Introduction

China feeds approximately 18% of the global population (United Nations, Department of Economic and Social Affairs, Population Division, 2022) with 5% of the world’s renewable water resources and 8% of its farmland (FAO, 2001). Agricultural production in China is characterized by a dominance of small farms (Duan et al. 2021) and faces various issues in the future, including resource scarcity (Qi et al. 2022; Song et al. 2023), excessive carbon emissions (Wang and Ge, 2020), and environmental degradation (Yu et al. 2023). These crises pose formidable challenges in terms of meeting the food demands of a growing affluent population (Duan et al. 2021).

From 1978 to 2020, China’s agricultural production increased more than twofold (National Bureau of Statistics, 2024). However, the increased production has come at a considerable cost in terms of resource inputs. The total capacity of agricultural machinery has increased by a factor of six over the last three decades, which coincides with a threefold increase in the use of agricultural fertilizers (National Bureau of Statistics, 2024). Although agricultural water consumption has stayed relatively constant over the last two decades, it constitutes over 60% of China’s overall water demand (National Bureau of Statistics, 2024). Additionally, low resource-use efficiency (RUE), a crucial indicator closely related to the food-energy-water (FEW) nexus, has intensified the resource crisis.

The concept of the FEW nexus arises from the interdependence of these three significant resources. Water is utilized throughout the food processing chain, which is essential for the generation, transformation, and conveyance of all forms of energy. In the same way, energy is required for the production, transportation, and distribution of food, in addition to water extraction, pumping, lifting, collection, delivery, supply, drainage, and treatment (Flammini et al. 2014). As demands for environmental protection, resource availability, and food security have increased, the concept of the FEW nexus has been extensively employed and diversely defined (Bleischwitz et al. 2018; Huntington et al. 2021; Liu et al. 2018), serving to develop conceptual tools for improving the environmental sustainability of food, energy, and water flows. These efforts include the FEW Index (Willis et al. 2016), quantitative assessments (Conway et al. 2015), and scenario development (Rabia Ferroukhi et al. 2015). Furthermore, the nexus mode of thinking is designed to highlight the assessment of critical interconnections among sector interventions (Flammini et al. 2014), especially synergies and trade-offs (Bleischwitz et al. 2018), in an integrated approach. Additionally, interlinkages in the nexus vibrate when factors such as technological advancement (continuous development and improvement of technologies over time, marked by the creation of new tools, techniques, systems or methods (Baum, 2001), population expansion, agricultural practices (Qian et al. 2018), and consumer behavior undergo changes. Coupled with its operability and problem-solving orientation (Liu et al. 2018), this nexus concept should always be considered during policy-making, management planning, and governance processes (Ferroukhi et al. 2015).

An assessment of RUE, similar to energy-use efficiency (Alluvione et al. 2011; Meul et al. 2007; Zhang et al. 2021) and energy efficiency (Adua et al. 2021; Wang et al. 2022; Zhang and Zhou, 2020; Zhen et al. 2023), based on the nexus concept can be conducted across two, three, or even more sectors at any scale, ranging from the factory to global level (Bleischwitz et al. 2018). Energy efficiency plays a dominant role in climate-change-related forecasts, models, and policies (Huntington and Smith, 2011; Saunders et al. 2021). Sharafi et al. (2023) utilized machine learning to conduct a relatively comprehensive assessment of agricultural RUE in Iran, and similar analyses are always available at the national level (Flammini et al. 2014). Many RUE analyses, however, are conducted within specific sectors. For example, Damerau et al. (2019) re-evaluated the global water-use efficiency of dietary nutrient production, while Song et al. (2018) measured province-level water resource efficiencies in China. Alluvione et al. (2011) developed efficiency scenarios for maize, wheat, and soybean, and Zhen et al. (2023) conducted a crop-specific efficiency assessment for China. Additionally, RUE may be assessed indirectly through other indicators (Adua et al. 2021; Feng et al. 2023; Zhang et al. 2023). Deficiencies still exist in direct representations of RUE and quantitative nexus analyses at more nuanced levels involving the integration of multiple resources, particularly in the context of agriculture in China.

Our research focuses on the spatial distribution of on-farm RUE, an output-input ratio, within the agriculture sector of China in the context of the FEW nexus. This assessment necessitated the use of a comprehensive dataset to uncover synergies, co-benefits, and unintended trade-offs. Drawing inspiration from the research conducted by Pellegrini and Fernández (Pellegrini and Fernández, 2018), we established a novel and relatively comprehensive on-farm output-input framework by converting crop production and the use of machinery, fuel, fertilizers, manpower, pesticide, seed, and water into the same thermal unit (GJ∙ha−1 ∙ y−1). Putting the focus on the energy flow, which is undeniably a core component in the assessment of resource use and of utmost importance in reducing consumption, costs, and emissions, we aimed to precisely quantify how resources are consumed in relation to crop production at the provincial level. Our goal was to contribute to the resolution of the following concerns: How can we accurately reflect RUE to help resolve the problem of opacity and delay in resource input information and explore the potential for resource conservation in agriculture? How can factors change and influence the resource input pattern to realize sustainable development in agricultural production?

