Evolution and Driving Factors of Ecosystem Service Value in the Henan Section of the Yellow River Basin at Different Grid Scales

Abstract

Advancing ecological civilization in the Yellow River Basin requires a nuanced understanding of the spatiotemporal evolution of ecosystem service value (ESV) and its underlying drivers, which are fundamental to regional sustainable development. This study examines the Henan section of the Yellow River Basin, applying the equivalent factor method to estimate ESV in 2020 at three grid scales: 3 km × 3 km, 5 km × 5 km, and 10 km × 10 km. Spatial patterns of land-averaged ESV at each scale are characterized using autocorrelation analysis, while the geodetector model is employed to identify and quantify the influence of driving factors on ESV spatial heterogeneity. The findings reveal that (1) ESV displays both consistent and variable spatial patterns, with higher values in the west and north, lower values in the east and south, and a distinct high-value belt along water bodies; (2) strong spatial positive correlation and aggregation of ESV are observed at all grid scales, though these effects weaken as grid cell size increases; and (3) human activities exert a significant influence on regional ESV, with the interaction of multiple factors providing robust explanatory power for ESV variation, which diminishes with increasing scale. These results offer insights for optimizing ecosystem management and promoting sustainable development in the Yellow River Basin.

1. Introduction

Ecosystem service value (ESV) quantifies the monetary benefits ecosystems provide, encompassing both tangible products and broader contributions to human well-being [1]. As a key indicator for evaluating ecosystem services, ESV reflects the impacts of human activities on natural systems [2]. It captures both economic and non-economic benefits essential for human survival and development, including provisioning (e.g., food, water), regulating (e.g., climate, hydrological regulation), supporting (e.g., soil formation, biodiversity maintenance), and cultural (e.g., recreation, aesthetic value) services. In large river basins such as the Yellow River—particularly the Henan section, which functions as a major grain-producing area, ecological barrier, and economic hub in central China—accurate ESV assessment is critical for balancing ecological conservation with socioeconomic development. However, ongoing economic growth, rising living standards, and the continuous expansion of urban and industrial land use have increasingly disrupted the structural and functional integrity of natural ecosystems [3], threatening the long-term sustainability of ecosystem services [4]. Thus, investigating the spatiotemporal evolution and driving mechanisms of ESV in the Henan section of the Yellow River Basin holds significant academic and practical value for advancing ecological conservation and green development in both the Henan section and the broader basin.

Since the pioneering global quantitative assessment of ESV by Costanza et al. [5], foundational models and methods have been established. Xie et al. [6,7] further refined these approaches by incorporating China-specific ecosystem conditions and introducing an ESV equivalence factor tailored to the national context. The equivalence factor method has since become a widely adopted standard, providing a reference for in-depth ESV studies across diverse regions [8,9].

A variety of methodologies have been employed in ESV research. For instance, Serkan used the travel cost approach to estimate the recreational value of migratory bird habitat at Lake Manyas in Turkey [10], while Qi and Zhang et al. [11,12] applied the hedonic pricing method to evaluate the livability of green spaces. Ma [13] adopted experimental approaches to assess farmland ESV and proposed policy recommendations for ecological compensation. Currently, ESV assessments often draw upon the work of Costanza, Xie, and others, with the InVEST model frequently used to evaluate ecosystem service functions. For example, Pan Tao et al. analyzed the spatiotemporal variation in water supply services using the InVEST model [14]. To identify drivers of ESV changes, spatial autocorrelation models [15], geographic detector models [16,17], and geographically weighted regression (GWR) models [18] are commonly applied. Studies have utilized various analytical units—including provinces [19,20], watershed [21], and the urban area [22]—to examine key drivers such as natural environmental factors (e.g., elevation, terrain), climatic variables (e.g., precipitation, temperature), and anthropogenic influences (e.g., land use change, socioeconomic conditions). Notably, grid-based analysis enables intuitive visualization of ESV spatial distribution and supports quantitative, precise analyses of spatiotemporal ESV changes across different grid scales. Geographic detectors further facilitate the identification of core factors driving the spatiotemporal evolution and spatial differentiation of ESV. Accordingly, grid scales such as 1 km × 1 km [22], 2.5 km × 2.5 km [23], 3 km × 3 km [24], 5 km × 5 km [25] have been adopted as basic analytical units. Research based on grid-scale analysis often yields more accurate results than studies using administrative units, highlighting the multifaceted and complex nature of ESV research. Collectively, these studies have explored the spatial differentiation and driving mechanisms of ESV using various evaluation units, offering substantial reference value for the present research. However, there remains a lack of studies examining spatial variation and visualization of ESV within the same region across multiple grid scales, which constrains the development of differentiated strategies in practical applications. Furthermore, research on the spatiotemporal interaction mechanisms between human activities and ecosystem services is still limited, and studies specifically addressing the impact of human activities on ecosystem services remain insufficient.

