Trade-offs between agricultural production and ecosystem services under different land management scenarios in the Loess Plateau of China

Abstract

Land management practices play a crucial role in balancing agricultural production and ecosystem services. This study evaluated the trade-offs between provisioning ecosystem services (crop yields) and other key ecosystem services including regulating services (water yield, soil conservation, carbon sequestration) and supporting services (biodiversity) under three land management scenarios in the Loess Plateau of China. An integrated assessment framework combining biophysical models, economic valuation, and trade-off analysis was applied. The results showed significant trade-offs between provisioning ecosystem services (agricultural production) and regulating and supporting ecosystem services (water yield, soil conservation, carbon sequestration, and biodiversity). The ecological restoration scenario maximized regulating and supporting services but reduced agricultural output by 15%, while the sustainable intensification scenario increased agricultural production by 15% with moderate ecosystem service provision. The business-as-usual scenario showed intermediate performance for both. Trade-offs were driven by land use intensity, landscape configuration, biogeochemical cycles, and hydrological processes. We propose strategies including sustainable intensification practices, landscape multifunctionality enhancement, ecosystem-based adaptation, participatory land-use planning, and improved monitoring systems. These findings highlight the need for integrated approaches that balance food production and environmental sustainability, providing insights for land management policies aligned with UN Sustainable Development Goals.

Introduction

Agriculture is the foundation of human civilization, providing food, fiber, and other essential resources for human survival and development. However, with the increasing demand for agricultural products driven by population growth and changing consumption patterns, the pressure on agricultural land has intensified globally^1–3, leading to widespread ecosystem degradation and biodiversity loss^4,5. This trend is particularly evident in developing countries where agricultural expansion often comes at the expense of natural habitats⁶. Sustainable agricultural land management practices have been proposed as a solution to balance agricultural production and ecosystem services, but the trade-offs between them remain poorly understood⁷.

Ecosystem services are the benefits that humans derive from ecosystems, including provisioning services (e.g., food and water), regulating services (e.g., climate and disease regulation), cultural services (e.g., spiritual and recreational benefits), and supporting services (e.g., nutrient cycling and soil formation)⁸. Agricultural production relies on ecosystem services, such as soil fertility, pollination, and pest control, but it can also have negative impacts on these services through land-use change, soil erosion, and the use of agrochemicals⁹.

To achieve sustainable agricultural land management, it is essential to understand the complex interactions between agricultural production and ecosystem services and to identify the trade-offs and synergies between them. This requires a comprehensive assessment of the impacts of different land management scenarios on both agricultural production and ecosystem services, taking into account the spatial and temporal scales, as well as the socio-economic and environmental contexts.

The purpose of this study is to evaluate the trade-offs between agricultural production and ecosystem services under different land management scenarios. Specifically, we aim to answer the following research questions:

1. What are the impacts of different land management scenarios on agricultural production and ecosystem services?

2. How do the trade-offs between agricultural production and ecosystem services vary at different administrative levels (county and township) and over the simulation period (2020–2040)?

3. What are the key drivers and mechanisms underlying the trade-offs between agricultural production and ecosystem services?

Understanding the trade-offs between agricultural production and ecosystem services is crucial for informing sustainable land management decisions and policies. By identifying the key drivers and mechanisms underlying these trade-offs, we can develop targeted interventions to minimize the negative impacts of agriculture on ecosystems while maximizing the benefits for human well-being. This study contributes to the growing body of literature on sustainable land management and provides valuable insights for policymakers, land managers, and researchers working towards a more sustainable future.

Materials and methods

Study area

The study area is located in the Loess Plateau of China, which covers an area of approximately 640,000 km² and is situated in the upper and middle reaches of the Yellow River¹⁰. The Loess Plateau is characterized by a semi-arid continental monsoon climate, with an average annual temperature of 4.3 °C to 14.3 °C and an average annual precipitation of 150 mm to 750 mm, most of which falls between June and September¹¹. The region is known for its highly erodible loess soils, which are prone to soil erosion and land degradation.

The land use in the Loess Plateau is dominated by agricultural land, including cropland and grassland, which accounts for approximately 60% of the total area. The main crops grown in the region are wheat, maize, and soybean, while the grasslands are used for livestock grazing and hay production. In recent years, the Chinese government has implemented a series of ecological restoration projects in the Loess Plateau, such as the Grain for Green Program, which aims to convert steep slopes and degraded cropland into forests and grasslands¹².

Despite these efforts, the Loess Plateau still faces significant challenges in balancing agricultural production and ecosystem services. The intensive agricultural practices, such as overgrazing and excessive fertilizer use, have led to soil erosion, water pollution, and biodiversity loss¹³. Moreover, the region is vulnerable to climate change, with increasing temperatures and changing precipitation patterns affecting agricultural production and ecosystem services.

The topography of the study area is characterized by rolling hills with elevations ranging from 800 to 2000 m above sea level. Slopes vary from 0° to 25°, with approximately 40% of the area having slopes greater than 15°. The entire region is covered by loess deposits, though soil depth varies from 20 to 200 m. Current agricultural intensity is moderate to high, with average nitrogen fertilizer application rates of 180 kg/ha/year and irrigation water use of 450 mm/year¹⁴.

Given its unique geographical, climatic, and land-use characteristics, the Loess Plateau provides an ideal setting for studying the trade-offs between agricultural production and ecosystem services under different land management scenarios. Understanding these trade-offs is crucial for developing sustainable land management strategies that can support both agricultural production and ecosystem health in the region.

Data sources and processing

To evaluate the trade-offs between agricultural production and ecosystem services under different land management scenarios, we used a combination of remote sensing data, field observations, and socio-economic data. The remote sensing data included Landsat 8 Operational Land Imager (OLI) images, which were acquired from the United States Geological Survey (USGS) Earth Explorer platform¹⁵. The images had a spatial resolution of 30 m and covered the entire study area for the years 2015, 2017, and 2019. The images were preprocessed using the ENVI software (Environment for Visualizing Images, version 5.6, L3Harris Geospatial Solutions, https://www.l3harrisgeospatial.com/Software-Technology/ENVI), which involved radiometric calibration, atmospheric correction, and image mosaicking.

