Measurement, Dynamic Evolution, and Spatial Convergence of the Efficiency of the Green and Low-Carbon Utilization of Cultivated Land Under the Goal of Food and Ecological “Double Security”: Empirical Evidence from the Huaihe River Ecological Economic Belt of China

Abstract

Under the “double security” goal of achieving both food security and ecological protection, this study explores the green and low-carbon utilization efficiency of cultivated land (GLCUECL) in the Huaihe River Ecological Economic Belt (HREEB). This study identifies the spatiotemporal evolution characteristics and trends, promoting the green, low-carbon, and sustainable utilization of arable land resources in the HREEB, thus contributing to regional and national food and ecological security. Using a global super-efficiency EBM framework that accounts for undesirable outputs, as well as the GML index, the researchers measured and decomposed the GLCUECL in 25 prefecture-level cities of the HREEB from 2005 to 2021. The Theil index and kernel density estimation were applied to analyze regional disparities and changing developmental traits. Spatial convergence and divergence were assessed using the coefficient of variation and spatial convergence models. Key findings include the following: (1) Over time, the GLCUECL in the HREEB exhibited an overall upward trend and a non-equilibrium characteristic, namely the “East Sea-river-lake Linkage Area (ESLA) > Midwest Inland Rising Area (MIRA) > Huaihe River Ecological Economic Belt (HREEB) > North Huaihai Economic Zone (NHEZ)”. The increase in the GML index of the GLCUECL is mainly attributable to a technical progress change. (2) The overall difference in the GLCUECL tends to decline, which is mainly attributable to the intra-regional differences. (3) The overall kernel density curves for the HREEB and its three sub-regions exhibited a “rightward shift” trend. Except for the expansion and polarization of the absolute difference in the GLCUECL in the NHEZ, the absolute difference in GLCUECL in other regions, such as the HREEB, ESLA, and MIRA, exhibited a decreasing trend. (4) Spatial convergence analysis revealed that only the NHEZ lacks σ-convergence, whereas all regions exhibited β-convergence. Moreover, factors such as rural economic development level, cultivated land resource endowment, agricultural subsidy policy, crop planting structure, and technological input exerted a heterogeneous effect on the change in the GLCUECL. Based on these findings, this study offers recommendations for improving GLCUECL in the HREEB. Our recommendations include the implementation of the concept of green new development, optimization of the institution supply, establishing a regional cooperation mechanism for green and low-carbon utilization of cultivated land, and formulation of differentiated paths for improving the green and low-carbon utilization efficiency of cultivated land according to local conditions.

1. Introduction

Farmland is a critical resource for agricultural production and plays a vital role in safeguarding both food and environmental stability [1,2,3]. The availability of grain is directly linked to national peace and stability: “Having grain in hand ensures a peaceful mind.” As of 2021, China has achieved “eighteen consecutive bumper harvests”, with the output consistently stabilizing above 650 million metric tons for seven years. However, the agricultural sector faces significant structural challenges, particularly the overuse of agricultural inputs and an excessive focus on cultivation at the expense of conservation. These practices have led to soil degradation, greenhouse gas emissions, and widespread agricultural pollution, compromising the country’s food and ecological security [4,5]. Food and ecological security are national priorities [6], and transitioning to greener, low-carbon methods on land is essential for achieving these objectives [7]. In 2021, the Ministry of Agriculture and Rural Affairs, along with five other national departments, issued the “14th Five-Year National Agricultural Green Development Plan,” which outlined strategies to strengthen agricultural resource protection and promote sustainable development. The Huaihe River Ecological Economic Belt (HREEB) is a major grain-producing region and a crucial ecological barrier in China [8]. Therefore, assessing the GLCUECL in the HREEB and understanding its spatiotemporal dynamics and convergence patterns is crucial for ensuring the long-term sustainability of arable land and supporting the region’s food and ecological security.

Since the 18th National Congress of the Communist Party of China (CPC) in 2012, there has been a growing emphasis on environmental sustainability and advancement of efficient agricultural practices, significantly broadening research on farmland utilization efficiency. Scholars have increasingly examined the environmental impacts of cultivated land use, integrating them into broader agricultural efficiency frameworks. Many studies adopt the “input–output” model, which includes both desirable and undesirable outputs to construct efficiency evaluation systems. For instance, Liang Liutao [9] utilized agricultural pollution emissions as “non-intended” output in the evaluation framework of agricultural land production efficiency, and Lu Xinhai et al. [10] factored in carbon emissions from cultivated land use as undesired output in efficiency measurement models. Feng Yonggang et al. [11] and Wen Gaohui et al. [12] considered non-point source pollution and introduced carbon emission factors to assess farmland productivity, respectively, leading to the “greening” measurement and “low-carbonization” investigation of cultivated land use [7,13]. These studies emphasize that farmland functions as both a carbon emitter and a carbon sink [14,15]; thus, focusing solely on agricultural pollutants and greenhouse gas emissions without considering the land’s carbon sink function leads to a biased assessment of land use efficiency under food and ecological security constraints. In response, some scholars have expanded the measurement system by incorporating the carbon sink as a desirable output, alongside non-point source pollution and carbon emissions as detrimental outputs, defining the resultant index as GLCUECL [7,13,16,17,18]. For efficiency measurement methods, scholars mainly utilized the super-efficiency SBM model [16,19,20], the global reference super-efficiency SBM model of undesirable output [21], the EBM model [22], the global DEA-EBM model [13], the EBM-GML model [23], and the super-efficiency EBM model [24] for estimation and calculation. To examine the spatiotemporal characteristics in cultivated land efficiency, scholars have employed tools such as spatial autocorrelation, kernel density estimation, the Dagum Gini coefficient, spatial convergence models, geographical detectors, the global differentiation index (GDI), geographically weighted regression (GWR), geographically and temporally weighted regression (GTWR), dynamic panel system GMM estimation, and Tobit regression [10,19,25,26,27,28]. Research has often focused on specific regions or economic zones [29] or distinctive economic zones with unique geographical attributes. Examples include the Yangtze River Economic Belt [3], the major grain-producing areas [30], the karst area in southwest China [31], the black soil regions of northeast China [32], and the Yellow River Basin [33].

In summary, while there is substantial literature on cultivated land use efficiency, several gaps remain: (1) Much of the research is focused on carbon emissions, diffuse pollution, and the aspiration for maximum carbon emissions and carbon neutrality, often overlooking the integration of green and low-carbon attributes. (2) Most studies focus on a single perspective—either green or low-carbon land utilization—neglecting a comprehensive assessment of the efficiency of green low-carbon utilization. (3) Traditional radial DEA and non-radial SBM models are commonly used, but the EBM model, capable of integrating both radial and non-radial distance functions, offers a more robust approach. However, there is a dearth of literature applying the EBM approach to evaluate the GLCUECL. (4) While cultivated land use efficiency has been examined across geographic regions, few studies have specifically addressed the GLCUECL in the HREEB.

This study addresses these gaps by constructing a holistic index system for measuring GLCUECL, which includes the carbon sink of cultivated land, crop output, diffuse agricultural pollution, and carbon discharge. The global super-efficiency EBM model, which accounts for undesirable outputs, is applied to measure GLCUECL in the HREEB and its three sub-regions, namely the ESLA, NHEZ, and MIRA. This study further introduces the GML index method to analyze dynamic changes in efficiency levels and their sources. Additionally, nonparametric kernel density estimation and the Theil index are used to examine regional differences and sources of GLCUECL, while σ- and β-convergence models are employed to explore trends in spatial convergence–divergence patterns and influencing factors.

2. Study Method and Data Description

2.1. Overview of the Study Area

The Huaihe River Ecological Economic Belt (HREEB) is located in mid-eastern China, encompassing cities across Jiangsu, Shandong, Anhui, Henan, and Hubei provinces (30°93′–36°13′ N,111°55′–120°45′ E) (Figure 1). Encompassing a planned area of 243,000 square kilometers, the region is a vital grain production zone and features extensive plains. It is situated in a cross-climatic region between northern and southern China, with a stable ecosystem, rich biodiversity, and strategic importance for ecological conservation. As of 2021, the region’s Gross Domestic Product attained CNY 9.18 trillion, with CNY 1.02 trillion derived from agriculture, accounting for 13.05% of the country’s gross agricultural production. The crop sown area is approximately 23,770.31 thousand hectares, and the grain sown area accounts for approximately 77.16% of the agricultural production area. The grain output is 116 million tons, contributing 17.04% of China’s grain output, highlighting its importance in ensuring both food availability and ecological security. However, rapid economic and social growth has introduced challenges such as rising carbon emissions, diffuse pollution sources, and declining soil fertility. These issues hinder sustainable land use and compromise long-term ecological revitalization, posing risks to both food and ecological security. The strategic plan for the Huaihe River Ecological Economic Belt underscores the importance of consolidating agricultural production capacity, enhancing food security, controlling agricultural non-point source pollution, and establishing experimental zones for sustainable agricultural development. This study uses cities within the HREEB as case studies to explore the spatiotemporal evolution and future development trends of the GLCUECL, offering theoretical support for maintaining both regional grain and ecological security.

