural Products, the China Rural Statistical Yearbook, and the China Statistical Yearbook. For variables such as total machinery power, pesticide and diesel usage, and effective irrigated areas in maize production, we followed the methodology of Liu et al. [48], which disaggregates agricultural totals based on maize’s proportion of total crop planting area. To eliminate price fluctuations, all monetary values were adjusted to constant 2001 prices. Missing data were imputed using Bayesian models where applicable.

3. Results

3.1. Temporal Characteristics of CEE in Maize Production

Based on measurements using the MAXDEA (R2022a) software (see Figure 2), the average carbon emission efficiency of maize production in China during the study period was 0.86, indicating an efficiency loss of approximately 14%. The temporal trend reveals a distinct “W-shaped” fluctuation, with 2009 and 2016 serving as critical turning points. Meanwhile, total carbon emissions exhibited a general upward trajectory despite fluctuations. The highest efficiency value was recorded in 2016 at 1.05, whereas the lowest was observed in 2009 at 0.68. The phased characteristics presented in the research results are basically consistent with the trends described in the 2025 China Agricultural and Rural Low-Carbon Development Report released by the Ministry of Agriculture and Rural Affairs. The report points out that from 2016 to 2022, China was in a key stage of shifting from traditional farming methods to more sustainable practices. In particular, the year 2015 marked a turning point in the intensity of greenhouse gas (GHG) emissions from crop production. Between 2015 and 2020, the GHG emission intensity of crop production fell by 16%. The use of technologies such as water-saving irrigation, straw return, precision fertilization, crop rotation, and intercropping has helped reduce emissions during farming activities and improve soil’s ability to store carbon.

3.1.1. Phase I: 2001–2009—Rapid Decline in Efficiency with Rising Emissions

During this initial phase, carbon emission efficiency declined markedly from 1.00 to 0.68, with a negative slope of −0.04. Concurrently, carbon emissions increased significantly from 44.58 million tons to 59.53 million tons, corresponding to an average annual growth rate of 14.84%. This period was characterized by an extensive, input-driven production model, where increased inputs failed to translate into proportional output gains, resulting in declining efficiency and escalating emissions—a typical feature of unsustainable, high-emission agricultural development.

3.1.2. Phase II: 2010–2016—Transition Toward Low-Carbon Production

In the second stage, maize carbon emission efficiency improved significantly, rising from 0.73 in 2010 to a peak of 1.05 in 2016, with an average annual growth rate of 3.75%. Carbon emissions also continued to grow, albeit at a slower pace, increasing from 62.97 million tons to 73.83 million tons (2.46% annually). This phase reflects a transitional period during which low-carbon technologies began to be adopted, contributing to gains in efficiency. However, limited experience in sustainable farming practices constrained the immediate effectiveness of emission control.

3.1.3. Phase III: 2017–2022—Post-Transition Volatility and Recovery

From 2017 to 2022, the carbon emission efficiency exhibited a “U-shaped” trajectory. Efficiency experienced a sharp decline in 2017–2018 relative to the 2016 peak, largely attributed to setbacks in the adoption of low-carbon practices as some farmers reverted to conventional methods amid transitional challenges. However, from 2019 onward, efficiency gradually rebounded, rising from 0.88 to 1.05 by 2022, with an average annual growth rate of 4.83%, suggesting growing adaptation to low-carbon agricultural practices and technological learning. Notably, carbon emissions during this phase remained relatively stable, fluctuating around 85 million tons, indicating preliminary success in decoupling emissions from production growth.

Figure 2. Evolution of carbon emissions and efficiency of maize in China from 2001 to 2022.

Figure 2. Evolution of carbon emissions and efficiency of maize in China from 2001 to 2022.

3.2. Kernel Density Estimation Analysis of Regional CEE

Figure 3 illustrates the kernel density estimation curves of maize carbon emission efficiency across China’s five major production regions. Over the sample period, the primary peaks of the distributions have noticeably shifted to the right, indicating a general upward trend in carbon emission efficiency and suggesting that low-carbon transition in maize production has begun to yield positive results. This improvement is closely associated with the iterative upgrading of agricultural technologies and the optimization of resource allocation driven by policy interventions.

