How Do Digitalization and Scale Influence Agricultural Carbon Emission Reduction: Evidence from Jiangsu, China

Abstract

In order to alleviate the constraints of global warming and sustainable development, digitalization has made significant contributions to promoting agricultural carbon reduction through resources, technology, and platforms. Under this situation, China insists on developing agricultural scale management. However, what impact will scale management in agricultural digital emission reduction have on mechanisms and pathways? Based on three rounds of follow-up surveys conducted by the Digital Countryside Research Institute of Nanjing Agricultural University in Jiangsu Province from 2022 to 2024, in this study a total of 258 valid questionnaires on the rice and wheat industry were collected. Methods such as member checking and audit trail were employed to ensure data reliability and validity. Using econometric approaches including Tobit, mediation, and moderation models, this study quantified the Scale Management Level (SML), examined the mechanism pathways of digital emission reduction in a scaled environment, further demonstrated the impact of scale management on digital emission reduction, and verified the mediating and moderating effects of internal and external scale management. We found that: (1) In scale and carbon reduction, the SBM-DEA model calculates that the scale of agricultural land in Jiangsu showed an “inverted S” trend with SML and an “inverted W” trend with the overall agricultural green production efficiency (AGPE), and the highest agricultural green production efficiency is 0.814 in the moderate scale range of 20–36.667 hm². (2) In digitalization and carbon reduction, the Tobit regression model results indicate that Network Platform Empowerment (NPE) significantly promotes carbon reduction (p < 1%), but its squared terms exhibit an inverted U-shaped relationship with agricultural green production efficiency (p < 1%), and SML is significant at the 5% level. From a local regression perspective, the strength of SML’s impact on the three core variables is: NPE > DRE > DTE. (3) Adding scale in agricultural digital emission reduction, the intermediary mechanism results showed that the significant intensity (p < 5%) of the mediating role of Agricultural Mechanization Level (AML) is NPE > DTE > DRE, and that of the Employment of Labor (EOL) is DRE > NPE > DTE. (4) Adding scale in agricultural digital emission reduction, the regulatory effect results showed that the Organized Management Level (OML) and Social Service System (SSS) significantly positively regulate the inhibitory effect of DRE and DTE on AGPE. Finally, we suggest controlling the scale of land management reasonably and developing moderate agricultural scale management according to local conditions, enhancing the digital literacy and agricultural machinery training of scale entities while encouraging the improvement of organizational level and social service innovation, and reasonably reducing labor and mechanization inputs in order to standardize the digital emission reduction effect of agriculture under the background of scale.

1. Introduction

Realizing digital effective carbon emission reduction is the endogenous need of high-quality agricultural development in China, which is of great significance to alleviating global warming and promoting sustainable agricultural development. Agriculture, as the second largest carbon source in the world, has the property of acting as a carbon sink [1]. With its rapid scale development, high inputs such as pesticides, fertilizers, agricultural machinery, and high emissions of livestock manure have increased agricultural carbon emissions. The 2023 Low Carbon Development Report of China Agriculture and Rural Areas showed that the total carbon emissions of agriculture in China are 828 million tons of carbon dioxide equivalent, accounting for 6.7% of national carbon emissions, and the agricultural “carbon deficit” (emission–absorption > 0) will exist for a long time [2]. In order to solve the problem of resource and environment constraints and achieve sustainable development, in 2022, China put forward the strategic plan of “actively and steadily promoting carbon neutrality in peak carbon dioxide emissions”, and has contributed more than 10% of emission effects to various industries with the help of digital models [3,4]. However, environmental issues arising from non-agriculturalization, non-grain orientation, and intensive large-scale production have posed serious challenges to the green and low-carbon transformation of agriculture. These practices affect the synergy between agricultural carbon sequestration and digital development [5,6], and result in an inverted U-shaped relationship between grain yield and sown area, as well as between farm size and the level of agricultural low-carbon development level [7,8]. Collectively, these factors have exacerbated the pressure on agricultural carbon emission reduction [9]. Therefore, it is urgent to find the best way to develop agricultural scale management. In-depth research on scale in agricultural digital emission reduction will become an important issue that urgently needs to be solved to promote future agricultural mechanization, digitization, standardization, industrialization, and green development.

Against this backdrop, the academic community has been conducting long-term research and exploration on digitization, scaling up, and carbon reduction. Existing scholars have conducted in-depth research on carbon emission reduction of agriculture digitalization, and explored the mechanism of and nonlinear relationship between digitization and carbon emissions [10,11], which proved that agriculture digitalization can significantly inhibit the intensity of carbon emissions [12,13], and its technology spillover will reduce the intensity of carbon emissions, which is persistent. Differences in the usage of digital technology will have an impact on environmental regulation and carbon emissions [14]. Some scholars have also found that digital technology significantly reduces the utilization rate of pesticides and fertilizers in rice production, especially among large-scale and part-time farmers [15]. About research on scale and carbon emissions, scholars [16] verified that the scale of agricultural land has a positive impact on the adoption of green technologies, and proved that socialized services can break the labor constraints by promoting the scale effect, thus reducing the intensity of agricultural carbon emissions [17], and the scale of planting land in major rice and wheat producing areas in China has a threshold effect on agricultural carbon emissions. Among them, the change of carbon emissions from chemical fertilizers is the main one [8,18], and the expansion of agricultural land management scale significantly increased foreign exchange emissions [19]. About digitalization and agricultural scale, some scholars have found that farmland circulation promotes the development of digital agriculture by transferring labor [20], and the popularization of digital agricultural technology has a greater impact on small-scale farmers’ adoption of green protection and control technology [21], and land scale management and service could positively affect agricultural production efficiency [22]. Some scholars also found that agricultural digitalization can improve fertilizer utilization efficiency by expanding the scale of farmland management, promoting the development of agricultural socialized services, and using green technology [23]. Farmland circulation has a significant positive mediating effect between digitalization and income of large-scale farmers, and has a significant relationship with agricultural low carbon [24,25]. Furthermore, studies indicate that moderate-scale management at the farmer level is generally increasing, showing no strong correlation with plot size but being significantly influenced by factors such as scale, quantity, distance, and land fragmentation [20,26]. Separately, social services have been found to break labor constraints by leveraging scale effects, thereby reducing agricultural carbon emission intensity [17].

Based on the existing research results mentioned above, the academic community still needs to conduct deeper explorations in the following areas: (1) In theory, most studies only focus on the bilateral relationship between digitization and carbon reduction, scale and carbon reduction, or digitization and scale. It is worth noting that there is a lack of research that integrates all three aspects into a comprehensive framework. In addition, there are few specific discussions on the digital emission reduction effects of agriculture under the background of scale, and there are in-depth discussions on the trend brought by the square term. (2) Methodologically, DEA or Topsis are widely used to measure the level of scale management. However, there is relatively limited research on the appropriate level of agricultural scale management, and even fewer studies examine it under low-carbon constraints. (3) In terms of data sources, the majority of analyses rely on macro-level data such as statistical yearbooks. In this context, findings derived from micro-level survey data could, to some extent, corroborate existing macro-level patterns. Therefore, the research on digitalization and scale reduction of agricultural carbon emissions is urgently needed for current academic achievements, further filling the existing research gap and exploring the role of scale reduction in agricultural digital emission reduction.

Based on the existing research deficiencies, the marginal contributions of this article mainly manifest in the following aspects: (1) This study innovatively incorporates “carbon” into the measurement system of scale management level. Based on the entropy weight method and practical survey data, both scale efficiency and scale frequency were taken into account in the calculation. (2) Based on field survey data, it classifies the managemental land scale intervals in Jiangsu Province and systematically reveals the evolving relationships between land scale, scale management level, and agricultural carbon reduction. (3) Using the Tobit model, the relationship between digitalization and carbon reduction under scale management was examined, and the impact of digitalization on agricultural carbon emission intensity and the influence of scale management level as an environmental variable were further studied. (4) This study analyzes the mechanisms and pathway effects of both internal and external scale management in the process through which digitalization influences carbon reduction. Building on this foundation, the primary objective of this study is to construct an analytical framework to examine how digitalization affects carbon reduction within the context of agricultural scale management. Further research was conducted on the impact of digitization on agricultural carbon emission intensity and the role of scale management in agricultural digital emission reduction, verifying the mediating and regulatory mechanisms played by internal and external scale management. The findings aim to provide evidence-based support for the policy formulation of China’s agricultural scale transformation and digital decarbonization.

