Driving Factors of Spatial–Temporal Differences in Agricultural Energy Consumption Evolution in the Yellow River Basin: A Perspective of Water–Energy–Food–Land–Population Nexus

Abstract

The Yellow River Basin (YRB) is a core region for agricultural production in China; however, its agricultural energy consumption exhibits significant spatial–temporal differences, and it is confronted with the practical demand for the coordination of low-carbon transition and food security. Investigating the driving factors of agricultural energy consumption in the YRB is crucial for optimizing its agricultural energy structure, advancing low-carbon agricultural development, and offering targeted support for regional agricultural sustainability. Based on the data of YRB from 2000 to 2021, this paper employs the Logarithmic Mean Divisia Index (LMDI) method to decompose the driving factors of agricultural energy consumption in the basin by examining the interrelationships among five key factors: water, energy, food, land, and population. The results showed the following: (1) Per capita food production efficiency effect is the main factor driving the increase in agricultural energy consumption, followed by the water consumption output efficiency effect, the effective irrigation rate effect, the actual irrigation ratio effect, and the population scale effect. (2) The agricultural employment structure effect, the energy consumption output efficiency effect, the intensity of agricultural acreage effect, and the irrigation quota effect have reduced agricultural energy consumption. (3) Specifically, in Inner Mongolia, Shanxi and Henan, the largest incremental effect is the per capita food production efficiency effect. However, the primary driver in the remaining six provinces is the water consumption output efficiency effect. Regarding the reduction effect, the largest driver in Gansu, Shanxi and Shandong is the energy consumption output efficiency effect. Further, this paper analyzes the drivers of spatial differences in agricultural energy consumption in nine places. The research results can provide theoretical support and practical references for formulating targeted regional policies for the low-carbon transition of agricultural energy in the YRB.

1. Introduction

Against the global backdrop of escalating energy supply–demand conflicts that have left many countries grappling with energy crises, the pressure on energy resources in key economic and production sectors has become increasingly prominent. Notably, data from the Food and Agriculture Organization (FAO) shows that the global agri-food system alone consumes approximately 30% of the world’s total energy, underscoring the critical role of agricultural energy use in the broader global energy balance. Within this context, China, now the world’s top energy consumer, has seen its energy demand keep expanding amid economic development, with 2024 energy consumption hitting 5.96 billion tons of standard coal and a 4.3% year-on-year increase. In response to this severe energy situation, China has put forward the “peak carbon” and “carbon neutrality” goals [1], making the green and low-carbon energy transition a focus of its high-quality development [2]. This process in turn exerts far-reaching impacts on the adjustment of agricultural energy consumption patterns globally and within the country.

Agriculture is an essential part of China’s economy. For a long time, the problem of agricultural energy consumption has not attracted much attention because of the relatively low share of energy consumption (average value between 1978 and 2012 of about 6%). However, with the accelerated transition in agriculture, the use of agricultural machinery has led to a rapid increase in China’s agricultural energy consumption. In 2021, China’s agricultural energy consumption reached 96.61 million tons of standard coal, an increase of 128.23% compared with 2000, far exceeding the growth rate of grain production. Increased agricultural energy consumption will lead to a series of serious challenges to China’s agricultural development, such as resource shortages, increased carbon emissions and ecological degradation. This requires agriculture to accelerate the transformation of its development mode and take the path of efficient, safe, and green development. The agricultural energy use in the YRB is of vital importance to China’s overall energy consumption. As a core grain-producing area, the basin contributes over one-third of China’s total grain output, with large-scale agricultural activities. Agricultural water use in the basin accounts for more than 60% of its total water resources, and energy-intensive processes such as irrigation require substantial energy input. As a key national energy base, the basin’s total energy consumption accounts for over 30% of China’s total. Meanwhile, energy consumption from agricultural machinery and other agricultural sectors has continued to rise with the development of agricultural production. Consequently, the basin’s agricultural energy use is not only a crucial component of its own total energy consumption, but also directly impacts the optimization of China’s national energy consumption structure and the progress toward achieving the “dual-carbon” goals.

The Yellow River, with an approximate length of 5464 km, is the sixth longest river in the world. Originating from the Bayan Har Mountains in Qinghai Province, it finally discharges into the Bohai Sea in Dongying City [3]. It is one of China’s important agricultural and industrial bases. The region along the Yellow River is rich in agricultural resources, such as abundant production of wheat, corn and other crops. At the same time, the region is also an important source of energy resources such as coal, oil and natural gas in China. In 2021, the grain output of the YRB was 238.679 million tons, accounting for 35.33% of the total national grain output. The agricultural security of the YRB is the agricultural security of China, which shows the importance of its agricultural development. Meanwhile, significant disparities exist in agricultural energy consumption among various provinces (autonomous regions) within the YRB. In 2021, Qinghai’s energy consumption was 195,300 tons of standard coal, while the energy consumption in Shandong reached 3,592,800 tons of standard coal, which is 18.4 times higher than that of Qinghai’s. Thus, there is a significant spatial difference in energy consumption.

Environmentally, the spatial–temporal disparities in agricultural energy consumption within the basin are highly consistent with the carbon emission pattern. Major grain-producing areas such as the Huang–Huai–Hai Plain in the lower reaches have significantly higher carbon emissions than the upper reaches due to high input levels, and the mismatch among water–energy–food elements further amplifies environmental impacts. Economically, there exists an inverted U-shaped relationship between agricultural economic development and energy consumption: regions with high economic levels in the lower reaches have earlier reduced unit energy consumption through technological upgrading, while the upper reaches witness a higher growth rate of energy consumption during the scale expansion stage, and inter-provincial differences in energy consumption widen alongside economic divergence. Policy-wise, policies such as the “dual-carbon” goal since 2012 have driven down the growth rate of energy consumption, government intervention and mechanization upgrading have curbed high emissions, yet the divergent policy focuses of ecological protection in the upper reaches, water-saving and efficiency improvement in the middle reaches, and grain security in the lower reaches have exacerbated the spatial–temporal disparities in agricultural energy consumption.

This paper adds two factors, namely land and population, to the original water–energy–food system [4,5,6]. It puts energy at the center, making water, energy, food, land, and population a tightly connected system. Land serves as a carrier, and water, energy and food are both inputs and outputs, which leads to the whole system interacting and influencing each other. The relationship diagram is shown in Figure 1.

