Integrating Future Multi-Scenarios to Evaluate the Effectiveness of Ecological Restoration: A Case Study of the Yellow River Basin

Abstract

Ecological restoration is an important strategy for mitigating environmental degradation, and the effectiveness evaluation of ecological restoration is of profound significance for the scientific implementation of restoration projects. This study improved the Patch-generating Land Use Simulation (PLUS) model. It was used to simulate the land use patterns under multi-scenarios such as natural development (ND), economic priority (EP), and ecological restoration (ER) in 2030. An evaluation framework covering ecological “Restoration–Monitoring–Effectiveness” (RME) was proposed. Based on 30 m high-resolution remote-sensing data from 2000 to 2020, the land use distribution, landscape pattern changes, and ecosystem services under different scenarios were evaluated and predicted in the Yellow River Basin of Sichuan to verify the effectiveness of the evaluation framework. The results showed the following: (1) Under the ER scenario, the transfer of land use types in 2020–2030 was mainly characterized by an increase in the area of wetlands and a decrease in the area of built-up land. (2) There were obvious differences in land use and landscape patterns under different scenarios. Compared with the ND and EP scenarios, the growth of the construction rate was suppressed in the ER scenario, and the coverage of grassland and wetlands increased significantly. (3) The mean values of ecosystem services in the ER scenario were higher than those in the ND and EP scenarios. These findings clearly indicate that the RME evaluation system can accurately evaluate the ecological restoration effects under multi-scenarios in the future, providing a new perspective for ecological restoration evaluation in other regions.

1. Introduction

In recent years, with the rapid development of the economy and society, the exploitation and utilization of resources in the Yellow River Basin (YRB) has continued to deepen [1]. For example, water resource development, hydropower construction, mineral resource extraction, and other activities have had varying degrees of impact on the ecological environment, causing a number of environmental problems, including soil erosion [2,3], water pollution [4,5], forest destruction [6,7], and reduction in biodiversity [8,9], among others. Beginning in 2016, China launched the “Shan-shui Initiative in China” to complete ecological restoration and management of an area of 80 million acres by 2023. It is important to establish a scientific assessment framework to maximize the effect of ecological restoration. Land use simulation technology can predict the future land distribution pattern [10], which can be integrated into the existing assessment framework to achieve the future ecological restoration effectiveness analysis, more scientific and comprehensive assessment, and planning of ecological restoration work.

At present, the evaluation methods of ecological restoration effectiveness are mainly based on ecosystem structure, ecosystem function, ecological recovery, socio-economic impact, and other aspects. To solve the problems of an incomplete index system and a single evaluation scale in the current ecological effectiveness evaluation, Ye [11] constructed a complex assessment indicator system for ecological restoration of city water pollution on the basis of the AHP multi-scale approach, and the quality assessment of ecological restoration was completed. Han et al. [12] proposed a comprehensive assessment system using the “P-S-R” perspective to develop LSER strategies. It combines landscape, ecology, and GIS approaches. Liu et al. [13] proposed a comprehensive assessment index system for ecological restoration effectiveness based on “pattern–quality–service” from the perspective of ecosystem patterns. Yuan et al. [14] developed a comprehensive model for evaluating the ecological restoration effects of specific coal mines. The model selects evaluation factors from the ecological projection zone and the impact zone. Xiao et al. [15] proposed a multi-scale effectiveness evaluation system consisting of three scales: the sub-project scale, conservation and restoration unit scale, and watershed scale.

In summary, existing studies have extensively assessed the effectiveness of ecological restoration using historical and current land use data [16,17,18]. Although these studies have added to the assessment framework and content, most of them have not considered the future benefits of ecological restoration. In conclusion, there are few studies that assess the expected benefits of restoration projects by using land use simulation technology.

