Towards Sustainable Agriculture: How Does Agricultural Scale Operation Affect the Cultivated Land Green Utilization Efficiency? The Empirical Evidence from China

Abstract

Promoting cultivated land green utilization efficiency (CLGUE) through agricultural scale operation is critical for reconciling the conflict between food security and sustainable land use. Based on panel data from 30 provinces in China (2007–2022), this paper calculates CLGUE using the Super-efficiency SBM-Undesirable model and empirically examines the impact mechanisms and nonlinear characteristics of scale operation using Tobit and threshold models. The findings reveal that: (1) Agricultural scale operation has a significant positive impact on CLGUE, but this effect is non-linear and characterized by diminishing marginal returns, validating the “moderate scale” operation theory. (2) Substantial heterogeneity exists across different functional grain production zones and geographic regions. (3) Mechanism analysis identifies technological innovation as a key transmission channel through which scale operation boosts CLGUE. (4) A significant double-threshold effect is observed in fiscal support for agriculture; specifically, the positive enabling effect of scale operation is maximized only when fiscal support intensity is maintained within a specific rational range. Consequently, this study suggests that policymakers should prioritize “moderate scale” strategies, tailor policies to regional conditions, and optimize the allocation of fiscal funds to foster a transition toward green and sustainable agriculture.

1. Introduction

1.1. Research Background and Significance

Amidst tightening global resource constraints and environmental degradation, promoting the green transformation of agriculture has become an imperative for achieving sustainable development. As the world’s largest developing country, China is at a critical juncture, transitioning from prioritizing yield growth to emphasizing quality and ecological benefits. According to the China Agricultural Green Development Report 2023 (released in August 2024), China’s agricultural green development index steadily rose from 75.19 in 2015 to 77.90 in 2022. Despite this progress, severe challenges persist. The Central Economic Work Conference in late 2024 explicitly called for the “synergistic promotion of carbon reduction, pollution control, green expansion, and growth.” Furthermore, the Plan for Accelerating the Construction of a Strong Agricultural Country (2024–2035), issued by the State Council in 2025, emphasized that “green, circular, and low-carbon development must become the defining feature of an agricultural powerhouse.” In this context, cultivated land serves as the core element of agricultural production, and its utilization efficiency is intrinsically linked to the dual imperatives of national food security and ecological safety.

Concurrently, driven by rapid urbanization and the labor exodus to non-agricultural sectors, China’s agricultural operation models are undergoing a profound transformation. The acceleration of land transfer has facilitated a shift from fragmented smallholder farming to moderate and large-scale operations [1]. Theoretically, scale operation facilitates the adoption of modern technologies and yield improvement; for instance, the average yield of grain-producing family farms is approximately 20% higher than the national average. However, a core question arises: Does the expansion of agricultural operation scale naturally lead to the “greening” of cultivated land utilization? This remains a “black box” that necessitates rigorous verification.

Existing literature regarding the environmental impact of agricultural scale operation has not yet reached a consensus, generally diverging into three perspectives. The first view supports the “Scale Bonus Hypothesis.” Proponents argue that scale operation optimizes factor allocation and internalizes environmental costs [2,3]. Empirical evidence indicates, for example, that a 1% increase in cultivated land area per household significantly reduces fertilizer and pesticide application intensities by 0.3% and 0.5%, respectively [4]. Conversely, the second view supports the “Scale Trap Hypothesis.” Critics argue that to maximize short-term profits, large-scale operators may intensify agrochemical inputs, thereby exacerbating non-point source pollution and ecosystem stress [5,6,7,8]. The third view proposes a “Non-linear Relationship.” Recent studies suggest a “U-shaped” or inverted “U-shaped” relationship between scale operation and fertilizer intensity or green total factor productivity [9,10], implying that the scale effect is contingent upon specific thresholds or external conditions.

Despite the rich body of existing research, several gaps remain. First, most studies focus on single indicators (e.g., fertilizer reduction), lacking a systematic assessment of the cultivated land green utilization efficiency (CLGUE)—a comprehensive metric incorporating resource inputs, desirable outputs, and undesirable outputs. Second, the transmission mechanisms and boundary conditions under which scale operation affects CLGUE remain underexplored. To bridge these gaps, this study utilizes panel data from 30 Chinese provinces spanning 2007 to 2022 and employs the Super-efficiency SBM-Undesirable model to measure CLGUE. We theoretically analyze and empirically test the non-linear relationship, transmission paths, and threshold characteristics of scale operation on CLGUE. The marginal contributions of this paper are as follows: (1) Perspective Reconstruction: Shifting from “Single-Factor Reduction” to “Total-Factor Systemic Evaluation.” While existing studies acknowledge the environmental implications of scale operation, they are largely confined to examining the reduction in single input factors such as fertilizers and pesticides, thereby failing to comprehensively measure the holistic green performance of agricultural production systems. Departing from this limitation, this study employs the Super-SBM-Undesirable model to construct the Cultivated Land Green Utilization Efficiency (CLGUE) indicator system. By explicitly endogenizing “environmental costs” as undesirable outputs, this study provides a more precise depiction of the comprehensive synergistic performance of scale operation in balancing “economic growth” and “ecological burden reduction.” (2) Mechanism Deconstruction: Shifting from “Linear Exogenous Variables” to “Nonlinear Endogenous Thresholds.” This study challenges this conventional setting by innovatively identifying the level of fiscal support as a critical threshold variable that constrains the realization of the green effects of scale operation. We unveil the “double-edged sword” effect of fiscal support. This finding theoretically corrects the cognitive bias that “more subsidies are always better” and fills the gap regarding the mechanisms of policy intensity boundaries. (3) Theoretical Integration: Reconciling the Academic Divide between “Scale Dividends” and “Scale Traps.” Addressing the persistent debate regarding whether scale operation exerts a “positive promotion” or “negative exacerbation” on environmental outcomes, this study provides robust evidence confirming that these views are not contradictory but rather represent different stages of an inverted U-shaped trajectory. This finding successfully reconciles conflicting theoretical perspectives, proving that the release of environmental dividends is subject to strict scale constraints, thereby providing a unified theoretical framework for understanding the ecological boundaries of appropriate scale operation.

