Rethinking the Evaluation of Agricultural Eco-Efficiency in the North China Plain, Incorporating Multiple Greenhouse Gases

Abstract

The reduction of substantial agricultural greenhouse gases (GHGs) emissions can make a significant contribution to climate change mitigation and regional sustainable development. Given that most of the current studies about eco-efficiency only considers CO₂, while ignoring other GHGs, such as CH₄ and N₂O, this study analyzes the spatiotemporal characteristics of CO₂, CH₄, and N₂O, and considers them as undesirable outputs to assess the agricultural eco-efficiency (AEE) in the North China Plain from 2004 to 2022, respectively, including AEECO₂, AEECH₄, AEEN₂O, and AEEGHG. The results show that (1) Agricultural GHGs emissions increased significantly before 2018 and slightly decreased after 2018, due to the enforcement of energy-saving and emission-reducing policies. Spatially, GHG emissions are higher in the north but lower in the south. (2) The study demonstrated that incorporating CH₄ and N₂O significantly affects efficiency (p < 0.01). AEECH₄ and AEEN₂O are higher than AEEGHG, while AEECO₂ is lower than AEEGHG, indicating that only considering a single emission will result in an inefficient outcome. (3) With significant regional heterogeneity, AEEGHG is higher in Henan, Beijing, and Tianjin, while it is the lowest in Hebei. Specific suggestions are proposed to promote sustainable agricultural development. This study presents a novel perspective for comprehensively assessing AEE and offers scientific evidences for agricultural policy formulation to promote climate mitigation.

1. Introduction

In recent years, the issue of global climate change has attracted widespread attention worldwide and is one of the major threats that humanity must confront. From 2021 to 2022, global GHGs emissions increased by 1.2%, reaching a new high of 574 billion tons of CO₂ equivalent [1]. China, the world’s largest emitter of CO₂, is undergoing a rapid process of economic development and faces huge challenges in reducing GHG emissions [2]. In 2020, China announced that it would reached the carbon emissions peak by 2030 and would achieve carbon neutrality by 2060 [3]. Since China is a large agricultural country, the agriculture sector accounts for 17% of China’s total GHG emissions [4]. The continuous increase in global GHG concentrations has led to global warming and a decline in agricultural sustainability [5]. In addition, in the total global emissions of methane (CH₄) and nitrous oxide (N₂O), agriculture contributes approximately 60% [6], which is very likely to lead to global warming and intensify the greenhouse effect [7]. Consequently, agriculture must urgently address non-CO₂ GHG emissions while ensuring synchronized mitigation of CH₄ and N₂O.

Agricultural eco-efficiency (AEE, also known as agricultural ecological efficiency) refers to the ability of an agricultural production system to achieve the highest output with the least resource input and environmental cost [8]. It highlights the coordinated optimization of economic benefits and environmental protection. Improving AEE is of great significance in achieving sustainable agricultural development and has become one of the important strategies all over the world [8]. Accurate assessment of AEE can help to diagnose the current status of agricultural sustainable development and predict the potential limiting factors that may arise in the future [9]. Nowadays, scholars have increasingly focused on assessing AEE using various methods. Data envelopment analysis (DEA) has been widely applied to assess eco-efficiency. Some scholars have used the DEA model to study the agricultural efficiency of farmland [10]. Meanwhile, slacks-based measure data envelopment analysis (SBM-DEA) is used to assess the green AEE [11]. In contrast to DEA, stochastic frontier analysis (SFA), as a parametric approach, can also estimate the relationship between agricultural input and output variables [12,13]. Both of the parametric and non- parametric methods made significant contributions to the assessment of eco-efficiency. The SBM-DEA model advances conventional DEA methodology through non-radial and non-angular integration, effectively resolving inherent input redundancy and output deficiency issues [9]. Specifically, SBM-DEA possesses a distinct advantage in handling multiple inputs and outputs, enabling a robust quantification of AEE [2].

Although there has been some progress in the research on AEE, there are still several problems to be examined. Firstly, traditional DEA model only focuses on the correspondent relationship between input and output variables, ignoring both radial and angular dimensions, which ultimately leads to slack and inaccuracy in efficiency measurement [14]. Moreover, most studies ignore undesirable outputs in their index systems for evaluating AEE [15,16]. The measurement of undesirable outputs mainly focuses on CO₂ emissions, but ignores non-CO₂ GHGs [17,18]. As for the agricultural GHG emissions, CO₂ plays an essential ecological role in photosynthesis and the carbon cycle. CH₄ and N₂O are partially process-based and biologically unavoidable, as they result from fundamental processes like soil nitrogen turnover. However, these GHG emissions would be considered undesirable outputs when human activities lead to excessive emissions beyond the ecosystem’s natural balance. Therefore, our study integrates CO_2, CH₄ and N₂O as undesirable outputs within the SBM-DEA at a finer city level to capture more detailed spatial heterogeneity and support targeted policy-making.

