Introduction

Ecosystem services (ESs) encompass a spectrum of natural environmental conditions and services that ecosystems generate and sustain, essential for human well-being (Costanza et al. 1997; Jia et al. 2021; Loomes and O’Neill 1997). Since Costanza et al. introduced the concept of ecosystem service (ES) in 1997, various types and values of ESs have been generally recognized and studied across different scales, serving as a bridge between ecosystems and socio-economic systems (Loomes and O’Neill 1997; Wu 2013). ESs comprise provisioning, supporting, regulating, and cultural services (MA 2005). In recent years, approximately 60% of ESs globally have suffered severe degradation due to intense anthropogenic disturbances and specific demands on ES types (Deng et al. 2022b), posing a serious threat to ecosystem security. In this context, adopting reasonable management measures becomes crucial to regulate and optimize ESs overall, thereby improving ecosystem sustainability. However, various ES types are interconnected through complex and non-linear relationships. Enhancing one ES often involves trade-offs with other ESs, presenting challenges in achieving both service diversification and maximizing overall benefits (Gou et al. 2021). Ecosystem management should aim not only for a single ES benefit but also to balance and harmonize multiple ESs to maximize their combined benefits. Broadly speaking, trade-offs can involve non-uniform, one-directional changes in rates, as well as shifts in the nature of one aspect compared to another (Lu et al. 2014). Their impact is often more significant than synergistic effects (Howe et al. 2014), highlighting the importance of focusing on ES trade-offs to maintain ecological balance and avoid unreasonable trade-offs (Qiang et al. 2017). While ES trade-offs have long been recognized, managing multiple ESs simultaneously in complex social-ecological systems while minimizing adverse effects is one of the most critical challenges in sustainability research (Wu 2013, 2017, 2021).

To enhance regional landscape sustainability, numerous studies have investigated ES trade-offs using various methods, scales, and perspectives (Cord et al. 2017a; Dade et al. 2019; Dou et al. 2020; Huang and Wu 2023). Currently, research on ES trade-offs predominantly relies on qualitative analysis, such as spatial mapping (Bennett et al. 2009; Carreo et al. 2012; Jia et al. 2014; Paul et al. 2005). However, these methods offer discrete insights, limiting a deeper understanding of ES relationships and their impact mechanisms (Liu et al. 2022). In contrast, the root mean squared error (RMSD) method offers a practical approach to quantifying ES trade-offs and enables quantitative assessments (Bojie and Dandan 2016). Regarding the scale of investigation, most existing ES studies have been conducted at the municipal administrative unit level due to data accessibility (Quintas-Soriano et al. 2019; Rla et al. 2019; Spake et al. 2017). Conversely, examining the watershed scale, a naturally defined geographic unit presents distinct advantages in addressing the misalignment between human management and ecological processes scales. Furthermore, the watershed scale allows for revealing spatial and temporal characteristics of ES trade-offs, aiding in making specific spatially targeted decisions for managing ES trade-offs.

Currently, research applied at the watershed scale remains insufficient (Zhao et al. 2018). From a research standpoint, the focus of ES trade-offs primarily centers on a static perspective, analyzing spatial patterns and driving factors at specific times to offer recommendations for ecosystem management (Hd et al. 2020). It is important to note that ES trade-offs may vary over time (Deng et al. 2022b). To foster ecosystem sustainability, we must analyze the trade-offs across multiple periods of ESs while also identifying the shifts in spatial and temporal variations among these relationships. This approach enables the development of strategies conducive to preserving ecosystems (Wu 2013, 2021). Studies examining the spatial and temporal dynamics of ES relationships and their influencing factors are still relatively scarce.

ES trade-offs are typically influenced by multiple factors, such as climate (Dai et al. 2017), socioeconomic factors (Liu et al. 2019), and land use change (Vanniera et al. 2019). However, solely focusing on a single factor makes it difficult to unveil the reasons for heterogeneous changes in ES trade-offs (Qiu et al. 2020) and can introduce uncertainty into ecosystem management decisions. Only an integrated analysis that considers the impacts of both human activities and climate change on ES trade-offs can clarify the influencing mechanisms of ES relationships and promote the sustainability of regional ecosystems (Hong et al., 2020).

