Spatio-Temporal Distribution and Driving Factors of Ecosystem Service Value in a Fragile Hilly Area of North China

Abstract

Ecosystem services (ESs) are essential for human society, and maintaining harmony between ecosystems and humanity to mitigate ES degradation is the fundamental basis for achieving a sustainable state. However, due to the influence of land use and land cover (LULC) and other ecological-economic factors, the quality and capacity of ESs supporting human welfare continue to decline, and the specific processes involved in this decline are still unclear. In this paper, a dynamically reconstructed assessment model was designed based on the equivalent factor method to estimate the Ecosystem service value (ESV) and to reveal its spatiotemporal response to LULC in a hilly area located in the Economic Circle of Beijing-Tianjin-Hebei during the period from 2000 to 2020; the structural equation model was also used to detect the detailed contribution of ecological-economic factors on ESV. Results showed that due to the decrease in forest land and the sprawl of orchard and construction land between 2000 and 2020, the spatial richness of various ecosystem services reduced, resulting in a decline in the total value of regional ecosystem services. The interaction of LULC, ecological and economic factors increased the regional heterogeneity of ESV. LULC caused a significantly direct impact on ESV (0.543), while economic factors had a negative direct impact on ESV (−0.26). Moreover, terrain factors affected the ESV through LULC and meteorological factors (0.259). The results of this study enrich our understanding of the detailed spatiotemporal variation and driving factors underlying the diminishing ESV in a rapidly developing hilly area, which has substantial guiding implications for land planning and ecosystem protection policies.

1. Introduction

Ecosystem services (ESs) reflect ecological functions and processes with benefits that human beings can obtain directly or indirectly from ecosystems [1], which are the basis for human beings’ survival and development. As the linkages between the earth’s ecosystem and human society [2,3], ecosystems’ sustainable capacity of goods and services provision critically depends on land resource management without damaging the natural resource base [4]. Land use changes usually affect the ecological structure and function of ecosystems by infecting the biophysical processes and the socio-economic flows of materials and energy, constrained by all kinds of societal and environmental factors [5,6,7]. The feedback between these factors and ecosystems is so complicated that the subsequent ESs change mechanism is still unclear. Therefore, exploring the dynamics of ESs across spatial and temporal scales in the identification of critical pathways of factors on ESs is of great significance for future land management and environmental protection [8,9].

Quantifying or valuing ESs and measuring their contribution to human society is the basis in land management decision-making [1,10]. Ecosystem service value (ESV) is a quantitative assessment of potential ecosystem service capacity, which encapsulates the trade-offs and synergies between ecosystem services and human appropriation [11]. There are multiple approaches to assessing ESV [1], and two ESV evaluation methods are widely used currently: the value transfer method and the ecological model method. The value transfer method, proposed by Costanza [12] and revised and improved by Xie et al. [13,14], assigns monetary value to different LULC types to determine ESV, assuming that the same land use type has the same ecosystem service capacity (called equivalent factor). As a result of its simplicity and practicality, it was widely used to quantify the overall economic loss or gain due to spatiotemporal changes in land use on global or national scales [12,15]. However, this equivalent factor method entirely depends on the spatial homogeneity of land use types, and its results are insensitive to the spatiotemporal differences and uncertainties in ESV. In fact, ESs change with external form and internal structure of the ecosystem [16], and the spatial agglomeration level of various services will affect the performance of ESs [17]. The ecological model, similarly to the InVEST, performs well in evaluating a single or several ESV [18] by predicting and mapping the spatial distribution for its functional modeling process, but is difficult for monetizing ESs to identify the spatial gains and losses. In this context, more effort is needed to explore the mosaic pattern of ESV in the same land use type, especially its spatiotemporal responses to land use changes.

Although ESV effectively reveals the capacity of ESs supporting human society, its feeble explanation of the complex feedback mechanism between societal and environmental components makes it insufficient when drawing a comprehensive environmental protection decision. The relationship between human-natural complex systems has drawn many research interests [19,20,21], however, most of them only focused on the relationship between single factor and ecosystem services [22,23], or the confirmation of main influencing factors and their individual effects on ESV [21,24]. Such analyses ignore the joint effects of factors on ESV changes, lacking a systematic perspective to comprehensively understand the mechanism of actors on ESV. Sannigrahi et al. [25] denoted that the joint effects of influencing factors are much higher than their individual effects. Structural equation modeling (SEM) is an appropriate analytical approach for inferring cause and effect between variables [26]. Different from traditional statistical models that emphasize the linear relationships between variables, SEM incorporates graphical and mathematical models to quantitatively analyze relationships between multivariate data and exhibit their direct and indirect effect networks [27]. Therefore, it is widely used in kinds of scenarios for solving complex system problems [28,29].