The Henan section of the Yellow River Basin serves as a vital ecological corridor linking the Loess Plateau and the North China Plain, playing a crucial role in maintaining regional ecological stability and ensuring national food security. Against this backdrop, the present study focuses on the Henan section of the Yellow River Basin, utilizing 2020 land cover data, socioeconomic data, and other relevant datasets to systematically assess the spatiotemporal evolution of ecosystem services from a multi-grid scale perspective. This research clarifies the spatial interactions among natural, climatic, and anthropogenic factors and ecosystem services across different grid scales; quantifies indices of human impact; applies the geographic detector model to identify dominant factors influencing ESV differentiation and their scale-specific variations; and refines the theoretical framework for understanding ESV influence mechanisms. Ultimately, this study provides a scientific foundation for improving the ecological environment of the Henan section of the Yellow River Basin and enhancing the quality and stability of its ecosystems.

2. Overview of the Study Area and Data Sources

2.1. Overview of the Study Area

The Henan section of the Yellow River Basin (33°41′–36°06′ N, 110°21′–116°06′ E) lies in the middle and lower reaches of the basin (Figure 1), serving as a core region for socioeconomic development. This area includes eight prefecture-level cities in Henan Province: Zhengzhou, Kaifeng, Luoyang, Xinxiang, Jiaozuo, Puyang, Sanmenxia, and Jiyuan. The basin stretches approximately 711 km and covers about 36,200 km². Topographically, it slopes from southwest to northeast, with mountainous and hilly terrain in the west and predominantly flat land in the east, resulting in a pronounced elevation gradient. The region experiences a temperate monsoon climate, with a frost-free period of 210–220 days, annual sunshine of 2300–2600 h, mean annual temperatures between 13.0 and 15.0 °C, and annual precipitation ranging from 600 to 800 mm. Precipitation is unevenly distributed throughout the year.

2.2. Data Sources and Preprocessing

To construct the driving factor indicator system, data were selected based on the study area’s characteristics and the principles of accessibility and scientific validity. Nine datasets were used: digital elevation model (DEM), slope, normalized difference vegetation index (NDVI), precipitation, temperature, per capita GDP, land use type, population density, and human impact index. Data sources are detailed in

Table 1. Data and sources.

Driving Factors	Data	Resolution/m	Data Source	Methods
Natural environmental factors	DEM	30	Geospatial Data Cloud (https://www.gscloud.cn/) accessed on 12 May 2025.	Extracting elevation information for the study area using ArcGIS 10.8 for clipping and resampling.
	Land use data	30	China Land Cover Dataset (https://www.resdc.cn/) accessed on 12 May 2025.	Align with DEM data space and unify the projection coordinate system.
	NDVI	1000	National Science and Technology Resource Sharing Service Platform (https://www.nesdc.org.cn/) accessed on 12 May 2025.	The max value synthesis method integrated the monthly data into annual NDVI data.
Climate factors	Precipitation	1000	National Glacier, Permafrost, and Desert Scientific Data Center (https://www.ncdc.ac.cn/) accessed on 12 May 2025.	Converted to raster data using interpolation methods.
	Temperature	1000	National Glacier, Permafrost, and Desert Scientific Data Center (https://www.ncdc.ac.cn/) accessed on 12 May 2025.
Socioeconomic factors	GDP per capita	1000	National Qinghai–Tibet Plateau Scientific Data Center (https://data.tpdc.ac.cn/) accessed on 15 May 2025.	Processing using site interpolation
	Population Density	1000	Landscan Population Dataset (https://landscan.ornl.gov/) accessed on 15 May 2025.	Crop to the study area and resample to a spatial resolution consistent with other drivers.
Socioeconomic Data	Crop acreage and yield data	-	National Compilation of Agricultural Product Cost and Revenue Data (https://www.stats.gov.cn/) accessed on 15 May 2025.	Extraction of the study area relied on the clipping function in ArcGIS.
	Food price data	-	Henan Provincial Statistical Yearbook (https://tjj.henan.gov.cn/tjfw/zxfb/) accessed on 15 May 2025.
Human activity factors	Human Activity Index	30	Based on land use type data	Determine human activity intensity parameters for each land use type based on existing research, then spatially overlay these parameters onto grid cells to obtain the HAI value for each grid.