2.2. Research Methods

2.2.1. Global Super-Efficiency EBM Framework Considering Undesirable Output

Data Envelopment Analysis (DEA) is a commonly utilized approach for measuring efficiency, particularly in assessing agricultural land use. Traditional DEA frameworks include radial models, mainly the CCR and BCC models, and non-radial models, primarily the SBM model. Although these models exhibit strong explanatory power in assessing efficiency, they have certain limitations, such as the inability to account for slack variables and retain the novel proportional structure of the evaluated frontier [34]. To address these shortcomings, Tone and Tsutsui [35] introduced the EBM (Epsilon-based Measure) model, a hybrid distance function that combines the characteristics of both radial and non-radial models. To enable the ranking and intertemporal comparison of efficient decision-making units, the super-efficiency method introduced by Andersen et al. [36] was integrated with the EBM model. This resulted in a super-efficient EBM model that incorporates unwanted by-products, such as greenhouse carbon emissions and non-point source pollution, which are typically present in agricultural production processes [37,38]. Pastor et al. [39] proposed the concept of global benchmarking, which involves constructing a global production frontier using data from the entire sample period to replace traditional period-specific frontiers. This approach ensures the feasibility and consistency of intertemporal comparisons and avoids the incomparability problems that arise when traditional models’ frontiers shift over time.

This study utilizes cities and counties in the HREEB as the decision-making units, assuming multiple $\mathrm{t}$ periods and $n$ decision-making units (DMUs), with each DMU having $m$ types of inputs $x_{i0}(i=\mathrm{1,2},\dots ,m)$, $s$ types of desirable outputs $y_{r0}(r=\mathrm{1,2},\dots ,s)$, and $p$ types of undesirable outputs $b_{q0}(q=\mathrm{1,2},\dots ,p)$. Furthermore, following Oh’s approach [40], this study constructs the global technology production possibility set as follows:

$$P=\left\{\begin{array}{c}\left(\overline{x},\overline{y},\overline{b}\right)|{\sum }_{t=1}^T{\sum }_{j=1,j\ne 0}^n{x}_j^t{\gamma }_j^t\le {\overline{x}}^t;{\sum }_{t=1}^T{\sum }_{j=1,j\ne 0}^n{y}_j^t{\gamma }_j^t\ge {\overline{y}}^t;\\ {\sum }_{t=1}^T{\sum }_{j=1,j\ne 0}^n{b}_j^t{\gamma }_j^t\le {\overline{b}}^t;{\sum }_{t=1}^T{\sum }_{j=1,j\ne 0}^n{\gamma }_j^t=1;\gamma \ge 0\end{array}\right\}$$

(1)

where $P$ is the set of production possibilities; $x_j$, $y_j$, and $b_j$ represent inputs, favorable outputs, and unfavorable outputs, respectively; and $\gamma$ corresponds to the weight variable assigned to each decision-making unit when establishing the production possibility set, $\left(\overline{x},\overline{y,}\overline{b}\right)$, which serves as the optimal solution for the production possibility set. Building on existing research, this study employs a super-efficiency EBM model with constant returns to scale and a non-oriented framework to assess the GLCUECL in the HREEB. Based on the above production possibility set, we constructed the global reference super-efficiency EBM model, incorporating undesirable outputs as follows:

$${\eta }^{*}=\underset{\theta ,\phi ,\gamma ,s^{+},s^{-}}{\min }{\displaystyle \frac{\theta +{\epsilon }_x{\sum }_{i=1}^m\frac{w_i^{-}{s}_i^{-}}{x_{i0}}}{\phi -{\epsilon }_y{\sum }_{r=1}^s\frac{w_r^{+}{s}_r^{+}}{y_{r0}}-{\epsilon }_b{\sum }_{q=1}^p\frac{w_q^{b^{-}}{s}_q^{b^{-}}}{b_{q0}}}}$$

(2)

$$s.t.\left\{\begin{array}{c}{\sum }_{t=1}^T{\sum }_{j=1,j\ne 0}^n{x}_{ij}^t{\gamma }_j^t-s_i^{-}\le \theta x_{i0},i=\mathrm{1,2},\dots ,m\\ {\sum }_{t=1}^T{\sum }_{j=1,j\ne 0}^n{y}_{rj}^t{\gamma }_j^t+s_r^{+}\ge \phi y_{r0},r=\mathrm{1,2},\dots ,s\\ {\sum }_{t=1}^T{\sum }_{j=1,j\ne 0}^n{b}_{qj}^t{\gamma }_j^t-s_q^{b^{-}}\le b_{q0},q=\mathrm{1,2},\dots ,p\\ {\sum }_{t=1}^T{\sum }_{j=1,j\ne 0}^n{r}_j^t=1\\ \gamma \ge 0;s_i^{-}\ge 0;s_r^{+}\ge 0;s_q^{b^{-}}\ge 0\end{array}\right.$$

In the preceding equation, ${\eta }^{*}$ denotes the model’s optimal efficiency value, whereas $w_i^{-}$, $w_r^{+}$, and $w_q^{b^{-}}$ denote the weights assigned to inputs, desirable outputs, and undesirable outputs. Additionally, $s_i^{-}$, $s_r^{+}$, and $s_q^{b^{-}}$ denote the non-zero slack measures for inputs, desirable outputs, and undesirable outputs. The parameters $\theta$ and $\phi$ denote the radial condition efficiency values. The parameter $\epsilon$ denotes the degree of combination between radial and non-radial models, ranging from [0,1]; when $\epsilon =0$, it corresponds to a radial framework, and when $\epsilon =1$, it corresponds to an SBM model. The term ${\eta }^{*}<1$ denotes efficiency loss in a decision-making unit, whereas ${\eta }^{*}\ge 1$ indicates that the decision-making unit is efficient [41,42].

2.2.2. GML Metric Method

To examine the evolving efficiency patterns of each decision-making unit, this research applies the Global Malmquist–Luenberger (GML) index method, based on the measurement results from the global super-efficiency EBM framework. The GML index effectively addresses issues such as unsolvability and insufficient transitivity in traditional Malmquist–Luenberger (ML) index linear programming. It offers advantages such as transitivity, cyclic accumulation, and the ability to conduct intertemporal comparisons [34,43]. As outlined in Oh’s study [40], the GML index is expressed as follows:

$$\begin{gathered} {GML}^{t,t+1}\left(x^{t+1},y^{t+1},b^{t+1},x^t,y^t,b^t\right)={\displaystyle \frac{E\left(x^{t+1},y^{t+1},b^{t+1}\right)}{E\left(x^t,y^t,b^t\right)}} \\ ={\displaystyle \frac{E^{t+1}\left(x^{t+1},y^{t+1},b^{t+1}\right)}{E^t\left(x^t,y^t,b^t\right)}}\left[{\displaystyle \frac{E\left(x^{t+1},y^{t+1},b^{t+1}\right)}{E^{t+1}\left(x^{t+1},y^{t+1},b^{t+1}\right)}}\times {\displaystyle \frac{E^t\left(x^t,y^t,b^t\right)}{E\left(x^t,y^t,b^t\right)}}\right] \\ ={GEC}^{t,t+1}\times {GTC}^{t,t+1} \end{gathered}$$

(3)

In this formula, $E\left(*\right)$ denotes the comprehensive efficiency value for green and low-carbon land use of the decision-making unit under global reference. $x^t$, $y^t$, and $b^t$, along with $x^{t+1}$, $y^{t+1}$, and $b^{t+1}$, denote the inputs, beneficial outputs, and adverse outputs of the evaluated unit during periods $t$ and $t+1$, respectively. Additionally, the index can be further decomposed into ${GEC}^{t,t+1}$ and ${GTC}^{t,t+1}$, representing the shift in technical efficiency and technological progress, respectively. A value of ${GML}^{t,t+1}>1$ indicates an improvement in GLCUECL, while ${GML}^{t,t+1}$ signals a decline in efficiency.

2.2.3. Theil Index and Its Decomposition

The Theil index, initially used to measure income inequality, is now widely employed to analyze regional disparities in various contexts. The index varies between 0 and 1, where smaller values indicate less disparity between regions. Based on the studies by Theil [44] and Zheng et al. [45], this study adapts the Theil index to assess the regional differentiation in GLCUECL within the HREEB. The equations are presented as follows:

$$Theil={\displaystyle \frac{1}{n}}\sum _{i=1}^n\left({\displaystyle \frac{{Te}_{it}}{{AveTe}_t}}ln{\displaystyle \frac{{Te}_{it}}{{AveTe}_t}}\right)$$

(4)

$${Theil}_p={\displaystyle \frac{1}{n_p}}\sum _{p=1}^{n_p}\left({\displaystyle \frac{{Te}_{pit}}{{AveTe}_{pit}}}ln{\displaystyle \frac{{Te}_{pit}}{{AveTe}_{pit}}}\right)$$

(5)

$${Theil}_{wt}=\sum _{p=1}^{n_p}\left({\displaystyle \frac{n_p}{n}}{\displaystyle \frac{{AveTe}_{pt}}{{AveTe}_t}}\right){Theil}_p$$

(6)

$${Theil}_{bt}=\sum _{p=1}^{n_p}{\displaystyle \frac{n_p}{n}}\left({\displaystyle \frac{{AveTe}_{pt}}{{AveTe}_t}}\right)ln\left({\displaystyle \frac{{AveTe}_{pt}}{{AveTe}_t}}\right)$$

(7)

$$Theil={Theil}_{wt}+{Theil}_{bt}$$

(8)

$${CR}_{pt}={\displaystyle \frac{n_p}{n}}{\displaystyle \frac{{AveTe}_{pt}}{{AveTe}_t}}{\displaystyle \frac{{Theil}_p}{Theil}}$$

(9)

$${CR}_{wt}={\displaystyle \frac{{Theil}_{wt}}{Theil}}$$

(10)

$${CR}_{bt}={\displaystyle \frac{{Theil}_{bt}}{Theil}}$$

(11)

where $Theil$ denotes the aggregate Theil index of GLCUECL in the HREEB. ${Theil}_p$ refers to the Theil index in the p-th region. ${Theil}_{wt}$ denotes the intra-regional differences; ${Theil}_{bt}$ represents the inter-regional differences. ${Te}_{it}$ denotes the efficiency in the i-th city in the period $t$.${AveTe}_t$ denotes the average value of the efficiency. $n_p$ denotes the sample size for the p-th (p = 1, 2, 3) region, while ${AveTe}_{pt}$ denotes the average value of GLCUECL in the p-th region.