Qianxi County, located in the Beijing-Tianjin-Hebei Economic Circle, is a typical mountainous and hilly area in north China. The complex terrain conditions and high forest vegetation cover make it a vital ecoregion for soil and water conservation and the drinking water sources for Tianjin and Tangshan City. Its ecological strategic position draws more attention both from academia and governments. Recently, the rapid economic development results in the sprawl of mining land and chestnut economic forest, encroaching and destroying numerous primary forest and other ecological lands. This irrational land use transformation reduces masses of ESs, aggravating soil erosion, river depots siltation and water pollution, and then threatens the sustainable development of Qianxi County itself and even the whole Beijing-Tianjin-Hebei Economic circle. Therefore, this paper takes Qianxi County as the case study attempting to explore the two unclear issues: (1) how to improve the ESV evaluation accuracy and exhibit its spatiotemporal variability more exactly? (2) how to quantify the contributions of factors on ESs? To answer the proposed questions, we aim to develop a dynamically reconstructed ESV assessment model in this study to reveal its spatiotemporal changes and response to LULC, and to make a deep understanding of the mechanism of factors on ESV including their contributions and pathways by the SEM model. Thus, we can provide practical suggestions for decision-makers to mitigate the conflicts between socio-economic development and eco-environmental conservation.

2. Materials and Methodology

As shown in Figure 1, the research framework consisted of four parts. Firstly, the spatiotemporal evolution and internal transformation of LULC over the past 20 years (of the study area, see Section 2.1) were analyzed; secondly, the modified equivalent coefficient and spatial richness index were used for static calculation and dynamic reconstruction of ESV in 2000, 2010 and 2020, respectively, to dissect the spatiotemporal patterns of ESV and its responses to LULC; thirdly, the Structural Equation Modeling (SEM) was used to quantify the driving mechanism of ecological-economic factors on ESV. Then, the strategies of ecological environmental protection and regional sustainable development were proposed in the end.

2.1. Study Area

Qianxi County, located in the Beijing-Tianjin-Hebei Economic circle, is a typical mountainous and hilly area in the northeast part of Tangshan City with a geographical coordinate of 118°6′~118°37′ E, 39°57′~40°27′ N (Figure 2). The terrain feature is high around and low in the middle. The rivers there crisscross and the water resource is rich, in addition to the Luan River and its tributaries, the two famous reservoirs, Daheiting Reservoir and Panjiakou Reservoir, guarantee the municipal water of Tianjin City and Tangshan City. It is honored as “Hometown of Chinese Chestnut” and the largest H-shaped steel production base in China, with iron ore reserves of 470 million tons. In recent decades, the extensive economic development model accompanying with the drastic sprawl of chestnut economic forest and mining land makes it a sensitive and fragile ecological economic zone.

2.2. Data Sources and Land Use Classification

In this study, the Landsat series satellite images from May to September with cloud coverage less than 5% in 2000, 2010 and 2020 were chosen in the GEE platform to extract the land use types by supervised classification. The training samples were collected through Google Earth and the land use were classified into seven types: cropland, forest, orchard, grassland, water, construction land and unused land. The overall classification accuracy of the three periods were higher than 85%, meeting the data accuracy requirement. Other geospatial data required in this study were as shown in

Table 1. Data sources and description.

Data	Source	Year	Description
Landsat images	Google Earth Engine	2000, 2010, 2020	raster (30 m)
DEM	Geospatial Data Cloud platform (http://www.gscloud.cn/, accessed on March 2022)	2020	raster (30 m)
NPP	the Resource and Environmental Science Data Centre of the Chinese Academy of Sciences (https://www.resdc.cn/, accessed on October 2022)	2000, 2010, 2020	raster (500 m)
Soil conservation		2000, 2010, 2020	raster (30 m)
Population density		2020	raster (1 km)
Precipitation	«China Water Resources Bulletin», «Tangshan Statistical Yearbook»	2000, 2010, 2020	numeric
Precipitation	National Earth System Science Data Center (https://www.geodata.cn/, accessed on July 2022)	2000, 2010, 2020	raster (30 m)
Temperature		2020	raster (30 m)
Evapotranspiration		2000, 2010, 2020	raster (500 m)
Road	Open Street Map (https://export.hotosm.org/en/v3/, accessed on October 2022)	2020	Vector (1:1 million)
Tourism and Culture Center	Hebei Culture and Tourism Department (https://www.hebeitour.gov.cn/, accessed on July 2022)	2020	coordinates
Grain yield	«Tangshan Statistical Yearbook»	2000, 2010, 2020	numeric
Crop prices	«Hebei Rural Statistical Yearbook»	2000–2020	numeric

.

2.3. Dynamic Change in Land Use

The land use transfer matrix, which reflects the evolution direction and degree of various land use types, is an important method for quantifying the structural changes in land use during the study time period. In this study, spatial overlay analysis completed in ArcGIS 10.6 was used to create three land use transfer matrices: 2000–2010, 2010–2020 and 2000–2020. These matrices were then visualized to obtain the final land use transfer Sankey map.

2.4. Estimation of ESV

Ecosystem services are constantly changing with the internal structure and external conditions of ecosystems and estimating ESV using the equivalent factor method cannot accurately capture the spatial variation of ESV. For example, ES has been greatly changed from a regulating service into a provisioning service, when large area of primary forest land changed into secondary economic forest land, but the equivalent factor method generally set them the same value, missing some important information. Considering the influence of the spatial agglomeration level of ESs on their functions, this study designed a dynamically reconstructed assessment model by introducing the spatial richness index of all kinds of ecosystem services to the total static ESV of grid in 2000, 2010 and 2020 to detect the spatial gain and loss, which portrayed the spatial differences of ESV in mountainous hilly area more accurately.