. Land use data were obtained from the CLCD dataset and reclassified into six categories—arable land, forest land, grassland, water bodies, built-up land, and unused land—according to actual land use conditions.

3. Research Methods

3.1. Research Approach

A multi-scale analysis of the spatial distribution of land-based ESV clarifies spatiotemporal patterns and identifies dominant drivers of ESV spatial heterogeneity, as well as the intensity of their interactions. This approach reveals scale-dependent differences in influencing factors, providing a scientific basis for differentiated optimization strategies and green development. The research framework is illustrated in Figure 2.

3.2. Calculation of Ecosystem Service Values

The equivalent factor method, revised by Xie in 2015 [6] and adapted to regional characteristics, was used to estimate ESV in the Henan section. For farmland, the average values of dryland and paddy fields from China’s unit-area ESV equivalent table were adopted; for forest land, averages of coniferous, coniferous-broadleaf mixed, broadleaf, and shrub forests were used; for water bodies, averages of water systems and wetlands; for grasslands, averages of grasslands, shrublands, and meadows; for unused land, desert values; and for built-up land, the value coefficient was set to zero [26,27]. This process established the unit-area ecosystem service equivalent table for the study area.

The standard equivalent factor is defined as one-seventh of the annual economic value of grain produced per hectare of farmland in the study area. The economic value of one standard equivalent factor in the Henan section is 2058.20 yuan/hm². Data on unit-area grain yield, planting area, and unit price of major crops (wheat, corn, and rice) were collected and averaged. The economic value of grain yield per unit area is calculated as

$$E_a={\displaystyle \frac{1}{7}}{\sum }_{j=1}^n{\displaystyle \frac{m_j{p}_j{q}_j}{M}}(\mathrm{j}=1,2,3,\dots \mathrm{n})$$

(1)

where E_a is the economic value of food production per unit area, j is the crop type, p_j is the average price of the j-th crop, q_j is the yield of the j-th crop, m_j is the sown area of the j-th crop, and M is the total planted area of all crops.

The value of each ecosystem service per unit area and the total ESV are calculated as:

VC_k = V × C_k

(2)

$$\mathrm{ESV}={\sum }_{k=1}^n\left(A_k\times V{C}_k\right)$$

(3)

where VC_k is the value of i-th ecosystem service per unit area provided by the k-th class of land, ESV is the total ecosystem service value, A_k is the area of the k-th land use type, and k denotes the land use type.

The coefficients of ESV per unit area in the study area after correction according to the calculation are shown in

Table 2. Value coefficient of ecosystem services (yuan·hm⁻²·a⁻¹).

Primary Type	Secondary Type	Farmland	Forest	Grassland	Water	Unutilized Land
Provisioning service	Food production	2274.31	519.69	480.25	1348.12	20.58
	Raw materials	504.26	1193.75	706.65	751.24	61.75
	Water production	−2685.95	617.46	391.06	11,196.59	41.16
Regulating service	Gas production	1831.79	3926.01	2483.56	2747.69	226.40
	Climate production	957.06	11,747.15	6565.65	6061.39	205.82
	Purify environment	277.86	3442.33	2167.97	9416.25	638.04
	Hydrological regulation	3077.00	7687.36	4809.32	130,150.03	432.22
Support service	Soil conservation	1070.26	4780.16	3025.55	3334.28	267.57
	Maintenance nutrients	319.02	365.33	233.26	257.27	20.58
	Biodiversity	349.89	4353.08	2751.12	10,723.20	246.98
Cultural service	Aesthetic landscape	154.36	1908.98	1214.34	6812.63	102.91
Total		8129.87	40,541.32	24,828.71	182,798.69	2264.02

.

