Trade-offs, synergies and scale effects of ecosystem services in the middle reaches of the Yellow River

Ecosystems form the foundation of sustainable human development by delivering a wide range of services that directly or indirectly shape human well-being. The provision of these services is influenced not only by ecological processes but also by the complexity of socio-economic systems. As a result, interactions among different ecosystem services are often intricate, typically expressed as trade-offs—where gains in one service come at the expense of another—or as synergies, where multiple services are enhanced simultaneously¹. Such trade-off and synergy relationships are not limited to a single region but exhibit universal spatial and temporal characteristics across global terrestrial ecosystems, especially the coupling relationship between carbon cycle components and water cycle components, which has become a core research focus in global ecosystem service studies.

Considerable progress has been made in ecosystem service research, with studies focusing on valuation², functional accounting³, spatiotemporal evolution⁴, driving factors⁵, and trade-off/synergy relationships⁶. At the global and interregional scale, numerous studies have confirmed the typical trade-off or synergy between carbon gain and water loss in terrestrial ecosystems: Kim et al.⁷, revealed that land use change in global grassland ecosystems led to a significant trade-off between carbon sequestration capacity and water yield, with the expansion of artificial grassland increasing carbon storage but reducing regional water retention capacity; Zhang et al.⁸ further found in the study of global semi-arid ecosystems that the increase in vegetation carbon uptake was accompanied by a significant rise in water use efficiency, forming a positive synergy between carbon and water cycle components, and this synergy was closely related to regional precipitation and vegetation type differentiation. In the Asian monsoon region, Chakraborty et al.⁹, explored the response of ecosystem carbon and water use efficiencies to groundwater variability in India, pointing out that human-induced groundwater overexploitation would break the original carbon-water synergy and lead to the decline of both carbon sequestration and water yield services, which provided an important reference for the study of carbon-water interaction in ecologically fragile areas with intense human activities.

In China, valuation has often relied on the improved equivalent factor method developed by Xie et al.¹⁰. However, its heavy dependence on this method and classification has limited its comprehensiveness. Recent approaches increasingly emphasize integration with ecological processes and the adoption of dynamic, model-based frameworks¹¹, and the application of remote sensing technology in the study of carbon and water use efficiencies has become a mainstream trend—by quantifying key parameters such as NDVI, evapotranspiration and net primary productivity (NPP), remote sensing can effectively characterize the spatiotemporal variation of ecosystem carbon and water cycle processes, and further reveal the intrinsic mechanism of carbon-water trade-off and synergy^{8, 9}. Among them, the InVEST model has become one of the most widely applied tools for ecosystem service assessment worldwide, which can effectively couple remote sensing data with ecological process models to realize the quantitative evaluation of carbon and water-related ecosystem services.

Research on trade-offs and synergies has employed various methods, including statistical analysis¹², spatial mapping¹³, and scenario simulation¹⁴. However, Chinese studies have largely favored correlation-based methods, often without adequately addressing methodological differences or applicability, and most regional studies focus on the internal characteristics of ecosystem service interactions in China, with insufficient connection to global universal laws and interregional comparison references. Moreover, most assessments adopt a static perspective, overlooking the historical dynamics of ecosystem service bundles. Analyses are also frequently conducted at the raster scale, with limited attention to administrative units such as counties, which are more directly relevant for policy and management. These limitations underscore the need for methodological frameworks that can more accurately capture the trade-offs and synergies of ecosystem services at scales meaningful to governance, and further combine global carbon-water interaction laws and remote sensing-based carbon water use efficiency research to improve the regional and global relevance of the study results.

When ecosystem service dynamics are analyzed at a single scale, the complex cross-scale interactions of ecological processes are often overlooked, which may lead to a mismatch between ecosystem service management strategies and the diversified supply of services. Therefore, it is necessary to adopt a multi-scale perspective to examine the trade-offs and synergies among ecosystem services, identify the relevant driving factors, and clarify the scale effects that influence these dynamics. Such an approach is crucial for guiding decision-makers in formulating sound management strategies that are suited to local ecological conditions.

The middle reaches of the China Yellow River are among the regions in the basin where human–environment conflicts are most acute, and as a typical semi-arid to semi-humid transitional zone, its carbon and water cycle processes and ecosystem service interactions are highly consistent with the research context of global semi-arid ecosystems⁸. Under the China national strategy of ecological protection and high-quality development, there is an urgent need to assess the spatiotemporal dynamics of ecosystem services and their interrelationships in this region. To address this gap, this study evaluates four key ecosystem services—grain production, water yield, soil conservation, and carbon sequestration—across 228 counties in six provinces for the years 2000, 2010, and 2020. Using the InVEST model in combination with correlation analysis and comparative approaches, we examine both static and dynamic trade-offs and synergies, and further compare the research results with global and other regional carbon-water interaction studies to reveal the regional characteristics and universal laws of ecosystem service trade-offs and synergies in the study area. The findings are intended to provide scientific evidence to support ecological management and policy-making in the area, and also offer a regional case for global research on carbon-water cycle coupling and ecosystem service interactions.

The middle reaches of the Yellow River are characterized by highly diverse and complex topography, and have long been plagued by severe soil erosion as well as frequent flood and drought disasters, resulting in an extremely fragile ecological environment. Taking this region as the study area (Fig. 1), its landscape features are marked by significant elevation differences and a mosaic distribution of multiple geomorphic units, including the Loess Plateau, the Linfen Basin, and the Guanzhong Plain.

In practical applications of ecological function management and data collection, watershed units are reasonable from the perspective of physical geography. However, they present certain challenges at the implementation level, such as administrative ambiguity and high coordination costs. In contrast, administrative units—with their clearly defined boundaries and established governance structures—enable more efficient and orderly implementation of ecological function regulation measures.

To ensure that the spatial units used in this study reflect both the natural continuity of watershed boundaries and the administrative consistency of county-level management, we adopted the natural watershed extent of the middle reaches of the Yellow River as delineated by the Yellow River Conservancy Commission. Within this boundary, a total of 228 county-level administrative units—including counties, county-level cities, banners, and municipal districts—from six provinces (autonomous regions)—Inner Mongolia, Ningxia, Gansu, Shanxi, Shaanxi, and Henan—were incorporated. These units collectively encompass the core ecological zone of the middle reaches of the Yellow River, covering a total area of approximately 343,800 km².

To balance the computational burden against the need for spatial detail and align with the analytical scale of macro-ecological patterns in the middle reaches of the Yellow River, two grid resolutions—5 km × 5 km and 10 km × 10 km—were adopted in the raster-scale analysis. The 5 km × 5 km grid was designed to capture the topographic fragmentation and intensive soil erosion in the core Loess Plateau, thereby better reflecting the driving effects of micro-landform units on the spatial differentiation of ecosystem services. The 10 km × 10 km grid, by contrast, was intended to mitigate local noise and reveal the overall trends of trade-offs and synergies among ecosystem services at a macro-scale, supporting regional ecological governance planning. Given the large study area (approximately 343,800 km²), both resolutions effectively reduced model runtime and data storage costs while maintaining analytical accuracy. Considering practical implementation requirements, county-level administrative units were incorporated as the administrative scale. Accordingly, a total of 24,540, 6,322, and 228 spatial units were generated across the three scales, respectively (Fig. 1).