The field observations were conducted in the study area during the growing seasons of 2015, 2017, and 2019. The observations included measurements of crop yields, biomass, and soil properties, such as soil moisture, soil organic carbon, and soil nutrients¹⁶. The field data were used to validate the remote sensing data and to calibrate the ecosystem service models.

The socio-economic data were obtained from the statistical yearbooks of the counties in the study area¹⁷. The data included information on population, GDP, agricultural inputs, and land-use policies. The data were used to analyze the drivers of land-use change and to assess the impacts of different land management scenarios on agricultural production and ecosystem services.

To process the data, we first conducted a land-use classification using the Landsat 8 OLI images. The classification was performed using the random forest algorithm¹⁸, a machine learning approach widely applied in land-use classification due to its robustness and accuracy^19,20. The algorithm was implemented using the RandomForest package (version 4.7-1.2, https://cran.r-project.org/package=randomForest) in R (version 4.2.2, https://www.r-project.org/) with 500 trees and default parameters for other settings. The algorithm was trained using a set of reference data, which were obtained from the field observations and high-resolution Google Earth images. The classification resulted in a land-use map of the study area, which included five main land-use types: cropland, grassland, forest, water bodies, and built-up areas.

Next, we calculated a set of ecosystem service indicators using the land-use map and the field observations. The indicators included net primary productivity (NPP), soil conservation, water yield, and habitat quality²¹. The NPP was estimated using the Carnegie-Ames-Stanford Approach (CASA) model, which is a light use efficiency model that uses remote sensing data to estimate vegetation productivity. The soil conservation was estimated using the Revised Universal Soil Loss Equation (RUSLE), which is an empirical model that predicts soil erosion based on rainfall, soil properties, and land-use factors. The water yield was estimated using the InVEST (Integrated Valuation of Ecosystem Services and Tradeoffs) model, which is a spatially explicit model that simulates the hydrological cycle and estimates the amount of water that is available for human use. The habitat quality was estimated using the InVEST model (Integrated Valuation of Ecosystem Services and Tradeoffs, version 3.15.1, The Natural Capital Project, https://naturalcapitalproject.stanford.edu/software/invest), which assesses the suitability of different land-use types for biodiversity conservation.

Finally, we integrated the agricultural production and ecosystem service indicators using a multi-criteria decision analysis (MCDA) approach²². The MCDA approach allowed us to evaluate the trade-offs between different land management scenarios based on their impacts on agricultural production and ecosystem services. The scenarios included business-as-usual, ecological restoration, and sustainable intensification. The MCDA was performed using the Analytic Hierarchy Process (AHP), which is a structured technique for organizing and analyzing complex decisions based on pairwise comparisons of the criteria.

Construction of the assessment indicator system

To comprehensively evaluate the trade-offs between agricultural production and ecosystem services under different land management scenarios, we constructed an assessment indicator system based on the research objectives and the data availability. The indicator system included two main components: agricultural production indicators and ecosystem service indicators.

Following the Millennium Ecosystem Assessment framework⁸, we classified crop yield as a provisioning ecosystem service, while water yield, soil conservation, and carbon sequestration were classified as regulating services, and biodiversity support as a supporting service. Net primary productivity (NPP) was included as an ecological process that underpins multiple ecosystem services rather than as a service itself.

Figure 1 presents the hierarchical structure of the Assessment Indicator System Framework used in this study. This hierarchical framework consists of three levels for evaluating land management impacts:

Fig. 1.

Assessment indicator system framework assessment indicator system framework.

Level 1: Overall goal—Sustainable land management

Level 2: Criteria categories:

• Provisioning services

• Regulating services

• Supporting services

Level 3: Specific indicators:

• Provisioning services: Crop yield, economic benefit

• Regulating services: Water yield, soil conservation, carbon sequestration

• Supporting services: Biodiversity, habitat quality

The agricultural production indicators were selected to reflect the quantity and quality of agricultural outputs, as well as the efficiency of agricultural inputs²³. The indicators included crop yield (kg/ha), crop quality (% of grade A products), and input–output ratio (kg/yuan). The crop yield was calculated as the total weight of harvested crops per unit area, while the crop quality was assessed based on the national standards for agricultural products. The input–output ratio was calculated as the ratio of crop yield to the total cost of agricultural inputs, such as fertilizers, pesticides, and labor.

The ecosystem service indicators were selected to reflect the multiple benefits that ecosystems provide to human well-being, including provisioning, regulating, and cultural services²⁴. The indicators included net primary productivity (NPP, gC/m^2/yr), soil conservation (t/ha/yr), water yield (m^3/ha/yr), and habitat quality (dimensionless). The NPP was calculated using the CASA model, which estimates the amount of carbon that is fixed by vegetation through photosynthesis²⁵. The soil conservation was calculated using the RUSLE model, which estimates the amount of soil that is retained by vegetation and prevented from erosion²⁶. The water yield was calculated using the InVEST model, which estimates the amount of water that is available for human use after accounting for evapotranspiration and surface runoff²⁷. The habitat quality was calculated using the InVEST model, which estimates the suitability of different land-use types for biodiversity conservation based on their proximity to human disturbances and their connectivity to other natural habitats²⁸.

To integrate the agricultural production and ecosystem service indicators, we used a weighted sum approach, which assigns different weights to each indicator based on its relative importance²⁹. The weights were determined using the analytic hierarchy process (AHP), which is a structured technique for organizing and analyzing complex decisions based on pairwise comparisons of the criteria³⁰. The AHP was conducted through a questionnaire survey of experts in the fields of agriculture, ecology, and land management. The experts were asked to compare the importance of each indicator pair-wise using a nine-point scale, where 1 represents equal importance and 9 represents extreme importance. The pairwise comparisons were then used to calculate the weights of each indicator using the following formula:

where is the weight of indicator , is the pairwise comparison value of indicator to indicator , and is the total number of indicators.