2. Theoretical Framework and Research Assumptions

2.1. Theoretical Framework Analysis

This study constructed a theoretical framework from the perspectives of economies of scale and technological innovation to examine the mechanisms and pathways through which digitalization and scale affect agricultural carbon reduction, as shown in Figure 1. In the context of continuously promoting agricultural scale management in Jiangsu Province, this framework aims to clarify the inherent logical relationship between digitization, scale and carbon reduction, and explore how digitization in this environment affects agricultural carbon reduction, and further understand the impact of scale management level on agricultural digital emission reduction, as well as the intermediary and regulatory mechanisms of internal and external scale management in agricultural digital emission reduction.

Based on the Figure 1, we established the theoretical basis of this study as follows:

Based on the characteristics of the surveyed subjects, this study is situated within the context of agricultural scale management to investigate the internal logical relationships among digitalization, scale, and carbon reduction.

First, agricultural scale management in China is steadily advancing. Under these circumstances, however, several challenges have emerged: (1) The disruption of organic material cycles and the continued use of chemical pesticides and fertilizers have led to a factor locking effect in agricultural development. (2) Scale diseconomies and over-reliance on agricultural machinery have resulted in a mechanical substitution effect. (3) Declining biodiversity and large-scale monoculture farming have contributed to a structural convergence effect in agriculture. (4) Regional disparities in Jiangsu, including locational advantages, geographical conditions, production materials, and equipment infrastructure have further led to heterogeneity in regional resource differences.

Second, Digital Empowerment (Data Resource Empowerment, Digital Technology Empowerment, Network Platform Empowerment) driven appropriately scaled agricultural management by technological innovation, thereby agricultural carbon emissions can be achieved by improving technology utilization efficiency. Most existing research has confirmed that digitalization promotes agricultural carbon reduction. In the context of scale management, we need to further explore how digitalization specifically affects agricultural carbon reduction [27,28,29].

Furthermore, Internal Scale Management (Agricultural Mechanization Level, Employment of Labor) regulates the impact of digitization on scale through technology transmission mechanism, while internal scale managements will transmit the effects of digitization to carbon emissions through factor production efficiency. At this point, internal scale managements mediate the path relationship between digitization and carbon reduction. In this process, the extensive use of pesticides and fertilizers by scale entities, as well as their dependence on machinery, will also affect carbon emissions [30,31]. At this point, controlling moderate scale managements through marginal effect of scale may bring unexpected effects. The mediating role of internal scale management between digitalization and carbon reduction requires further empirical investigation.

Lastly, External Scale Management (Organized Management Level, Socialized Service System) regulates the relationship between digitization and scale through technology diffusion mechanism, and affects the relationship between scale and carbon reduction by improving resource allocation efficiency, thereby regulating the path relationship between digitization and agricultural carbon reduction. At this time, under factor agglomeration effect, moderate scale management may have unexpected effects. Through production, management, and managemental activities, as well as rational resource utilization, large-scale managements indirectly affect surrounding farms. Moreover, large-scale contiguous farming and regional resource heterogeneity may also influence agricultural carbon emissions [32]. Therefore, the regulatory role of external scale management between digitalization and carbon reduction warrants in-depth study.

2.2. Research Hypothesis

Hypothesis 1.

Under the background of large-scale operation, digital empowerment may promote agricultural carbon reduction.

The impact of digital empowerment on agricultural carbon emissions stems primarily from investments in data, technology, and platform resources related to agricultural production, management, and managements. Existing academic research generally proves that digitalization significantly inhibits the intensity of agricultural carbon emissions [33,34], effectively promote agricultural carbon emission reduction [35], thus promoting the realization of the double carbon target [36]. There is a nonlinear relationship between digitalization and agricultural carbon emissions [11,37]. The impact of digital application on agricultural green production efficiency will be different under different land scales. Small-scale management makes the effect of digital emission reduction slow, while large-scale management has greater investment and risk, so it is necessary for us to control an optimal land scale. However, under scale managements, how digitalization will affect agricultural carbon reduction is a key issue that we need to discuss next.

Hypothesis 2.

By enhancing the input of internal scale management, the digitalization may have a negative impact on agricultural carbon reduction through mediating internal scale management.

The influence of internal scale management on agricultural digital emission reduction mainly lies in the effect of resources such as labor and machinery invested in agricultural production and management, and further affects agricultural carbon emission by changing the input of labor and machinery. Digital development has a positive impact on improving land circulation [38], and the expansion of land management scale will significantly increase foreign exchange emissions [19]. The expansion of cultivated land and the promotion of green technology will contribute to carbon reduction [39]. The labor transfer plays an intermediary role between digitization and agricultural sustainable development [40]. Mechanization not only has a positive impact on agricultural carbon reduction, but existing scholars have shown that mechanization may hinder the achievement of carbon neutrality [41,42].

Hypothesis 3.

By strengthening the external scale management, the digitization may have a positive impact on agricultural carbon reduction by regulating external scale management.

The influence of external scale management of agriculture on agricultural digital emission reduction mainly lies in the effect of organized management and socialized service in the process of agricultural production and management, through the regulation of large-scale management and the investment of scale economy, the consumption of chemical inputs is reduced, thus further reducing agricultural carbon emissions. The cooperative management can significantly improve the efficiency of agricultural land use [43]. Social services can support the reduction of agricultural chemical fertilizers [44], and effectively improve agricultural total factor productivity [45,46]. Moreover, socialized services can break labor constraints by promoting scale effect, thus reducing the intensity of agricultural carbon emissions [17].

3. Methods and Data

3.1. Research Methods

The method mainly includes two aspects: firstly, using the SBM-DEA model of unexpected output through the MaxDEA 8 software to measure the agricultural green production efficiency under different land scales, and the calculation method of scale management level under carbon constraints was designed by economies of scale and technological innovation. Second, we explored the agricultural digital emission reduction mechanism under the environment of scale management level. The effect of digital empowerment on agricultural carbon emission reduction was quantified using Tobit model through software Stata 17 under the influence of scale management level. Finally, we explored the intermediary mechanism of internal scale management through the intermediary effect model, and the regulation mechanism of external scale management by using the regulation model in agricultural digital emission reduction.

3.1.1. SBM-DEA Model of Unexpected Output

Based on the perspectives of scale effects and technological innovation, this study selects the SBM-DEA model of unexpected output (a derivative of traditional data envelopment analysis) [47,48] to estimate the explained variable. Unlike conventional radial CCR or BCC models, which struggle to incorporate undesirable outputs directly, the SBM-DEA model of unexpected output explicitly integrates undesirable outputs, into its analytical framework. Given that agricultural carbon emissions are treated as an undesirable output in this research, the SBM-DEA model of unexpected output provides a more accurate and appropriate measure of agricultural green production efficiency. The specific formulation of the model is presented below.

$$\begin{gathered} AGPE\ast =\mathit{min}{\displaystyle \frac{1-\frac{1}{m}{\Sigma }_{i=1}^{\eta }\frac{S_{io}^{-}}{x_{io}}}{1+\frac{1}{s}\left({\Sigma }_{r=1}^{S_1}\frac{S_r^g}{y_{ro}^g}+{\Sigma }_{r=1}^{S_2}\frac{S_r^b}{y_{ro}^b}\right)}} \\ Subjectto \\ \left\{\begin{array}{l}{x}_o=X\lambda +s^{-}\\ y_o^g=Y\lambda -s^g\\ y_o^b=Y\lambda +s^b\\ s^{-},s^g,s^b,\lambda \ge 0\end{array}\right. \end{gathered}$$

(1)

In the formula (Equation (1)), AGPE represents the value of agricultural green production efficiency, $m$, s₁, s₂ represent the index numbers of input, expected output and unexpected output respectively, s represents the slack of input and output, s⁻, s^g, s^b represent the input redundancy, insufficient expected output and excessive expected output respectively, and $\lambda$ represents the weight vector. In the judgment of efficiency value, when $\rho <1$ it means that the efficiency of the production unit is invalid, and the production efficiency of the production unit needs to be improved and optimized. When $\rho \ge 1$, the efficiency of the production unit is effective, and the larger the value, the higher the efficiency value.