3.2.1. Distributional Shifts: From Multimodal to Unimodal Structures

The distributional morphology reveals a transition from bimodal or multimodal patterns toward unimodal forms in all five regions, indicating increasing intra-regional convergence in efficiency levels. This shift reflects the gradual dismantling of barriers to low-carbon technology diffusion and the refinement of supporting policy mechanisms. As a result, heterogeneous production units within regions are converging toward the mean efficiency level. Notably, the Northern Spring Maize Region, the Huang-Huai-Hai Summer Maize Region, and the Southwest Mountainous Region exhibit increases in peak density, while the Southern Hilly Region and the Northwest Irrigated Region still exhibit relatively dispersed efficiency distributions.

3.2.2. Distribution Spread: Right Tail and Extreme Values

In terms of tail characteristics, right-skewed (positive skewness) distributions are observed in the Northwest Irrigated Region, the Southern Hilly Region, and the Southwest Mountainous Region, indicating the presence of outliers with exceptionally high carbon emission efficiency. Conversely, the Northern Spring Maize Region and the Huang-Huai-Hai Region exhibit more compact and symmetric distributions with minimal presence of outliers. This suggests relatively homogeneous efficiency performance. The differences in the above distribution characteristics mainly stem from imbalances in technology popularization, variations in policy coordination effectiveness, and differentiation in resource endowments across regions. Regions with right-skewed distributions may have gaps in low-carbon technology application, where some production units take the lead in achieving efficiency leaps, thus forming extreme high values. Regions with concentrated distributions, due to sufficient technology diffusion, unified policy implementation, and balanced resource allocation, show convergence in efficiency levels with less differentiation.

3.3. Spatial Pattern Analysis of Maize CEE

To examine the spatial distribution characteristics of maize carbon emission efficiency across China’s major production regions, a quantile-based classification method was employed. Efficiency levels were categorized into four tiers: super-efficiency (>1), high efficiency (≤1 and >0.70), moderate efficiency (≤0.70 and >0.35), and low efficiency (≤0.35), as illustrated in Figure 4.

In 2001, the overall carbon emission efficiency of maize production was relatively low across regions. The Northern Spring Maize Region predominantly exhibited moderate to low efficiency, forming a typical “moderate–low” spatial cluster. The other four major maize-producing regions also showed limited high-efficiency zones, with a clear tendency toward “moderate–high” clustering. Notably, prior to 2015, the Southwest Mountainous Region demonstrated relatively high carbon emission efficiency, leading the national transition toward low-carbon maize production. By 2022, the spatial landscape had experienced significant structural changes, with all regions achieving varying degrees of improvement in carbon emission efficiency. The Northern Spring Maize Region achieved high and even super-efficiency levels in low-carbon maize development, becoming a national leader. In contrast, the Southwest Mountainous Region showed limited improvement, with its efficiency advantage gradually declining.

3.4. Decomposition Analysis of CEE Differences

3.4.1. Within-Region Disparities in CEE

To assess intra-regional disparities in maize carbon emission efficiency, the Dagum Gini coefficient was applied to the five major maize-producing regions (see

Table 4. Differences in Dagum Gini coefficient within a region.

Year	Northern Spring Maize	Huang-Huai-Hai Summer Maize	Southern Hilly Maize	Northwest Irrigated Maize	Southwest Mountain Maize
2001	0.108	0.092	0.092	0.092	0.092
2004	0.095	0.130	0.130	0.130	0.130
2007	0.100	0.083	0.083	0.083	0.083
2010	0.045	0.186	0.186	0.186	0.186
2013	0.058	0.118	0.118	0.118	0.118
2016	0.068	0.101	0.101	0.101	0.101
2019	0.061	0.108	0.108	0.108	0.108
2022	0.073	0.091	0.091	0.091	0.091
average	0.088	0.073	0.110	0.101	0.136

). Results indicate that the Southwest Mountainous Region exhibited the highest within-group disparity, with an average Gini coefficient of 0.136—substantially above the national average of 0.102. This was followed by the Southern Hilly Region (average = 0.110), while the Huang-Huai-Hai Summer Maize Region showed the lowest level of internal disparity (average = 0.073). Overall, intra-regional differences contributed relatively little to total disparity, with an average annual contribution of less than 20%.