Quantification of ecosystem services

The classification of ecosystem services has been explored extensively in the literature¹⁵. Conventionally, ecosystem services are divided into four categories: provisioning services (e.g., grain production, raw materials, and water yield), regulating services (e.g., carbon storage, water purification, waste treatment, and soil conservation), supporting services (e.g., biomass production and biodiversity), and cultural services (e.g., education, recreation, and non-commercial use). Considering that the middle reaches of the Yellow River are among the regions facing the greatest environmental pressure¹⁶, and fastest growth in resource demand, this study selected ecosystem service functions that are closely related to human well-being.Taking into account model applicability, statistical techniques, and data availability, four key services—grain production, water yield, soil conservation, and carbon sequestration—were quantified and spatially mapped.

The region serves as a critical grain-producing base in northern China and an important ecological security barrier. It is dominated by forest–grassland ecosystems, which play a major role in national carbon sequestration and storage. Based on this context, the four core ecosystem service functions were systematically evaluated.

Grain production

Previous studies¹⁷, have shown a notable linear correlation between grain yield and NDVI (Normalized Difference Vegetation Index). The calculation formula is as follows:

In the formula,(:{text{FP}}_{text{ij}}) is the grain yield of the jth cultivated land unit in the ith county district (t) (:{text{FP}}_{text{i}}) is the total grain output of the i th county (:{text{NDVI}}_{text{ij}}) is the NDVI of the j th cultivated land grid unit in the i th county and (:{text{NDVI}}_{text{i}}) is the total NDVI of cultivated land in the i th county.

Water yield service

In the InVEST model, the water yield module is drawing upon the principle of water balance, where water yield is calculated as precipitation minus water storage and evapotranspiration losses (Eq. 4). In this study, the water yield module of InVEST was applied to estimate water yield. Using meteorological data, including mean annual precipitation and potential evapotranspiration, the annual water yield Yx for each grid cell x was estimated as follows:

In the formula, AETx is the actual annual evapotranspiration in grid cell x (mm); Px is the annual precipitation in grid cell x (mm); PETx is the potential evapotranspiration (mm); ωx represents the climate–soil parameter; Z is the seasonality factor; and AWCx is the plant-available water content (mm). The parameter Z captures local precipitation patterns and additional hydrogeological characteristics¹⁸. A value of 1.25 is used as the minimum threshold for ω, which corresponds to bare soil conditions (rooting depth = 0)¹⁹. The biophysical factors used in this module were modified on the basis of previous studies and expert knowledge^{20, 21}.

Carbon sequestration service

The carbon storage module in the InVEST model reflects both the quantity and variation of carbon storage, as well as how land-use change alters the amount of stored and sequestered carbon across time and space. Based on previous studies²², the biological carbon density of different land-use types per unit area was obtained, and the total carbon storage was then calculated using the following formula:

In the formula CS_total is the total carbon storage (t); CS_above is the aboveground carbon storage (t); CS_below is the belowground carbon storage (t); CS_soil is the soil carbon storage (t); and CS_dead is the dead organic matter carbon storage (t).

Soil conservation service

The Revised Universal Soil Loss Equation (RUSLE) was applied to estimate the erosion control capacity of the ecosystem at a spatial resolution of 1 km × 1 km for a given period (one year in this study). Specifically, potential and actual soil erosion were calculated based on land-use patterns, and soil conservation was derived using the following equations:

USLE_pi is the potential annual soil erosion per unit area (t/ha); USLE_a.i. is the actual annual soil erosion per unit area (t/ha); USLE_i is the annual soil conservation (t/ha); R_i is the rainfall erosivity factor of grid i (MJ·mm/(ha·h)); K_i is the soil erodibility factor of grid i ((t·ha·h)/(ha·MJ·mm)); LS_i is the slope length and slope steepness factor of grid i; C_i is the vegetation cover factor of grid i; and P_i is the soil conservation practice factor of grid i.

Data sources and data processing

The datasets used in this study included (1) land-use data of the study area for the years 2000, 2010, and 2020 and (2) environmental and climatic data covering topography, climate, soil type, river network, slope, and the Normalized Difference Vegetation Index (NDVI). Detailed data sources are listed in Table 1.

All maps in this article are sourced from the Standard Map Service website of the Ministry of Natural Resources of China (http://bzdt.ch.mnr.gov.cn/). The base maps have not been modified. Map Approval No.: GS (2019) No. 1822.The data was then mosaicked, cropped using vector boundaries, and projected into the Xi’an 80 coordinate system, with a spatial resolution of 1 km × 1 km.All images were processed using ArcGIS 10.7.

Methods for trade-off and synergy analysis of ecosystem services

In this study, the correlation coefficient method was applied to examine the static trade-off and synergy relationships of ecosystem services in the middle reaches of the Yellow River, while the difference comparison method was used to analyze their dynamic trade-off and synergy relationships.

The static trade-off/synergy relationships were calculated using the following formula:

ES₁ and ES₂ represent two types of ecosystem services; r is the correlation coefficient; n is the length of the time series; and i and j denote the row and column indices of the data, respectively.

Based on the difference comparison method^{23, 24}, synergistic relationships were further divided into positive synergy and negative synergy, as expressed in the following formula:

A_T₁ and A_T₂ represent the values of ecosystem service type A in periods T1 and T2, respectively; B_T₁ and B_T₂ represent the values of ecosystem service type B in periods T1 and T2, respectively; and ΔA and ΔB denote the changes in ecosystem service types A and B from T₁ to T₂. If C = 0, the relationship is compatible; if C < 0, a trade-off relationship exists; if C > 0 and both ΔA and ΔB are positive, the relationship is positive synergy; and if C > 0 and both ΔA and ΔB are negative, the relationship is negative synergy.

Spatiotemporal evolution of ecosystem services

According to the calculation methods and data presented above, the specific changes in various ecosystem service functions are shown in Table 2.

Grain production

Between 2000 and 2020, grain production in the middle reaches of the Yellow River increased notably. The average grain yield per unit area rose from 0.892 t/ha in 2000 to 1.179 t/ha in 2020, an increase of 32.17%. The coefficient of variation decreased from 1.258 to 1.140, indicating reduced spatial disparities. The number of low-value areas decreased from 121 counties in 2000 to 101 in 2020, while high-value areas increased from 11 to 23. Spatially, grain production exhibited a distinct pattern of being high in the southeast and low in the northwest. Regions such as the southern Shanxi Basin, the western Henan Basin, and the Guanzhong Plain exhibited strong productivity due to favorable water and heat conditions, whereas parts of the Longdong Plateau improved grain production through the expansion of dry farming.