3.1.2. Calculation and Analysis of Agriculture Scale Management Level

Based on the former research and combined with the practical experience during the investigation, this study designed the measurement method of the agriculture scale level from the perspective of managemental efficiency. The product of scale efficiency (SE, derived from calculation results of DEA model) and scale frequency (ST) is emphasized to represent the Scale Management level (SML) with carbon constraints.

$$ST\left(n\right)=S\left(n\right)\times P_r\left(n\right)/C_v\left(n\right)$$

(2)

In the formula (Equation (2)), s is the scale size, $P_r$ is the scale proportion, $C_v$ is the median value of the scale interval group, and n is the sample number.

$$SML\left(n\right)=ST\left(n\right)\times SE\left(n\right)$$

(3)

In the formula (Equation (3)), SML stands for scale management level with carbon constraints, SE stands for scale efficiency, and ST stands for scale frequency, and n is the sample number.

3.1.3. Measurement Model

First of all, this study takes all components of digital empowerment (DE) as explanatory variables (data resource empowerment, digital technology empowerment, network platform empowerment) and agricultural green production efficiency (AGPE) as explained variables. Due to the consideration of the decomposition and quantification of marginal effects, consistency in unbiased coefficient estimation, as well as the rationality of data boundaries and the reliability of assumed results, this study constructs a Tobit benchmark model to evaluate the effectiveness and robustness of agricultural digital emission reduction. The form of this model is as follows.

$${AGPE}_{\dot{i}}=\left\{\begin{array}{c}{AGPE}_{\dot{i}}\ast =0,{AGPE}_{\dot{i}}\ast \le 0\\ {AGPE}_{\dot{i}}\ast ={\alpha }_i+{\beta }_{i}DE+{\epsilon }_i,0<{AGPE}_{\dot{i}}\ast \le 1,i=1,2,\dots ,n\\ {AGPE}_{\dot{i}}\ast =1,{AGPE}_{\dot{i}}\ast >1\end{array}\right.$$

(4)

In the formula (Equation (4)), ${AGPE}_{\dot{i}}$ represents the agricultural green production efficiency value, ${AGPE}_{\dot{i}}$* represents the truncated dependent variable, ${\alpha }_i$ represents the intercept term, ${\beta }_i$ represents the correlation coefficient, ${DE}_i$ represents the components in digital empowerment, and ${\epsilon }_i$ represents the random error term, and it obeys the normal distribution N (0, ${\sigma }^2$).

Secondly, on the basis of the above analysis, this study examines whether the factors in digital empowerment can achieve effective carbon reduction in agriculture by increasing the factors in internal scale management (ISM) through the intermediary effect model. The factors of internal scale management (agricultural mechanization level, employment of labor) are used as intermediary variables, and the step-by-step method is combined with Sobel-Goodmen method to test the significance of the intermediary effect.

$$\begin{gathered} {AGPE}_{\dot{i}}={\alpha }_0+{\alpha }_1{DE}_i+{\alpha }_2{Control}_i+{\epsilon }_1 \\ {ISM}_{\dot{i}}=b_0+b_1{DE}_i+b_2{Control}_i+{\epsilon }_2 \\ {AGPE}_{\dot{i}}=c_0+c_1{DE}_i+c_2{ISM}_{\dot{i}}+c_3{Control}_i+{\epsilon }_3 \end{gathered}$$

(5)

In the formula (Equation (5)), ${AGPE}_{\dot{i}}$ represents the agricultural green production efficiency of the I-th new business entity with scale management level, ${DE}_i$ represents the input of digital enabling factors (including data resource empowerment, digital technology empowerment, network platform empowerment), ${ISM}_{\dot{i}}$ represents the internal scale management and use of the I-th new business entity with scale management level (including agricultural mechanization level, employment of labor), and ${Control}_i$ represents other control variables.

Finally, on the basis of the formula (Equation (5)), this study examined whether the digital empowerment can achieve effective carbon reduction in agriculture by regulating the factors of external scale management (ESM, such as organized management level and socialized service system) taken as regulatory variables, and the regulatory effect model is set based on the verification of the intermediary model to test the significance of the regulatory effect.

$${AGPE}_{\dot{i}}=d_0+d_1{DE}_i+d_2{ESM}_i+d_3{{DE}_i\times ESM}_{\dot{i}}+d_4{Control}_i+{\epsilon }_4$$

(6)

In the formula (Equation (6)), ${ACEE}_{\dot{i}}$ represents the agricultural green production efficiency of the I-th new business entity with scale management level, ${{DE}_i\times ESM}_i$ represents the interaction between digital empowerment and external scale management and ${Control}_i$ represents other control variables.

3.2. Selection and Treatment of Variables

3.2.1. Agricultural Green Production Efficiency as Explained Variable

(1)

Calculation of agricultural carbon emissions

Because of the radial relationship between carbon absorption and crop yield in

Table 1. Carbon emission coefficient to adopt.

Carbon Emission Source	Corresponding Index Name	Carbon Emission Coefficient	Reference Source
Chemical fertilizer	Fertilizer application rate	0.8956 kgC/kg	T.O.West, Oak Ridge National Laboratory, USA
Pesticide	Pesticide consumption	4.934 kgC/kg	T.O.West, Oak Ridge National Laboratory, USA
Agricultural plastic sheeting	Usage of agricultural plastic film	5.18 kgC/kg	IREEA Nanjing Agricultural University Institute of Resources and Ecological Environment
Diesel	Consumption of agricultural diesel oil	0.5927 kgC/kg	IPCC United Nations INPErgovernmental Panel of Experts on Climate Change
Turn over	Sowing area of grain crops	3.126 kgC/hm2	CABCAU College of Agriculture and Biotechnology, China Agricultural University
Irrigate	effective irrigation area	266.48 kgC/hm2	Duan, H. et al. [51]

, the carbon emission of this study mainly considered crop production processes, mainly include the carbon emissions produced by the consumption of agricultural materials such as fertilizers, pesticides, agricultural films and diesel oil [49,50], as well as the loss of organic carbon caused by ploughing and the carbon emissions caused by irrigation. The absolute consumption of carbon emission sources obtained through social research is multiplied by their respective carbon emission coefficients, and the multiplied carbon emissions are summed up to obtain agricultural carbon emissions. The specific carbon emission coefficients are as follows.

The specific calculation formula is as follows:

$$Ace=\sum {Ace}_i=\sum C_i\times {\sigma }_i$$

(7)

In the formula (Equation (7)), $Ace$ is agricultural carbon emission, ${Ace}_{\mathrm{i}}$ is agricultural carbon emission caused by carbon emission source i, $C_i$ and ${\sigma }_i$ are absolute consumption and carbon emission coefficient of carbon emission source i.

(2)

Calculation of agricultural green production efficiency

This study comprehensively designed the input–output index system to calculate agricultural green production efficiency (AGPE) that fits this study [52]. Input indicators include Land Inputs, Agricultural Inputs and Machinery Inputs, and output indicators include expected output and unexpected output. Grain yield and Agricultural output value are regarded as expected outputs to reflect the social and economic benefits generated by crop planting activities, and agricultural carbon emissions are regarded as unexpected outputs to reflect the negative environmental impact brought by crop planting activities. SBM-DEA model with unexpected output is selected to measure the agricultural green production efficiency, and the specific expression is shown in

Table 2. Design of input–output index system for measuring agricultural green production efficiency.

Indicator Name	Indicator Division		Specific Indicators	Measured Indicator Variable
Agricultural green production efficiency	Input index		Land input	Scale of land management
			Agricultural input	Sum of inputs of agricultural production materials (pesticides, fertilizers, films)
			Mechanical input	The sum of input costs for agricultural owned machinery and leased machinery
	Output index	Expected output [53,54]	Grain yield	The total amount produced by rice and wheat cultivation in that year
			Grain output value	The economic benefits obtained from rice and wheat cultivation in the past
		Unexpected output [55]	Carbon emissions	Calculated from each indicator variable in Table 1

.