3.3. Human Impact Index for ESV

The Human Activity Index (HAI) quantifies anthropogenic impact on regional landscapes [28]. The index ranges from 0 to 1, with higher values indicating greater human disturbance. The calculation is as follows:

$$\mathrm{HAI}={\sum }_{i=1}^N{\displaystyle \frac{A_i{P}_i}{TA}}$$

(4)

where HAI is the anthropogenic impact index, N is the number of landscape types, A_i is the total area of the i-th land use type, P_i is the anthropogenic impact intensity parameter corresponding to the i-th land use type, and TA is the total area of the landscape. Based on previous research [28,29] and regional characteristics, the human impact intensity parameters for the Henan section were set as follows: farmland (0.61), forest land (0.12), grassland (0.11), water bodies (0.12), built-up land (0.95), and unused land (0.05).

3.4. Spatial Autocorrelation Analysis

3.4.1. Global Spatial Autocorrelation

Global spatial autocorrelation of ESV was assessed using Moran’s I index, which characterizes the overall trend of ESV spatial correlation in the Yellow River Basin [30,31] Calculate the Global Moran’s I index to analyze the spatial autocorrelation of the ESV. This index, a statistical measure for spatial patterns in Geographic Information Systems (GIS), ranges from −1 to 1. Positive values indicate the clustering of similar ESV levels (e.g., adjacent areas of high-ESV forests), while negative values indicate the clustering of dissimilar ESV levels. The formula is

$$I={\displaystyle \frac{{\sum }_{i=1}^n{\sum }_{j=1}^n{W}_{ij}\left(x_i-x\right)\left(x_j-x\right)}{S^2{\sum }_{i=1}^n{\sum }_{j=1}^n{W}_{ij}}}$$

(5)

where I is the global Moran index; x_i and x_j are the values for grid cells i and j; (x_i − x) is the deviation from the mean for the i-th grid cell; W_ij is the normalized spatial weight matrix; n is the number of grid cells and S² is the variance.

3.4.2. Local Spatial Autocorrelation

Local spatial autocorrelation and aggregation of ESV were analyzed using LISA cluster maps to examine spatial correlation between regions and their neighbors [9]. The calculation formula is

$$I_i={\displaystyle \frac{\left(x_i-x\right){\sum }_{j=1}^n{W}_{ij}\left(x_i-x\right)}{S^2}}$$

(6)

$$S^2={\displaystyle \frac{1}{n}}{\sum }_{i=1}^n{\left(x_i-x\right)}^2$$

(7)

3.5. Sensitivity Analysis

To validate the accuracy of revised ESV calculations for different land types, a sensitivity analysis was conducted [32]. The value coefficient (VC) for unit-area ESV of each land type was adjusted by ±50% to assess the sensitivity index and examine the dependence of each ESV coefficient. The calculation is:

$$CS=\left|{\displaystyle \frac{\left({\mathrm{ESV}}_j-{\mathrm{ESV}}_i\right)/{\mathrm{ESV}}_i}{\left({\mathrm{VC}}_{jk}-{\mathrm{VC}}_{ik}\right)/{\mathrm{VC}}_{ik}}}\right|$$

(8)

where CS is the sensitivity index; VC_i is the original value coefficient and VC_j is the adjusted coefficient; ESV_i and ESV_j are the original and adjusted ESV values, respectively; and k is the land use type.

3.6. Geodetector

Geographic detectors are statistical methods for identifying driving factors by detecting spatial variations, revealing intra-regional similarities and inter-regional differences [16,33]. he factor detection and interaction detection modules were used to analyze the extent to which influencing factors exert single and interactive effects on ESV. Factor detection assessed the explanatory power of individual factors, while interaction detection evaluated the combined effects of two factors on ESV, determining the strength and type of their interactions. The calculation formulas are:

(1)

The factor detector uses ESV as the dependent variable (Y) and each influencing factor as the independent variable (X). The explanatory power of spatial differentiation for ESV is quantified by the q value:

$$q=1-{\displaystyle \frac{1}{N{\sigma }^2}}{\sum }_{h=1}^n{N}_h{\sigma }_h^2$$

(9)

where q is the indicator for explaining ESV spatial divergence (q∈ [0, 1]), with higher values indicating greater explanatory power, and vice versa. n is the number of strata within the study area, N is the number of samples, and σ² is the variance for the entire area.