3.4.2. Between-Region Disparities in CEE

To further explore regional heterogeneity, inter-regional disparities were also evaluated using Dagum inter-group Gini coefficients (see

Table 5. Differences in Dagum Gini coefficient between regions.

Main Producing Area	2001	2004	2007	2010	2013	2016	2019	2022
Spring—Summer	0.145	0.135	0.120	0.092	0.071	0.059	0.067	0.073
Spring—Hills	0.124	0.118	0.122	0.096	0.104	0.088	0.108	0.111
Spring—Irrigation	0.191	0.134	0.278	0.129	0.104	0.078	0.115	0.155
Spring—Mountain	0.116	0.119	0.141	0.111	0.097	0.107	0.092	0.120
Summer—Hills	0.172	0.138	0.124	0.064	0.113	0.096	0.119	0.121
Summer—Irrigation	0.256	0.182	0.294	0.104	0.150	0.101	0.133	0.162
Summer—Mountain	0.203	0.144	0.160	0.093	0.126	0.115	0.110	0.130
Hills—Irrigation	0.186	0.116	0.264	0.102	0.148	0.118	0.150	0.157
Hills—Mountain	0.125	0.110	0.125	0.091	0.132	0.123	0.130	0.136
Irrigation—Mountain	0.163	0.117	0.220	0.112	0.107	0.124	0.133	0.151

). The results show a declining trend in spatial inequality. The Gini coefficient fell from 0.0934 in 2001 to 0.0594 in 2022, indicating that regional differences in maize carbon emission efficiency have gradually narrowed over time. Among all region pairs, the Northwest Irrigated Maize Region consistently exhibited the largest efficiency gap compared to the other regions and remains the dominant contributor to inter-regional disparities. In contrast, the Northern Spring Maize Region and the Huang-Huai-Hai Region showed relatively minor differences and demonstrated a clear trend of convergence. The significant efficiency gap between the Northwest Irrigated Region and other regions is shaped by the combined effects of multiple factors. In terms of natural endowments, the arid climate in this region leads to insufficient water resources, making agricultural production highly reliant on irrigation systems. However, the inadequate improvement of irrigation facilities and limited application of water-saving technologies exacerbate inefficient water use. Meanwhile, land ecological conditions restrict production stability, widening the efficiency gap. At the technical level, the insufficient regional diffusion of low-carbon agricultural technologies results in significant efficiency differentiation between traditional production models and the application of advanced technologies, highlighting systemic efficiency gaps compared with other regions where technology popularization is more balanced. In terms of policies and operational conditions, weak agricultural infrastructure, insufficient targeting and intensity of policy support, coupled with a low level of large-scale operation, limit the capacity of production entities to adopt and apply advanced technologies, making it difficult to form an overall driving force for efficiency improvement, thus leading to the persistence of efficiency differences.

3.4.3. Overall Disparity and Decomposition of Maize CEE

Using the Dagum Gini coefficient decomposition framework, the overall disparity in maize carbon emission efficiency was decomposed into three components: within-region disparity, net between-region disparity, and the intensity of overlapping distributions across regions. The contribution of each component over the sample period is presented in

Table 6. Dagum’s overall Gini coefficient and decomposition.

Year	Gini Coefficient				Contribution Rate (%)
	Overall	Within-Region	Inter-Region	Super-Variable Density	Within-Region	Inter-Region	Super-Variable Density
2001	0.1593	0.0236	0.0934	0.0423	14.84	58.61	26.56
2004	0.1254	0.0206	0.0599	0.0448	16.47	47.81	35.72
2007	0.1788	0.0277	0.1150	0.0360	15.51	64.33	20.15
2010	0.0935	0.0162	0.0347	0.0426	17.30	37.12	45.59
2013	0.1054	0.0153	0.0595	0.0306	14.51	56.47	29.02
2016	0.0950	0.0162	0.0295	0.0492	17.10	31.08	51.82
2019	0.1078	0.0182	0.0435	0.0462	16.86	40.32	42.83
2022	0.1230	0.0203	0.0594	0.0432	16.55	48.31	35.14