(3)

Calculation of agricultural carbon emission intensity

Agricultural carbon emission intensity = total agricultural carbon emissions/total agricultural output value.

3.2.2. Digital Empowerment Taken as Core Explanatory Variables

Regarding the construction of a digital indicator system as the core explanatory variable, this study starts from the essential connotations and characteristics of each indicator, and combines theories such as agricultural economics, geographical environment, and human behavior to design a digital weighting (DE) factor indicator system [56,57]. The digital empowerment is constructed from three variables in

Table 3. Index System design of digital empowerment.

Primary Index	Secondary Index	Three-Level Index	Indicator Description
Digital empowerment	Data resource empowerment	Resource integration ability	Product online sales ability, agricultural online purchase ability, and access to business information.
		Data sharing level	Order channel push effect, expert resource push effect and four new technologies push effect.
	Digital technology empowerment	Online perception level	Perception of production environment, monitoring of agricultural productivity, and visualization of production process.
		Fine management level	Grid management level, refined management level, remote control level, automatic execution level, etc.
		Intelligent decision-making level	Ambient intelligence’s early warning ability, process intelligent diagnosis ability and production intelligent decision-making ability.
	Network platform empowerment	Application of digital platform for industrial chain	Digitization of enterprise-driven model, cooperative comanagement model and broker-driven model.
		Business Support Digital Platform Services	Service level of financial digital platform, insurance digital platform and training digital platform.
		Application of agricultural machinery service digital platform	Agricultural machinery dispatching service level, technical guidance service level, and technical achievement display level.
		Supervise the application of digital platform	Agricultural input management ability, product quality traceability management ability

: Data Resource Empowerment, Digital Technology Empowerment, and Network Platform Empowerment, and the contribution of digital empowerment in three dimensions is studied. Finally, we perform weighted averaging on the selected indicators to remove the final values and obtain the variable data required for this study.

3.2.3. Internal Scale Management Taken as Intermediate Variable

The index system of scale management is constructed, this article starts from the essential connotation and characteristics of each indicator, and combines theories such as agricultural economics, geographical environment, and human behavior to comprehensively design an internal indicator system for scale management [16,18]. This study takes Internal Scale Management (ISM) as an intermediary variable, included two dimensions in

Table 4. Design of index system for internal scale management.

Primary Index	Secondary Index	Three-Level Index	Indicator Description
Internal scale management	Employment of labor	Labor input	Labor quantity of new business entities
		educational level of workers	Overall quality education of new business entities
		Labor cost	Average daily wage and employment days of employed workers
	Agricultural mechanization level	Mechanical management level	Proportion of investment in leased equipment such as cultivated land and sowing to all inputs

: Employment of Labor (EOL) refers to the cost paid by hiring labor during production, operation and management activities, that is, the capital invested by hiring labor to participate in the production, operation and management of rice and wheat. Agricultural Mechanization level (AML) refers to the ratio of agricultural rental equipment input to all inputs during production, operation, and management activities, that is, the utilization rate of agricultural machinery. Study the intermediary mechanism of internal scale management under three dimensions, respectively. Finally, we perform weighted averaging on the selected indicators to remove the final values and obtain the variable data required for this study.

3.2.4. External Scale Management Taken as Moderator Variables

The index system of scale management is constructed, this article starts with the basic connotations and characteristics of each indicator, and combines theories such as agricultural economics, geographical environment, and human behavior to comprehensively design an external indicator system for scale management [17,58], to take as the moderator variable, and studies the moderator mechanism for digital carbon emission in

Table 5. Design of index system for external scale management.

Primary Index	Secondary Index	Three-Level Index	Indicator Description
External scale management	Organized management level	Value co-creation	Are you willing to cooperate with other farmers and join cooperatives
		Pooling-of-interest	Agricultural insurance premium income, order contract
		Risk sharing	Whether to obtain a stable sales channel, whether to provide safety monitoring of agricultural products, whether to unify the postpartum quality satisfaction of agricultural materials, and whether to use chemical fertilizers and pesticides in accordance with regulations.
	Socialized service system	Land trusteeship	The actual number of links to obtain land custody services
		Position condition	Kilometers between land and farm
		Commercialized service	Whether technical guidance, field guidance and frequency of technical guidance are provided by cooperatives, and whether centralized training is provided and the frequency of training is provided.

. Organized Management refers to the business risks and benefits of the business entities, that is, the values, risks, and benefits in the process of rice and wheat production. Socialized service refers to the network system formed with various services provided by social and economic organizations to meet the needs of agricultural production, that is, land trust, government support, training services and so on in the process of rice and wheat production, operation, and management. Finally, we perform weighted averaging on the selected indicators to remove the final values and obtain the variable data required for this study.

3.2.5. Environmental and Control Variables

The Scale Management level (SML) is taken as the environmental variable to study the boosting effect of digital agricultural carbon emission reduction under this different levels of management scale influence. This study selects three variables [59,60], namely, the willingness to adopt digital technology, the willingness to expand land scale and the average annual agricultural income, as control variables from the individual characteristics of new business entities that conduct moderate scale managements, which will control and influence whether they adopt agricultural carbon emission efficiency.

3.3. Regional Selection and Data Sources

Jiangsu Province is selected as the research area, as a representative of a strong agricultural province in the Yangtze River Delta and even the whole China. Its grain area has been stable at more than 50 million hm² for 14 consecutive years, and its total output has been stable at more than 0.35 billion kg for 10 consecutive years. In 2024, it reached 0.36 billion kg, completing the task of building 0.08 million hm² of high-standard farmland and significantly improving the level of agricultural modernization. In 2023 year, the work report of Jiangsu government mentioned that the contribution rate of agricultural scientific and technological reached 71.8%, and the mechanization rate reached 85%, of which 95% were rice, wheat and corn, and the land transfer area reached 16.5 million hm², with a transfer rate of 62%. The moderate scale management of agriculture was in good condition. The data of China.com showed that the contribution rate of agricultural science and technology in the Yangtze River Delta has reached 72%, and digital technology covers over 60% of the large-scale business entities. Therefore, Jiangsu, as the research area, has a good representation.

Based on Figure 2, the research data came from the social practice survey conducted by the Digital Rural Research Institute of Nanjing Agricultural University in Jiangsu Province between 2022 and 2024. This study focuses on new types of agricultural operators engaged in large-scale farming, with particular emphasis on the rice and wheat industries. It investigates the current development status and existing challenges related to digitalization, scaling, and carbon emissions in these operations (refer to Appendix A for specific survey questionnaires). Based on the consideration of the differences in Jiangsu’s topography, scale, environment, digitalization and carbon emissions, the research team was divided into three teams in 2023 to conduct a random sampling survey of 13 prefectures and cities in Jiangsu. Nanjing has launched a survey in July 2022, and in 2024. It also launched a supplementary survey in northern Jiangsu. Through the analysis of the survey data, we find that the data samples have little change trend in time, so they can be combined into a whole analysis (refer to Appendix B, Figure A1).

This study conducted random sampling on 13 prefecture level cities in Jiangsu Province, ensuring that 3–5 counties and districts were selected in each prefecture level city. A total of 263 questionnaires on the rice and wheat industry were collected, and the samples were finally screened for validity, resulting in 258 valid questionnaires on the rice and wheat industry. Based on the survey data, an indicator system for each variable was constructed according to the framework outlined in Section 3.2. Specific indicators under each variable were categorized, weighted, averaged, and standardized to derive the variables used in the analysis. This approach enabled an examination of the mechanisms and pathway effects linking digitalization, scale managements, and carbon emissions within the context of agricultural scale management. Descriptive statistics of the specific variables are presented in the

Table 6. Descriptive statistical analysis of main variables.