The weighted sum of the agricultural production and ecosystem service indicators was then calculated using the following formula:

where is the overall score of the land management scenario, is the weight of indicator , is the normalized value of indicator , and is the total number of indicators. The normalized value of each indicator was calculated by dividing its actual value by its maximum value across all land management scenarios, so that all indicators have a value between 0 and 1.

The assessment indicator system provides a comprehensive and quantitative framework for evaluating the trade-offs between agricultural production and ecosystem services under different land management scenarios. By using a combination of remote sensing data, field observations, and expert knowledge, the indicator system can capture the complex interactions between human activities and natural processes in agricultural landscapes. The results of the assessment can inform decision-making on sustainable land management practices that balance the needs of food production and environmental conservation.

Scenario setting

Scenario implementation in models

The implementation of different management practices in each scenario was parameterized in our models as follows:

Business-as-usual (BAU):—Current fertilizer application rates maintained (N: 180 kg/ha, P: 90 kg/ha)—Conventional tillage practices (plowing depth: 20–25 cm)—Current irrigation schedules (data source: local agricultural bureau records, 2015–2019)—Existing crop rotation patterns (data source: field surveys and agricultural census, 2018–2019).

Ecological restoration (ER):—Conversion of cropland on slopes > 15° to forest/grassland (identified using 30 m DEM)—Reduced fertilizer application (N: 120 kg/ha, P: 60 kg/ha)—Implementation of buffer zones (30 m width) along water bodies—Increased landscape connectivity through ecological corridors (minimum width: 50 m).

Sustainable intensification (SI):—Precision agriculture implementation: * Variable rate fertilizer application (N: 120–200 kg/ha based on soil testing) * Smart irrigation systems with soil moisture monitoring (triggering at 60% field capacity) * GPS-guided machinery operations (reducing overlap by 5–8%)—Conservation tillage: * Minimum tillage practices (plowing depth: 10–15 cm) * Crop residue retention (30% coverage, modeled as increased soil organic matter)—Integrated pest management: * Biological control agents (modeled as 15–30% reduction in pest damage coefficients) * Crop rotation for pest suppression (modeled using modified susceptibility factors) * Regular pest monitoring (modeled as 20% reduction in pesticide application).

Model implementation methodology: The implementation of these practices in our biophysical models involved adjusting specific parameters that control agricultural processes and ecosystem functions. For precision agriculture, we modified the spatial and temporal distribution of inputs rather than changing total amounts. For conservation tillage, we adjusted soil disturbance factors, erosion parameters, and soil organic matter dynamics. Integrated pest management was modeled through modifications to crop damage functions and reduced chemical inputs.

Scenario analysis and comparison

To analyze and compare the trade-offs between agricultural production and ecosystem services under different land management practices, we designed three scenarios based on the current land-use patterns and the potential land-use changes in the study area. The scenarios included:

1. Business-as-usual (BAU) scenario: This scenario assumes that the current land-use patterns and management practices will continue into the future without any significant changes. The BAU scenario serves as a baseline for comparing the impacts of different land management practices on agricultural production and ecosystem services³¹.

2. Ecological restoration (ER) scenario: This scenario assumes that the government will implement a series of ecological restoration projects in the study area, such as the Grain for Green Program, which aims to convert steep slopes and degraded cropland into forests and grasslands. The ER scenario represents a land-use change that prioritizes ecosystem services over agricultural production³².

3. Sustainable intensification (SI) scenario: This scenario assumes that the government will promote sustainable intensification practices in the study area, such as precision agriculture, conservation tillage, and integrated pest management. The SI scenario represents a land-use change that aims to increase agricultural production while minimizing the negative impacts on ecosystem services³³.

For each scenario, we estimated the changes in land-use patterns and management practices based on the government’s land-use planning documents and the expert opinions from the local agricultural and environmental authorities. We then simulated the impacts of each scenario on agricultural production and ecosystem services using the assessment indicator system described in “Construction of the assessment indicator system” section.

To ensure the reliability and validity of the scenario analysis, we conducted a sensitivity analysis to test the robustness of the results to the changes in the input parameters and assumptions. We also validated the results using the field observations and the remote sensing data from the study area.

The scenario analysis provides a useful tool for exploring the potential trade-offs and synergies between agricultural production and ecosystem services under different land management practices. By comparing the impacts of different scenarios on the assessment indicators, we can identify the land management practices that can maximize the benefits for both agricultural production and ecosystem services, and minimize the negative impacts on the environment and human well-being. The results of the scenario analysis can inform the decision-making on sustainable land management policies and practices in the study area and beyond.

Model validation and sensitivity analysis

Model validation was conducted using independent field data collected from 30 sampling sites during 2018–2019. The validation metrics included:

• Crop yield: R² = 0.85, RMSE = 0.42 t/ha

• Soil erosion: R² = 0.78, RMSE = 1.2 t/ha/year

• Water yield: R² = 0.82, RMSE = 45 mm/year

Sensitivity analysis was performed using a Monte Carlo approach with 1000 iterations. Key parameters were varied within ± 20% of their baseline values. The most sensitive parameters were:

• Crop coefficients (affecting water yield)

• C-factor in RUSLE (affecting soil erosion)

• Carbon pool values (affecting carbon sequestration)

Quantitative assessment methods

To quantitatively assess the agricultural production and ecosystem services under different land management scenarios, we used a combination of process-based models and empirical models based on the remote sensing data and field observations.

Carbon sequestration assessment

Carbon sequestration was estimated using the InVEST carbon storage and sequestration model. The model calculates the amount of carbon stored in four carbon pools: aboveground biomass, belowground biomass, soil organic carbon, and dead organic matter. The carbon storage for each land use type was determined based on literature values and field measurements³⁴.

Biodiversity assessment

Species richness was assessed using field surveys and the Shannon–Wiener diversity index. Sampling plots (20 m × 20 m) were established in different land use types, and plant species were identified and recorded. The Shannon–Wiener index was calculated as:

where pi is the proportion of individuals belonging to species i.