Based on NDVI values and county-level grain output, the spatial pattern of annual grain production services at the grid scale in the middle reaches of the Yellow River was obtained using NDVI values and county-level grain output (Fig. 2). The overall distribution revealed a distinct regional differentiation of being high in the southeast and low in the northwest. High-value areas were concentrated in cropland-dense regions, primarily in the central and southeastern parts of the study area, while most of the northwestern region was dominated by grassland, woodland, and unused land, with relatively low grain supply capacity.

Water yield service

From 2000 to 2020, the water yield service in the middle reaches of the Yellow River exhibited an overall increasing trend. The mean annual water yield rose from 201.33 mm in 2000 to 234.41 mm in 2010, and further to 275.03 mm in 2020, representing a total growth of 36.61%. In terms of growth stages, the growth from 2000 to 2010 was 16.43%, while the growth from 2010 to 2020 was 17.33%. During this period, regional disparities in water yield decreased, with the coefficient of variation declining from 0.409 to 0.173, a reduction of 136.42%.

At the grid scale, the spatial pattern of water yield service displayed a distinct trend of being high in the south and low in the north, which was largely consistent with precipitation distribution and inversely related to evapotranspiration. High-value areas were concentrated in the Qinling Mountains and western Henan Basin, where annual precipitation is abundant, potential evapotranspiration is relatively low, and forest cover is dense, resulting in strong water retention capacity. In contrast, low-value areas were primarily located in the Ordos Plateau and the northern Loess Plateau, where low precipitation, high potential evapotranspiration, and sparse vegetation cover limit water retention, leading to a weak water yield service.

At the county scale (Fig. 3) the water yield service exhibited distinct spatial heterogeneity, consistent with the grid-scale pattern of high in the south and low in the north. The lowest values were recorded in Yanchi County, located at the southeastern edge of the Mu Us Desert in Ningxia, where natural conditions are extremely harsh and fragile. The highest values were observed in Song County and Luanchuan County, both situated in the mountainous region of western Henan. These areas are characterized by complex terrain, relatively high elevation, a temperate continental monsoon climate, and abundant forest resources, all of which contribute to a strong water yield service.

Carbon sequestration service

From 2000 to 2020, carbon storage in the middle reaches of the Yellow River remained generally stable but exhibited a slight downward trend. The mean value declined from 66.73 t/ha in 2000 to 64.72 t/ha in 2020, a decrease of 3.10%. In the first decade (2000–2010), carbon storage fell marginally from 66.73 t/ha to 66.20 t/ha (−0.79%), while in the following decade (2010–2020) it decreased further from 66.20 t/ha to 64.70 t/ha (−2.32%). Over the same period, the coefficient of variation increased from 0.186 to 0.213, indicating widening regional disparities.

At the grid scale, the spatial distribution of the carbon sequestration service remained largely unchanged over the 20-year past. High-value areas were concentrated in the Qinling Mountains, the Yan’an uplands, the Lüliang Mountains, and the Taihang Mountains, where extensive and contiguous forest cover provides strong carbon sequestration capacity. In contrast, low-value areas were primarily distributed across the Ordos Plateau and around large urban agglomerations, where intensive land use and urban expansion have substantially reduced carbon storage capacity.

From the county-level perspective (Fig. 4), the carbon sequestration service exhibited slight local expansion or contraction but no notable overall change. For instance, during 2000–2010, counties such as Minxian and Zhangxian on the Longdong Plateau shifted from medium- to high-value categories, while Kelan and Hengqu declined from high-value to relatively high-value categories. Between 2010 and 2020, the Huazhou District of Weinan City shifted from high- to relatively high-value, whereas Weiyuan County declined from relatively high- to medium-value.

Extremes were also evident: the lowest carbon storage values were consistently found in the urban districts of Xi’an (e.g., Lianchi and Weiyang), which are dominated by dense populations and extensive built-up land—characterized by the lowest carbon density among land-use types. By contrast, the highest values were recorded in Taibai County, a region with complex geomorphology, high elevation, a sparse population, and extensive mountain forests, where forestland carbon density reaches up to 298 t/ha, the highest among all land-use categories.

Soil conservation service

From 2000 to 2020, the soil conservation service in the middle reaches of the Yellow River exhibited a distinct increasing trend. The mean soil retention capacity rose from 249.84 t/ha in 2000 to 395.24 t/ha in 2020, an increase of 58.19%. In the first stage (2000–2010), the average value grew sharply from 249.84 t/ha to 375.59 t/ha (+ 50.33%), while growth was more moderate in the second stage (2010–2020), rising from 375.59 t/ha to 395.24 t/ha (+ 5.23%). During this period, regional disparities declined, with the coefficient of variation dropping from 1.180 to 0.928, a reduction of 27.16%.

At the grid scale, the soil conservation service displayed pronounced spatial heterogeneity, with high-value areas primarily distributed along the northern slopes of the Qinling Mountains. By contrast, large parts of the Longdong Plateau, the Ordos Plateau, and the northern Loess Plateau exhibited weak soil conservation capacity due to high elevation, an arid climate, and severe soil erosion, soil conservation values were also relatively low in the Guanzhong–Weihe Plain, where cropland dominates and human activity is intensive.

At the county scale (Fig. 5), spatial differences were notable. High-value and relatively high-value counties were concentrated in the southern Qinling Mountains and the eastern Lüliang Mountains, while low-value and relatively low-value counties were clustered in the Ordos Plateau and the northern Loess Plateau. The lowest values were consistently found in urban districts of Xi’an, reflecting the dominance of built-up land with poor soil retention capacity. In contrast, the highest values were recorded in Taibai and Jiaocheng County, where mountainous terrain, extensive forest cover, and favorable ecological conditions contributed to strong soil conservation capacity.

Trade-offs and synergies among ecosystem services

Static trade-off synergy relationship analysis

Overall (Table 3), the six ecosystem service pairs exhibit significant static trade-off or synergy relationships. From a temporal perspective, the relationship between grain production and water yield is synergistic throughout the study period, but the degree of synergy shows a gradually decreasing interannual trend. Grain production and soil retention show a trade-off relationship in all years, with the trade-off gradually weakening over time. Grain production and carbon sequestration remain in a trade-off relationship, with little interannual variation. The relationship between water yield and soil retention is consistently synergistic during the study period, with the strength of synergy gradually declining over time. Water yield and carbon sequestration also maintain a synergistic relationship over the study period, again with a gradually decreasing interannual trend. In contrast, soil retention and carbon sequestration are synergistic throughout the study period, and their synergy shows a gradually increasing trend over time.