Land use simulation technology is often used to predict and analyze future land use changes [19]. By simulating different scenarios, their impact on ecosystems and biodiversity can be predicted. Therefore, this technology is an important means to build a future-oriented ecological restoration evaluation framework, which can make up for the shortcomings of the traditional evaluation system. There are many land use prediction models, including the CA-Markov model [20,21,22,23], the CLUE-S model [24,25,26], the GeoSOS-FLUS model [27,28,29,30], etc. However, traditional models have limitations in the study of the drivers and simulation accuracy associated with land use transition. It is difficult to predict both quantitative and spatial patterns simultaneously. The PLUS model combines the random forest algorithm and cellular automata to explore the driving factors of land expansion and forecast the evolution of landscape patches [31,32,33,34], which helps to identify and evaluate possible ecological changes in the future and provide a basis for formulating ecological protection and restoration measures.

As an important ecological barrier and water conservation area in China [35], the YRB in Sichuan is a key area for the implementation of the “Shan-shui Initiative in China”. In this paper, a Restoration–Monitoring–Effectiveness (RME) assessment framework was constructed using the InVEST model and the PLUS model. It was used to assess and predict land use distribution, landscape pattern changes, and ecosystem services under different scenarios in the Yellow River Basin in Sichuan from 2020 to 2030. We integrated land use simulation technology into the effectiveness assessment by simulating multiple land use scenarios and predicting their impacts on the ecosystem. This will help optimize land use allocation and identify risks to ecological restoration in advance by simulating potential environmental changes, which will help government departments develop appropriate response strategies. In conclusion, this study fills the gap in ecological restoration research that requires full consideration of future land use changes. It lays the foundation for guiding the direction of future land use and the development of land planning and management.

2. Materials and Methods

2.1. Study Area

The YRB in Sichuan sits on the southeastern edge of the Qinghai–Tibetan Plateau, at the junction of the Hengduan Mountain Range and the Northwest Sichuan Mountain Canyon. Its specific geographical location is 101°39′ E–103°23′ E and 32°09′ N–34°05′ N (Figure 1). The average altitude of this area is about 3500 m, and it has typical plateau climate characteristics. The main channel of the YRB in Sichuan spans 174 km and passes through five counties: Aba, Hongyuan, Ruoergai, and Songpan in Aba Prefecture, and Shiqu in Ganzi Prefecture. The watershed area covers 18,700 square kilometers, accounting for 2.4% of the YRB [36]. The land use type is mainly plateau meadows and wetlands, and it is a key area for ecological restoration and protection [37,38,39]. This area is also crucial to the water supply of the YRB, contributing 40% of the water volume of the Yellow River in the dry season and 26% of the water volume in the wet season. As a national wetland nature reserve and an important water conservation and replenishment area, the ecological environment protection of this area is of great significance to the ecological security of the YRB and even the whole country.

2.2. Data Sources

The data sources utilized for the research mainly include land use and land cover change (LUCC), driving factors, and the restoration dataset, as outlined in

Table 1. Data sources for this research.

Type	Data	Resolution	Source
Land use	CNLUCC	30 m	http://www.resdc.cn/ (accessed on 21 July 2023)
Natural factors	DEM	30 m	http://www.resdc.cn/ (accessed on 25 July 2023)
	Slope	30 m	Calculated from DEM
	Aspect	30 m
	Annual mean temperature	1 km	http://www.resdc.cn/ (accessed on 25 July 2023)
	Annual mean precipitation	1 km
	Soil type	1 km
Social factors	Population density	1 km	http://www.resdc.cn/ (accessed on 28 July 2023)
	GDP	1 km
Location factors	Distance from government	30 m	Open Steet Map (https://www.openstreetmap.org/ (accessed on 2 August 2023))
	Distance from highway network	30 m
	Distance from water	30 m
Restoration dataset	Restoration sites and water surfaces	/	Based on visual interpretation and field verification

