Spatiotemporal Evolution and Influencing Factors of Carbon Emission Efficiency of Apple Production in China from 2003 to 2022

Abstract

Understanding the carbon emission efficiency of apple production (APCEE) is critical for promoting green and low-carbon agricultural development. However, the spatiotemporal dynamics and driving factors of APCEE in China remain inadequately explored. This study employs life cycle assessment, super-efficiency slacks-based measures, and a panel Tobit model to evaluate the carbon footprint, APCEE, and its determinants in China’s two major production regions from 2003 to 2022. The results reveal that: (1) Producing one ton of apples in China results in 0.842 t CO₂e emissions. Land carbon intensity and total carbon emissions peaked in 2010 (28.69 t CO₂e/ha) and 2014 (6.52 × 10⁷ t CO₂e), respectively, exhibiting inverted U-shaped trends. Carbon emissions from various production areas show significant differences, with higher pressure on carbon emission reduction in the Loess Plateau region, especially in Gansu Province. (2) The APCEE in China exhibits a W-shaped trend (mean: 0.645), with overall low efficiency loss. The Bohai Bay region outperforms the Loess Plateau and national averages. (3) The structure of the apple industry, degree of agricultural mechanization, and green innovation positively influence APCEE, while the structure of apple cultivation, education level, and agricultural subsidies negatively impact it. Notably, green innovation and agricultural subsidies display lagged effects. Moreover, the drivers of APCEE differ significantly between the two major production regions. These findings provide actionable pathways for the green and low-carbon transformation of China’s apple industry, emphasizing the importance of spatially tailored green policies and technology-driven decarbonization strategies.

1. Introduction

As the foundation for human survival and development, the sustainability of food systems plays a crucial role in addressing climate change [1]. However, the food system accounts for 34% of global greenhouse gas emissions, with agriculture and land use responsible for 71% of this share. Agriculture has become one of the significant sources of greenhouse gas emissions [2]. As one of the world’s largest agricultural producers, China supports nearly 20% of the global population with less than 7% of the world’s arable land. However, this achievement is underpinned by the extensive use of chemical synthetics [3], resulting in severe environmental burdens, such as increased agricultural carbon emissions [4]. Under the dual-carbon targets, the agricultural sector faces severe pressure to reduce emissions.

In 2023, China’s apple cultivation area reached approximately 2 × 10⁶ ha and a production volume of 49.60 × 10⁶ t, accounting for 43% and 51% of global totals, respectively. Both metrics ranked first worldwide [5]. Undoubtedly, the apple industry has made significant contributions to increasing farmers’ incomes and promoting rural revitalization. However, the excessive use of chemical fertilizers and pesticides, coupled with insufficient organic fertilizer application, remains a major issue in China’s apple production [6]. In the short term, this may increase yields, but in the long run, it results in severe soil acidification, alkalinization, and a reduced capacity for water and nutrient retention. This not only diminishes apple yield and quality but also leads to significant greenhouse gas emissions, threatening both farmers’ health and the green development of agriculture. Therefore, promoting carbon reduction in China’s apple industry is crucial for achieving green and low-carbon agricultural development. In July 2024, the government’s opinion on accelerating the comprehensive green transformation of economic and social development emphasized the implementation of agricultural emission reduction and carbon sequestration actions, aiming to reduce and improve the efficiency of agricultural inputs such as fertilizers and pesticides. As an important indicator of green and low-carbon production [7], clarifying the carbon emission efficiency of apple production (APCEE) in China and exploring its influencing factors will help agricultural management authorities formulate more effective carbon reduction strategies and promote green and low-carbon transformation of the apple industry.

The main contributions of this study are as follows: First, from a research perspective, this study is the first to measure the carbon emissions of apple production and APCEE in China from 2003 to 2022. Meanwhile, comparative analyses of the changes in carbon emissions and APCEE in different times and different regions are conducted. This dynamic study not only provides a comprehensive understanding of the carbon footprint of apple production and APCEE in China but also provides important insights for regional synergistic carbon reduction. Second, the improvement of research methods. This study uses the life cycle assessment (LCA) method to measure carbon emissions from apple production. Unlike the existing literature that uses the carbon emission factor method to measure regional carbon emissions [8,9], this study uses the LCA method to measure the carbon footprint of apple production at both the national and main production region levels. This approach provides a solid database for a more comprehensive and accurate assessment of APCEE. Third, based on measuring carbon emissions from apple production using the LCA method, this paper combines the super-efficiency slacks-based measure (Super-SBM) and panel Tobit methods to study the key factors affecting China’s APCEE, so this study innovatively puts forward suggestions to promote the transformation of China’s apple industry into a green and low-carbon production at the regional level.

The remainder of this study is structured as follows. Section 2 demonstrates a literature review. Section 3 presents the materials and methods, including data sources, variable selection, research methods, and model construction. Section 4 provides the results analysis, which includes the spatiotemporal evolution of carbon emissions and APCEE in China, as well as an analysis of the factors influencing APCEE. Section 5 presents the discussion, and Section 6 presents the conclusions and limitations.

2. Literature Review

As an important indicator to measure green and low-carbon agriculture, agricultural carbon emission efficiency (ACEE) has received extensive attention and has formed a wealth of research. First, in terms of research content, most of the literature focuses on studying ACEE at the provincial level [8,10,11,12]. A small part of the literature focuses on micro-level ACEE. Wang et al. (2023) studied the effect of intercropping soybean, corn, and wheat on carbon emission efficiency [13], while Han et al. (2024) examined the carbon emission efficiency of state-owned farms in China [9]. Second, in terms of measurement methods, most scholars use the carbon emission coefficient method to calculate agricultural carbon emissions as the basis of non-desired outputs and measure the ACEE by constructing a Super-SBM model [9,11,14]. A few scholars have utilized other methods to measure the ACEE. Dong et al. (2025) used the ratio of net carbon sinks to carbon emissions to measure ACEE [15]. Yang et al. (2024) used the ratio of the value added of the primary industry in each province to the total amount of carbon emissions in each province to measure the ACEE [7]. Wang et al. (2023) measured the ACEE by the carbon emissions per unit of grain output [13]. Third, in terms of regional differences, China’s ACEE shows significant regional differences. Wang et al. (2025) showed that the ACEE in the Yangtze River Economic Belt is the best in the lower reaches, the second in the middle reaches, and the worst in the upper reaches [10]. Han et al. (2025) reported that the ACEE in half of the provinces in the sample was higher than the average, and the efficiency value in the east was higher than that in the west [14]. According to Wang and Feng (2021), China’s ACEE is generally on the rise, with the highest in the east, the middle in the center, and the lowest in the west [16]. Fourth, in terms of the influencing factors of ACEE, agricultural environmental supervision, agricultural investment, agricultural industry structure, crop damage, agricultural output share, agricultural mechanization level, digital technology application, agricultural environmental technology practice, and urbanization level positively affect ACEE [10,12,14,17]. By contrast, pesticide application [10], per capita GDP [12], farmers’ education level, aging of the rural labor force, and crop damage negatively influence the ACEE [17].

Apple production in China shows the characteristics of “double high”. First, the amount of chemical fertilizers and pesticides is high. The amount of fertilizer applied in orchards is seven times that in Europe and six times that in the United States. The amount of pesticides used also far exceeds the recommended value [18], resulting in a large amount of greenhouse gas emissions. Second, the labor input is high, and the mechanization level of apple production in China is low. Apple fertilization, spraying, pruning, bagging, picking, and other links rely on a large amount of labor, resulting in increased production costs [19]. Therefore, it is important to clarify the APCEE in China for the green and low-carbon transformation of the apple industry. The existing literature mainly studies the measurement of carbon emissions from apple production and the exploration of the factors affecting carbon emissions. First, in the measurement of carbon emissions from apple production. In terms of research scale, most of the literature focuses on the micro level [20,21,22], while only a few of the literature studies apple production carbon emissions at the regional level [23]. In terms of measurement methods, most of the literature utilizes LCA to measure the carbon footprint of apple production links [24,25,26] as well as other links [27,28]. A few researchers used the natural experiment approach [29] and the carbon emission coefficient method [30] to assess carbon emissions. In terms of period, most of the research samples were obtained through short-term farmer research [31] and natural experiments [21], and usually with a sample period of 1–3 years. Second, in terms of factors affecting carbon emissions from apple production, organic fertilizers replacing chemical fertilizers [6,32], application of biochar [33], optimized irrigation [34], and clean energy use [35] can reduce carbon emissions from apple production.