The Huang-Huai-Hai Plain (HHHP) represents a typical agricultural region in China but grapples with notable ecological and environmental challenges, characterized by heightened concentrations and intensification of issues. The swift pace of urbanization exacerbates these pressures, resulting in tensions over land use and imbalances in the provisioning of ESs, notably leading to water shortages (Deng et al. 2022a). Such issues arise from the interplay and conflicts among various ESs, driven by diverse factors. Understanding the heterogeneous changes in ES trade-offs and their determinants within the HHHP is crucial for fostering coordinated and sustainable development, vital for both natural ecology and food security. This knowledge holds substantial theoretical and practical implications. Hence, the objectives of this study are threefold: to evaluate four key ESs—food provision (FP), carbon sequestration (CS), soil conservation (SC), and water yield (WY) —in the HHHP during 2000, 2010, and 2020; quantify the strengths of ES trade-offs; and investigate the spatiotemporal dynamics of trade-off strengths among ESs, aiming to establish a scientific foundation for optimizing ESs and enhancing regional landscape sustainability.

Data and methods

Study area

The HHHP comprises 246 sub-basins, spanning the cities of Beijing and Tianjin and the five provinces of Shandong, Hebei, Jiangsu, Anhui, and Henan (Fig. 1), covering an area of about 390,000 km2 (112°43’-112°71’E, 32°49’-40°57’N). The region experiences an average annual precipitation of around 700 mm and a temperature of approximately 13.9 °C. Its mid-latitude monsoon climate results in uneven spatial and temporal distribution of precipitation, with about 70% of the annual rainfall concentrated in summer (Wang et al. 2020). The HHHP is characterized by low and flat terrain, with most areas below 50 m in elevation, representing a typical alluvial plain and serving as China’s largest agricultural area (Mw et al. 2021). Over the past two decades, rapid urbanization has significantly transformed the HHHP. Between 2000 and 2020, there was a noticeable increase in land occupied by built-up areas, rising from 13.14 to 18.13%, consequently exerting mounting pressure on the ecological environment. The conflict between promoting regional socio-economic development and ensuring ecological environmental conservation has emerged as a significant constraint on the HHHP’s sustainable development, leading to adverse consequences for the local community welfare (Deng et al. 2022b).

Fig. 1
figure 1

Geographic location of the study site

Data sources

The data utilized in this study primarily encompass sub-basin data (from the Global Drainage Basin Database), remote sensing data, topographic surveys, soil property records, meteorological datasets, land use data records (for 2000, 2010, and 2020), and socio-economic information (such as Gross Domestic Product (GDP), population figures, and Nighttime Light Data (NLD) pertinent to the HHHP. NLD proves valuable for characterizing urbanization levels due to its strong correlation with human economic activities. For detailed information regarding all data sources and descriptions, refer to (Deng et al. 2022a, b). Through the application of interpolation and resampling techniques in ArcGIS 10.6, the data was standardized into a consistent 250 × 250 m grid. Drawing from established landscape studies (Chen et al. 2021; Jia et al. 2021; Zhang et al. 2021b), this study opted for Patch Density (PD), Shannon’s Diversity Index (SHDI), Landscape Shape Index (LSI), and Contagion (CONTAG) to characterize the landscape pattern within the HHHP across four dimensions: density, diversity, shape, and aggregation, respectively. Refer to Table S1 for the calculation methodologies.

Modelling the spatiotemporal dynamics of ESs

In the HHHP, rapidly expanding construction land has encroached on the surrounding ecological space, resulting in the imbalance of regional ecosystem functions. According to the regional characteristics of the HHHP, four key ESs of FP, CS, SC, and WY were selected, which could be calculated with the following methods (Table 1). In this study, we verified the two ESs of CS (above-ground part) and WY based on the measured data collected from the literature (Li Xu et al. 2018) to ensure the accuracy of the simulation results. Since the calculation process of FP has already used statistical data on grain production, no additional verification is needed. Due to the lack of observational data on soil conservation, this study did not verify SC and CS (below-ground part).

