China’s grassland ecological compensation policy achieves win-win goals in Inner Mongolia

Approximately 10% of China’s grasslands are severely degraded and 90% of them are overgrazed. To protect ecosystems and boost human well-being, payments for ecosystem services programs have been implemented to generate win-win outcomes for pastoralists and the grasslands. Taking a payment for ecosystem services program in Damao County, Inner Mongolia as an example, our study evaluated the ecological effects of the Grassland Ecological Compensation Policy (GECP) based on historical trends at the pixel, parcel, and county levels. We also evaluated the socioeconomic effects of GECP using both objective and subjective well-being at the household level. Our results show that: (1) at the pixel level, the percentages of additionally increased Normalized Difference Vegetation Index (NDVI) and Net Primary Production (NPP) were 93.4% and 93.3% after GECP implementation, corresponding to the average additional effects of 0.08 and 58.99 g C/m2, respectively. At the parcel level, the GECP additionally increased NDVI between 0.02–0.17 (average of 0.08) and increased NPP between 28.36–115.15 (average of 60.30) g C/m2, respectively. At the county level, the GECP additionally increased grassland NDVI and NPP by 0.07 (∼3.4% annually) and 53.63 g C/m2 (∼4.5% annually) from 2008 to 2020, respectively; and (2) the GECP implementation significantly improved pastoralists’ objective well-being (P < 0.01) while the effects on subjective well-being indices were mixed. Our results also show that GECP effects on objective and subjective well-being significantly differ from households with large rangeland to those with small rangeland. We further discussed the experience, challenges, and opportunities of GECP. The long-term sustainability of GECP, particularly socioeconomic sustainability, still remains challenging and relies on guiding pastoralists to find alternative livelihoods. For future research and policy improvement, we call for the establishment of a better policy compensation mechanism that jointly considers the ecological effectiveness, economic efficiency, and social equity.