. Driving factors are the key factors affecting land use change. By selecting relevant driving factors, land use change can be simulated and predicted more accurately. Based on existing investigations [40,41], we selected 11 relevant driving factors and categorized them into natural, social, and location factors. We combined land use data with field investigation and divided it into cropland, woodland, grassland, water body, built-up land, and wetland. The restoration dataset (Figure 2) was produced through remote-sensing image interpretation and UAV field verification, which was composed of field-collected ecological restoration sites and the water surface. This dataset would be used as a conversion constraint for the simulation of the ER scenario. In order to ensure the accuracy and consistency of the research data and achieve reliable simulation analysis and visualization, the coordinate system of all data in this paper was projected to Krasovsky_1940_Albers, and raster data of different resolutions were resampled to a spatial resolution of 30 m × 30 m using the Kriging interpolation method in ArcGIS 10.8. Through the above steps, data with different scales can be effectively analyzed from a cartographic perspective, providing a solid foundation for further spatial analysis.

2.3. Research Framework

The research framework of this paper consists of three parts (Figure 3). (1) Land use simulation: First, the Markov model was used to design three future scenarios. Then, combined with the selected driving factors and constraint areas, the PLUS model was used to simulate future land use changes and generate land use distribution maps under each scenario. (2) Evaluation framework construction: According to the sensitivity of landscape pattern and ecosystem service function, appropriate evaluation indicators were selected, such as construction rate, wetland coverage, and water conservation, and these indicators were integrated to propose an RME evaluation framework. (3) Incorporating future LUCC data into the RME framework to evaluate the effectiveness of ecological restoration projects in the YRB of Sichuan.

2.4. Land Use Simulation

2.4.1. Scenario Design

Three scenarios were developed to simulate future land use by adjusting the probability of transitions in the land use transition matrix to match possible scenarios for future development: natural development (ND), economic priority (EP), and ecological restoration (ER). On the basis of prior investigations and the current conditions [42,43], the details of each scenario were designed as follows:

(1)

ND scenario: It is assumed that there is no change in the trends of land use from 2020 to 2030. This scenario does not introduce new policies or external interventions, and the change in land use type mainly depends on the existing socio-economic and natural change trends.

(2)

EP scenario: This scenario aims to promote economic development and expand the scale of urbanization. During the forecast period, the probability of grassland, wetland, and woodland being converted into built-up land and farmland is increased. The probability of built-up areas being converted into other landscape types is reduced. In this scenario, land use change is mainly based on maximizing economic benefits, which may lead to more natural land being developed into urban or agricultural land.

(3)

ER scenario: The goal of this scenario is to implement ecological restoration projects to protect local ecosystems and reduce human interference. During the forecast period, the probability of other land use types being converted to built-up land is reduced, and the probability of built-up land being converted to ecological land (such as grassland, wetlands, and woodlands) is increased. The restoration dataset is used as an ecological constraint to ensure that ecological restoration measures are effectively implemented. This scenario emphasizes environmental protection and ecological restoration and improves the ecological environment by reducing development pressure and increasing the area of ecological land.

2.4.2. PLUS Model

The PLUS model is an advanced land use simulation tool that integrates the land expansion analysis strategy (LEAS) and the cellular automaton model based on a multi-type random seed mechanism (CARS) [33]. LEAS analyzes land use conversion drivers by selecting extended areas of transition in land cover and using random forest classification algorithms. The CARS module uses adaptive coefficients to simulate the expected demand for land use and the spatiotemporal dynamic transition in land cover on the basis of the development probability. In comparison to the FLUS model, the PLUS model can perform a superior analysis of the relationship between LUCC and driving factors.

Combined with the characteristics of the study area, this study selected 11 driving factors (Figure 4): DEM, slope, aspect, annual average temperature, annual average precipitation, and soil type as natural factors; social factors were also considered, including population density and GDP; finally, we also considered locational factors, such as the distance to the government, highway network, and water.