Table 1 Methods and main references for evaluating ESs

Calculation of the ES trade-offs

Within the scope of this study, we employed the RMSD index to evaluate the trade-offs (FP_SC, FP_CS, CS_WY, CS_SC, FP_WY, and WY_SC) in the HHHP during the time periods of 2000, 2010, and 2020. Furthermore, we utilized this index to conduct a quantitative analysis of the spatial characteristics of these ES trade-offs (Fig. 2). The RMSD, being a convenient and effective measure, facilitated a quantitative evaluation of trade-off dynamics within ESs (Bradford et al. 2012). It extended the interpretation of trade-offs beyond a negative correlation (as conventionally perceived) to encompass to the rate of inhomogeneity of variation between ESs in the same direction. This extension allowed for a more precise assessment of the extent of interaction among ESs (Bradford et al. 2012). A larger RMSD value indicated stronger trade-off strength, while a smaller value suggested weaker trade-off strength.

When the RMSD value is 0, there is a synergistic relationship between the two ESs. To eliminate the effect of the magnitude, data normalization is required before calculating RMSD. Bradford et al. 2012 defined ES normalization as:

$${ ES}_{std}=\frac{({ES}_{sim}-{ES}_{min})}{({ES}_{max}-{ES}_{min})},$$
(1)

Where\({ ES}_{std}\) presents the normalized ES, \({ES}_{sim}\) denotes the simulated ES, \({ES}_{min}\) and \({ES}_{max}\) signifiy the minimum and maximum values of ES, respectively. The RMSD value is as follows:

$$\text{R}\text{M}\text{S}\text{D}=\sqrt{\frac{1}{n-1}\times \sum _{i=1}^{n}{({ES}_{i}-\widehat{ES})}^{2}},$$
(2)

Here, ESi represents the ES standard value corresponding to i, and \(\widehat{ES}\) is the average value.

Fig. 2
figure 2

(modified from (Fu and Yu 2016). Both the blue and red dots in the figure represents coordinate points of paired ESs. The degree of ES trade-offs increases with distance from the 1:1 line, where there are no trade-offs. The magnitude order of the trade-offs is C = D > A > B. Above the 1:1 line is the dominant area for ES-1, while below the 1:1 line is the dominant area for ES-2. For example, point C indicates that ES-1 is the gaining side and point D indicates that ES-2 is the gaining side

Schematic diagram of the trade-off relationship strengths among ESs

Identifying key drivers of ES trade-offs

Automatic linear modelling (ALM)

Considering the natural characteristics and socio-economic conditions of the HHHP, and drawing from prior studies (Deng et al. 2022b), this study identified several potential driving factors (refer to Table 2) and utilized the ALM model (Zhang et al. 2021a) to analyze the primary factors influencing the ES trade-offs. Through the removal of irrelevant variables and the management of multicollinearity among factors, the ALM model produced optimal regression outcomes. The model was expressed as follows:

$$Y = \alpha _{0} + \alpha _{1} X_{1} + \alpha _{2} X_{2} + \cdots + \alpha _{i} X_{I} + \delta ,$$
(3)

Where Y represents RMSD, Xi represents the potential driving factor, \({\alpha }_{i}\) signifies the model coefficient; \({\alpha }_{0}\) denotes the intercept, and δ indicates the error term.

Table 2 Description of potential driving factors investigated in this study

Spatial autocorrelation of ES trade-offs at the sub-watershed level

Moran’s I index can provide insights into the spatial characteristics of trade-off relationship strengths among ESs (Diniz-Filho et al. 2003; Ren et al. 2020). A positive Moran’s I index value indicates a spatially correlated distribution pattern of trade-off strengths among ESs. As the value increases, the spatial correlation becomes more pronounced, suggesting a spatial aggregation effect within the study area. Conversely, a decrease in the value implies spatial dispersion characteristics in the study area. A Moran’s I index value of 0 indicates a random spatial distribution, suggesting a random pattern in the distribution of trade-off strengths. The significance level was determined based on the Z value. When |Z| > 1.96 with a P-value < 0.05, the Moran’s I index was considered significant. When |Z| > 2.58 with a P-value < 0.01, it was considered highly significant.