The North China Plain (NCP), with about 18 million hectares of cultivated land, represents roughly 19% of China’s total agricultural land. It is the second largest plain in China and serves as a primary grain production base, which contributes to over 75% of the national wheat output and 32% of maize yield [19]. Since the 1970s, the multiple cropping index in NCP has generally increased. As a result, the double cropping system of winter wheat and summer maize becomes the optimal planting pattern [20,21]. Intensive farming practices have led to the intensification of environmental pollution and degradation, posing huge challenges to agriculture in the NCP. The overapplication of agricultural inputs such as chemical fertilizers and pesticides poses significant threats to the environment [22]. Therefore, this study focuses on AEE evaluation in the NCP. To better serve agricultural management, we targeted prefecture-level cities as the minimum research unit, since this can avoid the over-generalization of provincial-scale data and mitigate data fragmentation at the county level. Additionally, data at the level of prefecture-level city offer strong accessibility and continuity, supported by comprehensive official statistical systems, such as statistical yearbooks and economic bulletins, which can facilitate data collection and utilization.

Unlike previous studies that mainly focused on CO₂ or provincial/national levels, this study integrates CO₂, CH₄ and N₂O as undesirable outputs within the SBM-DEA, and evaluates AEE in the NCP by adopting the panel data of 75 cities from 2004 to 2022. Firstly, we considered CO₂, CH₄, N₂O, and all three emissions as the undesirable output, produced AEECO₂, AEECH₄, AEEN₂O, and AEEGHG, respectively. Then, we clarified the temporal and spatial variations in AEE in the NCP, and analyzed the underlying causes of the variations. Finally, corresponding policy recommendations were provided for regional agricultural sustainable management. The findings reveal dynamic variations of AEE and provide practical supports for devising sustainable agricultural policies.

The paper is organized as follows: Section 1 outlines the study’s background; Section 2 specifies the study area and data sources; Section 3 details the methodological framework; Section 4 reports the empirical results; Section 5 interprets the findings within a broader context; and Section 6 summarizes the main conclusions.

2. Study Area and Data Sources

The NCP is one of the three major Chinese plains, located at 32°~40° N and 114°~121° E, spanning seven provinces and municipalities directly under the central government: Beijing, Tianjin, Hebei, Shandong, Henan, Jiangsu, and Anhui (Figure 1). NCP has flat terrain, numerous rivers, and lakes. The NCP is suitable for planting wheat, corn, and other crops mainly based on dryland farming, and serves as an important production area of grain crops and cash crops. The wheat output accounts for about half of the country. Its fertile alluvial soils and extensive irrigation systems support intensive agricultural production. However, this also leads to high resource inputs and considerable GHG emissions, mainly from fertilizer application and rice cultivation. Due to rapid economic development, population density and intensive farming, the NCP faces increasing pressure to balance agricultural productivity and environmental sustainability. In this study, the NCP was chosen as the study area not only for its agricultural importance but also because it exemplifies the challenges of improving eco-efficiency under resource and environmental constraints. The analysis focuses on prefecture-level cities within the NCP, enabling a more detailed spatial examination of agricultural GHG emissions and eco-efficiency variations.

The primary data are from the China Statistical Yearbook (2005–2023), the China Price Yearbook (2005–2023), the National Compilation of Agricultural Product Cost and Income Data (2005–2023), and the related regional statistical yearbooks in the North China Plain (2005–2023). Jiyuan was established after 2004 and is excluded with much missing data. Therefore, a total of 75 cities (73 prefecture-level cities and two municipalities) in NCP from 2004 to 2022 were selected to construct the panel dataset.

3. Methods

3.1. Construction of Indicator System

In the process of agricultural production, there are various inputs, including crop planting area, pesticides, fertilizers, agricultural machinery use, etc., and inevitable outputs, including, CO₂, CH₄, and N₂O and other emissions. While GHG is a vital part of the natural carbon cycle and photosynthesis, excessive anthropogenic GHG emissions contribute to climate change, which are detrimental to the agricultural sustainability. Therefore, GHG is treated as an undesirable output in terms of anthropogenic emissions generated by agricultural inputs and practices in this study (e.g., fertilizer application, pesticide use, fuel combustion). We selected grain production as the desirable output, and selected CO₂, CH₄, and N₂O emissions as the undesired outputs to assess AEE (

Table 1. Input and output indicators to assess AEE.

Indicator Type	Indicator Name	Explanation of Indicators	Unit
Input indicators	Crop planting area (CPA)	The area of wheat, maize, and paddy	105 hectares
	Effectively irrigated area (EIA)	The area of farmland that can be irrigated normally	103 hectares
	Total power of agricultural machinery (TPM)	The sum of the power of various power machines	105 kWh
	Pesticide (PEI)	The sum of the converted amounts of various pesticides	ton
	Chemical fertilizer (CFI)	The sum of the purified amounts of various fertilizers	ton
Desirable output indicators	Crop production	The total yield of wheat, maize, and paddy	105 tons
Undesirable output indicators	GHG emissions	CO2 emissions	105 tons
		CH4 emissions	105 tons
		N2O emissions	105 tons

).

Specifically, we selected five input indicators: crop planting area (CPA), effectively irrigated area (EIA), total power of agricultural machinery (TPM), pesticide (PEI), and chemical fertilizer (CFI). CPA can reflect the planting scale of the place. EIA can reflect the local irrigation degree. TPM is also crucial to food production, which can reflect the degree of local mechanization. PEI and CFI are necessary inputs for modern agricultural food production. Among the undesired outputs, we divided CO₂, CH₄, and N₂O as the outputs. We assessed AEE using CO₂, CH₄, and N₂O as a single undesired output, named AEECO₂, AEECH₄, and AEEN₂O, respectively. We assessed the comprehensive AEE using all three GHGs, named AEEGHG.