Materials and methods

Description of the framework

We followed the approach taken by Pellegrini and Fernández (Pellegrini and Fernández, 2018) to assess on-farm energy-use efficiency, as measured by the output-input ratio. To do so, we constructed an output-input framework (Fig. 1) from the FEW nexus perspective to calculate on-farm RUE based on panel data of food output and resource inputs (including fertilizer, machinery, fuel, manpower, pesticide, seed, irrigation, and land use) for 31 provinces by the best available energy conversion factors (\({E}_{{cf}}\), energy required (GJ/mass) for the input side, and the energy content of the output side) and specialized equations. RUE was calculated as follows:

$${RUE}={E}_{{output}}/{E}_{{input}}={E}_{f}/({E}_{e}+{E}_{w})$$
(1)

where \({E}_{{output}}\) (GJ∙ha−1 ∙ y−1) is food output, equal to the energy for “food” (\({E}_{f})\); \({E}_{{input}}\) (GJ∙ha−1 ∙ y−1) is the resource input, which is equal to the sum of energy for “energy” (which includes fertilizer, machinery, fuel, manpower, pesticide, and seed) \(({E}_{e})\) and “water” (primarily irrigation) (\({E}_{w}).\)

Fig. 1
figure 1

Conceptual framework of our study from the perspective of the FEW nexus, with on-farm RUE and RS (resource saving) as key considerations. Energy flows throughout the entire ecosystem, influenced by technological advancement and societal adjustments.

Full size image

Conversion methods

On the output side, the physical amount of cereals (rice, wheat, maize, millet, barley, sorghum, and other cereals), pulses, tubers, cotton, oil plants (peanuts, rapeseed, sesame, and other oil plants), sugar cane, sugar beets, tobacco, vegetables, tea, and fruits for our study period were sourced from the National Bureau of Statistics of China (National Bureau of Statistics, 2024). Because the cotton weight data recorded is for ginned cotton and the corresponding energy conversion factor (\({E}_{{cf}-c}\)) denotes cottonseed, this study adopted an average lint percentage of 32.8% (Zhao et al. 2022) to convert cottonseed from ginned cotton. \({E}_{f}\) can be calculated as follows:

$${E}_{f}=W1\times {E}_{{cf}-c}/{L}_{u}$$
(2)

where \(W1\) is the yield (kg), \({E}_{{cf}-c}\) is the energy conversion factors for “food” (i.e., crop production), and \({L}_{u}\) is the total land use (ha).

As input, our framework incorporates “energy” in the nexus, which is represented as fertilizer, machinery, fuel, manpower, pesticide, and seed. In a manner analogous to the process mode of output, this study adopted the original data and converted them to an identical thermal unit. The energy consumption of the fertilizers was categorized as follows: N-fertilizer, P-fertilizer, and K-fertilizer, compound fertilizer quantity included. The number of machines used for each province was determined as the sum of tractors, harvesters, and threshers. The energy consumption for manpower was derived by multiplying the hours of work per hectare with the \({E}_{{cf}-{mp}}\) (Table 1) and the corresponding land use. Different crops require different working hours, and the energy consumption for work for cereals (rice, wheat, maize, millet, sorghum, and barley), pulses, oil plants (peanuts and rapeseed), cotton, sugar cane, sugar beet, tobacco, vegetables, and fruit were taken into consideration in this study. The energy equivalents of the selected crop seeds were assumed to be equal to the energy equivalent of the crop product itself, except for sugar beets (Table 2). Energy input from selected seeds of crop commodities was calculated by multiplying the seeding rate by its \({E}_{{cf}-c}\) and the seeded area. In our study, cereals (rice, wheat, maize, millet, sorghum, and barley), pulses, tubers, cotton, oil plants (peanuts, sesame, and rapeseed), sugar beet, tobacco, vegetables, and fruit were considered in the calculation (Table 2). Notably, tea and sugar cane were excluded from seed consumption analysis, primarily owing to their reliance on propagation mechanisms distinct from conventional seed-based reproduction but also because of incomplete data coverage. \({E}_{e}\) can be calculated as follows:

$$\begin{array}{l}{E}_{e}={E}_{{fertilizer}}+{E}_{{machinery}}+{E}_{{fuel}}+{E}_{{pesticide}}+{E}_{{manpower}}+{E}_{{seed}}\\ \quad\,\,=\left(\right.\!{W2}\times {E}_{{cf}-{fer}}+N\times M1\times {E}_{{cf}-{mac}}+N\times M2\times {E}_{{cf}-{fuel}}\\\quad+\,\mathop{\sum }\limits_{i=1}^{m}{h}_{i}\times {l}_{i}\times {E}_{{cf}-{mp}}/1000\\ \quad+\,\,W3\times {E}_{{cf}-p}+\mathop{\sum }\limits_{i=1}^{n}{r}_{i}\times {l}_{i}\times {E}_{{cf}-c}/1000\left.\!\right)/{L}_{u}\end{array}$$
(3)

where \({E}_{{fertilizer}},\) \({E}_{{machinery}},\) \({E}_{{fuel}}\), \({E}_{{manpower}}\), \({E}_{{pesticide}}\), and \({E}_{{seed}}\) (GJ∙ha−1 ∙ y−1) are the energy consumption for fertilizer, machinery, fuel, manpower, pesticide, and seed, respectively, \({E}_{{cf}-{fer}}\)(GJ/t), \({E}_{{cf}-{mac}}\)(GJ/t), \({E}_{{cf}-{fuel}}\)(GJ/t),\({E}_{{cf}-{mp}}\)(MJ/h), \({E}_{{cf}-p}\)(GJ/t), and \({E}_{{cf}-c}\)(GJ/t) are the energy conversion factors for fertilizer, machinery, fuel, manpower, pesticide, and seed, respectively, \(W2\) (t) is fertilizer consumption (fertilizer nutrient content), \(N\) is the number of machines, \(M1\) (t) and \(M2\,\)(t) are the unit weights of machinery and unit fuel consumption, respectively, \(W3\)(t) is the pesticide consumption, \(m\) and \(n\) refer to the crop commodities considered in the calculation of manpower and seed, \({h}_{i}\) (h/ha) is the hours of work per hectare for a specific crop; \({r}_{i}\) (kg/ha) is the seeding rate, \({l}_{i}\) (ha) is the land use for a specific crop, and \({L}_{u}\) (ha) is the total land use.