Geographically weighted regression (GWR) analysis

By mitigating the issue of multicollinearity among variables, the ALM model facilitated the selection of essential influencing factors as explanatory variables in the GWR model. Unlike traditional regression models, GWR accommodates the non-smoothness of geographic space. Its primary advantage lay in assigning distinct regression coefficients to each study unit, thereby capturing heterogenous drivers more effectively (Yu et al. 2020; Zhu et al. 2020). The GWR analysis was performed using the equations within the ArcGIS 10.6 tool as follows:

$${Y}_{i}={\beta }_{0}\left({u}_{i},{v}_{i}\right){+\sum _{k=1}^{i}{\beta }_{k}({u}_{i},{v}_{i})X}_{ik}+\tau$$
(4)

Where \({Y}_{i}\) represents the RMSD of sub-basin i; \({X}_{ik}\) denotes the value of the kth explanatory variable on sub-basin i; \(\left({u}_{i},{v}_{i}\right)\) denotes the coordinates of basin i; \({\beta }_{0}\left({u}_{i},{v}_{i}\right)\) signifies the spatial intercept at sub-basin i; \({\beta }_{k}({u}_{i},{v}_{i})\) represents the regression coefficient of the kth explanatory variable on basin i, locally estimated using weighted least squares; and \(\tau\) indicates the residual.

Results

Spatio-temporal patterns of ESs in the Huang-Huai-Hai plain

In this study, we verified the simulated and measured ESs, and the fitting R2 reached more than 0.65, which indicated the simulation results were credible (Fig. S1 and Fig. S2). Between 2000 and 2020, the four ESs in the HHHP demonstrated relative temporal stability with no significant overall changes. However, discernible spatial heterogeneity was evident, indicating diverse patterns across different areas (Fig. 3). The spatial distribution of FP exhibited a gradual decrease from the southern to the northern regions within the HHHP. High-value regions for FP were predominantly clustered in the southwest, akin to WY. The regional mean value surged from 276.95 mm in 2000 to 381.77 mm in 2010, followed by a decline to 350.66 mm in 2020, indicating noticeable temporal fluctuations. The distribution pattern of CS and SC consistently showcased higher values in mountainous areas and lower values in plain areas, reflecting the area’s spatial characteristic. Specifically, the central and eastern mountainous regions within the HHHP exhibited a high-value distribution of SC, often exceeding 200 t ha−1. In contrast, the surrounding plain areas displayed a low-value distribution of SC, generally below 30 t ha−1. The overall spatial distribution characteristics of CS did not exhibit significant changes. Generally, SC demonstrated an increasing trend from 14.18 t ha−1 in 2000 to 23.16 t ha−1 in 2020, while the change trend in CS was less pronounced.

Fig. 3
figure 3

Spatial patterns of FP, SC, CS, and WY in 2000, 2010 and 2020

Trade-offs for paired ESs

From 2000 to 2020, the trade-off intensity between FP_SC remained relatively consistent across space (Fig. 4). The concentrated area of strong trade-offs, with RMSD value surpassing 0.4, was primarily situated in the southwestern region. Moreover, a noticeable spatial gradient in the trade-off intensity was observed, gradually increasing from the northeast to the southwest. Overall, the majority of regions were FP-dominated, with fewer regions dominated by SC and a lower degree of trade-offs in these areas. Over this period, a declining trend in the intensity of the FP_SC trade-off within the HHHP was observed, dropping from 0.381 to 0.348. Conversely, for FP_CS, significant changes were noted in the spatial pattern of trade-off strengths from 2000 to 2020. In 2000, the southwest region predominantly exhibited FP dominance, while CS held a dominant position in the northeast. By 2020, the strong trade-off area shifted towards the northern part of the HHHP, and the FP-dominated area in the southwest decreased. In terms of temporal changes, the intensity of the FP_CS trade-off increased (0.112 in 2000, 0.170 in 2010, and 0.187 in 2020). By 2010 and 2020, the scope of the strong trade-off area was further reduced, and most areas exhibited weak trade-offs (RMSD < 0.3). Areas dominated by WY gradually increased and concentrated in the eastern part of the HHHP. The trade-off intensity between CS_SC was notably high (RMSD > 0.4) across most HHHP areas. CS prevailed in most regions, particularly in 2020, where CS-dominated areas encompassed the entire HHHP. The southern portion primarily displayed weak trade-off areas in WY_CS, with WY as the dominant factor. However, by 2020, some strong trade-off areas (RMSD > 0.4) transitioned into moderate trade-off areas (RMSD < 0.4), resulting in a substantial decrease in WY-dominated areas and a decrease in trade-off intensity from 0.257 in 2000 to 0.240 in 2020.