In summary, the existing literature presents several gaps. First, in terms of research subjects, most studies on ACEE focus on the macro agricultural level, with limited research on the APCEE and its influencing factors. Given that different crops exhibit significant differences in resource input and output efficiency, it is essential to construct life cycle inventories reflecting regional characteristics for specific crops to accurately assess their carbon emissions and ACEE. Second, in terms of research methodology, the aforementioned scholars generally use the carbon emission coefficient method to estimate the total agricultural carbon emissions and then apply the Super-SBM model to calculate ACEE, treating total carbon emissions (TCE) as undesirable outputs. This approach, however, overlooks the greenhouse gas emissions associated with agricultural input production and cannot account for indirect emissions from carbon sources, which may result in an underestimation of the TCE. The LCA method, on the other hand, can account for both direct and indirect emissions from different production stages, and its results can be further characterized as needed, thus providing a comprehensive and accurate assessment of carbon emissions. However, few studies have combined the LCA method with the Super-SBM model to measure ACEE. Third, in terms of spatial scale, existing studies have mainly focused on the carbon footprint of apple production in specific regions, with limited research on national-level carbon emissions from apple production. Due to the significant heterogeneity in natural endowments, production methods, and technological levels across different regions in China, there are notable variations in carbon emissions and APCEE. Moreover, existing literature lacks comparative analyses of carbon emissions and APCEE at both national and regional levels. Fourth, in terms of temporal scale, current studies primarily rely on cross-sectional data from apple production inputs at the farm level or short-term experimental data. Few studies have used multi-year data to examine the dynamic evolution of carbon emissions and carbon emission efficiency in China’s apple production.

Therefore, to address these gaps in the literature, this study first employs the LCA method to measure the cradle-to-gate carbon footprint of apple production in China and its major production regions from 2003 to 2022. Second, the Super-SBM model was used to assess APCEE and analyze its spatial and temporal evolution by considering both desirable and undesirable outputs. Third, a panel Tobit model is used to explore the main factors affecting APCEE. Finally, this paper proposes policy recommendations to promote the green and low-carbon transformation of China’s apple industry.

3. Materials and Methods

3.1. Data Sources

Given that the “Compilation of National Agricultural Product Cost and Benefit Data” only provides 8 provincial level apple production input data, and considering the continuity of the data, this study selected the production input data of 8 major apple-producing provinces in China from 2003 to 2022. The specific sample provinces are shown in Figure 1, which includes the Loess Plateau region (comprising Shaanxi, Gansu, Henan, and Shanxi provinces) and the Bohai Bay region (comprising Beijing, Shandong, Hebei, and Liaoning provinces). In order to increase the sample capacity, the mean values of the two main apple-producing regions are also included in the sample in this paper. Therefore, the data composition of this study is 10 provincial units over a period of 20 years, totaling a sample size of 200. Moreover, in 2022, the apple cultivated area and production volume in these two major apple production regions accounted for 81.81% and 87.59% of the national totals, respectively. Therefore, the research sample is highly representative.

The data required for this study, including sources of carbon emissions, labor inputs, and capital inputs to apple production, were obtained from the tables of apple cost benefits, expenses, and labor and fertilizer inputs by region in the “Compilation of National Agricultural Product Cost and Benefit Data”. The carbon sources and emission coefficients needed for carbon emission estimation are obtained from the Ecoinvent V3 of the life cycle inventories database and existing literature, with detailed information provided in Appendix A.1. Data on apple yield, apple output value, apple planting area, orchard area, total cropped area, disaster-affected area, total agricultural machinery power, agricultural output value, and rural electricity consumption are sourced from the “China Rural Statistical Yearbook”. Expenditures on agriculture, forestry, and water resources, as well as total fiscal expenditure, is obtained from provincial statistical yearbooks. The rural consumer price index is provided by the National Bureau of Statistics. Data on workers with junior high school education and above in rural areas are from the “China Population and Employment Statistical Yearbook”. Green patent data are sourced from the China Research Data Service (CNRDS).

3.2. Variable Selection

3.2.1. Dependent Variable

The dependent variable in this study is the APCEE. The Super-SBM model, which accounts for both desired and undesirable outputs, is used to measure APCEE. Specifically, the input indicators include three factors: labor, land, and capital. These are measured by the labor inputs, the land area, and the total cost of materials and services required to produce 1 ton of apples. The total cost of materials and services includes both direct and indirect costs. Direct costs consist of expenses related to fertilizers, pesticides, leased operations, tools and materials, and maintenance, while indirect costs include insurance fees and administrative expenses. Moreover, output indicators encompass both desired and undesirable outputs. The desired output is represented by the value of 1 ton of apples produced, while the undesirable output is represented by the carbon emissions generated from producing 1 ton of apples. To smooth out temporal differences, capital costs and the value of apple production are adjusted to constant prices with 2003 as the base year.

3.2.2. Independent Variables

Investigating the factors influencing the APCEE is a crucial step toward facilitating the green and low-carbon transformation of the apple industry. Based on the factors identified in the existing literature regarding ACEE.

Table 1. Description and definition of variables.

Variable Type	Variable Name	Variable Symbol	Variable Definition	Reference Source
Dependent variable	Apple production’s carbon emission efficiency	APCEE	The Super-SBM model is used to measure APCEE	[8]
	Apple cultivation structure	Culstruc	Apple cultivated area/total orchard area	[36]
	Apple industry structure	Indstruc	Apple output value/total agricultural output value	[9]
	Educational level of rural labor force	Education	Rural laborers with a middle school education or higher/total rural population	[37]
	Crop damage extent	Damage	Crop area affected by disasters/total sown area	[12]
Independent variables	Degree of agricultural mechanization	Macpower	Total agricultural machinery power/total sown area	[7]
	Rural electricity consumption	Electric	Total electricity consumption in rural areas/total rural population	[4]
	Green technology innovation	Greenpat	The sum of the green invention patents and the green utility model patents	[38]
	Agricultural subsidy	Agriexp	Agricultural, forestry, and water expenditures/total fiscal expenditure	[12]

shows the specific definitions of the variables.

3.3. Research Methods

3.3.1. Calculation Method for Carbon Emissions in Apple Production

This study employs the LCA method to calculate carbon emissions in apple production. LCA is an approach used to assess the environmental impacts of a product, service, or process throughout its entire life cycle. Compared to other methods for calculating carbon emissions, LCA offers several key advantages. First, it provides comprehensive coverage of the product life cycle. LCA can account for carbon emissions across all stages, including raw material extraction, manufacturing, transportation, use, and end-of-life disposal. In contrast to methods that focus solely on one stage while neglecting others, LCA offers a more accurate representation of the product’s overall carbon footprint. Second, LCA considers a wide range of carbon sources. In addition to evaluating carbon emissions from primary energy consumption, LCA also includes other indirect emission sources, such as emissions from raw material production and the use of auxiliary materials. Third, LCA follows a scientific evaluation framework. By integrating the environmental impacts of each stage, LCA clarifies the interrelationships between stages, which can contribute to the development of more scientifically sound carbon reduction strategies. Furthermore, LCA adheres to the ISO 14040–ISO 14044 standards and is widely used for evaluating environmental impacts on agricultural production [39,40]. Therefore, this study employs the LCA method to calculate carbon emissions in apple production. The specific steps include selecting the functional unit, defining system boundaries, conducting inventory analysis, and calculating carbon emissions.

(1)

Functional unit. In agricultural LCA studies, the functional unit is typically defined in terms of the output of a unit of agricultural product. In this study, the functional unit is set as the production of 1 ton of apples.

(2)

System boundaries. The system boundaries of this study encompass the production input processes within a one-year operational cycle of an apple orchard, specifically, including the agricultural input and production stages (Figure 2). The agricultural input production stage primarily covers the production processes of fertilizers, pesticides, diesel, agricultural films, and electricity. The agricultural production stage includes key operations such as weeding, fertilization, pesticide application, flower thinning and fruit shaping, bagging, and harvesting.

(3)

Inventory analysis. Life cycle inventory analysis quantifies the inputs (e.g., raw materials, energy) and outputs (e.g., emissions) at each stage of a product or service’s life cycle. The data collection for apple cultivation is divided into two parts: first, the data collection for the agricultural inputs production stage, which includes production data on fertilizers, chemicals, agricultural films, and electricity; second, the data collection for the agricultural production stage, which involves the resource inputs and environmental impacts of processes such as fertilization, pesticide application, and weeding. The life cycle inventories of carbon emissions during the agricultural production stage of apple production are presented in

Table 2. Life cycle inventories of the agricultural production stage of apple production in China.