From the perspective of scale effects, grain production and water yield are synergistic at the 5 km × 5 km, 10 km × 10 km, and county scales, and their correlation coefficients first increase and then decrease as the scale increases. Grain production and soil retention exhibit a trade-off relationship at the 5 km × 5 km and 10 km × 10 km scales; at the county scale, the correlation does not pass the significance test in 2000 and 2010, and shifts to a trade-off relationship in 2020. Grain production and carbon sequestration are in a trade-off relationship across all scales, with no obvious pattern of variation among scales. Water yield and soil retention are synergistic at the 5 km × 5 km, 10 km × 10 km, and county scales, and their correlation coefficients similarly show a pattern of first increasing and then decreasing with increasing scale. Water yield and carbon sequestration are synergistic at the 5 km × 5 km and 10 km × 10 km scales; at the county scale, the correlation is not statistically significant in 2000, becomes weakly synergistic in 2010, and turns into a significant synergistic relationship in 2020. Soil retention and carbon sequestration are synergistic at all scales, and their correlation coefficients increase with scale, indicating a progressively stronger synergy.

Dynamic trade-off synergy relationship analysis

Grain production and water yield services

From 2000 to 2010, the spatial synergy pattern between grain production and water yield in the middle reaches of the Yellow River was overwhelmingly dominant, with positively synergistic areas extensively distributed at all scales. During this period, the initial promotion of agricultural water-saving technologies and the implementation of ecological conservation policies provided strong support for the benign interaction between grain production and water yield. Trade-off relationships appeared only sporadically in a few ecologically sensitive areas, accounting for a very small proportion and exerting limited influence on the regional water–food relationship (Fig. 6).

From 2010 to 2020, the pattern underwent a marked transformation. Positively synergistic areas contracted significantly, and some originally synergistic regions gradually shifted to compatible or trade-off relationships, closely related to the intensified agricultural development and increased water consumption during this period. At the same time, trade-off areas expanded considerably, especially at the county scale, where regions with negative synergy became clustered in certain counties and continued to enlarge, clearly reflecting the growing disturbance of the water–food relationship by human activities.

At the fine scale of 5 km × 5 km, positive synergy between grain production and water yield shows a fragmented, patchy distribution, mostly confined to micro-regions such as small irrigation districts and local plain plots. At this scale, the synergy pattern is strongly influenced by micro-topography, field-level irrigation management, and other local factors, resulting in very high spatial heterogeneity. Trade-off areas are likewise scattered, mainly concentrated in the transition zones between agricultural land and ecological land, indicating the fine-scale shaping of the water–food relationship by the micro-environment.

At the medium scale of 10 km × 10 km, the spatial continuity of positively synergistic areas is significantly enhanced, forming relatively contiguous synergistic zones along valley plains such as those of the Fen River and Wei River. This pattern highlights the close linkage between agricultural spatial layout and water resource allocation at the medium scale. Trade-off areas form relatively concentrated patches, mostly located in the transition belts from hilly areas to plains; their distribution is directly related to the planting structure and regional water allocation schemes at this scale, presenting a clear regional pattern.

At the county scale, positive synergy patterns display distinct regional characteristics aligned with administrative boundaries. In major agricultural production areas, such as certain counties in the Yuncheng Basin and the Guanzhong Plain, synergistic relationships are particularly pronounced. Temporal changes in trade-off patterns are more evident at this scale: from 2010 to 2020, in some counties of Yulin (Shaanxi) and Shuozhou (Shanxi), where agricultural development intensity is high and water resources are severely constrained, trade-off areas expanded in a concentrated manner.

Grain production and soil conservation services

From 2000 to 2010, the relationship between grain production and soil conservation in the middle reaches of the Yellow River was dominated by positive synergies, while trade-off and negative-synergy areas were relatively scattered. This pattern indicates a strong degree of coordination between the two services during this period. However, from 2010 to 2020, trade-off and negative-synergy areas expanded markedly—particularly at the county scale—where negative synergies became more spatially concentrated and their extent increased. This shift reflects the growing disturbance of agricultural development, land-use change, and other human activities on the relationship between grain production and soil conservation (Fig. 7).

At the fine 5 km × 5 km scale, the spatial patterns were highly fragmented, with positive synergy, trade-off, and negative-synergy patches dispersed across the landscape. This fine-scale heterogeneity captures the influence of micro-topography and localized land-use practices on the interaction between the two services.

At the intermediate 10 km × 10 km scale, the spatial patterns became more integrated, revealing clearer regional regularities. For example, positive synergies were relatively concentrated in valley plains, whereas trade-off areas became more contiguous in hilly transition zones. This reflects the shaping effects of terrain gradients and regional cropping structures at the medium scale.

At the county scale, the spatial patterns exhibited pronounced macro-regional characteristics aligned with administrative boundaries. From 2000 to 2010, positive synergies were prevalent in major agricultural production areas, such as portions of the Fenwei Plain. From 2010 to 2020, however, trade-off and negative-synergy areas expanded rapidly in ecologically fragile and intensively cultivated counties—such as parts of northern Shaanxi and western Shanxi—providing a macro-level spatial basis for regional land management and agricultural planning.

Grain production and carbon sequestration services

From 2000 to 2010, positive synergies between grain production and carbon sequestration dominated across the middle reaches of the Yellow River, with extensive spatial coverage at all scales. During this period, the interaction between traditional farming practices and the carbon sequestration capacity of natural vegetation remained relatively stable, forming a benign coupling between carbon sink functions and agricultural production. Negative synergies appeared only sporadically in areas subjected to extremely intensive agricultural development, exerting minimal disturbance on the regional carbon–grain relationship (Fig. 8).

However, a marked shift occurred between 2010 and 2020. Positive synergy areas contracted significantly, and some regions gradually transitioned into trade-off or negative-synergy relationships—closely associated with increasing agricultural intensification and changes in land-use practices during this decade. Meanwhile, negative synergy areas expanded notably, especially at the county scale, where such areas emerged in clusters and continued to enlarge. This trend clearly reflects the growing disturbance of human agricultural activities on the carbon–grain relationship.

At the fine 5 km × 5 km scale, positive synergies appeared as fragmented patches, mostly confined to small farmland parcels with fertile soils, fine-scale cultivation management, or agroforestry composite systems. Synergy patterns at this scale were strongly affected by micro-topography, crop selection, and field-level carbon management practices, resulting in pronounced spatial heterogeneity. Negative synergy areas were similarly scattered, primarily distributed in steeply cultivated slopes and zones where cropland intermingles with built-up land, illustrating the micro-scale shaping effects of local land-use practices on the carbon–grain relationship.

At the intermediate 10 km × 10 km scale, the spatial continuity of positive synergies increased substantially, forming relatively contiguous synergy zones in major grain-producing areas such as the Fen River Plain and the Wei River Plain. This pattern highlights the close linkage between agricultural production layouts and vegetation carbon sink capacity at the medium scale. Trade-off areas formed relatively concentrated patches, typically located in transitional zones between dryland farming areas and forestlands. Their distribution was directly related to planting structure and agricultural input intensity at this scale, exhibiting distinct regional regularities.

At the county scale, positive synergy patterns displayed pronounced regional characteristics aligned with administrative boundaries. Strong synergies were particularly evident in key agricultural counties within the Yuncheng Basin and the Guanzhong Plain. Temporal changes in negative synergies were more striking: from 2010 to 2020, counties in parts of Yulin (Shaanxi) and Lüliang (Shanxi)—where agricultural development intensity was high and land-use transformation was rapid—experienced significant and concentrated expansion of negative synergy areas.