2.4.1. Estimating the Static ESV

The core essence of the equivalence factor method is that the economic value of natural food production per hectare of farmland was used to represent the standard equivalent factor [13], and the equivalence factors of other ecological services are defined by the magnitude of their contribution relative to the standard equivalent. We used the equivalent factor table in China [14] and the average market value of grain in the Qianxi County to construct the regional equivalent factor table. The calculation formula is as follows:

$$E = 1/7{\sum }_i^n\frac{m_i{p}_i{q}_i}{M}$$

(1)

where E denotes the economic value of the standard equivalent factor (CNY/hm²); i refers to the crop type; p_i is the average price of crop i in the past 20 years; q_i and m_i represent the yield and area of crop i, respectively; M is the total area of cropland for n types.

Maize, cereals and soybeans were selected as the main crops in this study. The average price of each crop from 2000 to 2020 was extracted to eliminate the impact of price fluctuation on E (Formula (1)). The value of one standard unit of equivalent factor in Qianxi County from 2000 to 2020 was 1302.05 CNY/hm². In addition, the equivalent factor table in China reflects the average annual value of various ESs in the whole country scale, and it should be corrected when using in a local scale. Relative correction models and the correction coefficient were encapsulated in

Table 2. Correction models and correction coefficient of equivalent factor in Qianxi County.

Formula	Implication
$F_{ni}=\{\begin{array}{l}{P}_i\times F_a\\ R_i\times F_b\\ S_i\times F_c\end{array}$ $\begin{array}{l}{P}_i=B_i/\overline{B}\\ R_i=W_i/\overline{W}\\ S_i=E_i/\overline{E}\end{array}$	Fni refers to the modified unit area value equivalent factor of the n-th ecological service type in year i; Pi, Ri, Si refer to the spatiotemporal adjustment factors of NPP, precipitation, and soil conservation, respectively, in year i; Fa, Fb and Fc refer to the basic equivalent factors of ecological service value related to NPP, precipitation and soil conservation services (Table 3); B, W and E are the physical quantity of NPP, precipitation and soil conservation of Qianxi in year i while $\overline{B}$, $\overline{W}$ and $\overline{E}$ refer to the same of nationwide average [14].
Correction Factors	Correction Coefficient
	2000	2010	2020
NPP	0.97	0.88	0.91
Precipitation	0.94	1.17	1.04
Soil Conservation	0.93	1.08	1.15

.

Considering the fact that the chestnut economic forest (orchard) was the dominant land use type while there was no corresponding equivalent factor value in the equivalent factor table in China, this paper took the average equivalent factor value of forest land and cropland as its value based on the fact that the chestnut economic forest is a kind of forest which grows on cropland. Thus, we formed the basic equivalent factor value per unit area of Qianxi County (

Table 3. Basic equivalent factor value of per unit area in Qianxi County.

Ecosystem Service Type		Crop	Forest	Orchard	Grassland	Water	Unused Land
Provisioning services	Food Production	1.11	0.25	0.68	0.23	0.66	0.00
	Raw Material Production	0.25	0.58	0.41	0.34	0.37	0.00
	Water Supply	−1.31	0.30	−0.50	0.19	5.44	0.00
Regulating services	Gas Regulation	0.89	1.91	1.40	1.21	1.34	0.02
	Climate Regulation	0.47	5.71	3.09	3.19	2.95	0.00
	Purifying Environment	0.14	1.67	0.90	1.05	4.58	0.10
	Hydrological Regulation	1.50	3.74	2.62	2.34	63.24	0.03
Supporting services	Soil Conservation	0.52	2.32	1.42	1.47	1.62	0.02
	Nutrient Cycling	0.16	0.18	0.17	0.11	0.13	0.00
	Biodiversity Conservation	0.17	2.12	1.14	1.34	5.21	0.02
Cultural services	Aesthetic Landscape	0.08	0.93	0.50	0.59	3.31	0.01

).

The total ESV can be calculated by the following formula:

$$ES{V}_S={\displaystyle \sum A_n\times F_{ni}}\times E$$

(2)

where ESV_s is the total static value of ESs; A_n represents the area of land use type n (hm²); F_ni refers to the modified equivalent factor value (Table 2); E is the economic value of the standard equivalent factor (1302.05 CNY/hm² in Qianxi County).

2.4.2. Calculating the Spatial Richness Index

The aim of this module is to adjust the ESV in a regional scale to a grid scale to exhibit the spatial agglomeration degree of various ESs. Considering the different basic data granularity, the research region was divided into a grid of 500 m × 500 m as the basic unit to calculate the spatial richness index so as to preserve the data information to the maximum extent [30]. The methods and relevant parameters are shown in

Table 4. Methods for spatial agglomeration degree of ESs.