Each of these factors can form corresponding subsystems with each other. In the subsystem water–energy, energy can be used as a source of power to transport water from rivers and lakes to farmland through pipelines, and water can be extracted from water to produce hydroelectric energy, but it is generally only converted in large reservoirs and dams [7]. In the food-energy subsystem, energy can power harvesters to harvest agricultural products. The harvesting of food also leaves by-products such as oil, starch, sugar, straw, algae, trees, livestock manure, etc., which are burned, extracted, microbially decomposed, etc., and turned into fuels or fertilizers for the land [8]. In the land-energy subsystem, energy can be used as machine power to reclaim and cultivate land, and land can be used as a carrier that can provide a site for some conveniently stored energy. In the population-energy subsystem, energy provides the population with the basics of life. In addition to the energy-centered subsystem, between the water-food subsystem, water provides the elements necessary for the growth of crops, which can also contaminate water bodies if not properly treated [9]. Between the food-land subsystems, land stores food, and food fertilizes the land and provides energy. Land and water provide the necessities of life for the population, and the population pollutes the land and water.

This paper takes the agricultural energy consumption in the YRB as the research object and explores its driving factors from the perspectives of the water–energy–food–land–population nexus. Exploring the driving factors of the spatial–temporal differences in agricultural energy consumption in the YRB can help deepen the understanding of its agricultural energy transformation issues and narrow the regional energy consumption gap. Furthermore, based on the findings, this paper proposes policy suggestions to support sustainable agricultural development in the YRB and provide a decision-making basis for policymakers.

2. Literature Review

Studying the drivers of agricultural energy consumption can lead to a better understanding of trends in agricultural energy consumption [10]. A large strand of literature has targeted the drivers of agricultural energy consumption. The research methods can be categorized into econometric models, Structural Decomposition Analysis (SDA) method, and index decomposition analysis (IDA). For the first category, econometric modeling, scholars have mainly applied regression analysis to study the effects of various factors on agricultural energy consumption. Qianling, Chunyang [11] found through regression analysis that intra-industry changes in energy intensity were the main factors influencing the changes in energy intensity in China. Guo, Zhang [12] used a multivariate path analysis model to find the key drivers of coal consumption during China’s urbanization process from 1978 to 2014. Shi, Xia [13] utilized the Tobit regression model to explore the drivers of energy efficiency changes in China’s three major economic zones. Econometric models are widely applied in the research on driving factors of energy consumption, as they can identify key influencing variables through significance tests, but they rely on strict theoretical assumptions. However, agricultural energy consumption in the YRB is affected by the nonlinear coupling effects of water–energy–food–land–population factors. Such models struggle to go beyond linear frameworks to analyze the dynamic correlations among multiple factors, and their focus on correlation analysis fails to meet the research need of accurately quantifying the independent contributions of each factor from the Nexus perspective.

The second category, structural decomposition analysis (SDA), is based on input-output data. Because it is more complex and difficult to obtain input-output table data, it is applied less often. Lin and Raza [14] used LMDI and Input-Output (I-O) modeling to explore the drivers of coal consumption in Pakistan over the period 1999–2018 and found that trendiness in energy intensity, energy mix and industrial structure effects are the main drivers of coal consumption growth. Wang, Zhang [15] used the LMDI and SDA method to find that economic growth is the largest driver of the growth of coal consumption. Huang, Hong [16] synthesized the spatial autocorrelation lag model (SLX) and SDA method to investigate the driving factors of energy savings. The SDA is well-established in inter-sectoral linkage analysis based on input-output tables, with an emphasis on revealing the impacts of economic structure changes. Nevertheless, it depends on frequently updated input-output data, which conflicts with the availability of detailed agricultural data in the YRB. Additionally, its sector-aggregated analysis makes it difficult to accurately capture the micro-mechanisms of factors such as water and land, resulting in low adaptability to the refined decomposition of spatial–temporal differences required by this study.

The third category, index decomposition analysis (IDA), has the advantages of lower data requirements, simple calculations, ease of understanding, etc. Among them, the Logarithmic Mean Divisia Index method (LMDI) is one optimal approach [17]. It can solve the problems of residual and zero in the decomposition calculation, facilitating the decomposition of quantitative indicators. Thus, it is widely used in decomposing the influencing factors.

Liu, Zhou [18] introduced the IDA method into the PDA method and found that potential economic development is the biggest driver of energy consumption growth. Zhang and Bai [19] applied the LMDI method to decompose the total domestic energy consumption in Shandong Province into five drivers, namely, energy structure, energy intensity, residents’ income, demographic, and total population effect. They found that the energy intensity effect played an important role in reducing the energy consumption of urban residents. Chenjun, Yuze [20] constructed a two-stage decomposition model to study the driving factors of total electricity consumption in the Yangtze River Delta. Bu, Wang [21] quantified the drivers of natural gas consumption and found that the economic effect and fossil energy structure effects are positive drivers of it. Hu, Chi [22] applied the LMDI method to decompose the drivers of decoupling agricultural energy consumption from economic growth in China. Wu and Ding [23] explored the drivers of the change in China’s agricultural energy intensity from 1981 to 2017. Zhao, Zhang [24] employed the LMDI method to decompose the driving factors of electricity consumption in the Yangtze River Delta region.

The LMDI method demonstrates high suitability for decomposing the driving factors behind the spatial–temporal differences in agricultural energy consumption within the YRB. First, in terms of decomposition characteristics, the LMDI method enables accurate quantification of the contribution of various water–energy–food–land–population (WEFLP) nexus-related factors, such as irrigation water use efficiency, grain planting scale, land use intensity, and rural population size, to changes in agricultural energy consumption, with no residual terms. This allows for a clear revelation of the intrinsic link between the WEFLP Nexus and the evolution of agricultural energy consumption. Second, regarding spatio-temporal adaptability, the LMDI method supports both cross-sectional decomposition and time-series decomposition, which is highly aligned with the research need to explore the spatial–temporal disparities of agricultural energy consumption. Finally, in terms of synergy with the Nexus framework, the LMDI method can integrate multi-dimensional WEFLP factors into a unified decomposition framework. It effectively uncovers the impact of the synergistic or restrictive effects of interactions among these factors on agricultural energy consumption differences, thereby providing a robust analytical tool for deciphering the complex driving mechanisms underlying the evolution of agricultural energy consumption from the perspective of the WEFLP Nexus.

