Spatiotemporal Evolution Characteristics and Driving Forces of Ecosystem Services in the Heihe River Basin Under Ecological Water Conveyance

WANG Jiawei¹, DONG Guotao², JIANG Xiaohui¹, NIE Tong¹, LI Yuehong¹
¹College of Urban and Environmental Sciences, Northwest University, Xi'an, Shaanxi, China
²Heihe Water Resources and Ecological Protection Research Center, Lanzhou, Gansu, China

Abstract

The implementation of ecological water diversion policies in the Heihe River Basin has effectively alleviated the trend of ecological environmental degradation, resulting in significant improvements in environmental quality. However, existing research on ecosystem services in this basin has been limited by short study periods, restricted regional scope, and a lack of qualitative and quantitative analysis of the impacts of environmental governance policies and other driving factors. This study examines the spatiotemporal evolution characteristics of water yield depth, habitat quality, carbon storage, and soil conservation in the Heihe River Basin from 1990 to 2022. The InVEST-PLUS coupled model is employed to evaluate the impact of ecological water conveyance on downstream ecosystem services, and geographic detectors are used to analyze the driving factors. The results indicate that: (1) After 2000, carbon storage and habitat quality in the basin showed an overall increasing trend, while water yield depth and soil conservation exhibited a pattern of initial increase followed by decrease. Spatially, ecosystem services displayed a stepped distribution pattern of "high in the south, low in the north," with high values concentrated in the Qilian Mountains and low values distributed in the desert zones of the middle and lower reaches. (2) Downstream carbon storage and habitat quality were significantly positively correlated with the annual runoff at Zhengyi Gorge (P<0.05). Compared with natural development scenarios, ecological water conveyance caused both indicators to show a year-by-year increasing trend under actual conditions. (3) Among all driving factors in the geographic detector, digital elevation, temperature, precipitation, and potential evapotranspiration were dominant. The results also demonstrated that the interactive explanatory power of factors on ecosystem services was higher than that of single factors. These findings provide a scientific basis for ecological governance and water resource allocation in the Heihe River Basin.

Keywords: ecosystem services; InVEST-PLUS coupled model; driving forces research; Heihe River Basin

1 Introduction

1.1 Study Area Overview

The Heihe River Basin is located in the central part of the Hexi Corridor, with geographical coordinates between 38°~42°N and 98°~101°E. As the "mother river" of the Hexi Corridor, the basin covers a total area of approximately 14.29×10⁴ km² and spans Qinghai, Gansu, and Inner Mongolia provinces. From south to north, it encompasses semi-arid, arid, and extremely arid climate zones, making it the second largest inland river basin in northwest China. The downstream area of the Heihe River Basin is predominantly desert. This study focuses on the river-lake coastal areas and oases in the downstream region to analyze the spatiotemporal distribution of ecosystem services. By establishing a buffer zone along the downstream river coast and removing most desert areas, the study area schematic diagram is obtained [FIGURE:1]. The basin is divided into upper, middle, and lower reaches from south to north. The terrain slopes downward from south to north, with the Qilian Mountains in the southern upstream area, plateaus and plains distributed in the middle reaches, and deserts dominating the downstream area. The Zhangye Oasis in the middle reaches and the Ejina Oasis in the downstream area serve as important ecological barriers preventing the expansion of the Badain Jaran Desert and reducing sandstorm disasters.

1.2 Data Sources

Table 1 [TABLE:1] presents the primary data used in this study, including resolution and sources. Monthly average precipitation, monthly potential evapotranspiration, monthly temperature, and land use type data were obtained from the National Tibetan Plateau Data Center (https://data.tpdc.ac.cn) at 1 km×1 km resolution. Land use data utilized the GLC_FCS30D dataset, reclassified into six categories (cropland, forestland, grassland, etc.) at 1 km×1 km resolution. Soil data were sourced from the Food and Agriculture Organization of the United Nations at 1 km×1 km resolution. Digital elevation model (DEM) data were obtained from the Geospatial Data Cloud at 90 m×90 m resolution. Root depth data were derived from Scientific Data (Nature) at 1 km×1 km resolution. Normalized Difference Vegetation Index (NDVI) data were obtained from the Science Data Bank at 30 m×30 m resolution. Population density and gross domestic product (GDP) data were provided by Beijing Shugu Information Technology Co., Ltd. at 1 km×1 km resolution.