ES Type	Key Indicators	Quantification Methodology
Food Production	Ratio of cropland and orchard	Assigned the value of cropland and orchard in the land use data to 1 and 0.5, respectively, and other types to 0, then counted the proportion.
Raw Material Production	Forest area ratio	Assigned the forest ‘1’, and other land use types as ‘0’, then counted the proportion of forest in each grid.
Water Supply	Water area ratio	Assigned the water bodies ‘1’, and other land use types as ‘0’, then counted the proportion of water bodies in each grid.
Gas Regulation	Carbon storage	With reference to IPCC parameters, the carbon storage and sequestration module of the Invest model were used to calculate the biophysical amount of carbon storage in four carbon pools (aboveground plant biomass, belowground plant biomass, soil organic matter, and dead organic matter) based on LULC maps.
Climate Regulation
Purifying Environment	Surface runoff	Based on the 30 m × 30 m digital elevation model (DEM), it is obtained by using the tools of fill, flow direction and flow accumulation in the ArcGIS hydrological analysis module.
Hydrological Regulation	water conservation	The water balance equation was used to calculate water conservation. Formula: ∆S = P − R − E, ∆S, P, E, R, respectively represent water storage, precipitation, evapotranspiration; R = P × α, α is the runoff coefficient, refer to the value of zhang et al. [31].
Soil Conservation	soil conservation	Soil conservation data were used to characterize, the data source was shown in Table 1.
Nutrient Cycling	Habitat quality	The Habitat Quality module of the Invest model was used to determine the weight values of threat and sensitivity factors and the maximum impact distance of the threat source on the habitat, and to calculate the habitat quality distribution map.
Biodiversity Conservation
Aesthetic Landscape	Importance of tourism and culture	The forest park, large city park, important public service centers and cultural activity centers were assigned as 1, 0.7, 0.5 and 0.3, respectively, and the importance distribution map of tourist attractions was obtained by the inverse distance interpolation method in ArcGIS.

.

The weights of cropland and orchard were assigned 1 and 0.5, respectively, when characterizing the food production function considering the function transformation of chestnut forest. Habitat quality influenced biodiversity [32], which in turn affects the nutrient cycle of the ecosystem, so it was used to characterize these two functions. Water area and water conservation capacity were used to characterize the water supply and hydrological regulatory functions. The spatial richness index of each grid was obtained by standardizing the calculated results into the domain of 1 and the whole spatial richness index of each service in the region can be calculated by the formula as follows:

$$R_{ki}=\frac{C_{ki}}{{\overline{C}}_i}(k=1,2,\cdots n)$$

(3)

where R_ki is the spatial richness index of service i in grid k; C_ki is the spatial agglomeration degree; ${\overline{C}}_i$ is the average of the spatial agglomeration degree of service i; n is the number of grids.

2.4.3. Dynamic Reconstruction of ESV

The process of dynamic reconstruction of ESV was introducing the spatial richness index of each ES to the static ESV calculating and then overlaying each grid ES layer to exhibit the spatiotemporal distribution patterns of ESV. The calculation formula was as follows:

$$ES{V}_d={\displaystyle \sum _{i=1}^{11}({\displaystyle {\sum }_{k=1}^n{R}_{ki}\times \frac{ES{V}_{si}}{n}})}$$

(4)

where ESV_d represents the total value of the reconstructed ecosystem service; ESV_si is the static value of ecosystem service i; n is the number of grids.

2.5. Structural Equation Modeling

The Structural Equation Model (SEM) was constructed here to quantify the complex pathway relationships and interactions between ecological-economic factors and ESV. The SEM was composed of a measurement model and a structural model. The measurement model was used to measure the covariation relationships between latent variables and observed indicators, and the structural model was used to reveal the structural relationships of latent variables. Relative published research has proved that there is a causal relationship between ESV and LULC, natural factors, and socio-economic factors [28,33], so economic development (ECO), meteorological conditions (MET), terrain factors (TER), and LULC were selected in this study as the four latent variables in the model. Population density and distance to the highway were used to indicate the latent variable ECO; total annual precipitation and average annual temperature represented latent variable MET; elevation and slope were selected to represent the latent variable TER; the manifest variables of LULC were forest area and cropland area. Then, the relevant hypotheses (

Table 5. Hypotheses in structural equation model.

Number	Hypotheses
H1	ECO may have a negative effect on ESV [34,35].
H2	ECO may affect LULC and MET in the region [36].
H3	MET may have a positive effect on ESV [37,38].
H4	MET may affect LULC [15,39].
H5	LULC is the main driver affecting ESV [33].
H6	TER may impact the ESV [40,41].
H7	TER can affect MET and LULC [42].

) were proposed, and the SEM theoretical model (Figure 3) was constructed.

Composite Reliability (CR) and factor loading were used to evaluate the reliability of the measurement model, and the Average Variance Extraction (AVE) was used to assess the convergent validity. While in the structural model, the coefficient of multiple determination (R²) and predictive relevance (Q²) were calculated to measure how well the independent variables performed in explaining the variance in the dependent variable and the predictive relevance, respectively. Goodness of Fit (GOF) was calculated to measure the performance of the whole model. Since the variables in this study were not normally distributed, the partial least squares structural equation model (PLS-SEM) was chosen and all data processing was completed in the software SmartPLS 3.3.9. The value ranges of parameters are shown in

Table A1. Description of indicator parameters.