Water yield and soil conservation services

From 2000 to 2010, positive synergies between water yield and soil conservation overwhelmingly dominated the middle reaches of the Yellow River, with extensive spatial coverage across all scales. During this period, ecological restoration efforts promoted a benign interaction between hydrological regulation and soil retention, while negative synergies appeared only sporadically in a few ecologically fragile areas, exerting limited influence on the regional water–soil relationship. However, from 2010 to 2020, a notable shift occurred. Positive synergy areas contracted significantly, and some regions transitioned into trade-off or negative-synergy states, largely due to intensified human activities that increased water consumption and exacerbated soil disturbance. Meanwhile, negative synergy areas expanded considerably—especially at the county scale—where such areas emerged in clusters and continued to grow, clearly indicating an escalating disturbance of the water–soil relationship by human activities (Fig. 9).

At the fine 5 km × 5 km scale, positive synergies appeared as fragmented patches, often confined to micro-regions such as small catchments with good vegetation cover and localized terraced fields. Synergy patterns at this scale were strongly influenced by micro-topography and vegetation types, resulting in pronounced spatial heterogeneity. Negative synergy areas were likewise scattered and primarily concentrated in steeply cultivated zones and areas adjacent to built-up land, reflecting the fine-scale shaping effects of micro-geomorphology and land-use practices on the water–soil relationship.

At the intermediate 10 km × 10 km scale, the spatial continuity of positive synergies increased substantially, forming relatively contiguous synergy zones in valley plains and in medium-scale watersheds with notable vegetation restoration. This pattern highlights the close association among vegetation configuration, water conservation, and soil retention at the medium scale. Trade-off areas formed relatively concentrated patches, typically located in transition zones between agricultural land and forestland. Their distribution was directly related to land-use structure and the spatial arrangement of soil and water conservation measures at this scale, exhibiting clear regional regularities.

At the county scale, positive synergy patterns displayed distinct regional characteristics aligned with administrative boundaries. Counties with strong ecological governance outcomes demonstrated particularly pronounced synergies. Temporal changes in negative synergies were more striking: from 2010 to 2020, counties in the loess hilly and gully regions, as well as those with intensive agricultural development, experienced concentrated and significant expansion of negative synergy areas.

Water yield and carbon sequestration services

From 2000 to 2010, water yield and carbon sequestration were predominantly characterized by positive synergies, with broad spatial coverage across the study area. Trade-off and negative-synergy zones appeared only sporadically in a few localized regions, indicating that the overall water–carbon relationship during this period was largely synergistic. However, from 2010 to 2020, the pattern underwent a pronounced transformation: positive synergy areas contracted substantially, while trade-off areas expanded markedly and negative synergy areas also increased significantly. At the county scale in particular, clusters of negative synergy zones emerged and continued to expand, revealing the intensifying disturbance of the water–carbon relationship caused by human activities (Fig. 10).

At the fine 5 km × 5 km scale, positive synergies appeared as fragmented patches, mostly confined to micro-sites with relatively good vegetation cover. These patterns were strongly influenced by fine-scale land-use types and vegetation structure, exhibiting extremely high spatial heterogeneity. Trade-off and negative-synergy patches were also scattered, largely concentrated in transition zones between agricultural and ecological land. This reflects the micro-environmental control exerted by localized land-use configurations on the water–carbon relationship.

At the intermediate 10 km × 10 km scale, the spatial continuity of positive synergies increased significantly, forming relatively contiguous synergistic zones in valley plains and in areas with concentrated ecological restoration. This pattern indicates a close association between vegetation configuration, agricultural production, water resources, and carbon sequestration at the medium scale. Trade-off areas formed relatively coherent clusters, typically located in transition belts between croplands and forestlands. Their distribution was directly linked to planting structure and the spatial arrangement of ecological engineering projects at this scale, exhibiting clear regional regularities.

At the county scale, positive synergies displayed distinct regional characteristics aligned with administrative boundaries, and were particularly prominent in counties where ecological governance had achieved notable success and where agricultural development was well aligned with ecological objectives. Temporal variations in trade-off and negative-synergy patterns were even more pronounced: from 2010 to 2020, counties with high agricultural development intensity and severe water scarcity—such as portions of the loess hilly–gully region—experienced concentrated and substantial expansion of trade-off and negative-synergy areas.

Soil conservation and carbon sequestration services

From 2000 to 2010, soil conservation and carbon sequestration were predominantly characterized by positive synergies, with broad spatial coverage across the study area. Trade-off and negative-synergy zones appeared only sporadically in a few localized regions, indicating that the overall soil–carbon relationship during this period was largely synergistic. However, from 2010 to 2020, the pattern underwent a pronounced shift: positive synergy areas contracted substantially, while trade-off areas expanded markedly and negative synergy areas also increased significantly. At the county scale in particular, clusters of negative synergy zones emerged and continued to enlarge, highlighting the growing disturbance of human activities on the soil–carbon relationship (Fig. 11).

At the fine 5 km × 5 km scale, positive synergies appeared as fragmented patches, mostly confined to micro-sites with relatively good vegetation cover. These patterns were strongly influenced by micro-topography and vegetation structure, resulting in extremely high spatial heterogeneity. Trade-off and negative-synergy patches were similarly scattered, largely concentrated in transition zones where agricultural land intersects with ecological land. This reflects the fine-scale shaping effects of localized land-use practices on the soil–carbon relationship.

At the intermediate 10 km × 10 km scale, the spatial continuity of positive synergies increased markedly, forming relatively contiguous synergy zones in areas with concentrated ecological restoration and in valley plains. This pattern indicates a close association among vegetation configuration, soil and water conservation measures, and carbon sequestration at the medium scale. Trade-off areas formed relatively coherent clusters, typically located in transition belts between croplands and forestlands. Their distribution was directly related to planting structure and the spatial arrangement of ecological restoration projects at this scale, exhibiting clear regional regularities.

At the county scale, positive synergy patterns displayed distinct regional characteristics aligned with administrative boundaries, and were particularly prominent in counties where soil conservation and ecological carbon sequestration had been effectively coordinated. Temporal variations in trade-off and negative-synergy patterns were more pronounced: from 2010 to 2020, counties with high agricultural development intensity and drastic land-use transitions experienced concentrated and substantial expansion of trade-off and negative-synergy areas.