1.2. Theoretical Analysis and Research Hypothesis

1.2.1. The Direct Effect of Agricultural Scale Operation on CLGUE

In rural China, the characteristics of cultivated land fragmentation are prevalent. Farmers tend to move between scattered plots to spray fertilizers and pesticides, which not only consumes physical energy and increases labor hours but also hinders the precise control of dosage and spraying range. To ensure coverage of crops at the edges of plots, farmers inevitably increase the application intensity per unit area. Agricultural scale operation can ameliorate the previous fragmented land management model by promoting land contiguity [11]. Contiguous operation facilitates farmers in improving the precision of fertilizer and pesticide application, promoting reduction in chemical usage, and thereby enhancing CLGUE [12].

Agricultural production is a complex process involving the allocation of multiple factors, primarily labor, land, and capital. Adopting green agricultural production behaviors requires not only a sufficient resource endowment but also the rational allocation of these resources [13]. Scale operation in agriculture can optimize the allocation between land and labor, as well as between land and capital. This optimization resolves the issue of low resource utilization caused by land fragmentation, appropriately enhances agricultural production efficiency, and influences green agricultural development [14]. Furthermore, specialized service organizations provide outsourcing services by treating specific production stages, such as soil preparation, planting, and harvesting, as intermediate products. This approach optimizes labor factor allocation, meets production demands, and further improves production efficiency.

When agricultural scale operation reaches a moderate level where production factor allocation is optimal, the best operating benefits are achieved. However, when the allocation of production factors deviates from the optimal combination, returns begin to show a trend of diminishing marginal returns [15]. Specifically, as the scale of cultivated land operation expands, the precision of farmers’ fertilizer and pesticide application gradually improves. However, with further expansion of scale, if the allocation of factors such as capital and labor fails to optimize synchronously with land scale, farmers may increase the intensity of chemical inputs to cope with production risks and maintain yields. Although this behavior may maintain or increase yields in the short term, in the long run, it exacerbates agricultural non-point source pollution and reduces CLGUE [16,17].

Hypothesis 1. 

There is an inverted “U-shaped” non-linear relationship between agricultural scale operation and the cultivated land green utilization efficiency.

1.2.2. The Indirect Effect of Agricultural Scale Operation on CLGUE

Land transfer facilitates the shift of production factors to more efficient operators, thereby improving agricultural economies of scale. Research indicates that economies of scale originate from the economies of division of labor; through the specialization of production links, the production efficiency of factors can be improved [18]. Small-scale farms have been regarded as a primary cause of poor agricultural performance, as they hinder technological progress and reduce opportunities for credit access [19]. While land transfer improves the efficiency of farmland resource allocation, it is accompanied by the expansion of farmers’ cultivation area, leading to changes in agricultural production methods and technologies. Therefore, agricultural scale operation may indirectly affect green development by influencing the level of technological innovation.

Agricultural scale operation can achieve sufficient market capacity to meet the adoption thresholds for machinery and pro-environment production technologies, thereby realizing agricultural green production [20]. Through centralized management, agricultural scale operation can form a demonstration effect for green production, encouraging surrounding farmers to improve production technologies and promoting sustainable regional agricultural development [21]. Since the technology adoption costs for large-scale farmers can be spread over a larger operating area, their propensity to adopt technologies for reducing fertilizers and pesticides is increased [22], ultimately enhancing CLGUE.

Hypothesis 2. 

Agricultural scale operation promotes the CLGUE by fostering technological innovation.

1.2.3. Threshold Effect Analysis of Fiscal Support Policy for Agriculture

Fiscal support for agriculture refers to the financial support provided by the government through public finance functions to the “agriculture, rural areas, and farmers” sector [23]. It serves as an important foundation for breaking through agricultural development bottlenecks and has a significant impact on green agricultural production. In actual agricultural production activities, government fiscal policy support and subsidies are typically the primary drivers stimulating the production enthusiasm of agricultural operators [24]. Fiscal support can release policy incentive effects through its internal operating mechanisms, promoting agricultural technological innovation [25], talent cultivation, and infrastructure construction [26], further enhancing agricultural green production efficiency. However, from a long-term perspective, the management system of fiscal support still needs optimization; only by rationally allocating funds can the sustainable advancement of agricultural green development be promoted [27].

Through capital investment and policy support, fiscal support for agriculture can effectively incentivize agricultural operators to undertake technological innovation and equipment upgrades [28], thereby improving CLGUE. As the scale of agricultural production expands, operators require more technical support and efficient equipment to achieve optimal resource allocation. Subsidies and development funds provided by fiscal support promote farmers’ income growth [29], help alleviate financial pressure on R&D and equipment procurement, and facilitate the promotion and application of agricultural technologies, thus enhancing CLGUE. Particularly when purchasing large-scale agricultural machinery and implementing advanced production methods, fiscal support provides necessary financial guarantees, which are crucial for improving the efficient utilization of cultivated land resources.

Nevertheless, the effect of fiscal support is not linearly increasing. Excessive fiscal funds may lead to inefficient and irrational resource allocation [30], causing farmers to develop a dependency on new technologies and agricultural equipment. This can result in an excessively high degree of mechanization, causing environmental pollution and limiting the exploration of agricultural green development paths [31].

Hypothesis 3. 

A threshold effect regarding the level of fiscal support for agriculture exists in the process of agricultural scale operation affecting CLGUE.

1.3. Cultivated Land Green Utilization Efficiency: Concept and Progress

1.3.1. Concept and Connotation of CLGUE

Cultivated land green utilization is a production mode that pursues the harmonious co-evolution of the rural economy, cultivated land utilization, and ecological protection, emphasizing the minimization of “undesirable” outputs while simultaneously obtaining economic benefits [32,33]. Cultivated land green utilization efficiency (CLGUE), in contrast, serves as a quantitative evaluation of the performance of this utilization process, defined as the ratio of production factor inputs to various outputs under environmental constraints [34]. Distinct from cultivated land green utilization, which prioritizes process analysis, CLGUE places greater emphasis on outcome analysis; its core objective lies in measuring the level of coordination between the maximization of economic output and the optimization of ecological benefits.