. Between 2001 and 2022, the overall Gini coefficient declined from 0.1593 to 0.1230, with an average annual reduction rate of 3.51%, indicating a gradual narrowing of carbon efficiency disparities at the national level. On average, net between-region disparity accounted for 46.66% of total inequality, making it the primary driver of overall efficiency differences. However, its contribution declined from 58.61% in 2001 to 48.31% in 2022, suggesting a weakening dominance of regional gaps in determining national carbon efficiency variance. The contribution of transvariation density exhibited a U-shaped trend—rising from 26.56% in 2001 to a peak of 51.82% in 2016, before falling to 35.14% in 2022. The contribution of within-region disparity remained relatively stable throughout the sample period, fluctuating within a narrow band of 14.51% to 17.30%. This indicates a limited influence of intra-regional variation on national-level efficiency inequality, consistent with earlier findings on regional homogenization driven by centralized environmental governance and standardized low-carbon policies.

3.5. Analysis of Impact Mechanisms

3.5.1. Temporal Heterogeneity Analysis of Impact Mechanisms

At the industrial level, the influence of agricultural mechanization on maize carbon emission efficiency shows a wide range from −0.25 to 0. After 2012, the abnormal value increased, and the overall trend was fluctuating (Figure 5a), which is the core factor affecting maize carbon emission efficiency. In the early stage of the development of mechanization, the production mode dominated by traditional high energy consumption agricultural machinery, due to the low energy conversion rate of power equipment the energy consumption in the operation process is concentrated, directly increasing total carbon emissions and forming a significant carbon effect. With the promotion of energy-saving agricultural machinery and intelligent optimization of the operation process and precise adjustment of mechanical utilization intensity in the later stage, the energy consumption intensity of unit output is significantly reduced and its negative impact is gradually weakened. The trend in efficiency improvement is related to the iteration of agricultural machinery technology and the improvement of management refinement. The influence coefficient of crop planting structure fluctuated between −0.005 and 0.015, showing a positive impact in most years. Its mechanism is not only reflected in the optimization of resource matching by adjusting the planting proportion of maize and other crops but also reflected in the coordinated improvement of land productivity and carbon emission capacity through reasonable layout. When planting structure tends to be intensive and regional, it can reduce resource waste and carbon emission redundancy caused by scattered plots. When the structure is unbalanced, it may weaken the positive effect of soil carbon pool disturbance caused by single crop continuous cropping. The overall change is closely related to policy guidance in agricultural distribution and the dynamic balance of market supply and demand regulation. At the social level, rural economic development showed a significant positive impact in most years (Figure 5c), with a coefficient range of −0.0002 to 0.0006 and weak dispersion and significance. After 2012, it converged to a positive value close to zero, which has not yet constituted a core influencing factor. Rural economic development has a positive impact through technical investment and facilities improvement and converges to a positive value close to zero with the weakening of marginal effect. During the study period, the negative impact of urbanization gradually weakened (Figure 5d) and the coefficient was between −0.04 and 0. In the early stage of urbanization, it will squeeze agricultural production space and resources, resulting in negative effects. Later, with the promotion of the coordinated development of urban and rural areas, the negative impact gradually weakened. At the government level, the effectiveness of environmental regulation showed a significant dispersion before convergence after 2010, with a coefficient range of −0.01 to 0.005 (Figure 5e). In the early stage of environmental regulation, due to serious differentiation in policy implementation intensity, its effect shows discrete characteristics, and in the late stage it tends to converge with the improvement in the system, but on the whole the significance of environmental regulation is weak and it is not the core influencing factor in maize carbon emission efficiency. The impact of fiscal agricultural support (Figure 5f) is similar to the mode of environmental regulation, with a coefficient between −0.002 and 0.003, which is weak and tends to converge after 2008. Due to the limitation in intensity and investment direction, its positive role is relatively limited.