1.3.1 InVEST Model

The InVEST (Integrated Valuation of Ecosystem Services and Tradeoffs) model, developed by the Natural Capital Project, is a comprehensive assessment tool applied to evaluate various ecosystem services. This study employs the InVEST model to assess four ecosystem services in the Heihe River Basin: water yield depth, carbon storage, habitat quality, and soil conservation.

Water Yield Module: The InVEST water yield module primarily uses annual average precipitation and actual evapotranspiration for calculation and simulation. The main formula is:

$$Y_x = \text{AET}_x$$

where $Y_x$ is the annual water yield depth value for a grid cell (mm), and $\text{AET}_x$ is the annual actual evapotranspiration for a grid cell (mm).

Habitat Quality Module: The InVEST habitat quality module estimates habitat quality based on land use data. The main formula is:

$$Q_{xj} =$$

where $Q_{xj}$ is the habitat quality (dimensionless, ranging from 0 to 1) of pixel $x$ in land use type $j$; $H_j$, $k$, $D_{xj}$, and $z$ represent habitat suitability factor, half-saturation constant, habitat degradation degree, and normalization constant, respectively. In this study, the $z$ value is set to 2.5.

Soil Conservation Module: The InVEST soil conservation module is based on the Revised Universal Soil Loss Equation (RUSLE). The main formulas are:

$$\text{usle}_x = R_x \times \text{rlkls}_x = C_x \times K_x \times LS_x$$
$$R_x \times \text{sc}_x = \text{rlkls}_x - \text{usle}_x$$

where $\text{usle}_x$ is the actual soil erosion amount (t·km⁻²·yr⁻¹); $\text{rlkls}_x$ is the potential soil erosion amount (t·km⁻²·yr⁻¹); $P_x$, $LS_x$, $R_x$, $C_x$, and $K_x$ represent vegetation cover and crop management factor, slope length and steepness factor, rainfall erosivity factor, soil conservation practice factor, and soil erodibility factor, respectively.

Carbon Storage Module: The InVEST carbon storage module calculates carbon storage by integrating land use data and carbon pool data. The main formula is:

$$C_{\text{total}} = C_{\text{above}} + C_{\text{below}} + C_{\text{soil}} + C_{\text{dead}}$$

where $C_{\text{total}}$, $C_{\text{above}}$, $C_{\text{below}}$, $C_{\text{soil}}$, and $C_{\text{dead}}$ represent total carbon storage, aboveground carbon storage, belowground carbon storage, soil carbon storage, and carbon storage in dead organic matter, respectively (Mg·hm⁻²).

1.3.2 Spatial Characteristic Analysis

This study employs global Moran's I index to test spatial autocorrelation, while local Moran's I index measures the correlation between the attribute of a unit region and the same attribute values of neighboring regions. Using ArcGIS Pro, a 2 km×2 km fishnet was created to extract mean values of various ecosystem services through zonal statistics, which were then visualized as LISA (Local Indicators of Spatial Association) cluster maps.

The principle and calculation formulas are as follows:

$$I_{\text{global}} = \frac{n}{S_0} \frac{\sum_i \sum_j w_{ij} z_i z_j}{\sum_i z_i^2}$$

$$I_{\text{local}} = \frac{z_i}{m_2} \sum_j w_{ij} z_j$$

where $I_{\text{global}}$ is the global Moran's I index; $I_{\text{local}}$ is the local Moran's I index; $n$ is the total number of features; $i$ represents a certain ecosystem service of pixel unit; $z_i$ is the deviation of feature $i$'s attribute from its corresponding mean; $z_j$ is the deviation of feature $j$'s attribute from its corresponding mean; and $w_{ij}$ is the spatial weight between features $i$ and $j$. The value of $I$ ranges from [-1, 1]. A positive $I$ value indicates global positive correlation, while a negative value indicates global negative correlation. An $I$ value of 0 indicates random distribution, and the larger the absolute value of $I$, the greater the spatial heterogeneity.