Nevertheless, the LMDI model has certain limitations. For instance, it exhibits strong data dependence, as the accuracy of the model is highly reliant on the quality and completeness of input data. When studying agricultural energy consumption in the YRB, it is necessary to ensure the accuracy and completeness of relevant data, including those related to water, energy, food, land, and population, to guarantee the reliability of the decomposition results generated by the LMDI model. Additionally, the LMDI model is based on the assumption of factor additivity, which decomposes the total change into the algebraic sum of the contributions of individual factors, while also defaulting to the independence of these factors. Therefore, when selecting variables, efforts should be made to satisfy these assumptions as much as possible to mitigate their potential impact on the decomposition results.

To summarize, existing studies have mainly used the IDA to decompose the drivers of agricultural energy use evolution, but there are fewer studies on the drivers of agricultural energy consumption. Meanwhile, there are fewer studies on the selection of influencing factors under a unified theoretical framework. Based on the linkage of water–energy–food–land–population, the study quantitatively analyzes the drivers of agricultural energy use from both temporal and spatial dimensions through the LMDI method. It enriches the research on the drivers of agricultural energy use evolution, and the results offer theoretical references for the scientific progress of agricultural energy conservation in the YRB.

This paper contributes to the literature in three ways. (1) For the first time, this paper decomposes the drivers of agricultural energy consumption by examining the correlation of five factors: water resources, energy, food, land, and population. (2) The drivers of agricultural energy consumption are evaluated from both temporal and spatial dimensions, thus providing a more comprehensive understanding of agricultural energy consumption. (3) Based on the results, this paper offers suggestions for the implementation of regional agricultural energy conservation policies.

3. Model and Data

3.1. Kaya Equation

The Kaya equation relates anthropogenic carbon dioxide to economic, policy, and demographic factors to establish a constant relationship [25]. The equation is usually expressed as:

$$E_{C{O}_2}={\displaystyle \frac{E_{C{O}_2}}{E}}\times {\displaystyle \frac{E}{GDP}}\times {\displaystyle \frac{GDP}{P}}\times P$$

(1)

Here, $E_{C{O}_2}$ represents greenhouse gas emissions, $E$ represents energy consumption, $GDP$ represents Gross Domestic Product, $P$ represents total population.

3.2. Temporal Difference Decomposition Model

LMDI is one of the index decomposition analysis methods proposed by Ang BW [17], which has the advantage of complete decomposition and no residue. Initially, it was often used to study the trend of industrial electricity consumption, and then gradually extended to carbon emissions, water use, energy consumption [10,26]. This paper uses the LMDI model to explore the drivers of agricultural energy consumption. Based on Equation (1), this paper establishes a decomposition model of agricultural energy consumption drivers.

$$\begin{array}{l}E={\displaystyle \frac{E}{G}}\times {\displaystyle \frac{G}{AW}}\times {\displaystyle \frac{AW}{AI}}\times {\displaystyle \frac{AI}{EI}}\times {\displaystyle \frac{EI}{CL}}\times {\displaystyle \frac{CL}{FO}}\times {\displaystyle \frac{FO}{AP}}\times {\displaystyle \frac{AP}{P}}\times P\\ =C\times W\times I\times R\times T\times L\times F\times S\times P\end{array}$$

(2)

Here, $E$ represents total agricultural energy consumption, $G$ represents agricultural value added, $AW$ represents the agricultural water use, $AI$ represents the actual irrigated area, $EI$ represents the effective irrigation area, $CL$ represents the cropland area, $FO$ represents the food production, $AP$ represents the employed population in agriculture, $P$ represents total population, $C$ represents the energy consumption output efficiency effect, $W$ represents the water consumption output efficiency effect, $I$ represents the irrigation quota effect, $R$ represents the actual irrigation ratio effect, $T$ represents the effective irrigation rate effect, $L$ represents the intensity of agricultural acreage effect, $F$ represents the per capita food production efficiency effect, $S$ represents the agricultural employment structure effect, $P$ represents the population scale effect.

Assuming that the total amount of agricultural energy consumption in the base year is $E_0$, the total amount of agricultural energy consumption in year $t$ is $E_t$ and the change in the total amount of agricultural energy consumption from the base year to year $t$ is $△E_t$. According to the LMDI model, combined with the driving force decomposition formula, the factors affecting changes in agricultural energy consumption can be decomposed in the following equation:

$$△E_t=E_t-E_0=△E_{C_t}+△E_{W_t}+△E_{I_t}+△E_{R_t}+△E_{T_t}+△E_{L_t}+△E_{F_t}+△E_{S_t}+△E_{P_t}$$

(3)

where $△E_{C_t}$, $△E_{W_t}$, $△E_{I_t}$, $△E_{R_t}$, $△E_{T_t}$, $△E_{L_t}$, $△E_{F_t}$, $△E_{S_t}$, $△E_{P_t}$ are the energy consumption output efficiency effect, the water consumption output efficiency effect, the irrigation quota effect, the actual irrigation ratio effect, the effective irrigation rate effect, the intensity of agricultural acreage effect, the per capita food production efficiency effect, the agricultural employment structure effect and the population scale effect to the changes in the total amount of energy consumption in agriculture, respectively. The LMDI decomposition equation for the effects of changes in the factors is:

$$△E_{C_t}={\displaystyle \frac{E_t-E_0}{\mathit{ln}{E}_t-\mathit{ln}{E}_0}}\times \mathit{ln}\left({\displaystyle \frac{C_t}{C_0}}\right)$$

(4)

$$△E_{W_t}={\displaystyle \frac{E_t-E_0}{\mathit{ln}{E}_t-\mathit{ln}{E}_0}}\times \mathit{ln}\left({\displaystyle \frac{W_t}{W_0}}\right)$$

(5)

$$△E_{I_t}={\displaystyle \frac{E_t-E_0}{\mathit{ln}{E}_t-\mathit{ln}{E}_0}}\times \mathit{ln}\left({\displaystyle \frac{I_t}{I_0}}\right)$$

(6)

$$△E_{R_t}={\displaystyle \frac{E_t-E_0}{\mathit{ln}{E}_t-\mathit{ln}{E}_0}}\times \mathit{ln}\left({\displaystyle \frac{R_t}{R_0}}\right)$$

(7)

$$△E_{T_t}={\displaystyle \frac{E_t-E_0}{\mathit{ln}{E}_t-\mathit{ln}{E}_0}}\times \mathit{ln}\left({\displaystyle \frac{T_t}{T_0}}\right)$$