2.2.4. Nonparametric Kernel Density Estimation

Kernel density estimation is an approach that does not rely on predefined parameters to approximate the probability density function of a random variable. Given the density function of the random variables $f\left(x\right)$, the formulas for estimating their probability density at points are as follows:

$$f\left(x\right)={\displaystyle \frac{1}{nh}}\sum _{i=1}^{n}k\left({\displaystyle \frac{x_i-\overline{x}}{h}}\right)$$

(12)

$$k(x)={\displaystyle \frac{1}{\sqrt{2\pi }}}exp\left[-{\displaystyle \frac{x^2}{2}}\right]$$

(13)

where $n$ denotes the total count of observations, and denotes independent and identically distributed random variables, whereas denotes the mean, while signifies the kernel function. The bandwidth $h$ influences both the kernel density’s estimation accuracy and the smoothness of the resulting density curve. Its selection follows the approach outlined by Tian et al. [46]. Specifically, the article employs the Gaussian kernel density estimation technique to analyze the distribution of the GLCUECL in the HREEB. By examining the number of peaks in the kernel density curve and how the curve shifts, researchers can assess the distribution patterns of the GLCUECL as well as their trends over time.

2.2.5. Spatial Autocorrelation Analysis

To explore whether there is spatial interdependence in the GLCUECL across the HREEB, Moran’s I index is used. This statistical measure varies between −1 and 1, where a positive value signifies a spatially positive correlation (regions with similar values are clustered together). A negative value indicates spatial negative correlation (high values are adjacent to low values, and vice versa). A value approximating zero suggests no spatial autocorrelation [47]. The global Moran’s I index is calculated as follows:

$$Global-Mora{n}^{\prime }sI={\displaystyle \frac{\sum _{i=1}^n\sum _{j=1}^n{W}_{ij}\left(X_i-\overline{X}\right)\left(X_j-\overline{X}\right)}{S^2\sum _{i=1}^n\sum _{j=1}^n{W}_{ij}}}$$

(14)

where $S^2={\displaystyle \frac{1}{n}}\sum _{i=1}^n{\left(X_i-\overline{X}\right)}^2$; $\overline{X}={\displaystyle \frac{1}{n}}\sum _{i=1}^n{X}_i$; $X_i$; and $X_j$ are the observed values of the i-th and j-th regions, respectively; $n$ denotes the number of evaluation units within the study area; and $W_{ij}$ denotes the spatial weight matrix. To explore the spatial correlation of the GLCUECL across different regions pertaining to geographical distance and economic factors, this study constructs an economic geography nested matrix. This matrix incorporates a geographic distance component—calculated based on the latitude and longitude of each region—along with an economic factor, which is measured by the per capita GDP. The relationship is expressed as $W_{ij}=W_d\ast diag\left(\overline{Y_1}/\overline{Y},\overline{Y_2}/\overline{Y},\dots ,\overline{Y_n}/\overline{Y}\right)$, where $W_d$ denotes the geographical distance matrix, $\overline{Y_i}$ signifies the mean per capita GDP of the region $i$ over the observation period, and $\overline{Y}$ denotes the average per capita GDP across all $n$ regions during the observation period.

2.2.6. Spatial Convergence Test

To analyze the convergence or divergence in GLCUECL across the HREEB, both σ-convergence and β-convergence are applied.

• σ-Convergence

σ-convergence refers to the tendency of the deviations in the GLCUECL to gradually decrease. Following the approaches of Zhang et al. [48] and Liu et al. [49], this study employs the coefficient of variation to measure σ-convergence, with the formula provided as follows:

$$\sigma ={\displaystyle \frac{\sqrt{\sum _{i=1}^n{\left({TE}_{it}-\overline{{TE}_{it}}\right)}^2/n}}{\overline{{TE}_{it}}}}$$

(15)

where ${TE}_{it}$ denotes the GLCUECL in the city $i$ during the year $t$. If the value of σ exhibits a gradually decreasing trend, it indicates the presence of σ-convergence.

2.

β-Convergence

β-convergence is analyzed in both absolute and conditional forms. Absolute β-convergence signifies that all regions tend to reach a uniform steady-state level of GLCUECL, whereas conditional β-convergence implies that each region approaches its own equilibrium after accounting for factors influencing efficiency. Recognizing the potential for spatial effects among regional efficiency levels, spatial elements are incorporated into the standard convergence model to mitigate estimation bias. To test for spatial β-convergence, an econometric framework incorporating spatial dynamics is developed. Commonly used models include the spatial lag model (SLM), the spatial error model (SEM), and the spatial Durbin model (SDM), with the SDM representing a general approach that includes both the SLM and SEM. Building on the research techniques of Zhang et al. [48] and Yu et al. [50], the absolute β-convergence model is structured as follows:

$$\mathrm{SLM}\text{: }ln{\displaystyle \frac{{TE}_{i,t+1}}{{TE}_{i,t}}}=\alpha +\beta ln{TE}_{i,t}+\rho {\sum }_{j=1}^n{W}_{ij}ln{\displaystyle \frac{{TE}_{i,t+1}}{{TE}_{i,t}}}+{\nu }_i+{\mu }_t+{\epsilon }_{it}$$

(16)

$$\mathrm{SEM}\text{: }ln{\displaystyle \frac{{TE}_{i,t+1}}{{TE}_{i,t}}}=\alpha +\beta ln{TE}_{i,t}+{\nu }_i+{\mu }_t+{\epsilon }_{it};{\epsilon }_{it}=\lambda {\sum }_{j=1}^n{W}_{ij}{\epsilon }_{it}+{\eta }_{it}$$

(17)

$$\mathrm{SDM}\text{: }ln{\displaystyle \frac{{TE}_{i,t+1}}{{TE}_{i,t}}}=\alpha +\beta ln{TE}_{i,t}+\rho {\sum }_{j=1}^n{W}_{ij}ln{\displaystyle \frac{{TE}_{i,t+1}}{{TE}_{i,t}}}+\theta {\sum }_{j=1}^n{W}_{ij}ln{TE}_{i,t}+{\nu }_i+{\mu }_t+{\epsilon }_{it}$$

(18)

where $ln\left({TE}_{i,t+1}/{TE}_{i,t}\right)$ denotes the yearly growth rate of the GLCUECL in the $i$ region of the HREEB from $t$ to $t+1$; $\alpha$ denotes the constant term, and $\beta$ denotes the spatial convergence effect. If $\beta$ is negative and passes the significance test, there is a convergence trend, and the convergence rate is $\phi =-ln\left(1+\beta \right)/t$; $\rho$, $\lambda$, and $\theta$ denote the spatial lag coefficient, spatial error coefficient, and the spatial lag coefficient of the independent variables, respectively; $W_{ij}$ denotes the spatial weight matrix; ${\nu }_i$ and ${\mu }_t$ denote individual fixed effects and time fixed effects, respectively; and ${\epsilon }_{it}$ denotes the random disturbance term for the $i$ region. All three models—the SLM, SEM, and SDM—include spatial lag terms. Conditional $\beta$ is achieved by extending the absolute convergence model to include a series of control variables that influence the GLCUECL, resulting in the conditional convergence model.

2.3. Indicator Selection and Data Sources

2.3.1. Selection of Measurement Indicators for the GLCUECL

The GLCUECL reflects the degree to which farmland is utilized in a green and low-carbon manner [7,27]. This study regards farmland utilization as a dynamic process of “input and output”. In light of the dual imperatives of food and ecological security, the concepts of “green” and “low-carbon” are incorporated into cultivated land utilization, defining GLCUECL as the comprehensive capacity to maximize socioeconomic outputs and ecological benefits of farmland—while minimizing resource consumption and environmental pollution—under a given level of agricultural technology. Drawing on existing studies [11,16,17] and incorporating the Cobb–Douglas production function—while accounting for the fundamental requirements of “pollution reduction and carbon mitigation” alongside “yield increase and carbon sequestration” in the cultivated land use process—this study constructs a measurement framework for green and low-carbon utilization efficiency of farmland that comprises inputs, desirable outputs, and undesirable outputs, as detailed in

Table 1. Measurement indicator system for the GLCUECL.

Indicators	Specific Indicators	Indicator Description (Units)	Reference Sources
Inputs	Cropland	Grain sown area/103 hm2	Feng et al. [11]; Ke et al. [16]; and Chen et al. [51]
	Labor	Agricultural practitioners/104 persons
	Fertilizer	Usage of chemical fertilizer by concentration/104 t
	Pesticides	Usage of pesticides/104 t
	Mechanical	The aggregate capacity of agricultural machinery/104 kW
	Irrigation	Effective irrigation area/103 hm2
Desirable Outputs	Economic	Total agricultural production value/CNY 108
	Social	Total grain production/104 t
	Ecological	Overall carbon sink of cultivated land use/104 t	Zhang et al. [52]; Li et al. [53]; and Xie et al. [54]
Undesirable Outputs	Non-point source pollution	Standard discharge of non-point source pollutants from cultivated land, including fertilizers and agricultural solid waste (COD, TN, TP)/108 m3	Yang et al. [55,56]; Huang et al. [57]
	Carbon emissions	Overall carbon release from cultivated land use/104 t	Tian et al. [58]; Min et al. [59]; and Zhang et al. [60]

.

For input indicators, grain sown area, agricultural practitioners, usage of chemical fertilizer by concentration, usage of pesticides, the aggregate capacity of agricultural machinery, and effective irrigation area were selected to represent land, labor, and capital inputs in the cultivated land utilization process, respectively.