1.3.2. Research Progress of CLGUE

Existing scholarship on cultivated land green utilization efficiency (CLGUE) predominantly focuses on key aspects including efficiency measurement under environmental constraints, spatiotemporal evolution, and the analysis of driving factors [35,36,37,38]. Regarding evaluation index systems, the prevailing paradigm integrates environmental constraints into traditional input–output frameworks. Specifically, carbon emissions are frequently treated as undesirable outputs to align with low-carbon objectives [39,40]. Moreover, recent studies have expanded this scope to encompass agricultural non-point source pollution as a concurrent undesirable output alongside carbon emissions [41,42]. Methodologically, measurement techniques have evolved from simple composite indicators [43,44] to advanced frontier analysis methods, such as Data Envelopment Analysis (DEA) and Stochastic Frontier Analysis (SFA) [45,46,47]. Notably, the Super-SBM model has gained prominence for its ability to handle multiple inputs and outputs without requiring pre-specified functional forms [48]. Crucially, this model effectively resolves the issue of input–output slacks, establishing it as one of the most robust tools in the field. Concerning determinants and policy implications, research has identified key drivers including land endowment [49,50,51], technological advancement [52,53], and policy orientation [54]. By employing methodologies such as spatial econometrics, Geodetectors [55], and Geographically Weighted Regression, scholars have dissected these drivers at both global and local scales, providing a scientific basis for differentiated optimization strategies.

2. Materials and Methods

2.1. Measurement of Cultivated Land Green Utilization Efficiency

To circumvent the potential estimation bias inherent in traditional Data Envelopment Analysis (DEA) methods caused by radial and angular choices, and to effectively distinguish efficiency differences among multiple Decision Making Units (DMUs), this study draws upon the methodology of Tone [56] and employs the Super-efficiency SBM-Undesirable model to measure the CLGUE. Assume there are n DMUs in the utilization of cultivated land. Each DMU contains m input vectors $x\in R^m$, s₁ desirable output vectors $y^g\in R^{s_1}$, and s₂ undesirable output vectors $y^{\mathrm{b}}\in R^{s_2}$. The matrices are defined as $X=\left[x_1,\dots ,x_n\right]\in R^{m\times \mathrm{n}}>0,$ $Y^g=\left[y_1^g,\dots y_n^g\right]\in R^{s_1\times \mathrm{n}}>0,$ $Y^b=\left[y_1^b,\dots ,y_n^b\right]\in R^{s_2\times \mathrm{n}}>0$, respectively. The model expression is as follows:

$$\begin{array}{l}{\rho }^{*}=\min {\displaystyle \frac{1+\frac{1}{m}{\displaystyle \sum _{i=1}^m\frac{s_i^{-}}{x_{ik}}}}{1-\frac{1}{s_1+s_2}\left({\displaystyle \sum _{r=1}^{s_1}\frac{s_r^g}{y_{rk}^g}+{\displaystyle \sum _{t=1}^{s_2}\frac{s_t^b}{y_{tk}^b}}}\right)}}\\ s.t.\left\{\begin{array}{l}{x}_{ik}\ge {\displaystyle \sum _{j=1,j\ne k}^n{x}_{ij}{\lambda }_j-s_i^{-}}\\ y_{rk}^g\le {\displaystyle \sum _{j=1,j\ne k}^n{y}_{rk}{\lambda }_j+s_r^{+}}\\ y_{tk}^b\ge {\displaystyle \sum _{j=1,j\ne k}^n{x}_{tj}^b{\lambda }_j-s_t^{b-}}\end{array}\right\}\\ 1-{\displaystyle \frac{1}{s_1+s_2}}\left({\displaystyle \sum _{r=1}^{s_1}\frac{s_r^g}{y_{rk}^g}+{\displaystyle \sum _{t=1}^{s_2}\frac{s_t^b}{y_{tk}^b}}}\right)>0\\ i=1,2,\dots ,m;j=1,2,\dots ,n;\\ r=1,2,\dots ,s_1;t=1,2,\dots ,s_2;\\ j=1,2,\dots ,n(j\ne k)\\ s^{-},s^g,s^b,\lambda \ge 0;\end{array}$$

(1)

In the equation, $s^{-},s^g,s^b$ represent the slack variables for inputs, desirable outputs, and undesirable outputs, respectively; $\lambda$ denotes the weight vector; and $k$ represents the specific DMU under evaluation. ${\rho }^{*}$ stands for the efficiency score of the DMU. When $\rho \ge 1$, the efficiency of CLGUE achieves an effective state. Conversely, when $0\le \rho <1$, the efficiency is in an ineffective state, indicating the presence of input redundancy or output shortfalls.

2.2. Variable Selection

2.2.1. Explained Variable

The explained variable in this study is the Cultivated Land Green Utilization Efficiency (CLGUE), which is measured based on the input and output quantities per unit of cultivated land area. Unlike traditional cultivated land use efficiency, CLGUE is an efficiency metric that seeks to maximize both economic and ecological benefits while minimizing resource consumption [57]. When measuring CLGUE, it is essential to consider not only socio-economic outputs but also environmental outcomes. Therefore, this study incorporates environmental factors into the evaluation indicator system to analyze CLGUE more comprehensively from the perspective of integrating socio-economic and ecological dimensions.

Specific calculation methods for key indicators are as follows:

(1) Desirable Outputs: In addition to the economic output, this study explicitly incorporates agricultural carbon sinks as an environmental output. The agricultural carbon sink was estimated based on the coefficients and methods Jin et al. [58], Cui et al. [59], and Li [60]. The calculation formula is as follows:

$$C_{S_i}={\displaystyle \sum _{i=1}^m{C}_{S_i}}={\displaystyle \sum _{i=1}^m\left\{\left[C_i\times Y_i\times \left(1-W_i\right)\right]/H_i\right\}}$$

(2)

where $C_{S_i}$ represents the total carbon sink of cultivated land; $m$ denotes the number of crop types; and $C_i$, $Y_i$, $w_i$, and $H_i$ refer to the carbon absorption rate, yield, moisture content, and economic coefficient of the i-th crop, respectively. The specific values for these coefficients are listed in Table 1. While we acknowledge that these biological coefficients may exhibit slight spatial variations, utilizing standard coefficients remains the dominant approach in macro-level empirical studies to ensure comparability across provinces.

Table 1. Moisture content, economic coefficient, and carbon absorption rate of crops.

Crops	Moisture Content	Economic Coefficient	Carbon Absorption Rate
Wheat	12%	0.40	0.485
Rice	12%	0.45	0.414
Maize	13%	0.40	0.471
Millet	12%	0.42	0.450
Sorghum	12%	0.35	0.450
Tubers	70%	0.70	0.423
Cotton	8%	0.10	0.450
Peanut	10%	0.43	0.450
Rapeseed	10%	0.25	0.450
Tobacco	85%	0.55	0.450
Vegetables	90%	0.60	0.450

(2) Undesirable Outputs: Undesirable outputs primarily include agricultural carbon emissions and non-point source pollution.