3.5.2. Spatial Heterogeneity Analysis of Impact Mechanisms

The GTWR results show that the impact of agricultural mechanization varies from weak positive correlation to strong negative correlation, with coefficients distributed between −0.2134 and 0.0007, showing significant regional differences (Figure 6a). Among them, the positive correlation between agricultural mechanization and maize carbon emission efficiency is evident in the spring maize region of Northeast China, the summer maize region of Huang-Huai-Hai, and the hilly maize region of Southern China, indicating that agricultural mechanization is the core positive influencing factor for low-carbon maize production in these areas. In other regions there is a negative correlation. The impact of crop planting structure on maize carbon emission efficiency in various regions reflects the influence of agricultural mechanization, with a coefficient range of −0.04 to 0.007 (Figure 6b). The Huang-Huai-Hai Summer Maize Region and the Southern Hilly Maize Region show significant positive correlations in Figure 6b, indicating that crop planting structure is the core factor for low-carbon maize production in these two regions, while other regions show the opposite trend. Rural economic development (Figure 6c) and fiscal agricultural support (Figure 6d) have overall improved the carbon emission efficiency of maize in various regions, although the impact is mostly weak (coefficient: −0.0001–0.0011) and not statistically significant. Among them, the Huang-Huai-Hai Summer Maize Region is most positively affected by rural economic development in Figure 6c; The Southern Hilly Maize Region is most affected by financial support, as shown in Figure 6d, which is an important supporting factor for the low-carbon transformation of maize in the region. Environmental regulation shows significant spatial heterogeneity, with a clear negative band (coefficient: −0.0018–0.0033) in the Central Region (Figure 6e). The Northeast Spring Maize Region and the Southern Hilly Maize Region show significant positive regulatory effects in Figure 6e, with environmental regulation becoming the core regulatory factor for low-carbon maize production in these two regions, while other regions show negative correlations. Urbanization has had a consistent negative impact on the carbon emission efficiency of maize (coefficient: −0.02–−0.004) (Figure 6f). From the spatial distribution in Figure 6f, urbanization is the core constraint factor faced by low-carbon maize production in various regions.

4. Discussion

Under the “Dual Carbon” policy framework, research on maize carbon emissions plays an important role in advancing the green transformation of agriculture. The findings of this study, particularly regarding the current status, spatiotemporal distribution, and driving mechanisms of maize carbon emission efficiency, are largely consistent with the existing literature on agricultural carbon emissions. However, several notable differences emerge that warrant further discussion.

In terms of methodology, this study employs the non-oriented super-efficiency SBM model under constant returns to scale (CRS) to estimate the carbon emission efficiency of maize production, which may differ slightly from existing research. A possible explanation is that the core of the non-oriented super-efficiency SBM model lies in optimizing both inputs and outputs together—it aims not only to reduce input redundancy but also to increase desirable outputs and decrease undesirable outputs, which aligns well with the reality of maize production and the core objectives of this study. Regarding returns to scale, after long-term development the five major maize-producing regions in China have reached a relatively mature stage of large-scale production, showing a convergent trend overall. In most regions, factor allocation has become more balanced, and the relationship between scale and output has become relatively fixed. Additionally, this study aims to analyze the spatiotemporal differences and driving mechanisms of carbon emission efficiency. The assumption of constant returns to scale (CRS) allows us to directly calculate overall efficiency values, making it easier to compare different regions and to explore the reasons for regional differences. In contrast, using variable returns to scale (VRS) breaks efficiency into parts, which may distract from the main issues. Additionally, the data in this study come from official statistical yearbooks covering many years with stable production levels. The CRS model does not need to separate scale efficiency, so it handles data fluctuations better and produces more reliable results.

Zhao et al. [49] suggested that the overall agricultural sector in China reached its carbon emission peak in 2016. In comparison, Huan et al. [50] also showed that China’s agriculture achieved carbon peaking in 2015. Our results indicate that maize production has yet to achieve carbon peaking and still possesses significant potential for improvement in emission efficiency. This discrepancy can be attributed to the sectoral heterogeneity within agriculture; different sub-sectors vary in their sensitivity to external policy and market signals, creating a complementary dynamic that enables overall carbon peaking at the macro level. In contrast, maize production exhibits relative stability and inertia, with slower responsiveness to systemic changes. As a result, the peaking of carbon emissions in maize production is likely to occur significantly later than in the broader agricultural sector.