Carbon Source	N kg	P2O5 kg	K2O kg	Pesticide kg	Diesel kg	Film kg	Electricity kW·h	Labor h
2003	8.58	8.58	8.58	1.54	1.09	0.37	26.48	304.8
2004	11.09	5.57	4.41	1.14	1.00	0.03	27.47	341.6
2005	12.98	7.14	7.25	1.02	0.43	0.92	25.36	318.56
2006	12.23	6.84	7.26	1.40	1.10	0.15	31.93	333.12
2007	15.75	7.38	8.34	1.45	1.32	9.63	30.66	290.24
2008	17.37	6.15	5.94	1.49	1.63	0.31	22.25	315.28
2009	15.31	7.09	6.98	1.71	1.90	2.02	36.68	338.4
2010	17.44	8.49	8.99	1.81	2.07	1.07	34.77	349.52
2011	16.14	8.78	8.70	1.39	2.69	0.91	19.85	322.56
2012	14.23	7.47	7.53	1.29	2.95	1.10	17.12	322.96
2013	16.68	8.10	7.84	1.33	2.36	1.14	17.23	303.12
2014	17.91	8.91	9.07	1.37	4.18	1.01	20.44	320.8
2015	15.13	7.82	8.15	1.18	3.12	0.84	16.42	299.12
2016	13.37	7.41	7.76	1.03	3.27	0.85	17.90	300.4
2017	12.28	7.17	7.54	1.03	2.19	0.73	19.36	283.84
2018	14.74	9.21	9.65	1.34	3.31	0.92	23.75	270.8
2019	10.58	6.91	7.30	0.96	3.84	0.78	17.94	298.8
2020	12.34	7.67	8.03	0.99	3.52	0.93	14.38	297.44
2021	12.40	8.72	9.29	1.18	2.85	0.91	15.11	262.08
2022	13.06	9.29	9.78	1.24	3.39	1.34	17.03	255.84

.

(4)

Carbon emission accounting: SimaPro 9.1.0.11 is a widely used software tool for LCA, incorporating multiple LCA methodologies such as the IPCC guidelines, IMPACT, and the ReCiPe. Given that this study evaluates the carbon emissions across the entire life cycle of apple production, the IPCC 2013 GWP 100a V1.03 method, specifically designed for carbon emission calculations in SimaPro 9.1.0.11, is employed to estimate the greenhouse gas emissions throughout the apple production life cycle, with emissions expressed in CO₂e.

3.3.2. Super-SBM Model

This study employs the Super-SBM model to assess the APCEE. The Super-SBM model, which is based on non-desired outputs, is a non-parametric method used to evaluate the efficiency of Decision-Making Units (DMUs) [41]. It is an extension of Data Envelopment Analysis (DEA). Compared to other models, Super-SBM offers the following advantages. First, it is particularly suitable for cases involving undesirable outputs. Super-SBM directly incorporates them as negative variables without requiring prior weighting assumptions, thus avoiding the bias introduced by traditional DEA models that treat emissions as “desirable” outputs. Second, for apple production inputs, which exhibit both redundancy and insufficiency, Super-SBM simultaneously measures radial and non-radial slack. This enables a more precise assessment of the efficiency of DMUs. Based on the study by Elahi et al. (2024) [8], the model is constructed as follows:

$$APCEE=min{\displaystyle \frac{\frac{1}{m}\sum _{i=1}^m\frac{\overline{x_i}}{x_{ik}}}{\frac{1}{q_1+q_2}\left(\sum _{r=1}^{q_1}\frac{\overline{y_r}}{y_{rk}}+\sum _{t=1}^{q_2}\frac{\overline{y_t^b}}{y_{tk}^b}\right)}}$$

(1)

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

(2)

In Equation (1), $APCEE$ represents the efficiency value of the DMU. A DMU is considered fully efficient when $APCEE\ge 1$, and exhibits efficiency loss when $APCEE<1$. Vectors $x_{ik}$, $y_{rk}$, and $y_{tk}^b$ correspond to inputs, desired outputs, and undesired outputs, respectively. The variables $\overline{x_i}$, $\overline{y_i}$, and $\overline{y_t^b}$ represent the slack variables for inputs, desired outputs, and undesired outputs. $k$ refers to the $k$-th DMU out of $n$ total DMUs, and $m$ represents the number of DMUs. $q_1$ and $q_2$ are the numbers of desired and undesired outputs, respectively. In Equation (2), $\lambda$ denotes the weight vector. This paper was calculated using MAXDEA 9 Ultra software, and returns to scale were calculated using the assumption of constant returns to scale.

3.3.3. Tobit Model

To explore the factors influencing APCEE, this study adopts the following steps to select the empirical model. First, to prevent multicollinearity among the variables, a multicollinearity test is conducted. The results show that the mean variance inflation factor (VIF) is 3.11, with the maximum VIF being 5.40, both of which are below 10, indicating no multicollinearity issues among the variables. Second, to avoid spurious regression caused by non-stationary variables, a unit root test is performed on the panel data. The results indicate that all variables pass the unit root test. Third, considering that the APCEE values in this study range from 0.2 to 1.6, making it a constrained dependent variable, the panel Tobit model is selected [42]. Given that the fixed effects Tobit model fails to capture sufficient statistics for individual heterogeneity and cannot perform conditional maximum likelihood estimation and adding dummy variables for panel units in the mixed Tobit model results in inconsistent fixed effects estimates, the study proceeds with the random effects panel Tobit model. Finally, the likelihood ratio (LR) test for the random effects Tobit model shows a p-value of 0.00, strongly rejecting the null hypothesis and suggesting the presence of individual effects. Therefore, the random effects panel Tobit model is chosen. The model is constructed as follows:

$$\begin{array}{c}{APCEE}_{i,t}={\alpha }_0 +{\alpha }_1{Culstruc}_{i,t}+{\alpha }_2{Indstruc}_{i,t}+{\alpha }_3{Education}_{i,t}+{\alpha }_4{Damage}_{i,t}\\ +{\alpha }_5{Macpower}_{i,t}+{\alpha }_6{Electric}_{i,t}+{\alpha }_7{Greenpat}_{i,t}+{\alpha }_8{Agriexp}_{i,t}+{\epsilon }_{i,t}\end{array}$$

(3)

$$\begin{array}{c}{APCEE}_{i,t}={\beta }_0 +{\beta }_1{Culstruc}_{i,t}+{\beta }_2{Indstruc}_{i,t}+{\beta }_3{Education}_{i,t}+{\beta }_4{Damage}_{i,t}\\ +{\beta }_5{Macpower}_{i,t}+{\beta }_6{Electric}_{i,t}+{\beta }_7{Greenpat}_{i,t-1}+{\beta }_8{Agriexp}_{i,t-1}+{\epsilon }_{i,t}\end{array}$$

(4)

$$\begin{array}{c}{APCEE}_{i,t}={\gamma }_0 +{\gamma }_1{Culstruc}_{i,t}+{\gamma }_2{Indstruc}_{i,t}+{\gamma }_3{Education}_{i,t}+{\gamma }_4{Damage}_{i,t}\\ +{\gamma }_5{Macpower}_{i,t}+{\gamma }_6{Electric}_{i,t}+{\gamma }_7{Greenpat}_{i,t-3}+{\gamma }_8{Agriexp}_{i,t-3}+{\epsilon }_{i,t}\end{array}$$

(5)

In Equation (3), $APCEE$ is the dependent variable. $Culstruc$, $Indstruc$, $Education$, $Damage$, $Macpower$, $Electric$, $Greenpat$, and $Agriexp$ represent independent variables. ${\alpha }_0$ denotes the constant, while ${\alpha }_1$ to ${\alpha }_8$ are the regression coefficients, $\beta$ and $\gamma$ have the same meaning as $\alpha$. $\epsilon$ is the random disturbance term, $i$ represents a specific sample province, and $t$ denotes a specific year. In Equations (4) and (5), considering the time lag effect of green patent applications in contributing to emission reduction [43] and the time delay in the impact of government subsidies on environmental changes [44], and acknowledging that apple trees typically bear fruit after about three years, this study incorporates a one-year and a three-year lag for green technology innovation and agricultural subsidies, respectively. The meanings of the other variables remain consistent with those in Equation (3).

4. Results

4.1. Spatiotemporal Evolution of Carbon Emissions in Apple Production in China

4.1.1. Analysis of National Carbon Emissions in Apple Production

This study employs the LCA model to estimate the carbon footprint of apple production in China from 2003 to 2022. The specific values and trends are presented in Figure 3 and Figure 4, and

Table 3. The table of YCI, TCE, LCI, and RCI values for apple production in China.