Variable Type	Variable Names and Symbols	Minimum Value	Maximum Value	Average Value	Standard Deviation	Median
Explained variable	Agricultural carbon emission efficiency (AGPE)	0.004	1.046	0.780	0.309	0.913
Core explanatory variable	Data resources empowerment (DRE)	0.000	0.833	0.324	0.191	0.321
	Digital technology empowerment (DTE)	0.000	0.893	0.385	0.298	0.500
	Network platform empowerment (NPE)	0.000	0.825	0.389	0.160	0.395
Moderator variable	Employment of labor (EOL)	0.037	0.750	0.382	0.104	0.375
	Agricultural mechanization level (AML)	0.000	1.046	0.180	0.221	0.108
Regulatory variable	Organized management level (OML)	0.400	1.000	0.835	0.098	0.841
	Socialized service system (SSS)	0.400	1.000	0.914	0.082	0.935
Environment variable	Scale management level (SML)	0.223	1.000	0.827	0.174	0.861
Control variable	Willingness to adopt digital technology (DTA)	0.200	0.600	0.379	0.126	0.400
	Willingness to expand land scale (LSE)	0.500	1.000	0.572	0.176	0.500
	Annual agricultural income (AAI)	0.000	1.000	0.485	0.235	0.400

Note: In the part, the data of independent variables are collected and normalized by using the 5-level Likelihood Scale.

.

4. Results and Discussion

Using data analysis, this study investigates the relationships among land operation scale, scale management intensity, and agricultural green production efficiency. We employ a framework of scale management to analyze digital emission reduction mechanisms, specifically assessing the influence of management intensity on digital emission reduction, and revealing the mediating and moderating effects of internal and external scale operations. The specific findings are detailed below.

4.1. The Relationship Between Land Management Scale, Scale Management Level, and Agricultural Green Production Efficiency

Figure 3a showed that, in Jiangsu, the land management scale and scale management level (SML) with carbon constraints showed an “inverted S-shaped” trend, that is, with the increase of land scale, SML showed a trend of rising at first, falling and finally rising. Therefore, within the established scale of 66.667 hm², when we divide the land with moderate agricultural scale into 10–26.667 hm², SML is the best and the management efficiency is better. Figure 3b showed that, the relationship between land scale and AGPE is inverted ‘W-shaped’, and SML and AGPE between 20–36.667 hm² are in a relatively stable and effective state. SML and agricultural green production efficiency (AGPE) showed a downward trend as a whole, that is, the higher SML is, the lower its AGPE is, and the smaller its AGPE is. When SML is at the highest level, the AGPE is 0.814. From this, we find that the land scale rice and wheat industry is controlled within the range of 20–36.667 hm², and its economic benefits and green benefits will reach a good state, which is regarded as the moderate scale range in this study. On the basis of existing research on scaling and carbon reduction trends, this study aims to more clearly define the appropriate range of moderate-scale land management in Jiangsu Province, thereby providing a reference for better regulating the scale of agricultural managements in the region. Compared with existing research, this result reveals a complex nonlinear relationship between SML and AGPE, characterized by an “inverted S” and “inverted W” shape, and it provides a more detailed graph, indicating that there may be multiple efficiency equilibrium points within a moderate scale range, which provides more flexible space for policy-making rather than pointing to an absolute optimum. Furthermore, it seeks to delineate the boundaries of moderate-scale land management for the Yangtze River Delta and even the entire country, offering a scientific basis for future digitalization, scaling, and low-carbon development.

4.2. Agricultural Digital Emission Reduction Mechanism in a Scale Environment

4.2.1. Impacts of Digital Empowerment to Agricultural Green Production Efficiency

Based on the research framework in Figure 1, we divide the variable of digital empowerment into data resource empowerment, digital technology empowerment, and network platform empowerment, and explore the impact of digitalization on agricultural green production efficiency using the Tobit regression model with formula (Equations (2) and (3)).

As shown in

Table 7. Tobit regression model analysis of digital empowerment on agricultural green production efficiency.

	Coefficient	Std. Err.	t	[95% Conf. Interval]
_cons	1.136 ***	0.117	9.730	0.906	1.366
DRE	−0.567	0.354	−1.600	−1.265	0.131
DTE	−0.344	0.240	−1.440	−0.816	0.128
NPE	2.035 ***	0.519	3.920	1.012	3.058
DRE2	0.683	0.478	1.430	−0.258	1.623
DTE2	0.281	0.325	0.860	−0.360	0.921
NPE2	−2.870 ***	0.629	−4.560	−4.110	−1.630
DTA	−0.326 **	0.143	−2.280	−0.606	−0.045
LSE	−0.658 ***	0.106	−6.220	−0.866	−0.450
AAI	0.050	0.068	0.740	−0.084	0.185
var(e.c1)	0.063	0.006		0.052	0.075

Note: “*” in the table is significant, in which ** p < 0.05 *** p < 0.001.

, both NPE and NPE² are highly significant at the 1% level, with a coefficient of 2.035 for NPE and −2.870 for NPE². Among them, Network platform empowerment exerts a significantly positive impact on agricultural green production efficiency, indicating that NPE substantially promotes agricultural carbon reduction. Specifically, a 10% increase in NPE leads to a 2.035% improvement in digital emission reduction, thus providing partial support for Hypothesis 1. Mechanistically, NPE facilitates precise monitoring and management of agricultural production through technologies such as the Internet of Things and big data, thereby reducing the excessive input of fertilizers, pesticides, and energy from the source. It also integrates upstream and downstream resources to optimize supply chains and provides data support for government departments to implement targeted ecological compensation and policy assistance. However, the significantly negative coefficient of the squared term of NPE suggests a potential nonlinear relationship. This may be attributed to deviations in the dissemination process of green digital literacy among large-scale producers. Even when acquiring knowledge on green production techniques via platforms, some farmers may struggle to fully absorb and implement them, leading to diminished carbon reduction effects in later stages. Another possible explanation is the underutilization of agricultural machinery and digital equipment purchased through platforms, which could also weaken the overall emission reduction performance. In contrast, data resource empowerment and digital technology empowerment show an inhibitory tendency toward agricultural carbon reduction. Although their effects are not statistically significant, they reflect regional resource heterogeneity within the sample. One plausible reason is the high dependence of large-scale agriculture on energy and machinery inputs. As core enabling tools, data resources and digital technologies may initially intensify resource consumption, resulting in a short-term rebound effect that increases agricultural carbon emissions. Nevertheless, we anticipate that over the long term, their positive impact on agricultural carbon reduction will gradually become evident. In addition, DTA has a significant effect on AGPE at the 5% level, with a coefficient of effect of −0.326, while LSE has a significant control effect on AGPE at the 1% level, with a coefficient of effect of −0.658, while AAI has no significant control effect on AGPE, and by observing the upper and lower bounds of the confidence interval, we can find that the results of this study have a good accuracy. To sum up, digital empowerment will promote agricultural carbon emission reduction, and the promotion of NPE is more significant and effective. Consistent with previous research findings, our study further strengthens existing evidence. In addition, our research clarifies the mechanism by which online platforms promote effective carbon reduction, revealing significant differences in the impact of different dimensions of digital empowerment (NPE, DRE, DTE) on AGPE. Not all forms of digital empowerment can immediately promote emission reduction, and by incorporating land size (LSE) as a control variable into the regression model of digital empowerment and AGPE, a clever connection was established among the three. That is, assume that the H1 part holds.

In order to ensure the robustness of the empirical results, this study uses the methods of reducing sample size, replacing explanatory variables and shortening the research period to conduct Tobit regression analysis again. The results showed that NPE and NPE² were highly significant at the 1% level in Model 1–3, and the degree of influence was not significantly different from the regression coefficients and the results in Model 7; The NPE in Model 4 is not significant, while NPE² is significant at the 5% level with a regression coefficient of −0.823. This may be due to slight differences in the replacement of network platforms, but the overall impact and regression direction are still similar to the original results; The NPE in Model 5 is significant at the 5% level with a coefficient of 1.878, while NPE² is highly significant at the 1% level with a coefficient of −2.766, which is similar to the original results. The significance of DTA and LSE in Model 1–5 is the same as that in Table 7, and the overall coefficient difference is not significant. Therefore, the significance of the three core variables is basically unchanged, and the action direction has not changed, and the basic conclusions are still similar to the previous ones, which shows that the regional selection and time limit selection of the research are reliable, and the above analysis results are highly stable, further enhancing the credibility and persuasiveness of the research results. The regression results are shown in the

Table 8. Robustness test results.

