Evaluation of Multiple Ecosystem Service Values and Identification of Driving Factors for Sustainable Development in the Mu Us Sandy Land

Abstract

Exploring the evolution of ecosystem services value (ESV) and its drivers is pivotal for optimizing the land-use structure and improving the value of ecosystem services. Using the 1980–2020 land-use/land-cover (LULC) dataset of the Mu Us Sandy Land, this study quantitatively evaluated ESV through LULC change, analyzing the spatiotemporal evolution characteristics of ESV and its driving forces. The results showed that (1) the LULC changes were stable from 1980 to 2020, and the ESV showed a slight downward trend in general. Grassland and water ecosystem services predominantly influenced ecosystem service function value fluctuations across the study area. (2) ESV demonstrated strong positive spatial autocorrelation, with high-value areas concentrated primarily in Red Alkali Nur, Dawa Nur, Batu Bay, and Ulanmulun Lake and low-value areas mainly distributed in unused land and certain agricultural zones. (3) The land-use degree and human activity intensity index were the main factors leading to the differentiation of ESV. The synergistic effects of human activities, landscape pattern changes, and natural factors led to the spatial differentiation of ESV in the study area. Beyond artificial ecological restoration projects, policies for ecosystem service management should pay more attention to the role of geodiversity in service provision.

1. Introduction

Ecosystem services (ESs) are the direct or indirect benefits ecosystems provide to human societies, encompassing the full range of ways the natural environment supports human survival and development [1]. These services are not only a matter of human well-being but also directly impact the Earth’s ecological balance and sustainable development [2,3]. The Millennium Ecosystem Assessment (MEA) assessed that about 60% of the world’s ecosystem services were degraded because of anthropogenic disturbances and emphasized the importance of geodiversity [4]. Geodiversity plays a crucial role in the effective management of ecosystems and the protection of sustainable development by providing and maintaining ecosystem services [5,6]. Ecosystem service value (ESV) is an important indicator for assessing ecosystem services. The examination of ESV accounting serves as a fundamental framework for constructing ecological security patterns, ecological function zoning, and ecological construction planning. Costanza et al. (1997) first calculated global ecosystem service value (ESV) equivalence factors in 1997 [1]. Subsequently, many scholars have drawn on the methodology to assess the value of ecosystem services in different countries and regions [7,8,9]. Changes in land use/land cover (LULC) reflect the interaction of human activities with the natural ecosystems [10]. Such changes directly affect ecosystem structure and functioning [11,12,13], which, in turn, causes changes in ESV. Thus, it is now necessary for land-use planning and regional management to optimize land use and conduct a scientific evaluation of ESV [14,15]. Cabral et al. (2016) quantified the impact of land-use change on ecosystem services and found that ESV was influenced by land-use change in Bordeaux, France [16]. Niquisse et al. (2017) conducted their monetary assessment of ecosystem services and changes resulting from land-cover changes in Mozambique, identifying cropland as a key factor influencing ESV [17].

Desertification refers to land degradation caused by many factors such as human activities and climate change [18,19,20]. According to the United Nations, 70% of the drylands in the world have undergone varying degrees of degradation. A total of 41% of the global land area is subject to desertification, affecting nearly two-thirds of the countries and 38% of the population in the world [21]. Desertification has become one of the globally significant environmental problems because its impacts threaten the basic survival needs of human beings. Therefore, advancing research on desertification assessment is crucial for enhancing ecological environments and can serve as an essential reference for desertification prevention and management initiatives [22]. China is deeply endangered by desertification, especially in Northwest China. Desertification is becoming more severe in these regions due to the combined strain of the delicate ecological environment and the growing population, which has seriously threatened the long-term viability of the environment and the social economy [23]. While the assessment of ecosystem service value (ESV) has emerged as a frontier hotspot in ecological research, studies on the long-term temporal dynamics of ESV in arid regions under the dual contexts of climate change and land use/land cover (LULC) remain scarce. In-depth exploration of the dynamic evolution of ESV in arid regions under such contexts, along with the revelation of its driving mechanisms, can provide solid theoretical support and decision-making basis for the scientific conservation and management of dryland ecosystems, thus holding significant academic and practical value.

The Mu Us Sandy Land is located at the junction of arid and semi-arid regions of Eurasia and the agricultural and pastoral intertwined zone in northern China [24,25]. The region is a key barrier to maintaining ecosystem stability in China and a priority area for ecological restoration. Since the Early Quaternary, the ecological environment of the Mu Us Sandy Land has been fragile and highly sensitive to climate variability [26]. The desertification of the area has been worsening due to the impacts of climate and human activities [27,28]. To reverse the desertification trend, since 1978, the Chinese government has implemented major ecological engineering projects such as the “Three-North” Shelter Forest Program (TNSFP), Natural Forest Protection Project (NFPP), and Beijing–Tianjin Sandstorm Source Control Project (BTSSCP) in the Mu Us Sandy Land, which has significantly increased the vegetation cover of the Mu Us Sandy Land and gradually improved the ecological environment [29,30,31]. Therefore, studying the value of ecosystem services in the Mu Us Sandy Land is significant in guiding desertification control.

In this context, based on land-use data from 1980 to 2020, we explored the spatiotemporal evolution of ESV in the Mu Us Sandy Land and then quantified the effects of climate change, human activities, and landscape patterns on ESV. The specific objectives of this study are as follows: (1) to analyze the response of ecosystem service values to land-use changes from 1980 to 2020; (2) to explore the spatial variation characteristics of ESV; (3) to quantify the drivers of spatial ecosystem differentiation. This study provides an essential theoretical basis for the restoration and sustainable utilization of land resources in arid and semi-arid zones.