Following Zhang et al. [61] and Lu et al. [62], Carbon emissions were calculated using the carbon emission factor method, covering six major sources: chemical fertilizers, pesticides, agricultural film, agricultural diesel, plowing, and irrigation (Table 2). Specifically, agricultural carbon emissions are estimated using the following formula:

$$Z=\sum Z_i=\sum \left(K_i\times {\lambda }_i\right)$$

(3)

where $Z_i$ represents the carbon emissions from the i-th carbon source, and $K_i$ and ${\lambda }_i$ denote the quantity and carbon emission coefficient of the i-th carbon source, respectively.

Table 2. Carbon source and carbon emission coefficient.

Carbon Emission Source	Emission Coefficients	Reference
Chemical fertilizers	0.8956 kgC/kg	Oak Ridge National Laboratory, USA
Pesticides	4.9341 kgC/kg	Oak Ridge National Laboratory, USA
Agricultural film	5.180 kgC/kg	Institute of Agricultural Resources and Ecological Environment, Nanjing
Agricultural diesel	0.5927 kgC/kg	IPCC2013
Agricultural Tillage	312.60 kgC/km2	School of Biology and Technology, China
Agricultural irrigation	266.48 kgC/hm2	West et al. [63]

Non-point source pollution in crop cultivation primarily stems from the excessive application of fertilizers, pesticides, and agricultural films. Adopting the inventory analysis framework by Wang et al. [64] and the unit analysis method by Lai et al. [65], this study quantifies pollution levels via fertilizer nitrogen and phosphorus loss, pesticide runoff, and agricultural film residues. Specifically, pollution loads are estimated by multiplying the input quantity of each source by its corresponding coefficient. To ensure calculation accuracy and capture regional heterogeneity, all coefficients are rigorously derived from the First National Pollution Source Census: Handbook of Fertilizer Loss, Pesticide Loss, and Farmland Film Residue Coefficients for Agricultural Pollution Sources.

An evaluation indicator system comprising inputs, desirable outputs, and undesirable outputs was constructed. Furthermore, the Super-efficiency SBM-Undesirable model, based on variable returns to scale, was employed to measure CLGUE. Specific details are presented in Table 3.

Table 3. Evaluation Index System of Cultivated Land Green Utilization Efficiency.

Indicator Category	Variable	Unit	Variable Description/Measurement
Input	Labor Input	person/ha	Number of employees in the primary industry × (Gross output value of planting/Gross output value of agriculture, forestry, animal husbandry, and fishery)/Total cultivated land area
	Irrigation Input	%	Effective irrigated area/Total cultivated land area
	Machinery Input	kW/ha	Total power of agricultural machinery/Total cultivated land area
	Pesticide Input	kg/ha	Pesticide application amount/Total cultivated land area
	Fertilizer Input	kg/ha	Chemical fertilizer application amount/Total cultivated land area
	Agricultural Film	kg/ha	Agricultural film usage amount/Total cultivated land area
Desirable Output	Economic Output	CNY/ha	Gross output value of planting/Total cultivated land area
	Environmental Output	kg/ha	Total carbon sink of cultivated land/Total cultivated land area
Undesirable Output	Carbon Emission	kg/ha	Total carbon emission of cultivated land/Total cultivated land area
	Pollution Emission	kg/ha	Non-point source pollution in crop cultivation/Total cultivated land area

2.2.2. Explanatory Variable

The core explanatory variable, agricultural scale operation, is operationalized as the ratio of land contractual management rights certificates to primary industry practitioners [66]. We employed this specific proxy to capture the consolidation of operation rights induced by labor migration: given that household land entitlements remain relatively rigid, a rising ratio signifies that the diminishing agricultural workforce is managing increasingly consolidated land resources. Furthermore, clear property rights serve as the institutional underpinning for contract stability [67], a mechanism empirically validated to drive physical farm upsizing in China [68].

2.2.3. Mechanism Variable

To precisely capture innovation activities within the agricultural sector, we utilize Agricultural R&D Expenditure as the proxy variable for technological innovation. According to Yu et al. [69], internal agricultural R&D expenditure is a core metric of innovation inputs, as it directly reflects the intensity of resource allocation committed to technological progress in this domain. Moreover, Feng et al. [70] identify agricultural R&D investment as a pivotal mechanism linking external drivers to agricultural green development performance. Consequently, this indicator enables us to effectively delineate the transmission pathway through which scale operation influences CLGUE via technological innovation. To minimize data skewness and mitigate heteroscedasticity, the data were logarithmically transformed.

2.2.4. Threshold Variable

The threshold variable is fiscal support for agriculture, measured by the ratio of expenditure on agriculture, forestry, and water conservancy to total fiscal expenditure.

2.2.5. Control Variables

To minimize omitted variable bias, this paper selects natural disasters, trade openness, industrialization, soil erosion control, and rural per capita disposable income as control variables. The specific measurements are as follows: ① Natural disasters: Measured by the ratio of the total affected area to the total crop sown area. ② Trade openness: Proxied by the ratio of total import and export volume to the regional GDP. ③ Industrialization: Represented by the share of industrial value added in the regional GDP. ④ Soil erosion control: Measured by the area of soil and water loss control. ⑤ Rural per capita disposable income: Obtained directly from statistical yearbooks. ⑥ Specific variable selections and descriptive statistics are presented in Table 4.

Table 4. Descriptive statistics of variables.

Variable Type	Variable Name	Code	N	Mean	Sd	Min	Max
Explained variable	Cultivated land green utilization efficiency	Clgue	480	0.677	0.232	0.296	1.179
Explanatory Variable	Agriculture scale operation	Land	480	0.950	0.496	0.034	3.903
Mechanism	Technological innovation	Tech	480	12.638	1.184	8.766	14.914
Threshold	Fiscal support for agriculture	Fina	480	0.154	0.061	0.029	0.368
Control	Natural disaster	Disa	480	0.176	0.142	0.001	0.696
	Trade openness	Open	480	0.291	0.322	0.008	1.670
	Industrialization	Ind	480	0.348	0.085	0.100	0.574
	Soil erosion control	Soil	480	0.040	0.032	0	0.167
	Rural per capita disposable income	Dpi	480	1.186	0.668	0.249	3.973

2.3. Data Sources

This study utilizes panel data from 30 provinces in China (excluding Tibet, Hong Kong, Macao, and Taiwan due to data unavailability) spanning the period from 2007 to 2022. The data were primarily sourced from the China Statistical Yearbook, China Rural Statistical Yearbook, Annual Statistical Report on Rural Operation and Management in China, China Population and Employment Statistics Yearbook, China Statistical Yearbook on Science and Technology, and various provincial statistical yearbooks. To eliminate the impact of inflation, all monetary variables were deflated to 2007 constant prices. Missing values were imputed using the linear interpolation method.