Table 1 Best available \({E}_{{cf}}\) for fertilizers, machinery, fuel, manpower, pesticide, and seed for different periods from 1987 to 2020.
Full size table
Table 2 Seeding rates \({r}_{i}\)–the quantity of seed utilized per unit of sown area (Mekasha et al. 2022; Vahabzadeh et al. 2023; Wu et al. 2017). (unit: kg/ha).
Full size table

For water, our goal was to convert water usage into thermal units. The energy for water can be largely attributed to the energy-intensive artificial water supply related to irrigation systems (Chiarelli et al. 2020). We examined the two most prevalent types of irrigation systems: sprinkler and surface types (Rosa et al. 2021). We assumed that the fundamental condition of each hectare that is efficiently irrigated is equivalent, a supposition that deviates from reality but is feasible for theoretical calculations. The energy consumption was calculated according to the respective extents of coverage within China. The water-energy conversion equation is as follows (Daccache et al. 2014; Rosa et al. 2021):

$$\begin{array}{l}{E}_{w}={E}_{{sur}}+{E}_{{spr}}\\\qquad=\left.\left(\frac{{V}_{{sur}}\times {TH}}{1.3212\times {\mu }_{{pump}}\times {\mu }_{{motor}}}+\frac{{V}_{{spr}}\times {TH}}{1.3212\times {\mu }_{{pump}}\times {\mu }_{{motor}}}\right)\right/{L}_{u}\end{array}$$
(4)

where \({E}_{{sur}}\) and \({E}_{{spr}}\) (GJ∙ha−1 ∙ y−1) are the energy consumption for surface and sprinkler irrigation systems, respectively, \({V}_{{sur}}\) and \({V}_{{spr}}\) (m3) are the volumes of surface and sprinkler irrigation water, respectively, \({TH}\) (m) signifies the total pressure head (or pumping head), \({\mu }_{{pump}}\) and \({\mu }_{{motor}}\) represent the pump and the motor efficiencies, respectively, and \({L}_{u}\) (ha) is total land use.

We assumed a theoretical efficiency of about 60% for surface irrigation. For sprinkler systems, the assumed efficiency is 80% (C. Brouwer, 1989; Daccache et al. 2014). The water volume for surface and sprinkler irrigation systems can be calculated as follows:

$${V}_{{sur}}=\frac{0.8\times {P}_{{sur}}\times {V}_{o}}{0.6\times {Pspr}+0.8\times {Psur}}$$
(5)
$${V}_{{spr}}=\frac{0.6\times {P}_{{spr}}\times {V}_{o}}{0.6\times {Pspr}+0.8\times {Psur}}$$
(6)

Where \({P}_{{sur}}\) and \({P}_{{spr}}\) are the covered area ratio of surface and sprinkler irrigation water, respectively; \({V}_{o}\) (m3) is the volume of total irrigation water.

The total pressure head needs to be calculated according to the type of irrigation. For a surface irrigation system, an average pressure head of 3 m was used for calculation (Rosa et al. 2021). For a sprinkler irrigation system, the total pressure head (\({TH}\)) required was calculated as follows (Rosa et al. 2021):

$${TH}=H+F+L$$
(7)

where H (m) is the nominal operating pressure of the sprinkler system, F (m) is the friction loss associated with the water distribution system, and L (m) is the lift height. We assumed that the lift height necessary to extract water from a surface source was inconsequential for the purposes of this study, so only the groundwater abstraction height was incorporated.

Study area

This study was conducted at the provincial level. We employed a method of sustainable agricultural zoning derived from the National Agricultural Sustainable Development Plan (2015–2030) to better illustrate trend differences, represent the substantial influence of regional heterogeneity on efficiency issues, and support the notion of localized development (Ministry of Agriculture and Rural Affairs of The People’s Republic of China, 2023). As shown in Table S1, this approach categorizes the 31 provinces into seven regions based on area ratio. The region abbreviations (detailed in the Fig. 2 and Table S1) are used in the latter part.

Fig. 2
figure 2

The seven regions of sustainable development agricultural zoning in China (Northeast Region: NE; Huang-Huai-Hai Region: HHH; the Middle and Lower Yangtze: MLY; Southern China Region: SC; Northwest and along the Great Wall Region: NW & GW; Southwest Region: SW; Qinghai-Tibet Region: QT). The 31 provinces are also shown.