For desirable outputs, agricultural output value, grain production, and cultivated land carbon sequestration were selected to reflect the economic, social, and ecological benefits of land use. The carbon sink in arable land value is mainly the amount of carbon stored by all grain crops in the cultivated land ecosystem during the whole growth period; this study calculates the carbon sink in arable land value for five crops: wheat, maize, rice, tubers, and legumes.

The undesirable outputs generated during the cultivated land utilization process are primarily non-point source pollution and carbon emissions. For non-point source pollution emissions, the unit investigation and evaluation method is usually selected to estimate the output of chemical oxygen demand (COD), total nitrogen (TN), total phosphorus (TP), and other pollutants generated by chemical fertilizer and agricultural solid waste (the method mainly measures pollution in straw of food crops) during farmland use, and converts the calculation results into equal standard discharge. According to the Environmental Quality Standard for Surface Water (GB3838-2002) [61], the discharge evaluation standards for COD, TN, and TP pollutants are set at 20 mg/L, 1 mg/L, and 0.2 mg/L, respectively. For cultivated land carbon emissions, this study focuses on emissions from agricultural materials, methane (CH₄) emissions from rice paddies, and nitrous oxide (N₂O) emissions from soil. Agricultural materials primarily include fertilizers, pesticides, agricultural films, diesel fuel, plowing, agricultural machinery usage, and cropland irrigation. Based on the existing studies [52,62], the carbon emission coefficients of these sources are as follows: fertilizers, 0.8956 kg C/kg; pesticides, 4.3941 kg C/kg; agricultural films, 5.18 kg C/kg; diesel fuel, 0.5927 kg C/kg; plowing, 312.6 kg C/km²; agricultural machinery power, 0.18 kg C/kW; and irrigation, 20.476 kg C/hm². Considering the actual conditions and data availability in the study area, mid-season rice is predominant in Jiangsu, Anhui, Henan, and Shandong; therefore, this study mainly utilizes the methane emission coefficient of medium-season rice for calculation. Referring to the existing studies [59], the CH₄ emission factors during the rice growing season for each province are as follows: Jiangsu Province, 53.55 g/m²; Shandong Province, 21 g/m²; Anhui Province, 51.24 g/m²; and Henan Province, 17.85 g/m². All of the calculation formulas and specific coefficients for farmland carbon sinks, carbon emissions, and non-point source pollution are drawn from the references listed in Table 1.

For the convenience of analysis, methane and nitrous oxide were uniformly converted into standard carbon. As outlined in the Fifth Assessment Report of IPCC, 1 ton of methane is equivalent to 34 tons of carbon dioxide (9.2727 tons of standard carbon), and 1 ton of nitrous oxide is equivalent to 298 tons of carbon dioxide (81.2727 tons of standard carbon) in regard to the greenhouse effect.

2.3.2. Selection of Control Variables for Conditional Convergence

When analyzing the conditional convergence of GLCUECL, it is essential to control for other factors influencing this efficiency. Based on the findings of Jiang et al. [63], Li et al. [26], and Lyu et al. [64], variables such as rural economic development level (AE), cultivated land resource endowment (LD), agricultural subsidy policy (SP), crop planting structure (ST), and technological input (TH) are selected as the control variables of conditional convergence. Rural economic growth is measured by the per capita disposable income of rural households, while cultivated land resource endowment is assessed using the per capita arable land area. Agricultural subsidy policies are represented in the share of government expenditure allocated to agriculture, forestry, and water management relative to total fiscal expenditure. Crop planting structures are determined by the proportion of land used for grain production relative to the total cultivated area. Lastly, technological input is assessed through regional investment in science and technology. Additional details are provided in

Table 2. Statistical overview of variables.

Variable	Code	Obs	Mean	Std. Dev.	Min	Max
Agricultural Subsidy Policy	lnSP	425	2.4114	0.4351	1.0325	7.5790
Rural Economic Development Level	lnAE	425	9.0424	0.6082	7.6425	11.4605
Cultivated Land Resource Endowment	lnLD	425	−0.9393	0.3538	−2.1579	−0.0687
Crop Planting Structure	lnST	425	4.3180	0.1102	3.9294	4.5534
Technological Input	lnTH	425	−0.0220	1.0480	−6.1921	2.3405

.

2.3.3. Data Sources

Following the Huaihe River Ecological Economic Belt Development Plan, this research divides the HREEB into three major regions, namely the ESLA, MIRA, and NHEZ. Specifically, the ESLA includes Huai’an, Yancheng, Yangzhou, Taizhou, and Chuzhou. The NHEZ contains ten cities, namely Suqian, Xuzhou, Lianyungang, Zaozhuang, Jining, Linyi, Heze, Huaibei, Suzhou, and Shangqiu. The main cities of the MIRA are Bozhou, Bengbu, Huainan, Lu’an, Pingdingshan, Luohe, Xinyang, Zhoukou, and Zhumadian. Due to data availability, the analysis utilizes panel data from 25 cities, including Huai’an, within the Huaihe River Ecological Economic Belt from 2005 to 2021 (Tongbai County of Nanyang City, Suizhou County of Suizhou City, Guangshui City, and Dawu County of Xiaogan City are not included in the research scope) as a sample. The data required for calculating GLCUECL and assessing influencing factors were primarily obtained from the statistical yearbook of Henan, the Anhui Statistical Yearbook, the Jiangsu Statistical Yearbook, the Shandong Statistical Yearbook, and related city statistical yearbooks from 2006 to 2022. Additionally, data were sourced from the Lianyungang Yearbook, the Yangzhou Yearbook, the Henan Yearbook (2006–2016), the Jiangsu Rural Statistical Yearbook (2006–2019), the China Regional Economic Statistical Yearbook (2006–2014), as well as publications from provincial agricultural department websites and other official government websites. Partial missing data were estimated using linear interpolation or the median method. To adjust for price variations, the total agricultural output value was adjusted using the GDP index, with 2005 as the base year. DEM data was obtained from the Geospatial Data Cloud (http://www.gscloud.cn/, accessed on 24 December 2024).

3. Findings and Evaluation

3.1. Evaluation of the Measurement Results of GLCUECL in the HREEB

3.1.1. Temporal Characteristics of GLCUECL in the HREEB

This study employed the constant returns to scale and non-directional global super-efficiency EBM model, utilizing iDEA Ultra V5 data envelopment analysis software (software copyright registration number: 2023SR1344621) to calculate the GLCUECL across 25 prefecture-level cities within the HREEB. The efficiency values for cities in different regions—namely the HREEB, ESLA, NHEZ, and MIRA—were averaged to illustrate the temporal evolution trend of GLCUECL, as presented in Figure 2.

The results indicate that during the sample period, the average GLCUECL in the HREEB increased from 0.6326 in 2005 to 0.9404 in 2021, showing an overall upward trend with an average increase of approximately 48.65%. Meanwhile, the average growth rates in the ESLA, NHEZ, and MIRA were approximately 47.04%, 31.63%, and 71.77%, respectively. Notably, the average change trend of the GLCUECL in the HREEB and its three sub-regions was generally consistent and relatively stable, which fluctuated in a small range from 2005 to 2014, and steadily increased from 2016 after experiencing a brief “up-and-down” from 2014 to 2016. Additionally, the average GLCUECL in the ESLA consistently surpasses that of the entire HREEB, as well as the NHEZ and MIRA. The average GLCUECL for the MIRA is slightly higher than the overall average for the HREEB, whereas the NHEZ consistently lags behind the overall HREEB average. The average GLCUECL in the ESLA is higher than that of the entire HREEB, the NHEZ, and the MIRA. Meanwhile, the MIRA maintains a slightly greater efficiency than the HREEB average, while the NHEZ remains below it. The average gap in GLCUECL across the eastern, northern, and central-western regions gradually narrowed during the sample period. This trend was particularly notable in the NHEZ, which rapidly reduced its efficiency gap with the MIRA and even surpassed it in 2021, approaching the overall average efficiency of the HREEB.

This result indicates that, although the GLCUECL in the HREEB has generally followed an upward trajectory, regional disparities remain pronounced, revealing a marked spatial imbalance. The underlying causes include the “southeastern paddy fields and northwestern dry land” clustering pattern of cultivated land resources within the HREEB [65], whereby disparate topographical conditions and variations in water–heat regimes exert significant influences on green and low-carbon land utilization. Moreover, uneven and inadequate high-quality regional economic development, differences in local governance capacities, tiered agricultural policy incentives, and the absence of effective collaborative mechanisms have collectively contributed to the spatial heterogeneity of GLCUECL.

3.1.2. Spatial Features of GLCUECL in the HREEB

To better analyze the spatial evolution of GLCUECL in the HREEB from 2005 to 2021, drawing on the experience of the existing literature [30] and combining the measurement results, the GLCUECL was divided into five types of regions: low level [0, 0.7); relatively low level [0.7, 0.8); medium level [0.8, 0.9); relatively high level [0.9, 1); and high level [1, 2). Then, ArcGIS 10.6 software was used for visualization analysis, as shown in Figure 3.