2.4. Benchmark Model Specification

Since the Cultivated Land Green Utilization Efficiency (CLGUE) values derived from the super-efficiency SBM model are non-negative and truncated, they constitute a limited dependent variable. Applying standard estimation methods to such data may lead to biased estimates. The Tobit model possesses distinct advantages in handling censored data and limited dependent variables. Consequently, this study constructs a Tobit model to address potential inconsistencies in regression estimation and to empirically examine the impact of agriculture scale operation on CLGUE. Furthermore, to investigate the potential non-linear relationship between agriculture scale operation and CLGUE, a quadratic term of agriculture scale operation is incorporated into the model. The benchmark econometric model is specified as follows:

$${\mathit{Clgue}}_{it}={\alpha }_0+{\alpha }_1{\mathit{Land}}_{\mathit{it}}+{\alpha }_2{X}_{\mathit{it}}+{\epsilon }_{\mathit{it}}$$

(4)

$${\mathit{Clgue}}_{it}={\beta }_0+{\beta }_1{\mathit{Land}}_{\mathit{it}}+{\beta }_2{\mathit{Land}}_{\mathit{it}}^2+{\beta }_3{X}_{\mathit{it}}+{\epsilon }_{\mathit{it}}$$

(5)

In Equations (4) and (5), the subscripts i and t denote the province and year, respectively, Clgue_it represents the Cultivated Land Green Utilization Efficiency. Land_it denotes the level of agriculture scale operation, while ${\mathit{Land}}_{\mathit{it}}^2$ is the quadratic term introduced to capture potential non-linear relationships. $X_{it}$ represents a vector of control variables. α₀ and β₀ are the constant terms, and α_k and β_k are the parameters to be estimated. ${\epsilon }_{it}$ represents the random error term.

To further examine the underlying mechanism through which agriculture scale operation influences CLGUE, this study follows the mechanism testing procedure proposed by Jiang [71]. The econometric model is specified as follows:

$${\mathit{Tech}}_{it}={\theta }_0+{\beta }_1{Land}_{it}+{\beta }_2{Land}_{it}^2+{\beta }_3{X}_{it}+{\lambda }_i+{\mu }_i+{\epsilon }_{it}$$

(6)

In the equation, ${\mathit{Tech}}_{it}$ represents the mechanism variable, measuring the level of technological innovation. ${\lambda }_i$ and ${\mu }_t$ denote the province fixed effects and year fixed effects, respectively, and ${\epsilon }_{it}$ is the random error term.

3. Results and Analysis

3.1. Benchmark Regression Analysis

Table 5 presents the regression results regarding the impact of agriculture scale operation on cultivated land green utilization efficiency (CLGUE). Specifically, Column (1) reports the estimates without control variables, while Columns (2) through (6) display the results with the stepwise inclusion of control variables, including the severity of natural disasters, trade openness, industrialization, severity of soil erosion, and rural per capita disposable income. The results across all columns indicate that, irrespective of the inclusion of controls, the coefficients of both the linear and quadratic terms of agriculture scale operation are statistically significant at the 1% level. Notably, the linear term is significantly positive while the quadratic term is significantly negative. This confirms a non-linear, inverted U-shaped relationship between agriculture scale operation and CLGUE. These findings imply that agriculture scale operation initially promotes CLGUE but begins to exert an inhibitory effect after exceeding a certain threshold. Thus, Hypothesis 1 is supported.

Table 5. Benchmark Regression Results.

	(1)	(2)	(3)	(4)	(5)	(6)
Land	0.420 *** (0.043)	0.390 *** (0.045)	0.385 *** (0.044)	0.311 *** (0.047)	0.298 *** (0.048)	0.251 *** (0.054)
Land2	−0.055 *** (0.012)	−0.050 *** (0.012)	−0.053 *** (0.012)	−0.041 *** (0.012)	−0.038 *** (0.012)	−0.035 *** (0.012)
Disa		−0.096 ** (0.045)	−0.082 *** (0.045)	−0.034 (0.046)	−0.023 (0.047)	0.005 (0.048)
Open			−0.164 *** (0.046)	−0.106 ** (0.048)	−0.104 ** (0.048)	−0.065 (0.051)
Ind				−0.544 *** (0.134)	−0.529 *** (0.134)	−0.368 (0.156)
Soil					0.677 (0.530)	0.459 (0.539)
Dpi						0.043 ** (0.021)
Constant	0.341 *** (0.054)	0.381 *** (0.056)	0.435 *** (0.057)	0.654 *** (0.076)	0.629 *** (0.078)	0.555 *** (0.086)
N	480	480	480	480	480	480

Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05.

3.2. Robustness Checks and Endogeneity Tests

3.2.1. Robustness Checks

To verify the reliability of the benchmark results, this study conducted robustness tests using two distinct approaches. First, exclusion of municipalities. Compared to other provinces, the four municipalities (Beijing, Tianjin, Shanghai, and Chongqing) possess distinct advantages in terms of economic development, industrial structure, and infrastructure. These unique characteristics grant their agricultural sectors a specialized status that may introduce potential bias. Consequently, these four municipalities were excluded, and the model was re-estimated using the remaining 416 observations to eliminate the interference of outliers. Second, alternative estimation strategy. To further test the sensitivity of the results, the Two-way Fixed Effects model was employed to re-analyze the data. As presented in Table 6, the regression results from both robustness checks are consistent with the benchmark findings, confirming that the impact of agriculture scale operation on CLGUE is robust and reliable.

Table 6. Endogeneity and Robustness Tests.