Full size image

Data sources

We sourced crop production and land use data from the National Bureau of Statistics of China (National Bureau of Statistics, 2024). The total land use encompasses the combined area of sown area, orchards, and tea gardens, except for the flax sown area. The corresponding available energy conversion factors \({E}_{{cf}-c}\) for converting crop production to thermal units (GJ∙ha−1 ∙ y−1) were obtained from FAO (FAO, 2001). The energy demands associated with the industrial synthesis of ammonia declined annually due to advancements in N-fertilizer synthesis (Pellegrini and Fernández, 2018; Smith et al. 2020). This study adopted the \({E}_{{cf}-{fer}}\) results of Pellegrini & Fernández (Pellegrini and Fernández, 2018) and Smith et al. (2020) for N-fertilizer synthesis for China between 1987 and 2020 (Table 1). By taking the industrial energy efficiency increase in synthesizing ammonia into consideration, we obtained the different actual energy consumption in N-fertilizer throughout our study period and reported these results as “updated” (Table 1). Energy consumption results converted by the constant \({E}_{{cf}-{fer}}\) 69.1 GJ/t N are correspondingly reported as “outdated”. This study applied constant values of \({E}_{{cf}-{fer}}\) on P and K fertilizer energy consumption for the entire study period (Pellegrini and Fernández, 2018), as their data are limited and the influence on the energy input is minimal (Table 1). Given that China is a major consumer and producer of compound fertilizer (Ludemann et al. 2022), the NPK content in compound fertilizer cannot be ignored. This study used the widely accepted ratio 1:1:1 to determine the NPK fertilizer nutrient content in compound fertilizer (Zhang et al. 2009). Machinery statistics are from the National Bureau of Statistics of China (National Bureau of Statistics, 2024). Machinery (mass units) and fuel consumption per unit (\(M1\) and \(M2\), respectively) were estimated using the Stout standards: 6 t/unit and 3 t∙unit−1 ∙ y−1, respectively, for Asia (Pellegrini and Fernández, 2018; Stout, 1990). There are two distinct categories of tractor vehicles in terms of quantity: (1) large- or medium-sized and (2) small-sized (National Bureau of Statistics of China, (National Bureau of Statistics, 2024). It is reasonable to apply 6 t/unit to the first category (Stout, 1990). But for the small-sized tractors, the number of which is three times that of the large- or medium-sized in China, this study adopted 1.2 t/unit (Xiao et al. 2018). The energy requirements for repairs (44.5 GJ/t) and manufacturing (80.9 GJ/t) were the same as those used in the peer research (Pellegrini and Fernández, 2018). These values were divided by 10- and 30-year lifespans for conversion. For \({E}_{{cf}-{fuel}}\), an average value of 45.5 GJ/t was used (Pellegrini and Fernández, 2018) (Table 1). For \({E}_{{cf}-p}\) (\({E}_{{cf}}\) for pesticide), an average value of 120 GJ/t was picked from the literature (Wu et al. 2017). Pesticide usage data were derived from the National Bureau of Statistics of China (National Bureau of Statistics, 2024). Manpower data or working hours per hectare for a specific crop were from Cost-benefit of Agricultural Products in China (in Chinese) (National Development and Reform Commission (NDRC), 2021). Seeding rates are shown in Table 2.

Country-area data regarding surface and sprinkler irrigation \({P}_{{sur}}\) and \({P}_{{spr}}\) were from FAO (FAO, 2023). We adopted Zhou et al. (2020)’s nationwide survey-based reconstruction dataset (Zhou et al. 2020) for irrigation water volume for 1987 to 2013. In instances of missing data, we acquired the agricultural water use data from the National Bureau of Statistics of China (National Bureau of Statistics, 2024) and inferred that irrigation water volume accounts for 90% of the total agricultural water use (Zhai et al. 2023). There is an overlap in the years covered by the two data sources, specifically 2004–2013, with a relative deviation of approximately 5%. As efficiency values, 0.8 and 0.65 were assigned to the pump and motor, respectively (Daccache et al. 2014). Pumping lift data were derived from the global water table depth dataset (Fan et al. 2013) and weighted depending on the mix of abstraction sources for China, as identified by the global water information system (FAO, 2023). Sprinklers’ nominal operating pressures vary in accordance with their type and configuration. Typical operating pressures of a medium-sized sprinkler (3 bar, 30.59 m of water column) (\(H\)) were assumed. 20% of nominal operating pressure (\(H\)) was applied to offset friction losses associated with the piped systems (\(F\)) (Daccache et al. 2014).

Results

Output-input framework analysis

During the study period, as national crop production steadily increased, regional food output rose from 20–40 GJ∙ha−1 ∙ y−1 to 40–80 GJ∙ha−1 ∙ y−1 (Fig. 3a). The regions roughly split into three production groups (Table S1): NE and HHH regions had the highest average food output (close to 80 GJ∙ha−1 ∙ y−1 in 2020), MLY and NW & GW regions were in the middle (approximately 55 GJ∙ha−1 ∙ y−1), and SC, SW, and QT regions held relatively low average food output (approximately 40 GJ∙ha−1 ∙ y−1). Nationwide, food output increased by approximately 80% (Table S6).

Fig. 3: Food output and resource input details.
figure 3

a Trends of food output and resource inputs for seven regions from 1990 to 2020. (The selected data points: 1990, 1995, 2000, 2005, 2010, 2015, and 2020; data missing for Chongqing before 1997 and Hainan in 1987). b The average proportion of “energy” (including fertilizer [N, P, and K], machinery [Mac], fuel, pesticide [PC], seed, and manpower [MP]) and “water” components of energy consumption from 1987 to 2020. c Average energy consumption of the surface and sprinkler irrigation systems from 1987 to 2020.

Full size image

Excluding Beijing and Guizhou, all provinces showed output growth from 1987 to 2020 (Figs. 4 and S1). Generally, the data show a decline spanning from 1997 to 2003, which aligns with the low food prices and under-investment during that period (Wang et al. 2022). Zhejiang, Fujian, Chongqing, and Qinghai all maintained relatively stable trends throughout the study period (Fig. S1), whereas Jilin and Hainan showed relatively larger fluctuations. Notably, Xizang’s food output had no obvious increase during the final two decades of the study period after having a large increase during the first decade. The food output for Tianjin, Hebei, Inner Mongolia, Heilongjiang, and Henan displayed a consistent linear increase (ca. 200%).

Fig. 4: Annual resource inputs and food output.
figure 4

Annual resource inputs in the study area in (a) 1997, (b) 2017, and (c) 2020; annual food output in the study area in (d) 1997 and (e) 2020.

Full size image

During the study period, the resource inputs (energy and water in the FEW nexus) increased nationwide from 1987 to 2020, but there was substantial heterogeneity across regions and provinces (Table 3, Figs. 3a and 4). The resource inputs for the regions under investigation varied from 10 to 40 GJ∙ha−1 ∙ y−1 during the study period (except in the QT Region, where it exceeded 60 GJ∙ha−1 ∙ y−1 in 2020) (Fig. 3a). This trend exhibited an initial increase followed by a subsequent decline. Significantly, the resource inputs for MLY and SC regions had a relatively smaller increase when evaluating the values from 1987 to 2020. This observation might be construed as an earlier embrace of green revolution technologies in these regions (Pellegrini and Fernández, 2018). Agricultural intensification, which is closely linked to the concept of the FEW nexus, refers to the strategy of increasing inputs per unit area to improve agricultural production and maintain yield levels (FAO, 2004). As evidenced by our results, this approach has been widely adopted and embraced throughout China.