3.2. Measurement of Agricultural Greenhouse Gas Emissions

3.2.1. Calculation of CO₂ Emissions

In the process of crop growth and planting, framers usually apply chemical fertilizers, pesticides, and other inputs to ensure the normal growth of crops or increase their yield. We employed the carbon emission coefficient method to calculate the CO₂ emissions generated by agricultural inputs as follows [23].

$$C_{ait}=\sum T_{kit}\ast a_k\ast (44/12)$$

(1)

In Equation (1), $C_{ait}$ represents all the carbon emissions from agricultural inputs of city i in year t (2004–2022), tons. $T_{kit}$ represents the actual input amount of carbon sources k of city i in year t, 10⁴ tons. As shown in

Table 2. Carbon emissions factor of agricultural activities.

Carbon Source	Carbon Emission Factor	Source
Nitrogenous fertilizer	2.116 kg/kg	[24]
Phosphate fertilizer	0.636 kg/kg	[24]
Potassium fertilizer	0.18 kg/kg	[24]
Compound fertilizer	0.381 kg/kg	[23]
Pesticide	4.93 kg/kg	[25]
Diesel fuel	0.59 kg/kg	[26]
Irrigate	20.476 kg/hm2	[27]

, $a_k$ represents the emission coefficients for the k carbon sources [23,24,25,26,27]. $44/12$ denotes the conversion factor to translate C into CO₂. Chemical fertilizer and pesticide are calculated as the net usage quantity of agricultural activities. EIA is considered in the calculation.

3.2.2. Calculation of CH₄ Emissions

This study focused on CH₄ emissions from rice cultivation because rice paddies are the dominant anthropogenic source of CH₄ in the NCP. We referred to the Guidelines for Provincial Greenhouse Gas Inventories from National Development and Reform Commission of the People’s Republic of China (NDRC) to perform the calculation of CH₄ emission [23].

$$C_{rit}=S_{it}\ast I\ast 25$$

(2)

where, $C_{rit}$ represents the total amount of rice carbon emissions of city i in year t, tons. $S_{it}$ represents the rice planting area of city i in year t, hectares. I represents the CH₄ emission factor of rice, which is 153.3 kg∙hm⁻² [28]. 25 is the conversion factor, i.e., CH₄ equivalent to CO₂ [29].

3.2.3. Calculation of N₂O Emissions

For N₂O emissions, we considered wheat, maize, and rice, which are the major crops receiving nitrogen fertilizers. Other potential sources, such as manure management and crop residues, were excluded due to data limitations and consistency with previous studies. Agricultural N₂O emissions consist of direct emissions resulting from nitrogen inputs, and indirect emissions arising from the volatilization of nitrogen compounds, which subsequently contribute to atmospheric deposition and nitrogen losses via leaching or surface runoff [30].

$$C_{Nit}={ (N}_{dit}+N_{init}) \ast 273$$

(3)

where, $C_{Nit}$ represents the carbon emission volume of the farmland of city i in year t. $N_{dit}$ is the direct N₂O emissions of city i in year t. $N_{init}$ represents the indirect N₂O emissions of City i in year t. 273 is the conversion factor, which is the conversion of N₂O equivalents to CO₂ [31].

The N in the direct N₂O emissions mainly comes from the nitrogen in nitrogen fertilizers and compound fertilizers ($N_{cf1it}$) as well as the nitrogen from straw application ($N_{s1it}$), and direct N₂O emissions from agricultural land. $N_{dit}$ is calculated according to Equations (4)–(9) [32,33,34].

$$N_{dit}={\mathrm{N}}_{\mathrm{c}\mathrm{f}1\mathrm{i}\mathrm{t}}+{\mathrm{N}}_{\mathrm{s}1\mathrm{i}\mathrm{t}}$$

(4)

$$N_{cf1it}={(N}_{Nit}+N_{fit}/3)\ast EF$$

(5)

$$N_{s1ij}=RCro{p}_{it}\ast EF$$

(6)

$$RCro{p}_{it}=\sum _{s=1}^3\left(NRoo{t}_{its}+NStr{a}_{its}\ast f^{rute}\right)$$

(7)

$$NRoo{t}_{its}=Pro{d}_{its}\ast f_s^{dry}\ast {\displaystyle \frac{f_s^{rb}}{f_s^{hary}}}\ast f_s^{rbrate}$$

(8)

$$NStr{a}_{its}=Pro{d}_{its}\ast f_s^{dry}\ast \left({\displaystyle \frac{1}{f_s^{harv}}}-1\right)\ast f_s^{rbrate}$$

(9)

where, $N_{nit}$ is the net amount of nitrogen fertilizer converted to nitrogen of city i in year t. $N_{fit}$ represents the net fertilizer usage (after conversion) of city i in year t [33]. $EF$ is a direct emission factor for N₂O. $RCro{p}_{it}$ is the nitrogen content of straw returning to the field of city i in year t. $NRoo{t}_{its}$ is the nitrogen content in the roots of crops (wheat, corn, and rice, respectively) in year t. $NStr{a}_{its}$ represents the nitrogen concentration in crop-specific straw and stubble residues of city i and year t. $f^{retu}$ is the proportion of the corresponding type of crop straw returned to the field. $f_s^{dry}$ is the dry matter fraction of the corresponding crop type. $f_s^{rb}$ represents the shoot-to-root ratio for the specific crop variety. $f_s^{harv}$ is the crop harvest index for the corresponding crop type. $f_s^{rbrate}$ is the nitrogen content of the root biomass of the corresponding crop type. The emissions coefficients of major crops are shown in

Table 3. Emissions coefficients of major crops.