Our analysis reveals that neither chemical fertilizer application nor education level has a statistically significant impact on maize carbon emission efficiency—this diverges from prior studies. For example, Xu et al. [51] and Rehma et al. [52] found a significant negative relationship between fertilizer use and agricultural carbon emissions, while education level was positively correlated with efficiency. Conversely, Chen et al. [16] reported a negative impact of education level on soybean carbon efficiency. The lack of a significant relationship between chemical fertilizer use and maize carbon emission efficiency in this study can be explained by the specific characteristics of the research context and subjects. In this study, “chemical fertilizer use” specifically refers to the pure chemical fertilizer input in the maize production process, which is separated from the total agricultural use through the proportion of planting area and is more in line with the actual input situation in crop production. However, previous studies have taken the total chemical fertilizer use in the entire agricultural industry as an indicator, which may include inputs for other crops or non-productive purposes. Such differences in the connotation of variables may weaken correlations. We conducted a comparative re-evaluation of previous studies and found that many prior works adopted total input-based measures or assumed linear marginal productivity across regions. Our study instead focuses on efficiency-oriented outcomes, emphasizing carbon emission intensity per unit of output. We also note that, in recent years, major maize production regions in China have undergone significant input optimization due to policy interventions (e.g., fertilizer reduction and efficiency improvement programs), which may have reduced the additional effect of fertilizer use in the later part of our study, this temporal variation helps explain the divergence in findings. Similarly, for the education level variable, the apparent inconsistency with the earlier literature may stem from shifts in the role of education in agricultural technology adoption. In earlier periods, education served as a primary enabler of access to information and innovation. However, in recent years, widespread extension services, digital platforms, and standardized technical support may have reduced the differentiating impact of education alone, especially in the context of large-scale agricultural policy implementation. Moreover, our sample consists largely of regions where smallholder operations dominate, and the interplay between education and practical adoption behavior may be more nuanced than captured by a single linear term. In this study, “education level” is measured by the average years of education of rural laborers. However, most maize growers have a junior or senior high school education, and their educational background is more related to basic production skills, with a weak direct connection to the adoption of low-carbon technologies. In previous studies, education level often included experiences of professional agricultural technical training, and such targeted education has a more significant effect on improving carbon efficiency. In regions with complex terrain such as the Southwest Mountainous Area, farmers’ production relies more on traditional experience, which limits the role of education level in transmitting low-carbon policies. In contrast, the role of education level is more likely to be manifested in large-scale planting areas, which also leads to differences between this study and previous research.

The findings also underscore the relatively low CEE in the Northwest Irrigated Maize Region and the Southwest Mountainous Region. These inefficiencies likely arise from constraints related to topography and underdeveloped infrastructure, which may in turn impair the optimal spatial allocation of maize production and impede the effective implementation of low-carbon policies. Moving forward, these regions should be prioritized as focal zones for coordinated low-carbon development. Key interventions may include (1) development of small-scale, terrain-adapted agricultural machinery suitable for long-distance transportation and mountainous terrain; (2) enhanced fiscal support for agriculture, including targeted subsidies for low-carbon technologies; and (3) integrated strategies to overcome economic and geographic barriers to sustainable intensification.

It is worth noting that some influencing factors in the GTWR model of this paper show weak significance, which mainly stems from the following two reasons: first, the limitation of data characteristics. Most of the influencing factors involved in the study are variables with long-term effects, and their short-term marginal effects are inherently weak, which may weaken the impact intensity of the real fluctuations of variables on carbon efficiency. Second, the limitation of model adaptability. Although GTWR can capture spatiotemporal heterogeneity, agricultural carbon efficiency is affected by multiple hidden factors, and the existing indicator system fails to fully cover the factors of micro subjects, resulting in limited explanatory power of observable variables.

Although this study provides a relatively comprehensive depiction of the spatiotemporal evolution and influencing mechanisms of carbon emission efficiency in China’s major maize-producing regions, it still has the following limitations:

(1) Due to factors such as data availability and technical constraints, the study relies on fixed coefficients commonly used in previous research when calculating carbon sink amounts without incorporating localized variables such as soil organic carbon content, differences in biomass among maize varieties, and straw returning. This may lead to deviations in estimation and affect the accuracy of expected outputs.

(2) Restricted by the sample size and the smallholder production model of maize, the study did not achieve satisfactory results in terms of spatial spillover effects, so no in-depth analysis was conducted on this aspect in the paper.

(3) The sensitivity and robustness tests of the super-efficient SBM model and GTWR model are still insufficient, and there is still room for improvement in the credibility of the research results.