Table 3 Cropland use and resource inputs (1987–2020) at the country and regional levels.
Full size table

Inputs of resource categories other than N-fertilizer, manpower, seed, and irrigation demonstrated a nationwide increase from 1987 to 2020 (Table 3). In 2020, fuel was the largest consumption sector, followed by fertilizer (sum of NPK fertilizers), machinery, seed, manpower, pesticide, and irrigation (Table 3). The use of machinery and fuel in 2020 tripled compared to the initial value, but patterns varied by geographical region. Notably, the SW and QT regions witnessed an increase of over 1000% (Table 3). This surge was partially attributed to the implementation in 2004 of the Law of the People’s Republic of China on Promotion of Agricultural Mechanization (Ministry Of Agriculture And Rural Affairs, 2004). Fuel consumption increased markedly mainly because of the increased operation of machinery. Concurrent with the rapid increase in the machinery input was a decline in manpower consumption from 2.9 × 108 to 1.6 × 108 GJ (Table 3). Overall, N-fertilizer consumption first rose and then fell across all regions, reaching a maximum of 1.16 × 109 GJ in 2009 (Table 3; Fig. S4; Table S6). Except for HHH and NW & GW, all regions consumed less N-fertilizer energy in 2020 than in 1987. K-fertilizer had the largest relative increase (ca. 800%) from 1987 to 2020 (Table 3). Increases in P-fertilizer and K-fertilizer were nearly linear (Table S6) before 2015, partially due to the increase in capital input and its substitution for labor(Lei et al. 2023). After 2015, the decrease may be ascribed to new agricultural subsidies, increasing utilization rates, and other measures (Fan et al. 2023). In HHH and NW & GW, P-fertilizer increases exceeded 400% and K-fertilizer increases exceeded 1500%. Pesticide use experienced a similar trend as N-fertilizer and the turning points occurred around 2010. National Seed consumption remained at a relatively stable level (ca. 1.1–1.6 GJ∙ha−1 ∙ y−1) (Fig. S4 and Table S6). The average energy consumption for sprinkler irrigation was about 10 times that of surface irrigation, but it was still almost negligible in the total resource input (Fig. 3c). Irrigation covers a small part of our energy-based calculations, so water conservation should be the primary concern from our research angle. Agricultural land use, an important element related to the FEW nexus, increased slowly in recent years, but not in all regions. For example, agricultural land use in NE, NW & GW, and SW increased by about 50% from 1987 to 2020, while that of SC slightly decreased (Table 3).

Excluding Beijing, Shanghai, Jiangsu, Zhejiang, and Hainan, every province showed growth in overall inputs between 1987 and 2020 (Fig. 4, Tables S4 and S5). However, the resource input growth stopped almost entirely around 2017 (Fig. 4), signaling a light weakening of intensification. The resource input of Chongqing, Sichuan, Guizhou, Gansu, Xizang, and Qinghai experienced an increase of approximately 300% (Tables S3, S4 and Figs. S2, S3) as the overall center of grain production in China moved northward and the primary tuber-producing regions continued to be Sichuan, Yunnan, and Guizhou (Meng, 2018); notably, Xizang’s increase exceeded 1000% as the QT region is confronted with an immense agricultural intensification challenge that threatens its ecological environment (Hao et al. 2023). In contrast, Beijing, Shanghai, Jiangsu, Zhejiang, and Hainan succeeded in reducing resource inputs (Fig. 4, Tables S4 and S5), consequently positioning them as demonstration provinces for saving resources (Mi and Sun, 2021).

RUE analysis

Food output was always greater than pooled inputs with the participation of solar energy, and thus RUE was greater than 1 most of the time (Fig. 5). A higher RUE generally represents a higher level of resource utilization and sustainable development in the context of modern agriculture. The aggregate trend of China’s on-farm RUE is U-shaped (Fig. 6), showing signs of improvement in recent years. Considerable heterogeneity was observed among the different regions and provinces. Our analysis revealed that the trends in these regions and provinces may be positioned at varying stages of the U-shaped trajectory (i.e., our study period captured a specific phase of the U-shaped trend within a particular region). To enhance the clarity of our findings, we divided the general trends into four categories: U-shaped (a decline followed by an increase), downward, fluctuating, and upward.

Fig. 5: RUE results.
figure 5

RUE (the ratio of food output to resource input) for (a) seven regions and (b) 31 provinces from 1987 to 2020. Different colors denote different regions.

Full size image
Fig. 6: Food output, resource input, and RUE of China.
figure 6

a, b Total annual outdated resource input, food output and updated resource input (GJ·ha−1·y − 1) of China from 1987 to 2020. c, d “Updated” means that the actual energy consumption of N-fertilizer were included in the overall energy calculation; “outdated” means that energy consumption results of N-fertilizer were converted by the same constant 69.1 GJ/t N. Resource input and RUE are presented for two machinery lifespan scenarios: broken line for 10 years and solid line for 30 years.

Full size image

Among the seven regions studied, HHH, MLY, and SC regions exhibited the most distinct U-shaped pattern. We consider the NE and NW & GW regions as indicative of a fluctuating trend, while the SW and QT regions displayed downward trends. Fluctuating and downward trends usually commenced at comparatively elevated levels (around 3), whereas the U-shaped took a moderate altitude (around 2). The majority of the data-represented asymptotic values fell within the range of 1–4 (except the QT region). Notably, the SW and QT regions exhibited overall declining trends (Fig. 5a), indicating that they are either in the lagging part of this U-shaped trajectory or have failed to maintain a comparatively high RUE level. The persistent reduction in RUE observed in the SW and QT regions may indicate the existence of resource overuse and misuse within the agricultural FEW nexus. Alternatively stated, the time series data imply a decrease in the efficiency of nutrient utilization and a reduction in energy returns (Pellegrini and Fernández, 2018).

