The environmental impact assessment of China’s ecological migration from a social–ecological perspective

Abstract

When accounting for the social–ecological impact of an ecological restoration program, both objective environmental contexts and people’s subjective perceptions are required. While this kind of environmental impact assessment lacks a comprehensive perspective. We use the difference-in-differences model to evaluate the effect of the greenness of the landscape after ecological migration in the Qilian Mountains in China; and analysis of variance and fixed effects models are used to evaluate the effects of such ecological restoration programs on local people’s perceptions. The results show that the ecological migration program in the Qilian Mountains has been successful at not only significantly improving remotely sensed greenness at the landscape scale, but also at enhancing immigrants’ environmental perceptions. These findings demonstrate the environmental impacts of ecological migration from a social–ecological perspective, and can provide methodological implications for landscape planning to support a better understanding of ecological restoration programs in the drylands.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13280-024-02011-w.

Introduction

Ecological restoration for sustainable development has attracted global attention after the calls from the United Nations Decade of Ecological Restoration, the Nature-based Solution, and the live on land proposal in the sustainable development goals (SDGs) (Albert et al. 2019; Badura et al. 2021; Yurui et al. 2021; Liu et al. 2023a). At the landscape scale, human well-being is tightly linked to the environment, so through regional conservation and development programs, both ecological and social progress should be achievable without detracting from the primary objectives of these programs (Ding et al. 2021; Liu et al. 2023b). To ensure the successful implementation of ecological restoration programs, it is necessary to simultaneously assess their social and ecological effectiveness (Estrada-Carmona et al. 2014; Jepson 2022; Wang et al. 2023a, 2023b). Therefore, appropriate monitoring from a social–ecological perspective is necessary for achieving the implementation of ecological restoration programs (Cilliers et al. 2018; Zhao et al. 2023; Zhang et al. 2023; Liu et al. 2023a, b, c, d) in a way that helps to determine whether the benefits of these programs can both objectively improve environmental quality and be subjectively recognized by the local people (Kong et al. 2022; Liu et al. 2023c).

Ecological restoration programs have an impact on people’s environmental perceptions, but not all ecological restoration programs can successfully improve the positivity of people's perceptions. For example, in addition to most farmers having a positive attitude toward the ecological protection afforded by the Grain for Green Program since, they find that flood levels are reduced, the environment has improved, and their incomes have increased following the implementation of the program (Li et al. 2022a, b). However, some farmers are dissatisfied with the program and are unwilling to participate in future programs, because they think that it did not contribute to improving water quality or protecting biodiversity (Zhang et al. 2019). A survey of fishermen's views on marine reserves shows that the vast majority feel that the impact was either neutral or positive for both fish abundance and habitat quality, while some have more negative perceptions of marine reserve outcomes (Bennett et al. 2019). Similarly, farmers in southwestern France have mixed views on the EU’s Common Agricultural Policy, as they cited both 32 positive and 25 negative contributions of rural forests (Blanco et al. 2020). Therefore, the results of public perception and the biophysical evaluation of landscapes can be inconsistent. If the vegetation greening and ecosystem service improvement reflected by remote sensing is inconsistent with people’s perceptions, then their level of participation in sustaining ecological restoration programs will be affected (Wang et al. 2022).

Many studies have proven that the implementation of ecological restoration has resulted in large-scale greening at the landscape scale (Birdsey and Pan 2015; Li et al. 2022a, b; Lin et al. 2022). Due to factors such as temperature rise and the CO₂ fertilization effect, however, it cannot simply be considered that the greening of vegetation in ecological restoration areas is the direct effect of ecological restoration programs, e.g., the warming of high latitude areas and the wetting of arid areas also contribute to the improvement of the local environment, which may be wrongly attributed to ecological restoration programs (Piao et al. 2019; Qian et al. 2019; Gao et al. 2021). Similarly, if only people who have experienced ecological restoration programs are selected as respondents, the positive perception effect of ecological restoration programs cannot be proven. Therefore, to exclude the social and environmental context that influences greening and social perception at the landscape scale, a combination of subjective and objective assessments based on statistical methods is needed within the control group and experimental group to comprehensively and quantitatively monitor the effectiveness of regional ecological restoration programs (Bennett 2016; Huang et al. 2019; Sheng et al. 2019; Fischer et al. 2021; Ding et al. 2022).