	Model 1	Model 2	Model 3	Model 4	Model 5
_cons	1.129 *** (t = 8.820)	1.128 *** (t = 9.660)	1.135 *** (t = 9.720)	1.418 *** (t = 14.700)	1.126 *** (t = 7.510)
DRE	−0.517 (t = −1.400)	−0.462 (t = −1.390)	−0.579 (t = −1.640)	−0.438 (t = −1.100)	−0.270 (t = −0.720)
DTE	−0.403 (t = −1.600)	−0.318 (t = −1.300)	−0.296 (t = −1.590)	−0.354 (t = −1.430)	−0.242 (t = −0.990)
NPE	2.043 *** (t = 3.660)	1.836 *** (t = 3.580)	2.014 *** (t = 3.890)	0.784 (t = 2.390)	1.878 ** (t = 3.230)
DRE2	0.595 (t = 1.200)	0.302 (t = 0.690)	0.712 (t = 1.510)	0.146 (t = 0.300)	0.109 (t = 0.210)
DTE2	0.390 (t = 1.130)	0.309 (t = 0.940)	0.204 (t = 0.980)	0.187 (t = 0.560)	0.201 (t = 0.620)
NPE2	−2.911 *** (t = −4.310)	−2.531 *** (t = −3.960)	−2.831 *** (t = −4.530)	−0.823 ** (t = −2.470)	−2.766 *** (t = −3.930)
DTA	−0.332 ** (t = −2.180)	−0.321 ** (t = −2.250)	−0.344 ** (t = −2.410)	−0.329 ** (t = −2.360)	−0.305 ** (t = −2.080)
LSE	−0.648 *** (t = −5.890)	−0.630 *** (t = −5.820)	−0.653 *** (t = −6.200)	−0.780 *** (t = −7.500)	−0.602 *** (t = −4.890)
AAI	0.050 (t = 0.680)	0.057 (t = 0.830)	0.054 (t = 0.780)	0.079 (t = 1.120)	−0.012 (t = −0.170)
var(e.c1)	0.070	0.062	0.063	0.066	0.056

Note: “*” in the table is significant, in which ** p < 0.05 *** p < 0.001. Model 1 is the robustness test result of reducing the sample size, and samples are randomly selected within the scale range of (0, 20], (20, 36.667] and (36.667, 66.667] for elimination. Models 2–4 is the robustness test result of replacing the core explanatory variables, and replaces the three core explanatory variables in turn. Model 5 is the result of robustness test to shorten the research cycle, that is, to remove the research data of one year.

.

4.2.2. Impacts of Digital Empowerment to Agricultural Carbon Emission Intensity

To further investigate the relationship between digitalization and carbon reduction, we conducted an additional Tobit regression analysis. This model specified agricultural carbon emission intensity as the dependent variable and digital empowerment as the explanatory variable, aiming to provide deeper insights into the effects of digital emission reduction within the context of scaled agriculture.

From

Table 9. Tobit regression model analysis of digital empowerment on agricultural carbon emission intensity.

	Coefficient	Std. Err.	t	[95% Conf. Interval]
_cons	−0.005	0.040	−0.130	−0.085	0.074
DRE	−0.290 **	0.121	−2.400	−0.528	−0.052
DTE	0.051	0.083	0.610	−0.112	0.214
NPE	−0.345 *	0.179	1.930	−0.007	0.697
DRE2	0.291 *	0.163	1.780	−0.030	0.612
DTE2	−0.059	0.113	−0.530	−0.281	0.163
NPE2	−0.269	0.217	−1.240	−0.697	0.158
DTA	0.071	0.049	1.450	−0.025	0.168
LSE	−0.086 **	0.037	−2.310	−0.159	−0.013
AAI	0.107 ***	0.024	4.570	0.061	0.154
var(e.c1)	0.008	0.001		−0.085	0.074

Note: “*” in the table is significant, in which * p < 0.1 ** p < 0.05 *** p < 0.001.

, it can be observed that DRE and NPE have a significant negative impact on carbon emission intensity, indicating that DRE and NPE significantly reduce agricultural carbon emission intensity. By comparing with Table 7, it is found that the asymmetric impact of data resources on agricultural green production efficiency and carbon emission intensity may be due to the different properties of these two indicators. As carbon emission intensity is a single indicator, it can further demonstrate that data resources directly affect precision agriculture. Reduce fertilizer application directly and quickly while ensuring yield. This further illustrates the rebound effect and time lag effect. Due to the fact that green production efficiency is a comprehensive indicator, the role reflected by data resources will add new costs and management complexity, making it difficult to demonstrate its significance. Moreover, the cost of production materials saved from fertilizers and other sources may also be invested by farmers in other production processes that may generate carbon emissions (such as leasing more agricultural machinery), thereby weakening the net positive impact on the final overall efficiency. The promoting effect of data resources on green production efficiency may take longer to manifest. In addition, the significant positive coefficient of the weighted square term of data resources indicates a significant U-shaped relationship with agricultural carbon emission intensity. The mitigating effect of data resources on carbon emission intensity may weaken and eventually reverse after exceeding a certain threshold. In the initial stage, data resources promote precision agriculture, directly reducing input waste and emissions. However, with the increase in data-driven efficiency leading to the expansion of production scale, the related increase in energy consumption for machinery, irrigation, and transportation may offset the initial carbon reduction. In addition, in highly digitized situations, the energy footprint of maintaining extensive data infrastructure itself may become significant, leading to an upward trend in carbon intensity. Compared with Table 7, the negative non significant coefficient of the NPE squared term here indicates that even if the overall efficiency decreases, the direct emission reduction benefits of the network platform may continue to exist, although its marginal effectiveness weakens. These findings collectively underscore that merely accumulating data resources or expanding platform functions is insufficient. Synergistic policies focusing on enhancing digital literacy, promoting clean energy, and guiding sustainable intensification are crucial to harnessing the long-term decarbonization potential of digital agriculture. Compared with existing research results, the above findings reveal that different dimensions of digital empowerment can trigger varying degrees and forms of rebound effects, and further point out that policy interventions must go beyond “promotion platforms” and focus on how to help farmers bridge the gap from “adoption” to “mastery”. This part of the research results not only did not negate previous findings, but also collectively depicted a more complex, three-dimensional, and realistic picture. It reveals the multiple balances, temporal dynamics, and unexpected consequences in the process of digital technology empowering agricultural green transformation, providing valuable directions and hypotheses for subsequent research.

4.2.3. The Role of Management Level in Boosting Agricultural Digital Emission Reduction

In formula (Equations (2) and (3)), we added moderate scale management (SML) as environment variable to explore how the effect of digital agricultural emission reduction will change under the difference of scale management level (SML) used formula (Equations (2) and (3)).

As shown in

Table 10. Tobit regression analysis on the boosting effect of scale management level.

	Model 1	Model 2	Model 3	Model 4
_cons	1.258 *** (t = 9.930)	1.581 *** (t = 14.510)	1.645 *** (t = 15.830)	1.291 *** (t = 10.160)
SML	−0.263 ** (t = −2.380)	−0.253 ** (t = −2.430)	−0.226 ** (t = −2.170)	−0.270 ** (t = −2.730)
DRE	−0.471 (t = −1.330)	−0.032 (t = 0.120)
DRE2	0.547 (t = 1.150)	−0.598 (t = −1.520)
DTE	−0.281 (t = −1.180)		−0.131 (t = −0.550)
DTE2	0.198 (t = 0.610)		−0.154 (t = −0.470)
NPE	2.004 *** (z = 3.890)			1.311 ** (t = 3.330)
NPE2	−2.841 *** (t = −4.550)			−2.337 *** (t = −4.670)
DTA	−0.276 * (t = −1.930)	−0.341 ** (t = −2.420)	−0.363 ** (t = −2.620)	−0.239 * (t = −1.670)
LSE	−0.620 *** (t = −5.850)	−0.752 *** (t = −7.340)	−0.852 *** (t = −8.620)	−0566 *** (t = −5.420)
AAI	0.098 (t = 1.390)	0.113 (t = 1.520)	0.096 (t = 1.300)	0.088 (t = 1.240)
var(e.c1)	0.061	0.068	0.068	0.063

Note: “*” in the table is significant, in which * p < 0.1 ** p < 0.05 *** p < 0.001. Among them, Model 1 is the overall boosting result, and Models 2–4 is the local boosting result.