(8)

$$△E_{L_t}={\displaystyle \frac{E_t-E_0}{\mathit{ln}{E}_t-\mathit{ln}{E}_0}}\times \mathit{ln}\left({\displaystyle \frac{L_t}{L_0}}\right)$$

(9)

$$△E_{F_t}={\displaystyle \frac{E_t-E_0}{\mathit{ln}{E}_t-\mathit{ln}{E}_0}}\times \mathit{ln}\left({\displaystyle \frac{F_t}{F_0}}\right)$$

(10)

$$△E_{S_t}={\displaystyle \frac{E_t-E_0}{\mathit{ln}{E}_t-\mathit{ln}{E}_0}}\times \left({\displaystyle \frac{S_t}{S_0}}\right)$$

(11)

$$△E_{P_t}={\displaystyle \frac{E_t-E_0}{\mathit{ln}{E}_t-\mathit{ln}{E}_0}}\times \left({\displaystyle \frac{P_t}{P_0}}\right)$$

(12)

3.3. Spatial Difference Decomposition Model of Agricultural Energy Consumption

Assuming that there are two regions, r1 and r2, the difference in agricultural energy consumption $△E_r$ can be formulated as:

$$\begin{array}{l}\mathsf{\Delta }{E}_r=E_{r2}-E_{r1}=C_{r2}\times W_{r2}\times I_{r2}\times R_{r2}\times T_{r2}\times L_{r2}\times F_{r2}\times S_{r2}\times P_{r2}\\ -C_{r1}\times W_{r1}\times I_{r1}\times R_{r1}\times T_{r1}\times L_{r1}\times F_{r1}\times S_{r1}\times P_{r1}\end{array}$$

(13)

where $C_{r1}$ and $C_{r2}$ represent the energy consumption output efficiency effect of regions r1 and r2. $W_{r1}$ and $W_{r2}$ represent the water consumption output efficiency effect. $I_{r1}$ and $I_{r2}$ represent the irrigation quota effect of regions r1 and r2. $R_{r1}$ and $R_{r2}$ represent the actual irrigation ratio effect of regions r1 and r2. $T_{r1}$ and $T_{r2}$ represent the effective irrigation rate effect of regions r1 and r2. $L_{r1}$ and $L_{r2}$ represent the intensity of agricultural acreage effect of regions r1 and r2. $F_{r1}$ and $F_{r2}$ represent the per capita food production efficiency effect of regions r1 and r2. $S_{r1}$ and $S_{r2}$ represent the agricultural employment structure effect of regions r1 and r2. $P_{r1}$ and $P_{r2}$ represent the population scale effect of regions r1 and r2.

The spatial difference in agricultural energy consumption $\mathsf{\Delta }{E}_r$ can be decomposed into the following effects due to the region changing from r1 to r2.

$$\mathsf{\Delta }{E}_r=△E_{C_r}+△E_{W_r}+△E_{I_r}+△E_{R_r}+△E_{T_r}+△E_{L_r}+△E_{F_r}+△E_{S_r}+△E_{P_r}$$

(14)

They can be broken down into:

$$△E_{C_r}={\displaystyle \frac{E_{r2}-E_{r1}}{\mathit{ln}{E}_{r2}-\mathit{ln}{E}_{r1}}}\times \mathit{ln}\left({\displaystyle \frac{C_{r2}}{C_{r1}}}\right)$$

(15)

$$△E_{W_r}={\displaystyle \frac{E_{r2}-E_{r1}}{\mathit{ln}{E}_{r2}-\mathit{ln}{E}_{r1}}}\times \mathit{ln}\left({\displaystyle \frac{W_{r2}}{W_{r1}}}\right)$$

(16)

$$△E_{I_r}={\displaystyle \frac{E_{r2}-E_{r1}}{\mathit{ln}{E}_{r2}-\mathit{ln}{E}_{r1}}}\times \mathit{ln}\left({\displaystyle \frac{{I_I}^{r2}}{{I_I}^{r1}}}\right)$$

(17)

$$△E_{R_r}={\displaystyle \frac{E_{r2}-E_{r1}}{\mathit{ln}{E}_{r2}-\mathit{ln}{E}_{r1}}}\times \mathit{ln}\left({\displaystyle \frac{{I_{RA}}^{r2}}{{I_{RA}}^{r1}}}\right)$$

(18)

$$△E_{T_r}={\displaystyle \frac{E_{r2}-E_{r1}}{\mathit{ln}{E}_{r2}-\mathit{ln}{E}_{r1}}}\times \mathit{ln}\left({\displaystyle \frac{{I_{RE}}^{r2}}{{I_{RE}}^{r1}}}\right)$$

(19)

$$△E_{L_r}={\displaystyle \frac{E_{r2}-E_{r1}}{\mathit{ln}{E}_{r2}-\mathit{ln}{E}_{r1}}}\times \mathit{ln}\left({\displaystyle \frac{{E_L}^{r2}}{{E_L}^{r1}}}\right)$$

(20)

$$△E_{F_r}={\displaystyle \frac{E_{r2}-E_{r1}}{\mathit{ln}{E}_{r2}-\mathit{ln}{E}_{r1}}}\times \mathit{ln}\left({\displaystyle \frac{{E_F}^{r2}}{{E_F}^{r1}}}\right)$$

(21)

$$△E_{S_r}={\displaystyle \frac{E_{r2}-E_{r1}}{\mathit{ln}{E}_{r2}-\mathit{ln}{E}_{r1}}}\times \mathit{ln}\left({\displaystyle \frac{{S_A}^{r2}}{{S_A}^{r1}}}\right)$$

(22)

$$△E_{P_r}={\displaystyle \frac{E_{r2}-E_{r1}}{\mathit{ln}{E}_{r2}-\mathit{ln}{E}_{r1}}}\times \mathit{ln}\left({\displaystyle \frac{P^{r2}}{P^{r1}}}\right)$$

(23)

3.4. Data Source

The research period of this study is 2000–2021. Data are mainly from the China Statistical Yearbook, China Water Resources Bulletin, China Rural Statistical Yearbook, China Energy Statistics Yearbook, and Statistical Yearbooks of nine provinces in the YRB. Relevant indicators are explained as follows:

(1)

Agricultural GDP (primary sector value added). In 2000, the agricultural GDP was adjusted according to the constant price to eliminate the effect of price factors.