Crop	$\mathit{E}\mathit{F}$ (kg/kg) [32]	${\mathit{f}}^{\mathit{r}\mathit{e}\mathit{t}\mathit{u}}$ [34]	${\mathit{f}}_{\mathit{s}}^{\mathit{d}\mathit{r}\mathit{y}}$ [34]	${\mathit{f}}_{\mathit{s}}^{\mathit{r}\mathit{b}}$ [34]	${\mathit{f}}_{\mathit{s}}^{\mathit{h}\mathit{a}\mathit{r}\mathit{v}}$ [34]	${\mathit{f}}_{\mathit{s}}^{\mathit{r}\mathit{b}\mathit{r}\mathit{a}\mathit{t}\mathit{e}}$ [34]
Wheat	0.0057/0.0109	0.14	0.83	0.20	0.37	0.10
Corn	0.0057/0.0109	0.14	0.40	0.17	0.44	0.10
Rice	0.0057/0.0109	0.14	0.83	0.13	0.43	0.10

Note: $EF$ (the direct emission factor) of N₂O is 0.0057 kg/kg for cities in Beijing, Tianjin, Hebei, Shandong, and Henan, while 0.0109 kg/kg for cities in Jiangsu and Anhui.

.

Agricultural N₂O indirect emissions are calculated as follows [34].

$$N_{init}=N_{lit}+N_{Sit}$$

(10)

$$N_{lit}=\left(N_{nit}+{\displaystyle \frac{N_{fit}}{3}}+RCro{p}_{it}\right)\ast 10\mathrm{\%}\ast 0.01$$

(11)

$$N_{sit}=\left(N_{nit}+{\displaystyle \frac{N_{fit}}{3}}+RCro{p}_{it}\right)\ast 20\mathrm{\%}\ast 0.0075$$

(12)

where, $N_{lit}$ represents the N₂O emissions in city i due to atmospheric deposition over the period from 2004 to 2022. The volatility rate is set at the recommended value of 10%. The emission factor is taken as the IPCC emission factor of 0.01. $N_{sit}$ represents the nitrogen loss in city i due to leaching and runoff from farmland in year t. This loss is calculated at 20% [32].

3.3. Slack-Based Measure Data Envelopment Analysis

DEA epitomizes a non-parametric methodology germane to the evaluation of efficiency that integrates multiple inputs and outputs without the need to pre-assign or define weights. By assigning relative proportions of different inputs, it identifies the optimal input combination to evaluate the efficiency of each decision-making unit (DMU) [9]. This study used SBM-DEA [35,36], which is built upon the traditional DEA framework, and introduces significant improvements by incorporating non-radial and non-angular features. These enhancements address the common limitations of traditional DEA models, such as excessive inputs or insufficient output, helping to avoid inefficiency caused by input redundancy or output scarcity [9]. The SBM-DEA model provides a powerful framework for quantifying AEE by solving these problems, while maintaining the accuracy of the assessment, even when dealing with undesirable outputs.

To calculate AEE, we will characterize each DMU by three vector components: the input vector $X\in R^m$, the desired output vector $y^e\in R^a$, and the undesired output vector $y^n\in R^b$ (m, a and b represent the types of input, desired output and undesired output elements, respectively). According to the model under the constant returns to scale, the production set is defined as $\mathrm{P}\left(\mathrm{x},{\mathrm{y}}^{\mathrm{e}},{\mathrm{y}}^{\mathrm{n}}\right)=\left(\mathrm{x},{\mathrm{y}}^{\mathrm{e}},{\mathrm{y}}^{\mathrm{n}}\right)\mathrm{x}⩾\mathrm{X}\mathsf{\lambda },{\mathrm{y}}^{\mathrm{e}}⩾{\mathrm{Y}}^{\mathrm{e}\mathsf{\lambda }},{\mathrm{y}}^{\mathrm{n}}⩾{\mathrm{Y}}^{\mathrm{n}\mathsf{\lambda }}$. The SBM-DEA model is expressed as [9]:

$${\gamma }^{\ast }=\mathrm{m}in {\displaystyle \frac{1-\frac{1}{m}{\sum }_{i=1}^m\frac{s_i^{-}}{x_{i_0}}}{1+\frac{1}{s_1+s_2}\left({\sum }_{h=1}^{s_1}\frac{s_h^a}{y_{h_0}^a}+{\sum }_{r=1}^{s_2}\frac{s_r^b}{y_{r_0^b}}\right)}}$$

(13)

$$s.t\left\{\begin{array}{c}{x}_0=X\lambda +s^{-}\\ y_0^a=Y\lambda -s^a\\ y_0^b=Y^b\lambda +s^b\\ s^{-}\ge 0, s^a\ge 0, s^b\ge 0, \lambda \ge 0\end{array}\right.$$

(14)

In Equation (13), $m$ is input, $s_1$ is the desired output, $s_2$ is the undesired output. $s_i^{-},s_h^a,s_r^b$, represent the input, desired output, and undesired output, respectively. $r$ is the weight vector. ${\gamma }^{\ast }$ is the efficiency value, ranging from $0\le {\gamma }^{\ast }\le 1$. When ${\gamma }^{\ast }=1$ or $s_i^{-}=s_h^a=s_r^b$, it indicates that the DMU is operating efficiently. When ${\gamma }^{\ast }<1$, it means that this DMU has not reached the optimal efficiency and still needs improvement.