Only five provinces exhibited a constantly increasing trend: Hebei, Tianjin, Shanghai, Jiangsu, and Zhejiang. Ten provinces display the characteristics of a U-shaped trajectory: Henan, Shandong, Anhui, Jiangxi, Guangdong, Fujian, Hainan, Gansu, Neimeng, and Yunnan, which is in agreement with the literature that shows that Asia’s on-farm RUE in agriculture gradually decreased after the 1980s (Pellegrini and Fernández, 2018), with a turning point after 2010. Except for Hebei (circa 1990s), Yunnan (circa 2020 s), and Hainan (circa 2020 s), China experienced a majority of turning points in the 2000s. The early increase in RUE in Hebei aligns with existing literature on agricultural development utilizing diverse methods (Li et al. 2020; Wei and Jing, 2016). This trend may be attributed to increasing nitrogen utilization efficiency alongside the pressures of resource scarcity, environmental pollution, and ecological degradation (Li et al. 2020). Eight provinces showed a fluctuating trend: Liaoning, Jilin, Heilongjiang, Beijing, Ningxia, Shanxi, Xinjiang, and Xizang, generally with the RUE around 2 or 3. The majority of provinces with fluctuating trends are located in the northern part of China. Beijing exhibited a winding trend, mainly because of a decline in RUE after 2010. The rationale behind our perception of Xizang as having a fluctuating trend differs from that of the others. Xizang’s RUE trend represents early abrupt fluctuations and later constant low values (<1) (Fig. 5). Researchers found that most pastoral and populous regions in the Tibetan Plateau would struggle to achieve agricultural self-sufficiency amid the concurrent peril of extreme weather (Rui et al. 2022). Additionally, the unusual trend may be ascribed in part to the fact that this study did not account for herbage, given the intense grazing in the QT region (Zhu et al. 2023). The downward trend observed in the remaining provinces is predominantly concentrated within the Moderate Developing Zones and Protective Developing Zones (i.e., NW & NE, SW, and QT regions) entirely (Fig. 5 and Table S1). It is important that these downward provinces reach the turning point soon, preferably through nexus-mode strategies. These strategies should incorporate, but not be limited to, technological advancements, dietary and agricultural restructuring modifications (Qian et al. 2018), and reductions in farm loss (Ren et al. 2023).

Impact factor analysis

Technological advancements have contributed to improvements in on-farm RUE, leading to increased yields and resource savings. The results were obtained by comparing the energy consumption of N-fertilizer synthesis (updated and outdated) and two machinery lifespans (10 y and 30 y) (Fig. 6). China almost halved its resource input with the advancement of synthesizing ammonia (average RUE rose from 1.20 to 1.92 in 2020), and improved machinery longevity contributed to a 21% increase in RUE (updated RUE rose from 1.75 to 2.10 in 2020). Pellegrini and Fernández (2018) observed a global increase in on-farm energy-use efficiency and argued that even after discounting the improvements in N-fertilizer synthesis, today’s farms can be as energy-efficient as at the beginning of the green revolution. However, this does not hold true for China, where agriculture has historically been heavily dependent on fertilizer and on-farm machinery use. Once the advancements in N-fertilizer synthesis are deducted, China’s on-farm RUE is falling beneath the threshold of one (Fig. 6c).

China’s RUE trend followed a U-shaped trajectory; that is rebounded but at a relatively modest level (highest RUE: 2.10). Technological advancement did improve RUE greatly, but it did not guarantee resource-saving or land-sparing. Rebound effects similar to those that have impacted climate change mitigation (Zhen et al. 2023) are clear: as RUE crossed the turning point and started to rise, resource input did not rapidly decline (Fig. 6). Total land use increased from 1.50 × 108 to 1.83 × 108 ha, indicating that land conservation may not have been effectively implemented overall.

Discussion

Significance of RUE

Our results highlight the importance of resource inputs in the food system. For example, during the study period, total food output stopped growing after the resource input declined slightly (Fig. 6a, b), although under the same conditions, several provinces witnessed an increase in food output that contributed to a consistent improvement in RUE (Hebei, Henan, Shandong, Tianjin, Anhui, Jiangsu, Zhejiang, and Guangdong). The resource input changes from increasing to declining reflected achievements in resource-saving, but the ensuing food output decreases highlight the dependence of the food system on these resources. This trend can lead to several negative consequences, including increased greenhouse gas emissions, higher food prices, and reduced food security (Pelletier et al. 2011). Recognizing this relationship is critical for understanding, mitigating, and reversing the potential negative effects of reducing resource inputs.

There are several factors contributing to regional differences. Resource endowment is a major influencing factor which are fundamental to agricultural production and thus to the spatial pattern, prompting us to focus on trends rather than absolute regional data differences. China’s terrain, dominated by mountains and plateaus, slopes gradually from west to east. The vast plains on the east coast, which are conducive to farming (Wang et al. 2020), led to an early increase in RUE of regions like NE and HHH. Soil texture also matters significantly. For instance, the NE region has soil with high organic matter content (Sheng et al. 2019) and humidity (Zhuang et al. 2022). Although the SW region’s soil has a fair amount of organic matter and humidity, karst rocky desertification hampers local agricultural development due to substantial nutrient loss and high bare rock exposure (Li et al. 2021). Resource investment is also decisive, given the food system’s reliance on resource inputs (Fig. 6a, b). Pesticide input was partially influenced by the implementation of the Pesticide Industrial Policy (in Chinese) (Ministry of Agriculture and Rural Affairs of The People’s Republic of China, 2010). This may also help explain the RUE increase after 2010 for NE, HHH, MLY, and SC regions. Machinery and fuel, the largest input components, were more likely to first extensively used in the economically developed eastern areas (Fig. 3b). Eastern regions like NE, HHH, and MLY achieved better economies of scale partially through agricultural growth, as agriculture is closely linked to overall economic growth (Rohne Till, 2022). RUE can also be affected by economies of scale (E.g., Shandong, Henan, and Hebei are located in North China Plain, while Zhejiang might have a more developed land market that facilitates land transfer (Wang et al. 2010), further explaining contemporary agricultural development disparities. Additionally, agricultural investment has no immediate returns, and the lagged effect is evident (Eberly and Rebelo et al. 2012). A key policy management goal is to shorten the revenue-gaining time and balance regional differences.