Since the twenty-first century, the Ecological Migration Program (EMP) has been widely applied in ecologically fragile or degraded areas to solve ecological degradation. In particular in China, the EMP has effectively improved environmental quality (Zhang et al. 2022) and provided a good reference for poverty alleviation (Hu et al. 2018). The Qilian Mountains area, a key area of ecological construction along the economic belt referred to as the Silk Road, is an important ecological security area in China’s dryland (Zhao et al. 2021; Wang and Zhou 2022). To protect the environment of Qilian Mountain National Park, a large-scale EMP has been implemented in recent years to reduce the impact of human activities on the ecosystem (Qian et al. 2019; Li et al. 2022a, b). The quantitative social and ecological performance of the EMP as assessed by the objective environmental changes and subject perceptions are key to determining the viability of the subsequent programs, available experience, and the optimization processes of regional ecological restoration programs (Wang and Zhou 2022). However, as with many assessments of ecological restoration programs worldwide, the existing research evidence cannot prove that the greening of the Qilian Mountains is caused by ecological restoration programs and cannot reflect the positive contribution of these programs to migrants’ subjective ecological perceptions (Geng et al. 2019).

This study focuses on two research gaps. One is that ecological benefit has often not been compared with social perception in an environmental impact assessment at the landscape scale. The other is that the environmental context is not statistically controlled when analyzing the environmental and social effects of ecological restoration programs. Accordingly, we propose a technical method for the overall assessment of the ecological restoration program from a social–ecological perspective. Taking the EMP at the northern foot of the Qilian Mountains in China as an example, we pose the following specific scientific questions: Does the Qilian Mountain EMP have a positive environmental impact from a social–ecological perspective? By answering this question, we plan to achieve two main research objectives: determine the (1) program effects on vegetation greenness under the environmental context and (2) program effects on migrants’ perceptions under the social context.

Materials and methods

Study area

According to China’s 13th 5-Year Plan in Gansu Province, more than one million people will emigrate from ecologically fragile areas, hoping to alleviate the increasingly severe ecological and poverty issues in these areas. The plan in Gansu Province is to organize a total of 234 780 households and 1.12 million ecological migrants by 2018, with a total investment of approximately 40 billion yuan. In Gansu Province, Wuwei city is the main battlefield of ecological protection and restoration in the Qilian Mountains, among which Gulang County and Tianzhu Tibetan Autonomous County are key areas for the relocation of Qilian Mountain immigrants. We regard Gulang and Tianzhu Zangzu Autonomous County as representative study areas and select 59 large communities at the landscape scale, including 18 immigration sites, 20 emigration sites, and 21 control sites, to explore the impacts of the EMP on the environment. These sites have similar characteristics in terms of topography, transportation, culture, and livelihoods, which satisfy the requirements of the sample for the study (Fig. 1).

Fig. 1.

Location of the study area. The image shows the average NDVI in 2021 and the three kinds of community points, including the immigration group, the emigration group, and the control group

Data and processing

The data used in this study include both remotely sensed data (Table S1) and questionnaire survey data (Table S2). The normalized difference vegetation index (NDVI) is a commonly used indicator of vegetation coverage that is used to evaluate vegetation status based on greenness (Huete et al. 2002; Piao et al. 2006; Gong et al. 2015). We use the MODIS NDVI and EVI product (MOD13Q1) with a spatial resolution of 250 m and the LAI product (MOD15A2H) with a spatial resolution of 500 m. We use two climatic control variables—precipitation and temperature—taken from the Climate Hazards Group Infrared Precipitation with Station data (CHIRPS) and the European Centre for Medium-Range Weather Forecasts Reanalysis v.5 (ERA5) dataset, respectively. Land surface temperature (LST) (MOD11A1) with a resolution of 1 km is also applied as a supplementary control variable at the landscape scale. We set up a 1 km buffer zone for the community point as the research area and extract the annual mean values of NDVI, LST, temperature and precipitation of the emigration group, immigration group and control group from 2001 to 2022.