In Equation (14), the conditions $X_0,Y_0^g,y_0^b,s^{-},s^a, s^b$ that need to be satisfied are listed. $X_0$ can be calculated using $X$, $\lambda$, $s^{-}.$ $s^{-}$ representing the possible reduction in input that the DMU can achieve without reducing its output. $Y_0^g$ can be calculated using $Y$ and $\lambda$, $s^a$. $s^a$ represents the slack variable for the desired output. $s^b$ represents the slack variable for the undesired output. $s^{-}, s^a$ and $s^b$ are non-negative numbers.

To assess the robustness of our SBM-DEA results, we performed two additional analyses. A sensitivity analysis was conducted by excluding TPM from the input set and recalculated the efficiency scores [37]. Moreover, we applied SFA using a Cobb-Douglas production function to compare AEE rankings [38]_. Since the AEE from DEA are bounded and not normally distributed, we applied the Wilcoxon signed-rank test, a non-parametric test for paired samples, to compare the different AEEs.

4. Results

4.1. Temporal Variations of Agricultural Greenhouse Gas Emissions

Figure 2 shows the trend and structure of GHG emissions in the NCP from 2004 to 2022. The total agricultural GHG emissions (total CO₂) in the North China Plain showed a fluctuating upward trend followed by a slight decrease from 2004 to 2022. It reached the peak of 8042.15 million tons in 2018. From 2004 to 2018, GHG emissions increased by 2347.27 million tons, approximately 87.1%. Specifically, the emission volumes of CO₂, CH₄, and N₂O also had their own variation characteristics. The CO₂ emission volume was generally at a high level with some fluctuations, while the CH₄ and N₂O emission volumes were slightly lower than CO₂ but also had their own variation ranges. The greatest contribution to the increase in total GHG emissions was the increase in CO₂ emissions, which was mainly due to the increase in fertilizer and pesticide inputs. This increase in fertilizer and pesticide usage led to the increase in CO₂ emissions [39,40]. From 2018 to 2022, the overall GHG emissions showed a decreasing trend, reducing by 599.42 million tons. The emission reduction effect is primarily attributable to nationwide energy−saving policies, the promotion of green development concepts, and the adoption of technologies like precision fertilization and green pest control in recent years.

4.2. Spatial Heterogeneity of Agricultural Greenhouse Gas Emissions

The total amount of GHG emissions in each city of the North China Plain from 2004 to 2022 are shown in Figure 3. Overall, GHG emissions show a pattern of lower in the north and higher in the south. This is mainly due to the differences in cropping pattern, and the crop types. Some crops are grown once a year in the north, while they are grown twice in the south, resulting in variations in GHG emissions. Overall, the GHG emissions in Yancheng City are the highest, with the maximum emission volume reaching 294.80 million tons. In 2022, the city with the smallest GHG emissions was Beijing, with an emission volume of 14.03 million tons. The city with the largest change in GHG emissions from 2004 to 2022 was Huainan City, with an increase of approximately 100 million tons.

In Figure 4, we list the nine prefecture-level cities with the highest GHG emissions. The GHG emissions of most cities have fluctuated, but overall, they show an increasing trend. Anqing and Liuan were different from other cities. The GHG emissions of Anqing and Liuan have been decreasing continuously since 2014, but increasing in 2017. The reason for the decrease in GHG emissions in 2014 was that Anqing strengthened the governance of agricultural pollution and promoted biogas projects after 2014. At the same time, Liuan and Anqing promoted sustainable farming restructuring, such as promoting integrated rice–fish farming, reducing the use of fertilizers and pesticides, and increasing the comprehensive utilization rate of straw. In 2017, the area of cultivation was expanded after several years of soil fertilization and increased crop yields in Anqing and Liuan. It led to an increase in GHG emissions. Moreover, Liuan suffered from extreme weather conditions such as frequent floods and droughts in 2016, which led to adjustments in agricultural production patterns in 2017 and an increase in GHG emissions.

4.3. Differences Among the Agricultural Eco-Efficiency

Through the Wilcoxon signed-rank comparison, it was found that there were significant differences among AEECO₂, AEECH₄, and AEEN₂O, indicating that incorporating CH₄ and N₂O into the efficiency assessment system would have a significant impact on the eco–efficiency results (

Table 4. Wilcoxon signed-rank results for the AEEGHG, AEECO₂ and AEECH₄ in the North China Plain.

Comparison	Test	p-Value	Significant
AEECO2 vs. AEECH4	Wilcoxon signed-rank	<0.05	Yes
AEECO2 vs. AEEN2O	Wilcoxon signed-rank	<0.05	Yes

). This is because several GHGs have similar sources and generation processes, but their specific formation processes and temporal and spatial distribution patterns are different. This divergence can be largely attributed to differences in crop types and planting methods, which lead to changes in the input–output relationship in the agricultural production process, resulting in different efficiency values.