1.3.3 PLUS Model

The PLUS (Patch-generating Land Use Simulation) model is a cellular automaton model that simulates land use changes at the patch scale. It primarily consists of two modules: Land Expansion Analysis Strategy (LEAS) and a Cellular Automata (CA) model. The InVEST model operates based on land use data, which significantly influences model assessment results. Previous studies have confirmed the feasibility of coupling InVEST with PLUS to assess ecosystem services under different scenarios. The model requires inertia coefficients for different land use types, which can be obtained through the LEAS module to achieve optimal parameter selection.

1.3.4 Geographic Detector

The geographic detector, proposed and developed by Wang Jinfeng's team, is an emerging spatial analysis tool for analyzing spatial heterogeneity and its influencing factors. It can effectively avoid endogeneity issues in regression analysis, prevent multicollinearity effects, and qualitatively analyze driving factors while identifying their pairwise interactions. This study uses the geographic detector to analyze driving factors of ecosystem service changes in the Heihe River Basin.

(1) Factor Detection: This function analyzes the spatial heterogeneity of factor Y and measures the explanatory power of factor X on the spatial differentiation of Y (q-value). The expression is:

$$q = 1 - \frac{\sum_{h=1}^{L} N_h \sigma_h^2}{N \sigma^2}$$

where $h$ is the stratum of factor X and variable Y; $N_h$ is the number of units in stratum $h$; $N$ is the total number of units; $\sigma_h^2$ is the variance of Y values in stratum $h$; and $\sigma^2$ is the variance of Y values in the entire study area. The q-value ranges between [0, 1], with values closer to 1 indicating stronger explanatory power of factor X on the spatial heterogeneity of factor Y.

(2) Interaction Detection: This function determines whether two factors have interactive effects and identifies the characteristics of these interactions by calculating and comparing the explanatory power of individual factors and factor pairs.

Since the ecosystem stability in the study area is relatively low and sensitive to natural factors such as climate, vegetation, and human activities, and drawing on driving force analysis results from the Yellow River Basin, this study selected precipitation (PRE), potential evapotranspiration (PET), temperature (TEM), normalized difference vegetation index (NDVI), digital elevation model (DEM), population distribution (POP), and gross domestic product (GDP) as driving factors.

2 Results

2.1.1 Temporal Variation Characteristics

Analysis of temporal changes in ecosystem services across the study area (Figure 2 [FIGURE:2]) reveals that water yield depth showed an initial increase followed by a decreasing trend, with the highest average value of 58.3 mm occurring in 2018 and the lowest value of approximately 19.8 mm in 2022. Carbon storage exhibited a continuous growth trend overall (Figure 2 [FIGURE:2]), reaching its lowest value of about 133.7 t·hm⁻² in 1990 and peaking at 135.4 t·hm⁻² in 2022. Habitat quality was at its lowest in 1990 at approximately 0.457, reached its highest level of about 0.474 in 2000, and remained significantly higher after 2000 than before, showing a fluctuating pattern (Figure 2 [FIGURE:2]). Soil conservation displayed similar fluctuation patterns, peaking at 4574.2 t·hm⁻² in 2018 and reaching its lowest value of about 2473.1 t·hm⁻² in 1990.

To further understand the spatial relationships and changes in ecosystem services at the local scale, the study area was divided into subregions for statistical analysis, revealing temporal changes in ecosystem services in the upper, middle, and lower reaches across years (Table 2 [TABLE:2]). In terms of spatial differentiation characteristics, ecosystem services in the same year showed a gradient decreasing pattern (upper > middle > lower). Due to the high proportion of desert area in the lower reaches, its service supply capacity was significantly limited. The temporal evolution trends of various services in the upper and middle reaches were consistent with the overall basin pattern, showing synchrony. Downstream carbon storage and habitat quality showed continuous growth trends, while water yield depth and soil conservation showed initial increase followed by decrease.