The questionnaire survey data were obtained from sampling household surveys in the cities of Wuwei and Zhangye in Gansu Province in 2021. The study used the principle of stratified sampling to randomly select 44 villages in 11 townships (Table S2). Overall, we collected 923 questionnaires, from which 833 were considered valid after excluding the incomplete questionnaires. Our questionnaires contained two parts (Appendix S1). The first part recorded people’s perceptions of changes to their environment, including plants, water resources, soil quality, blowing sand weather, air moisture, and summer temperature. The second part recorded several household characteristics as control variables, including age, gender, education, and income (Table S3). We detail the questionnaire data process in the supplementary material.

The framework for social–ecological assessment

Indicated by the elements of a social–ecological system (Ostrom 2009; McGinnis and Ostrom 2014; Falk et al. 2018), four kinds of corresponding elements were provided in the EMP context: ecological restoration program (governance system), perception of immigrants (users), vegetation greenness (resource units), and environmental and social context (resource systems) (Fig. 2). We suspect that ecological restoration program implementation cannot always promote social and ecological progress simultaneously. On the one hand, when pursuing the goal of environmental improvement, a win–win situation between people and nature might be achieved, or the interests of the people might be sacrificed. On the other hand, environmental improvement is not necessarily only caused by programs but may be influenced by other factors. In particular, the EMP exerts impacts on the immigration areas and the emigration areas that are not necessarily positive. To understand, the social–ecological effects of the EMP, we refined four elements into several indicators (Table 1) to explore the impact of each element on the social ecosystem through ecological restoration programs.

Fig. 2.

The framework for social–ecological assessment in ecological migration. The one-way arrows in the diagram represent the impact of one element on another, and the two-way arrows represent the interaction between the two elements. This image shows the impact of ecological restoration programs on (1) social context, (2) environmental context, (3) perception of people, and (4) vegetation greenness; it also illustrates the interaction between (5) perception of immigrants and vegetation greenness, (6) vegetation greenness and environmental context, (7) social context and environmental context, and (8) immigrants’ perception and social context

Table 1.

The meaning of the elements in the framework

Type	Meaning	Element	Indicator
Governance system	Rules governing the environment	Ecological restoration program	Ecological migration program
	Social composition	Social contexts	Age, gender, education, and income
Resource units	Variability of Resource units	Vegetation greenness	NDVI, LAI, and EVI
Resource systems	Combination of resource units	Environmental contexts	Plant
			Water resource
			Soil quality
			Blowing sand weather
			Air moisture
			Summer temperature
Users	Users’ perceptions of the social–ecological system	Perceptions of people	Environmental perceptions

Assessment of the effects of the EMP on vegetation greenness

To assess the impacts of the EMP on vegetation greenness, we use a difference-in-differences (DID) method based on remotely sensed data. The advantage of the DID model lies in its capacity to effectively control unobservable factors by differencing out individual and time-fixed effects, thereby mitigating endogeneity concerns. Furthermore, the DID model can alleviate issues stemming from selection bias by estimating causal effects based on the change in the same cohort at different time points. Nevertheless, the efficacy of the DID model hinges on the critical assumption of parallel trends, so treatment and control groups need to exhibit similar trends before treatment initiation; deviations from this assumption may yield erroneous results. After comprehensively evaluating the feasibility and scientificity of this method, we apply the DID model to the data analysis of both the immigrant group and the emigrant group in Stata 17 to better compare the impact of EMP.