Fig. 4
figure 4

Spatial distribution of ES trade-off intensity in 2000, 2010 and 2020. The grid coverage area was a certain ES dominant area

The scatter plot displays distinct variations in trade-off relationships and degrees between paired ESs (Fig. 5). Notably, these trade-offs exhibit greater strength and variability (Fig. 6), with FP showing a larger relative gain. In 2000, the FP_CS trade-off points were mainly dispersed on both sides of the 1:1 line in the scatter plot, indicating a concentrated distribution and lower trade-off degrees. However, in 2010 and 2020, most FP_CS points were positioned above the 1:1 line, highlighting increased relative gains for CS. For FP_WY, the trade-off points were primarily located in the lower section of the scatter plot in 2000, suggesting a higher relative gain for FP. However, in 2010 and 2020, the trade-offs were predominantly aligned along the 1:1 line, lacking a discernible trend. The CS_SC trade-off points notably deviated from the 1:1 line (Fig. 5), indicating a more concentrated range in trade-off intensity, higher overall trade-offs (Fig. 6), and a larger relative gain for CS.

Regarding WY versus SC, in 2000, points were dispersed on both sides of the 1:1 line, suggesting lower trade-off levels. However, in 2010 and 2020, the majority of scatter points fell below the 1:1 line, indicating WY’s dominance in the WY_SC trade-off. Conversely, for WY_CS, trade-off points primarily clustered in the upper section of the scatter plot, signifying a higher relative gain for CS.

Fig. 5
figure 5

Scatter plot matrices of paired ESs (standardized) in 2000 (a), 2010 (b), and 2020 (c)

Fig. 6
figure 6

The error bar of ES trade-off intensity among ESs

Key driving factors of trade-offs among ESs

Using the automatic linear model, this study analyzed the factors influencing the trade-off strengths among ESs. In 2000, the trade-off relationship strength between FP_SC was primarily dominated by the forested area, which accounted for up to 45.7% of the contribution. The proportion of grassland area and NDVI were the additional factors following the forested area (Table 3). Conversely, NDVI became the dominant factor in 2010 and 2020. The relationship between FP_WY was primarily shaped by precipitation, which played a significant role with a contribution of 52.8% in 2000 and 33.7% in 2010. However, in 2020, the influence of PPT became negligible. The trade-off strength between WY_SC was primarily governed by PPT, which played a pivotal role with contributions of 40.7% in 2000, 65.9% in 2010, and 76.6% in 2020, respectively.

Table 3 Effects of driving factors on ES trade-offs

Analyzing the spatial heterogeneity of ES trade-off drivers

Spatial autocorrelation of trade-offs among ESs

At the watershed scale, the trade-offs between pairs of ESs exhibited a spatial clustering trend and demonstrated a positive spatial correlation (Table 4). Between 2000 and 2020, the spatial aggregation of the trade-offs between FP_SC increased noticeably. Moran’s I values rose from 0.48 in 2000 to 0.72 in 2020. Similarly, the FP_CS relationship showed an escalating trend in spatial aggregation. In contrast, the trade-off relationships involving WY_SC and WY_CS consistently displayed highly aggregated spatial patterns, with Moran’s I values consistently exceeding 0.7.

Table 4 Moran’s I and Z-value of RMSD

Spatial patterns of the key drivers for ES trade-offs

Based on the spatial regression findings, the GWR model was deemed more effective in capturing the variability of factors (Fig. 7). FP_SC was chiefly influenced by NDVI, emerging as the dominant factor shaping this relationship. Between 2000 and 2020, its dominant area proportions were 89.51%, 70.52%, and 76.33%, respectively. Over time, there was a shift in the predominant area, with temperature gradually assuming the most significant role, transitioning from the southwest to the southeast.