2.1.2 Spatial Distribution Characteristics

Based on the spatial distribution and changes of ecosystem services from 1990 to 2022 (Figure 3 [FIGURE:3]), the study area exhibited the following patterns: For water yield depth, high-value areas (≥84 mm) were continuously concentrated in the upstream Qilian Mountains, with a maximum value of 403 mm; values in the middle and lower reaches remained stable below 84 mm, with no significant fluctuations in desert areas. The spatial pattern showed no significant change, with only local area increases reaching up to 409 mm. High-value areas of soil conservation (24.2×10³ t·km⁻²·yr⁻¹) were concentrated in the upper and middle reaches, with maximum absolute change of 13.36×10³ t·km⁻²·yr⁻¹. Carbon storage distribution in the middle reaches presented a scattered pattern, with a maximum change value of 31.3 t·hm⁻², generally higher than 22.3 t·hm⁻² in the upper and middle reaches, while downstream desert areas mostly showed zero values. Habitat quality high-value areas expanded in 2022 compared to 1990, further strengthening the south-high-north-low gradient pattern, with overall increases showing upstream > middle > downstream.

The global Moran's I index for ecosystem services from 1990 to 2022 (Table 3 [TABLE:3]) shows that the P-values for all ecosystem services were less than 0.05, passing significance tests, and all showed extremely strong positive spatial autocorrelation, indicating that ecosystem services have significant clustering characteristics in space. Compared with 1990, the global Moran's I for water yield depth in 2022 increased substantially, while those for soil conservation, carbon storage, and habitat quality decreased to varying degrees.

Local LISA cluster analysis of the study area produced cluster maps (Figure 4 [FIGURE:4]). Water yield depth, soil conservation, and carbon storage showed high-high clusters mainly distributed in the southern Qilian Mountains, while low-low clusters were primarily distributed in the northern desert belt, with central transition zones mostly showing non-significant clusters. Soil conservation, carbon storage, and habitat quality showed non-significant clustering distribution around southern rivers and lakes. From 1990 to 2022, the area of low-low clusters for water yield depth and soil conservation in the north increased, while high-high clusters for habitat quality in the south shifted southwestward.

Statistical analysis of spatial heterogeneity characteristics of ecosystem services (Table 4 [TABLE:4]) shows that for high-high cluster areas, water yield depth proportion increased from 12.94% (1990) to 25.34% (2022), habitat quality increased from 12.69% (1990) to 23.62% (2022), while soil conservation and carbon storage remained stable. For low-low cluster areas, soil conservation proportion decreased from 44.48% (1990) to 38.84% (2022), carbon storage and habitat quality decreased by 39.76% and 0.77% respectively, while water yield depth showed little change.

2.2.1 Impact of Ecological Water Conveyance on Downstream Ecosystem Services

Considering the impact of water allocation policies on the downstream area, the relationship between water discharge at Zhengyi Gorge and downstream ecosystem services was analyzed. Based on the evolution of water allocation policies, annual runoff changes at Zhengyi Gorge from 1990 to 2022 were divided into four periods: baseline period (before 2000), emergency scheduling period (2000-2005), conventional scheduling period (2006-2017), and ecological scheduling period (2018-2022). With the transformation of different water diversion modes, the discharge at Zhengyi Gorge showed a stage-by-stage increasing trend. During the baseline period before ecological water conveyance, the maximum annual runoff occurred in 1990 at 7.96×10⁸ m³. The emergency scheduling period peaked in 2003 at 1.216×10⁹ m³. The conventional scheduling period showed strong fluctuation trends, with a peak of 1.128×10⁹ m³ in 2016. The ecological scheduling period peaked in 2020 at 1.592×10⁹ m³, with an overall trend of initial increase followed by decrease.

Combined with changes in precipitation and potential evapotranspiration (Figure 5 [FIGURE:5]), the study area maintained high levels of potential evapotranspiration with small fluctuation amplitude. During the baseline period before 2000, changes in Zhengyi Gorge runoff were basically consistent with natural precipitation trends. During the emergency scheduling period, runoff changes showed some lag compared to precipitation changes. During the conventional and ecological scheduling periods, runoff changes showed opposite trends to precipitation in some years—for example, precipitation decreased significantly in 2018 while Zhengyi Gorge runoff increased slightly. These results indicate that after the implementation of ecological water conveyance projects in the lower Heihe River Basin, the discharge at Zhengyi Gorge under human regulation has reduced the impact of natural condition changes to some extent, becoming a comprehensive factor affecting the downstream ecological environment. Therefore, to distinguish it from natural conditions and human activities, it is treated as a separate driving factor.