Despite the fact that almost every province with positive RUE trends is endowed with natural or economic advantages, there are statistically positive and negative examples of different agricultural development patterns. Jiangsu and Shandong, which prioritize maximizing output while minimizing resource costs, are examples of positive RUE trends. Jiangsu’s agricultural economy has steadily improved, making a notable leap from 2003 to 2020 (Fig. 5), indicating better sustainability, which is highly likely to continue. Jiangsu seems to change the way of crude growth based on resource consumption, reducing the dependence on water and soil resources (Zhang et al. 2022). Zhejiang, which is in the upward phase, has managed to reduce resource consumption while maintaining stable output. It has also played an instrumental role as a leading province promoting digital agriculture, which utilizes digital and geospatial technologies to monitor, assess, and manage soil, climate, and genetic resources (The People’s Government of Zhejiang Province, 2023). This approach may help to facilitate sustainable development in agriculture (Shen et al. 2022). As demonstrated by our findings (Figs. 5b and S2), through reduced fertilizer application and machinery use, which corresponds with a decrease in fuel consumption, Zhejiang has achieved great success in agricultural resource-saving. Zhejiang presumably has successfully altered its resource input structure to establish a more sustainable pattern. However, whether digitization further improves sustainability still requires deeper exploration and must still be implemented in a responsible and ethical manner (Basso and Antle, 2020).

China’s recent energy conservation efforts have been fruitful but not enough to ensure food security, achieve the carbon-reduction target (Zhang et al. 2023), and maintain a developed RUE level. According to our food output data, China’s per capita calorie consumption in 2020, excluding other industrial uses, was 4957 kcal. This value exceeded the projected result (Ren et al. 2023), partially indicating a potential to reduce consumption by turning to a more sustainable diet (Kozicka et al. 2023). China had seen a significant rise in dairy and meat production during the study period, causing pollution and a decline in RUE (Qian et al. 2018). Two possible improvements are reducing meat and milk consumption through a shift to a more vegetable diet and cutting farm pollution with the shift to adequate large-scale cultivation (Qian et al. 2018). Moreover, transitioning to a healthier diet with increased intake of fruits, vegetables, and whole grains benefits both health and the environment. Policy-making could focus on subsidizing healthier food and imposing carbon taxation on high-resource-intensity products (He et al. 2024).

Higher RUE is associated with less pollution. Considering the correlation between RUE and pollution, it is natural to find that the trend of RUE and the environmental Kuznets curve seem to be intrinsically related (Wang et al. 2024). The quest to link the two and other similar explorations may help further create efficiencies that better reflect sustainability. Besides, the improvement of RUE is a pivotal factor in safeguarding food security and promoting the sustainability of agriculture. The overall U-shaped trends in China’s RUE serve as a call to action. High-performing regions are urged to maintain their RUE levels, while low-performing regions should be encouraged to boost their confidence in enhancing RUE. For each province, it is neither practical nor sustainable to solely prioritize one particular approach or pattern over others. In the context of national RUE rebounding, cross-sectoral and regional-specific policy-making, which operates according to local conditions and nexus thinking, as well as an exploration of the digital-agriculture development path, are urgently needed to further improve RUE.

Nexus depiction

The FEW nexus depicts a balanced and uniform circulation with the interconnectedness of food, energy, and water systems (Huntington et al. 2021). Nonetheless, our analysis appeared to merely reflect a single-track energy flow. One unit of “food” energy (\({E}_{f}\)) consumes about 0.5 units of “energy” input (\({E}_{e}\)), as well as approximately 0.005 units of “water” energy (\({E}_{w}\)) (Tables S3–5). However, these resources will not take on proportionate importance according to our computation naturally. Irrigation plays a vital role in agriculture and has been given concern over sustainability, as it accounts for a substantial portion of consumptive water usage (Huntington et al. 2021). Because we focused on the energy flow while ignoring the material flow, our limited research perspective may have introduced bias into the analysis, resulting in an imbalance of nexus elements. Nexus thinking is not an immutable concept, the value of which lies in its ability to provide a comprehensive understanding and management of complex systems and identify potential trade-offs or synergies within those systems, so as to establish more sustainable and resilient communities that are better equipped for the challenges of the future.

Role of technology in RUE

We cannot overlook the large contribution of technology in enhancing sustainability gains. However, it is important to recognize that relying solely on technology or single-process efficiency gains is not desirable (Adua et al. 2021). Technological advancements in fertilizers, machinery, pesticides, breeding, irrigation systems, digital agriculture, and other measures contribute to higher yield (Liu et al. 2022) (i.e., higher RUE). Digital agriculture is a potential source of future gains (Shen et al. 2022), and further exploration of a proper implementation plan is desirable (Basso and Antle, 2020). Manure treatment and utilization techniques with adequate large-scale cultivation serve as a potential pollution abatement method (Qian et al. 2018). RUE nearly doubled since ammonia synthesis technology was upgraded. Agricultural mechanization is a resource-intensive way to improve agricultural labor productivity, and as evidenced by our results, there was an evident decrease in manpower input over the study period (Fig. 3b and Fig. S4). However, smallholder farms dominate China’s agriculture (Duan et al. 2021), which can result in low-quality management, inferior support structures for power and supporting machinery, an insufficient supply of multi-function intelligent operation machinery, and problems with the integration of on-farm machinery and agriculture itself (Yang and Wei, 2023).