(2)

Actual irrigated area. Data on the actual irrigated area have been missing for some years. So, this paper divides the irrigated water consumption of farmland by the actual average irrigation water consumption per acre to obtain the actual irrigated area.

(3)

Cropland area. The cultivated land area has not been updated since the third land census in 2018. Therefore, this paper uses exponential smoothing to fill in the missing data. We calculate the average change rate of the two periods: 2016–2017 and 2017–2018. Then, this paper takes 2018 as the basic period and multiplies the 2018 data by the average change rate to calculate the cultivated land area in 2019–2021.

(4)

Agricultural energy consumption. The total energy consumption of agriculture, forestry, livestock, and fisheries is summarized by multiplying the amount of energy of different types by the corresponding conversion coefficient to standard coal.

4. Results

4.1. Trend Analysis of the Evolution of Agricultural Energy Use

4.1.1. Overall Trend of Agricultural Energy Use in the Yellow River Basin

Figure 2 shows the trend of total agricultural energy consumption in the YRB from 2000 to 2021. It shows a fluctuating upward trend and decreases in 2019 before rising slowly. The agricultural energy consumption rose from 13,064,800 in 2000 to 24,416,000 tons of standard coal in 2021, an increase of 11,351,200 tons. During 2000–2021, the agricultural energy consumption in the YRB changed greatly, with the change rate in 2003, 2007, 2008, 2013, 2018 and 2020 being negative. Among them, 2005 and 2018 had the largest changes in agricultural energy consumption compared to the previous year, which were 38.3% and −20.6%, respectively, with the absolute value of the rate of change exceeding 20%. The former may be due to a significant increase in grain production in Shandong in 2005, an increase of 400.7 million tons compared to 2004. The main reason for the latter is that the implementation of the 2018 Inner Mongolia Autonomous Region National Economic and Social Development Plan, as well as further reforms in the supply-side of agricultural and animal husbandry industry.

4.1.2. Trend of Agricultural Energy Use in the Yellow River Basin at the Provincial Scale

Figure 3 shows the trend of agricultural energy consumption in the nine provinces. (1) There are large differences in agricultural energy consumption in different provinces, which is consistent with the research results of Sun, Liu [27]. The size of the agricultural energy consumption in the nine provinces is in the order of Shandong, Henan, Inner Mongolia, Shanxi, Sichuan, Gansu, Shaanxi, Ningxia, and Qinghai. The agricultural energy consumption is largest in Shandong, with an average total energy consumption of 4,352,800 tons of standard coal, and the smallest is Qinghai, with an average total energy consumption of 143,100 tons of standard coal, with the former 30.42 times more than the latter. The average value of agricultural energy consumption in Shandong is the largest, followed by Henan, both of which reach more than 3.5 million tons of standard coal. The types of energy consumption are mainly raw coal and diesel fuel, indicating the high demand for energy consumption in agricultural production activities in these two provinces. (2) Agricultural energy consumption in most provinces increases and then decreases, with the exception of Shandong Province, which shows multiple trends of increasing and then decreasing. Agricultural energy consumption varied considerably between the same provinces at both the 2008 and 2017 time points. From 2000 to 2021, Shandong Province had the largest total agricultural energy consumption of 775,400 tons of standard coal, and Henan had the largest elevated total agricultural consumption of 2,640,500 tons of standard coal. (3) There are differences in the average annual change in agricultural energy consumption among the provinces. The absolute value of the annual average rate of change in Sichuan, Ningxia, and Inner Mongolia all reached more than 6%, at 6.69%, 10.86%, and 6.15%, respectively, while Shanxi Province had the smallest annual average change in total agricultural energy consumption, at −0.03%.

4.2. Decomposition of the Temporal Difference

4.2.1. Overall Decomposition Results of the Yellow River Basin

Figure 4 shows the contribution values of the drivers of agricultural energy use change in the YRB from 2000 to 2021. If the effect contributes positively to agricultural energy use, then the effect leads to an increase in agricultural energy use, and if the contribution is negative, then the effect favors a decrease. The value of the increase or decrease is the absolute amount of the effect’s contribution [28]. From Figure 4, the impact of each driving effect on the total amount of agricultural energy use is dynamic; there are overall positive and negative fluctuations, and the different effects have greater inter-annual volatility.

In terms of pro-increase effects, the water consumption output efficiency effect, actual irrigation ratio effect, effective irrigation rate effect, per capita food production efficiency effect and population scale effect are positive and increase the agricultural energy use. Among them, the per capita food production efficiency effect has the greatest effect on agricultural water consumption in the YRB. It was negative during 2000–2001 and 2002–2003, and the annual average effect was the largest, reaching 1,037,961 tons of standard coal. It shows that the more food produced per unit of agricultural employees, the more it will lead to an increase in total agricultural energy consumption. Although advances in agricultural technology can help increase food production per unit, the reality is generally that the larger the area planted with food, the greater the food production and the greater the energy consumption required for harvesting. The water consumption output efficiency effect has the second highest pro-increase effect, with an average annual effect of 909,742 tons of standard coal. It shows that the increase in agricultural value added per unit of agricultural water consumption can be produced, and the corresponding required agricultural energy power will also increase. The actual irrigation ratio effect, the actual irrigation ratio effect and the population scale effect are relatively small, at 64,299, 115,772, and 50,177 tons of standard coal, respectively.

Regarding pro-decrease effects, the energy consumption output efficiency effect, the irrigation quota effect, the intensity of agricultural acreage effect and the agricultural employment structure effect are negative, thus reducing agricultural energy use. Among them, the agricultural employment structure effect is always negative, and the annual average effect is the largest, reaching −637,311 tons of standard coal. This indicates, in terms of agricultural development in the YRB, that a greater proportion of the employed agricultural population contributes to reducing agricultural energy use. This relates to the excessive lack of agricultural employment-population caused by the inflow of agricultural employment population to cities during the process of urbanization. This is followed by the energy consumption output efficiency effect and the intensity of agricultural acreage effect, with annual average effects of −479,674 tons of standard coal and −373,420 tons of standard coal, respectively. It indicates that improving energy consumption output efficiency and land use efficiency can obtain more output and benefits with limited resources. This is conducive to reducing agricultural energy use. The irrigation quota effect is the smallest in reducing agricultural energy use, with an annual average of −341,394 tons of standard coal.