	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
YCI (t CO2e/t)	0.69	0.65	0.77	0.77	1.04	0.89	0.92	1.03	0.93	0.83	0.92	1.00	0.86	0.79	0.74	0.91	0.68	0.76	0.80	0.85
TCE (106 t CO2e)	31.41	35.67	40.09	42.31	52.69	52.53	55.29	61.40	60.00	57.25	61.63	65.20	62.59	46.54	45.75	43.51	41.74	44.06	39.17	39.53
LCI (t CO2e/ha)	16.53	19.01	21.21	22.28	26.86	26.37	26.98	28.69	27.55	25.66	27.13	28.26	26.88	23.92	23.50	22.44	21.10	22.10	19.83	20.21
RCI (t CO2e/104 yuan)	6.57	6.28	5.75	5.32	5.09	6.56	4.26	3.66	4.14	3.96	4.66	4.42	5.09	5.58	4.89	4.19	4.70	5.06	6.55	4.66

. First, as shown in Figure 3a, the YCI (defined as carbon emissions per ton of apples produced) in China exhibits an “M” shaped pattern. The likely explanation is that apple production alternates between high-yield and low-yield years. In high-yield years, farmers tend to increase inputs such as agricultural materials, labor, and machinery, while in low-yield years, inputs are reduced. Consequently, YCI fluctuates according to the changes in farmers’ resource inputs. Specifically, the carbon emissions shared by stage and source in the production of 1 ton of apples in China in 2022 are shown in Figure 4. It can be observed that the carbon emissions from the agricultural inputs production stage account for 51%, while the agricultural production stage accounts for 49%. Similar findings have been reported by Zhang et al. (2023) and Cheng et al. (2022) [31,45]. Among all carbon sources during the production cycle, fertilizers contribute the largest share, accounting for 80%, a result consistent with previous studies [18,31]. The carbon emission distribution by stage and carbon source in the Loess Plateau and Bohai Bay regions is similar to the national average.

Second, as shown in Figure 3b, the TCE of apple production exhibits an inverted “U” shape and peaked at 65.2 × 10⁶ t CO₂e in 2014. Third, as shown in Figure 3c, the LCI (defined as carbon emissions per hectare of land) of apple production also follows an inverted “U” shape, with trends similar to those observed in TCE. LCI peaked at 28.69 t CO₂e/ha in 2010. Fourth, as shown in Figure 3d, the RCI (defined as carbon emissions per 10⁴ yuan CNY worth of apples) exhibits fluctuating patterns within a certain range. In 2015, the Ministry of Agriculture set the target of zero growth in the use of fertilizers and pesticides by 2020 and also proposed the “one control, two reductions, and three basics” goal for agricultural non-point source pollution. By the end of the same year, the Chinese government officially incorporated green development as one of its new development concepts, marking the beginning of the green transformation of agriculture. As shown in Figure 4 and Figure 5 the dominant role of fertilizer use in the carbon footprint of apple production, coupled with the peak fertilizer application per hectare in 2014, contributed to the peak of TCE in that year. With the decline in both the apple planting area and LCI, the TCE is expected to further decrease.

4.1.2. Analysis of Carbon Emissions of Apple Production in Major Production Regions

This section further analyzes the carbon emissions from apple production in the two major production regions, encompassing eight major production provinces. The relevant indicators include YCI (

Table A2. YCI in the two major production regions (t CO₂e/t).

Regions	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
Bohai Bay	0.75	0.81	0.83	0.88	1.05	1.00	1.03	1.09	1.02	0.94	1.05	1.11	0.89	0.74	0.73	0.89	0.71	0.73	0.74	0.78
Beijing	0.82	0.76	0.74	0.71	1.42	1.16	0.77	0.89	0.88	0.65	0.88	1.05	0.83	0.43	0.36	0.82	0.40	0.42	0.61	0.76
Hebei	0.75	0.77	0.78	0.88	0.96	0.90	1.30	1.14	0.91	0.94	1.02	1.03	0.98	0.77	0.91	0.90	0.88	0.99	0.70	0.78
Liaoning	0.71	0.90	0.75	1.06	0.67	0.94	0.91	1.20	1.04	1.07	1.06	1.13	0.74	0.85	0.67	0.88	0.58	0.66	0.73	0.72
Shandong	0.71	0.80	1.04	0.87	1.16	1.00	1.13	1.14	1.27	1.09	1.26	1.25	1.00	0.92	0.99	0.96	0.96	0.86	0.91	0.89
Loess plateau	0.65	0.59	0.83	0.76	0.84	0.72	0.89	1.01	0.85	0.82	0.91	1.02	1.02	1.02	0.96	1.21	0.83	0.98	0.87	0.98
Shanxi	0.57	0.50	0.65	0.63	0.72	0.66	0.78	0.96	0.74	0.95	1.02	1.03	1.04	0.85	0.80	1.10	0.71	0.86	0.86	0.88
Henan	0.69	0.65	1.17	1.01	0.92	0.70	0.93	1.01	0.85	0.74	0.72	0.82	0.91	0.87	0.82	0.93	0.79	0.92	0.73	0.77
Shaanxi	0.71	0.65	0.74	0.78	0.96	0.95	1.11	1.27	1.01	0.76	0.85	0.96	1.00	1.00	1.02	1.20	0.92	0.89	0.85	0.92
Gansu	0.62	0.54	0.76	0.63	0.75	0.56	0.73	0.81	0.79	0.83	1.05	1.27	1.13	1.35	1.21	1.59	0.89	1.26	1.05	1.34

), TCE (

Table A3. TCE in the two major production regions (10⁶ t CO₂e).

Regions	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
Bohai Bay	4.70	4.84	4.96	4.62	5.70	5.25	5.96	5.76	5.71	5.83	6.47	6.03	5.79	4.00	3.99	3.69	3.84	3.70	3.29	3.34
Beijing	0.21	0.22	0.20	0.19	0.23	0.26	0.16	0.20	0.18	0.16	0.17	0.19	0.20	0.10	0.08	0.14	0.08	0.08	0.06	0.05
Hebei	5.54	5.93	5.91	5.64	6.71	6.13	7.78	5.63	5.56	5.82	6.81	6.97	6.83	2.70	3.16	2.97	3.20	3.50	2.74	3.19
Liaoning	1.29	2.17	1.76	2.18	2.43	3.10	3.63	4.06	3.89	3.59	4.07	3.59	2.88	2.55	2.45	2.28	1.84	2.16	1.83	1.71
Shandong	11.75	11.05	11.97	10.48	13.42	11.51	12.29	13.15	13.22	13.76	14.84	13.39	13.26	10.67	10.29	9.35	10.23	9.06	8.51	8.39
Loess plateau	3.24	4.07	3.78	5.67	6.23	5.98	7.99	8.61	8.32	7.94	8.72	9.71	10.34	8.80	8.74	8.17	7.54	8.25	6.96	6.87
Shanxi	2.22	2.45	2.31	3.47	2.97	3.07	3.54	3.63	3.62	4.67	4.79	4.59	4.71	4.45	4.46	2.96	3.57	3.50	3.27	2.81
Henan	2.59	4.36	3.59	5.95	4.57	4.26	5.33	5.32	5.20	5.05	4.22	4.90	6.50	5.38	4.76	4.15	3.30	4.14	2.75	2.96
Shaanxi	6.30	6.96	6.46	9.69	12.90	12.59	18.36	19.25	17.78	15.07	16.27	18.16	18.61	16.22	16.48	17.21	16.17	16.84	15.19	14.45
Gansu	1.86	2.51	2.77	3.58	4.49	4.01	4.74	6.24	6.68	6.96	9.58	11.19	11.53	9.13	9.26	8.37	7.12	8.52	6.64	7.26

), LCI (

Table A4. LCI in the two major production regions (t CO₂e/ha).

Regions	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
Bohai Bay	20.27	22.93	22.54	24.35	30.43	30.75	31.94	31.75	30.28	29.76	29.78	30.44	29.43	23.78	23.21	22.45	22.16	22.81	19.86	21.27
Beijing	16.22	17.17	18.72	19.70	22.55	28.35	19.99	24.22	23.49	20.29	23.07	27.14	29.52	14.78	11.75	21.81	12.17	12.67	10.19	11.27
Hebei	20.05	22.24	22.40	22.28	26.86	25.15	33.03	21.23	23.50	24.71	28.70	28.92	28.15	22.92	25.87	24.90	25.51	27.82	23.87	27.73
Liaoning	11.23	19.45	16.00	19.99	22.67	27.19	29.75	32.22	29.05	25.83	26.26	22.75	17.88	18.06	17.51	16.61	13.49	15.54	13.72	13.19
Shandong	32.90	32.46	34.95	33.70	44.01	41.67	45.44	49.69	47.84	49.20	48.90	43.94	44.24	39.45	38.76	36.25	41.47	36.77	35.00	34.93
Loess plateau	14.44	18.34	19.39	24.21	22.53	21.59	26.16	28.12	26.84	26.88	28.19	30.72	33.75	33.29	32.95	29.91	26.91	30.51	25.63	26.24
Shanxi	14.39	16.02	15.26	23.74	20.57	20.69	24.38	26.36	25.04	30.99	31.06	28.94	30.27	29.08	29.34	20.05	24.46	24.24	23.48	21.07
Henan	15.77	26.47	21.65	35.47	25.04	24.62	30.32	29.93	28.79	28.22	23.89	28.47	38.17	34.11	32.28	32.14	27.68	35.20	26.09	28.39
Shaanxi	15.70	16.89	15.16	20.96	26.61	23.71	32.51	32.00	28.53	23.35	24.46	26.64	26.77	28.14	28.12	28.80	26.31	27.15	24.45	23.45
Gansu	11.10	14.48	15.08	17.28	18.12	16.26	18.13	23.24	24.32	24.53	33.02	37.97	39.13	39.67	40.20	35.72	29.54	34.28	26.38	28.31

), and RCI (

Table A5. RCI in the two major production regions (t CO₂e/10⁴ yuan).