Variable	Excluding Municipalities (1)	Alternative Model (2)	IV-Tobit
			Land (3)	Land2 (4)	Clgue (5)
Land	0.235 *** (0.061)	0.242 ** (0.095)			0.506 *** (0.090)
Land2	−0.029 *** (0.014)	−0.033 ** (0.017)			−0.079 * (0.022)
L.land			0.669 *** (0.087)
L.land2				0.517 *** (0.092)
F-statistic			1544.47 ***	1674.07 ***
Controls	Yes	Yes	Yes	Yes	Yes
_Cons	0.544 *** (0.091)	0.569 *** (0.144)	0.237 (0.159)	−0.046 (0.718)	0.372 *** (0.107)
Province fixed effect		Yes
Year fixed effect		Yes
Wald F test			110.87 ***	47.31 ***
Wald test					12.90 ***
N	416	480	450	450	450

Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

3.2.2. Endogeneity Issues

The relationship between agriculture scale operation and CLGUE may suffer from reverse causality, which could lead to simultaneity bias and inconsistent parameter estimates. To mitigate potential endogeneity concerns in the benchmark model, this study employs the Instrumental Variable (IV) approach for robustness testing. Specifically, the one-period lagged terms of the linear and quadratic agriculture scale operation variables were selected as instrumental variables, and the IV-Tobit model was utilized for estimation. The results are presented in Table 6. As indicated by the first-stage regression results in Columns (3) and (4), the regression coefficients of the instrumental variables (L.land and L.land²) are statistically significant at the 1% level, demonstrating a strong correlation between the instruments and the endogenous explanatory variables. Regarding the validity tests, the Wald test rejects the null hypothesis of exogeneity at the 1% level, confirming the presence of endogeneity. Furthermore, the Wald F-statistics significantly exceed the empirical rule-of-thumb value of 10, strongly rejecting the null hypothesis of weak instruments. Most importantly, as shown in Column (5), the coefficients of both the linear and quadratic terms of agriculture scale operation remain significant at varying levels. These findings indicate that the IV-Tobit model passes all validity tests, further verifying Hypothesis 1 and demonstrating the stability of the conclusions. The benchmark regression results remain valid.

3.3. Heterogeneity Analysis

According to the classification standards of the National Bureau of Statistics, the sample was subdivided into the Eastern, Central, and Western regions. The results in Columns (1) and (3) of Table 7 indicate that a significant inverted “U-shaped” relationship exists between agriculture scale operation and CLGUE in the Eastern region, whereas the relationships in the Central and Western regions are not statistically significant. This regional disparity can be attributed to specific geographic and economic conditions. The Eastern region benefits from flat terrain that is conducive to scale expansion and mechanized farming; additionally, farmers there typically possess stronger environmental awareness and value technical training, which effectively improves efficiency. In contrast, although the Central region serves as a major agricultural hub, its heavy reliance on chemical inputs and machinery to maximize agricultural output results in a trade-off between desirable and undesirable outputs, rendering the impact of scale management insignificant. Similarly, the Western region, despite its vast territory, suffers from limited arable land resources, severe land fragmentation, complex topography, and high susceptibility to natural disasters. These factors constrain agricultural production efficiency, thereby neutralizing the potential benefits of scale management on CLGUE.

Table 7. Heterogeneity Regression Results.

Variable	Eastern Region (1)	Central Region (2)	Western Region (3)	Main Grain-Producing Areas (4)	Non-Main Grain-Producing Areas (5)
Land	0.238 *** (0.093)	−0.072 (0.203)	0.478 ** (0.221)	−0.172 (0.185)	0.322 *** (0.087)
Land2	−0.048 *** (0.019)	0.085 ** (0.084)	−0.160 (0.113)	0.107 (0.076)	−0.057 *** (0.018)
Controls	Yes	Yes	Yes	Yes	Yes
_Cons	0.424 *** (0.167)	0.658 *** (0.182)	−0.013 *** (0.155)	0.588 *** (0.145)	0.209 (0.141)
N	208	96	176	208	272

Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05.

Furthermore, based on the Outline of the Medium- and Long-Term Plan for National Grain Security (2008–2020), the sample was divided into main grain-producing areas and non-main grain-producing areas. The results presented in Columns (4) and (5) of Table 7 reveal a distinct disparity: the relationship between agricultural scale operation and CLGUE is statistically insignificant in main grain-producing areas, whereas a significant inverted U-shaped trajectory is observed in non-main grain-producing areas. This paradox in main producing areas can be attributed to the yield-centric rigidity inherent in China’s core grain regions, where the overarching mandate of national food security compels operators to prioritize output stability over input optimization. Consequently, large-scale farmers frequently maintain high-intensity applications of fertilizers and pesticides as an insurance mechanism, neutralizing the potential scale dividend of resource reduction through policy-induced input stickiness. Additionally, the standardized monoculture typical of these areas accelerates soil nutrient depletion, necessitating continued reliance on agrochemicals to sustain productivity. Moreover, as these regions represent the frontier of agricultural intensification, incremental scale expansion tends to increase undesirable outputs, such as non-point source pollution, at a rate that counterbalances marginal economic gains. In sharp contrast, non-main grain-producing areas, less burdened by strict yield targets, initially benefit from diversified crop structures and demand elasticity, allowing the greening effects of scale to manifest; however, as scale exceeds a critical threshold, constraints such as topographical complexity and managerial diseconomies likely set in, leading to the observed decline in efficiency.

3.4. Mechanism Analysis

If agricultural scale operation significantly influences technological innovation, this provides empirical support for the premise that technological innovation acts as a key transmission mechanism affecting CLGUE. Table 8 presents the regression results regarding the impact of agricultural scale operation on the level of technological innovation. The coefficients of both the linear and quadratic terms are statistically significant at the 5% and 10% levels, respectively, indicating that moderate agricultural scale operation effectively fosters technological progress. Theoretically, agricultural technological innovation reduces the reliance on labor, energy, and other resources during land utilization and optimizes the configuration of traditional input factors [72]. Compared to traditional practices, technological innovation facilitates more scientific planning and rational layout of cultivated land resources. This not only enhances local factor utilization efficiency and agricultural productivity but also improves the production environment and elevates the quality of agricultural development [73]. Ultimately, this achieves a dynamic balance between expanding production capacity and ecological protection, thereby enhancing CLGUE.

Table 8. Regression Results of Mechanism Analysis.

Variable	Tech
Land	0.222 ** (0.094)
Land2	−0.042 * (0.022)
Controls	Yes
_Cons	11.074 *** (0.224)
Year fixed effect	Yes
Province fixed effect	Yes
N	480

Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1.