As shown in Figure 5, we can observe that AEECO₂ was overestimated, while AEECH₄ and AEEN₂O were underestimated compared to AEEGHG. This is consistent with some research results [9]. In 2004, CO₂ was overestimated the most, by approximately 11.53%, while the overestimation of CH₄ occurred most frequently in 2008, by approximately 0.57%. In 2022, N₂O was underestimated by approximately 0.59%, making it the year with the highest underestimation compared to overestimation over the past five years. In Figure 5, we can observe that the overall AEEGHG has fluctuated and increased. There was a slight decline in 2007 and 2010 due to sudden natural conditions. In 2007, a flood occurred in the Huai River Basin [41], and in early 2010, a freeze disaster occurred, both of which led to a decrease in grain production. After 2013, the country issued an important document, aiming to expedite modern agriculture development and reinforce the driving forces of rural growth. The relevant policies promoted the coordinated development of agriculture and environment, laying the foundation for sustainable agriculture and improving AEE.

4.4. Robustness Checks

As shown in

Table 5. Robustness check results.

Robustness Checks	Method	Spearman’s Rho	p-Value
Sensitivity analysis	SBM-DEA vs. modified SBM-DEA	0.98	<0.01
Methodological comparison	SBM-DEA vs. SFA	0.63	<0.01

, the Spearman rank correlation between the original DEA scores and the adjusted scores (excluding TPM) was very high (rho = 0.98, p < 0.01), indicating strong robustness to input choice. The comparison with SFA results yielded a Spearman rank correlation of 0.63 (p < 0.01), suggesting moderate to strong consistency in efficiency rankings between DEA and SFA models. These can prove that SBM-DEA is a reasonable method and the obtained results are reliable.

4.5. Spatial Heterogeneity of Agricultural Eco-Efficiency

The AEEGHG shows significant spatial heterogeneity, as shown in Figure 6. Based on the equal interval definition, five regions were delineated, namely the Lag area (0.00–0.20), Starting area (0.20–0.40), Progress area (0.40–0.60), Leading area (0.60–0.80), and Coordination area (0.80–1.00). In Figure 6, it can be observed that Hebei is mostly in the Lag area, while Henan and Shandong, as major agricultural provinces, are mostly in the Leading area and the Coordination area. Due to their diverse natural conditions and abundant agricultural resources, they rank among the top of AEEGHG across the NCP. Other regions with higher AEEGHG are mostly economically developed areas, such as Beijing and Tianjin. Anhui has abundant agricultural resources, but its AEEGHG is relatively low, and most of it is in the Lag area. There is a large potential for improvement in AEE, which will help enhance agricultural sustainable development and promote the rational allocation of agricultural resources.

In the NCP, the AEEGHG of Henan is the highest, while that of Hebei is the lowest. Among them, the overall AEEGHG of Jiangsu is relatively balanced. In 2022, 46.2% of the cities in Jiangsu were in the Leading area, and 53.8% were in the Coordination area. In Anhui, the AEEGHG varies greatly. Only 25% of the cities are in the Coordination area, 6.25% are in the Starting area, and the proportion of the Lag area is as high as 68.73%. Moreover, the proportions of the Lag area and the Coordination area are relatively the largest, indicating that the differences in AEEGHG in the NCP region are significant. In the temporal dimension, AEEGHG shows a generally good trend as a whole, but there are still many prefecture-level cities in the Lag area. From 2004 to 2022, the number of cities in the Starting area and the Progress area further decreased, while the Leading area and the Coordination area have increased accordingly.

5. Discussion

5.1. Spatial Heterogeneity of Greenhouse Gas Emissions

The spatial pattern of GHG emissions in the NCP is characterized by relatively “lower emissions in the north and higher emissions in the south”. This is the result of the combined effects of crop types, climate conditions, management practices, and policy interventions. Regarding agricultural planting structure and crop types, northern NCP mainly adopts a wheat-corn rotation system with higher mechanization but more intensive fertilizer inputs. Emissions mainly come from nitrogen fertilizer application and machinery utilization. By comparison, there is a higher proportion of rice cultivation in southern NCP (e.g., Jiangsu, Anhui). Moreover, the flooded paddies significantly increase CH₄ emissions. Since CH₄ has a global warming potential (GWP) around 28 times higher than that of CO₂ [42], this leads to elevated emissions.

The humid climate promotes the activity of soil microorganisms and accelerates the decomposition of organic matter in southern NCP. Meanwhile, the waterlogged conditions promote the anaerobic CH₄ production process in the paddy fields, which can increase CH₄ emissions. This is consistent with previous results [41]. Furthermore, in terms of policies, some studies suggest that GHG emissions in northern NCP are mainly driven by “carbon neutrality” policies, e.g., optimizing fertilization and biochar application, and strict regulations on agricultural carbon sink management [43], which aligns with our research findings.

GHG emissions exhibit a pattern of rising first and then falling over time in the NCP, reaching the peak in 2018. The results shows that although the GHG emissions decreased briefly in 2012, they were still higher than those in 2004. In fact, the GHG emissions have been fluctuating and increasing during 2004–2018, with the peak in 2018. In 2012, China released relevant policies to promote agricultural mechanization and replace chemical fertilizers and pesticides with organic fertilizers. However, the rapid implementation of this policy may have outpaced agricultural production practices. In 2018, the Ministry of Agriculture and Rural Affairs of China issued the “2018 Key Financial Policies for Strengthening Agriculture and Benefiting Farmers”, with the aim of promoting the research and development of low-consumption, ecological, cost-saving, and safe green technologies, precise fertilization, integrating with the smooth implementation of the agricultural policy in 2012. As a result, GHG emissions further decreased after 2018.