2.4.3. Model Verification

It was essential to validate the precision of the PLUS model before simulating the LUCC. In this paper, the distribution of land use in 2020 was calculated using the 2000 and 2010 land use data with 11 driving factors and constraints. A comparison of the result with the actual land use in 2020 has been made. The kappa coefficient is widely used to evaluate the accuracy of land use simulation [44]. The kappa formula is as follows:

$$kappa=\frac{p_0-p_e}{1-p_e}$$

(1)

where $p_0$ is the proportion of simulated correct grids, $p_e$ is the percentage of correct grids simulated under random conditions, and 1 is the ideal percentage of simulated correct grids. The kappa value ranges from 0 to 1. The higher the value, the more accurate the simulation.

However, the kappa coefficient has limitations in reflecting the true performance of the model, and the spatial distribution of errors is not considered in the evaluation process. Therefore, we also used the overall accuracy to supplement the conclusion of the kappa coefficient and compared the simulation prediction results with the actual land use map to provide a more comprehensive verification from the spatial distribution.

2.5. Evaluation Framework Construction

2.5.1. Remote-Sensing Monitoring

Remote-sensing monitoring is a crucial component of ecological environment monitoring systems [45]. It provides essential technical support for enhancing supervisory capacity for protecting the environment and evaluating the effect of environment restoration [46,47,48]. In this study, the construction rate, eco-land rate, grassland coverage, and wetland coverage were chosen to assess the proportion of land use types quantitatively. The construction rate is the percentage of the area of built-up land. The eco-land rate calculates the sum of the woodlands, grasslands, water bodies, and wetlands as a percentage of the area covered by all land use types.

To effectively portray the landscape pattern in the research region, four landscape pattern indices were selected and calculated using Fragstats4.2. Patch density (PD) measures the number of patches per unit area, indicating landscape fragmentation. The contagion index (CONTAG) assesses the extent to which different patch types are aggregated or clumped. Shannon’s diversity index (SHDI) evaluates the diversity of the landscape by accounting for the abundance and evenness of patch types. The aggregation index (AI) measures the degree to which patches of the same type are spatially aggregated.

These indices (

Table 2. Remote-sensing monitoring indicators.

Evaluation Content	Evaluation Indicators	Meaning of Indicators
Land use change	Construction rate	Proportion of built-up land out of total area
	Eco-land rate	Proportion of eco-land out of total area
	Grassland coverage	Proportion of grassland out of total area
	Wetland coverage	Proportion of wetland out of total area
Landscape patterns	Patch density	Water conservation
	Contagion index	Soil conservation
	Aggregation index	Carbon storage
	Shannon diversity index	Habitat quality

) were used to characterize human disturbance, landscape fragmentation, the diversity of the landscape, and the aggregation of patches. The indicators were selected based on their sensitivity to urbanization processes and ecosystem changes and their applicability to remote-sensing data.

2.5.2. Effectiveness of Ecosystem Services

The InVEST model is a tool for ecosystem service assessment and quantification. It contains multiple model components for calculating various ecosystem service indices to support decision making on land use and natural resource management [49,50]. This study selected four ecosystem service indicators from the InVEST model to characterize the expected ecological effectiveness. The annual average rainfall was used to replace the rainfall in each future scenario, and the annual average evapotranspiration (ET) was used to replace the ET in future scenarios. Then, the data were imported into the water conservation model of InVEST; after debugging the model parameters, the model was run to obtain the spatial distribution of expected water conservation. The expected soil conservation calculation was aligned with expected water conservation. We used the annual average rainfall to calculate the future rainfall erosivity factor. To calculate the expected habitat quality and carbon storage, we input the results of the land use simulation into the InVEST model, the threat factor and carbon density parameters were set, and the model was debugged and run to obtain the expected spatial distribution of habitat quality and carbon storage. The formulas for calculating expected ecosystem services are expressed as follows:

$$W={\sum }_{i=1}^j\left(P_i-R_i-{ET}_i\right)\times A_i\times {10}^3$$

(2)

where $W$ indicates the expected water conservation (${\mathrm{m}}^3$), $P_i$ represents the annual average rainfall (mm), $R_i$ represents surface runoff (mm), ${ET}_i$ indicates the annual average evapotranspiration (mm), $A_i$ is $i$ the ecosystem area (${\mathrm{k}\mathrm{m}}^2$), $i$ is the $i$ ecosystem type, and $j$ is the total number of ecosystem types.