3.5. Threshold Effect Analysis

To further explore the potential non-linear characteristics of the relationship, this study selects fiscal support for agriculture as the threshold variable. Drawing on the panel threshold regression framework, we construct a piecewise function, as shown in Equation (5), to empirically test the threshold effect of agriculture scale operation and estimate the specific threshold values. The model is specified as follows:

$$\begin{array}{c}{C\mathit{lgue}}_{it}={\theta }_0+{\theta }_1{\mathit{Land}}_{it}I\left({\mathit{Fina}}_{it}\le {\sigma }_1\right)+{\theta }_2{\mathit{Land}}_{it}I\left({\sigma }_1<{\mathit{Fina}}_{it}\le {\sigma }_2\right)+\\ {\theta }_3{\mathit{Land}}_{it}I\left({\mathit{Fina}}_{it}>{\sigma }_2\right)+{\theta }_4{X}_{it}+{\epsilon }_{it}\end{array}$$

(7)

where ${\theta }_1$, ${\theta }_2$ and ${\theta }_3$ denote the threshold-specific coefficients, quantifying the influence of agricultural scale operation on the green use efficiency of cultivated land under varying conditions. The terms involving I(·) represent indicator functions that delineate the sample based on the threshold variable, fiscal agricultural expenditure $\left({\mathit{Fina}}_{it}\right)$. Specifically, $I\left({\mathit{Fina}}_{it}\le {\sigma }_1\right)$, $I\left({\sigma }_1<{\mathit{Fina}}_{it}\le {\sigma }_2\right)$ and $I\left({\mathit{Fina}}_{it}>{\sigma }_2\right)$ correspond to the regimes where fiscal support is below the lower threshold ${\sigma }_1$, located between the two thresholds, and above the upper threshold ${\sigma }_2$, respectively.

Before estimating the threshold effects, it is essential to verify the existence and determine the number of thresholds. This study employs the Bootstrap resampling method with 1000 replications, specifying the level of fiscal support for agriculture as the threshold variable under investigation. Table 9 presents the results of these existence tests. The F-statistics for single and double thresholds are statistically significant at the 5% and 1% levels, respectively, whereas the triple threshold effect is insignificant. These findings confirm the presence of a double-threshold effect regarding fiscal agricultural support. The corresponding threshold estimates are reported in Table 10, identified as 0.090 and 0.169. Furthermore, to validate the consistency of these estimates, a Likelihood Ratio (LR) test was conducted, as illustrated in Figure 1. The dashed lines in the figure delineate the 95% confidence interval. It can be observed that at the estimated threshold values of 0.090 and 0.169, the LR statistic approaches zero and falls well within the 95% confidence interval. This indicates that the derived threshold estimates are consistent with the true values.

Table 9. Tests for Panel Threshold Effects.

Threshold Variable	Bootstrap Replications	Thresholds	F	p	10%	5%	1%
Fina	1000	Single	35.10 **	0.012	20.718	25.154	38.677
	1000	Double	28.41 ***	0.009	18.247	22.076	28.290
	1000	Triple	13.64	0.518	28.879	34.867	46.875

Note: Standard errors in parentheses. *** p < 0.01, ** p < 0.05.

Table 10. Panel Threshold Estimates and Confidence Intervals.

Threshold Variable	Thresholds	Threshold Value	95% Confidence Interval
Fina	Single	0.090	[0.089, 0.091]
	Double	0.169	[0.167, 0.170]

Figure 1. Likelihood Ratio (LR) Statistics for Threshold Estimates. Note: The dashed lines indicate the critical values corresponding to the 95% confidence interval.

Table 11 presents the threshold regression results, confirming that the influence of agricultural scale operation on the green use efficiency of cultivated land (CLGUE) is non-linear and contingent upon the level of fiscal agricultural support. Below the first threshold (${\mathit{Fina}}_{it}$ $\le$ 0.090), the impact of agricultural scale operation on CLGUE is statistically insignificant. This indicates that large-scale farming may still be in a nascent stage of development, where its potential to enhance efficiency has not yet materialized due to inadequate external funding. A significant positive relationship emerges in the intermediate regime ($0.090<{\mathit{Fina}}_{it}\le 0.169$), with a regression coefficient of 0.156 (p < 0.05). Within this specific interval, moderate government intervention fosters an enabling environment for farm scale expansion through infrastructure provision, policy guidance, and targeted subsidies. This supportive context facilitates the adoption of advanced agricultural technologies and managerial expertise, thereby significantly enhancing CLGUE. In the regime where ${\mathit{Fina}}_{it}>0.169$, the promoting effect of agricultural scale operation on the green use efficiency of cultivated land diminishes considerably, with the coefficient dropping to 0.051. This attenuation suggests that as fiscal support levels rise, large-scale farming gradually enters a maturity stage or approaches market saturation, leading to the onset of diminishing marginal returns. Furthermore, excessive fiscal subsidies risk distorting market incentives, potentially influencing operators to prioritize yield maximization over ecological protection, which ultimately weakens the positive impact on efficiency.

Table 11. Regression Results of the Panel Threshold Model.

Variable	Green Use Efficiency of Cultivated Land
	Fina ≤ 0.090	0.090 < Fina ≤ 0.169	Fina > 0.169
Land	0.023 (0.079)	0.156 ** (0.068)	0.051 * (0.029)
Controls	YES	YES	YES
Year fixed effect	YES	YES	YES
Province fixed effect	YES	YES	YES
N	480	480	480

Note: Standard errors in parentheses. ** p < 0.05, * p < 0.1.

4. Discussion

4.1. Comparison with Previous Studies

This study empirically confirms a significant inverted U-shaped nonlinear relationship between scale operation and CLGUE. This finding integrates and refines two seemingly contradictory mainstream viewpoints in the existing literature.

On the one hand, the results support the arguments of Frank et al. [2] and Wu et al. [4] regarding the positive environmental effects of scale operation. Specifically, within a moderate scale range, land contiguity can indeed achieve “reduction and efficiency enhancement” by optimizing factor allocation and reducing the per-unit intensity of chemical fertilizers and pesticides. On the other hand, the trajectory on the right side of the inverted U-shaped curve validates the concerns of some scholars [5,6]. Once the scale expands excessively beyond a specific threshold, operating entities—driven by risk aversion and the need to reduce management complexity—tend to increase chemical inputs as a substitute for labor-intensive management, thereby leading to a decline in green efficiency. Distinct from existing studies that focus solely on the U-shaped relationship of chemical inputs [9] or the inverted U-shaped relationship of agricultural green total factor productivity [10], this paper constructs a comprehensive evaluation system incorporating carbon emissions and non-point source pollution based on Tone’s [56] Super-SBM-Undesirable model. Theoretically, this proves that the environmental dividends of scale operation are not released linearly; rather, there exists a dynamic trade-off interval between “economic output growth” and “environmental cost internalization,” thereby expanding the ecological connotation of appropriate scale operation in agriculture.