(2)

The interaction detector evaluates whether the combined effect of two driving factors enhances or weakens their interaction with ESV changes. Interaction types are categorized as shown in

Table 3. Probe factor interaction types.

Judgment Basis	Interaction
q(A ∩ B) < Min(q(A), q(B))	Non-linear attenuation
Min(q(A), q(B)) < q(A ∩ B) < Max(q(A), q(B))	One-factor nonlinear weakening
q(A ∩ B) > Max(q(A), q(B))	Two-factor enhancement
q(A ∩ B) = q(A) + q(B)	Independent
q(A ∩ B) > q(A) + q(B)	Nonlinear enhancement

.

4. Results and Analysis

4.1. Spatial Patterns of ESV

To accurately represent the study area, a grid-based approach in ArcGIS 10.8 was used to generate three grid scales—3 km × 3 km, 5 km × 5 km, and 10 km × 10 km—covering the Henan section of the Yellow River Basin. Specifically, generated a 3 km×3 km grid scale—dividing the entire study area (Henan section of the Yellow River Basin) into small, uniform squares (each covering 9 km²), which helps capture fine-scale ESV differences that might be hidden in larger administrative units like cities or counties. ESV was spatially visualized, and the natural breakpoint method classified ESV into five intervals: low, lower, medium, higher, and high-value zones (Figure 3). A common GIS data classification tool that automatically groups data into levels based on natural gaps in values, ensuring each level has distinct and meaningful ESV ranges. This classification clarified the spatial distribution of ESV across different scales.

Across all grid scales, ESV displays a consistent spatial pattern: higher values cluster in the western and northern regions, while lower values dominate the east and south. A distinct high-value ESV belt aligns with water bodies, especially in areas rich in aquatic, forest, and grassland resources. In contrast, farmland, built-up land, and unused land—subject to greater human disturbance—exhibit lower ESVs. Although these spatial patterns are broadly consistent across scales, scale-dependent differences are evident. As grid scale increases, the overall spatial distribution characteristics of ESV become more pronounced; however, spatial resolution decreases, and the degree of spatial variation diminishes accordingly.

4.2. Spatial Autocorrelation Analysis of Ecosystem Services

4.2.1. Global Spatial Autocorrelation Analysis

GeoDa was used to examine the clustering relationships and spatial distribution characteristics of ESV across the three grid scales in the Henan section of the Yellow River Basin, with results summarized in

Table 4. Moran’s I of ESV in the study area under different grid scales.

Grid Scale	Moran’s I	E(I)	Z(I)	Z(I)	P
3 km grid	0.766	−0.000	89.019	0.001	0.000
5 km grid	0.759	−0.000	51.927	0.001	0.000
10 km grid	0.671	−0.000	34.589	0.001	0.000

E(I) is the theoretical expectation; Z(I) is the standard deviation; P is the probability.

. A spatial weight matrix was constructed, and Moran’s I index was calculated for univariate spatial autocorrelation of ESV. The index exceeded zero and passed the significance test at the 99.9% confidence level. These findings indicate that ESV in the study area exhibits strong positive spatial correlation and aggregation effects.

Comparative analysis across the three grid scales reveals that Moran’s I is highest at the 3 km scale (0.766) and lowest at the 10 km scale (0.671). This trend suggests that as grid scale increases, Moran’s I for ESV decreases, implying a slight weakening of spatial clustering and positive spatial autocorrelation.

4.2.2. Local Spatial Autocorrelation Analysis

Figure 4 illustrates the spatial distribution of ESV at various scales using LISA clustering. The spatial aggregation pattern of ESV remains consistent across all scales, with four aggregation types identified: high–high, low–low, low–high, and high–low. The high–high and low–low clusters are particularly prominent, exhibiting a distinct clustered spatial distribution.