Fig. 7
figure 7

The localized R2 for paired ESs

In 2000, the FP_CS relationship was predominantly influenced by the positive effect of LSI (54.57%), followed by the positive impact of TEM (17.51%) and the negative effect of arable land area (15.52%). By 2010, a significant portion of the southern part of the HHHP (63.28%) was primarily affected by the arable land area. However, by 2020, the arable land’s role was predominantly negative (7.98%) and limited to a small part of the southwest. In 2010, the primary influencing factor in the northern part accounted for 33.85% and was represented by NDVI. By 2020, the area significantly influenced by NDVI expanded to 51.77%. Notably, the positive influence of NDVI was mainly observed in the eastern region, while the western region experienced a negative impact.

In the year 2000, precipitation significantly influenced the interaction between CS and SC within the central and northern HHHP. It constituted the most substantial proportion, encompassing 48.93% of the total area, followed by NDVI (30.72%) and the percentage of grassland area (14.46%). By 2010, the dominant factors in the trade-off relationship gradually shifted to PD (53.90%) and forest area (45.19%), and by 2020, the entire region was entirely influenced by PD. Concerning the trade-off association between WY_SC, regions predominantly influenced by PPT constituted 92.62% (2000), 83.17% (2010), and 100% (2020) of the total area, respectively. Moreover, in 2000, approximately 7.37% of the region was predominantly characterized by construction land, primarily located in the eastern sector. By 2010, around 10.04% of the area exhibited significant influence from NDVI, primarily concentrated in the eastern region of Shandong Province, as depicted in Fig. 8.

Fig. 8
figure 8

Spatial distribution of dominant factors of trade-offs among ESs in 2000, 2010 and 2020. The size of these points represents the degree of drivers’ influence on the ES trade-offs, with larger points representing greater influence and vice versa

Discussion

Spatiotemporal heterogeneity of trade-off intensity among ESs

In the context of escalating natural resource scarcity, the enhancement of one ES supply often comes at the expense of other services, leading to trade-offs between ESs (Cord et al. 2017b; Li et al. 2021; Zeng et al. 2020). Clarifying ES trade-offs is a fundamental prerequisite for effective ES management. However, current assessments of ES trade-offs lack spatiotemporal heterogeneity analysis (Kragt and Robertson 2014; Sun et al. 2019). These assessments concentrate on assessing ES interactions by treating the entire region as a uniform entity, thus failing to account for spatial heterogeneity (Groot et al. 2007). Conversely, some analyses utilize multi-period data and examine the spatial heterogeneity of trade-off relationships but overlook the dynamics of these relationships over time (Huang and Wu 2023; Jopke et al. 2015; Li and Luo 2023). Thus, this study employs the RMSD approach to quantify the spatiotemporal dynamics of trade-offs among ESs.

In the temporal aspect, among the six pairs of ES trade-offs, the strength of the trade-off relationship between CS_SC exhibited the most notable change (Figs. 4 and 6), displaying the highest temporal variability. The average value of the RMSD ranged from under 0.4 in 2000 to over 0.5 in 2020. This change was primarily due to a significant upward trend in soil retention from 2000 to 2020 (Fig. 3). This trend suggests an enhancement in crop growth and the soil and water conservation capacity in the HHHP region. However, studies by Wen et al. (2022) and Huang et al. (2019) have demonstrated that agricultural areas are primarily associated with carbon emissions, where improved crop growth tends to result in higher carbon emission, leading to a decline in soil carbon. To achieve a mutually beneficial scenario between SC_CS and to bolster regional landscape sustainability, actions need to be taken to amplify carbon sequestration while minimizing agricultural carbon emissions. For instance, employing practices such as conservation tillage and straw-returning can improve the carbon sequestration potential of agricultural land (Maraseni and Cockfield 2011).