Figure 3 illustrates significant variations in the spatiotemporal dynamics of GLCUECL in 25 prefecture-level cities in the HREEB from 2005 to 2021. In 2005, the GLCUECL was generally low. However, in 2011, 2016, and 2021, a distinct “basin shape” spatial distribution pattern emerged, characterized by “higher efficiency levels around the periphery and lower levels in the central area”. Especially, in 2005, the GLCUECL in the HREEB was mainly at a low level, accounting for 76%, followed by a relatively low-level efficiency area, and the relatively level area did not appear. In 2011, the count of low-efficiency areas decreased significantly, while the number of medium-level or above increased by 9, which were concentrated in the ESLA and MIRA. Bozhou, Yangzhou, Huaian, and Taizhou became high-level efficiency areas. In 2016, only one region remained in the high-efficiency category, while low, medium, and relatively high-efficiency regions increased to 8, 5, and 4, respectively, which were distributed in the RSLA, NHEZ, and MIRA. The number of low-efficiency areas decreased from 19 in 2005 to 7 in 2016. In 2021, the GLCUECL in the HREEB was mainly at a high level, accounting for 68%, and was distributed in contiguous areas, showing a distinct agglomeration trend.

From the above assessment, it is evident that while the overall GLCUECL in the HREEB has improved, the speed and extent of improvement differ across cities. This has resulted in notable disparities in efficiency levels among different urban areas, and such differences are likely to persist for a certain period in the future.

3.1.3. GML Index and Its Decomposition of GLCUECL in the HREEB

Building on the temporal evolution analysis of the GLCUECL, the GML index model was used to calculate the GML index of GLCUECL for the entire HREEB and its sub-regions from 2005 to 2021. This index was further decomposed into the global efficiency change index (GEC) and the global technological change index (GTC). It is important to note that the GML index follows a chain index approach. For clarity, the ratio of the GLCUECL in 2005 to 2006 is defined as the 2006 index, with similar definitions for subsequent years. The findings are depicted in Figure 4.

Figure 4a illustrates that the GML index of the GLCUECL in the HREEB exhibited a fluctuating growth trend throughout the study duration. The decomposition results reveal that the GTC index followed the same trend as the GML index, while the GEC index remained generally lower than the GML index. The average values of the GTC and GEC indices were 1.0346 and 1.0002, respectively, indicating that the growth in GLCUECL was driven by a combination of technological progress and efficiency improvements, with technological progress playing a predominant role.

Figure 4b–d depicts the GML index and its decomposition for three sub-regions of the HREEB, highlighting variations in efficiency trends over time. While all three regions exhibited fluctuating growth in efficiency indices, notable differences emerged. Specifically, the average GML and GTC indices for the ESLA and MIRA were greater than 1, while the GEC indices were less than 1, suggesting that the efficiency growth in these regions was primarily driven by technological progress, which offset the negative effects of lower technical efficiency. In the NHEZ, the GTC index was higher than the GEC index and exhibited a trend consistent with that of the GML index, indicating that technological innovation, rather than improvements in technical efficiency, played a central role in efficiency growth. Additionally, since the average values of the GML, GEC, and GTC indices all exceeded 1, it can be concluded that both efficiency gains and technological advancements jointly promote the growth of the GLCUECL in the NHEZ. In summary, the GML index of GLCUECL in the HREEB and its three sub-regions generally exhibited fluctuating growth. In summary, the GML index for GLCUECL in the HREEB and its sub-regions exhibited fluctuating growth, driven mainly by technological progress.

3.2. Regional Differences and Decomposition of GLCUECL in the HREEB

This study further calculated the Theil index for GLCUECL in the HREEB from 2005 to 2021 and then analyzed the differences in the GLCUECL among its various regions. The specific computation outcomes are depicted in

Table 3. The Theil index and its decomposition of GLCUECL in the HREEB.

Year	Overall Differences	Intra- Regional Differences	Inter- Regional Differences	Differences in the ESLA	Differences in the NHEZ	Differences in the MIRA	Contribution Rate of Intra-Regional Differences (%)	Contribution Rate of Inter-Regional Differences (%)
2005	0.5015	0.4539	0.0476	0.0193	0.0290	0.3669	90.5060	9.4940
2006	0.4160	0.3362	0.0798	0.0478	0.0609	0.2174	80.8162	19.1838
2007	0.3542	0.2566	0.0977	0.0477	0.0474	0.1557	72.4245	27.5755
2008	0.3529	0.2738	0.0791	0.0339	0.0653	0.1699	77.5916	22.4084
2009	0.3547	0.2623	0.0924	0.0298	0.0285	0.1930	73.9471	26.0529
2010	0.3925	0.2930	0.0995	0.0320	0.0357	0.2149	74.6589	25.3410
2011	0.3758	0.2659	0.1099	0.0282	0.0413	0.1893	70.7490	29.2510
2012	0.3060	0.2227	0.0833	0.0381	0.0341	0.1455	72.7765	27.2235
2013	0.3554	0.2753	0.0800	0.0525	0.0342	0.1816	77.4751	22.5249
2014	0.3485	0.2706	0.0779	0.0442	0.0262	0.1960	77.6392	22.3608
2015	0.2519	0.2113	0.0406	0.0322	0.0577	0.1167	83.8728	16.1272
2016	0.3144	0.2825	0.0319	0.0354	0.0723	0.1725	89.8569	10.1431
2017	0.2750	0.2464	0.0286	0.0320	0.0696	0.1440	89.6077	10.3923
2018	0.1662	0.1367	0.0295	0.0144	0.0326	0.0894	82.2484	17.7516
2019	0.1900	0.1703	0.0197	0.0087	0.0380	0.1231	89.6473	10.3527
2020	0.2488	0.2150	0.0338	0.0100	0.0759	0.1309	86.4107	13.5893
2021	0.2215	0.1969	0.0246	0.0107	0.0717	0.1163	88.9145	11.0856

.

Figure 5 illustrates the changes in regional differences and their contribution rates to the GLCUECL in the HREEB from 2005 to 2021.

The overall difference in the GLCUECL in the HREEB exhibits a downward trend. The Theil index exhibits a downward trend, dropping from 0.5015 in 2005 to 0.2215 in 2021, representing a 55.83% decline. This indicates a narrowing of GLCUECL disparities across the HREEB. While the intra-regional Theil index follows a similar trend, the inter-regional Theil index fluctuates in an “M” shape, initially rising, then falling, and increasing again. As shown in Figure 5 and Table 3, intra-regional differences have consistently contributed more to the overall GLCUECL variation than inter-regional disparities, with intra-regional contributions above 70%, peaking at 90.51% in 2005 and reaching a low point of 70.75% in 2011. The contribution of inter-regional differences remained above 10%, peaking at 29.25% in 2011. This indicates that the GLCUECL disparities in the HREEB are primarily attributed to intra-regional differences, with inter-regional differences playing a secondary role.

From a regional perspective, the intra-regional disparities within the three sub-regions of the HREEB exhibit distinct trends. The Theil index for the ESLA and MIRA decreased from 0.0193 and 0.3669 in 2005 to 0.0107 and 0.1163 in 2021, representing reductions of 44.56% and 68.30%, respectively. Contrastingly, the Theil index for the NHEZ demonstrates an initial decline followed by an increase, rising from 0.0290 in 2005 to 0.0717 in 2021, a 59.55% increment. This indicates that the differences in GLCUECL levels are narrowing in the ESLA and MIRA, whereas the level of differences in the NHEZ is expanding. Second, the intra-regional differences remained relatively steady in the ESLA and NHEZ than those in the MIRA, and the overall fluctuation range is smaller. In addition, it is estimated that from 2005 to 2021, the average Theil index values for GLCUECL in the ESLA, NHEZ, and MIRA are 0.0304, 0.0483, and 0.1720, respectively, which further indicates that disparities within the MIRA exert a greater impact on the overall differences.

3.3. Dynamic Evolution Characteristics of GLCUECL in the HREEB

To investigate the evolutionary trends of GLCUECL across the entire HREEB and its sub-regions, a nonparametric kernel density estimation (KDE) technique was employed. This method allows for a comprehensive analysis of the evolution of GLCUECL efficiency over time. Kernel density curve graphs were created for the years 2005, 2009, 2013, 2017, 2019, and 2021, as illustrated in Figure 6.

From 2005 to 2021, the kernel density curve for the entire HREEB shows a rightward shift along the horizontal axis, indicating a general upward trend in the GLCUECL. Specifically, from 2005 to 2017, the main peak is positioned on the left, whereas the secondary peak is on the right. However, in the period from 2019 to 2021, this pattern reverses, with the main peak moving rightward and the secondary peak shifting to the left. Additionally, the distance between the main and secondary peaks has narrowed, and the height of the main peak initially declines before increasing again, while the width of the peaks becomes narrower. These patterns indicate that the GLCUECL in the HREEB is evolving towards regional polarization, although the degree of polarization has gradually weakened. The internal disparities initially widened and subsequently narrowed, reflecting a gradual convergence towards stability.

The kernel density curves for the ESLA, NHEZ, and MIRA all exhibit a trend of rightward shifts to varying degrees on the horizontal axis, which indicates that the levels of GLCUECL in the three regions exhibit an upward trend to different extents from 2005 to 2021. In regard to shape, the kernel density curve for the GLCUECL in the ESLA indicates an initial decline in the main peak’s height, followed by a subsequent increase, while the peak’s width initially expands and subsequently narrows. By 2021, the curve demonstrates a noticeable low-level reversal, forming a “U” shape, with efficiency levels clustering around higher values. This indicates that the absolute differences in the GLCUECL within the ESLA initially increased and subsequently decreased, whereas the degree of clustering increased. The kernel density curve for the GLCUECL in the NHEZ evolves from a single-peaked to a double-peaked distribution; the peak width becomes wider, and the height of the main peak gradually declines, indicating the emergence of polarization in green and low-carbon utilization efficiency in this region. This trend reflects a gradual expansion of internal disparities. The MIRA essentially maintains a bimodal structure with a dominant peak and a secondary peak. Over time, the main peak height steadily increases while the width of the peak contracts and the right tail extension diminishes. This indicates that the GLCUECL in this region is similar to that of the entire HREEB, exhibiting the characteristics of regional agglomeration alongside elements of polarization and decentralization, with the internal gap gradually narrowing. These observations highlight distinct differences in the specific performance of the GLCUECL in various regions of the HREEB.