Landscape aesthetics assessment

Scenic beauty was evaluated using a combination of landscape metrics and expert scoring. Key metrics included landscape diversity, naturalness, and visual quality. A panel of 15 experts in landscape ecology and planning rated the aesthetic value of different landscape types using a 5-point Likert scale.

Economic benefit calculation

The economic benefit was calculated as:

where EB is the economic benefit, Yi is the yield of crop i, Pi is the market price of crop i, and Ci is the production cost including labor, fertilizer, and other inputs.

The agricultural output value was calculated considering both price trends and market demand:

where Yi = yield of crop i, Pi = base price of crop i, Di = demand adjustment factor, r = annual price change rate (estimated at 2%), t = years from baseline.

For agricultural production, we used the Agricultural Production Systems Simulator (APSIM) model (version 7.10, https://www.apsim.info/) to estimate the crop yields and nutrient balances under different land management practices³⁵. The APSIM model is a process-based model that simulates the crop growth and soil processes based on the daily weather data, soil properties, and management practices. The model has been widely used and validated in various agricultural systems around the world³⁶. In this study, we parameterized the APSIM model using the local soil and weather data, and the crop management data from the field surveys and the government’s agricultural census. We then ran the model for each land management scenario and estimated the crop yields and nutrient balances for the major crops in the study area, including wheat, maize, and soybean.

For ecosystem services, we used a combination of process-based models and empirical models to estimate the key indicators, including net primary productivity (NPP), soil conservation, water yield, and habitat quality. For net primary productivity (NPP), which is an ecological process that underpins multiple ecosystem services rather than a service itself, we used the Carnegie-Ames-Stanford Approach (CASA) model (version 3.0, Potter et al.25). This process-based model estimates the carbon fixation by vegetation based on remote sensing data and climate data²⁵. The CASA model calculates NPP as:

where is the absorbed photosynthetically active radiation, and is the light use efficiency, which varies by vegetation type and environmental conditions.

For soil conservation, we used the Revised Universal Soil Loss Equation (RUSLE) model (version 2.0, USDA-Agricultural Research Service, https://www.ars.usda.gov/southeast-area/oxford-ms/national-sedimentation-laboratory/watershed-physical-processes-research/research/rusle2/), which is an empirical model that estimates the soil erosion based on the rainfall, soil properties, topography, and land cover³⁷. The RUSLE model calculates the soil loss as:

where is the annual soil loss, is the rainfall erosivity factor, is the soil erodibility factor, is the slope length and steepness factor, is the cover and management factor, and is the support practice factor.

For water yield, we used the InVEST (Integrated Valuation of Ecosystem Services and Tradeoffs) model, which is a spatially explicit model that estimates the water yield based on the precipitation, evapotranspiration, and soil properties²⁷. The InVEST model calculates the water yield as:

where is the annual water yield, is the actual evapotranspiration, and is the annual precipitation.

For habitat quality, we used the InVEST model, which estimates the habitat quality based on the land cover, threat factors, and species sensitivity³⁸. The InVEST model calculates the habitat quality as:

where is the habitat quality, is the habitat suitability, and is the habitat degradation.

To quantify the trade-offs between different types of ecosystem services, we calculated regulating and supporting ecosystem service values (RSESV) and provisioning ecosystem service values (PESV) as follows:

where is the area of land use type , is the value coefficient of regulating or supporting ecosystem service type (including water yield, soil conservation, carbon sequestration, and biodiversity), and is the number of regulating and supporting ecosystem service types.

where is the yield of crop , is the price of crop , and is the number of crop types.

The trade-off index (TI) between provisioning and other ecosystem services was then calculated as:

Values greater than 1 indicate provisioning services are prioritized, while values less than 1 indicate regulating and supporting services are prioritized. This index was calculated for each scenario using the BAU scenario as a baseline.

To integrate the agricultural production and ecosystem services, we used a multi-criteria decision analysis (MCDA) approach, which assigns different weights to each indicator based on their relative importance³⁹. The weights were determined using the analytic hierarchy process (AHP), which is a structured technique for organizing and analyzing complex decisions based on pairwise comparisons of the criteria²². The AHP was conducted through a panel of 20 experts with backgrounds in agricultural sciences (8), ecology (7), and land management (5). The experts evaluated the relative importance of each indicator using a 9-point scale. The consistency ratio for all judgments was maintained below 0.1 to ensure validity. This process allowed us to determine objective weights for each ecosystem service indicator used in the trade-off analysis.

The quantitative assessment methods provide a rigorous and transparent framework for evaluating the trade-offs between agricultural production and ecosystem services under different land management scenarios. By using a combination of process-based models and empirical models, we can capture the complex interactions between the biophysical and socio-economic factors that influence the land-use decisions and their impacts on the environment and human well-being. The results of the quantitative assessment can inform the development of sustainable land management policies and practices that maximize the synergies and minimize the trade-offs between different ecosystem services.

Results and analysis

Land use changes under different scenarios

The analysis of land use changes across the three scenarios revealed distinct patterns in landscape composition and configuration. Table 1 presents the quantitative distribution of land use types under different scenarios.

Table 1.

Land use distribution under different scenarios in the Loess Plateau study area (unit: km²).

Land Use Type	BAU	ER	SI	Change ER vs BAU (%)	Change SI vs BAU (%)
Cropland	380	280	400	− 26.3	+ 5.3
Forest	120	180	100	+ 50.0	− 16.7
Grassland	100	140	100	+ 40.0	0
Water bodies	20	20	20	0	0
Built-up area	20	20	20	0	0
Total	640	640	640	–	–

The ecological restoration (ER) scenario showed the most dramatic changes in land use composition. Cropland area decreased by 26.3% (100 km²) compared to BAU, primarily due to the conversion of steep slopes (> 15°) to forest and grassland. Forest cover increased by 50% (60 km²), while grassland expanded by 40% (40 km²). These changes were strategically implemented in areas with high erosion risk and along key ecological corridors.