2. Methodology and Data Source

2.1. Study Area

The Mu Us Sandy Land is in the junction of the Inner Mongolia Autonomous Region, Ningxia Hui Autonomous Region, and Shaanxi Province (37°30′–39°20′ N and 107°20′–111°30′ E), with a total area of 48,000 km² and an elevation of 1000–1600 m a.s.l., decreasing from northwest to southeast [24,32] (Figure 1). The climate here is a temperate continental monsoon climate, as the area belongs to the edge of the Lower Summer Monsoon Zone in East Asia. The average annual precipitation decreases from 470 mm in the southeast to 260 mm in the northwest, the average temperature ranges from 6 to 9 °C, the average evaporation ranges from 1800 to 2500 mm, and the annual relative humidity ranges from 47% to 51% [33]. The region is covered by farmland, woodland, and grassland, with a harsh natural environment, and the dual pressure of ecology and population restricts local agricultural production and socio-economic development.

2.2. Data Sources

All data used in the study were as follows: (1) land-use type data came from the Resources and Environmental Science Data Center (https://www.resdc.cn/, accessed on 10 March 2024), with a spatial resolution of 30 m and a classification accuracy of more than 90%. We reclassified the land use/land cover into six first-level classes using ArcGIS 10.8 (including farmland, woodland, grassland, water, built-up land, and unused land). (2) The factors influencing the value of ecosystem services, including the mean annual temperature (TEM), the mean annual precipitation (PRE), and the mean annual potential evapotranspiration (PET), were obtained from the Institute of Tibetan Plateau Research, Chinese Academy of Sciences (https:// data.tpdc.ac.cn/, accessed on 20 March 2024), all with a spatial resolution of 1 km; elevation data (DEM) were obtained from the US Geological Survey (https://glovis.usgs.org, accessed on 5 March 2024) with a spatial resolution of 1 km, and slope was obtained using the ArcGIS 10.8 slope tool; net primary productivity (NPP) was obtained from the National Aeronautics and Space Administration (https://www.nasa.gov/, accessed on 6 May 2024) with a spatial resolution of 500 m; gross domestic product (GDP) and population density (POP) data were obtained from the Resources and Environmental Science Data Center (https://www.resdc.cn/, accessed on 9 May 2024), both with a spatial resolution of 1 km; Normalized Difference Vegetation Index (NDVI) was downloaded from National Earth System Science Data Center with a spatial resolution of 1 km; human activity intensity (HAI) and land-use intensity (La) were computed using ArcGIS 10.8 [34,35], with detailed calculation procedures and their implications provided in Supplementary Materials (S1) and (S2), respectively. In addition, we calculated the mean shape index (MSI), landscape division index (DIVISION), and Shannon diversity index (SDI) by using the Fragstats software 4.2. Specifically, MSI reflects the complexity of patch shapes, which is typically related to the ratio of patch perimeter to area. DIVISION quantifies the degree to which the landscape is fragmented or divided into separate patches. SDI, rooted in information theory, assesses the heterogeneity of land-use/cover types by integrating both class richness and proportional abundance, providing a comprehensive quantification of landscape diversity across the study region. (3) Other socio-economic data were obtained from the China Agricultural Yearbook and the National Compendium of Cost and Benefit Information of Agricultural Products. All spatial data were standardized to the same coordinate system and resampled to a 1 km resolution using the nearest neighbor method to maximize the preservation of the original data values. The spatiotemporal analysis part of the ESV in this study was conducted on a 1 km raster scale, balancing the accuracy of the data analysis and the efficiency of the model computation. The study area consists of 10,537 grids with a grid of 3 km × 3 km as the basic evaluation unit, and the mean value of all the variables in each grid was calculated and used in the driver analysis part of this study.

2.3. Research Methods

2.3.1. Measurement of Land-Use Change

Parameters such as degree of single dynamics (K), degree of composite dynamics (K_S), and transfer matrix (S_ij) were used to describe the land-use changes in the study area [36].

(1)

Changes in land categories were analyzed through the land-use momentum, attitude, and land-use transfer matrix. Land-use momentum (K) can compare land transformation differences in different periods and is an important indicator for analyzing land-use change [12].

$$K\left(\%\right)={\displaystyle \frac{U_b-U_a}{U_a}}\times {\displaystyle \frac{1}{T}}\times 100$$

(1)

where $U_a$ and $U_b$ are the area of a land category at the end and beginning of the study period; T represents the number of annual intervals in the different monitoring period.

(2)

The degree of integrated land-use dynamics can characterize the rate of land-use change throughout the study area.

$$K_s\left(\%\right)={\sum }_{i=1}^n\left|u_{bi}-u_{ai}\right|/2{\sum }_{i=1}^n{u}_{ai}\times {\displaystyle \frac{1}{T}}\times 100$$

(2)

where K_s is the degree of integrated land-use dynamics, and $u_{ai}$ and $u_{bi}$ are the area of the specific land-use type at the start and end dates, respectively. T is the monitoring period, and n is the number of LULC types.

(3)

With the support of ArcGIS tools, the land-use transfer matrix was obtained through interactive overlay calculation to analyze the dynamic transformation process of each land-use type in the Mu Us Sandy Land.

$$S_{ij}=\left[\begin{array}{cccc}{S}_{11}& S_{12}& \dots & S_{1n}\\ S_{21}& S_{22}& \dots & S_{2n}\\ \dots & \dots & \dots & \dots \\ S_{n1}& S_{n2}& \dots & S_{nn}\end{array}\right]$$

(3)

where S is the area; n is the number of different types of land use in the Mu Us Sandy Land; and i, j (i, j = 1, 2, 3, ···, n) are the types of land use before and after the land transfer, respectively.