Validity Type	Parameter	Description	Empirical Value
Internal consistency	Composite reliability (CR)	Reflects whether all measured indicators in each latent variable consistently explain this latent variable. The larger the value of CR, the higher the internal consistency [67].	CA > 0.700. Values must not be lower than 0.600.
Factor reliability	Factor loading	Measures how much of the indicator’s variance is explained by the corresponding latent variable [68].	Values should be significant at the 0.050 level and higher than 0.700.
Convergent validity	Average variance Extracted (AVE)	Attempts to measure the explanatory power of variation of latent variables, and a larger AVE indicates a higher convergent validity of the model [69].	Proposed threshold value: AVE > 0.500.
Model validity	Coefficient of Determination (R2)	represents the extent to which the independent variable explains the dependent variable [68].	Values of approximately 0.670 are considered substantial, values around 0.333 moderate, and values around 0.190 weak.
Model validity	Predictive Relevance (Q2)	A measure of the predictive relevance of a block of manifest variables [70,71].	The proposed threshold value is Q2 > 0.
Model validity	Goodness of Fit (GOF)	Characterizes the performance of the whole model [72]. $\mathrm{GOF}=\sqrt{\overline{AVE}\times \overline{R^2}}$	GOFsmall = 0.1; GOFmedium = 0.25; GOFlarge = 0.36.

.

3. Results

3.1. Dynamic Changes in LULC

3.1.1. Structural Changes in LULC

Forest and orchard land were the two primary land use types in Qianxi County (

Table 6. The area and proportion of different land use types.

LULC	2000		2010		2020
	Area/hm2	Propotion/%	Area/hm2	Propotion/%	Area/hm2	Propotion/%
Cropland	14,845.89	10.29	13,245.97	9.18	13,742.86	9.52
Forest	69,634.21	48.26	61,762.92	42.81	51,578.72	35.75
Orchard	35,871.82	24.86	41,289.78	28.62	50,411.42	34.94
Grassland	5879.34	4.07	4968.38	3.44	4851.64	3.36
Water	5284.69	3.66	4989.47	3.46	5022.9	3.48
Construction land	11,538.09	8	16,917	11.72	17,868.82	12.38
Unused land	1228.05	0.85	1108.58	0.77	805.72	0.56

). In 2000, forest land as the largest proportion of land use types constantly shrank, while orchard land continually expanded and was similar to forest land by 2020. The change rate of forest and orchard land was −35.01% and 28.84%, respectively. The cropland decreased first and then increased slightly, from 14,845.89 hm² in 2000 to 13,245.97 hm² in 2010 and then climbing to 13,742.86 hm² by 2020. Grassland constantly decreased from 5879.34 hm² in 2000 to 4851.64 hm² in 2020. From 2000 to 2010, the construction land sprawled dramatically with a growth rate of 46.6%. The change rate of all land use types from 2010 to 2020 significantly decreased compared with that in the previous decade, which was highly correlated with the resource exhaustion conditions and the implementation of the Ecological Civilization Strategy in recent years.

Spatial distribution patterns of land use types in Qianxi County (Figure 4) were closely related to its topography (Figure 2). Cropland and construction land were primarily distributed in the middle plain area, whereas forest land was primarily distributed in the high-altitude regions of the south and northeast. The forest land shrank and fragmented dramatically due to the sprawl of orchard (chestnut economic forest) and construction land (residential and mining construction), especially in the north of Luanhe River. The residential construction land was expanding constantly with the process of urbanization, obvious in the south of Luanhe River, occupying large amount of cropland.

3.1.2. Conversion between LULC Types

The overall trend of land use transfer in Qianxi County from 2000 to 2020 was that the forest land was transferred into orchard and mining construction land, and cropland was changed into orchard and residential construction land (Figure 5).

During the period of 2000 to 2010, the area of forest and cropland decreased significantly while the area of orchard and construction land increased substantially. As the main outflow type, the forest mainly transferred into orchard, construction land and cropland and the quantity was 10,358.26 hm² with a proportion of 57.8%, 25.4% and 7.75%, respectively. The area of cropland transferred out was 3512.28 hm², 54.9% into construction and 22.2% into forest. The inequality between outflow and inflow of cropland resulted in its total area decline. Orchard land increased by 6837.64 hm² with the 87.6% proportion came from forest and was the main inflow type. Construction land increased by 6071.77 hm², mainly from forest land and cropland.

During the period of 2010 to 2020, the land transfer trend was similar to that of the previous decade, but the expansion rate of orchard accelerated with an increment of 12,549.18 hm² including 11,294.27 hm² transferred from forest. A total of 2780.04 hm² of construction land formed mainly transferred from forest land and cropland with the proportion of 36.8% and 25.6%, respectively. Compared with the 6071.77 hm² of the previous decade, this transfer rate moderated noticeably due to the depletion of mineral resources. Meanwhile, 1828.2 hm² of construction land mainly transferred out into forest and cropland, 674.9 hm² and 518.63 hm², respectively.

3.2. Static ESV Analysis

3.2.1. Changes in the Total ESV

The total ESV of Qianxi County was 3257.56, 3146.4 and 2971.13 million CNY in 2000, 2010 and 2020, respectively (

Table 7. The total static ESV in 2000, 2010 and 2020.