Regions	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
Bohai bay	6.48	6.20	4.82	4.21	4.95	6.44	4.53	3.64	3.91	3.82	4.22	4.00	4.29	4.00	3.58	3.18	3.48	3.85	5.37	4.04
Beijing	6.79	6.52	3.20	2.13	3.53	5.18	1.92	1.84	1.97	1.49	1.89	2.43	2.46	1.45	1.05	2.06	1.06	1.18	2.09	1.92
Hebei	6.66	7.19	5.83	5.57	5.88	6.55	8.31	4.77	4.42	4.60	5.79	5.38	6.00	5.09	5.62	4.37	5.64	7.04	7.14	5.22
Liaoning	5.48	5.74	4.69	5.29	4.01	7.17	5.05	4.98	5.99	6.92	6.83	5.78	5.37	6.96	4.94	5.40	5.87	6.42	8.04	5.23
Shandong	6.51	5.70	6.26	5.51	5.51	6.95	5.77	4.61	5.40	5.83	6.80	4.42	5.97	6.09	5.84	4.10	5.80	5.61	7.18	4.76
Loess plateau	7.29	8.03	7.61	6.50	5.12	7.13	7.09	4.96	4.86	5.12	5.77	5.16	6.38	7.17	6.58	5.84	6.41	6.34	6.62	5.31
Shanxi	6.15	7.58	5.26	5.67	5.00	6.93	7.60	6.18	4.93	7.09	8.57	6.86	8.79	7.74	6.85	5.59	6.91	5.74	7.71	5.14
Henan	8.21	13.36	9.55	10.23	6.94	9.50	9.55	5.31	7.12	6.63	6.14	5.78	8.64	8.01	7.71	6.81	6.77	7.33	5.26	4.75
Shaanxi	6.28	6.08	4.53	5.85	5.69	7.69	6.99	5.90	5.09	3.51	3.91	3.58	4.48	4.99	5.02	4.52	5.38	4.90	6.00	4.28
Gansu	9.55	6.41	8.99	4.73	3.49	4.70	4.86	3.18	3.30	4.30	5.74	5.26	5.34	8.06	6.76	6.15	6.92	7.32	7.72	6.70

).

Table A2 presents the YCI in the two major production regions from 2003 to 2022. From 2003 to 2022, the YCI increased by 50.77% in the Loess Plateau region and only 4% in the Bohai Bay region. From 2003 to 2022, YCI decreased by 7.32% in Beijing, while in the other seven major production provinces, it increased. The YCI in the Loess Plateau region was significantly higher than in the Bohai Bay region. As shown in Figure 6a, before 2014, YCI in the Loess Plateau region was lower than that in the Bohai Bay region. After 2014, they were consistently higher. This can likely be attributed to the relatively higher fertilizer inputs in the Bohai Bay region before 2014 (Figure 5), after which fertilizer use in the Loess Plateau region surpassed that in the Bohai Bay region.

Table A3 presents the TCE in the two major production regions from 2003 to 2022. Except for the period from 2003 to 2005, the TCE in the Loess Plateau region was higher than that in the Bohai Bay region (Figure 6b). In the Bohai Bay region, except for Liaoning Province, where TCE in 2022 increased by 32.56% compared to 2003, the other three provinces saw a reduction in TCE. In the Loess Plateau region, all four provinces experienced an increase in TCE, with Gansu Province showing the largest increase at 290.32% and Henan Province showing the smallest increase at 14.29%. This phenomenon can likely be attributed to two factors: first, apples, as an economic crop, have seen efforts from local governments to moderately expand the planting area in the context of rural revitalization, thus increasing farmers’ income. Second, the Loess Plateau region’s advantageous geographical resources and relatively low input costs have led to a gradual shift in apple production toward this region, resulting in a significant increase in carbon emissions from apple production in the Loess Plateau.

Table A4 presents the LCI in the two major production regions from 2003 to 2022. From 2003 to 2022, the LCI in the Loess Plateau region increased by 81.72%, while the Bohai Bay region increased by only 4.93%. As shown in Figure 6c, before 2014, the LCI in the Bohai Bay region was higher than the national average and that of the Loess Plateau, peaking at 31.94 t CO₂e/ha in 2009. This can be attributed to the region’s relatively stronger economic foundation, which led to higher resource inputs per unit of land area, resulting in higher LCI. However, after 2014, the Loess Plateau region surpassed Bohai Bay in terms of LCI, reaching its peak of 33.75 t CO₂e/ha in 2015. Afterwards, the LCI fluctuated and decreased, but it remained higher than both the national average and that of Bohai Bay. On the one hand, the expansion of the planting area in the Loess Plateau may have led to increased use of agricultural inputs such as fertilizers and pesticides, thereby raising LCI. On the other hand, the region may still face challenges related to relatively underdeveloped apple planting technologies and management practices, lacking advanced agricultural techniques and management models to support greener production.

Table A5 presents the RCI in the two major production regions from 2003 to 2022. From 2003 to 2022, the RCI in the Loess Plateau region decreased by 27.16%, while in the Bohai Bay region, it decreased by 37.65%. Compared to 2003, all eight major production provinces exhibited a decrease in RCI by 2022, with the largest reduction occurring in Beijing (71.72%) and the smallest in Liaoning Province (4.56%). The average decrease across all provinces was 30.63%. Notably, the reduction in RCI was more significant in the Bohai Bay region than in the Loess Plateau region. As shown in Figure 6d, the Loess Plateau region has a higher RCI than both the national average and the Bohai Bay region, confirming the earlier analysis of LCI.

Due to the significant correlation between carbon emissions and the area of apple cultivation in each major production region, this section focuses on classification analysis based on YCI, RCI, and LCI [46]. The classification results are shown in Figure 7. This study divides the period from 2003 to 2022 into four intervals. The carbon emissions in the major production regions are classified into eight types: the “LLL” type, which represents low YCI, low RCI, and low LCI; the “LLH” type, representing low YCI, low RCI, and high LCI; and the “HHH” type, representing high YCI, high RCI, and high LCI. The remaining types follow this pattern.

In Figure 7a, between 2003 and 2007, Hebei and Henan provinces belong to the “HHH” type, Shandong province is categorized as “HLH”, and Liaoning province is classified as “LLL”. The remaining provinces are categorized as having “two low and one high” type. The carbon emissions in the Loess Plateau region were overall better than those in the Bohai Bay region. In Figure 7b, between 2008 and 2012, Liaoning and Shaanxi provinces are classified as “HHH”, Henan and Shandong provinces belong to the “two high and one low” type, Gansu and Beijing provinces are categorized as “LLL”, and Hebei and Shanxi provinces are classified as “two low and one high”. The carbon emissions in Shaanxi and Liaoning provinces showed the fastest decline compared to the previous period. In Figure 7c, between 2013 and 2017, Gansu and Shanxi provinces are categorized as “HHH”, Henan and Shandong provinces are “two high and one low”, Beijing and Hebei provinces belong to the “LLL” type, and Liaoning and Shaanxi provinces are classified as “two low and one high”. Gansu province experienced a faster decrease than in the previous period, and the carbon emissions in the Bohai Bay region were generally better than those in the Loess Plateau region. In Figure 7d, between 2018 and 2022, Gansu province is classified as “HHH”, Shaanxi, Shanxi, Henan, and Shandong provinces are “two high and one low”, Beijing and Hebei provinces remain as “LLL”, and Liaoning is categorized as “LHL”. Compared to the previous period, the Bohai Bay region showed no change, while carbon emissions in the Loess Plateau region continued to increase, indicating an upward trend in carbon reduction pressure, particularly in Gansu province.

4.1.3. Scenario Analysis

As the economy develops and people pursue higher quality dietary standards, green and low-carbon apples are favored. To meet the market demand, some fruit farmers in China have started to shift to green and low-carbon cultivation. Therefore, in this paper, concerning [47], the low-carbon scenarios (LCS) are set up such that the quantities of nitrogen, phosphate, and potash fertilizers should be reduced by 33%, 38% and 35%, respectively, and pesticides by 99% under the 2022 input scenario. As can be seen in Figure 8, the YCI decreases significantly in all regions under the LCS. At the national level, the YCI under the LCS is 0.55 t CO₂e/t, a 35% decrease compared to 2022. The Loess Plateau and Bohai Bay regions show similar decreasing trends.

4.2. Spatiotemporal Evolution of APCEE in China

4.2.1. Analysis of National APCEE

This study evaluates the APCEE using the Super-SBM model. Figure 9 illustrates the trend in APCEE nationally and in major production regions from 2003 to 2022.