Accuracy of the results

The results of this study show that grain production, water yield, and soil conservation service in the middle reaches of the Yellow River all increased during 2000–2020, which is consistent with the findings of related studies²⁵. However, in the process of ecological restoration, trade-offs often emerge between grain production and other regulating services such as soil conservation, water retention, and carbon storage²⁶. Large-scale ecological restoration programs, such as the “Grain to Green” project, have notably improved vegetation cover, water retention, and soil conservation in the region²⁷; at the same time, they have reduced the land available for agriculture, thereby diminishing the material basis for grain production and causing declines in overall grain output^{28, 29}. Some studies, however, have suggested that such programs can simultaneously enhance multiple ecosystem services, with grain production, water retention, and soil conservation showing notable synergies at the plot scale³⁰. Moreover, the reduction in agricultural land has encouraged more intensive land use, improving the quality of the remaining cropland and raising unit grain yields³¹.

In this study, both static and dynamic approaches were employed to analyze trade-off and synergy relationships among ecosystem services. The static approach emphasizes the overall study area, using single-period data at the scale of the entire middle reaches of the Yellow River, and evaluates interrelationships based on the ecosystem service values of all assessment units within the region. In contrast, the dynamic approach focuses on individual assessment units and is based on temporal changes. Using 5 km × 5 km and 10 km × 10 km grids as well as county-level units, it characterizes dynamic trade-offs and synergies by computing the differences in ecosystem service values between two time periods.

The results obtained from the two approaches differ significantly, indicating that the selection of an appropriate method should be aligned with the specific evaluation objective. A combined application of both methods may also be adopted to obtain a more comprehensive understanding—minimizing trade-offs and enhancing synergies as much as possible—thereby supporting the achievement of sustainable development³².

In addition, the quantitative evaluation of carbon sequestration and water yield services in this study is based on remote sensing data such as NDVI, precipitation and land use, which is consistent with the remote sensing-based research framework of ecosystem carbon and water use efficiency adopted in global studies^{8, 9}. The slight decline in carbon storage and the continuous increase in water yield in the study area from 2000 to 2020 reflect the specific characteristics of carbon-water interaction in the semi-arid transitional zone of the Yellow River middle reaches: compared with the significant positive synergy between carbon uptake and water use efficiency in global semi-arid ecosystems⁸, the study area is affected by intensive human agricultural activities, and the carbon sequestration capacity of vegetation has not increased with the improvement of water yield capacity, which is similar to the research conclusion of Chakraborty et al.⁹, in India that human activities break the original carbon-water synergy. This also verifies the rationality of the study’s conclusion on the dominant driving effect of human activities on ecosystem service relationships, and the research results are highly consistent with the global research context of carbon-water cycle coupling.

To further verify the reliability of the assessment results, we conducted a quantitative validation of grain production, water yield, carbon storage, and soil conservation by combining county-level statistical data, hydrological observation records, published literature/inventory data, and soil erosion monitoring data. For grain production, the simulated county-level yield was compared with the statistical data from the China County Statistical Yearbook³³, showing a high linear correlation (R²=0.82, p < 0.001) with a root mean square error (RMSE) of 0.09 t/ha, which confirms the accuracy of the NDVI-based grain production estimation method in this study. For water yield, the annual simulated runoff of the main river basins was validated against the long-term hydrological observation data from key hydrological stations (Longmen, Sanmenxia) in the Yellow River basin³⁴, with a relative error (RE) of 12.3% within the acceptable range of ecological model simulation (± 15%). For carbon storage, the simulated results were cross-checked with the regional carbon storage inventories and published literature in the Loess Plateau^{12, 22}, and the average value difference was only 4.2 t/ha, with consistent spatial distribution characteristics of high carbon storage in mountain forest areas and low values in urban and arid grassland areas. For soil conservation, the simulated results were validated using the soil erosion monitoring data from the Loess Plateau Soil and Water Conservation Scientific Observation and Experiment Station^{3, 27}, showing a significant positive correlation (R²=0.76, p < 0.001) with an RMSE of 18.6 t/ha, which indicates that the RUSLE model effectively captures the spatial pattern and temporal variation of soil conservation capacity in the study area. The above validation results fully demonstrate the credibility of the ecosystem service quantification results in this study.

Uncertainty of model application and data support

This study quantifies core ecosystem services based on the InVEST and RUSLE models, and the reliability of results is inevitably affected by parameter sensitivity, structural model uncertainty and data resolution limitations, which are common challenges in model-based ecosystem service assessment^{11, 13}.

In terms of parameter sensitivity, the biophysical parameters of the InVEST and RUSLE models used in this study were modified based on previous studies and expert knowledge^{20, 21}, and the lack of localized calibration may lead to estimation deviations. For the InVEST water yield module, the plant-available water content (AWCx​) and seasonality factor (Z) are highly sensitive to simulation results¹⁹, and the unified parameter setting ignores the spatial heterogeneity of soil and climate in the study area. For the RUSLE model, the vegetation cover factor (Ci​) and rainfall erosivity factor (Ri​) are key sensitive parameters, and the simple linear correlation between Ci​ and NDVI fails to reflect the difference in soil conservation capacity of different vegetation types³. Single-factor sensitivity analysis shows that a 20% change in these key parameters will lead to a 5%~8% fluctuation in the simulation results of water yield and soil conservation.

Structural model uncertainty is derived from the simplification of complex natural ecological processes by the model itself¹⁸. The InVEST water yield module is constructed based on the water balance principle, which only considers the vertical hydrological process and ignores the horizontal connectivity such as groundwater recharge and watershed water exchange, leading to the underestimation of water yield in the Guanzhong Plain with abundant groundwater²². The carbon storage module of InVEST only calculates the static carbon pool storage and does not involve the dynamic carbon cycle process, which cannot reflect the real carbon sequestration capacity of the ecosystem. The RUSLE model is mainly applicable to the simulation of sheet erosion and rill erosion, and neglects the gully erosion which is dominant in the Loess Plateau, resulting in the underestimation of potential soil erosion in the study area²⁸.

Data resolution limitations and scale mismatch further increase the uncertainty of the assessment results¹². The study uses 30 m resolution land use and DEM data, which are aggregated to 5 km, 10 km and county scales for analysis, and the micro-topographic and land use heterogeneity information is lost during the aggregation process, leading to the smoothing of the spatial differentiation characteristics of ecosystem services. In addition, the meteorological and NDVI data with 1 km resolution cannot match the fine-scale land use data, and the resampling operation distorts the spatial pattern of driving factors⁴. Static soil type and river system data also fail to reflect the temporal changes of the underlying surface from 2000 to 2020, which affects the accuracy of dynamic simulation of ecosystem services.

It should be noted that although the above uncertainties exist, the overall spatiotemporal evolution characteristics of ecosystem services and the multi-scale law of trade-off/synergy relationships revealed in this study are consistent with the actual ecological and socio-economic development of the middle reaches of the Yellow River, and the research conclusions still have reliable scientific reference value.

Scale effects of ecosystem services and their interrelationships

Scale exerts a significant regulatory influence on the trade-off and synergy relationships among ecosystem services in the middle reaches of the Yellow River. Across different spatial scales (5 km × 5 km, 10 km × 10 km, and county level), the strength, significance, and spatial patterns of service interrelationships exhibit distinct variations, which are directly linked to the dominant influencing factors characteristic of each scale.