Correlation analysis between annual average runoff at Zhengyi Gorge and various ecosystem services (Table 5 [TABLE:5]) shows that the Pearson and Spearman correlation coefficients between Zhengyi Gorge runoff and downstream carbon storage and habitat quality were both above 0.5 and passed significance tests (Table 6 [TABLE:6]), indicating significant positive correlations. The correlation between runoff changes and habitat quality and carbon storage changes was further verified by quantifying changes in habitat quality and carbon storage in the lower Heihe River under ecological water conveyance scenarios. Using the InVEST-PLUS coupled model, a natural development scenario was simulated and compared with actual regulation scenarios to assess the effects of ecological water conveyance. The 2020 land use prediction results were used for model calibration, showing a Kappa coefficient of 0.82 and overall accuracy of 0.88, meeting reliability requirements for spatial simulation results.

Comparison of simulated and actual downstream carbon storage and habitat quality from 1990 to 2022 (Table 7 [TABLE:7]) shows that under ecological water conveyance and actual conditions, both carbon storage and habitat quality in the lower Heihe River showed continuous increasing trends, while under natural development scenarios they showed overall decreasing trends. The simulated results were significantly lower than actual results, particularly in 2022, when actual carbon storage was 35.16 t·hm⁻² compared to 30.39 t·hm⁻² under natural scenarios, and actual habitat quality was 0.473 compared to 0.457 under natural scenarios. These differences indicate that after ecological water conveyance implementation, carbon storage and habitat quality in the lower Heihe River Basin have improved period by period, and comparison with natural development scenarios reveals that environmental deterioration in the downstream area has been effectively curbed.

2.2.2 Geographic Detector Driving Force Analysis

Based on the factor explanatory power and ranking from the geographic detector model (Table 8 [TABLE:8]), analysis reveals that for water yield depth, precipitation (PRE) ranked first with a q-value of approximately 0.45, followed by potential evapotranspiration (PET). For soil conservation, DEM had the highest q-value of about 0.38, followed by precipitation. For carbon storage, precipitation showed the highest q-value of approximately 0.42, with temperature (TEM) ranking second. For habitat quality, temperature had the highest q-value of about 0.41, with precipitation ranking second. Overall, precipitation, temperature, potential evapotranspiration, and digital elevation were the main factors influencing the four ecosystem services. The q-values for NDVI and GDP in carbon storage and habitat quality were higher than those in water yield depth and soil conservation, indicating that NDVI and GDP have greater impacts on habitat quality and carbon storage.

Horizontal comparison shows that the explanatory power of any single driving factor was lower than that under dual-factor interaction (Figure 6 [FIGURE:6]). Vertical comparison reveals that the interaction between precipitation and other factors showed strong explanatory power. The interaction between precipitation and potential evapotranspiration (PRE∩PET) showed the strongest explanatory power for water yield depth. The interaction between precipitation and temperature (PRE∩TEM) showed the strongest explanatory power for soil conservation. The interaction between digital elevation and precipitation (DEM∩PRE) showed the strongest explanatory power for carbon storage and habitat quality.

3 Discussion

Previous research in this basin mostly focused on direct assessment of ecosystem services with limited service types, short time series, and restricted scope to sub-basins, with few studies analyzing the driving factors behind changes. This study systematically assessed ecosystem services across the Heihe River Basin over a relatively long time series. Water yield depth and soil conservation showed initial increase followed by decreasing trends. Habitat quality was generally low before 2000 but improved after 2000, showing fluctuating patterns. Carbon storage showed an overall increasing trend. The spatial distribution of ecosystem services is consistent with Wang et al.'s research, showing a stepped distribution from high in the south to low in the north, with downstream services mainly distributed along rivers, lakes, and oases. Moran's I spatial analysis results show significant spatial differentiation characteristics, with high values clustered in the southern Qilian Mountains and low values clustered in the northern desert areas.