From the temporal dimension, the total effect and various driving effects of agricultural energy consumption exhibit distinct dynamic evolutionary characteristics. During 2000–2004, the total effect fluctuates, and multiple effects such as the Energy consumption output efficiency effect and Water consumption output efficiency effect interact complexly, with no obvious single changing trend observed in agricultural energy consumption. From 2004 to 2005, the total effect rises sharply to a relatively high positive value, and effects like the Energy consumption output efficiency effect play a significant positive driving role, becoming the main force promoting the growth of agricultural energy consumption. From 2005 to 2017, the total effect fluctuates within a certain range, as the inhibitory effects of Energy consumption output efficiency effect, Effective irrigation rate effect and others interact alternately with the promoting effects of Population scale effect, Irrigation quota effect and others, keeping agricultural energy consumption in a relatively dynamic equilibrium state. From 2017 to 2018, the total effect plummets to a large negative value, and the inhibitory effects of Energy consumption output efficiency effect and others are greatly enhanced, leading to a significant reduction in agricultural energy consumption. From 2018 to 2021, the total effect gradually rebounds from the negative value and tends to stabilize, and various effects enter a new stage of mutual checks and balances and synergy, with agricultural energy consumption gradually developing towards a more stable state.

The annual average total effect of the drivers of agricultural energy consumption in the YRB is 346,151 tons of standard coal, which indicates that the total effect of each factor during the period of 2000–2021 contributes to the increase in total agricultural energy consumption.

4.2.2. Decomposition Results of the Yellow River Basin at the Provincial Scale

This paper analyzes the decomposition of the drivers of agricultural energy consumption in each province (autonomous region) from the time dimension. Considering the large number of drivers, to facilitate a comparison and analysis, this study only shows the decomposition results of the drivers of agricultural energy use in each province and city for 2000–2021. Figure 5 exhibits the decomposition results.

As can be seen from Figure 5, there are significant differences in the total effect and various driving effects of agricultural energy consumption among different regions. In major agricultural provinces such as Henan and Shandong [29], effects like the Water consumption output efficiency effect show a relatively strong promoting trend, while effects such as the Agricultural employment structure effect have obvious inhibitory effects. This is because major agricultural provinces have concentrated planting structures and high water use efficiency in irrigation and other aspects, while agricultural labor loss and unreasonable employment structures exist in some regions. It is suggested that major agricultural provinces continue to optimize water use technologies and promote water-saving irrigation; regions with labor loss should improve agricultural employment security to attract talents back. At the same time, policies should be adjusted in a targeted manner according to the performance of effects such as the Energy consumption output efficiency effect and Effective irrigation rate effect in different regions to narrow the gap in agricultural development between regions. In small agricultural provinces such as Qinghai and Ningxia, the overall impact of various effects is relatively small. Effects like the Energy consumption output efficiency effect mostly show weak promotion or weak inhibition, and effects such as the Actual irrigation ratio effect also have no prominent performance. The reason is that these provinces have small agricultural scales, limited application of mechanization and irrigation technologies [30], and small population scales, so the Population scale effect has a weak driving force. It is recommended to increase investment in agricultural science and technology in these provinces, promote small-sized energy-saving agricultural machinery and water-saving irrigation technologies, and at the same time, develop moderately scaled agriculture in combination with the local population situation to improve the efficiency of agricultural energy use and output levels.

Most of the total effects of the drivers of agricultural energy consumption in the nine provinces are positive. For example, in Qinghai, Sichuan, Gansu, Ningxia, Inner Mongolia, Shaanxi, and Henan, it is 106,553, 2,164,276, 440,394, 476,302, 1,753,608, 864,819, and 2,640,460 tons of standard coal, and the rate of change is 120%, 252%, 42%, 503%, 141%, 157%, and 122%, respectively. The total effect of agricultural energy consumption drivers in Shanxi and Shandong Province is negative, −401,839 and −775,394 tons of standard coal, respectively, with a rate of change of −15% and −18%. Among them, Shandong Province has the largest decrease in total agricultural energy consumption, which is worth learning from other provinces and cities.

The largest positive driving effect in Qinghai, Sichuan, Gansu, Ningxia, Shaanxi, and Shandong in absolute terms is the water consumption output efficiency effect, which is 157,963, 1,195,329, 1,676,021, 364,179, 990,671, and 4,956,332 tons of standard coal, respectively. The reason is that during the period of 2000–2021, the value added of agriculture in these provinces (autonomous regions) increased year by year, and the water consumption in agriculture decreased year by year, which in turn led to an increase in the efficiency effect of water output, resulting in an increase in the total amount of energy consumed in agriculture. The positive driving effect with the largest absolute value in Inner Mongolia, Shanxi, and Henan is the per capita food production efficiency effect, at 2,789,857, 2,468,602, and 5,232,670 tons of standard coal, respectively. This is due to the small increase in grain production and the large outflow of people working in agriculture, which caused an increase in total agricultural energy consumption.

The largest negative driving effect in Gansu, Shanxi and Shandong in absolute terms is the energy consumption output efficiency effect, which is −1,027,581, −2,383,694 and −4,070,569 tons of standard coal, respectively. This indicates that the ratio of total energy consumption and agricultural value added in these three provinces has been decreasing year by year, i.e., using limited resources to obtain more outputs and benefits, which is conducive to reducing the total agricultural energy consumption. The largest negative driving effect in Inner Mongolia and Shanxi in absolute terms is the intensity of the agricultural acreage effect, which is −1,348,720 and −590,978 tons of standard coal, respectively. The main reason is that the ratio of cropland area to food production in Inner Mongolia and Shanxi has been decreasing year by year, i.e., the improvement of land use efficiency is conducive to the reduction in agricultural energy consumption. The largest negative driving effect in absolute terms in the provinces (autonomous regions) of Qinghai, Sichuan, Ningxia, and Henan is the agricultural employment structure effect, which is −130,988, −977,072, −249,344, and −3,819,847 tons of standard coal. The reason for this may be that these four provinces have a more serious loss of population working in agriculture.

4.3. Decomposition of the Spatial Difference

From the spatial dimension, for the nine provinces and municipalities in the YRB, we select relatively large and small agricultural energy consumption as the object to be compared. The spatial analysis of driving factors constitutes a key advancement of this study compared with existing research findings [22,31]. Thus, there is a total of sixteen groups of spatial comparison objects in the region. Shandong, the province with the largest total agricultural energy consumption, and Qinghai, the province with the smallest total agricultural energy consumption, are taken as the compared objects, respectively. Due to the large amount of data in different groups in different years, we only compare the data in 2000 and 2021. Decomposition results are shown in

Table 1. Decomposition results of spatial differences in total agricultural energy consumption in the Qinghai group of the Yellow River Basin from 2000 to 2021 (10⁴ tons of standard coal).