Year	ESV/Million CNY						Total
	Cropland	Forest	Orchard	Grassland	Water	Unused Land
2000	75.76	1759.27	750.56	91.02	580.63	0.31	3257.56
2010	62.69	1525.41	843.46	75.32	639.24	0.27	3146.40
2020	64.95	1242.14	1005.07	71.61	587.18	0.20	2971.13
variation	−10.81	−517.13	+254.50	−19.41	+6.55	−0.12	−286.42

), and it has decreased by 8.79% in the past 20 years. The proportion of ESV composition in land use type from high to low was that forest > orchard > water > grassland > cropland > unused land and this order had no change in the 20 years (Figure 6). The ESV of forest, grassland and cropland decreased by 517.13, 19.41 and 10.81 million CNY, respectively, while the ESV of orchard increased by 254.5 million CNY, from 750.56 million CNY in 2000 to 1005.07 million CNY in 2020. The ESV change in water and unused land was little.

3.2.2. Value Change in Individual Ecosystem Service

During the past 20 years, the most prominent ES in Qianxi County was the regulating service, followed by the supporting service, provisioning service and cultural service (Figure 7). Among them, the hydrological regulation (HR) function contributed the most to the total ESV of Qianxi County, followed by the climate regulation (CR) function. Soil conservation (SC), biodiversity conservation (BC), gas regulation (GR), environmental purification (PE) and aesthetic landscape (AL) functions contributed relatively more while food production (FP), raw material production (RM), water supply (WS) and nutrient cycling (NC) functions contributed less. In terms of variation characteristics, FP and RM decreased first and then increased, while WS and HR ascended first and then descended, and all other ecological functions have declined downward in the past 20 years.

3.3. Dynamic Reconstruction of ESV

3.3.1. Spatial Richness of ESs

As shown in Figure 8, the higher the richness, the greater the ESV in the region. The richness of RW decreased obviously in the north of Luanhe River, while the richness of FP increased significantly in this same region. The area of water bodies had little change and the richness of WS varied scarcely, but the HR richness weakened noticeably especial in the part of the mainstream of Luanhe River and its tributaries. The richness weakening of GR, CR, NC and BC was quite obvious, especially in the area where orchard and mining land sprawled violently. The richness of SC declined slightly in the north, and the richness of PE and AL had no evident changes.

3.3.2. Spatial Distribution of ESV

To better display the spatial distribution and the evolution characteristics of ESV, the reconstructed ESV was divided into five levels using the natural breakpoint method in ArcGIS: Low level (0–0.38 million CNY), Sub-low level (0.38–0.53 million CNY), Intermediate level (0.53–0.63 million CNY), Sub-high level (0.63–0.72 million CNY) and High level (0.72–0.99 million CNY) (Figure 9). The High level was mainly concentrated in water bodies, including forests along rivers and some high-altitude forests with less human activities, but it declined much during the past 20 years. The whole ESV level was weakening, for the transformation of High to Sub-high, Sub-high to Intermediate, Intermediate to Sub-low, and Sub-low to Low was continuing (Figure 10).

3.4. Driving Factors of ESV

The factor loading coefficients of each observation index on its corresponding latent variables were all greater than 0.7 and the significance (P) were less than 0.01, indicating that the observation indexes had an excellent explanatory ability for the latent variables. The internal consistency of the measurement model was excellent because the average variance sampling (AVE) values were higher than 0.5 and the combined reliability (CR) of each variable were higher than 0.7. The R² values of ESV, LULC and MET in this study were 0.386, 0.474, and 0.519, respectively, indicating that the model could explain its variance to a moderate degree. The model had an outstanding predictive correlation as the Q² were greater than 0, and the GOF of the model was 0.60. Therefore, the quality of the measurement model and the structural model both met the requirement, and the SEM results were reliable (the parameters are shown in

Table A2. Indicator parameter of SEM.

Latent Variable	Observed Variable	Factor Loading (P)	CR	AVE	R2	Q2
TER	Elevation	0.932 ***	0.905	0.827	--	--
	Slope	0.886 ***
ECO	Population	0.808 ***	0.783	0.643	--	--
	Distance to the highway	0.796 ***
LULC	Forest	0.893 ***	0.795	0.662	0.474	0.308
	Crop	0.725 ***
MET	Precipitation	0.873 ***	0.884	0.791	0.519	0.407
	Temperature	0.906 ***
ESV	ESV	1.000 ***	1.000	1.000	0.386	0.240

Note: *** indicates passing the 0.001 significance test.

).

The PLS-SEM model (Figure 11) showed that LULC, with a path coefficient of 0.543, had the largest direct positive impact on ESV. As a typical mountainous and hilly area, the direct impact of TER on ESV was not significant, but it had a strong negative (−0.596) and positive (0.541) influence on MET and LULC, respectively, and thus it affected ESV indirectly with an indirect effect of 0.259 (

Table 8. Direct and indirect effects in the structural model.