Introduction
Approximately 42% of China's terrestrial areas are grassland supporting the livelihoods of 16 million pastoralists , Byrne et al 2020, Kemp et al 2020. During the past decades, around 90% of China's grasslands have been overgrazed or are considered to be degraded, of which 10% (∼40 million ha) were severely degraded (Kemp et al 2020, Zhou et al 2020. Approximately 80% of the degraded grasslands are in arid and semi-arid regions of China, where desertification of grassland resulting from climate change and human activities has become a major threat to pastoralists and grassland ecosystems , Kemp et al 2020. People living in those areas are among the poorest in China and conflicts between grassland protection and livelihood are remarkable (Kemp et al 2020, Li et al 2014. To meet the Sustainable Development Goals (SDGs) of the United Nations (UN) and create win-win goals of grassland restoration and poverty alleviation, effective mechanisms are urgently needed to incentivize adaptations of land use behaviors and improve grassland management outcomes (Byrne et al 2020, Li et al 2014, Reynolds et al 2007, UN 2015. Over the past two decades, the Chinese government has introduced several large-scale Payments for Ecosystem Services (PES) programs to combat grassland degradation, among which the Grassland Ecological Compensation Policy (GECP) is the world's largest in terms of its implementation scope, the number of participants, and total monetary transfer (Hou et al 2021, Li et al 2014, Yin et al 2019. The Chinese central government initiated the GECP in 2011 with the dual goals of grassland restoration and poverty alleviation (Hou et al 2021, MARA 2011. To protect grassland, the GECP divided it into two zones: a grazing ban zone where no grazing is allowed and a forage-livestock balance zone where only limited grazing is allowed (MARA 2011). The grazing ban policy mainly focuses on severely degraded grassland that is unable to support grazing. The foragelivestock balance zone is designed to balance grazing intensity and grassland resources of moderate degradation (MARA 2011). To alleviate poverty and moderate the negative impacts on the livelihoods of pastoralists, governments paid participating households for reducing their grazing intensity. The total investment of GECP was 12.00 billion USD (1 USD = 6.45 RMB as of 2021) and 14.54 billion USD during the first period (2011)(2012)(2013)(2014)(2015) and the second period (2016-2020), respectively (MFC 2016, NFGA 2020).
As the largest pastoralist-focused PES program in China and abroad, the effects of GECP have attracted growing attention over recent years (Hou et al 2021, Liu et al 2018. Yin et al (2019) found that although total income increased in Inner Mongolia, GECP negatively affected pastoralist net household income. Liu et al (2018) showed that the GECP significantly improved grassland ecological conditions from 2011 to 2014 in Inner Mongolia. Pan et al (2021) suggested that the GECP improved grassland ecosystem services and household income, although the extent depended on the types of subsidies provided. In the same year, Hou et al (2021) found that the GECP significantly improved both grassland ecological conditions and pastoralist income. On the other hand, Hou et al (2021) and Li et al (2021) further explored the effects of GECP on pastoralists' social equality and demonstrated that the GECP widened the income gap regionally and locally.
Although studies have evaluated the effects of GECP in different contexts, some gaps still exist in GECP evaluation. First, existing studies did not describe the effects of GECP on a desertified grassland ecosystem. Approximately 70% of the grassland in China is desertified due to climatological warming and long-term overgrazing (Akiyama andKawamura 2007, Zhao et al 2008). Grassland desertification has caused many environmental and socioeconomic problems . Hence, the reversal of grassland desertification is of great importance to achieve the SDGs, the win-win prospects, and the livelihoods of the poorest pastoralists (Akiyama and Kawamura 2007, Kemp et al 2020. Second, the evaluation indicator of socioeconomic benefit is simplistic and usually based on household income. More comprehensive indicators are required (e.g., human well-being indices) to properly evaluate the socioeconomic effects of GECP (Li et al 2014, Yang et al 2013a. Third, most existing studies consider a single scale (e.g., pixel, parcel, or county level) evaluation for GECP ecological effects. However, ecosystems generally show heterogeneity at varying scales in space and time (Levin 1992). Revealing how nature works needs a comprehensive understanding of different scales (Wu and Qi 2000). Multiple spatial scale evaluation can unravel the spatial heterogeneity of GECP ecological effects, and thus we can better understand the GECP effects.
With the urgent need for grassland protection and improvement in grassland policy recognition (Bardgett et al 2021), we systematically evaluated the ecological and socioeconomic effects of GECP in a desertified grassland ecosystem at multiple levels and assessed human well-being indicators. We selected one of the earliest county-level policy pilots and attempted to evaluate both the ecological and socioeconomic effects of the GECP. This pilot started GECP implementation at the end of 2007. Because the grasslands in Damao County were severely degraded, the government implemented a grazing ban policy (e.g., part of GECP) across the county. All of the pastoral households (6620 households) in our study area were involved in GECP from 2008 to 2020. The local government invested 13.95 USD/ha from 2008 to 2015 and the central government invested 15.12 USD/ ha from 2011 to 2015 and 17.44 USD/ha from 2016 to 2020. Combining climate data, remotely sensed data, and socioeconomic survey data from the past 30 years, our objectives were to (1) evaluate the ecological effects of the GECP at multiple scales; (2) analyze the socioeconomic effects of the GECP using human well-being indicators at the household scale.

Study area and GECP implementation
Damao County is one of the earliest areas enrolled in the GECP pilot program. This county is among the 33 typical livestock husbandry counties in Inner Mongolia located in the midwest of the Inner Mongolia Autonomous Region of China (figure 1). In most arid and semi-arid pastoral areas, overgrazing has caused grassland degradation, in turn damaging environmental sustainability and weakening socioeconomic development in the area (Cease et al 2015, Huang et al 2013, Liu et al 2018. To promote grassland protection and economic development, the government fully launched GECP to the entire pastoral areas on January 1st, 2008 (DMQ 2012). The general characteristics and investment of the GECP in Damao County are presented in table 1.