	Sichuan- Qinghai	Gansu- Qinghai	Ningxia- Qinghai	Inner Mongolia-Qinghai	Shaanxi-Qinghai	Shanxi-Qinghai	Henan-Qinghai	Shandong-Qinghai
Energy consumption output efficiency effect	−30.18 →−33.19	34.83 →16.81	−0.67 →31.04	20.56 →61.65	−0.98 →1.87	142.60 →92.44	−9.01 →−13.75	50.04 →−44.87
Water consumption output efficiency effect	45.24 →88.53	2.15 →13.87	−10.96 →−34.74	8.07 →7.00	22.68 →49.95	77.83 →41.24	97.01 →203.89	145.88 →164.23
Irrigation quota effect	−21.94 →−38.87	−10.19 →−20.61	3.89 →1.64	−23.62 →−81.11	−19.86 →−36.28	−99.56 →−95.59	−93.10 →−194.73	−114.55 →−146.86
Actual irrigation ratio effect	0.55 →3.31	9.67 →5.66	2.52 →6.30	4.69 →−15.32	−1.82 →1.67	10.24 →4.89	10.86 →0.38	3.10 →0.02
Effective irrigation rate effect	19.92 →33.28	−16.13 →−27.83	−0.22 →7.03	1.12 →−1.32	−3.74 →3.19	−16.28 →0.28	50.53 →92.64	92.01 →82.79
Intensity of agricultural acreage effect	−64.29 →−130.63	−7.99 →−12.81	−4.20 →−16.33	−13.87 →−56.07	−15.35 →−49.63	−34.93 →−53.82	−102.36 →−216.66	−169.95 →−172.91
Per capita food production efficiency effect	32.26 →42.17	17.09 →18.79	10.23 →37.00	63.71 →179.50	18.33 →16.86	67.98 →67.71	51.40 →181.64	105.75 →113.42
Agricultural employment structure effect	1.19 →45.15	5.21 →44.35	−0.62 →−1.36	−11.89 →42.52	−2.53 →17.57	−30.60 →−0.68	13.24 →2.98	1.95 →12.59
Population scale effect	94.36 →273.15	61.57 →91.35	0.63 →6.98	66.58 →143.21	49.33 →116.70	138.34 →148.29	189.19 →404.76	313.72 →331.35

Note: The arrows (→) in the table are used to show the comparison relationship of the corresponding decomposition effect values between different regions, facilitating a clearer view of the spatial differences in each decomposition effect.

and

Table 2. Decomposition results of drivers of spatial differences in total agricultural energy consumption in the Shandong group of the Yellow River Basin from 2000 to 2021 (104 tons of standard coal).

	Qinghai- Shandong	Sichuan- Shandong	Gansu- Shandong	Ningxia- Shandong	Inner Mongolia-Shandong	Shaanxi-Shandong	Shanxi- Shandong	Henan- Shandong
Energy consumption output efficiency effect	−50.04 →44.87	−290.22 →20.83	102.29 →154.97	−58.92 →208.86	3.71 →323.85	−91.08 →96.96	494.05 →425.94	−186.54 →120.58
Water consumption output efficiency effect	−145.88 →−164.23	0.92 →−181.62	−296.44 →−284.44	−281.59 →−394.26	−284.35 →−440.07	−79.30 →−139.38	−101.21 →−262.46	51.37 →3.57
Irrigation quota effect	114.55 →146.86	85.70 →291.21	181.90 →223.57	163.68 →214.50	124.94 →153.75	47.36 →156.44	−95.94 →34.17	−122.05 →−39.13
Actual irrigation ratio effect	−3.10 →−0.02	−2.60 →10.53	51.28 →21.19	27.49 →29.55	19.65 →−49.12	−18.45 →6.29	37.01 →16.65	43.59 →1.04
Effective irrigation rate effect	−92.01 →−82.79	−54.19 →−127.84	−291.54 →−273.97	−96.11 →−83.61	−201.90 →−237.38	−181.56 →−153.71	−362.02 →−202.36	−19.02 →−27.58
Intensity of agricultural acreage effect	169.95 →172.91	−74.73 →71.58	312.49 →306.20	121.44 →166.90	305.83 →307.32	173.12 →157.97	372.18 →240.82	−8.40 →−9.52
Per capita food production efficiency effect	−105.75 →−113.42	−2.77 →−186.07	−121.96 →−161.88	17.07 →13.89	123.01 → 255.60	−43.75 →−163.16	−20.62 →−47.26	−54.15 →120.77
Agricultural employment structure effect	−1.95 →−12.59	3.75 →108.72	27.03 →140.49	−9.56 →−24.13	−72.04 →100.74	−21.75 →41.47	−145.69 →−33.23	58.37 →−36.39
Population scale effect	−313.72 →−331.35	−16.68 →−64.21	−296.79 →−336.29	−310.86 →−433.87	−331.44 →−474.38	−166.49 →−220.74	−350.09 →−307.25	16.65 →−11.94

Note: The arrows (→) in the table are used to show the comparison relationship of the corresponding decomposition effect values between different regions, facilitating a clearer view of the spatial differences in each decomposition effect.

. Among various groups and years, the difference in agricultural energy use in the group Shandong-Qinghai is always the largest. The difference in agricultural energy consumption in the group Ningxia-Qinghai is always the smallest.

If Qinghai is used as the comparison object, the population scale effect is positive in all groups, and compared with other effects, the population scale effect has the largest value, with different degrees of increase from 2000 to 2021. This suggests that the population scale effect maximizes the gap in agricultural energy consumption between other provinces and Qinghai over time. The intensity of agricultural acreage effect is negative in all groups, and the value of the intensity of agricultural acreage effect is relatively minimal and decreases to different degrees in all groups. This indicates that over time, the intensity of agricultural acreage effect minimizes the gap in total agricultural energy consumption between other provinces and Qinghai. The role of other effects on total agricultural energy consumption varies across groups. In absolute terms, the water consumption output efficiency effect in Ningxia-Qinghai changes from −10.96 to −34.74, showing a pro-reducing effect. The water consumption output efficiency effect and per capita food production efficiency effect of the other groups widen the gap in total agricultural energy consumption between the provinces and the Qinghai to a greater extent, with an overall increasing effect secondary to the total population effect; The irrigation quota effect in Ningxia-Qinghai changes from 3.89 to 1.64, showing an increasing effect. The irrigation quota effect of the other groups is negative and narrows the gap in total agricultural energy consumption between the provinces and Qinghai to a greater extent, and the reduction effect is generally inferior to the intensity of agricultural acreage effect.