, Model 1 represents the overall regression of the three core variables after adding SML, where SML is significant at the 5% level and the regression coefficient is −0.263. NPE and NPE² are still highly significant at the 1% level, but compared to Table 7, the effect of network platforms empowering agricultural carbon reduction has decreased at this time. Model 2–4 shows separate regressions of the three core variables after adding SML. SML in Model 2 is significant at the 5% level with a regression coefficient of −0.253, SML in Model 3 is significant at the 5% level with a regression coefficient of −0.226, and SML in Model 4 is significant at the 5% level with a regression coefficient of −0.270. From this comparison, it can be seen that in local regression, the influence strength of SML on the three core variables is as follows: NPE > DRE > DTE. In depth analysis reveals that the scale management level has a significant negative impact on agricultural digital emission reduction, that is, in the environment of scale management, agricultural carbon emissions may increase. From models 1–4, it can be seen that the scale management level is always related to agricultural green production efficiency, and there is a potential inhibitory effect. This finding is consistent with the framework background of this study, which suggests that large-scale agricultural operations can lead to the extensive use of pesticides and fertilizers, as well as excessive dependence on agricultural machinery. Additionally, large-scale monoculture crop cultivation can also to some extent inhibit agricultural carbon emission reduction. Due to the differentiation of regional resources, Jiangsu will also have differences in energy supply. This means that the current scale management level of Jiangsu in various regions is relatively good, but overall it will have a reverse effect on agricultural carbon reduction. This discovery precisely proves the need to develop moderate scale agricultural managements, control the overall moderate scale scale, and achieve digital emission reduction in agriculture under long-term moderate scale operations. From Model 4, it can be observed that the relationship between NPE and AGPE is nonlinear. The positive sign of NPE and the negative sign of NPE² together represent an inverted U-shaped relationship, which means that when the land scale exceeds a certain critical point, its inherent management complexity, path dependence, and profit maximization logic will gradually weaken the empowering function of the network platform, leading to a decrease in emission reduction benefits. NPE itself has the potential for emission reduction, but scale management level determines whether this potential can be released. Therefore, we should develop moderate scale agricultural operations, so that SML and NPE can work together to promote emission reduction. The willingness to adopt DTA and LSE are important negative control variables. Compared with existing research results, the results of this article indicate that digitization and scaling are not simply a collaborative relationship, but rather have a profound tension. At the same time, it indicates that true “moderation” is the management scale that enables digital empowerment (especially NPE) to fully realize its green benefits. This points out the direction for future agricultural policies and theoretical research: it is necessary to simultaneously promote the moderation of the “physical scale” of land and the green transformation of the “management mode”.

4.3. Intermediary Mechanism of Internal Scale Management in Agricultural Digital Emission Reduction

According to the formula (Equation (5)), we used the Employment of Labor (EOL) and Agricultural Mechanization level (AML) to express internal scale management, and used Sobel-Goodman model to test and explore the intermediary role of internal scale management in agricultural digital emission reduction.

As shown in

Table 11. Analysis of the intermediary results of internal scale management in agricultural digital emission reduction.

	DRE		DTE		NPE
	EOL	AML	EOL	AML	EOL	AML
Sobel	−0.059 ** (z = −2.296)	−0.043 ** (z = −1.685)	−0.038 ** (z = −2.338)	−0.046 ** (z = −2.602)	−0.055 ** (z = −1.804)	−0.072 ** (z = −2.142)
Goodman-1 (Aroian)	−0.059 ** (z = −2.245)	−0.043 ** (z = −1.648)	−0.038 ** (z = −2.288)	−0.046 ** (z = −2.556)	−0.055 ** (z = −1.758)	−0.072 ** (z = −2.097)
Goodman-2	−0.059 ** (z = −2.350)	−0.043 ** (z = −1.726)	−0.038 ** (z = −2.393)	−0.046 ** (z = −2.651)	−0.055 ** (z = −1.855)	−0.072 ** (z = −2.190)
Indirect effect	−0.059 ** (z = −2.296)	−0.043 ** (z = −1.685)	−0.038 ** (z = −2.338)	−0.046 ** (z = −2.602)	−0.055 ** (z = −1.804)	−0.072 ** (z = −2.142)
Direct effect	−0.297 *** (z = −3.342)	−0.313 *** (z = −3.599)	−0.194 *** (z = −3.509)	−0.186 *** (z = −3.352)	−0.402 *** (z = −3.588)	−0.385 *** (z = −3.436)
Total effect	−0.356 *** (z = −3.982)	−0.356 *** (z = −3.982)	−0.232 *** (z = −4.175)	−0.232 *** (z = −4.175)	−0.457 *** (z = −4.003)	−0.457 *** (z = −4.003)
Proportion of total effect that is mediated	0.166	0.120	0.162	0.199	0.119	0.157

Note: “*” in the table is significant, in which ** p < 0.05 *** p < 0.001.

, both EOL and AML exhibit significant negative mediating effects in agricultural digital emission reduction, confirming all hypotheses under Hypothesis 2. However, their mediating roles vary across different pathways of digital decarbonization in agriculture. Specifically, increased investment in agricultural mechanization contributes to higher carbon emissions from data resources, digital technologies, and network platforms. The strength of its mediating effect follows the order: NPE > DTE > DRE. Similarly, increased investment in labor hiring also leads to elevated carbon emissions in these three domains, with a mediating intensity order of DRE > NPE > DTE. These results demonstrate that the carbon emission impacts of agricultural mechanization and labor hiring differ across data resources, digital technologies, and network platforms. These findings provide insights for achieving more effective carbon reduction strategies across various sectors. Compared with existing studies, this research reveals that over-reliance on agricultural mechanization under large-scale farming conditions may exacerbate agricultural carbon emissions. This indicates that the application of digital technologies (DRE, DTE, NPE) will indeed trigger the deepening of agricultural mechanization, which in turn leads to an increase in carbon emissions. In the process of agricultural scaling in Jiangsu, large-scale management of hundreds or even thousands of acres of land requires refined management, inevitably requiring the employment of more agricultural and technical workers. This has led to a short-term increase in carbon emissions, expanding the research perspective on agricultural carbon emissions from traditional “inputs of production materials” (agricultural machinery, fertilizers) to the field of “organizational and labor costs”. Nevertheless, the integration of scale and digital technologies aims ultimately to achieve “increased production, enhanced efficiency, and reduced emissions”.

4.4. Moderator Mechanism of External SCALE Management in Agricultural Digital Emission Reduction

Based on the formula (Equation (6)), we used organized management level (OML) and social service system (SSS) to express external scale management, and explore the regulatory role of external scale management in agricultural digital emission reduction through moderator model.

As shown in

Table 12. Analysis of the regulatory effect of external scale management in data resources enabling agricultural emission reduction efficiency.

	Model 1	Model 2	Model 3	Model 4	Model 5	Model 6
_cons	1.261 *** (t = 7.080)	0.941 *** (t = 4.010)
DRE	−0.392 *** (t = −4.180)	−0.343 *** (t = −3.870)
OML	0.173 (t = 1.020)
SSS		0.482 ** (t = 2.280)
DRE*OML	2.061 ** (t = 2.120)
DRE*SSS		1.069 (t = 1.120)
_cons			1.412 *** (t = 7.360)	1.039 *** (t = 4.380)
DTE			−0.224 *** (t = −3.680)	−0.220 *** (t = −3.980)
OML			0.072 (t = 0.400)
SSS				0.402 * (t = 1.890)
DTE*OML			0.154 (t = 0.270)
DTE*SSS				1.100 * (t = 1.760)
_cons					1.319 *** (t = 7.210)	1.110 *** (t = 4.450)
NPE					−0.446 *** (t = −3.770)	−0.425 *** (t = −3.730)
OML					0.156 (t = 0.900)
SSS						0.330 (t = 1.430)
NPE*OML					1.299 (t = 1.210)
NPE*SSS						1.706 (t = 1.550)

Note: “*” in the table is significant, in which * p < 0.1 ** p < 0.05 *** p < 0.001.