$$Q_i=R\times K\times LS\times (1-C)$$

(3)

where $Q_i$ indicates the expected soil conservation, $R$ represents the erosivity factor of the expected rainfall, $K$ indicates the erodibility factor of the soil, $LS$ represents the topographic fluctuation coefficient, and $C$ is the surface vegetation cover coefficient.

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

(4)

where $C_{total}$ is the total carbon stock, and $C_{above}$, $C_{below}$, $C_{soil}$, and $C_{dead}$ represent aboveground biogenic carbon, belowground biogenic carbon, soil organic carbon, and dead organic carbon, respectively.

$$Q_{xj}=H_j\times \left[1-\left(\frac{D_{xj}^z}{D_{xj}^z+k^z}\right)\right]$$

(5)

where $Q_{xj}$ is the habitat quality of grid $x$ in land use type $j$, $H_j$ is the habitat adaptability of grid $x$ for land use type $j$, $D_{xj}$ represents the degree of habitat degradation of grid $x$ in land use type $j$, $k$ is the half saturation coefficient, and $z$ is a normalized constant.

2.5.3. Proposed RME Evaluation Method

This paper simulates the future distribution of LUCC using the PLUS model. Monitoring indicators were chosen to quantitatively analyze land use changes. The InVEST model has been integrated for the evaluation of ecosystem service functions for each scenario. Ultimately, we proposed a long-term RME evaluation index system, incorporating a future perspective.

Table 3. Effectiveness evaluation indicator system of ecological restoration.

Indicator Type	Evaluation Content	Indicator Layer	Unit
Monitoring indicators	Land use change	Construction rate	%
		Eco-land rate	%
		Grassland coverage	%
		Wetland coverage	%
	Landscape patterns	Patch density	/
		Contagion index	/
		Aggregation index	/
		Shannon diversity index	/
Effectiveness indicators	Ecosystem services	Water conservation	mm
		Soil conservation	(t/ha)
		Carbon storage	/
		Habitat quality	(t/ha)

displays the specific evaluation indicators.

3. Results

3.1. Model Accuracy Verification

This study compared simulated LUCC in 2020 with actual LUCC (Figure 5). The simulation results were consistent with the spatial patterns in the current data. Referring to previous studies [51,52], a kappa coefficient > 0.8 is generally considered to be credible. The model accuracy evaluation results showed that the kappa coefficient was 0.8247 and the overall accuracy was 0.9231, indicating that the accuracy of the PLUS model meets the research requirements, can accurately reflect the changes in land use, and can be used to predict the LUCC in the study area in 2030.

3.2. Land Use Change during 2020–2030

The PLUS model was used to simulate LUCC in 2030 for the three scenarios (Figure 6). Grassland area accounted for the highest proportion, followed by wetlands and woodlands. In the EP scenario, the areas of built-up land and cropland continued to increase, while the areas of other ecological lands decreased. Built-up land had the largest increase, rising by 9.1% compared to the ND scenario. Conversely, the ER scenario restrained built-up land expansion, decreasing by 28.14% compared to the EP scenario, while the areas of grassland, wetlands, and woodlands increased.

For the purpose of further exploring the changes in land use structure, the transfer of land use types from 2020 to 2030 was analyzed (Figure 7). In the ND scenario, the most significant change in land types is the mutual transfer of grassland and woodland, followed by the transfer of each land type to built-up land. The area of grassland decreased by 13.01 km², while the built-up area increased by 9.42 km², expanding by 20.62%, mainly from the transfer of 8.25 km² from grassland to built-up land and 0.98 km² from cropland to built-up land.