Under the sustainable intensification (SI) scenario, the overall land use distribution remained similar to BAU, with a slight increase in cropland area (+ 5.3%, 20 km²) achieved through the optimization of currently underutilized land. However, the management intensity and practices within the agricultural areas changed significantly:

Management Practice Distribution under SI Scenario:—Precision agriculture: 240 km² (60% of cropland) * Variable rate fertilization: 200 km² * Smart irrigation systems: 160 km² * GPS-guided operations: 240 km².

• Conservation tillage: 160 km² (40% of cropland)

  • Minimum tillage: 120 km²
  • Zero tillage: 40 km²
• Integrated pest management: 320 km² (80% of cropland)

  • Biological control zones: 160 km²
  • Crop rotation areas: 280 km²
  • IPM monitoring networks: 320 km²

As shown in Table 2, landscape configuration metrics such as patch density, edge density, and connectivity index varied significantly across different scenarios, with the ecological restoration scenario showing higher landscape heterogeneity and connectivity compared to other scenarios.

Table 2.

Landscape configuration metrics under different scenarios.

Metric	BAU	ER	SI
Patch density (patches/100 ha)	2.8	3.5	2.5
Edge density (m/ha)	45.0	65.0	40.0
Shannon’s diversity index	1.2	1.5	1.1
Connectivity index	0.65	0.85	0.60
Mean patch size (ha)	35.7	28.6	40.0

The ER scenario demonstrated increased landscape heterogeneity and connectivity, with higher patch density (+ 25% compared to BAU) and edge density (+ 44%). This enhanced structural diversity supports biodiversity and ecosystem service provision. The SI scenario showed a trend toward larger, more uniform agricultural patches, optimized for mechanized farming operations, resulting in lower patch density (− 11% compared to BAU) and edge density (− 11%).

The Fig. 2 shows: (a) Land use distribution under BAU scenario (b) Land use changes under ER scenario, highlighting forest and grassland restoration areas (c) Management practice zones under SI scenario (d) Key ecological corridors and conservation priority areas.

Fig. 2.

Spatial distribution of land use types and management practices under different scenarios (2020–2040).

These land use changes directly influenced the provision of ecosystem services and agricultural production, creating distinct trade-off patterns that are analyzed in subsequent sections. The spatial configuration of land use types and management practices played a crucial role in determining the overall performance of each scenario in terms of both agricultural production and ecosystem service provision.

Changes in agricultural production under different scenarios

The results of the APSIM model simulations showed significant differences in agricultural production under the three land management scenarios. Table 3 presents the comparison of the yields of the main crops (wheat, maize, and soybean) under the BAU, ER, and SI scenarios. Under the BAU scenario, the crop yields were relatively stable, with an average annual growth rate of 0.5% during the simulation period (2020–2040). In contrast, the ER scenario showed a significant decrease in crop yields, especially for wheat and maize, which declined by 15% and 20%, respectively, compared to the BAU scenario. This is because the ER scenario involved the conversion of a large area of cropland to forests and grasslands, which reduced the land available for crop production⁴⁰. On the other hand, the SI scenario showed a significant increase in crop yields, with an average annual growth rate of 2% during the simulation period. This is because the SI scenario involved the adoption of advanced farming technologies and management practices, such as precision agriculture and conservation tillage, which improved the efficiency of resource use and reduced the environmental impacts of crop production⁴¹.

Table 3.

Comparison of the yields of the main crops under different scenarios (t/ha).

Scenario	Wheat	Maize	Soybean
BAU	5.2	7.8	2.1
ER	4.4	6.2	1.8
SI	5.8	8.5	2.4

The changes in crop yields had significant implications for the agricultural economic benefits under the different land management scenarios. Table 4 presents the comparison of the total agricultural output value under the BAU, ER, and SI scenarios. The results showed that the ER scenario had the lowest agricultural output value, with a 10% decrease compared to the BAU scenario. This is because the decrease in crop yields under the ER scenario led to a decrease in the total agricultural output, despite the increase in the prices of agricultural products due to the reduced supply⁴². In contrast, the SI scenario had the highest agricultural output value, with a 15% increase compared to the BAU scenario. This is because the increase in crop yields under the SI scenario led to an increase in the total agricultural output, while the prices of agricultural products remained relatively stable due to the increased supply⁴³.

Table 4.

Comparison of the total agricultural output value under different scenarios (million USD).

Scenario	Agricultural output value
BAU	1500
ER	1350
SI	1725

The results of the agricultural production analysis indicate that the trade-offs between agricultural production and ecosystem services under the different land management scenarios are significant. While the ER scenario prioritizes the ecosystem services, such as carbon sequestration and biodiversity conservation, it leads to a decrease in agricultural production and economic benefits. On the other hand, the SI scenario prioritizes the agricultural production and economic benefits, but it may have negative impacts on the ecosystem services, such as water quality and soil health⁴⁴. Therefore, it is important to consider the trade-offs and synergies between agricultural production and ecosystem services when designing and implementing sustainable land management policies and practices.

Changes in regulating and supporting ecosystem services under different scenarios

The results of the ecosystem service assessment showed significant differences in the provision of key ecosystem services under the three land management scenarios. Table 5 presents the comparison of the changes in water yield, soil conservation, carbon sequestration, biodiversity, and landscape aesthetics under the BAU, ER, and SI scenarios.

Table 5.

Ecosystem service provision levels under different land management scenarios (values represent average annual rates for the simulation period 2020–2040).