2.3.2. Measurement of Ecosystem Service Values

The equivalent factor method and the physical quantity-based method are two commonly used approaches for evaluating ecosystem service values in current research [10,37]. Compared with the physical quantity-based method, the equivalent factor method is simpler, more operational, and facilitates result comparison in practical applications, enabling rapid accounting of ecosystem service values. Therefore, this study primarily employs the equivalent factor method to evaluate ecosystem service values in the Mu Us Sandy Land [35].

(1)

The equivalent factor of ecosystem service value was revised by using the grain yield per unit area of cultivated land and the grain purchase price of Mu Us Sandy Land in 2020 as the baseline. Xie et al. (2003) and Pan et al. (2021) argue that the economic value corresponding to the per-unit-area equivalent factor for ecosystem service valuation should be set at one-seventh of the product derived from the average unit grain yield multiplied by the grain purchase price [10,38]. The assessment formula is as follows:

$$E_a={\displaystyle \frac{1}{7}}\times {\sum }_{i=1}^n{\displaystyle \frac{m_i{p}_i{q}_i}{M}}\left(i=1,2,3,\dots n\right)$$

(4)

where E_a is the economic value of providing food production services per unit area of farmland (CNY/hm²), p_i is the type of crop and is the average price of the crop (CNY/kg), q_i is the yield of crop i per unit area (kg/hm²), m_i is the total planting area of the i-th food crop (hm²), and M is the total area of food crops (hm²). The economic benefit of the study area is 1759.75 CNY/hm².

(2)

The formula for the value of ecosystem services per unit area (VC) of an ecosystem is provided as follows:

$$VC=E_a\times Q$$

(5)

where Q is the ecosystem service value equivalent per unit area proposed by Xie et al. (2015) [37].

(3)

The formula for calculating ESV per unit area for different land-use types in the Mu Us Sandy Land is as follows (

Table 1. Ecosystem service value per unit area of different land-use types in the Mu Us Sandy Land (unit: CNY/hm²).

Service Type Categories	Service Type Categories	Farmland	Woodland	Grassland	Water	Built-Up Area	Unused Land	Total
Provision services	Food production	1495.79	360.75	410.61	1152.64	0.00	17.60	3437.38
	Raw material production	703.90	835.88	604.18	642.31	0.00	52.79	2839.06
	Water resources supply	35.19	431.14	334.35	9573.04	0.00	35.19	10,408.92
Regulatory services	Gas regulation	1179.03	2736.41	2123.43	2349.27	0.00	193.57	8581.71
	Climate regulation	633.51	8182.84	5613.60	5182.46	0.00	175.97	19,788.39
	Environmental purification	175.97	2437.25	1853.60	8050.86	0.00	545.52	13,063.21
	Hydrological Regulation	475.13	5886.36	4111.95	111,277.77	0.00	369.55	122,120.77
Support services	Soil conservation	1812.54	3325.93	2586.83	2850.79	0.00	228.77	10,804.86
	Nutrient cycling maintenance	211.17	255.16	199.44	219.97	0.00	17.60	903.34
	Biodiversity protection	228.77	3035.57	2352.20	9168.30	0.00	211.17	14,996.00
Cultural services	Aesthetic landscape	105.58	1328.61	1038.25	5824.77	0.00	87.99	8385.21
	Total	7056.60	28,815.90	21,228.45	156,292.17	0.00	1935.72	215,328.84

Note: The unit prices in the table are the unit prices of 2020. “Gas regulation” refers to the process by which ecosystems regulate atmospheric composition through absorbing, releasing, or transforming gases such as carbon dioxide (CO₂), oxygen (O₂), and methane (CH₄).

):

$$ESV={\sum }_{i=1}^m{\sum }_k^n{A}_{ik}\times {VC}_{ik}$$

(6)

where $A_{ik}$ is the area of the i-th land-use type (hm²), ${VC}_{ik}$ is the value coefficient of the i-th land-use type corresponding to the k-th service (Table 1), i is the number of land-use types, k is the type of ecosystem service, m = 6 (including farmland, woodland, grassland, water, built-up land, and unused land), and n = 11 (including food production, water resources supply, etc., as listed in Table 1).

2.3.3. Measurement of Ecosystem Service Value Sensitivity Index

According to a paper by Pan et al. (2021) [10], which conducted relevant research in Northwest China (a region with similar ecological characteristics to our study area), considering the uncertainty in the value coefficient (VC), the VCs were adjusted by 50% for each land-use category, and the estimated ESV results were validated with sensitivity index coefficients [39]. The coefficient of sensitivity (CS) is used to measure the responsiveness of ecosystem service value (ESV) to changes in the value coefficient (VC), specifically reflecting the percentage change in ESV caused by a 1% adjustment in VC [10]. The formula is:

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

(7)

where ${ESV}_i$ and ${ESV}_j$ denote the initial and adjusted total estimated ESV, respectively, and VC_ik and VC_jk denote the initial and adjusted VC, respectively, for land-use type k. If CS > 1 (<1), then for every 1% increase or decrease in VC, the increase or decrease in ESV is greater (less) than 1%, indicating a high (low) elasticity of ESV concerning VC, reflecting a low (high) confidence in the evaluation results. The greater the proportional change in ESV relative to VC, the more critical it is to use accurate ecosystem VC [7].