5.2. Influence of CH₄ and N₂O on Agricultural Eco-Efficiency

CH₄ and N₂O are key sources of GHG emissions in agriculture system. However, existing studies often focus on CO₂ as the only undesirable output [18,44]. Research on the characteristics, driving factors, and decoupling effect of agricultural GHG emissions examines CO₂ emissions from agricultural inputs and rice cultivation, but ignores the significance of CH₄ [45]. Meanwhile, some studies have examined the regional differences and convergence of China’s agricultural carbon efficiency in the carbon sink effect [46], which also didn’t integrate CH₄ and N₂O as undesirable outputs. Comparatively, this study integrated different undesirable output indicators and proposed multidimensional insights on AEE improvement, providing references for the formulation of sustainable agricultural policies.

Through a comparative analysis of AEE, it was found that AEECH₄ and AEEN₂O were underestimated, which is consistent with research results [9]. This may stem from differences in the GWP of CO₂, CH₄, and N₂O [9]. However, all the three GHGs were converted into CO₂ equivalent (CO₂e) in our embedded model. CH₄’s GWP value is approximately 28, while N₂O’s GWP is about 273. The contributions of CH₄ and N₂O would be significantly amplified [1], resulting in a relatively high AEE when they are regarded as undesirable outputs. In the study, CO₂ mainly originates from agricultural inputs and energy consumption; CH₄ from paddy fields; and N₂O from fertilizer application [27]. These activities are indispensable parts of agricultural production and directly reflect the agricultural management efficiency [47]. The study emphasizes the significant impacts of different GHG emissions on AEE and provides a more comprehensive perspective for related research.

5.3. Spatiotemporal Variations of Agricultural Eco-Efficiency

In Figure 5, AEEGHG exhibited fluctuations between 2004 and 2022 but displayed an overall upward trend. Since 2000, China’s No. 1 Central Policy Document has consistently highlighted the significance of the agricultural sector, promoting circular and eco-friendly farming practices. This underscores the government’s dedication to advancing sustainable agriculture and enhancing ecological efficiency. In this study, we found that AEEGHG decreased in 2007 and 2010. The occurrence of floods in the Huaihai River Basin of China in 2007 [41] led to a decline in grain production in many areas of NCP, thereby resulting in a decrease in AEEGHG. In the same year, severe low-temperature freeze damage occurred in the affected regions surrounding the NCP, leading to a reduction in grain production and a consequent decrease in AEEGHG. Despite these natural influences, AEEGHG quickly returned to the average level. the results show that 2018 was the year when AEEGHG reached the peak. The Ministry of Agriculture and Rural Affairs of China released the “2018 Key Agricultural Policy for Strengthening and Benefiting Farmers” in 2018, aiming to promote the development of low-consumption, ecological, cost-saving, and safe green technologies, precise fertilization, which is conducive to improving AEE.

The AEEGHG in the NCP shows significant spatial heterogeneity (Figure 6), which is highly correlated with the factors such as planting conditions, planting types, and natural conditions. The different application rates of fertilizers and pesticides affect efficiency levels [48]. The results show that the AEEGHG of Henan is the highest within the NCP. Henan is the core area for grain production in China, possessing abundant agricultural resources and a high level of agricultural mechanization. The promotion of smart agricultural systems in Henan enables precise irrigation and variable fertilization, reduces water waste, and increases the crop yield. Moreover, the efficiency of agricultural machinery utilization was improved through land transfer and management [49].

The levels of AEEGHG in Beijing and Tianjin were relatively high during 2004–2022, which results from their intensive facility agriculture systems and high yield. Intensive agricultural production reduces energy consumption, and the emission intensity of GHGs is low (Figure 3), which aligns with the relevant results [50]. Southern Jiangsu (e.g., Nanjing) adopts precision agricultural technologies with advanced economy. In contrast, northern Jiangsu carried out land consolidation to reduce resource redundancy, which leads to a relatively small overall difference in AEEGHG within Jiangsu [51]. Moreover, the surrounding areas of Nanjing adopt ecological agriculture such as rice-shrimp co-culture, which effectively reduces GHG emissions, reduces nitrogen fertilizer usage, and optimizes water management. As a result, Nanjing ranks the highest AEEGHG level within Jiangsu. This is consistent with the research results [52,53].

The AEEGHG level in Shandong ranks fourth among all provinces. There is a significant variation in the overall AEEGHG level in Shandong. The research reveals that the ecological service supply is higher in the southern and southwestern regions but lower in other areas. The area of ecological sources has decreased, while fragmentation has intensified [54]. This further highlights the spatial imbalance of AEE.

The overall AEEGHG in Anhui varies significantly and ranks fifth among all provinces. The AEEGHG levels of most prefecture-level cities in Anhui are low. Some studies found that the GHG emissions from the agricultural fertilizers accounted for 43.99% of the total emissions [55], which leads to low efficiency [56], which is consistent with previous results. The northern part of Anhui is dominated by a small-scale agricultural economy. Moreover, land fragmentation constrains technology dissemination. The severe redundancy of resources leads to low efficiency, which aligns with the existing results [57,58]. Within the NCP, Hebei has the lowest AEEGHG with an overall low efficiency level. The over-exploitation of groundwater in Hebei is severe, and the agricultural production mode with high fertilizer input and low efficiency exacerbates the unsustainability of the agricultural ecosystem and reduces AEEGHG, which is consistent with the research results [49].