Data collection and analysis 2.2.1. Data collection
We obtained multiple sources of data including ecological, socioeconomic, and biophysical data to evaluate the ecological and socioeconomic effects of the GECP. For ecological assessment, we collected data including   figure S1). Also, we applied the Carnegie-Ames-Stanford Approach (CASA) model (Piao et al 2001, Zhu 2006) to generate aboveground time-series NPP data at a 30 m resolution from 1990 to 2020 (see supplementary data of A.2, table S2 and figure S1). For socioeconomic assessment, we conducted a face-to-face questionnaire survey in 2017 and randomly selected 251 households to obtain socioeconomic data, including basic family information (gender, marriage, and education), income and expenses, social network, pastoralist's self-perception, policy effect perception, and indicators of human well-being (see supplementary data of A.3, table S1). There were 188 of 251 households that had spatially explicit rangeland boundaries (i.e., parcel), and some of them had two rangelands (one for summer grazing and another for winter grazing). Thus, there were totally 194 parcels. Our study is part of the National Key Research and Development Project (2016YFC0503404), which was approved and funded by the Ministry of Science and Technology of China. We obtained permission from the government of Damao County for conducting household surveys inside the county. Since household heads or their spouses are most familiar with household activities, they were often selected as our interviewees. Only adults were chosen as interviewees in our study. As many of our interviewees are not literate or have very low levels of education, a verbal consent process was adopted. First, we read the verbal consent script to the selected interviewees. If they agreed, then we continued to interview them; or else, we did not collect any further information and turn to the next selected interviewee. Other socioeconomic, biophysical, and GECP-related data are listed in table S1. The details of preprocessing of ecological and socioeconomic data were listed in supplementary data of A.4.

Evaluation of ecological effects
To systematically understand the ecological effects of GECP, we conducted our analyses at the pixel, parcel, and county levels. We followed previous conservation program evaluations (e.g., FONAFIFO. et al 2012, Kalacska et al 2008, Yang et al 2013b and used the linear regression to create the historical trend before GECP implementation to construct the baseline in evaluating the ecological effects of GECP (supplementary methods of B.1). We chose this approach rather than other methods (e.g., matching) for three reasons. First, the GECP is a national conservation policy and has been fully implemented in Inner Mongolia's agricultural and animal husbandry counties since 2011 (MARA 2011). It was unlikely, if not impossible, to select an appropriate without-GECP baseline from similar adjacent sites or larger regional administrative units during the implementation period (Yang et al 2013b). Second, previous studies demonstrated that grazing and climate factors were the main driving forces influencing grassland ecological conditions in Inner Mongolia (Herrero-Jauregui and Oesterheld 2018, Li et al 2018, Yan et al 2020. However, we found that the effects of climate factors on NDVI and NPP were insignificant during the study period (supplementary methods and results of B.2, figure  S2 and figure S3). Third, the survey participants and local government officials indicated that GECP was the main driving force on grassland restoration and livelihood transformation in Damao County. Therefore, historical trends can reasonably predict what would occur in the ecosystem without the GECP.

Evaluation of socioeconomic effects
As the household is the basic administrative unit, we analyzed the socioeconomic effects of GECP at the household level. We adopted the widely used objective and subjective well-being indices (see details in Supplementary Materials A.4) to measure socioeconomic effects at the household level. We applied a matching approach to evaluate the approximate effects of GECP on changes in pastoralists' objective and subjective wellbeing (Sekhon 2011, Zhou et al 2022). The common approaches to program evaluation are quasi-experimental design and pre-versus post-comparative analysis based on historical extrapolation (Baylis et al 2016, Huettner et al 2009, Wunder 2005. Because of the full implementation strategy, we could not select appropriate control groups without GECP to apply a quasi-experimental design. Besides, unlike the evaluation of ecological effects, we did not have time series data available on well-being indices to build the pre-GECP trends and approximate the without-GECP baseline in the impact evaluation. However, we found that livestock number was positively and significantly correlated with household's rangeland size (figure S4). Moreover, time-series data before GECP implementation from 1991 to 2007 showed that animal husbandry income accounted for 96.7% of the household's total income (SBDC 2005(SBDC , 2010. Correspondingly, GECP-participated households were directly banned from animal husbandry activities, while GECP's subsidy is determined by the household rangeland size and was the main source of family income during the policy implementation period. In addition, previous studies have demonstrated that household income is a primary driver behind both objective and subjective wellbeing (Diener and Suh 1997, Western and Tomaszewski 2016). Therefore, it is reasonable for us to analyze the approximate effects of GECP on pastoralists' well-being based on the sizes of household rangeland using a matching approach (supplementary methods and results of B.3).
The main advantage of applying the matching approach is its ability to diagnose whether GECP plays a significantly positive or negative role in pastoralits' well-being change. The matching approach requires obtaining an adequate balance in covariates between large rangeland size groups and small rangeland size groups (table S6, figure S5-S9), which means a good control of confounding bias between large rangeland size households and small rangeland size households. After the matching, the obtained policy effect is corresponding to the difference in pastoralists' well-being between households with a large rangeland size and counterfactual expected change in households with a small rangeland size. The disadvantage of this approach is that the policy effect might be underestimated because the control group is not a completely blank control of without-GECP implementation.