The DID method uses the following specification:

$$Y_{i,t}=\beta \left(P_i\times T_t\right)+\theta P_i+\delta T_t+\rho X_{i,t}+{\epsilon }_{i,t},$$ 1

where Y_i,t is the vegetation greenness of village i as measured by the logarithm of its NDVI and recorded in the relevant period t. We chose the logarithm of NDVI as the dependent variable because the data were transformed to allow the clutter to approach the assumptions of the model (Hou et al. 2021). P_i is a binary indicator that takes the value of P_i = 1 if village i is a program village in 2018–2022 (treatment group) and P_i = 0 otherwise (control group). The period is denoted by T_t, in which T_t = 0 for periods before the EMP was implemented (2001–2017) and T_t = 1 for periods occurring after (2018–2022) that implementation. The coefficient β measures the average treatment effect on the treated villages, which represents the average change in vegetation greenness in the treatment group relative to that of the control group. The vector X_i,t includes a set of control variables, such as the LST, temperature, precipitation and NDVI from the previous year.

Assessment of the EMP effects on people’s perceptions

To assess the impacts of EMP on people’s environmental perceptions, we applied a fixed-effect (FE) model to the questionnaire survey data. Data analysis was also performed in Stata 17.

The FE method uses the following specification:

$$Y_{i,l}={\alpha }_{i,l}+\beta P_{i,l}+\rho X_{i,l}+{\mu }_l+{\epsilon }_{i,l},$$ 2

where Y_i,l represents people’s perception of village l, and P_i,l is a binary indicator that takes the value of P_i,l = 1 if people i are immigrants and that of P_i,l = 0 otherwise. β is the regression coefficient that indicates the EMP effect, and X_i,l is a set of control variables including age, gender, education, and income.

Results

Impact on vegetation greenness

Overall, vegetation greenness as measured by NDVI showed a positive trend following the EMP implementation (Fig. 3). The control group exhibited a relatively stable trend in NDVI variation, whereas the immigration area and the emigration area experienced more pronounced fluctuations. On average, the NDVI increased by 3.10% in the control group, 4.42% in the immigration group and 10.85% in the emigration group between the preprogram period (2000–2017) and the postprogram period (2018–2022), as shown in Fig. 3.

Fig. 3.

The temporal trends of mean NDVI among the three groups (control, emigration, and immigration groups). In the figure, red represents the NDVI variation in the control group, green corresponds to the emigrating group, and blue corresponds to the immigration group. The curves depict the average NDVI values for all research sites, with the upper boundary of the colored region representing the maximum NDVI value observed among all research sites, while the lower boundary represents the minimum NDVI value

Since, changes in vegetation greenness are not necessarily caused by the EMP, we apply DID to estimate the impacts on the vegetation greenness of the immigration group and that of the emigration group. The results show the impacts of the EMP on vegetation greenness as estimated by NDVI across model specifications with different sets of fixed effects and control variables (Table 2). The EMP has significant positive impacts on both the immigration group (0.041, R² = 0.518) and the emigration group (0.185, R² = 0.662) (Table 2, Columns 3 and 6).

Table 2.

Estimated impacts of the EMP on vegetation greenness per DID model

Overall impacts on lg (NDVI)	Immigration
Variables	(1)	(2)	(3)	(4)	(5)	(6)
P × T	0.146***	0.056**	0.041*	0.258***	0.248***	0.185***
T-statistic	6.44	2.43	1.88	13.34	12.93	9.14
Observations	858	858	858	943	943	943
R2	0.384	0.476	0.518	0.620	0.638	0.662
Code	39	39	39	41	41	41
Year FE	Y	Y	Y	Y	Y	Y
Temperature FE	N	Y	Y	N	Y	Y
Precipitation FE	N	Y	Y	N	Y	Y
LST FE	N	Y	Y	N	Y	Y
NDVI from the previous year’s FE	N	N	Y	N	N	Y