The high-value concentration zone centers on the Yellow River Basin, extending along the southwestern and northern peripheries of municipal areas. Specifically, this includes the southwestern parts of Sanmenxia and Luoyang, as well as the peripheral areas of Jiyuan, Jiaozuo, and Xinxiang north of the basin. These regions are dominated by forests, grasslands, and water bodies. Conversely, low–low value zones are mainly distributed in the eastern parts of Zhengzhou, Kaifeng, Jiaozuo, Xinxiang, Puyang, and Luoyang, as well as the central basin, where farmland and built-up land prevail and human disturbance is significant. Low–high and high–low clusters are less prominent, typically adjacent to high–high and low–low clusters, scattered along river basin edges and urban centers.

4.3. Sensitivity of ESV to Coefficient Variation

For sensitivity analysis, the service value coefficients for each ecosystem type in the Henan section were increased by 50%. The resulting sensitivity indices (

Table 5. Sensitivity index of ESV in the study area.

Land Use Type	Sensitivity Index CS	Land Use Type	Sensitivity Index CS
Farmland	0.27	Water	0.11
Forest	0.59	Build-up land	0.00
Grassland	0.03	Unutilized land	0.00

) are all below 1, indicating that ESV is inelastic to changes in value coefficients. Among land use types, forest land shows the highest sensitivity index (average 0.297), followed by farmland (0.221) and water bodies (0.121). Grassland and unused land have indices below 0.1. These results confirm that adjustments to value coefficients have limited impact on ESV, demonstrating the robustness of the coefficients used in this study.

4.4. Geographic Detector Analysis of ESV Drivers

4.4.1. Factor Detection Results

A geographic detector was applied to identify driving factors influencing ESV spatial differentiation at the three grid scales. The factor detector quantified the explanatory power of each factor (

Table 6. Results of detecting driving factors for spatial differentiation of ESV.

Grid Scale	3 km Grid		5 km Grid		10 km Grid
	q-Statistic	P	q-Statistic	P	q-Statistic	P
Elevation (X1)	0.376	0.000	0.342	0.000	0.272	0.000
Slope (X2)	0.162	0.000	0.146	0.000	0.087	0.000
Normalized vegetation index (X3)	0.031	0.000	0.024	0.000	0.018	0.000
Precipitation (X4)	0.099	0.000	0.092	0.000	0.081	0.000
Temperature (X5)	0.246	0.000	0.235	0.000	0.152	0.000
Land average GDP (X6)	0.043	0.000	0.042	0.000	0.016	0.000
Land use type (X7)	0.224	0.000	0.206	0.000	0.105	0.000
Population density (X8)	0.042	0.000	0.038	0.000	0.032	0.000
Human influence index (X9)	0.971	0.000	0.970	0.000	0.968	0.000

q denotes the explanatory power of a factor for the spatial heterogeneity of ESVs.

). All driving factors yielded p-values below 0.05, confirming their statistical significance in explaining ESV spatial differentiation.

The analysis indicates that ESV spatial heterogeneity is jointly driven by natural and socioeconomic factors, with the relative influence of each factor varying across grid scales. Notably, socioeconomic factors exert a more pronounced effect on the spatial heterogeneity of regional ESV. The Q-statistic analysis demonstrates that the HAI possesses the strongest explanatory power, underscoring the substantial impact of human activities on ESV spatial differentiation within the watershed. At the 3 km grid scale, the anthropogenic influence index explains 97.1% of the variance, while its explanatory power diminishes at the 10 km scale. This pattern suggests that the spatial analyzability of the ecosystem decreases as grid scale increases. Elevation, temperature, and land use type provide moderate explanatory power (q values 10.5–37.6%), serving as secondary factors. Precipitation, per capita GDP, population density, and normalized vegetation index consistently show the lowest explanatory power (<10.0%), indicating limited influence on ESV spatial differentiation.

4.4.2. Interaction Detection Results

Figure 5 can be interpreted as a scoring table for factor influence. The horizontal and vertical axes represent factors affecting ESV. The color and numerical value (q-value) in each cell of the table indicate the strength of influence for that specific factor combination. Interaction detection results show that interaction q-values for ESV spatial differentiation exceed those of individual factors across three grid scales. Among 45 interaction datasets, most factor pairs exhibit bivariate enhancement, with a few showing nonlinear enhancement. Notably, at all scales, the q-statistic values for interactions between HAI and other factors exceed 0.971, peaking at 0.976 for HAI and precipitation at the 3 km scale. These results underscore the strong association between ESV spatial heterogeneity and specific factor interactions, particularly those involving human activity.