Table 4. The table of APCEE values for the national and major production regions.

Regions	2003	2004	2005	2006	2007	2008	2009	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021	2022
China	0.72	0.74	0.74	0.60	0.68	0.65	0.73	0.76	0.75	0.62	0.61	0.60	0.67	0.60	0.42	0.55	0.57	0.56	0.62	0.70
Bohai Bay	0.78	0.78	0.79	1.04	0.82	0.67	1.04	1.06	1.03	0.78	0.76	0.77	0.90	0.87	0.50	0.75	1.04	0.88	0.77	0.84
Beijing	1.42	1.00	1.20	1.33	1.11	1.16	1.38	1.31	1.31	1.40	1.35	1.20	1.27	1.47	1.55	1.26	1.53	1.53	1.43	1.36
Hebei	0.85	0.78	0.80	0.75	0.66	0.66	0.58	0.59	0.76	0.68	0.65	0.61	0.68	0.81	0.46	0.82	0.82	0.64	0.76	1.02
Liaoning	1.06	1.12	1.17	1.09	1.16	0.65	1.03	0.70	0.65	0.43	0.47	0.42	0.54	0.42	0.36	0.36	0.37	0.40	0.41	0.49
Shandong	1.22	1.19	1.01	0.60	1.10	1.03	0.77	1.06	0.81	0.57	0.52	1.07	0.67	0.67	0.41	0.67	0.73	0.68	0.77	1.09
Loess plateau	0.48	0.52	0.44	0.46	0.63	0.52	0.57	0.69	0.78	0.62	0.60	0.61	0.66	0.62	0.44	0.58	0.61	0.69	0.80	0.76
Shanxi	1.01	0.54	0.63	0.70	0.63	0.52	0.48	0.56	0.75	0.49	0.45	0.45	0.45	0.54	0.37	0.43	0.58	0.62	0.61	0.66
Henan	0.45	0.45	0.45	0.36	0.49	0.41	0.43	0.64	0.54	0.46	0.53	0.49	0.50	0.53	0.36	0.47	0.60	0.60	1.07	0.84
Shaanxi	0.86	0.66	0.70	0.47	0.63	0.51	0.74	0.68	1.05	1.15	1.18	1.20	1.17	1.12	1.13	1.09	1.06	1.17	1.04	1.11
Gansu	0.26	0.65	0.22	1.07	1.27	1.26	1.37	1.15	1.20	0.66	0.57	0.59	1.03	0.51	0.40	0.48	0.43	0.53	0.62	0.56

presents the specific values for APCEE. The APCEE in China shows an overall “W” shaped trend. From 2003 to 2008, APCEE decreased from 0.72 to 0.65. Although the inputs of land, labor, and capital per unit value of apple production decreased by 37.44%, 34.78%, and 11.03%, respectively, carbon emissions increased by 21.12 × 10⁶ t CO₂e, or 67.23%, leading to a reduction in APCEE. From 2008 to 2010, the APCEE rose from 0.65 to 0.76. During this period, the inputs of land, labor, and capital per unit value of apple production decreased by 49%, 46.67%, and 15.2%, respectively, while carbon emissions only increased by 16.89%, and the value of apple production increased by 109.68%, thus improving APCEE. From 2010 to 2017, the APCEE sharply decreased from 0.76 to 0.42, a decline of 44.74%. Despite a 25.48% decrease in carbon emissions, the inputs of land, labor, and capital per unit value of apple production increased by 63.29%, 32.14%, and 10.38%, respectively, and the value of apple production fell by 44.29%, leading to a further decline in APCEE. From 2017 to 2022, APCEE increased from 0.42 to 0.70, a growth of 66.67%. Although the inputs of land and capital per unit value of apple production increased slightly by 10.75% and 3.05%, respectively, carbon emissions decreased by 6.23 × 10⁶ t CO₂e, a reduction of 13.61%, and labor inputs decreased by 0.18%, resulting in a noticeable improvement in APCEE.

4.2.2. Analysis of APCEE in Major Production Regions

The APCEE in the two major production regions is presented in Table 4. During the study period, except for 2021, the APCEE in the Bohai Bay region consistently exceeded that of the Loess Plateau region. This could be attributed to several factors. First, in terms of technological innovation, the Bohai Bay region, due to its economic development and rich technological accumulation, likely adopted high-efficiency, low-carbon technologies earlier and more widely, thereby improving APCEE. Second, in terms of production efficiency, the Bohai Bay region benefits from a favorable geographical location, convenient transportation, and well-developed agricultural infrastructure, which are more conducive to achieving large-scale and intensive production. As concluded earlier, the RCI in the Bohai Bay region is lower than that of the Loess Plateau and national levels, thus contributing to higher APCEE. However, in recent years, due to increased attention to green agricultural development by the Chinese government at all levels, significant technological advancements have been made in the Loess Plateau region. This has led to a decrease in apple production LCI and RCI, gradually improving the APCEE of the Loess Plateau and narrowing the gap between the two regions.

Additionally, the trends of APCEE in major provinces are shown in Figure 10. APCEE is categorized into five levels, ranging from 0.4 to 1.49, with the first level defined as 1.2–1.49. As illustrated, Beijing and Shaanxi have the best performance in APCEE in the Bohai Bay and Loess Plateau regions, respectively. Throughout the study period, the average APCEE in Beijing remained above 1.2, consistently maintaining the highest in China. Shaanxi province showed the most significant improvement, rising from the fourth level to the second level and consistently maintaining APCEE greater than 1. For the remaining provinces, the annual average growth rate of APCEE from 2018 to 2022 was positive, indicating a continuing upward trend. Notably, Liaoning and Gansu provinces exhibit considerable potential for further improvement in their APCEE.

4.3. Analysis of the Determinants of APCEE in China

Table 5. Estimation results for the driving factors of APCEE in national and major production regions.

	APCEE
	China			Bohai Bay Region			Loess Plateau Region
	(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)
Culstruc	−1.229 ***	−1.398 ***	−1.314 ***	−2.489 ***	−2.361 ***	−2.304 ***	−0.469	−0.709 **	−0.536 *
	(0.292)	(0.290)	(0.296)	(0.493)	(0.391)	(0.366)	(0.341)	(0.339)	(0.319)
Indstruc	1.540 ***	1.530 ***	1.455 ***	3.928 ***	5.335 ***	5.181 ***	1.351 ***	1.373 ***	1.361 ***
	(0.231)	(0.235)	(0.252)	(0.824)	(0.922)	(0.819)	(0.179)	(0.178)	(0.169)
Education	−0.882 *	−0.964 **	−1.138 **	−0.705	−0.479	−0.459	−1.156 ***	−1.291 ***	−1.554 ***
	(0.468)	(0.460)	(0.466)	(0.780)	(0.697)	(0.655)	(0.376)	(0.384)	(0.406)
Macpower	0.161 *	0.144 *	0.110	−0.221 *	−0.259 **	−0.182 *	0.159	0.147	0.080
	(0.082)	(0.083)	(0.084)	(0.129)	(0.102)	(0.106)	(0.103)	(0.104)	(0.108)
Damage	0.175	0.215	0.239	0.187	0.140	0.093	0.206	0.336	0.488 **
	(0.159)	(0.160)	(0.155)	(0.202)	(0.202)	(0.193)	(0.209)	(0.216)	(0.209)
Electric	−0.071	−0.071	−0.054	−0.279 ***	−0.296 ***	−0.270 ***	0.116	0.064	0.002
	(0.059)	(0.060)	(0.062)	(0.074)	(0.074)	(0.080)	(0.108)	(0.121)	(0.129)
Greenpat	0.075 ***			0.098 ***			0.043
	(0.025)			(0.031)			(0.045)
Agriexp	−2.148 **			−0.393			−1.954 *
	(0.878)			(1.449)			(1.045)
L.Greenpat		0.080 ***			0.142 ***			0.066
		(0.024)			(0.033)			(0.048)
L.Agriexp		−1.548 *			−1.970			−1.054
		(0.884)			(1.417)			(1.067)
L3.Greenpat			0.110 ***			0.147 ***			0.135 ***
			(0.024)			(0.024)			(0.044)
L3.Agriexp			−2.966 ***			−2.258 **			−3.197 ***
			(0.868)			(1.085)			(1.051)
_cons	1.359 ***	1.410 ***	1.386 ***	3.443 ***	3.142 ***	2.803 ***	0.194	0.439	0.792
	(0.400)	(0.409)	(0.475)	(0.629)	(0.619)	(0.721)	(0.428)	(0.473)	(0.569)
sigma_u	0.104 ***	0.106 ***	0.120 ***	0.093 **	0.015	0.000	0.000	0.000	0.000
	(0.029)	(0.029)	(0.035)	(0.046)	(0.075)	(0.024)	(0.021)	(0.020)	(0.025)
sigma_e	0.192 ***	0.185 ***	0.169 ***	0.164 ***	0.164 ***	0.156 ***	0.188 ***	0.183 ***	0.173 ***
	(0.010)	(0.010)	(0.010)	(0.012)	(0.013)	(0.012)	(0.013)	(0.013)	(0.013)
Wald test	73.35 ***	76.9 ***	91.39 ***	71.95 ***	144.55 ***	269.49 ***	97.49 ***	101.20 ***	119.01 ***
Log-likelihood	36.82	40.93	49.82	33.66	36.61	37.21	25.45	26.71	28.67
N	200	190	170	100	95	85	100	95	85

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

presents the estimation results for the driving factors of APCEE in China. Columns (1), (4), and (7) indicate results from the random effects. In the Tobit model, columns (2), (5), and (8) indicate results with one lag, while columns (3), (6), and (9) indicate results with three lags. Columns (1)–(3) present results for the national sample, columns (4)–(6) present results for the Bohai Bay region, and columns (7)–(9) present results for the Loess Plateau region.