The spatial distribution patterns of trade-offs among the three pairs of ESs, CS_SC, WY_SC, and WY_CS, exhibited distinct characteristics (Fig. 4). CS_SC presented a trend of lower trade-offs in the central mountainous region and higher trade-offs in the surrounding area. The weaker CS_SC trade-offs in the central mountainous region with more pronounced topography could be attributed to the association between higher levels of CS and enhanced vegetation growth, subsequently leading to increased soil retention capacity (Qiao et al. 2019). The degree of WY_SC trade-offs gradually decreased from the south to the north (Fig. 4). This decline might be attributed to higher precipitation in the southern region, surpassing the vegetation’s soil retention capacity, resulting in a higher WY_SC trade-off. As precipitation reduced further northward, the probability of soil erosion decreased, resulting in a lower WY_SC trade-off pattern. Conversely, WY_CS exhibited an opposite trend, showcasing a gradual increase from south to north. In the dry and water-scarce HHHP, increased precipitation in the south correlated with improved crop growth conditions and higher carbon emissions. Consequently, this led to decreased soil carbon sequestration, represented by a lower WY_CS trade-off. In contrast, the relatively dry northern region experienced limited crop growth due to drought, resulting in lower agricultural carbon emissions, thus presenting a higher WY_CS trade-off.

Effects of influencing factors on dynamic change of trade-off intensity among ESs

Examining the mechanism influencing trade-off relationships and determining the significance of each influencing factor could enhance our comprehension of methods and strategies to regulate this relationship. This understanding is crucial for formulating “win-win” policies that uphold ecological conservation and food security (Gou et al. 2021). With this goal, the current study employed quantitative analysis to investigate how natural factors and human activities impacted the spatial and temporal dynamics of trade-offs among ESs. The objective was to uncover the underlying mechanism driving these variations. The findings from our research suggest that precipitation negatively impacted the trade-offs associated with FP_WY, WY_CS, and CS_SC while exerting a positive influence on WY_SC. According to our study, precipitation negatively impacted the trade-off relationships among WY_CS, FP_WY, and CS_SC but had a positive effect on the trade-off relationship involving WY_SC. This finding is based on the definition of WY, which represents the difference between rainfall and evapotranspiration. Therefore, increased precipitation would have a positive impact on WY. The HHHP region features a mid-latitude monsoon climate, making it vulnerable to drought risks. Wang et al. (2017) identified precipitation as the primary factor limiting crop yields in this area. Hence, an increase in precipitation would contribute to the promotion of FP.

In 2000, most northern areas of HHHP experienced precipitation levels below 600 mm, where increased precipitation benefited vegetation growth and enhanced soil retention capacity. However, further increments in precipitation alongside intensified human activities might escalate soil erosion, thus diminishing soil retention. Between 2010 and 2020, NDVI predominantly exhibited adverse effects on the trade-off relationship of FP_WY and FP_CS, excluding FP_CS in 2010. Several factors contribute to this observation. On the one hand, precipitation significantly influences vegetation growth in HHHP, where higher NDVI implies increased precipitation, elevating FP and WY while reducing the trade-off magnitude. Conversely, regions with higher NDVI values tend to boast robust vegetation growth and superior vegetation carbon sequestration, thereby reducing the intensity of the FP_CS trade-off. The HHHP region consists predominantly of arable land. Acknowledging water’s significant impact on trade-offs among ESs, there is a necessity for substantial development in local agricultural irrigation technology. It is essential to optimize agricultural irrigation strategies to mitigate trade-offs among ESs and maximize the synergy.

Among the landscape configuration factors, SHDI predominantly exerted a negative influence on the spatial interrelation of ESs, aligning with the findings of Yushanjiang et al. (2018). In the central and southeastern areas of HHHP, LSI primarily benefited from the trade-off dynamics among ESs, while it predominantly had adverse effects in the southwestern region. This suggests that the intricate landscape structure in the central and southeastern areas did not support the cohesive development of ESs. Conversely, the complex landscape shape in the southwestern region facilitated the synergistic evolution of ESs. Considering the impact of spatial heterogeneity introduced by LSI, landscape pattern optimization should account for this influence. To effectively mitigate trade-offs among ESs, management measures should be tailored for each sub-region.