5.4. Policy Implications

Reducing GHG emissions can be achieved by optimizing agricultural practices and technologies. For regions with high efficiency, such as Henan, these measures can consolidate their strengths, including promoting advanced technologies such as smart agriculture, and optimizing precise irrigation and variable fertilization, thereby enhancing resource utilization efficiency. Shandong, as a major grain-producing area, can also learn from Henan. For regions with low efficiency, such as Hebei, they should focus on addressing the problems of excessive groundwater extraction and excessive use of chemical fertilizers, promoting water-saving agriculture and ecological planting models, and promoting water-saving irrigation technologies, such as drip irrigation and sprinkler irrigation to reduce groundwater over-extraction, while combining soil moisture condition monitoring to achieve precise irrigation.

Agricultural production should be supported by low-carbon technologies. In regions with suitable climates, low-carbon, mechanization and precision agricultural technologies should be vigorously promoted to reduce CH₄ and N₂O emissions, such as rice–shrimp co-culture and organic fertilizer substitution for chemical fertilizers. Based on the results of AEE in the NCP, policy recommendations should focus on regional differences and the actual needs for GHGs reduction. Further implementing policies can contribute to strengthening grassroots implementation, such as the “Circular Economy Development Strategy” and “Fiscal Key Support for Agriculture and Farmers’ Policies”, ensuring that subsidies and technical support truly benefit farmers, and guiding the green production transformation.

Regarding the fragmentation of land, promoting land transfer and large-scale operation can improve resource utilization efficiency. In resource-scarce areas, priority should be given to developing water-saving irrigation and efficient planting models to alleviate environmental pressure. Additionally, facilitating cooperation among small and medium-sized farms, such as through shared use of agricultural machinery, could also improve eco-efficiency without compromising the resilience of local food systems.

Agricultural management efficiency should be improved from multiple dimensions. For scientific decision-making, it is recommended to establish a comprehensive monitoring system covering CO₂, CH₄ and N₂O, assess AEE regularly, and adjust policy directions dynamically. Other efforts can be made to strengthen cooperation among provinces and cities in the NCP to achieve regional coordinated development, enhance agricultural sustainability, and support the realization of the carbon neutrality goal.

5.5. Limitations and Future Work

The study primarily relies on the data obtained from statistical yearbooks and government reports. There may be data omissions or measurement errors, especially for the data at the prefectural-level city level. The records for some years and regions may be incomplete. Linear interpolation is used to compensate for such gaps. Due to limitations in data availability, we employed average carbon emission factors in the calculation of GHG emissions. This approach may introduce uncertainties compared to site-specific or process-specific emission factors. However, these data can still effectively support macro policy analysis, and their systematic deviations will not significantly affect the scientific nature of the overall conclusions and policy recommendations. In the future, the spatial resolution will be improved to the county level to analyze regional differences more accurately and formulate targeted policies. Additionally, integrating multisource datasets (remote sensing, meteorological, socioeconomic) into the robust assessment frameworks for AEE can enhance the applicability and accuracy of the model.

In terms of the application of the static SBM-DEA model, the changes in the frontier over a certain period cannot be captured. In the future, we would conduct further discussion by combining it with production indices, such as Malmquist index. Moreover, we would further refine the total GHG emissions, exclude the inevitable GHG emissions before assessing input efficiency. This will enhance the accuracy of AEE assessment. Future research can explore more flexible models or data preprocessing methods to optimize the analysis results.

6. Conclusions

As an important grain production base in China, the improvement of AEE in the NCP has great significance for achieving green development and carbon neutrality goals. Using the SBM-DEA model and the panel data of 75 prefecture-level cities in the NCP, this study evaluated the AEE of the NCP by constructing a non-desired output index system including various GHGs (CO₂, CH₄, N₂O), and analyzed its spatial and temporal characteristics and impacting factors. This study provides a theoretical basis and practical references for regional agricultural policy formulation and resource optimization allocation. The main conclusions are as follows:

(1)

The total agricultural GHG emissions indicated an upward trend from 2004 to 2018, and then slightly decreased after 2018 in the NCP. CO₂ emissions were the main contributor, but CH₄ and N₂O emissions cannot be ignored. Spatially. The GHG emissions presented a spatial pattern of “lower emissions in the north and higher in the south”, which was closely related to crop types, climatic conditions, and agricultural management measures.

(2)

Through comparative analysis, it was found that if only CO₂ were regarded as the undesirable output, the AEEs would be an underestimation. While if only CH₄ or N₂O were considered, the AEEs would be overestimation. This indicates that the multi-GHGs embedded model can comprehensively reflect the level of AEE.

(3)

AEEGHG has generally shown a fluctuating upward trend, with policy promotion and technological progress being the main driving factors. Spatially, AEEGHG exhibits a heterogeneous characteristic of “lower values in the north and south, higher values in the east and west”, with the regions in Henan, Beijing, and Tianjin having higher AEEs, while the regions in Hebei and Anhui have lower AEEs.

(4)

Considering spatial heterogeneity of AEEs, this study suggests differentiated measures. In high-AEE regions, efforts should be made to consolidate existing achievements and promote smart agriculture and precise technologies. In low-AEE regions, the focus should be on addressing the issue of excessive resource consumption and promoting water-saving irrigation and ecological planting models. Meanwhile, regional collaboration, low-carbon agricultural technologies, and GHG emissions monitor should be strengthened to contribute to the realization of carbon neutrality goal and agricultural sustainable development.