4.3.1. Analysis of Factors Influencing the APCEE at the National Level

Columns (1)–(3) of Table 5 present the estimation results of factors affecting the APCEE at the national level. As shown in column (1), the apple industry structure, degree of agricultural mechanization, and green technology innovation have significant positive impacts on APCEE. Apple’s cultivation structure, the educational level of the rural labor force, and agricultural subsidies have significant negative impacts on APCEE. The crop damage extent and rural electricity consumption do not affect APCEE. In addition, as shown in columns (2) and (3), green technology innovation and agricultural subsidies with one and three lags have significant effects on APCEE. The longer the lag time of green technology innovation, the larger the impact coefficient, indicating a continuously improving effect on the APCEE. However, agricultural subsidies had a continuous negative effect on APCEE.

4.3.2. Analysis of Factors Influencing the APCEE in the Major Production Regions

Columns (4)–(9) of Table 5 present the estimation results of factors influencing APCEE in major production regions. As shown in columns (4) and (7), unlike the national sample, the educational level of the rural labor force in the Bohai Bay region has no impact on the APCEE. The degree of agricultural mechanization and rural electricity consumption has negative impacts on the APCEE. Agricultural subsidies have no impact on the APCEE in the current period but have a negative impact after a lag of three periods. In the Loess Plateau region, the degree of agricultural mechanization has no impact on the APCEE. Green technological innovation has no impact on the APCEE in the current period but has a positive impact after a lag of three periods.

4.3.3. Endogeneity Test

In this paper, the instrumental variables approach and the addition of omitted variables are used to mitigate the endogeneity problem.

First, the instrumental variables approach. Green technological innovation may both be the cause of changes in APCEE and be affected by APCEE, thus generating endogeneity problems. Therefore, this paper chooses research and development investment intensity (RDI) as an instrumental variable for green technological innovation. RDI directly determines the output and diffusion of green patents and low-carbon technologies, which in turn improves the APCEE and satisfies the correlation setting. Meanwhile, RDI mainly comes from regional science and technology policies and long-term strategies of enterprises, which are not directly related to the APCEE, satisfying the exogenous setting.

Table 6. The results of the instrumental variable test.

Dependent Variables	(1)	(2)
	Greenpat	APCEE
Greenpat		0.367 ***
		(0.071)
RDI	0.286 ***
	(0.046)
Controls	Y	Y
F	38.20
Uncentered R2	0.994
N	200	200

Notes: Standard errors in parentheses, *** p < 0.01.

reports the 2SLS regression results. The estimated coefficients of the first stage in column (1) of Table 6 are significantly positive at the 1% level, i.e., the RDI positively affects green technological innovation. The Cragg-Donald Wald F-statistic for the weak instrumental variable test is 38.2, which is significantly larger than the Stock Yogo critical value of 16.38 at the 10% level of significance for the weak instrumental variable; therefore, the weak instrumental variable hypothesis is rejected. The second stage estimation results are provided in column (2) of Table 6. The sign and significance of the dependent variable are consistent with the basic regression results. This indicates that the findings remain robust after addressing the endogeneity issue.

Second, adding an omitted variable. Considering that the growth of per capita farm household income may affect APCEE, this paper includes per capita farm household income and its lagged one-period and lagged three-period variables in the model to alleviate the possible endogeneity problem of the model. The estimation results are shown in columns (1)–(3) of

Table 7. The results of adding the omitted variable and the robustness test.

	APCEE
	(1)	(2)	(3)	(4)	(5)	(6)	(7)
Culstruc	−1.205 ***	−1.373 ***	−1.303 ***	−1.204 ***	−1.633 ***	−1.204 ***	−1.204 **
	(0.296)	(0.293)	(0.303)	(0.240)	(0.390)	(0.235)	(0.603)
Indstruc	1.520 ***	1.486 ***	1.316 ***	1.774 ***	1.763 ***	1.774 ***	1.774 **
	(0.240)	(0.255)	(0.284)	(0.168)	(0.307)	(0.164)	(0.774)
Education	−0.850 *	−0.923 **	−0.990 **	−0.928 ***	1.160	−0.928 ***	−0.928
	(0.471)	(0.467)	(0.503)	(0.313)	(0.877)	(0.306)	(0.666)
Macpower	0.177 *	0.154	0.094	0.192 **	0.382 ***	0.192 ***	0.192
	(0.096)	(0.097)	(0.102)	(0.075)	(0.141)	(0.073)	(0.140)
Damage	0.170	0.215	0.249	0.339 **	0.237	0.339 **	0.339
	(0.159)	(0.160)	(0.154)	(0.170)	(0.173)	(0.166)	(0.220)
Electric	−0.065	−0.066	−0.046	−0.065	−0.086	−0.065	−0.065
	(0.061)	(0.061)	(0.064)	(0.044)	(0.078)	(0.043)	(0.119)
Income	−0.043
	(0.122)
Greenpat	0.094			0.102 ***	0.318 ***	0.102 ***	0.102 **
	(0.060)			(0.023)	(0.079)	(0.022)	(0.047)
Agriexp	−2.143 **			−3.117 ***	0.867	−3.117 ***	−3.117 **
	(0.875)			(0.739)	(1.177)	(0.722)	(1.446)
L.Income		−0.064
		(0.134)
L.Greenpat		0.109 *
		(0.064)
L.Agriexp		−1.494 *
		(0.891)
L3.Income			−0.165
			(0.159)
L3.Greenpat			0.183 **
			(0.074)
L3.Agriexp			−2.790 ***
			(0.884)
cons	1.478 **	1.656 **	2.178 **	1.081 ***	−1.380 *	1.081 ***	1.081
	(0.699)	(0.779)	(0.937)	(0.347)	(0.813)	(0.340)	(0.756)
sigma_u	0.105 ***	0.109 ***	0.138 ***				0.000
	(0.030)	(0.031)	(0.043)				(0.023)
sigma_e	0.192 ***	0.185 ***	0.167 ***				0.215 ***
	(0.010)	(0.010)	(0.010)				(0.010)
Province					Y
Year					Y
Wald test	73.76 ***	77.15 ***	90.46 ***			201.48 ***	43.56 ***
Log-likelihood/R2	36.73	40.91	50.34	0.502	0.412	23.887	23.887
N	200	190	170	200	200	200	200

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

, and after adding the omitted variables again, the sign and significance of the estimated coefficients remain consistent with the basic regression results, and the estimation results are robust.

4.3.4. Robustness Test

The robustness results are shown in columns (4)–(6) of Table 7. Column (4) uses the random effects model, Column (5) uses the fixed effects model, Column (6) presents the mixed regression model, and Column (7) shows the robust standard errors for using provinces. It can be observed that the regression results are not significantly different from the benchmark results, again validating the robustness of the estimation results in this paper.

5. Discussion

5.1. The TCE and LCI in China Reached Their Peak During the Study Period

This study employs the LCA method to calculate the carbon footprint of apple production in China. The results show that, between 2003 and 2022, the average YCI was 0.842 t CO₂e. Due to differences in carbon sources, the results in both domestic and international literature vary. However, the findings of this study are generally consistent with the existing literature [31,47]. Additionally, a meta-analysis covering 161 studies, and 289 fruit carbon footprints indicates that the average carbon footprint for fruit from “cradle-to-gate” ranges from 0.138 to 0.868 t CO₂e [48]. Therefore, the results of this study are scientifically reliable. Based on these findings, the study calculates the TCE, LCI, and RCI in China and two major production regions from 2003 to 2022. The results indicate that the TCE exhibits an inverted “U” shape, peaking in 2014 at 6.52 × 10⁷ t CO₂e. Similarly, LCI also follows an inverted “U” shape, reaching its peak in 2010 at 28.69 t/ha. The RCI exhibits fluctuating patterns over the study period. These features are observed for the first time in this study. The emergence of the peak indicates that the expansion of China’s total apple production has been decoupled from carbon emissions, laying the groundwork for the next step towards “carbon neutrality”. At the same time, some orchards have shifted from carbon sources to carbon sinks, and the inclusion of apple production in carbon emissions trading or ecological compensation pilots has provided fruit farmers with tradable carbon sink revenues, forming a market-based mechanism whereby “increased sinks are increased revenues”, thus raising the income level of fruit farmers and promoting the green and low-carbon transformation of the apple industry.