The EP scenario intensified the shift to built-up land, with the area of built-up land increasing by 14.40 km², a growth of 31.51%. The expansion was mainly due to the transfer from woodland to built-up land by 9.57 km². In contrast, the ER scenario showed a decrease in built-up land of 2.49 km², mainly converted to grassland, woodland, and wetlands. Wetlands increased significantly by 5.09 km², mainly from water bodies and grasslands.

These findings emphasize how different scenarios affect LUCC dynamics and highlight the key role of controlling rapid urbanization and implementing ecological restoration in managing land use sustainability.

3.3. Implementation of the Evaluation Framework

3.3.1. Changes in Monitoring Indicators under Multiple Scenarios

The changes in land use types’ area ratios are depicted in Figure 8. In the ND and EP scenarios, the construction rate steadily increased, with the EP scenario showing the highest growth, rising from 0.27% to 0.35%. Conversely, the ER scenario curtailed the expansion trend of the construction rate compared to the other scenarios.

Both the eco-land rate and grassland coverage rate exhibited declining trends under the ND and EP scenarios. The eco-land rate decreased from 99.65% to 99.59% and 99.55%, while the grassland coverage rate decreased from 71.88% to 71.81% and 71.79%, respectively. Only the ER scenario showed an increase in these percentages, notably in wetland coverage, which rose from 19.376% to 19.406%.

Four landscape indices reflecting pattern changes under different scenarios are shown in Figure 9. PD and SHDI increased in the ND and EP scenarios, with the EP scenario showing the highest growth rates: PD rose from 0.3059 to 0.3365, and SHDI increased from 0.8147 to 0.8199. In contrast, the ER scenario showed much smaller variations, with PD increasing by only 0.0039 and SHDI by 0.0003.

The CONTAG and AI indices displayed consistent trends. In the ND scenario, both indicators decreased significantly: CONTAG dropped from 73.0251 to 72.8003, and AI decreased from 96.9157 to 96.8001. In the ER scenario, CONTAG and AI rebounded slightly, decreasing by only 0.0184 and 0.0283 compared to 2020.

These findings indicate that economic development, as seen in the ND and EP scenarios, intensified landscape fragmentation and increased landscape diversity and patch dispersion. The ER scenario, however, showed significantly reduced changes in these indicators, suggesting that ecological restoration mitigates landscape pattern variations compared to scenarios emphasizing economic growth.

3.3.2. Analysis of Ecosystem Services Changes in 2030

Through calculating the four ecosystem service indicators under the EP and ER scenarios, the spatial pattern of ecosystem services in the study area was obtained (Figure 10). The general pattern of ecosystem services in the two scenarios was similar. Water and soil conservation gradually declined from south to north, while habitat quality and carbon storage presented a general urban–rural difference in spatial distribution.

This study quantified changes in ecosystem services (

Table 4. Statistical analysis of the average values of ecosystem services in 2020 and 2030.

ES		Average				Change
		2020	ND	EP	ER	ND	EP	ER
WC	(mm)	287.8259	287.7307	287.6542	287.8568	−0.0952	−0.1717	0.0309
SC	(t/ha)	62.2719	62.2722	62.2703	62.2722	0.0003	−0.0016	0.0003
HQ	-	0.6671	0.6666	0.6664	0.6670	−0.0004	−0.0007	−0.0001
CS	(t/ha)	20.4151	20.4022	20.3986	20.4159	−0.0129	−0.0165	0.0008

). Water conservation declined under the ND and EP scenarios by averages of 0.0952 and 0.1717, respectively, compared to 2020. However, in the ER scenario, there was an average increase of 0.0309, indicating an improvement in water conservation. Soil conservation was only impaired under the EP scenario, with a mean decrease of 0.0016. Habitat quality decreased in the three scenarios but only decreased by 0.0001 in the ER scenario, which was less disturbed than the other two scenarios. The mean carbon storage also increased by 0.0008 only under the ER scenario and decreased in ND and EP scenarios.