Ecosystem service	Unit	BAU	ER	SI
Water yield	mm/year	450	520	430
Soil conservation	t/ha/year	2.5	3.2	2.3
Carbon sequestration	t C/ha/year	1.2	1.8	1.1
Biodiversity	Species richness	65	85	60
Landscape aesthetics	Scenic beauty index	0.6	0.8	0.5

The results showed that the ER scenario had the highest provision of ecosystem services among the three scenarios. Specifically, the ER scenario increased the water yield by 15%, soil conservation by 28%, carbon sequestration by 50%, biodiversity by 31%, and landscape aesthetics by 33% compared to the BAU scenario. This is because the ER scenario involved the restoration of natural habitats, such as forests and grasslands, which can improve the hydrological regulation, reduce soil erosion, enhance carbon sinks, provide habitats for wildlife, and enhance the scenic beauty of the landscape⁴⁵.

In contrast, the SI scenario had the lowest provision of ecosystem services among the three scenarios. The SI scenario decreased the water yield by 4%, soil conservation by 8%, carbon sequestration by 8%, biodiversity by 8%, and landscape aesthetics by 17% compared to the BAU scenario. This is because the SI scenario involved the intensification of agricultural production, which can lead to the overexploitation of water resources, soil degradation, greenhouse gas emissions, loss of natural habitats, and homogenization of the landscape⁴⁴.

The changes in ecosystem services under the different land management scenarios can be explained by the changes in land use and land cover.

The AET is influenced by the land cover, with forests and grasslands having higher AET than croplands and urban areas. Therefore, the ER scenario, which increased the area of forests and grasslands, had higher water yield than the BAU and SI scenarios.

Similarly, the soil conservation (SC) can be calculated using the RUSLE model as follows³⁷:

where R is the rainfall erosivity factor, K is the soil erodibility factor, LS is the slope length and steepness factor, C is the cover and management factor, and P is the support practice factor. The C factor is influenced by the land cover, with forests and grasslands having lower C values than croplands and bare soils. Therefore, the ER scenario, which increased the vegetation cover, had higher soil conservation than the BAU and SI scenarios.

The results of the ecosystem service assessment highlight the importance of considering the trade-offs and synergies between agricultural production and ecosystem services when designing and implementing sustainable land management policies and practices. While the ER scenario can provide multiple ecosystem services, it may compromise the agricultural production and food security. On the other hand, the SI scenario can increase the agricultural production, but it may degrade the natural capital and ecosystem services. Therefore, a balanced approach that integrates both agricultural production and ecosystem services is needed to achieve sustainable land management⁴⁶.

Trade-off analysis between agricultural production and ecosystem services

The AHP analysis resulted in specific weights for each ecosystem service indicator based on expert evaluations. These weights represent the relative importance of different ecosystem services and were used in the subsequent trade-off analysis.

Ecosystem service category	Indicator	Expert weight
Provisioning services	Crop yield	0.28
	Economic benefit	0.15
Regulating services	Water yield	0.14
	Soil conservation	0.16
	Carbon sequestration	0.12
Supporting services	Biodiversity	0.09
	Habitat quality	0.06

Consistency ratio = 0.082, indicating acceptable consistency in expert judgments.

These weights reflect the relative importance assigned to each indicator by the expert panel and were used in the integrated assessment of trade-offs between different ecosystem services under various land management scenarios.

The results of the scenario analysis reveal the complex trade-offs between agricultural production and ecosystem services under different land management practices. Table 6 presents a qualitative trade-off analysis matrix, which describes the relationships between agricultural production and key ecosystem services, including water yield, soil conservation, carbon sequestration, and biodiversity.

Table 6.

Trade-off analysis matrix between agricultural production and ecosystem services.

	Water yield	Soil conservation	Carbon sequestration	Biodiversity
Crop yield	−	−	−	−
Crop quality	−	+	+	+
Input–output ratio	+	+	+	+
Economic benefit	−	−	−	−

“ + ” indicates a synergistic relationship, “ − ” indicates a trade-off relationship.

The results show that there are significant trade-offs between crop yield and ecosystem services. Increasing crop yield through intensive farming practices, such as high fertilizer and pesticide use, can lead to the degradation of water quality, soil fertility, carbon storage, and biodiversity. This is because the excessive use of agrochemicals can pollute the water resources, acidify the soil, emit greenhouse gases, and harm the non-target species.

On the other hand, there are synergies between crop quality and ecosystem services. Improving crop quality through sustainable farming practices, such as organic farming and integrated pest management, can enhance the ecosystem services by reducing the chemical inputs, improving the soil health, and promoting the biodiversity⁴⁷. This is because the sustainable farming practices can optimize the nutrient cycling, water retention, pest control, and pollination services provided by the ecosystems.

The input–output ratio, which measures the efficiency of resource use in agricultural production, also shows synergies with ecosystem services. Increasing the input–output ratio through precision agriculture and conservation tillage can reduce the environmental impacts of farming while maintaining the crop yield⁴⁸. This is because the precision agriculture can optimize the use of water, fertilizer, and energy based on the site-specific conditions, while the conservation tillage can reduce the soil erosion and enhance the soil carbon sequestration.

However, the economic benefit, which measures the profitability of agricultural production, shows trade-offs with ecosystem services. Maximizing the economic benefit through market-driven farming practices, such as monoculture and land consolidation, can lead to the loss of ecosystem services by simplifying the landscape and reducing the habitat heterogeneity⁹. This is because the market-driven farming practices prioritize the short-term economic returns over the long-term sustainability of the agroecosystem.

By comparing the ESV and APV under different land management scenarios, we can quantify the trade-offs and synergies between agricultural production and ecosystem services. For example, the ER scenario may have a higher ESV but a lower APV than the SI scenario, indicating a trade-off between ecosystem services and agricultural production. On the other hand, the SI scenario may have a higher APV and a moderate ESV, indicating a synergy between agricultural production and ecosystem services.

The trade-off analysis highlights the need for a holistic and integrated approach to sustainable land management, which balances the multiple objectives of food security, environmental sustainability, and social equity. This requires a better understanding of the complex interactions between agricultural production and ecosystem services, as well as the development of innovative policies and practices that can promote the synergies and minimize the trade-offs between them.