Additionally, our research uncovered that PD negatively affected ES trade-offs. This could be attributed to fragmented landscapes impeding inter-regional energy flow, altering material and nutrient cycling processes, and significantly compromising ESs (Holt et al. 2015). For example, landscape fragmentation hindered the effective large-scale management of arable land and the rational allocation of agricultural resources, notably impacting FP efficiency (Ying-Chieh et al. 2018). Expanding the cropland area was associated with an increased trade-off level between FP and SC. This relationship stemmed from the rise in FP due to expanded cropland, where the soil conservation capacity of cultivated land was weaker compared to forestland and grassland, resulting in decreased SC as the cultivated land area increased.

The expansion of urban land resulted in decreased surface runoff infiltration (Suriya and Mudgal 2012), potentially accounting for the positive influence of urban areas on the trade-off relationship between WY and SC. Conversely, the expansion of construction land encroached upon substantial forest, grassland, and cropland areas, causing a decrease in SC. Regarding socio-economic factors, the increase in GDP spurred greater agricultural inputs and advanced agricultural technology. The adoption of drip irrigation on more arable land not only increased FP but also conserved water and enhanced WY capacity (Wu et al. 2013). Additionally, implementing agricultural practices like conservation tillage played a pivotal role in enhancing SC. Consequently, the rise in GDP corresponded to decreased trade-off levels between FP and SC as well as between FP and WY.

Implications for landscape sustainability

Building on our understanding of trade-offs among ESs, integrating inter-ESs relationships into landscape management is pivotal for enhancing regional landscape sustainability (Christensen and Walters 2004; Feng et al. 2021; Kanter et al. 2018). However, it is crucial to minimize trade-offs between the targeted ESs and other services when applying the trade-off relationships for regional landscape optimization. This alignment with regional characteristics and ecological function positioning aims to maximize the targeted ESs. For example, in the HHHP region, FP stands as the most crucial targeted ES, making it the primary focus for future planning.

Measures such as conservation tillage and water-saving irrigation should be extensively implemented to ensure the maximization of overall benefits among ESs. Badgley et al. conducted an analysis of organic agriculture and global food supply, demonstrating that applying measures such as conservation tillage, crop diversification and intensification, and biological control in farming systems could guarantee food production while preserving ESs (Badgley et al. 2007). This indicates the potential for transforming trade-offs between FP and other services into synergistic relationships through sustainable management practices. Conversely, Inner Mongolia serves as a crucial ecological security barrier (Hao and Yu 2018). Since the 1960s, wind and water erosion have extensively affected many parts of Inner Mongolia, aggravating the issue of sand and dust storms (Wu et al. 2015). When integrating trade-offs into landscape optimization, decision-makers should prioritize enhancing soil retention capacity.

Conclusions

In this study, we selected the major grain-producing area of HHHP as a representative region to quantify the spatial variability characteristics of four key ESs: FP, SC, CS, and WY. The RMSD method was utilized to assess and quantify the trade-offs among these ESs at the watershed level, aiming to comprehend the spatial and temporal variations in trade-off strengths and the influencing factors. The results revealed that although the types of interactions between ESs remained unchanged, there were alterations in the dominant ESs involved in the trade-offs. Trade-offs and their influencing factors exhibited significant spatial and temporal heterogeneity.

Specifically, there existed a high degree of trade-off between SC and other ESs, with none emerging as relative gainers, competing at a disadvantage. Meanwhile, CS competitively gained an advantage against most ESs, acting as a relative gainer, with CS dominance based on the reduction of SC. From 2000 to 2020, there was a notable trade-off relationship between WY and SC, where WY predominantly dominated in most regions, signifying a strong limiting effect of WY on SC in the HHHP. The distribution pattern of the trade-off relationship between FP and WY demonstrated a balanced distribution on both sides of the 1:1 line, indicating no absolute superiority or inferiority in the competition between FP_WY.

In the HHHP region, climate factors, vegetation characteristics, landscape patterns, and socio-economic factors collectively influence the relationships between ESs and their dominant types, displaying varying effects, both positive and negative, across time and space. Precipitation and NDVI were identified as pivotal factors. When optimizing the regional landscape pattern, incorporating the relationship between ES trade-offs and these influencing factors into governmental decision-making processes can lead to the comprehensive optimization of regional ESs. It is important to note the existence of spatial scale differentiation in ES trade-offs and their driving mechanisms. Therefore, future research efforts should undertake cross-scale studies to provide better insights for decision-making in landscape management.