In conclusion, the ER scenario demonstrated higher mean values of ecosystem services compared to the ND and EP scenarios, highlighting the importance of ecological restoration in enhancing water conservation, soil conservation, habitat quality, and carbon storage. Local governments should prioritize supporting and promoting ecological restoration projects to sustain healthy ecosystem services.

4. Discussion

4.1. Impact of the Restoration Dataset on Accuracy of the PLUS Model

In the PLUS model, the conversion constraint is a restriction or rule imposed on a particular type of land conversion when simulating land use change. These constraints can be policies, regulations, environmental protection requirements, land planning, or other factors. By applying these constraints, the model can better reflect the actual situation and avoid unreasonable land use change. For example, certain areas may be protected from conversion to other types of land.

In this paper, the restoration dataset was used as a conversion constraint to ensure that the PLUS model thoroughly considered the current situation of the YRB in Sichuan. As shown in Figure 11, the kappa coefficient and overall accuracy followed the same trend, with the values of both metrics showing a trend of restoration dataset > water surface > no-restriction zone. Moreover, the kappa coefficient and overall accuracy reach the maximum value when using the restoration dataset as a conversion constraint. The kappa coefficient increased from 0.8095 to 0.8247, and the overall accuracy increased from 0.9142 to 0.9231.

This conclusion was consistent with the research of Nie W. et al. [53] and the research of Lin Z. et al. [31], who also used the PLUS model in their research; especially after considering specific conversion constraints, the prediction accuracy of the model was significantly improved. In addition, their study also showed that the choice of conversion constraints has an important impact on the prediction results of the model, and different conversion constraints may lead to different simulation results. This further confirms the accuracy and reliability of this study.

4.2. Spatial Distribution Characteristics of Ecosystem Services in Each Scenario

For the detailed analysis in this study, Zoige County was selected to investigate changes in the spatial distribution of expected indicators for ecosystem service effectiveness. The county has a diverse ecological landscape, including woodlands, grasslands, and urbanized areas, which is suitable for evaluating the impact of different scenarios on ecosystem services. Figure 12 shows that water conservation and soil conservation had similar spatial patterns, with high values in the middle and south of the region with high vegetation cover. Precipitation mainly affected low values. Habitat quality and carbon storage were degraded around cities, towns, and the highway network. The four ecosystem services had the highest value in the areas with high vegetation coverage, which was consistent with the distribution of woodlands and grasslands. It shows that ecological service functions are best in places with intact vegetation. Areas with less precipitation are associated with lower values of these services, highlighting the impact of climate on the distribution of ecosystem services. Human activities also have a negative impact on habitat quality and carbon storage.

Under the EP scenario, trends of water conservation, habitat quality, and carbon storage degradation were intensified around Zoige County compared to the ND scenario; this is due to the convergence of the labor force in urban centers under the background of a vigorously developing regional economy. As human activities intensify, built-up land will rapidly increase in the future, leading to outward expansion and occupation of grassland and wetlands. This will result in a significant decline in water conservation, habitat quality, and carbon storage, as well as a significant increase in low-value areas. However, in the scenario of ER, the values of water conservation, habitat quality, and carbon storage in low-value areas were significantly lower compared to the scenarios of ND and EP. This suggests that ecological restoration projects can effectively protect ecological land, such as grassland and wetlands, and improve the situation of built-up land rapidly occupying ecological land, which plays a crucial role in maintaining the ecological stability of the YRB.

This study summarizes the spatial patterns observed under different scenarios and emphasizes the role of ecological restoration in reducing the negative impacts of ecosystem services. It highlights the importance of protecting areas with high vegetation cover, such as woodlands and grasslands, in maintaining ecosystem services under the pressure of urbanization. It recommends the formulation of policy measures that can balance economic development and ecological protection.