Paths	Direct Effects (P)	Indirect Effects (P)	Total Effects (P)
TER → ESV	−0.014	0.259 ***	0.245 ***
TER → LULC	0.541 ***	0.062 ***	0.603 ***
TER → MET	−0.596 ***	--	−0.596 ***
ECO → ESV	−0.260 ***	−0.058 ***	−0.319 ***
ECO → LULC	−0.131 ***	−0.023 ***	−0.154 ***
ECO → MET	0.220 ***	--	0.220 ***
LULC → ESV	0.543 ***	--	0.543 ***
MET → ESV	0.115 ***	−0.057 ***	0.058 ***

Note: *** indicates passing the 0.001 significance test.

). The direct influence coefficient of ECO on ESV was −0.26. Additionally, MET conditions had a small direct positive effect (0.115) and indirect negative effect (−0.057) on ESV. The results suggested that the spatial heterogeneity of ESV was caused by the complex interactions between natural and socio-economic factors.

4. Discussions

4.1. Changes in ESV and Its Response to LULC

Analyzing the response of ESV to LULC changes can improve and stabilize the regional ecosystems and provide a basis for optimizing of land use structure [43,44]. Consistent with previous works [45,46,47], this paper also proved that the LULC was the most important driving force for ESs changes. The total ESV decreased by 286.42 million CNY during the past 20 years, and this significant ESV shrinkage was caused by the dramatic loss of forest land and the sprawl of orchard and construction land. Different from other studies that found cropland expansion and urbanization to be the main reasons for the decline in regional ESV [48,49], the sprawl of orchard in Qianxi County took the major responsibility for the ESV decrease. The forest land transferred into orchard land by 5989.62 hm² and 11,294.27 hm² during 2000–2010 and 2010–2020, respectively (Figure 5), made the forest ESV shrink by 233.86 million CNY and 283.27 million CNY and the orchard ESV increase by 92.9 million CNY and 161.61 million CNY during the same period correspondingly (Table 7). The ESV obtained from the increase in orchard land expansion could not compensate for the ESV loss of forest land decrease. The sparse surface vegetation and homogeneous tree variety of artificial forest (chestnut economic forest in Qianxi County) resulted in its weak capacity of ecological services, such as soil and water conservation, carbon sequestration, nutrient retention, habitat quality, and so on [50,51]. From this point of view, localization of the basic equivalent factor value of per unit area (Table 3) was of great significance, not only being a useful supplement to the “Chinese terrestrial ecosystem service value per unit area equivalence table” updated by Xie et al. [14], but also improving the accuracy of regional ESV to make the land use policies more practical. In addition, the dramatic sprawl of mining lands destroyed large areas of forest land and the process of urbanization encroached on a large amount of cropland, which was also responsible for the total decline in ESV [52,53].

As for the change in individual ecological service, the food production (FP) was determined by cropland (Table 3). The cropland in Qianxi County declined first during the first period from 2000 to 2010 due to the sprawl of orchard and mining land and urbanization, but it then ascended during the second period from 2010 to 2020 for the mineral resource depletion and ecological civilization construction. So, the FP service fluctuated in the same way. The water bodies there in Qianxi County were relatively stable (Table 6), but the related ecosystem services, water supply (WS) and hydrological regulation (HR), showed a fluctuation of climbing up and then declining, which was the result of the irregular changing of the precipitation regulation factor. While other ecosystem services all declined in Qianxi County during the past 20 years for the reduction in and fragmentation of forest land (Figure 7). Habitat fragmentation led to the continuous decline in biodiversity [54], aggravated regional soil and water loss, reduced water conservation capacity and habitat quality, and highlighted regional ecological vulnerability [55,56].

The spatial distribution of ESV in Qianxi County (Figure 9) was closely related to its topography, similar to the findings of Li et al. [33], and the pattern was “high in the north and low in the south”. The terrain in the south was flat (Figure 2), suitable for cropland and residential land (Figure 3). Conversely, the terrain in the north was mainly low mountains and hills, and forest, grassland and water bodies agglomerated there. So, the topography factor influenced the ESV by affecting the distribution of land use types [57]. The forest in the north of Luanhe River was occupied by orchard and mining land extensively, while the cropland in the south of Luanhe River was converted into residential construction land accompanying by some forest in the flat area being converted into orchard (Figure 4). This transformation model indicated that the economic activities in Qianxi County were relatively booming and the disturbances on ecological processes were consequently intensive, resulting in the whole landscape fragmentation. The richness of FP increased but the richness of RM decreased. Although the orchard can enhance the food production function, it could also weaken the original regulating and supporting services, resulting in a series of problems such as the decline in biodiversity, soil conservation capacity and local climate disorder [58], exacerbating the imbalance between ecological and economic development. In addition, natural factors such as elevation and slope would increase the risk of soil erosion [40], which was also proved in this study. In fact, the richness index of individual service in Qianxi County showed a weakening trend in varying degrees except FP and WS (Figure 8), and the weakening of RM, GR, CR, NC and BC was quite obvious, which was already confirmed by the results shown in Figure 7. Water bodies maintained the high ESV level and had little change during the past 20 years (Figure 9), but the ESV level in other land use types was on the downgrade to a great extent. The negative land use conversion, similarly to forest to orchard land or to mining land, cropland to residential construction land and so on, accounted for this downgrade.