3.4. Spatial Convergence Analysis of the GLCUECL in the HREEB

3.4.1. Spatial Correlation Test

To assess the presence of spatial correlation in GLCUECL across cities within the HREEB, this study employs Moran’s I index for correlation analysis, with the findings presented in

Table 4. Global Moran’s I index for GLCUECL in the HREEB (2005–2021).

Year	Moran’s I	z-Value	p-Value	Year	Moran’s I	z-Value	p-Value
2005	−0.005	0.862	0.194	2014	0.103 ***	3.257	0.001
2006	0.050 **	2.083	0.019	2015	0.081 ***	2.717	0.003
2007	0.037 **	1.812	0.035	2016	0.037 **	1.757	0.039
2008	0.025 *	1.492	0.068	2017	0.042 **	1.847	0.032
2009	0.039 **	1.799	0.036	2018	0.058 **	2.234	0.013
2010	0.041 **	1.823	0.034	2019	0.056 **	2.187	0.014
2011	0.078 ***	2.645	0.004	2020	0.038 **	1.770	0.038
2012	0.123 ***	3.669	0.000	2021	0.002	0.984	0.163
2013	0.127 ***	3.762	0.000

Note: ***, **, and * denote significance levels of 1%, 5%, and 10%, respectively.

. Regarding overall spatial association characteristics, the global Moran’s I index for the GLCUECL in the HREEB from 2005 to 2021 was positive except for 2005. The Z-values and p-values met the 5% significance threshold for all years except for 2005 and 2021. This indicates that the GLCUECL is not randomly distributed among the HREEB regions, but has a significant positive spatial correlation. Therefore, when analyzing the convergence of GLCUECL across the HREEB, spatial correlation must be considered to ensure the precision and validity of regression model results.

3.4.2. σ Convergence Analysis

To analyze the patterns of convergence in GLCUECL within the HREEB and its three sub-regions—namely, the ESLA, NHEZ, and MIRA—convergence was assessed by calculating the coefficient of variation for the years 2005 to 2021, as outlined in Equation (15). The findings of this evaluation are illustrated in Figure 7.

The coefficient of variation for the HREEB indicates a noticeable decline over time, with a reduction of 0.1141 in 2021 than 2005, representing a reduction of approximately 47.32%. The coefficient of variation exhibits different patterns across the three sub-regions. In the ESLA, the overall trend can be described as “increase–slight decrease–increase–fluctuating decrease”. Specifically, there was a rapid rise from 2005 to 2006, followed by a slight decline from 2006 to 2011, a brief increase from 2011 to 2013, and subsequently a fluctuating decrease from 2013 to 2021. The coefficient of variation declined from 0.0966 in 2005 to 0.0720 in 2021, representing a reduction of approximately 25.47%. The NHEZ exhibits an overall trend characterized by “fluctuating increase–slight decrease–fluctuating increase.” Specifically, there was a brief period of fluctuating increase from 2005 to 2008, a slight decrease from 2008 to 2014, and a fluctuating rise from 2014 to 2021. The coefficient of variation increased from 0.0795 in 2005 to 0.1246 in 2021, representing an increase of approximately 56.73%. The MIRA exhibits a trend similar to the overall pattern, with the coefficient of variation demonstrating a general decline. Specifically, it decreased from 0.2936 in 2005 to 0.1562 in 2021, representing a reduction of approximately 46.80%. In conclusion, both for the entire Huaihe River Ecological Economic Belt and for its three sub-regions, only the GLCUECL in the NHEZ does not exhibit σ-convergence, which is consistent with the kernel density estimation results and the differences identified through the Theil index in the previous study.

3.4.3. β-Convergence Analysis

The spatial correlation in the GLCUECL in the HREEB was established in the earlier analysis. However, due to the potential presence of varying spatial effects across different regions, according to the spatial econometric model selection steps proposed by Elhorst [66], the LM test and Hausman test are utilized to select the optimal spatial econometric framework to investigate the convergence of the GLCUECL in the HREEB. The results of these model applicability tests for both absolute and conditional β-convergence are shown in

Table 5. Applicability test of the β-convergence model.

Convergence Type	Model Test	HREEB		ESLA		NHEZ		MIRA
		Statistic	p-Value	Statistic	p-Value	Statistic	p-Value	Statistic	p-Value
Absolute β-convergence	LM Error	117.3870 ***	0.0000	0.2710	0.6030	20.6550 ***	0.0000	50.2040 ***	0.0000
	R-LM Error	27.3650 ***	0.0000	0.0000	0.9830	16.0250 ***	0.0000	0.1480	0.7000
	LM Lag	90.5840 ***	0.0000	0.3330	0.5640	12.1960 ***	0.0000	53.0400 ***	0.0000
	R-LM Lag	0.5630	0.4530	0.0630	0.8020	7.5650 ***	0.0060	2.9840 *	0.0840
	Hausman Test	11.2100 ***	0.0037	14.7200 ***	0.0001	30.0900 ***	0.0000	17.3900 ***	0.0002
Conditional β-convergence	LM Error	85.2800 ***	0.0000	0.0760	0.7830	5.1600 **	0.0230	54.4100 ***	0.0000
	R-LM Error	2.7640 *	0.0960	0.0190	0.8900	0.3270	0.5670	0.6240	0.4300
	LM Lag	85.5800 ***	0.0000	0.1210	0.7280	7.5480 ***	0.0060	56.5360 ***	0.0000
	R-LM Lag	3.0650 *	0.0800	0.0650	0.7990	2.7150 *	0.0990	2.7500 *	0.0970
	Hausman Test	229.5100 ***	0.0000	8.7200 *	0.0686	98.1300 ***	0.0000	17.1800 **	0.0163

Note: ***, **, and * denote statistical significance at the 1%, 5%, and 10% levels, respectively.

.

The LM test outcomes determine that the SEM and SDM are the selected methods for analyzing absolute and conditional β-convergence, respectively. In the NHEZ, the SDM and SLM are utilized for the same analyses. The MIRA adopts the SLM for examining absolute and conditional β-convergence. In the ESLA, the β-convergence model does not satisfy the LM test; therefore, the OLS model is employed for the convergence analysis. The findings of the Hausman test reveal that the mixed fixed effects approach is most optimal for studying both absolute and conditional β-convergence.

• Absolute β-Convergence Assessment

Table 6. Absolute β-convergence test of the GLCUECL in the HREEB.

Model	HREEB	ESLA	NHEZ	MIRA
	Double-Fixed SEM	Double-Fixed OLS	Double-Fixed SDM	Double-Fixed SLM
β	−0.5592 ***	−0.4268 ***	−0.5086 ***	−0.4634 ***
	(−0.0433)	(−0.0813)	(−0.0739)	(−0.0569)
θ			0.4475 ***
			(−0.0882)
ρ/λ	0.7330 ***		0.4580 ***	0.5524 ***
	(−0.0523)		(−0.1004)	(−0.0770)
R2	0.1302	0.2611	0.1367	0.1543
Log-Likelihood	363.6477		163.8137	129.6924
City Fixed Effects	Yes	Yes	Yes	Yes
Time Fixed Effects	Yes	Yes	Yes	Yes
Convergence Speed φ(%)	5.1198	3.4783	4.4406	3.8906
N	375	80	150	150

Note: *** denote statistical significance at the 1% levels.

illustrates the results of the absolute β-convergence test for the GLCUECL in the HREEB. First, the convergence coefficients β for the HREEB and its three sub-regions are notably negative at the 1% significance level, indicating that, in the absence of other influencing factors, the GLCUECL in the region and its sub-regions align with their respective equilibrium levels. Although the coefficient of variation in the NHEZ exhibits an upward trend (as observed in Figure 6), a long-term convergence trend has already begun to emerge. Second, there are distinct variations in the convergence rates of green and low-carbon utilization efficiency across regions. Specifically, the convergence rate for the HREEB is 5.1198%, whereas the convergence rates for the ESLA, NHEZ, and MIRA are 3.4783%, 4.4406%, and 3.8906%, respectively. Under the circumstances of the high level of variation coefficient in the NHEZ, the cities within this region exhibit a significant spatial interaction effect, enabling the region to maintain a relatively high convergence speed [48,67]. Finally, the HREEB and its three sub-regions exhibit different spatial effects. The HREEB displays spatial error lags, and the spatial error coefficient λ is substantially positive at the 1% confidence level, highlighting a strong regional interdependence in the GLCUECL. In the NHEZ, both the explanatory variables and explained variables exhibit spatial lags, with ρ and θ showing notable positive values at the 1% level. This demonstrates that improvement in GLCUECL within neighboring cities significantly enhances local efficiency, showing a positive spillover effect. Similarly, in the MIRA, a spatial lag of explained variables exists, and the spatial autoregressive coefficient ρ is also significantly positive at the 1% confidence level, reinforcing the notion that the GLCUECL gains from positive spillover effects from neighboring cities. By contrast, no spatial effects are observed in the ESLA. The preceding conclusions were obtained under the assumption that although the endowment conditions of each region are similar, they are not aligned with actual conditions. New economic growth theory suggests that, considering the varying social and economic conditions of regions at different stages of development, the final convergence results will show significant changes [68]. It is necessary to take into account the influence of external environmental factors [69]. Therefore, it is necessary to further utilize conditional β-convergence for research and analysis.

2.

Conditional β-Convergence Evaluation

Table 7. Conditional β-convergence assessment for the GLCUECL in the HREEB.