Temporal changes in ecosystem services in the Heihe River Basin are closely related to natural conditions and policy implementation. The Qilian Mountains in the south block water vapor transport northward, and the basin's inland location under continental arid climate influences creates a chain of landforms from southeast to northwest: sequentially cold-wet mountains, dry plain oases, and extremely dry desert gobi. Therefore, differences in climate conditions, topography, and vegetation distribution are important reasons for the spatiotemporal distribution differences of ecosystem services. Additionally, to improve the ecological environment and socioeconomic quality in the middle and lower reaches, the government has implemented numerous policies, among which water allocation policies are crucial for improving the downstream ecological environment.

In the driving force analysis, focusing on the impact of ecological water conveyance downstream, water allocation policies increased the annual runoff at Zhengyi Gorge, which was significantly positively correlated with downstream habitat quality and carbon storage. Simulation of natural scenarios using the InVEST-PLUS coupled model compared with actual scenarios shows that after ecological water conveyance implementation, downstream carbon storage and habitat quality improved. In the geographic detector's dual-factor interaction analysis, the explanatory power after factor interaction was higher than single-factor explanatory power, indicating that ecological environment changes in the study area result from comprehensive effects of multiple driving factors. The strongest explanatory power for water yield depth came from PRE∩PET interaction, revealing that topographic gradients in the basin have significant amplification effects on water vapor redistribution. The strongest explanatory power for soil conservation came from PRE∩TEM interaction, highlighting that soil conservation is greatly affected by rainfall erosion. The strongest explanatory power for carbon storage and habitat quality came from DEM∩PRE interaction, as topography constrains vegetation distribution—different altitudes significantly affect vegetation distribution, which in turn affects carbon storage and habitat quality.

4 Conclusions

1) Carbon storage and habitat quality in the Heihe River Basin showed overall increasing trends after 2000, while water yield depth and soil conservation showed initial increase followed by decreasing trends. All ecosystem services exhibited significant spatial autocorrelation and differentiation characteristics, with high values mainly distributed in the southern Qilian Mountains and low values clustered in the northern desert areas, showing an overall south-high-north-low stepped distribution pattern.

2) Downstream carbon storage and habitat quality were significantly positively correlated with water discharge at Zhengyi Gorge. Simulated carbon storage and habitat quality under natural development scenarios showed逐年 decreasing trends, while actual scenarios under ecological water conveyance showed逐年 increasing trends.

3) In the geographic detector, digital elevation, temperature, precipitation, and potential evapotranspiration were dominant factors. The results showed that interactive driving effects of factors had greater impact on ecosystem services than single-factor effects.

5 Limitations and Outlook

Since 2000, the overall ecological environment of the Heihe River Basin has improved, which is closely related not only to natural conditions but also to policies such as water allocation and socioeconomic development. Therefore, studying the impact of water allocation on ecosystem services is significant, particularly the mechanisms behind these effects. However, due to data limitations, this study did not conduct year-by-year analysis of ecosystem service changes, resulting in insufficiently detailed temporal analysis. Additionally, the precision of land use type classification was not high enough, which may have affected results, and the underlying mechanisms of how influencing factors affect ecosystem service functions were not deeply explored. Future research should consider obtaining more comprehensive data, using higher-precision land use data for more accurate assessment of ecosystem services, while also considering trade-offs and synergies among ecosystem services and further investigating the mechanisms of influencing factors.

References

[1] Kang Tingting, Li Zeng, Gao Yanchun. Effectiveness of ecological restoration in the mountain oasis desert system of northwestern arid area of China[J]. Acta Ecologica Sinica, 2019, 39(20): 7418-7431.

[2] Zhao Jun, Ma Xiaoping, Wei Wei. Vegetation succession and ecological changes in the Heihe River watershed over the past 50 years[J]. Acta Prataculturae Sinica, 2014, 23(5): 61-68.

[3] Ma Z, Kang S, Zhang L, et al. Analysis of impacts of climate variability and human activity on streamflow for a river basin in arid region of northwest China[J]. Journal of Hydrology, 2008, 352(3/4): 239-249.

[4] Jiang Xiaohui, Xia Jun, Huang Qiang, et al. Adaptability analysis of the Heihe River water diversion scheme[J]. Acta Geographica Sinica, 2019, 74(1): 103-116.