For some service pairs, the strength of synergies or trade-offs changes systematically with increasing scale. The synergy between soil conservation and carbon sequestration strengthens consistently as the scale expands, with correlation coefficients at the county scale significantly higher than those at the fine and medium scales. In contrast, the synergy strengths of grain production versus water yield and water yield versus soil conservation follow a “rise–fall” pattern, peaking at the medium scale (10 km). Other service combinations show limited scale sensitivity: for example, grain production and carbon sequestration maintain a stable trade-off relationship across all scales, with only minor fluctuations in correlation coefficients and no evident trend of increase or decrease.

The significance of service relationships is more stable at fine and medium scales. For all six service pairs, correlation coefficients at the 5 km and 10 km scales generally pass the 0.001 significance level in most years, being only slightly affected by interannual variation. At the county scale, however, the significance levels fluctuate more widely. Some service pairs show a transition from non-significant to significant relationships only at the county level—for instance, the trade-off between grain production and soil conservation becomes significant only in 2020, while the synergy between water yield and carbon sequestration begins to appear significantly from 2010 onward.

The fine 5 km scale exhibits highly fragmented patterns: synergistic, trade-off, and negative-synergy zones appear as scattered patches with strong spatial heterogeneity, predominantly controlled by micro-scale factors such as fine topography, field-level management, and local vegetation composition. The medium 10 km scale shows regional integration: synergistic or trade-off areas become more spatially continuous, forming coherent clusters shaped primarily by regional planting structures, water resource allocation schemes, and the spatial configuration of ecological engineering projects at the medium scale. At the county scale, macro-regional characteristics dominate: distinct regional differentiation emerges along administrative boundaries. Service relationships at this scale strongly reflect county-level agricultural development intensity, ecological governance policies, and broader land-use transformation trends, resulting in spatial patterns with pronounced macro-level directional characteristics.

From the perspective of global and interregional comparison, the scale effect of carbon-water related ecosystem service trade-offs and synergies in the study area is consistent with the universal characteristics of global terrestrial ecosystems. Kim et al.⁷, pointed out in the study of global grassland ecosystems that the trade-off between carbon sequestration and water yield is more significant at the regional administrative scale, while the synergy is more obvious at the fine natural grid scale, which is highly consistent with the research results of this study that the carbon sequestration-water yield synergy at the 5 km and 10 km grid scales is stable and significant, and the relationship at the county scale is more sensitive to human activities and shows obvious temporal variation. In addition, Zhang et al.⁸, found that the carbon water use efficiency of global semi-arid ecosystems has obvious scale dependence, and the regional scale can better reflect the overall characteristics of carbon-water interaction by reducing micro-topographic noise, which also provides a theoretical support for the study’s conclusion that the 10 km medium scale can better reveal the regional integration pattern of ecosystem service trade-offs and synergies.

The remote sensing-based carbon and water use efficiency research method has also been verified in this study: the NDVI data used in the grain production and carbon sequestration evaluation, and the precipitation and evapotranspiration remote sensing inversion data used in the water yield evaluation, can effectively characterize the spatiotemporal variation of carbon and water cycle components in the study area⁹. The results show that the areas with high carbon water use efficiency in the study area are mainly concentrated in the Qinling Mountains and other forest areas with good vegetation cover, while the areas with low carbon water use efficiency are concentrated in the Ordos Plateau and northern Loess Plateau with intensive human activities and fragile ecological environment, which is consistent with the spatial differentiation law of carbon water use efficiency in global semi-arid ecosystems⁸, and further confirms the accuracy of the study’s ecosystem service quantification results and the rationality of the driving mechanism analysis.

Optimization strategies for the ecosystem in the study area

Combined with the research results of global and other regional carbon-water interaction and ecosystem service trade-offs^{7, 8, 9}, significant spatial differences were observed in ecosystem services and their trade-offs and synergies between the southeastern and northwestern parts of the middle reaches of the Yellow River: (1) The southeastern region, characterized by mountainous terrain, dense populations, and more favorable ecological conditions, generally exhibited stronger ecosystem services; (2) By contrast, the northwestern region, constrained by an arid climate, limited precipitation, low forest cover, and intensive human activities, has long faced severe ecological problems such as soil erosion and land degradation, resulting in weaker ecosystem functions. For the ecological management of the study area, it is necessary to refer to the global experience of carbon-water synergy improvement, and formulate targeted optimization strategies based on the regional characteristics of the Yellow River middle reaches.

Significant spatial differences were observed ecosystem services and their trade-offs and synergies between the southeastern and northwestern parts of the middle reaches of the Yellow River: (1) The southeastern region, characterized by mountainous terrain, dense populations, and more favorable ecological conditions, generally exhibited stronger ecosystem services; (2) By contrast, the northwestern region, constrained by an arid climate, limited precipitation, low forest cover, and intensive human activities, has long faced severe ecological problems such as soil erosion and land degradation, resulting in weaker ecosystem functions.

These findings suggest that ecological restoration in the middle reaches of the Yellow River must fully consider spatial heterogeneity. Restoration and protection strategies should be planned scientifically, taking into account differences in vegetation type and distribution, to enhance the overall supply of ecosystem services. Greater attention should be given to fostering synergies among multiple ecosystem services and preventing the expansion of trade-off-dominated areas. Given that much of the region lies within the Loess Plateau, restoration planning should be tailored to local geographical and ecological conditions to explore pathways for the synergistic development of multiple services.

Furthermore, the middle reaches of the Yellow River face strong pressures from population growth, resource exploitation, and environmental stress, making human activities a major driver of ecosystem service dynamics. It is therefore critical to enhance public awareness of ecological protection, especially in ecologically fragile areas, with a focus on soil and water conservation. Strengthening ecological education and promoting sustainable land-use practices can help improve ecological resilience while balancing the needs of local communities.

Research innovations and theoretical contributions

This study breaks through the limitations of traditional ES trade-off/synergy research based on pure correlation analysis and single-scale description, and its innovations and theoretical contributions are mainly reflected in three aspects: conceptual insights, mechanism analysis and theoretical framework construction, which realize the transformation of ES trade-off/synergy research from “statistical description” to “mechanism analysis” and from “regional case mapping” to “theoretical generalization”.

New conceptual insights beyond traditional correlation-based methods

Traditional correlation-based methods only focus on the statistical co-variation relationship between ES, and define trade-off/synergy simply by the positive and negative of correlation coefficients, which is a phenomenal description of ES interactions³². This study proposes the concept of “static-dynamic integration of ES trade-off/synergy”, which holds that the ES interaction relationship is not a static statistical result, but a dynamic evolution process affected by ecological and socio-economic processes. Static correlation reveals the overall interaction intensity of ES in a certain period, while dynamic difference reflects the temporal evolution characteristics of ES interactions at the micro-unit scale. The combination of the two methods can avoid the one-sidedness of single static or dynamic analysis, and realize the comprehensive characterization of ES interaction from “overall to local” and from “state to process”. In addition, this study clarifies the scale threshold effect of ES trade-off/synergy, which is a new conceptual insight beyond the traditional scale dependence description: ES interactions at fine and medium scales are dominated by natural ecological factors with stable statistical significance, while ES interactions at the county scale are affected by macro-policy and socio-economic factors with significant temporal variation of statistical significance, and some ES pairs will transform from non-significant to significant with the expansion of scale (e.g., grain production and soil conservation, water yield and carbon sequestration). This concept of scale threshold effect enriches the connotation of ES scale effect theory.