5.2. The Overall APCEE in China Is Relatively Low, Indicating Efficiency Loss

Based on the calculation of unintended outputs in apple production using the LCA method, this study employs the Super-SBM model to assess the APCEE in China and two major production regions. The results show that from 2003 to 2022, the APCEE in China exhibited a “W” shaped fluctuation, with an average APCEE of 0.645. Overall, the APCEE in China remains low, indicating efficiency losses. First, apple cultivation in China is still dominated by small-scale farmers, resulting in small production scales and low management efficiency. This makes it difficult to achieve economies of scale during production, leading to higher YCI. Furthermore, small-scale operations limit the adoption of new technologies and equipment, further affecting APCEE. Second, the application of chemical fertilizers per unit area in apple production in China remains high. Currently, the average educational level of apple growers is at the junior high school level, and there is a lack of awareness regarding green and low-carbon development. Additionally, the insufficient supply of affordable and convenient organic fertilizers contributes to the overuse of chemical fertilizers. Third, the aging of the apple farming population is increasing, resulting in higher labor costs. Labor-intensive processes in apple cultivation, such as tree pruning, branch training, flower thinning, fruit thinning, bagging, and unbagging, require substantial labor. As the population ages and the physical demands of apple farming increase, most young people are unwilling to engage in this labor, thus raising production costs. Existing literature indicates that the ACEE of Chinese remains at a low level [12,16,49].

From the perspective of the two major apple production regions, the Bohai Bay region outperforms both the Loess Plateau region and the national average in terms of APCEE, TCE, and RCI. Zhu et al. (2018) conducted a comparative study of the environmental impacts of traditional and organic apple production systems in the Loess Plateau of Shaanxi Province and the Bohai Bay of Shandong Province, reaching conclusions consistent with those of this study [47]. Specifically, they found that, due to the semi-arid conditions of the Loess Plateau, limited irrigation, and improper fertilizer use, yields are low, and environmental pollution is severe. Additionally, some studies indicate that ACEE in eastern China is higher than in western China [12,16], as well as higher than the national average [12].

5.3. The Factors Influencing APCEE in China

This study utilizes the Tobit model to examine the key factors influencing the APCEE in China. The results indicate that the structure of the apple industry, the degree of agricultural mechanization, and the green technology innovation significantly enhance APCEE. Industries with a higher proportion of output value generally hold greater economic importance in the local area and possess higher levels of specialization, leading to economies of scale and, consequently, improved ACEE [12]. Agricultural mechanization can enhance management practices, improve production efficiency, and reduce labor intensity, thereby increasing the efficiency of green agricultural production [50,51]. Green technology innovation not only improves resource utilization efficiency but also promotes the development of high-efficiency, low-toxicity agricultural inputs, thus improving ACEE [9,52]. Additionally, the level of green technology innovation demonstrates a significant lagged effect on carbon reductions [43].

The structure of apple cultivation, the educational level of rural labor, and agricultural subsidies all have a significant negative impact on the APCEE. The expansion of the crop cultivation scale requires a substantial input of resources, yet green production technologies are still not widely applied in this sector, leading to a decline in ACEE [52]. Farmers with higher education levels have easier access to loans and relevant subsidies, enabling them to expand production scales. Moreover, some highly educated farmers still adhere to the “high input, high output” mindset, which reduces the ACEE. Low-carbon agriculture is characterized by high inputs and slow results. Despite substantial government investment in a short period, no significant improvements have been achieved, and the increased agricultural input costs have consequently reduced ACEE. Furthermore, government agricultural subsidies exhibit a significant lagged effect on environmental impact [44].

The extent of crop damage and rural electricity consumption had no impact on the APCEE. Jin et al. (2024) argue that higher agricultural technology levels enhance the disaster resilience of agriculture, while a comprehensive agricultural insurance system effectively mitigates the impact of disasters, enabling a rapid recovery of productivity [17]. Therefore, no significant effect of disaster extent on ACEE was observed. According to Ma et al. (2022), increasing agricultural electricity consumption can reduce carbon emissions, but the high electricity costs render the effect of rural electricity consumption on ACEE insignificant [52].

6. Conclusions and Limitations

6.1. Conclusions

As a crucial component of Chinese agriculture, improving the APCEE holds significant implications for the green and low-carbon transformation of agriculture. This study employs LCA to measure the carbon emissions throughout the entire lifecycle of apple production in China from 2003 to 2022. Additionally, the Super-SBM model is applied to calculate the APCEE. The study analyzes APCEE across different periods and regions, followed by the use of a panel Tobit model to identify the key factors influencing APCEE. The results indicate that:

First, the YCI in China from 2003 to 2022 exhibited an “M” shaped pattern. Both LCI and TCE show an inverted “U” shaped trend, with LCI peaking in 2010 (28.69 t CO₂e/ha) and TCE reaching a peak in 2014 (6.52 × 10⁷t CO₂e). RCI fluctuates within a specific range. There are significant regional disparities in terms of YCI, TCE, LCI, and RCI across major apple production regions. Notably, Gansu province has consistently exhibited an “HHH” type pattern in carbon emissions over the past decade, highlighting particularly pronounced emissions reduction pressures.

Second, from 2003 to 2022, the overall APCEE in China exhibited a “W” shaped fluctuation, with relatively low-carbon efficiency values and the presence of efficiency losses. The APCEE in the Bohai Bay region was higher than that in the Loess Plateau region and the national average. Specifically, Beijing and Shaanxi had the highest APCEE in the Bohai Bay and Loess Plateau regions, respectively. The remaining provinces experienced a positive annual average growth rate in APCEE from 2018 to 2022, indicating a potential for further improvement. Notably, Liaoning and Gansu provinces have significant potential for enhancing APCEE.

Third, the structure of the apple industry, the degree of agricultural mechanization, and the level of green technological innovation have a significant positive impact on the APCEE. In contrast, the structure of apple cultivation, the education level of rural labor, and agricultural subsidies have a significant negative effect on APCEE. Additionally, both green technological innovation and agricultural subsidies exhibit significant lagged effects on APCEE. There are significant differences in the factors influencing the APCEE between the two major production regions.

In summary, to promote the green and low-carbon transition of apple production in China, this study proposes the following recommendations.

(1)

Given the significant regional differences in carbon emission characteristics of apple production, each region should leverage its natural resource endowments, technology, capital, and other advantages to adopt differentiated carbon reduction strategies, thereby promoting the green and low-carbon transformation of apple production. For the Loess Plateau region, it is recommended to promote the construction of high-standard orchards, manufacture mountainous, simplified, lightweight, and energy-saving agricultural machinery, and improve the mechanization rate of farmers. Increase investment in research and development of organic fertilizers and biological pesticides and give full play to the emission reduction and synergistic effects of green technology innovation. For the Bohai Bay region, it is suggested to continuously and steadily increase investment in the field of agricultural green technological innovation, particularly enhancing the green production efficiency of agricultural machinery.

(2)

Governments at all levels, along with apple industry associations, agricultural companies, and other organizations, should strengthen inter-regional communication and collaboration. Emphasis should be placed on leveraging the high APCEE in provinces such as Beijing and Shaanxi to drive and influence neighboring major production provinces. Efforts should be made to actively guide low-efficiency provinces, such as Liaoning and Gansu, to learn from the advanced experiences of high-efficiency provinces to improve the APCEE.

(3)

In the future, each production region should actively guide and train fruit farmers with low-carbon production awareness, continue to phase out backward apple orchards, and adopt intensive planting methods according to local conditions. This will enhance the degree of mechanization and reduce the input of labor and material resources. At the same time, governments at all levels should precisely allocate agricultural green subsidies for the green transformation of apple production and continuously increase investment in agricultural green technological innovation to improve APCEE.

6.2. Limitations

It is undeniable that this article has the following limitations. First, a single carbon emission factor is used for all carbon sources in a long research cycle. The changes in carbon emission factors from different sources with time and space is not considered, which need to be further investigated in future research. Second, there is no uniform standard for the calculation of the carbon footprint of agricultural products, which makes it difficult to form a comparative analysis. Furthermore, Jin et al. (2024) suggest that omitting agricultural carbon sinks leads to an underestimation of ACEE [17]. Therefore, future studies could integrate apple cultivation carbon sinks into the assessment of environmental efficiency.