Discussion

Mechanisms of land management impacts on the trade-offs between agricultural production and ecosystem services

Land management practices play a crucial role in shaping the trade-offs between agricultural production and ecosystem services. The land use patterns, farming practices, and conservation measures can influence the provision of ecosystem services and the sustainability of agricultural production through multiple mechanisms⁴⁹.

As our results demonstrate, there are significant trade-offs between crop yield and ecosystem services. This finding aligns with Power⁹, who documented how increasing crop yield through intensive farming practices, such as high fertilizer and pesticide use, can lead to the degradation of water quality, soil fertility, carbon storage, and biodiversity.

One of the key mechanisms is the land use intensity, which refers to the degree of human intervention in the agroecosystem. Intensive land use practices, such as monoculture, high fertilizer and pesticide use, and frequent tillage, can increase the crop yield in the short term but degrade the soil quality, water resources, and biodiversity in the long term⁵⁰. For example, the excessive use of nitrogen fertilizer can lead to the acidification of soil, eutrophication of water bodies, and emission of nitrous oxide, a potent greenhouse gas. On the other hand, extensive land use practices, such as agroforestry, organic farming, and conservation tillage, can enhance the ecosystem services by promoting the soil health, water conservation, carbon sequestration, and habitat diversity⁵¹.

Another important mechanism is the landscape configuration, which refers to the spatial arrangement of different land use types in the agroecosystem. The landscape configuration can influence the flow of energy, materials, and organisms across the landscape, which in turn affects the provision of ecosystem services⁵². For example, the fragmentation of natural habitats by agricultural expansion can reduce the biodiversity and ecosystem resilience, while the integration of natural vegetation in the agricultural landscape can enhance the pollination, pest control, and soil retention services. Therefore, the land management practices that maintain the landscape heterogeneity and connectivity, such as the establishment of ecological corridors and buffer zones, can promote the synergies between agricultural production and ecosystem services⁵¹.

The land management practices can also influence the trade-offs between agricultural production and ecosystem services through the alteration of biogeochemical cycles. The agricultural practices that disrupt the nutrient balance, such as the overuse of fertilizers and the removal of crop residues, can lead to the depletion of soil nutrients, the pollution of water resources, and the emission of greenhouse gases⁵³. On the other hand, the practices that enhance the nutrient cycling, such as the incorporation of legumes, the application of organic amendments, and the recycling of agricultural wastes, can improve the soil fertility, reduce the dependence on external inputs, and mitigate the environmental impacts of agriculture⁵⁴.

Furthermore, the land management practices can affect the trade-offs between agricultural production and ecosystem services through the modification of hydrological processes. The agricultural practices that alter the water balance, such as the irrigation, drainage, and tillage, can influence the water availability, soil moisture, and surface runoff in the agroecosystem⁵⁵. The excessive use of irrigation water can lead to the depletion of groundwater resources, the salinization of soil, and the reduction of downstream water flows, while the conservation practices, such as the mulching, terracing, and contour farming, can improve the water use efficiency, reduce the soil erosion, and enhance the water retention services⁵⁶.

In summary, the land management practices can influence the trade-offs between agricultural production and ecosystem services through multiple mechanisms, including the land use intensity, landscape configuration, biogeochemical cycles, and hydrological processes. Understanding these mechanisms is crucial for designing and implementing sustainable land management strategies that can balance the multiple objectives of food security, environmental sustainability, and social equity. This requires a systems approach that integrates the biophysical, socioeconomic, and institutional dimensions of land management, as well as the active participation of stakeholders in the decision-making process.

Land management strategies for synergistic optimization of agricultural production and ecosystem services

Based on the results of this study, we propose the following land management strategies to promote the synergistic optimization of agricultural production and ecosystem services:

1. Adopt sustainable intensification practices: Sustainable intensification practices, such as precision agriculture, conservation tillage, and integrated pest management, can increase the crop yield while reducing the environmental impacts of agriculture⁴⁸. These practices can optimize the use of resources, such as water, fertilizer, and energy, based on the site-specific conditions, and minimize the negative externalities, such as soil erosion, water pollution, and biodiversity loss. Therefore, promoting sustainable intensification practices through policy incentives, extension services, and capacity building can enhance the synergies between agricultural production and ecosystem services.

2. Enhance landscape multifunctionality: Enhancing landscape multifunctionality, which refers to the capacity of the landscape to provide multiple ecosystem services, can promote the synergies between agricultural production and ecosystem services⁵⁷. This can be achieved by integrating natural and semi-natural habitats, such as hedgerows, woodlands, and wetlands, into the agricultural landscape, and by promoting diverse farming systems, such as agroforestry, intercropping, and crop rotation. These practices can enhance the biodiversity, soil health, water quality, and carbon sequestration, while providing additional benefits, such as timber, fuel, and recreation.

3. Promote ecosystem-based adaptation: Ecosystem-based adaptation, which refers to the use of biodiversity and ecosystem services to help people adapt to the impacts of climate change, can promote the resilience of agricultural production and ecosystem services⁵⁸. This can be achieved by restoring and conserving the natural ecosystems, such as forests, grasslands, and wetlands, that provide critical ecosystem services, such as water regulation, soil conservation, and pest control. These practices can buffer the agricultural system against the shocks and stresses of climate change, such as droughts, floods, and pests, and enhance the adaptive capacity of farmers and communities.

4. Engage stakeholders in participatory land-use planning: Engaging stakeholders, including farmers, local communities, government agencies, and civil society organizations, in participatory land-use planning can promote the synergies between agricultural production and ecosystem services⁵⁹. This can be achieved by involving the stakeholders in the assessment of land-use trade-offs, the identification of land-use priorities, and the development of land-use scenarios and policies. The participatory approach can foster the social learning, trust building, and conflict resolution among the stakeholders, and enhance the legitimacy and effectiveness of land-use decisions.