2.3.4. Spatial Autocorrelation Analysis of ESV

Exploratory spatial data analysis (ESDA) involves using spatial analysis techniques to visualize the spatial distribution patterns of objects or phenomena [40]. In this paper, the spatial and temporal characteristics of ecosystem service values in the Mu Us Sandy Land from 1980 to 2020 were investigated using ArcGIS 10.8 with the ESDA method. In this study, the global spatial autocorrelation Moran’s I in the ESDA analysis method was chosen to reflect the spatial heterogeneity of ESV, which was calculated as follows [10]:

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

(8)

where n is the total number of ESV evaluation units; x_i (x_j) is the observed value of ESV evaluation unit i (j); and W_ij is the spatial weight between ESV evaluation units i and j, where x is the average value of ESV evaluation units. The Moran’s index I value is assigned the range of [−1, 1]. When the value of Moran’s I is close to 1, a clustering pattern occurs in this index. When the value of Moran’s I is close to −1, the indicator has a decentralized pattern. When Moran’s I value is 0, it indicates that the ESV values are randomly distributed within the region and do not have spatial autocorrelation.

The results can be obtained after completing the calculation of Z-scores and p-values. To examine the statistical significance of Moran’s I statistic, the formula was calculated as follows:

$$Z(I) ={\displaystyle \frac{1-E\left(I\right)}{\sqrt{VAR\left(I\right)}}}$$

(9)

where E(I) is the expected value of I, and $\sqrt{VAR\left(I\right)}$ is the expected variance of I. When the calculated p-value is greater than 0.05, the underlying assumption that the data values are spatially randomly distributed is accepted. When the p-value is less than 0.05 and the Z-score is negative, the underlying assumption of randomness is rejected, and it is inferred that the highest and lowest values in the dataset are spatially dispersed. Similarly, when the p-value is less than 0.05 and the Z-score is positive, the assumption of randomness is again rejected, and the inference is that the high/low data values are spatially clustered in geographic space.

2.3.5. Analysis of Driving Factors in ESV Spatial Heterogeneity

The GeoDetector model (GDM) is a set of statistical methods to detect spatial dissimilarity and reveal the driving force behind it, and its basic theory is to judge the similarity of the spatial distribution of two variables from the perspective of spatial stratified heterogeneity [41]. The model includes four types of detection functions: ecological detection, interaction detection, factor detection, and risk detection. In this study, we used the factor detection and interaction detection methods to analyze each driving factor’s influence on ESV’s spatial distribution.

(1)

Spatial and temporal variations in ESV are the result of a combination of factors [42]. The factor detector detects the spatial heterogeneity of the dependent variable (ESV in this study) and analyzes the effect of each driver on the spatial distribution of ESV by comparing the q-values:

$$q =1-{\displaystyle \frac{{\sum }_{h=1}^L{N}_h{\sigma }_h^2}{N{\sigma }^2}}=1-{\displaystyle \frac{SSW}{SST}}, h= 1, 2, \dots , L$$

(10)

$$\mathrm{SSW}={\sum }_{h=1}^L{N}_h{\sigma }_h^2$$

(11)

$$\mathrm{SST}=N{\sigma }^2$$

(12)

where h = 1, 2, …, L is the classification or zoning of variable ESV or factor X; $N$ and $N_h$ are the number of cells in the whole region and layer h, respectively; $\sigma$ and ${\sigma }^2$ are the variance of ESV values in the layer h and the whole region, respectively; and SSW and SST are the sum of within-layer variance and the total regional variance, respectively. q denotes the degree to which the spatial heterogeneity in ESV changes can be attributed to factor X, with its value ranging between 0 and 1. A higher value indicates a more pronounced influence of the driving factor on ESV.

(2)

Interaction detection is used to identify interaction between every two drivers, whether drivers X₁ and X₂ acting together increase or decrease their influence on ESV or whether the effects of these drivers on ESV are independent of each other. The interactions are categorized into five groups (

Table 2. The interaction type of each factor.

Interaction Type	q Value Relationship
Non-linear reduction	q(X1∩X2) < min(q(X1), q(X2))
Single-factor non-linear reduction	min(q(X1), q(X2)) < q(X1∩X2) < max(q(X1), q(X2))
Bi-factor enhancement	q(X1∩X2) > max(q(X1), q(X2))
Independent	q(X1∩X2) = q(X1) + q(X2)
Non-linear enhancement	q(X1∩X2) > q(X1) + q(X2)

).

3. Results

3.1. Variation Characteristics of the LULC

The results showed that grassland was the dominant land-use/cover type (about 57%), followed by unused land, farmland, woodland, water, and built-up land in 2020. Land-use changes in the study area mainly occurred in the southern and eastern regions, where human activities were intensive. Changes were particularly evident in built-up land and woodland (

Table 3. Area and dynamic degrees (K) of various land-use types in the Mu Us Sandy Land during 1980—2020.

LULC Classes	Area(km2)					K(%)
	1980	1990	2000	2010	2020	1980–1990	1990–2000	2000–2010	2010–2020	1980–2020
Farmland	13,462	13,472	13,946	13,443	13,246	0.01	0.35	−0.36	−0.15	−0.04
Woodland	2321	2322	2369	2784	2880	0.00	0.20	1.75	0.34	0.60
Grassland	52,450	52,414	52,781	53,514	52,065	−0.01	0.07	0.14	−0.27	−0.02
Water	1263	1219	1186	1055	1167	−0.35	−0.27	−1.10	1.06	−0.19
Built-up area	463	468	503	954	1863	0.11	0.75	8.97	9.53	7.56
Unused land	21,184	21,248	20,358	19,393	19,922	0.03	−0.42	−0.47	0.27	−0.15

and

Table 4. Land-use transition matrix in the Mu Us Sandy Land from 1980 to 2020 (Unit: km²).