This table provides the results from the DID model for the estimated impacts of the EMP on vegetation greenness (Eq. 1). The dependent variable is vegetation greenness as measured by the logarithm of NDVI. The treatment groups (P = 1) include the immigration group (18 villages) and the emigration group (20 villages). The control group (P = 0) includes 21 villages that had not experienced migration. The pre-program period (T = 0) is 2001–2017. The postprogram period (T = 1) is 2018–2022. We set different control variables in Columns (1) to (6), including precipitation, temperature, LST and NDVI, from the previous year. FE, Fixed effects; Y, Yes; and N, No. A two-sided t test is performed for each coefficient. **p < 0.05, ***p < 0.01

Next, we tested the parallel-trend assumption using an event study analysis to verify that the treatment group and the control group underwent similar change trends in vegetation greenness before program implementation. There were differences in average vegetation greenness between the treatment (immigration and emigration) and the control group for each period after controlling for context factors, including LST, temperature, precipitation and NDVI from the previous year (Fig. S1). The results indicate a clear difference before and after the program implementation, and there is a clear upward trend detected among the treatment group during the last implementation period (period = 3) of the EMP.

Impact on immigrant perceptions

Overall, the environmental perceptions of immigrants (mean value = 7.0) are higher than those of nonimmigrants (mean value = 1.5) (Fig. 4), and there are similarities across different types of personal attributes. We apply analysis of variance (ANOVA) to immigrants’ and nonimmigrants’ environmental perceptions alongside various factors. We also group these variables by gender (female, male), education (primary, secondary, higher), income (low, middle, high), and age (youth, middle-aged, old). Figure 5 shows that the perceptions of both female and male immigrants are significantly higher than those of nonimmigrants. Regardless of educational background, income and age, the average perception of immigrants is significantly higher than that of nonimmigrants. In addition, people with different genders, educational levels, income levels and ages report significant differences in their environmental perceptions (p < 0.01). These differences are as follows: first, the average perceptions of males were higher than those of females (Fig. 5a); second, the higher the education or income of a respondent is, the higher the perception (Fig. 5b and c); and last, the perception of young people is higher than that of middle-aged and elderly people (Fig. 5d).

Fig. 4.

Immigrants’ and nonimmigrants’ environmental perceptions. The figure shows the maximum value, minimum value, and mean value and distribution characteristics of the data

Fig. 5.

ANOVA results regarding immigrants’ and nonimmigrants’ environmental perceptions. a Gender: female or male; b education: primary (primary school or below), secondary (junior high school, senior high school, or technical secondary school), or postsecondary (junior college, college or above); c income: low (less than 20 000 Yuan), middle (20 000–70 000 Yuan), or high (above 70 000 Yuan); and d age: youth (15–45 years), middle-aged (46–65 years), or elderly (66 years or above). ***p < 0.01

To further confirm the evidence, we also use an FE model (Eq. (2) in Methods) to estimate the impacts of the EMP on people’s perceptions. The results show that the program has a positive and significant impact on people’s environmental perceptions (5.523, R² = 0.499) (Table 3, Column 3). These results indicate that people who have experienced the EMP are more likely to perceive changes to the environment.

Table 3.

Estimated impacts of the EMP on people’s perceptions: FE model

Overall impacts on perception	Perception
Variables	(1)	(2)	(3)
Treat	5.519***	5.542***	5.523***
T-statistic	23.66	23.96	23.78
Observations	833	833	833
R2	0.490	0.498	0.499
County FE	Y	Y	Y
Gender FE	Y	N	Y
Age FE	Y	N	Y
Education FE	N	Y	Y
Income FE	N	Y	Y

This table provides the results of the FE model regarding the estimated impacts of the EMP on people’s perceptions (Eq. 2). The dependent variable is the sum of people's perceptions of the change in ecological environment contexts. We set different control variables in Columns (1) to (3), including gender, age, education and income. FE, Fixed effects; Y, Yes; and N, No. A two-sided t test is performed for each coefficient. ***p < 0.01