4.2. Driving Factors of ESV

A deep, comprehensive understanding of the influence mechanisms of factors on ESV is the foundation for ecosystem management and sustainable development decision-making [59]. Existing studies have shown that ESV exhibits significant spatial heterogeneity, and such difference is not caused by one independent factor but the results of the interaction between natural conditions and human activities [11,60]. Consistent with other research results [33], LULC is the largest positive driving force of regional ESV, and the coverage rate of ecological lands, such as forest, grassland or water body, plays a decisive role in ESV [61,62]. Human activities are also critical to the ESV change [63]. While, different from flat regions with high population density and strong disturbance, the economic development in high-altitude areas is still full of challenges although the human activity interference there is weak, due to some other factors. Therefore, the mechanism of factors on ESV in a low mountain and hill area, such as Qianxi County, is quite complicated. As one of the dominant factors of LULC (Figure 11), the terrain changed the spatiotemporal distribution of habitats and resources by influencing regional land use patterns, and then affected the ESV indirectly [35]. Population growth and economic development usually reduce ESV in urban regions [44]. In this study, the total effect of ECO on ESV was −0.319 (Table 8), attributed to the fact that economic development accelerated the intensity of human exploiting natural resources [64], leading to the fragmentation of landscapes [36,59], and thus reducing the ecosystem service capacity. The spatial heterogeneity of meteorological factors in Qianxi County, such as precipitation and temperature, were not obvious due to the limited research scale [65] and the favorable climate background conditions, so its contribution to regional ESV was weak. MET had a direct positive effect on ESV, but it turned to be a weak indirect negative effect on ESV when interacting with LULC. Therefore, the effects of various factors on ESV were not invariable, and this abnormal instability would be violent in this fragile mountainous and hilly area, where the complicated interactions between various factors and ESs may change the direction of ecosystem service evolution.

4.3. Policy Suggestions

Based on the analysis above, the protection of forest and water bodies should be strengthened, and the occupation of forest by chestnut forest and mining sprawl should be reasonably controlled. We proposed that the local government should vigorously promote a series of ecological restoration and land planning projects to coordinate the conflicts between economic development and ecological protection. Basically, ecological restoration and remediation projects, such as soil erosion control, rehabilitation of abandoned mines, river channel dredging and riverbank afforestation and so on, should be extensively taken in the north of Luanhe River, and some land planning projects, such as territorial space planning, village planning and other similar programs, could be implemented in the whole county, with the prime farmland protection planning taken in the south of Luanhe River. The ecological economy model should become the mainstream to change the original extensive economic development model. The high vegetation coverage and richness of water resources could make this ecological economy model possible. The ecological tourism industry, specific farming industry (chestnut industrial chain, mushroom industrial chain), three-dimensional agriculture (Chinese herbs growing under forest and chestnut forest) and other organic agriculture, were all good ecological economy types, which could make Qianxi a county full of competitive advantage in the whole Beijing-Tianjin-Hebei Economic circle. This shift in economic development model could also greatly alleviate the conflicts between ecological protection and economic development, thereby promoting the regional sustainable development greatly. What was noteworthy was that the decision-makers should also pay more attention to the influence of natural and socio-economic factors on ESs to make relative policies more practical. For example, priorities should be given to the spatial differences between economic development model and geographical conditions when making land use planning decisions and differentiated and diversified strategies of coordinated regulation should be adopted to balance the trade-off and synergy between land use multi-functions and to protect the interests of stakeholders when implementing ecological restoration projects.

4.4. Limitations and Implications

In the SEM model, only two observation indexes were chosen, respectively, to represent the potential variables (Figure 3); some important variable, such as land use intensity, land use pattern, population structure, interactions between different ESs, and so on, were excluded in the model limited by data accessibility or the article space. This limitation may lead to certain deviations in the results. In addition, the different precision of the extracted spatial data could also exacerbate the uncertainty of the outcome to some extent. Although the richness index exhibited the spatial agglomeration effects of individual ESV, qualifying the interaction between different ESs on ESV needed some more efforts. In the future, linking the internal interaction between different ESs with external social and ecological factors to construct a more comprehensive ESV influencing mechanism would be of great significance, reducing the trade-offs and enhancing the synergies between multiple ESs to realize the regional coordinated development [66].

5. Conclusions

Alongside extracting the equivalent factor value of orchard that was not included in “Chinese terrestrial ecosystem service value per unit area equivalence table” and localizing all the basic equivalent factor value of per unit area, the main innovation of this study was dynamically reconstructing the ESV by introducing the spatial richness index into the static ESV assessing process. The reconstructed ESV broke the limitation of homogeneous distribution of various functions in the same land use type and the geospatial information became richer. It exhibited the spatial agglomeration effects of various individual ESs in the grid scale instead of the regional scale and revealed their spatiotemporal evolution pattern more precisely to some extent. In addition, the SEM model was used to quantify the influence mechanism of ecological-economic factors on ESV. All of these efforts made the land use policies and ecological conservation strategies more feasible and could be referenced for other similar regions. The relationship between human-natural systems and ESV is a hot research topic, and further research is required to decouple the interactions of natural and socioeconomic factors on ESV.