4.3. Analysis of the Overall Ecological Effects under Different Scenarios

In the EP scenario, built-up land has increased significantly, and this change was mainly attributed to the rapid expansion of towns. Rapid urbanization has led to large-scale ecological and development problems [54]. This rapid expansion has encroached on a large amount of ecological land, resulting in a reduction in grassland and woodland areas, seriously affecting ecosystem services, reducing soil and water conservation functions, and reducing habitat quality and carbon storage. In addition, rapid urbanization, while promoting economic development, has also brought serious problems of ecological environmental degradation and fragmentation of landscape patterns, challenging regional ecological balance and sustainable development.

Under the ER scenario, land use, landscape pattern, and ecosystem services all showed clear advantages. First, the ER scenario effectively reduced the expansion of built-up land and increased the area of wetlands, grasslands, woodlands, and protected ecological land. Second, the landscape pattern became more stable, landscape fragmentation and heterogeneity were significantly reduced, landscape aggregation and connectivity were improved, and the impact of human activities on the environment was reduced. In addition, ecosystem service functions were significantly improved, soil and water conservation functions were enhanced, and habitat quality and carbon storage were increased, indicating that ecological restoration projects not only help restore natural ecology but also enhance the sustainability of the regional environment and promote ecosystem health and stability [55].

In conclusion, this study highlights the profound impact of different development strategies on the ecological environment and land use and emphasizes the importance of ecological restoration in achieving sustainable development.

4.4. Limitations and Prospects

In this study, three single scenarios for future land use stimulation were created. It is important to note that economic development and ecological restoration are not entirely independent. When considering regional development, it is necessary to take into account both economic development and ecological protection.

When using the PLUS model, the parameters of each module were set based on the existing literature and research findings. This approach may need more objectivity. In order to improve simulation accuracy, repeated experiments can be conducted, although there is no unified standard for collecting driving factors.

It is very important to select appropriate and comprehensive evaluation indicators when evaluating the effectiveness of ecological restoration. Due to the limitation of data, this study only selected some commonly used ecological restoration evaluation indicators. In future studies, more characteristic indicators should be added to comprehensively evaluate the impact of land use change on the ecosystem. In addition, when assessing expected ecosystem services, we used multi-year averaging to obtain precipitation and evapotranspiration data for future scenarios as a proxy for actual values, which may bias the results.

Therefore, future research should aim to address these limitations, including developing models that can balance economic development and ecological protection measures to more fully understand land use change, establishing uniform standards for collecting and verifying driving factors to improve the objectivity and reliability of simulation models, developing advanced time series models and statistical methods to reduce bias in estimating ecosystem services, and creating multiple scenarios with different assumptions to capture a wider range of potential futures and better understand the robustness of research results. By addressing the limitations in these areas, future research can provide more accurate and comprehensive land use simulations for more precise land use planning and policy making.

5. Conclusions

In this paper, the PLUS model was used to forecast the spatial distribution of LUCC in three scenarios. Additionally, an RME evaluation framework of ecological restoration was constructed and used to assess effectiveness in the YRB of Sichuan.

The findings indicated the following: (1) Under the ND scenario, the built-up land area exhibited a slight expansion trend. Under the EP scenario, the expansion trend of built-up land intensified. Under the ER scenario, the speed of urban expansion was further restricted, effectively protecting the local ecology. (2) Under the ND and EP scenarios, urbanization will intensify in the future, leading to increased landscape fragmentation. Under the ER scenario, the ecological land use rate and wetland coverage rate will increase slightly, limiting the process of landscape fragmentation. (3) Water conservation function, soil conservation function, and carbon storage are all improved under the ER scenario, indicating that the ecological restoration project will improve the environmental conditions in the study area.

The research findings fully demonstrate that the RME evaluation system integrated with land use simulation technology can accurately evaluate the ecological restoration effect in future scenarios. It shows a wide range of potential applications and provides a solid scientific basis for future ecological restoration projects and policy formulation.