Year	Land-Use Type	2020
		Farmland	Woodland	Grassland	Water	Built-Up Area	Unused Land
1980	Farmland	11,649.0366	296.4366	1211.5458	31.7052	206.1954	66.7917
	Woodland	63.0819	1896.7761	207.0783	11.3427	67.2561	75.8079
	Grassland	1171.2042	526.8645	48,008.5488	96.8832	781.7463	1864.4283
	Water	42.7203	6.9903	91.53	966.6189	36.5607	118.9107
	Built-up area	15.0012	0.7371	13.923	0.8541	430.2234	1.9377
	Unused land	304.7229	152.325	2532.852	59.7609	340.6158	17,794.0881

, Figure 2). During the 41-year study period, the built-up land continued to expand, increasing by 1400 km², with an expansion rate of 34.14 km²/a and a dynamism of 7.56%. The increase in construction land comes mainly from grassland, unused land, and farmland (Figure 3). Woodland was distributed around built-up land and had grown significantly during the study period. The growth rate amounted to 24.08%, with most of the increment of woodland occurring between 2000 and 2010. Farmland and grassland, which were the main components of the study area, remained stable (the degree of dynamics was −0.04 and −0.02, respectively), while the areas of water and unused land decreased by 96 km² and 1262 km², respectively, over the 41 years, exhibiting fluctuating changes. Overall, land-cover changes during the 41 years were more complex and occurred mainly in areas with concentrated human activities.

3.2. Dynamics of the ESV

Based on the ecosystem service value per unit area of different LULC in the study area (Table 1) and the land-use area (Table 3), the ESV and ESV_f of the Mu Us Sandy Land were estimated. We find that the ESV in the Mu Us Sandy Land showed an overall decreasing trend, with a total reduction of CNY 1103.6 × 10⁶ (

Table 5. Change in ESV of Mu Us Sandy Land from 1980 to 2020 (×10⁶ CNY).

	Year	Farmland	Woodland	Grassland	Water	Built-Up Area	Unused Land	Total
Ecosystem services value	1980	9499.59	6688.17	111,343.21	19,739.70	0.00	4100.64	151,371.31
	1990	9506.65	6691.05	111,266.79	19,052.02	0.00	4113.03	150,629.53
	2000	9841.13	6826.49	112,045.87	18,536.25	0.00	3940.75	151,190.49
	2010	9486.18	8022.35	113,601.91	16,488.82	0.00	3753.95	151,353.22
	2020	9347.17	8298.98	110,525.91	18,239.30	0.00	3856.35	150,267.71
Change (%)	1980–1990	0.07	0.04	−0.07	−3.48	0.00	0.30	−0.49
	1990–2000	3.40	1.98	0.70	−2.78	0.00	−4.37	0.37
	2000–2010	−3.61	17.52	1.39	−11.05	0.00	−4.74	0.11
	2010–2020	−1.47	3.45	−2.71	10.62	0.00	2.73	−0.72
	1980–2020	−1.60	24.08	−0.73	−7.60	0.00	−5.96	−0.73

). Of these, grassland, which almost accounted for 59% of the study region, contributed the most to ESV (73.55–73.87%), followed by water (12.14–12.65%). At the same time, the contribution of other types of land use in ESV (13.48–14.31%) was also of interest. In terms of the trend of decrease in ESV, water was the major contributor (135.96%), followed by the grassland ecosystem (74.06%). The ESV displayed a fluctuating trend, decreasing and then increasing, followed by another decrease from 1980 to 2020.

There was an overall decreasing trend in the values of each ESV_f in the study area, but there were significant differences between functions. Among all services, hydrological regulation (>24%) and climate regulation (>21%) accounted for the largest portion of total ESV. The analysis of ESV_f at different stages showed that the decreases in total ESV from 1980 to 1990, from 2010 to 2020, and from 1980 to 2020 and the increase in total ESV from 1990 to 2010 are mainly due to increases and decreases in the value of regulatory services (including gas regulation, climate regulation, environmental purification, and hydrologic regulation) (

Table 6. The values of each ecosystem function in the study area from 1980 to 2020 (ESV_f) (×10⁶ CNY).

		1980		1990		2000		2010		2020
Service type categories	Service type categories		%		%		%		%		%
Provision services	Food production	4433.86	2.93	4428.95	2.94	4511.25	2.98	4464.28	2.95	4392.62	2.92
	Raw material production	4503.49	2.98	4499.61	2.99	4552.26	3.01	4582.32	3.03	4498.92	2.99
	Water resources supply	3184.76	2.10	3141.74	2.09	3122.98	2.07	3034.80	2.01	3098.88	2.06
Regulatory services	Gas regulation	14,066.51	9.29	14,051.22	9.33	14,172.91	9.37	14,333.36	9.47	14,065.27	9.36
	Climate regulation	33,222.74	21.95	33,182.30	22.03	33,424.05	22.11	34,058.37	22.50	33,378.39	22.21
	Environmental purification	12,697.19	8.39	12,659.00	8.40	12,671.71	8.38	12,741.76	8.42	12,612.13	8.39
	Hydrological regulation	38,410.25	25.37	37,909.25	25.17	37,710.24	24.94	36,738.63	24.27	37,455.82	24.93
Support services	Soil conservation	17,624.60	11.64	17,606.36	11.69	17,773.07	11.76	17,950.12	11.86	17,615.54	11.72
	Nutrient cycling maintenance	1454.61	0.96	1453.28	0.96	1469.51	0.97	1479.52	0.98	1452.31	0.97
	Biodiversity protection	14,955.10	9.88	14,908.18	9.90	14,970.57	9.90	15,116.97	9.99	14,914.63	9.93
Cultural services	Aesthetic landscape	6818.20	4.50	6789.64	4.51	6811.94	4.51	6853.07	4.53	6783.20	4.51
	Total	151,371.31	100.00	150,629.53	100.00	151,190.49	100.00	151,353.22	100.00	150,267.71	100.00

).