Interactions between HAI and climatic indicators further amplify their impact on ESV. Across all grid scales, the explanatory power of multi-factor interactions for ESV spatial heterogeneity exceeds 50%. Specifically, at the 3 km and 5 km scales, significant interactions occur between X4 and X1 and between X9 and X1–X9; at the 10 km scale, interactions primarily involve X9 and X1–X9. Although the contribution rates of other factor interactions remain below 50%, their explanatory power consistently surpasses that of single factors.

These findings confirm that ESV spatial differentiation is influenced by socioeconomic factors, consistent with the factor analysis. Across multiple grid scales, HAI exert a stronger influence than do interactions between socioeconomic and natural factors alone. Under the combined influence of natural ecosystems and human activities, the spatiotemporal heterogeneity of ESV within the watershed is likely to intensify. Although q-statistic values for interactions among other factors remain below 50%, multi-factor interactions still have a notable impact compared to individual factors. Therefore, in advancing regional development and utilization, it is essential to prioritize ecological environment protection and restoration, safeguard ecosystem stability, fully realize ecological value, and promote harmonious development between society and nature.

5. Discussion, Conclusions and Recommendations

5.1. Discussion

The observed decline in spatial clustering of ESV, as indicated by the reduction in Moran’s I from 0.766 (3 km grid) to 0.671 (10 km grid), is primarily attributable to the “scale averaging effect” within grid cells. This effect obscures fine-scale spatial heterogeneity in ESV. At the 3 km scale, grid cells closely correspond to actual land use distributions, enabling the detection of substantial ESV differences among land use types. Consequently, adjacent grid cells with similar land uses form distinct high–high clusters (e.g., forested mountainous areas in the west) or low–low clusters (e.g., built-up cities in the eastern plains), resulting in pronounced spatial clustering. As the grid scale increases to 5 km and 10 km, each grid cell encompasses a broader area and is more likely to contain multiple land use types with varying ESVs. The ESV of each cell thus becomes an average across all included land use types, diminishing differences between neighboring cells. This averaging reduces ESV similarity among grids, thereby weakening spatial positive correlations and clustering.

The relationship between ESV and its influencing factors demonstrates a marked scale dependence. As grid scale increases, the proportion and spatial concentration of each land cover type within a grid decrease, leading to reduced homogeneity and weaker spatial correlation of ESV. Moreover, the aggregation of data at coarser scales accentuates the spatial heterogeneity of influencing factors. Both ESV and its drivers are thus subject to scale effects, which intensify spatial heterogeneity and attenuate the influence of individual factors. At finer scales, interactions among influencing factors are stronger, and the relative importance of two-factor interactions shifts accordingly. Thus, selecting an appropriate analytical scale is critical for ecosystem optimization, requiring careful consideration of both the variability and interactive effects of influencing factors across scales.

The spatial pattern of ESV in the Henan section is characterized by higher values in the west and north, and lower values in the east and south. High-value areas are concentrated in the mountainous regions of Sanmenxia and western Luoyang, as well as along water corridors, while low-value areas are found in the built-up eastern plains (Zhengzhou, Kaifeng). For ecological barrier construction in the west and north, a strategy of “strict protection + ecological restoration” is recommended. Key measures include delineating ecological functional zones, prohibiting industrial and mining development, and implementing targeted projects—such as afforestation in hilly areas and restoration of riparian buffer zones—to sustain high ESV in forest, grassland, and aquatic ecosystems. This approach is supported by the finding that forest land exhibits the highest ESV sensitivity and serves as the principal driver of ESV growth. In the southeastern region, where low-ESV areas are dominated by farmland and built-up land (e.g., the eastern suburbs of Zhengzhou and the urban periphery of Kaifeng), optimization of land use with ecological considerations is recommended. Measures include integrating “ecological corridors” into urban planning, converting idle land into green spaces, and promoting agricultural models with low input and high ecological benefits to mitigate the adverse effects of intensive human activities. These recommendations align with the finding that human disturbance is the primary driver of ESV variation.