Model	HREEB	ESLA	NHEZ	MIRA
	Double-Fixed SDM	Double-Fixed OLS	Double-Fixed SLM	Double-Fixed SLM
β	−0.6139 ***	−0.5321 ***	−0.5889 ***	−0.5850 ***
	(−0.0437)	(−0.0985)	(−0.0744)	(−0.0609)
lnSP	0.0081	−0.0817	0.0064	0.0080
	(−0.0147)	(−0.0544)	(−0.0145)	(−0.0345)
lnAE	−0.0991 ***	0.0195	0.0744 ***	0.0001
	(−0.0363)	(−0.0343)	(−0.0289)	(−0.0234)
lnLD	0.0896 **	0.0652	0.2632 ***	0.0684
	(−0.0372)	(−0.0615)	(−0.1015)	(−0.0503)
lnST	0.2070	0.1940	0.2109	0.5447 *
	(−0.1400)	(−0.1344)	(−0.1545)	(−0.2964)
lnTH	−0.0023	−0.0049	−0.0186 **	0.0171
	(−0.0087)	(−0.0170)	(−0.0094)	(−0.0163)
θ	0.1804
	−0.1565
ρ/λ	0.6061 ***		0.3219 ***	0.5472 ***
	−0.0763		−0.1121	−0.0753
R2	0.1756	0.3419	0.1103	0.1489
Log-Likelihood	378.6246		172.3731	139.3055
City Fixed Effects	YES	YES	YES	YES
Time Fixed Effects	YES	YES	YES	YES
Convergence Speed φ(%)	5.9479	4.7469	5.5557	5.4967
N	375	80	150	150

Note: ***, **, and * denote significance at the 1%, 5%, and 10% thresholds, respectively.

depicts the conditional β-convergence test results of the GLCUECL in the HREEB. The analysis indicates the following: firstly, conditional β-convergence is evident across the HREEB and its sub-regions. The β-convergence coefficients for the HREEB and each of its sub-regions are significantly negative at the 1% confidence level, indicating that even when other heterogeneous influencing factors beyond initial values are considered, the GLCUECL in the region and its sub-regions still converges to the respective long-term equilibrium states. Secondly, although the speed of conditional β-convergence has accelerated to varying degrees by comparison with absolute β-convergence, the result still indicates the characteristics of the “entire HREEB > NHEZ > MIRA > ESLA”; this observation reveals that after considering other influencing factors, the convergence speed of GLCUECL also changes. Finally, the HREEB and its sub-regions exhibit different spatial effects, though the spatial effects within individual areas have changed compared with the absolute β-convergence assessment. The spatial correlation form of the entire HREEB shifts from an SEM to an SDM, with the spatial autoregressive coefficient ρ being significantly positive at the 1% threshold. This finding indicates that after controlling for other variables, the growth of the GLCUECL in the HREEB is positively influenced by spatial spillovers from neighboring cities. In the NHEZ, the model shifts from SDM to SLM, and the spatial lag coefficient ρ for both the NHEZ and MIRA remains significantly positive at the 1% confidence threshold, confirming that, after accounting for control variables, the GLCUECL in these regions is affected by the positive spatial spillover effects from neighboring cities. Similarly to absolute β-convergence, no spatial effects are observed in the ESLA.

The regression analysis of the control variables reveals variations in the coefficients and significance levels of the HREEB and its three sub-regions. Specifically, concerning the HREEB, the coefficient associated with the rural economic development level exhibits a significant negative correlation. This suggests that enhancing the rural economic development level within the HREEB will drive the GLCUECL towards a stable state, consequently reducing regional disparities. Conversely, the coefficient for cultivated land endowment is significantly positive, indicating that increasing the per capita cultivated land area in these regions enhances the GLCUECL but does not alleviate regional disparities. Notably, factors such as agricultural subsidy policies, crop planting structure, and scientific and technological input do not exert a significant influence on the rate of change in GLCUECL. However, it is important to underscore that this does not imply these factors have no impact on the GLCUECL in the HREEB, but rather that drawing a definitive conclusion on whether these factors inhibit or promote the regional disparities in GLCUECL is inconclusive.

For the three sub-regions, the impacts of agricultural subsidy policies, rural economic development levels, cultivated land resources endowments, crop planting structures, and technological inputs on the GLCUECL show significant spatial heterogeneity. Specifically, agricultural subsidy policies exhibit adverse effects on the GLCUECL in the ESLA. Conversely, they demonstrate positive effects in the NHEZ and MIRA. However, these effects do not reach statistical significance, making it challenging to ascertain their effectiveness in promoting the convergence of GLCUECL towards low or high values.

The rural economic development level significantly influences the GLCUECL in the NHEZ. Specifically, an increase in rural disposable income notably enhances the GLCUECL in this region. However, this increase also hinders the reduction in internal disparities. Conversely, the ESLA and MIRA did not exhibit statistically significant effects, suggesting a need for additional investigation into the impact of rural economic development on the GLCUECL in these regions. The endowment of cultivated land resources significantly enhances the GLCUECL in the NHEZ. This suggests that expansion of the operation scale of cultivated land can boost the GLCUECL but may further exacerbate inter-regional disparities. Crop planting structure has a significant positive effect on GLCUECL in the MIRA, indicating that adjusting and optimizing crop planting structure can promote GLCUECL in the MIRA. Interestingly, in the NHEZ, the coefficient associated with science and technology input exhibits a significant negative correlation. This suggests that the level of innovation in science and technology contributes to the convergence of GLCUECL in this economic zone.

3.4.4. Robustness Check

To test the reliability of the convergence model results, this study further conducts stability checks on the β-convergence results using the following methods: (1) replacing the economic–geographical nested spatial weight matrix with a geographical inverse distance squared spatial weight matrix; (2) recalculating the GLCUECL using the global super-efficiency SBM model; and (3) applying a 1% winsorization to the variables. The results of these checks are illustrated in

Table 8. Robustness check of β-convergence regarding the GLCUECL across the HREEB.

Method	Model	Absolute β-Convergence				Conditional β-Convergence
		HREEB	ESLA	NHEZ	MIRA	HREEB	ESLA	NHEZ	MIRA
		Double-Fixed SEM	Double-Fixed OLS	Double-Fixed SDM	Double-Fixed SLM	Double-Fixed SDM	Double-Fixed OLS	Double-Fixed SLM	Double-Fixed SLM
Change in Spatial Weight Matrix	β	−0.5376 ***	−0.6009 **	−0.4687 ***	−0.4492 ***	−0.5996 ***	−0.7045 **	−0.5495 ***	−0.5722 ***
		(−0.042)	(−0.1612)	(−0.072)	(−0.0564)	(−0.0426)	(−0.2231)	(−0.0728)	(−0.0599)
	θ			0.3960 ***		0.2243 **
				(−0.083)		(−0.1054)
	ρ/λ	0.6818 ***		0.5025 ***	0.5086 ***	0.5691 ***		0.3570 ***	0.5093 ***
		(−0.0489)		(−0.0789)	(−0.0693)	(−0.0625)		(−0.0853)	(−0.0671)
Global Super- Efficiency SBM	β	−0.6005 ***	−0.3896 ***	−0.5015 ***	−0.4664 ***	−0.6546 ***	−0.7293 **	−0.5817 ***	−0.6199 ***
		(−0.0443)	(−0.0771)	(−0.0759)	(−0.0579)	(−0.0448)	(−0.2314)	(−0.0768)	(−0.0627)
	θ			0.4547 ***		0.1808
				(−0.0863)		(−0.1604)
	ρ/λ	0.7514 ***		0.4615 ***	0.5403 ***	0.5980 ***		0.3514 ***	0.5188 ***
		(−0.0486)		(−0.0989)	(−0.0795)	(−0.0775)		(−0.1081)	(−0.0781)
1% Winsorization of Variables	β	−0.5318 ***	−0.4268 ***	−0.4829 ***	−0.4451 ***	−0.5892 ***	−0.7045 **	−0.5466 ***	−0.5691 ***
		(−0.0417)	(−0.0813)	(−0.0696)	(−0.0553)	(−0.0422)	(−0.2231)	(−0.0702)	(−0.0593)
	θ			0.4328 ***		0.1671
				(−0.0836)		(−0.1507)
	ρ/λ	0.7251 ***		0.4480 ***	0.5567 ***	0.5989 ***		0.3125 ***	0.5483 ***
		(−0.0532)		(−0.0987)	(−0.076)	(−0.0758)		(−0.1124)	(−0.0746)

Note: *** and ** indicate significance at the 1% and 5% levels, respectively.

. The findings indicate that the convergence model selection results remain consistent across all robustness tests. The β coefficients obtained under different scenarios remain negative and achieve statistical relevance at the 1% level, reinforcing that the HREEB still exhibits significant absolute and conditional β-convergence trends in the GLCUECL. Additionally, the spatial effect coefficients are significantly positive, whereas no spatial effects are observed in the ESLA, aligning with previous conclusions. Therefore, it is reasonable to propose that the β-convergence results for the GLCUECL across the whole HREEB and its sub-regions are robust.

4. Discussion

4.1. Review of Research Findings

Food and ecological security are critical components of national security strategies. Farmlands serve as the cornerstone of food security while also providing essential environmental benefits, including carbon absorption, oxygen production, water retention, soil erosion control, and habitat maintenance [70,71]. Currently, China’s efforts for land use and conservation focus on safeguarding food security, maintaining social stability, and enhancing ecological security [72]. Advancing the “greening” and “decarbonization” of cultivated land utilization is crucial for achieving and maintaining both food and environmental stability.