The CS values across the land-use types were less than 1 in different periods (

Table 7. Sensitivity index of service value for each land-use type in Mu Us Sandy Land.

Change in Valuation Coefficient	1980		1990		2000		2010		2020
	%	CS	%	CS	%	CS	%	CS	%	CS
Farmland VC ± 50%	3.14	0.0628	3.16	0.0631	3.25	0.0651	3.13	0.0627	3.11	0.0622
Woodland VC ± 50%	2.21	0.0442	2.22	0.0444	2.26	0.0452	2.65	0.0530	2.76	0.0552
Grassland VC ± 50%	36.78	0.7356	36.93	0.7387	37.06	0.7411	37.53	0.7506	36.78	0.7355
Water VC ± 50%	6.52	0.1304	6.32	0.1265	6.13	0.1226	5.45	0.1089	6.07	0.1214
Built-up area VC ± 50%	0.00	0.0000	0.00	0.0000	0.00	0.0000	0.00	0.0000	0.00	0.0000
Unused land VC ± 50%	1.35	0.0271	1.37	0.0273	1.3	0.0261	1.24	0.0248	1.28	0.0257

). These results indicated that the total ESV estimated in this study area exhibits higher reliability. In addition, these findings suggested that changes in the corresponding VC had a relatively small effect on ESV in the Mu Us Sandy Land. The CS values of grassland and water were higher than those of other land-use types, which indicated that these two land types had a greater influence on ESV (Table 7).

3.3. Spatial Characteristics of the ESV

Selecting an appropriate scale as the evaluation unit is crucial for clarifying the spatial differentiation characteristics of ecosystem values in the study area. For ecosystems with high spatial heterogeneity, smaller evaluation units are typically employed to effectively capture the spatial heterogeneity of their value. In contrast, larger evaluation units are more suitable for ecosystems with low spatial heterogeneity to achieve the same purpose [10].

In this paper, the evaluation unit scale of 3 km × 3 km was chosen to analyze the spatial distribution characteristics of ESV in the study area. The ESV results in the study area were categorized into five classes using the natural fracture zone classification of ArcGIS 10.8 software (Figure 4). The spatial distribution of ESV did not change significantly during the 41 years. Among them, the distribution of patches in the waters of Red Alkali Nur, Dawa Nur, Batu Bay, and Ulanmulun Lake had high (CNY 35.01–66.19, ×10⁶) and extremely high ESV values (CNY 66.19–134.88, ×10⁶). The medium-ESV (CNY 20.75–35.01, ×10⁶) areas were mainly located in the northeast and northwest of the Mu Us Sandy Land, where the Yellow River and its direct current flowed through. And the ESV values of remaining unused land and some agricultural land were at low (CNY 12.82—20.75, ×10⁶) and extremely low levels (CNY 0.14–12.82, ×10⁶). Land degradation, often caused by irrational water and land-use practices, results in the shrinkage of water areas and the conversion of exposed surfaces into unused land, which directly diminishes the ecosystem service value (Table 4). Meanwhile, the large-scale reclamation of grasslands into farmland, although increasing the supply of food production, leads to a significant decline in other ecosystem service values (such as soil conservation, climate regulation, and biodiversity maintenance). This imbalance, where the loss of multiple key ecosystem services outweighs the gain from the enhanced food supply, ultimately contributes to the overall reduction in ESV [36,38]. The results of the autocorrelation analysis of ESV at the 3 km × 3 km scale from 1980 to 2020 showed that Moran’s I was greater than 0 for all five time periods, with a significance level of 0.00%, indicating a strong spatial positive correlation of ESV (Table 7). It could be seen that Moran’s I had the smallest value in 2010 and increased in 2020. Most of the grid cells were located within the low–low or high–high aggregation zones. Low and high ESV aggregates were observed (

Table 8. Moran’s I of ESV in the Mu Us Sandy Land from 1980 to 2020.

Year	1980	1990	2000	2010	2020
Moran’s I	0.46	0.47	0.46	0.42	0.45
Z (I)	91.64	93.51	92.37	84.75	89.70
p-value	0.00	0.00	0.00	0.00	0.00

, Figure 4).

3.4. Driving Factors of Spatial Heterogeneity in ESV

The factor detector revealed that EVS was significantly influenced by anthropogenic, natural, and landscape factors, which passed the 0.1% significance test (

Table 9. The top four factors of ESV variation in the Mu Us Sandy Land.