As grid scale increases, both the spatial clustering of ESV and the explanatory power of its driving factors diminish. Targeted governance is therefore most effective at finer scales (3 km): municipal management should utilize 3 km grid ESV maps to identify micro-scale ecological risks. Establishing a grid-based supervision system, with management entities assigned to each 3 km grid, would facilitate localized monitoring and remediation of ecological damage. At coarser scales (10 km), coordination should focus on river basin management. Macro-level policies for the Henan section of the Yellow River Basin should be developed based on 10 km grid data. For instance, an ecological compensation mechanism could link high-ESV mountainous areas in the west with low-ESV plains in the east, requiring highly developed eastern cities (e.g., Zhengzhou) to fund ecological restoration in the west. This mechanism is grounded in the finding that basin ESV heterogeneity is driven by cross-regional human–land interactions.

Geodetector analysis confirms that HAI provides the strongest explanatory power for spatial variations in ESV, with built-up land and farmland identified as the primary sources of disturbance. Accordingly, it is recommended to control construction land use and require ecological facilities for new low-ESV projects to compensate for lost ESV. For farmland ecological optimization, implementing “ecological agriculture transformation” in farmland-dominated regions is advised. Specific measures include converting traditional farmland into ecological buffer zones and prohibiting excessive fertilization—both of which enhance ecological functions through targeted management. Furthermore, coordinated urban–rural development, rational land allocation, high-quality economic growth, and prevention of disorderly urban expansion are essential for sustaining ecosystem services.

5.2. Conclusions

This study examined the Henan section of the Yellow River Basin using multi-scale grids as evaluation units. Employing the equivalent coefficient method, sensitivity analysis, spatial autocorrelation analysis, and geographic detector, the research provided a comprehensive analysis of ESV spatial differentiation across three grid scales. Spatial differentiation patterns, dominant influencing factors, and their interactions were systematically explored. The main conclusions are as follows:

(1)

Across all grid scales, the spatial distribution of ESV in the Henan section exhibits distinct regional variation with overall consistency: higher values are found in the west and north, and lower values in the east and south. High-value areas are concentrated in ecologically sensitive zones, including water bodies, forests, and grasslands. Forests display the highest sensitivity to ESV and are the main ecological factor driving ESV growth.

(2)

ESV in the Henan section demonstrates significant global positive correlation (Moran’s I > 0.67) at all grid scales. High–high clusters are concentrated in water bodies and mountainous forested areas, while low–low clusters are found in the eastern regions and central plains, where human impact is pronounced. Spatial clustering is strongest at finer grid scales and weakens as the scale increases.

(3)

Geographic detector results reveal that ESV spatial differentiation is jointly shaped by natural environmental and socioeconomic factors, with pairwise interaction effects surpassing those of individual factors. Human activities contribute most significantly across all grid scales, particularly at the 3 km scale, confirming their role as the principal drivers of ESV spatial heterogeneity in the Henan section.

5.3. Recommendations

The Yellow River Basin has long faced imbalances between economic development and ecological protection. Enhancing the ecological environment requires precise regulation of ecological space, refined policy implementation, and sustained management. Optimizing the spatial structure of ecosystem services demands scientific rigor in spatial layout and careful consideration of effects on highly correlated groups. The identification of human impact factors as primary drivers of ESV spatial differentiation highlights the disruptive effects of intensive human activities on ecosystem service functions. To mitigate ecological and environmental pressures, establishing a systematic ecological protection policy framework is essential. Environmental access management should be strengthened, unregulated expansion of high-pollution and high-energy-consuming industries strictly curbed, and a full life-cycle land use supervision mechanism established. Ecological governance should be innovated through restoration projects, promotion of green production technologies, and establishment of a robust ecological compensation system—measures that enhance ecosystem resilience and systematically mitigate negative human impacts. Finally, deepening interdepartmental coordination, fostering consensus on ecological protection, integrating resources, and jointly advancing ecological structure optimization will support the sustainable improvement of regional ecosystem service functions and promote environmentally sustainable, high-quality economic and social development.