In-depth analysis of causal driving mechanism and emergent system behavior of ES interactions

This study goes beyond the pure statistical analysis of traditional correlation methods, and deeply reveals the hierarchical causal driving mechanism of ES trade-off/synergy spatial pattern through the combination of multi-scale analysis and factor identification: fine-scale (5 km) ES interaction pattern is dominated by micro natural factors (topographic relief, soil type, local vegetation cover), and the spatial pattern presents high fragmentation due to the heterogeneity of micro-geomorphic conditions; meso-scale (10 km) ES interaction pattern is jointly driven by regional ecological and agricultural configuration (planting structure, water resource allocation, ecological engineering layout), and the spatial pattern shows regional integration with the reduction of micro-noise; county-scale ES interaction pattern is dominated by macro socio-economic and policy factors (agricultural development intensity, ecological restoration policy, GDP density), and the spatial pattern presents obvious administrative differentiation and temporal evolution characteristics.

On this basis, this study further clarifies the emergent system behavior of ES interactions in the MRYR: with the expansion of spatial scale, the interaction intensity of partial ES pairs presents non-linear emergent characteristics that cannot be simply superposed by fine-scale results, such as the synergy intensity of soil conservation and carbon sequestration increases continuously with scale expansion, and the synergy intensity of grain production and water yield presents a “rise-fall” humped pattern peaking at the meso-scale. This emergent system behavior is the result of the cross-scale coupling of natural ecological processes and human socio-economic activities, which breaks the traditional understanding that “ES interaction intensity is a fixed value independent of scale”, and deepens the understanding of the systematic characteristics of ES interactions.

Construction of a universal multi-scale coupling analysis framework for ES trade-off/synergy

Aiming at the limitations of single-scale, pure statistical and case-based research in existing ES trade-off/synergy studies, this study constructs a multi-scale coupling analysis framework of ES trade-off/synergy that integrates static correlation analysis, dynamic difference comparison, hierarchical driving mechanism identification and emergent system behavior analysis. The framework takes multi-scale spatial units as the analysis carrier, takes the static-dynamic integration of ES trade-off/synergy as the core characterization, takes the hierarchical driving mechanism as the deep analysis, and takes the emergent system behavior as the theoretical generalization, forming a complete research chain from “pattern description” to “mechanism analysis” and then to “theoretical construction”.

This framework breaks the single-scale and pure statistical analysis paradigm of traditional ES research, and has the following universal theoretical value: (1) It is applicable to ES trade-off/synergy research in different types of regions (ecologically fragile areas, urban agglomerations, agricultural areas, etc.), and can realize the comprehensive characterization of ES interactions by adjusting the selection of spatial scales and driving factors; (2) It connects the micro natural factors, meso regional configuration and macro policy factors of ES interactions, and clarifies the cross-scale coupling relationship of ES driving factors, which provides a theoretical basis for the formulation of scale-adaptive ES coordinated regulation strategies; (3) It realizes the transformation of ES trade-off/synergy research from regional case mapping to theoretical system construction, and enriches the ES scale effect theory and system theory, which is an important supplement to the existing ES research system.

In study, four key ecosystem services—grain production, water yield, carbon sequestration, and soil conservation—were selected to quantify and evaluate the major ecosystem functions in the middle reaches of the Yellow River. Both static and dynamic methods were employed to examine trade-offs and synergies among these services. The primary conclusions are as follows:

1. (1)

   Spatiotemporal patterns of ecosystem services

Over the two decades, grain production, water yield, and soil conservation services generally increased, whereas carbon storage remained relatively stable with a slight decline. Spatially, grain production displayed a southeast–northwest gradient, while water yield showed a distinct south–north differentiation. High levels of soil conservation were concentrated in mountainous forest regions, including the Qinling and Lüliang Mountains, while low values persisted in the ecologically fragile Ordos Plateau and northern Loess Plateau. Carbon storage was primarily constrained by intensive land use and sparse vegetation cover in cropland and grassland areas.

1. (2)

   Static trade-offs and synergies: temporal and scale dynamics

The six ES pairs exhibited clear temporal evolution and scale dependence. Temporally, synergies weakened over time for grain production–water yield, water yield–soil conservation, and water yield–carbon sequestration. In contrast, the synergy between soil conservation and carbon sequestration strengthened, while trade-offs for grain production–carbon sequestration remained stable.

Across scales, grain production–water yield and water yield–soil conservation followed a humped pattern, peaking at the medium scale. Soil conservation–carbon sequestration showed increasingly strong synergies at coarser scales. Notably, grain production–soil conservation and water yield–carbon sequestration demonstrated scale threshold effects, transitioning from non-significant or weak relationships at fine scales to significant trade-offs or synergies at the county scale in the later period.

1. (3)

   Dynamic trade-offs and synergies: a two-stage shift

A striking two-stage differentiation characterized the dynamic ES relationships. The period 2000–2010 was dominated by positive synergies across all scales, driven by effective ecological restoration and moderate human activity. However, from 2010 to 2020, intensified agricultural activity and increased water consumption led to a contraction of synergistic areas and a substantial expansion of trade-off and negative synergy zones. This shift was most evident at the county scale, where negative interactions became increasingly concentrated.

1. (4)

   Dominant role of scale effects

Dominant role of scale effects and theoretical contribution of multi-scale coupling framework. Scale played a pivotal role in regulating the spatial pattern and temporal evolution of ES relationships, and the ES interaction pattern at different scales is driven by hierarchical factors with typical scale threshold effects. Fine scales (5 km) were characterized by high fragmentation, shaped primarily by microtopography and local management practices; medium scales (10 km) exhibited regional integration, reflecting the influence of planting structures and ecological engineering layouts; at the county scale, macro-level processes—such as agricultural development intensity and policy implementation—dictated the patterns, which showed the most pronounced temporal changes and clear administrative differentiation.

Based on the above findings, this study constructs a universal multi-scale coupling analysis framework of ES trade-off/synergy, which breaks the limitations of traditional pure correlation analysis and single-scale research, and clarifies the emergent system behavior of ES interactions with scale expansion. This framework enriches the scale effect theory and system theory of ecosystem services, and provides a new theoretical perspective and methodological reference for the coordinated regulation of ES in ecologically fragile areas with intense human-environment conflicts worldwide. For the Yellow River Basin, the research results can provide scientific support for the formulation of scale-adaptive ecological protection and high-quality development policies, and realize the balanced supply of multiple ES under the background of ecological restoration and agricultural development.