, both OML and SSS exert positive effects on agricultural digital emission reduction, indicating that increased investment in organizational management and socialized services promotes digital decarbonization in agriculture, thus providing partial support for Hypothesis 3. Specifically, OML has a significant positive impact on DRE, suggesting that enhanced organizational management practices can effectively facilitate carbon emission reduction through improved data resource utilization. However, its effects on DTE and NPE, while positive, are not statistically significant. This may reflect the limitations of the current agricultural organizational structure in Jiangsu, where traditional management methods have shown limited effectiveness in promoting digital decarbonization, indicating the need for more refined and targeted policy design. On the other hand, SSS shows a significant positive influence on DTE, indicating that strengthening socialized service systems substantially contributes to emission reduction via digital technology adoption. Nevertheless, its effects on DRE and NPE, though favorable, lack statistical significance. A possible explanation is that the current social service system prioritizes technical efficiency and economies of scale, without fully integrating environmental externalities into service valuations. Thus, while these services address the question of “who cultivates the land,” they have yet to adequately resolve “how to cultivate land in a greener manner.” These findings align with the broader literature confirming the positive roles of organizational management and socialized services in agricultural carbon reduction. We provide a “dual track drive” strategy for both internal and external efforts in agricultural green transformation: internally, we need to help large-scale farms improve their organizational and management capabilities to activate the potential of data resources. Externally, it is necessary to standardize and guide the socialized service system, making it a channel for the dissemination and implementation of digital emission reduction technology. Moreover, this study further clarifies the importance of such investments within the context of large-scale farming.

4.5. Heterogeneity Analysis of Moderate Scale Management Level of Different Land

In order to further and comprehensively explain the empowerment mechanism of digitalization on AGPE, this study conducted heterogeneity analysis on scale differences and regional differences respectively of SML.

The heterogeneity analysis in

Table 13. Analysis of heterogeneity test results of moderate scale management level.

	Model 1	Model 2	Model 3
_cons	1.811 *** (t = 13.860)	0.814 *** (t = 4.780)	0.992 *** (t = 4.980)
DRE	−0.448 * (t = −1.670)	0.016 (t = 0.070)	0.154 (t = 0.500)
DTE	0.017 (t = 0.120)	−0.198 (t = −1.490)	−0.082 (t = −0.520)
NPE	−0.214 (t = −0.670)	0.168 (t = 0.610)	−0.160 (t = −0.420)
DTA	−0.449 * (t = −1.860)	−0.117 (t = −0.550)	0.057 (t = 0.190)
LSE	−1.106 *** (t = −7.540)	0.047 (t = 0.220)	−0.264 (t = −1.060)
AAI	0.078 (t = 0.770)	0.150 (t = 1.290)	−0.006 (t = −0.040)
var(e.c1)	0.055	0.050	0.069

Note: “*” in the table is significant, in which * p < 0.1 *** p < 0.001. Among them, Model 1–3 is the result of heterogeneity analysis of regional differences, that is, comparative analysis of the differences in southern Jiangsu, central Jiangsu, and northern Jiangsu.

reveals significant differences in the factors affecting agricultural green production efficiency among the central, southern, and northern regions of Jiangsu. In Model 1 (Sunan), the willingness to expand land size (LSE) showed a significant negative impact on the outcome variables (β = −1.106, p < 0.001). Similarly, Data Resource Authorization (DRE) and Digital Technology Adoption Intention (DTA) also showed negative coefficients, significant at the 10% level. This indicates that in the economically developed and highly intensive agricultural system of southern Jiangsu, further scale expansion may backfire, leading to low efficiency or rebound effects, and hindering carbon emission reduction. In contrast, the results of Model 2 (in central Jiangsu) and Model 3 (in northern Jiangsu) were significantly insignificant. This indicates that the driving factors for agricultural green production efficiency in these regions are less apparent or may be different compared to the south. The lack of importance may be due to the low level of digitalization in agriculture, different agricultural structures, or different policy environments in central and northern Jiangsu. This obvious regional heterogeneity highlights the inconsistent impact of scale management and digital empowerment. The research findings emphasize the urgent need for policies tailored to specific regions, rather than a one size fits all approach. Strategies should be tailored according to the specific development stage, resource endowment, and operational background of each region to effectively promote agricultural carbon reduction.

5. Conclusions and Suggestions

5.1. Conclusions

This study developed a new method for evaluating scale management level from the perspectives of scale effect and technological innovation in the context of scale management, while considering the “carbon” factor, based on existing measurement models. Using Tobit, mediation, and moderation models to explore the effects of digital emission reduction mechanisms and scale management level in a scaled business environment, and further investigate the mediating mechanisms of internal scale management and the regulatory effects of external scale management. The specific conclusions are as follows: (1) In Jiangsu, the land scale and the scale management level showed an “inverted S-shaped” and an “inverted W-shaped” relationship with agricultural green production efficiency, respectively. Overall, both the appropriately scaled management level and agricultural green production efficiency demonstrate a declining trend. Within the range of 20–36.667 hm², economic and ecological benefits remain relatively stable. (2) The Tobit model results indicate that online platforms significantly promote carbon reduction, but their squared terms exhibit an inverted U-shaped relationship with agricultural green production efficiency, and SML is significant at the 5% level. From a local regression perspective, the strength of SML’s impact on the three core variables is: NPE > DRE > DTE. (3) Both labor employment and agricultural mechanization in internal scale management negatively significantly affect agricultural carbon emission reduction, in which the negative intermediary role played by labor employment is DRE > NPE > DTE, and the negative intermediary role played by agricultural mechanization is NPE > DTE > DRE. (4) Organized management and socialized service in external scale management have a significant positive impact on agricultural carbon emission reduction, in which organized management significantly positively regulates the effective carbon emission reduction of agricultural data resources and socialized service significantly positively regulates the effective carbon emission reduction of agricultural digital technology. In addition, the results are still valid under the robustness test and heterogeneity test, and have good reliability and persuasiveness.

5.2. Suggestion

According to the above conclusions, this paper puts forward the following suggestions for the future agricultural development in Jiangsu: (1) In view of exploring the scope of Jiangsu’s land scale, we should flexibly control the best scale of Jiangsu’s agricultural moderate scale management, refer to the standard of 20–36.667 hm² of land moderate scale management, guide the grain business entities to develop moderate scale management, and gradually improve the development level of standardization, digitalization, ecology and organization in southern Jiangsu, central Jiangsu and northern Jiangsu in a moderate range based on regional differences. (2) To achieve effective digital emission reduction in agriculture under the background of large-scale operation, investment in network platforms can be considered, and moderate scale should be closely combined with digital emission reduction to achieve effective emission reduction of data resources and digital technology after long-term operation. (3) In view of the intermediary role of internal scale management, we advocate combining the regional differentiation of Jiangsu to reasonably reduce the input of labor and mechanization, steadily improve the level of cultivated land fertility, enhancing the digital literacy and agricultural machinery training of scale entities, steadily improving the fertility level of farmland, and promoting the green transformation of agricultural production processes. (4) Based on the adjustment of external scale, we encourage efforts to improve the degree of organization of agricultural production and management, innovate social services, and provide production services such as labor introduction, land transfer and production custody for business entities. These suggestions will provide a theoretical basis and practical insights for promoting digital emission reduction in the process of agricultural scaling, and provide scientific support for achieving agricultural carbon emission reduction while balancing economic benefits and green development.

5.3. Limitations and Prospects

Although this study evaluates the agricultural digital emission reduction mechanism under the moderate scale management level, it still has certain limitations and deserves further study. First, a total of 258 valid samples were obtained from the survey data, and the number of samples was limited, among which the number of samples in some scale intervals was small, so the representativeness of simultaneously evaluating the three indicators of digital empowerment level, large-scale management and low-carbon agriculture was limited. Secondly, this paper takes Jiangsu Province as the key research area, which can well show the mechanism analysis and path effect of agricultural carbon emissions in the Yangtze River Delta region, but the research on agricultural carbon emissions covering a wider area of China needs to be deepened. Although this problem has not been solved in our current research, it is a direction of future research work.