[5] Zhang L, Dawes W R, Walker G R. Response of mean annual evapotranspiration to vegetation changes at catchment scale[J]. Water Resources Research, 2001, 37(3): 701-708.

[6] Daily G C. Nature's services: Societal dependence on natural ecosystems[M]. Washington: Island Press, 2013: 454-464.

[7] Costanza R, D'Arge R, De Groot R, et al. The value of the world ecosystem services and natural capital[J]. Nature, 1997, 387(6630): 253-260.

[8] Cao Qiwen, Wei Xiaomei, Wu Jiansheng. A review on the tradeoffs and synergies among ecosystem services[J]. Chinese Journal of Ecology, 2016, 35(11): 3102-3111.

[9] Zhang Jingjing, Zhu Wenbo, Zhu Lianqi, et al. Multi-scale analysis of trade-off/synergy effects of forest ecosystem services in the Funiu Mountain region[J]. Acta Geographica Sinica, 2020, 75(5): 975-988.

[10] Zhang Fuping, Li Xiaojuan, Feng Qi, et al. Spatial and temporal variation of water conservation in the upper reaches of Heihe River Basin based on InVEST model[J]. Journal of Desert Research, 2018, 38(6): 1321-1329.

[11] Li Fang, Zhang Jinlong, Yang Huan. Simulation of annual water yield in the upper Heihe River Basin from 1990 to 2018 based on InVEST[J]. Plateau Meteorology, 2022, 41(3): 698-707.

[12] Wang Bei, Zhao Jun, Hu Xiufang. Changes of spatial synergies or trade-offs of ecosystem services in Heihe River Basin[J]. Chinese Journal of Ecology, 2022, 41(3): 580-588.

[13] Wang Geng, Feng Yan. Spatiotemporal variation and scenario prediction of ecosystem service trade-offs/synergies in the Taizi River Basin, Liaoning Province[J]. Acta Ecologica Sinica, 2024, 44(1): 96-106.

[14] Zhu Chunxia, Zhong Shaozhuo, Long Yu, et al. Spatiotemporal variation of ecosystem services and their drivers in the Yellow River Basin, China[J]. Chinese Journal of Ecology, 2023, 42(10): 2502-2513.

[15] Han Wuhong. Evaluation of ecosystem services in Qilian Mountain National Park based on InVEST model and its driving forces[D]. Lanzhou: Northwest Normal University, 2022.

[16] Xu Baorong, Liu Yichuan, Dong Ying, et al. Evaluation of habitat quality in Lanzhou region based on InVEST model[J]. Journal of Desert Research, 2021, 41(5): 120-129.

[17] Bao Yubin, Li Ting, Liu Hui, et al. Spatial and temporal changes of water conservation of Loess Plateau in northern Shaanxi Province by InVEST model[J]. Geographical Research, 2016, 35(4): 664-676.

[18] Zhang Hua, Han Wuhong, Song Jinyue, et al. Spatial temporal variations of habitat quality in Qilian Mountain National Park[J]. Chinese Journal of Ecology, 2021, 40(5): 1419-1430.

[19] Li Cheng, Zhao Jie, Zhuang Zhicheng, et al. Spatiotemporal dynamics and influencing factors of ecosystem service trade-offs in the Yangtze River Delta urban agglomeration[J]. Acta Ecologica Sinica, 2022, 42(14): 5708-5720.

[20] He Jiaying, Jiang Xiaohui, Lei Yuxin. Effects of ecological engineering on spatio-temporal changes of key ecosystem services on the Loess Plateau: A case study in the Yanhe River Basin, China[J]. Acta Ecologica Sinica, 2023, 43(12): 4823-4834.

[21] Fu B J, Zhao W W, Chen L D, et al. Assessment of soil erosion at large watershed scale using RUSLE and GIS: A case study in the Loess Plateau of China[J]. Land Degradation & Development, 2005, 16(1): 73-85.

[22] Jiang Xiaofang, Duan Hanchen, Liao Jie, et al. Land use in the Gan-Gao region of middle reaches of Heihe River Basin based on a PLUS-SD coupling model[J]. Arid Zone Research, 2022, 39(4): 1246-1258.