If Shandong is taken as the comparison object, only the irrigation quota effect shows different degrees of increase in each group, indicating that the irrigation quota effect has widened the gap of total agricultural energy consumption between other provinces and Shandong with the passage of time; The role of other effects on total agricultural energy consumption varies in different groups. In terms of absolute value, the Henan-Shandong population scale effect changes from 16.65 to −11.94, showing a change from a pro-increase to a pro-decrease effect. The population scale effect in other groups is negative and narrows the gap in total agricultural energy consumption between provinces and Shandong to a greater extent. The Shanxi-Shandong energy consumption output efficiency effect changes from 494.05 to 425.94, with a slightly smaller increase over time. The energy consumption output efficiency effect in other groups all show different degrees of increase over time, widening the gap in total agricultural energy consumption between the provinces and Shandong to a greater extent.

When Qinghai is taken as the reference, the Population Scale Effect widens the gap in energy consumption, reflecting the direct correlation between population size and energy demand. In contrast, the Agricultural Acreage Intensity Effect narrows this gap, which reveals the restrictive role of land use scale on energy consumption. The Water Consumption Output Efficiency Effect expands the gap in some groups, confirming that water resource utilization efficiency is indirectly linked to energy consumption by influencing agricultural production [32], and it jointly acts on energy demand together with grain production efficiency.

When Shandong is used as the reference, the Irrigation Quota Effect broadens the gap, highlighting the driving role of water resource allocation (irrigation volume) in energy consumption. The growth of the Energy Consumption Output Efficiency Effect demonstrates the synergistic correlation between energy utilization efficiency, grain production, and land input. Overall, in the WEFLP system, population scale dominates the differences on the demand side, land use and water resource allocation regulate the intensity of energy consumption, and the interaction between energy and grain production efficiency further strengthens the inter-regional correlation differences. Together, these factors constitute the core mechanism underlying the regional differentiation of agricultural energy consumption.

5. Conclusions and Policy Implications

5.1. Conclusions

Taking the Yellow River Basin as the research object, this paper analyzes the drivers of agricultural energy consumption from the temporal and spatial perspectives using the LMDI method, and obtains the following conclusions:

(1)

The per capita food production efficiency effect is the main factor driving the increase in agricultural energy consumption in the YRB, followed by the water consumption output efficiency effect, the effective irrigation rate effect, the actual irrigation ratio effect, and the population scale effect. The biggest driving factor for the decrease in agricultural energy consumption in the YRB is the agricultural employment structure effect, followed by the energy consumption output efficiency effect, the intensity of agricultural acreage effect and the irrigation quota effect.

(2)

The individual drivers are heterogeneous across regions. The largest positive driver of Qinghai, Sichuan, Gansu, Ningxia, Shaanxi, and Shandong is the water consumption output efficiency effect, and that in other regions is the per capita food production efficiency effect. The largest negative driver of Gansu, Shanxi, and Shandong is the energy consumption output efficiency effect. The largest negative driver in Inner Mongolia and Shaanxi is the intensity of agricultural acreage effect, and that in other provinces (autonomous regions) is the agricultural employment structure effect.

(3)

As for Qinghai, which has the smallest agricultural energy consumption, the population scale effect maximizes the gap between other provinces and Qinghai, and the intensity of the agricultural acreage effect minimizes the gap between other provinces and Qinghai. As for Shandong, which has the largest agricultural energy consumption, the population scale effect minimizes the gap between other provinces and Shandong, and the energy consumption output efficiency effect maximizes the gap between the provinces and Shandong.

5.2. Recommendations

(1)

The driving effects of various factors on agricultural energy consumption manifest as both incremental and decremental. In practice, the decremental effects should be further enhanced while the incremental effects curbed. Given the heterogeneous contribution of different driving effects across regions, targeted energy conservation measures are required based on provincial realities: for example, upstream provinces with high animal husbandry proportions may prioritize promoting straw recycling technology and photovoltaic pastoral facilities to address high energy consumption and waste emissions in traditional livestock farming; downstream major grain-producing provinces should focus on subsidizing high-efficiency energy-saving agricultural machinery and optimizing planting structures to reduce mechanical energy consumption and agricultural input-related energy use in large-scale farming; provinces with fragile ecosystems and low agricultural productivity ought to prioritize subsidies for agricultural water-saving technologies and small-scale machinery renewal policies to balance grain production and energy conservation. Such region-specific strategies, aligned with provincial agricultural characteristics, emission intensities, and resource endowments, can improve the targeting and effectiveness of energy conservation policies.

(2)

For several provinces (autonomous regions) that are in the main grain-producing areas, such as Henan, Shandong, Sichuan, and Inner Mongolia, to safeguard grain production, traditional energy sources, such as coal and diesel, should continue to be the main source of agricultural production for the time being. However, wind, water, electricity, and bio-energy development can be accelerated and increased simultaneously. Under the conditions and cost-controllable circumstances, the degree of mechanization, intelligence and informatization should be improved to carry out large-scale agricultural planting activities to improve the efficiency of energy use.

(3)

For several provinces (autonomous regions) that are not in major food-producing areas, especially those with low total agricultural energy consumption, a small agricultural workforce, and no large-scale arable land, the route of energy export can be taken. Excess wind and water energy will be developed and delivered to other large agricultural production provinces in need, as shown in the text of the Qinghai Action Program for Building a National Clean Energy Industry Highland (2021–2030).

(4)

For other provinces such as Shaanxi and Shanxi, it is recommended that traditional energy sources such as coal and diesel are still used for agricultural production. The agricultural population is encouraged to recycle straw, utilize livestock and poultry waste resources, and promote energy-saving and environmentally friendly agricultural machinery, greenhouses, and other agricultural production tools. Local governments and village committees are carrying out activities to improve the rural environment, transforming from a traditional closed small-farm economy into a new green tourism countryside, and developing a new model that combines the small-farm economy with tourism.