	Variables	Code	q Statistic	p Value
X11	Land-use intensity	LA	0.1483	0.000
X10	Human activity intensity	HAI	0.0764	0.000
X14	Shannon’s Diversity Index	SHDI	0.0632	0.000
X13	Landscape Division Index	DIVISION	0.0557	0.000

). In general, anthropogenic factors and landscape factors played large roles in influencing ESV. It could be seen that the q values of LA, HAI, SHDI, and DIVISION were higher (q > 0.05), which indicated that LA, HAI, SHDI, and DIVISION were dominant factors influencing the spatial differentiation of ESV in the Mu Us Sandy Land. On the contrary, the values of the other drivers were less than 0.05, which had small effects on ESV. The influence order of 14 factors (Table S1) was LA (q = 0.1483) > HAI (q = 0.0764) > SHDI (q = 0.0632) > DIVISION (q = 0.0557) > PRE (q = 0.0269) > GDP (q = 0.0232) > MSI (q = 0.0194) > TEM (q = 0.0191) > PE (q = 0.0149) > ELEV (q = 0.0131) > NDVI (q = 0.0109) > POP (q = 0.0088) > SLP (q = 0.0075) > NPP (q = 0.0056).

Applying the interaction detector to assess the degree of influence of the interaction of different driving factors on the spatial distribution pattern of ESV, the results showed that the q-values of the interactions among other driving factors were larger than those of the individual factors, which showed bi-factor enhancement or non-linear enhancement (

Table 10. Interactions among driving factors of ESV in the Mu Us Sandy Land. Green represented bi-factor enhancement of both factors, and the rest, without color coding, were non-linear enhancement.

	TEM	PRE	PE	ELEV	SLP	NDVI	NPP	POP	GDP	HAI	LA	MSI	DIVISION	SHDI
TEM	0.0191
PRE	0.0758	0.0269
PE	0.0489	0.0605	0.0149
ELEV	0.0664	0.0524	0.0581	0.0131
SLP	0.0375	0.0409	0.0416	0.0337	0.0075
NDVI	0.0450	0.0799	0.0475	0.0470	0.0305	0.0109
NPP	0.0427	0.0682	0.0367	0.0395	0.0224	0.0445	0.0056
POP	0.0354	0.0514	0.0397	0.0358	0.0257	0.0354	0.0267	0.0088
GDP	0.0506	0.0704	0.0502	0.0462	0.0458	0.0578	0.0499	0.0393	0.0232
HAI	0.1368	0.1427	0.1123	0.1136	0.0920	0.1156	0.1083	0.1110	0.1174	0.0764
LA	0.1680	0.1790	0.1678	0.1733	0.1639	0.1732	0.1727	0.1708	0.1731	0.2904	0.1483
MSI	0.0420	0.0495	0.0369	0.0352	0.0297	0.0365	0.0314	0.0355	0.0483	0.1015	0.1681	0.0194
DIVISION	0.0836	0.0811	0.0733	0.0782	0.0719	0.0869	0.0756	0.0693	0.0849	0.1530	0.1974	0.0613	0.0557
SHDI	0.0963	0.0953	0.0889	0.0886	0.0803	0.0998	0.0876	0.0783	0.0957	0.1494	0.2001	0.0732	0.0741	0.0632

). The results indicated that the strongest interaction was between the LA and HAI, with a value of q = 0.2904. The interaction between LA and HAI with other factors significantly affected ESV, with mean q values exceeding 0.15. The spatial differentiation of regional ecosystem services (ESVs) resulted from the combined effects of anthropogenic, natural, and landscape factors, with anthropogenic influences being the most significant.

4. Discussion

4.1. Changes in LULC and ESV

Complex dynamic transformations had occurred between and within different land-use categories. This transformation mainly occurred in regions where human activities were concentrated (Table 3, Figure 3). This finding corresponded to the main characteristics of urbanization development in the arid zone of northern China [7,43,44]. The significant expansion of forestland and built-up land is primarily due to human-driven ecological restoration measures such as afforestation in the arid regions of the West since the end of the last century and the need for urban development and expansion (Figure 5) [45,46]. The implementation of a series of forestry and ecological projects, such as the “Three-North” Shelter Forest Program (TNSFP), Grain for Green Project (GGP), and Natural Forest Protection Project (NFPP), has facilitated the restoration of forested land in the Mu Us Sandy Land [3,25]. According to China’s annual forestry statistical yearbooks, the Mu Us Sandy Land Forest coverage rate ranged from 7.2% in 1980 to 32.92% in 2020. Since 1980, significant changes in unused land, built-up areas, and woodland have highlighted the increasing human impact on the ecological system in the region.

Changes in LULC have markedly affected ESV worldwide and the surface terrestrial biogeochemical cycle [47]. From 1980 to 2020, increases and decreases in ESV primarily resulted from woodlands, grasslands, and water, showing a fluctuating trend of decreasing, then increasing, and then decreasing (Table 5). Grassland was the dominant land-use type, accounting for more than 57% of the total, and the water had the highest ecosystem service value (156,292.17 CNY/hm²) per unit area. The reduction of grasslands and water bodies primarily caused the decline in ecosystem service values and led to a corresponding decrease in water resource supply and hydrological regulation services. Similar findings were obtained from ecosystem service studies in arid and semi-arid regions [25].

Among the different ecosystem types, woodland, grassland, and water had higher ESV values in the existing research, which were decisive for the increase or decrease in ESV in the region [15]. The grassland occupied the most extensive area, and water had the highest ESV per unit area. Changes to grassland and water ecosystems significantly influence the total ecosystem service value of the region, more so than alterations in other land-use types. The land-use transfer matrix results showed that the degradation of grasslands and the reduction in water body areas are the primary causes for the decline in ecosystem service value in this region. We observed that grassland types deteriorated into unused land, construction land, and cropland. Due to climate change and human activities, the water area in the Mu Us Sandy Land has decreased, with the exposed surface being in an unutilized state, which seriously affects the biodiversity of the regional ecosystem and ecosystem services. Unconstrained high-intensity development of land resources reduced regional ESV in the long run by decreasing the contribution of individual areas. The loss of services provided by natural ecosystems can lead to long-term degradation of environmental quality [48]. Overall, the total amount of ecosystem services in the study area showed a significant downward trend from 1980 to 2020 due to excessive human land use (Table 5). To address land degradation and the decline of ecosystem services, the government has initiated focused afforestation projects on degraded land in the area (Figure 5). While this measure has effectively increased the area of forest land and promoted a certain improvement in the value of regional ecosystem services, it has not yet recovered to the level observed at the initial stage of the study. Realizing synergies between regional economic development and ecological protection requires integrating regional socio-economic and natural resources to enhance the quality of the terrestrial ecosystem. To reduce uncontrolled development, it is essential to safeguard land and water sources while preventing land degradation caused by improper use.