Furthermore, a significant theoretical contribution of this study, beyond the descriptive results, lies in revealing the “threshold effect” of fiscal support for agriculture policy within the green effects of scale operation. While existing literature often treats fiscal policy as an independent influencing factor, this study identifies a significant double-threshold effect. Moderate fiscal support can reinforce the green spillovers of scale operation by alleviating financial constraints and promoting technology adoption. However, excessive fiscal subsidies (exceeding 0.169) may distort factor price signals, leading to “over-capitalization” and resource waste, which conversely weakens green efficiency. This finding theoretically clarifies the functional boundaries between “government support” and “market mechanisms” in agricultural green development. It explains why the environmental performance of scale operation varies significantly under different policy intensities, providing a crucial theoretical basis for optimizing agricultural subsidy policies.

4.2. Policy Recommendations

First, implement differentiated scale management strategies based on regional functional positioning. According to our heterogeneity analysis in Table 7, the Eastern region and non-main grain-producing areas exhibit a significant inverted U-shaped relationship. This finding indicates the existence of an efficiency ceiling for scale expansion. Therefore, local governments in these zones must establish a dynamic monitoring mechanism to strictly control scale within the increasing returns stage. Conversely, our results show a statistically insignificant relationship in main grain-producing areas. As discussed in Section 3.3, we attribute this null result to rigid yield constraints and high input dependence. Addressing this challenge requires a paradigm shift from physical land consolidation to “service scale operation.” By incentivizing socialized agricultural services, policy can break the path dependence of chemical overuse and unlock the green potential of these core regions.

Second, accelerate the digital transformation of agriculture to activate the technological innovation engine. Our mechanism analysis in Table 8 empirically confirms that technological innovation is the key transmission channel linking scale operation to CLGUE. However, the diminishing marginal returns observed in this mechanism suggest that the dividend from traditional mechanization is tapering off. Therefore, policy interventions must focus on unclogging this specific pathway. Instead of subsidizing traditional material inputs, fiscal resources should be redirected toward intelligent technologies, including agricultural big data, IoT, and variable-rate fertilization drones. By reducing the marginal cost of adopting green technologies, the “innovation effect” of scale operation can be fully maximized. This strategy ensures that land consolidation effectively translates into precise, automated, and low-carbon production models.

Third, optimize the allocation of fiscal funds strictly according to scientific efficiency intervals. Our threshold regression in Table 10 and Table 11 reveals that the positive enabling effect of fiscal support is maximized only within the specific interval of 0.090 to 0.169. This finding provides a precise quantitative benchmark for fiscal reform. Local governments must avoid the extremes of insufficient support below 0.090 or excessive subsidization above 0.169. For regions below the lo wer threshold, fiscal input should be increased to trigger the catalytic effect in green infrastructure development. However, for regions exceeding the 0.169 limit, strict budget constraints and auditing mechanisms must be introduced. These measures are essential to prevent policy dependency and resource misallocation, ensuring that support intensity remains within this optimal efficacy window.

4.3. Limitations and Future Research

Despite the theoretical and practical contributions of this study, three main limitations should be acknowledged.

First, the spatial scale of this study is constrained by data availability. The measurement of Cultivated Land Green Utilization Efficiency (CLGUE) was conducted at the provincial level. Given China’s vast geographic expanse, this macro-level aggregate approach aligns with the “Modifiable Areal Unit Problem” in spatial econometrics, which inevitably smooths out local variations. This aggregation may mask significant intra-provincial heterogeneity in natural resource endowments and input–output redundancies that would otherwise be observable at the county or village level. Consequently, the study may not fully capture the nuanced behavioral responses of individual farmers to scale expansion, which are often driven by micro-level incentive mechanisms.

Second, regarding the estimation of agricultural carbon sinks, this study employed the correlation coefficient method based on crop yields. We explicitly acknowledge that relying on fixed economic coefficients and carbon absorption rates for nationwide calculations constitutes a methodological limitation. Due to significant climatic differences across China, such as the contrast between subtropical and temperate zones, the use of uniform coefficients may introduce estimation bias in specific regions compared to plot-level field measurements. However, in the absence of continuous and authoritative official data for region-specific parameters across all provinces, utilizing standardized coefficients remains the prevailing approach in macro-level empirical research to ensure cross-provincial comparability.

Third, the generalizability of our findings is contextualized by China’s unique institutional setting. This study is deeply embedded in the Chinese context, characterized by collective land ownership and the urban–rural dual land system. Under these institutional arrangements, the mechanisms linking scale operations to green efficiency may differ from those in market economies with private land ownership. Therefore, caution should be exercised when extending these conclusions to countries with different land tenure regimes.

Future research should prioritize three key dimensions. First, to address spatial scale limitations, future studies should downscale the analysis to the county level and employ specific spatial econometric models, including the Spatial Durbin Model or Geographically Weighted Regression, to capture micro-level spatial spillovers and local heterogeneity more precisely. Second, to enhance environmental accounting precision, future research should integrate multi-source remote sensing data from platforms like MODIS or Landsat with process-based algorithms such as the CASA model. This integration would achieve pixel-level precision in carbon sequestration estimation, thereby replacing unified parameters. Third, regarding generalizability, future research could conduct comparative studies between China and other transition economies like Vietnam or market economies such as the United States to isolate the specific impact of land tenure institutions on the scale–efficiency relationship.

5. Conclusions

Based on China’s provincial panel data from 2007 to 2022, this paper empirically investigates the impact of agricultural scale operation on the green use efficiency of cultivated land (CLGUE) and its transmission mechanisms. Furthermore, it analyzes the threshold effect of the level of fiscal agricultural support within this influence process. The main conclusions are as follows:

(1) The relationship between agricultural scale operation and CLGUE is not a simple linear synergistic one. While the expansion of agricultural scale operation significantly improves CLGUE, it exhibits characteristics of diminishing marginal returns. (2) Heterogeneity analysis indicates a significant inverted “U”-shaped relationship between agricultural scale operation and CLGUE in eastern regions and non-major grain-producing areas. However, the relationship is not statistically significant in central and western regions and major grain-producing areas. (3) Mechanism analysis reveals that expanding agricultural scale operation promotes technological innovation, which in turn improves CLGUE, although this mediating effect also shows diminishing marginal returns. (4) There exists a significant threshold effect of fiscal agricultural support in the process by which agricultural scale operation influences CLGUE. A moderate level of fiscal support effectively unlocks the synergistic effects of scale operation and significantly enhances CLGUE. Conversely, fiscal support levels that are either too high or too low may weaken the effectiveness of green cultivated land development.