Discussion

EMP effects on vegetation greenness

Based on a DID model, we evaluate the impacts on vegetation greenness of the Qilian Mountains of Gansu Province’s EMP from 2001 to 2022. We find that the implementation of the program (which was completed in 2018) not only improved the greenness of the landscape emigration group but also improved the greenness of the landscape immigration group. At the landscape scale, the increased greenness of vegetation in the emigration area arises from the natural recovery of vegetation due to diminished human activity. Conversely, the growth of vegetation greenness in the immigration area can be attributed to farming bolstering vegetation growth in the arid area, characterized by a generally sparse vegetation environmental context. The ecological migration plan of Gansu Province also indicates that ecological construction and environmental protection should be carried out, specifically by demolishing old houses in the relocation area and engaging in environmental restoration, as well as greening and beautifying the environment of the immigration area, to benefit both the emigration area and the immigration area. However, some express that migration may trigger increased risk for those areas toward which migrants move (Hermans-Neumann et al. 2017) and may exacerbate local ecological and social problems such as regional water shortages (Fan et al. 2015; Kim et al. 2017; Ding et al. 2021). In comparison, the EMP in the Qilian Mountain area can comprehensively improve the greenness of vegetation at the landscape scale because the resultant reduction in human activities increases the greenness of the emigration area, and the application of reasonable artificial ecological management also promotes the vegetation growth of the landscape.

EMP effects on people's perceptions

We found that the respondents generally hold a positive environmental perception, but the value is not high, with 6.42% expressing negative views of the EMP. Regarding single environmental perceptions (including plants, water resources, soil quality, blowing sand weather, air moisture, and summer temperature), negative perceptions accounted for 2.52%, 6.96%, 4.80%, 8.64%, 14.05%, and 22.69% of the overall responses, respectively. Previous evidences showed some involuntary resettlement programs proved to be an unsuccessful experience (Alene 2021). More than 60% of families were dissatisfied with the programs (Moahmmadi Sarrafi 2018), and this discontent stemmed from factors such as income, safety, education and transportation (Ramakrishnan 2014; Nikuze et al. 2019). However, in some cases, resettlement can provide improved residential conditions for those relocated by offering decent homes and proper environmental quality (Kearns and Mason 2013), which is similar to our positive findings.

Some investigations have uncovered negative attitudes toward ecological restoration programs; for example, many farmer migrants have lost their confidence in the EMP in the Tianchi Scenic Area in Xinjiang (Tang et al. 2012). Other studies indicate that people perceive the benefits of ecological restoration programs and have a positive attitude toward them (Bennett et al. 2019; Yurui et al. 2021), and the overall satisfaction of rural households with the land consolidation program in rural China is 76.5% (Luo and Timothy 2017). Although these findings are positive, they do not prove that people’s positive attitudes are caused by their participation in ecological restoration programs. In this study, we interviewed both immigrants and nonimmigrants to compare the impact of participation programs on people’s perceptions and found that the level of familiarity with the program is higher among families who are returning farmland than it is among families who are not returning farmland (Zhang et al. 2022), which is similar to the conclusion of our findings.

The particular environment in dryland

Our study presents a universal methodology and framework for examining the impacts of ecological restoration programs on social–ecological systems in China. However, the results apply mainly to areas with similar climates and may not be universally extensible to different climatic zones. The variability arises from significant disparities in the natural environmental characteristics between drylands and humid regions. In humid regions, emigration contributes to vegetation greening, whereas immigration results in degradation. Conversely, in arid regions, immigration promotes vegetation greening, but emigration does not necessarily lead to observable greening. This contrast exists because cropland is greener than grassland in drylands. Furthermore, the phenomenon of population relocation facilitating vegetation restoration is more readily observed in humid areas but is challenging to detect in drylands. Therefore, more case studies are needed to explore the effectiveness of vegetation restoration in dryland regions.