In future research, more in-depth analyses should be carried out in the following aspects: (1) Establish a cross-regional long-term monitoring network in collaboration with agricultural research institutions, conduct fixed-point experiments to systematically collect data, and at the same time strengthen technical reserves and cooperation to improve the application capabilities of remote sensing and GIS technologies, improve relevant databases, complete the calibration of dynamic model parameters, and realize the accurate spatiotemporal accounting of carbon sink amounts. (2) In the future, the sample can be expanded to the county level, the observation period can be extended, and cross-regional interaction data can be included. Optimize the spatial weight matrix in combination with the characteristics of smallholder management to build an adaptive model and capture the spillover signals after policy coordination to improve the identification accuracy of spatial spillover effects. (3) In order to better address the lack of specialized sensitivity or robustness analysis for core models such as ultra-efficient SBM and GTWR models in research. Future research can further address this deficiency by replacing the core model of sensitivity analysis and conducting robustness testing by adjusting key variables to verify the stability of the influencing mechanism. In addition, potential hidden associations, such as the association between technology adoption and education level, can be verified by adding interaction terms between variables or nonlinear effects to better ensure the credibility of the article’s results.

5. Conclusions

This study employs a comprehensive methodological framework—including the super-efficiency SBM model, kernel density estimation, Dagum Gini coefficient, and spatiotemporal geographically temporally weighted regression (GTWR) model—to investigate the temporal evolution, spatial disparities, and driving factors of carbon emission efficiency in maize production across China and its five major producing regions. The key findings are as follows:

(1) From 2001 to 2020, China’s average carbon emission efficiency in maize production was 0.86, exhibiting a non-linear “W-shaped” trajectory with stronger regional convergence over time, as indicated by kernel density analysis.

(2) The Southwest Mountainous Region initially led in efficiency but was overtaken by the Northern Spring Maize Region due to scale and technology advantages. Inter-regional disparities were the main source of inefficiency, though the overall Dagum Gini coefficient declined. The Northwest Irrigated Region remains a critical target for improvement.

(3) Over time, cropping structure and economic development showed increasing positive effects, while the negative impacts of mechanization and urbanization weakened. Spatially, driver effects varied; mechanization and fiscal support were more effective in northern and hilly regions, while urbanization remained a consistent negative factor. These findings emphasize significant spatiotemporal heterogeneity and the need for region-specific low-carbon strategies.

To enhance carbon emission efficiency in maize production, this paper puts forward the following policy suggestions: (1) Implement differentiated regional carbon reduction measures, regionally differentiated approaches should be implemented based on local resource endowments and development conditions: In the Southwest Region, efforts should focus on talent cultivation and capacity building. This includes strengthening collaboration with agricultural universities and research institutions and providing targeted training in mountain-specific cultivation techniques and small-scale mechanization. The Northern Region should prioritize the research and development of new-energy agricultural machinery, aiming to enhance efficiency and reduce carbon intensity through cleaner energy use. For the Northwest Region, it is essential to advance high-efficiency, water-saving irrigation technologies. Emphasis should be placed on developing and deploying precision drip irrigation and intelligent water control systems suited for arid environments. The Huang-Huai-Hai Region should accelerate the construction of high-standard farmland infrastructure, focusing on the integration and optimization of irrigation and drainage systems to improve resource-use efficiency. In the Southern Hilly Region, agricultural development should be adapted to the complex terrain. This involves building small-scale terraced irrigation facilities and improving rural transportation infrastructure to address logistical constraints and ensure efficient market access. (2) Strengthen carbon sequestration practices. It is necessary to integrate technologies such as straw biochar production and conservation tillage, build an intelligent monitoring and measurement system, and improve the precision and coverage of technology application. At the same time, standardize accounting standards, incorporate carbon sequestration effects into the evaluation of maize carbon emission efficiency, enhance the carbon sequestration potential of farmland by optimizing tillage and planting structures, and form a full-chain improvement mechanism. (3) Optimize public intervention strategies. Public intervention should be based on regional differences, formulate targeted policies according to resource endowments, and precisely allocate emission reduction and carbon sequestration resources. Increase financial input to support low-carbon technologies such as new energy agricultural machinery and water-saving irrigation, strengthen industry–university research collaboration to cultivate talents, improve environmental regulations, incorporate carbon emission efficiency into evaluation systems, and promote government–enterprise collaborative green transformation.