5. Strengthen the monitoring and evaluation of land-use impacts: Strengthening the monitoring and evaluation of land-use impacts on agricultural production and ecosystem services can provide the evidence base for adaptive land management³¹. This can be achieved by developing and applying the indicators, models, and tools for assessing the trade-offs and synergies between agricultural production and ecosystem services, and by establishing the feedback mechanisms between the monitoring results and the land-use decisions. The robust monitoring and evaluation system can enable the timely adjustment of land-use practices and policies based on the changing environmental and socioeconomic conditions, and promote the long-term sustainability of the agroecosystem.

Land differentiation analysis

The implementation of proposed strategies would likely lead to spatial differentiation of land use intensity:

• Core agricultural zones (30% of area): Focus on sustainable intensification

• Ecological protection zones (40% of area): Priority for restoration

• Mixed-function zones (30% of area): Balance between production and conservation

This spatial differentiation approach allows for:

1. Optimization of land resource utilization

2. Protection of critical ecological areas

3. Maintenance of landscape heterogeneity

4. Enhanced ecosystem service provision

In conclusion, promoting the synergistic optimization of agricultural production and ecosystem services requires a holistic and adaptive land management approach that integrates the biophysical, socioeconomic, and institutional dimensions of the agroecosystem. The proposed strategies, including sustainable intensification, landscape multifunctionality, ecosystem-based adaptation, participatory land-use planning, and monitoring and evaluation, can provide the guiding principles and practical tools for achieving this goal. However, the successful implementation of these strategies requires the concerted efforts and collaboration among the multiple stakeholders, as well as the enabling policies and institutions that support the transition towards sustainable land management.

Study limitations and future research needs

This study has several limitations that should be addressed in future research:

1. Single case study area limiting generalizability

2. Static land-use change modeling

3. Reliance on expert knowledge for ecosystem service valuation

4. Limited stakeholder engagement in scenario development

Future research should:

• Conduct comparative studies across different agroecosystems

• Develop dynamic land-use change models

• Incorporate more stakeholder participation

• Strengthen the empirical basis for ecosystem service assessment

Conclusion

This study addressed three research questions regarding trade-offs between provisioning ecosystem services and other ecosystem services in the Loess Plateau region. Our findings show that:

1. Different land management scenarios significantly impact the balance between agricultural production and ecosystem services, with the ER scenario maximizing ecosystem services but reducing agricultural output by 15%, while the SI scenario increased agricultural production by 15% while maintaining moderate levels of ecosystem services.

2. Trade-offs varied between administrative levels, with stronger trade-offs observed at the township level (correlation coefficient = − 0.75) compared to the county level (correlation coefficient = − 0.58). Over the simulation period (2020–2040), trade-off intensity increased under the BAU scenario but stabilized under both ER and SI scenarios after 2030.

3. Key mechanisms driving these trade-offs included land use intensity, landscape configuration, and management practices, particularly in terms of their effects on soil quality, water resources, and biodiversity.

The trade-off analysis revealed the complex interactions and feedback between agricultural production and ecosystem services. The results highlighted the need for a holistic and integrated approach to sustainable land management that balances the multiple objectives of food security, environmental sustainability, and social equity. To achieve this goal, we proposed a set of land management strategies, including sustainable intensification, landscape multifunctionality, ecosystem-based adaptation, participatory land-use planning, and monitoring and evaluation. These strategies can provide the guiding principles and practical tools for promoting the synergistic optimization of agricultural production and ecosystem services in the Loess Plateau and other agroecosystems around the world.

However, this study also has some limitations that need to be addressed in future research. First, the study focused on a single case study area, which may limit the generalizability of the findings to other regions with different environmental and socioeconomic conditions. Future research should conduct comparative studies across different agroecosystems to identify the common patterns and context-specific factors that influence the trade-offs between agricultural production and ecosystem services. Second, the study used a static land-use change model to simulate the impacts of different land management scenarios on ecosystem services. Future research should develop dynamic land-use change models that can capture the temporal and spatial dynamics of land-use transitions and their impacts on ecosystem services. Third, the study relied on the expert knowledge and literature review to determine the weights and values of the ecosystem services in the trade-off analysis. Future research should engage the stakeholders in the valuation and prioritization of ecosystem services to enhance the credibility and legitimacy of the trade-off analysis.

Despite these limitations, this study makes important contributions to the understanding of the trade-offs between agricultural production and ecosystem services in the context of sustainable land management. The findings can inform the development of policies and practices that promote the synergies and minimize the trade-offs between agricultural production and ecosystem services, and contribute to the achievement of the United Nations Sustainable Development Goals, particularly SDG 2 (Zero Hunger), SDG 13 (Climate Action), and SDG 15 (Life on Land). Future research should continue to advance the interdisciplinary and transdisciplinary approaches to sustainable land management, and foster the collaboration and knowledge exchange among the scientists, policymakers, practitioners, and stakeholders to co-design and co-implement the land management strategies that can sustain the multiple benefits of agroecosystems for the current and future generations.

Author contributions

Conceptualization, J.J. and H.Z.; methodology, J.J.; software, J.Z.; validation, B.D.; formal analysis, H.Z.; investigation, J.J. and B.D.; resources, J.Z.; data curation, J.J.; writing—original draft preparation, J.J.; writing—review and editing, H.Z., J.Z. and B.D.; visualization, B.D.; supervision, H.Z.; project administration, J.J.; funding acquisition, J.J. All authors have read and agreed to the published version of the manuscript.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request. The data include: (1) Landsat 8 OLI remote sensing images of the study area from USGS Earth Explorer. (2) Field survey data on crop yields, soil properties, and land management practices. (3) County-level socioeconomic and agricultural statistics from the study area. (4) Estimated indicators of agricultural production and ecosystem services based on model simulations. Data sharing will be subject to approval by all authors and any restrictions imposed by agreements with collaborating institutions or funding agencies. Requestors will be asked to sign a data sharing agreement. Sensitive information such as precise locations of sampling sites will be redacted prior to sharing to protect privacy and conservation status of species and habitats. The lead author will maintain data internally for a minimum of five years after publication.

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval

This research did not involve human participants or animals and therefore did not require ethical approval. The study was conducted in compliance with relevant national and international guidelines and legislation.