EMPs have sparked heated debate globally, and many studies have shown their serious adverse impacts. Recent research has revealed that households relocated in Gondar City lost their neighborhood ties and social networks, and they also endured various adverse effects on their essential infrastructure (Alene 2021). Informal households displaced in Kigali have similarly faced adverse consequences on their physical, financial, social, and human livelihood assets (Nikuze et al. 2019). According to Moahmmadi Sarrafi (2018), the implementation of the Mina Resettlement Project successfully mitigated the risks associated with landslides and falling rocks, resulting in the construction of durable structures. However, these new structures proved to be inadequate due to a disregard for the cultural norms and affordability requirements of the low-income residents. In China, the impact of EMPs on the environment in relocation areas has been overwhelmingly detrimental (Hu et al. 2018). The socio–ecological assessment of EMP in this study focuses solely on the environment and does not account for the impact of economic and cultural aspects of migrants. This is due to the influence of China's policy system; environmental perception is more objective than social perceptions such as government support and satisfaction (Wang et al. 2023a, b, c; Zhang et al. 2023). Therefore, we investigated environmental perception, which ensures data reliability but inevitably lacks an exploration of residents' livelihoods, local cultural sentiments, and other related factors. It is essential not to overlook issues related to a decrease in income and cultural adaptation challenges associated with EMP (Webb et al. 2004; Adamo 2010; Warner 2010; Tang et al. 2012; Wang and Zhou 2022). This study has the limitation of relatively clustered sites for data collection. First, as the study area is located at the northern foot of the Qilian Mountains, uninhabitable terrain, such as mountains, is dominant, so the settlements themselves are relatively clustered. Second, when the government organized the relocation, it planned to build large-scale resettlement areas so that the stations of the relocation group would be close to each other. Finally, the control group needs to meet the environmental context similar to the immigration group and emigration group and meet the situation of no moving in the time series, making these research sites relatively clustered. Moreover, this study has limitations regarding the comprehensiveness of the sample types evaluated for social impacts, as it only includes immigrants and nonimmigrants in the immigration area. The EMP involves the relocation of entire villages, which means that nonimmigrants from the relocated villages are not identifiable in the research, thus limiting the diversity of the sample types. Future research could track migration patterns to delve deeper into the differences between immigrants and nonimmigrants.

Conclusion

We provide a technical means of statistically assessing the impact of the EMP from a social–ecological perspective, which can support the quantitative assessment of ecological restoration program effects connecting people and nature. The results show that China’s EMP had significant positive effects on vegetation greenness and people’s environmental perceptions in the Qilian Mountains. On the one hand, the application of the EMP has led to positive and significant impacts on the increase in greenness for both the immigration group and the emigration group. On the other hand, people who have experienced the EMP are more likely to perceive changes to their environment. The methodological contribution of this study is the application of a combination of subjective and objective statistical methods to assess the benefits of the ecological restoration program based on remote sensing and questionnaire data to assess the effect of ecological restoration program at the landscape scale. Although there are some limitations to the data we applied, particularly the lack of an exploration of residents' livelihoods and local cultural sentiments, the established framework provides a reference for assessing the social–ecological impacts of ecological restoration in other drylands.

Supplementary Information

Below is the link to the electronic supplementary material.

Biographies

Tianjing Wu

is a master student in Faculty of Geographical Science, Beijing Normal University, Beijing, China. Her research interests include landscape ecology and social–ecological system.

Yanxu Liu

is an associate professor in Faculty of Geographical Science, Beijing Normal University. His research interests include landscape pattern and ecological processes, ecosystem services and landscape management.

Xinhua Qi

is a professor in School of Geographical Sciences School of Neutrality Future Technology, Fujian Normal University, Fuzhou, China. His research interest is human geography and ecology.

Qing Zhang

is a master of geography, working for Yulin Bureau of Natural Resources and Planning, China. Her research interest is land resource management.

Ying Yao

is a Ph.D. student in Faculty of Geographical Science, Beijing Normal University, Beijing, China. Her research interests include ecosystem resilience, environmental impact assessment, and ecological restoration.

Jincheng Wu

is an undergraduate student in Faculty of Geographical Science, Beijing Normal University, Beijing, China. His research interest is ecological remote sensing.

Declarations

Conflict of interest

The authors declare no conflict of interest.