4.2. Spatial Characterization and Driving Mechanisms of ESV

The effectiveness of ecosystem services varies by area or scale, with the same services functioning differently across different locations [49,50]. This study utilized a 3 km × 3 km grid to capture more detailed small-scale variations and quantitatively analyze the drivers of ecosystem service value (ESV) in the study area. The factor detection results indicate that LA most strongly affects ESV, followed by HAI. This finding is consistent with existing research on the mechanisms driving the value of ecosystem services [10,51,52]. Human activities and land use alter spatial layouts and contribute to ecosystem degradation through overcultivation and overgrazing, ultimately impacting the provision of ecosystem services [13,53,54]. There is a significant interaction between ecosystem services and geodiversity; however, current management of ecosystem services often neglects the role of geodiversity in maintaining these functions and services [6,55]. The landscape factors SHDI and DIVISION significantly influenced ESV, following LA and HAI, with contributions of 6.32% and 5.57%, respectively (Table 9). The findings suggested that changing landscape patterns could lead to fluctuations and changes in ecosystem service values. Domestic and foreign scholars on landscape research found that landscape diversity could improve the stability of ecosystem services and resistance to disturbance, and landscape fragmentation and isolation would lead to a decrease in the value of ecosystem services [56,57,58]. The results of the interaction detector analysis showed that the anthropogenic, natural, and landscape factors combined to create spatial variation in ESV, among which anthropogenic factors were the most significant (Table 10). Therefore, future development should prioritize conservation, control anthropogenic activities such as land development and utilization, reduce the degree of landscape fragmentation, enhance landscape diversity, and account for the importance of geodiversity in the ecosystem’s sustainable management to achieve the beautiful vision of synchronized economic and social development with ecological protection.

4.3. Limitations and Recommendations

This study still has some limitations. First, the evaluation of ESV based on the equivalent factor method relies heavily on land-use data, which may exaggerate the influence of land-use type on ESV to a certain extent, and attempts should be made to use other ESV evaluation methods to further validate the importance of different drivers at a later stage. Second, the study utilized a grid scale with a resolution of 3 km × 3 km to analyze the drivers of ecosystem service value. However, the driving influences and interactions at different spatial scales are unclear, and the scale effect raises the uncertainty of the study’s conclusions. Future research should examine the drivers of ESV at various spatial scales. The population density dataset is derived from comprehensive interpolation using nighttime light datasets and other factors. Due to its limited data accuracy, its impact on ecosystem service value may be underestimated [10]. While the land-use classification data have a relatively high level of accuracy, they may still introduce some uncertainties into the research findings. In addition, this study has not considered the impact of policy factors on ecosystem service valuation (ESV). Therefore, future research should comprehensively address the influence of policy factors. Finally, geodiversity is threatened by human activities and environmental changes, further affecting the quality of ecosystem service values. However, it is challenging to quantify the scale of threats, from human activities to geodiversity, and the services they provide [55]. Threats such as urbanization, overgrazing, and land-use changes may disrupt geodiversity. The quality of ecosystem services in the same ecosystem type is dynamically changing in space and time, and this study has not quantified the spatiotemporal heterogeneity of ecosystem service values.

5. Conclusions

In this paper, the equivalent factor method was employed to evaluate the ESV in the Mu Us Sandy Land, and the spatial and temporal distribution characteristics of the ESV were analyzed. Based on this, we quantified the contributions of anthropogenic, natural, and landscape factors driving ecosystem service values (ESVs) using the GeoDetector model. The results showed that the land-use structure of the study area was generally stable, and land-use changes occurred in areas with intensive human activities. During the 41-year study period, the woodland and built-up land area continued to increase, while the area of all other land-use types decreased. The total area of arable land and grassland did not change much. However, significant conversions occurred between land parcels at different development levels, accompanied by an overall negative development trend, indicating a general degradation trend in the regional ecological environment. Services provided by grassland and water dominated the fluctuations in ESV across the study area. Among all services, regulatory services contributed the most to ESV. The ESV in the study area exhibited significant positive spatial autocorrelation, featuring the co-occurrence of high-value and low-value clusters across multiple spatial scales. The synergistic effects of human activities, landscape pattern changes, and natural factors led to spatial differences in ESV. Future efforts focus on strengthening the protection and management of ecological resources, the strict implementation of land-use planning, the targeted establishment of nature reserves, the scientific implementation of vegetation restoration projects, etc., to effectively improve the condition of sandy ecosystems and enhance their stability and resistance. Future sustainable ecosystem management requires strict enforcement of land-use planning, strategic establishment of nature reserves, scientific implementation of vegetation restoration programs, and explicit integration of geodiversity’s critical supporting role into ecosystem restoration frameworks. These measures will effectively improve the health of sandy ecosystems and enhance their stability and resilience to disturbances.