The findings of this study indicate an overall improvement in the GLCUECL in the HREEB in recent years, though significant regional disparities exist. This observation aligns with the findings of Ke et al. [16], who observed an increasing trend in the level of GLCUECL in China from 2000 to 2019, with significant regional disparities. Similarly, studies by Yang et al. [13] and Fu et al. [7] also reported an overall upward trend in the GLCUECL in the Yangtze River Economic Belt and major grain-producing areas, respectively. The upward trend can be explained by the Chinese government’s emphasis on cultivating land protection, which has gained increasing focus since the 18th National Congress of the Communist Party of China. Policies have progressively advanced a “trinity” approach to protecting land quantity, quality, and ecology while promoting environmentally sustainable and low-carbon growth.

From the perspective of spatial distribution, the GLCUECL exhibits apparent regional disequilibrium, which is specifically manifested in the ESLA > MIRA > NHEZ. Cultivated land, as the core resource underpinning both food and ecological security, is currently undergoing a profound green and low-carbon transformation. However, owing to the unique resource endowment and the stage of economic development in different regions, the GLCUECL shows a significant regional imbalance. This imbalance is not only reflected in the differences in efficiency values but also more profoundly indicates the potential accumulation of critical vulnerability risks in the social–technical system during the pursuit of efficiency optimization. The spatial disequilibrium of GLCUECL in the HREEB essentially stems from the differences in resource allocation and input–output in the process of cultivated land utilization and protection among various regions within the belt [73]. GLCUECL requires a balance between food security and ecological integrity. Optimization behaviors under such trade-offs may render cultivated land systems increasingly fragile, highlighting the need to be vigilant against the risks of homogenized competition [48]. Each region should act in accordance with the local conditions; strengthen the integration of cultivated land utilization and ecological protection; promote the green and low-carbon transformation of cultivated land utilization; and thereby achieve a long-term balance between food and ecological security.

In regard to convergence, the HREEB and its three sub-regions display distinct spatial effects on the GLCUECL. Areas with relatively lower efficiency exhibit a catching-up trend towards those with higher efficiency, ultimately stabilizing at their respective equilibrium levels. This observation aligns with the conclusions of scholars such as Xue et al. [23], Li et al. [26], and Lyu et al. [64], reaffirming that China’s cultivated land utilization is increasingly focused on greening and low carbonization [19]. It is particularly noteworthy that although the intra-regional disparity in the GLCUECL within the NHEZ has shown signs of widening, the efficiency levels are gradually converging toward a stable state. This phenomenon of seemingly contradictory polarization and convergence is the result of the dynamic interaction between the “diffusion effect” and “aggregation effect”. Given the disparities in regional economic development levels and resource endowments, areas with better conditions have great advantages in the input of agricultural production factors such as capital, technology, and policy, so as to improve agricultural production conditions and GLCUECL [74]. Moreover, through “learning-by-doing” effects, these areas reduce the cost of green technology and form a circular cumulative advantage, which will inevitably lead to the expansion of the internal absolute gap. Nevertheless, policy interventions that reduce steady-state disparities, coupled with technology diffusion and spatial spillovers from high-efficiency to low-efficiency regions, can promote the development of each region towards its own steady state [75], that is, the internal convergence of high-efficiency regions and low-efficiency regions, respectively.

China has achieved remarkable success in ensuring food security and promoting farmers’ income growth. However, this process has long been accompanied by excessive exploitation of the ecological sustainability of cultivated land, and the accumulated environmental costs have become undeniable. The excessive use of agricultural chemicals leads to the loss of soil organic matter, which not only weakens agricultural productivity but also intensifies carbon emissions and damages the ecosystem services of farmland, such as carbon sink capacity and biodiversity. The leaching of nitrogen and phosphorus from fertilizers causes water body eutrophication, damaging aquatic ecosystems while releasing greenhouse gases such as methane and nitrous oxide, which further exacerbate climate change, leading to rising temperatures, changes in precipitation patterns, and an increase in the frequency of extreme weather events, etc. [76,77]. In the long term, this may threaten food security and the well-being of rural residents [78]. Therefore, actively developing mitigation and adaptation farming practices, adopting protective tillage measures, precise nutrient management, scientific crop rotation or intercropping, and the application of biochar, etc., to promote the green and low-carbon transformation of farmland utilization is an important way to reduce pollution, lower carbon emissions, increase carbon sequestration, and enhance the agricultural ability to cope with climate change [77].

4.2. Policy Recommendations

Xi Jinping, General Secretary of the Communist Party of China (CPC) Central Committee, stated that China’s economic and social progress has transitioned into a phase of high-quality growth, characterized by an accelerated shift toward sustainability and low-carbon emissions. As a critical grain production base and an ecological security barrier, cultivated land utilization has a direct impact on ecological security in the HREEB, making the evaluation of GLCUECL vital for advancing high-quality agricultural development and ecological revitalization in the region. Based on the empirical findings of this study, a series of policy recommendations should be proposed to promote the sustainable and low-carbon use of farmland.

Firstly, the green development concept should be implemented; moreover, institutional supply should be optimized to enhance the efficiency of green and low-carbon utilization of cultivated land. Although the overall GLCUECL in the HREEB has shown a rising trend, there remains substantial room for improvement, with evident signs of regional polarization.

Secondly, a regional cooperative mechanism for the GLCUECL should be established to promote the synergistic development of regions in the HREEB and to reduce the regional differences in the GLCUECL. Research shows that the GLCUECL in the HREEB presents a regional unbalanced pattern. This requires cities in the HREEB to break through geographical restrictions and form an effective pattern of synergistic development. Through demonstration and trickle-down effects, the practices of regions with higher efficiency will radiate outward and influence regions with lower efficiency, preventing the “Matthew effect” of GLCUECL between regions caused by the cumulative causal loop, ultimately narrowing the differences in the GLCUECL between regions.

Finally, differentiated strategies for enhancing the GLCUECL should be implemented based on local conditions. The GLCUECL in the HREEB and its three sub-regions all exhibit a β-convergence mechanism, indicating that cities with lower efficiency have a tendency to catch up with those with higher efficiency. Each region should formulate targeted policies based on its natural resource endowment and economic development stage, avoiding a one-size-fits-all approach, and enhance the GLCUECL in accordance with local conditions. For instance, the ESLA should actively cultivate and develop new quality agricultural productivity, accelerate the transformation of agricultural scientific and technological innovation achievements, and provide technological support for enhancing the GLCUECL; the NHEZ should further promote the rural revitalization strategy, develop moderate-scale operations, and strengthen investment in agricultural scientific and technological innovation; and finally, the MIRA should enhance agricultural policy support, optimize crop planting structures, and promote green and low-carbon transformation methods for cultivated land utilization.

4.3. Research Limitations

While this study contributes significantly to the literature and provides useful policy insights, several limitations exist: First, the analysis is conducted at the meso level, concentrating on prefecture-level cities and relying on publicly available government data. Future research could explore micro-level data by conducting field surveys and including farmers as participants to provide a more comprehensive and accurate perspective. Second, the GLCUECL metric is influenced by the choice of evaluation indicators, which could lead to variations in results. Future studies could optimize input–output indicators, particularly by incorporating ecosystem service values to improve the accuracy of assessments.

5. Conclusions

This study utilized panel data from 25 cities in the HREEB from 2005 to 2021. A global super-efficiency EBM model, incorporating unwanted outputs, was constructed and combined with the GML index model to measure and analyze the GLCUECL under the dual objectives of food and ecological security. The Theil index was utilized to quantify and break down regional disparities in GLCUECL across the HREEB. Additionally, the kernel density estimation was applied to illustrate the dynamic evolution of green and low-carbon efficiency in cultivated land within the HREEB and its three sub-regions. Finally, σ-convergence and β-convergence tests were conducted to assess convergence in cultivated land green and low-carbon utilization efficiency in these regions. The key research findings are as follows:

(1)

From the measurement results, the GLCUECL in the HREEB exhibits a general upward trend throughout the study period. However, disparities persist in efficiency levels, exhibiting a trend of “ESLA > MIRA > HREEB > NHEZ”, though the efficiency gap is gradually decreasing. The growth in the GML index for the entire HREEB and its three sub-regions is driven by both technological progress (GTC) and changes in efficiency (GEC), with GTC being the primary driving force behind the improvement.

(2)

In regard to regional differences, the overall disparities of the GLCUECL within the HREEB exhibited a declining trend over the sample period. The efficiency differences between the ESLA and MIRA tend to decrease, whereas the disparities within the NHEZ tend to widen. The main driver of regional discrepancies is the intra-regional disparities, with intra-regional variance in the MIRA contributing most significantly to the overall disparities.

(3)

The kernel density curves for the HREEB and its three sub-regions have shifted significantly to the right, reflecting a general improvement in GLCUECL across the study period. However, there are notable differences in performance among specific regions. Apart from the NHEZ, which has exhibited an increase in absolute internal differences and a polarization phenomenon, the overall HREEB, along with the ESLA and MIRA, exhibits a trend of decreasing internal disparities.

(4)

The spatial convergence of the GLCUECL within the HREEB exhibits significant regional differences. Among the HREEB and its three sub-regions, only the NHEZ does not manifest σ-convergence. Additionally, the HREEB and its three sub-regions exhibit significant absolute β-convergence; the HREEB has the fastest convergence rate. After considering other variables, a significant trend of conditional β-convergence remains observable, and the convergence speed increases by different degrees. The HREEB, NHEZ, and MIRA exhibit different spatial effects, whereas the ESLA manifests no spatial effects. Factors such as rural economic development level, cultivated land resource endowment, agricultural subsidy policies, crop planting structure, and technological investment exert heterogeneous influences on the GLCUECL across different regions. This study provides important insights into the spatial distribution, trends, and drivers of GLCUECL in the HREEB and suggests targeted policies for promoting green and low-carbon land use.