Despite additional power consumption and friction loss, water-saving irrigation has great potential for wide application in the future in terms of water-saving. Nevertheless, the area equipped for sprinkler irrigation increased only from 0.54 × 106 to 3.42 × 106 ha during our study (i.e., <1% of the area covered by surface irrigation). Improving irrigation technology, strengthening water-saving management practices, and ensuring water-saving irrigation policies are all equally important (Cao et al. 2020). Given that sprinkler irrigation has been shown to consume ten times more energy compared to surface irrigation, an overall evaluation of the water-energy balance should be taken when planning and implementing a much more high-tech irrigation system.

Importance of land

All the impact factors (3.3) we discussed offset negative effects brought about by soil degradation through land intensification or extensification (Pellegrini and Fernández, 2018). However, land extensification reinforces the degradation of land and the loss of arable land brought about by urbanization. To achieve sustainable socio-ecological systems, it is imperative to implement cross-sectoral and region-specific policies that promote sustainable intensification (Zuo et al. 2018). Closely associated with sustainable development and the FEW nexus, agricultural land is a crucial resource in and of itself. Although the amount of agricultural land expanded at a low rate, the expansion covered more than 3.3×107 ha during the study period (Table 3). This trend is expected to continue, especially for the NE, NW & GW, and SW regions. In light of the fact that the degree of intensification in agriculture of China seemed to decline (Figs. 4 and 6), we encourage further quantitative research into a sustainable descent halt point in terms of the degree of agricultural intensification. The expansion or contraction of agricultural land, changes in agricultural land use, and management policies are constantly evolving, driven by various impact factors, including technological advancement. Technological advancement does make land-sparing possible, but a feasible, effective, and powerful policy is needed to substantially manage land use. Crop rotation entails the systematic planting of various crops within the same area over successive growing seasons or years. This agricultural practice has the potential to augment crop yields (i.e., increase RUE) and facilitate sustainable agricultural intensification (Smith et al. 2023; Yang et al. 2024). As such, it represents a prudent land management strategy that is particularly suitable for adoption in the southern regions of China. Under the dual pressure of rising food demands and environmental problems caused by land expansion, attaining sustainable intensification with minimal environmental impact is challenging. Land-sparing is unlikely to take place naturally (Perfecto and Vandermeer, 2012) even if RUE and other indicators continue to head in a positive direction. The implementation of land policies should be carefully considered because some highly regarded strategies, such as land sharing (the adoption of agro-ecological practices that combine crop production with the preservation of natural habitats or the promotion of biodiversity on the same piece of land) and cropping switching, may also have harmful effects on specific regions instead (He et al. 2022; Collas et al, 2022). To fully emancipate the productive forces, implementing energy-efficient technologies (Wang et al. 2012) may be one of the top priorities, but other important actions including land consolidation (a series of purposeful activities and measures aiming to optimize land use patterns, improve land utilization efficiency, and enhance land quality) (Deng et al. 2024; Duan et al. 2021) and cropping-lifestyle modification (Adua et al. 2021) cannot be taken lightly.

Limitations and further research

Several noteworthy limitations impede the global applicability of our framework’s RUE analysis. For example, original data are missing for several provinces, particularly about the proportions of irrigation systems and underground water use at the provincial level. In these instances, national ratios had to be used in our study. Although inconsistencies between the two water data sources may have been minor, there is a need for a standardized database for irrigation to reduce data errors. This need for standardization and comprehensiveness is equally urgent when it comes to other critical elements, such as pesticide use. Additionally, the substitution of water energy by irrigation impacted the data accuracy despite having a negligible influence on the trend analysis. The focus of this study is on concerns about integrated analysis and effective management of critical flow resources. To aid in the application of the framework, we propose the utilization of adaptable scale options. To establish a robust framework, it would be beneficial to incorporate other key elements into the nexus assessment, such as materials. Additionally, complementary strategies such as scenario analysis and modeling should be incorporated (Bleischwitz et al. 2018). By approaching the subject matter with academic rigor, this type of investigation can yield a more comprehensive assessment of the agricultural output-input system, which will aid in the resolution of complicated socio-ecological issues and the promotion of sustainable development.

Conclusion

Our study, based on the output-input framework of the FEW nexus, evaluated China's on-farm RUE on the regional and provinvial levels. The findings have important policy implications for achieving sustainable socio-ecological systems, which underscores the urgent needs of implementing cross-sectoral and region-specific policies to reach higher levels of RUE and sustainable intensification.

Arising from the management and improvement of food, energy, and water resources, our framework (Fig. 1) informs the energy flow through the agricultural production process, aiming to attain resource-saving and land-sparing. Our results indicate that both resource inputs and food output experienced a surge prior to 2017. Subsequently, the level of agricultural intensification declined and food output remained relatively stable (Fig. 6). On-farm RUE exhibited an overall U-shaped pattern, but varied across regions and provinces. China’s RUE demonstrated a promising trend of rebounding, albeit at a moderate level and with substantial prospects for regional enhancement. Notably, the value of RUE partially aligns with the spatial characteristics of the zoning we use. Underlying issues in the current agricultural and food systems, such as shifting cropping structures, adjusting cultivation scale, and modifying consumer lifestyles, are also required. Exclusive reliance on technological efficiency gains is not desirable because such gains typically boost the output of a single component, and maintaining overall increases in RUE necessitates holistic control. A comprehensive approach grounded in a nexus mode of thinking that includes both technological advancement and policy intervention is required to ensure the environmental sustainability of our agricultural systems and effectively address the food security demands of an expanding population.