[23] Chen Tongyao, Jia Yanfeng, Wang Jianan, et al. Current situation and function of soil conservation in National Nature Reserves in the Qilian Mountains based on InVEST model[J]. Arid Zone Research, 2020, 37(1): 150-159.

[24] Qin Zheng, Gao Yuxiao, Sarsenbay Samha. Multi-scenario land use change simulation and carbon storage assessment in Jinan City based on PLUS-InVEST model[J]. Yellow River, 2024, 46(5): 117-122.

[25] Wang J F, Zhang T L, Fu B J. A measure of spatial stratified heterogeneity[J]. Ecological Indicators, 2016, 67: 250-256.

[26] Meng Bin, Wang Jinfeng, Zhang Wenzhong, et al. Evaluation of regional disparity in China based on spatial analysis[J]. Scientia Geographica Sinica, 2005, 25(4): 393-400.

[27] Liang X, Guan Q F, Clarke K C, et al. Understanding the drivers of sustainable land expansion using a patch generating land use simulation (PLUS) model: A case study in Wuhan, China[J]. Computers Environment and Urban Systems, 2021, 85: 101569, doi:10.1016/j.compenvurbsys.2020.101569.

[28] Xue Xiaoyu, Wang Xiaoyun, Duan Hanming, et al. Multi-scenario simulation of land use change and its impact on ecosystem services in Beijing-Tianjin-Hebei region based on the FLUS-InVEST model[J]. Acta Ecologica Sinica, 2023, 43(11): 4473-4487.

[29] Ren Yinming, Liu Xiaoping, Xu Xiaocong, et al. Analysis on spatiotemporal evolution of habitat quality in Qilian Mountains based on land use change[J]. Bulletin of Soil and Water Conservation, 2020, 40(2): 278-284, 325.

[30] Wang Ya, Meng Jijun. Effects of land use change on ecosystem services in the middle reaches of the Heihe River Basin[J]. Arid Zone Research, 2017, 34(1): 200-207.

[31] Hu Feng, Zhang Yan, Guo Yu, et al. Spatial and temporal changes in land use and habitat quality in the Weihe River Basin based on the PLUS and InVEST models and predictions[J]. Arid Land Geography, 2022, 45(4): 1125-1136.

[32] Fang Kuangnan, Wu Jianbin, Zhu Jianping, et al. A review of technologies on random forests[J]. Journal of Statistics and Information, 2011, 26(3): 7.

[33] Breiman L. Random forests[J]. Machine Learning, 2001, 45: 5-32.

[34] Wang Jinfeng, Xu Chengdong. Geodetector: Principle and prospective[J]. Acta Geographica Sinica, 2017, 72(1): 116-134.

[35] Li Chuanzhe, Yu Fuliang, Liu Jia, et al. Research on land use/cover change and its driving force in midstream of the Heihe Mainstream Basin during the past 20 years[J]. Journal of Natural Resources, 2011, 26(3): 353-363.

[36] Huang Xueyu, Xiu Lina, Lu Zhixiang. Effects and driving factors of ecosystem service trade-offs in the Longdong Loess Plateau, China[J]. Arid Land Geography, 2025, 48(3): 480-493.

[37] Thevs N, Peng H, Rozi A, et al. Water allocation and water consumption of irrigated agriculture and natural vegetation in the Aksu-Tarim River Basin, Xinjiang, China[J]. Journal of Arid Environments, 2015, 112: 87-97.

[38] Yan F, Shangguan W, Zhang J, et al. Depth to bedrock map of China at a spatial resolution of 100 meters[J]. Scientific Data, 2020, 7(1): 2, doi:10.5194/essd-2018-103.

[39] Zhang Yichi, Yu Jingjie, Qiao Maoyun, et al. Effects of eco-water transfer on changes of vegetation in the lower Heihe River Basin[J]. Journal of Hydraulic Engineering, 2011, 42(7): 757-765.

[40] Shi Ying, Bie Qiang, Su Xiaojie, et al. Spatiotemporal variation of water conservation function evaluation based on InVEST model: A case of Lanzhou City[J]. Arid Land Geography, 2024, 47(9): 1518-1529.