Ecological effects of GECP
The ecological results of GECP were in line with our expectation, which showed that the GECP had significant (P < 0.01) and positive effects on the recovery of both NDVI and NPP (figures 2-4). Analyses at different spatial levels revealed the spatial heterogeneity of grassland restoration (figures 2 and 3). At the pixel level, only 1.9% and 1.1% of the total area had a significant increase in NDVI and NPP between 1990. From 2008.3% of the total area had significant (P < 0.05) gains in NDVI ranging between −0.17-0.32 (equivalent to an average additional effect of 0.08) after GECP implementation ( figure 2(A)). Approximately 94.4% of the total area had significant (P < 0.05) gains in NPP ranging between -112.66-253.73 (equivalent to an average additional effect of 58.99) g C/m 2 (figure 2(B)) with GECP implementation. In addition, pixels with significant (P < 0.05) gains in NDVI and NPP were mainly distributed in the west and southeast of Damao County, corresponding to townships of Mingan, Bayinhua, Daerhan, and Xilamuren, while the degraded areas were in the north of Damao County (figure 2).
At the parcel level, from 2008 to 2020, rangeland enrolled in GECP had all gained in NDVI and NPP, but to various extents (figure 3). The gains in the selected rangelands varied between 0.02 and 0.17 (equivalent to an average additional effect of 0.08) for NDVI and 28.36 to 115.15 (equivalent to an average additional effect of 60.30) g C/m 2 for NPP. The average additional effects of GECP at the pixel and parcel levels were very close to each other. The slight differences between the two spatial scales were mainly caused by the systematic sampling bias (pixels were taken over the entire study area, while parcels were samples of the entire area). Besides, parcels with significant (P < 0.05) gains in NDVI and NPP were mainly distributed in townships of Daerhan, Xilamuren, northeast of Bayinaobao, and north of Mingan ( figure 3).
At the county level, the with-GECP grassland NDVI and NPP in 2020 estimated by linear regression from 2008 to 2020 were significantly (P < 0.001) higher than the without-GECP grassland NDVI and NPP in 2020 estimated by the historical trend lines from 1990 to 2007 (figure 4). Before GECP implementation, the average value of grassland NDVI was 0.17. Because the slope of NDVI from 1990 to 2007 was statistically insignificant (P > 0.1), the estimated baseline NDVI index in 2020 in the absence of the GECP was also 0.17. The estimated NDVI index in 2020 in the presence of the GECP, as found from a linear regression analysis between 2008 and 2020, was 0.24. Therefore, the estimated increase in NDVI attributable to the GECP was 0.07 between 2008 and 2020. This gain was equivalent to a 41.2% of cumulative increase from 2008 to 2020 or a 3.4% of annual increase. The effect of the GECP on grassland NPP was consistent with grassland NDVI. The estimated aboveground grassland NPP with and without GECP in 2020 were 152.66 g C/m 2 and 99.03 g C/m 2 , respectively. Hence, the increased aboveground grassland NPP by the GECP between 2008 and 2020 was 53.63 g C/m 2 . The corresponding percentage increase in grassland NPP from 2008 to 2020 was 54.2% (∼4.5% annually). The effects in the three spatial levels were generally consistent with each other. The small differences between effects at pixel and county levels were caused by the differences in estimated effects in 2020. For example, at the pixel level, there were 1.9% and 1.1% of the total pixels of NDVI and NPP with significantly (P < 0.05) increased trends between 1990-2007, and thus the values of those pixels in 2020 were estimated via the linear extrapolation; however, the values in 2020 at the county level were estimated by the mean value from 1990-2007 because of the insignificant (P > 0.1) trend.

Socioeconomic effects of GECP
Overall, our results indicated that the objective well-being of pastoralists improved (table 2 and figure 5), yet most subjective well-being indices showed insignificant changes during the GECP implementation period (table 3 and figure 5). For objective well-being, our results showed that all indicators gradually improved after the GECP implementation ( figure 5(A)). Especially, indicators of floor area, transportation, and total income increased by 47.72%, 158.12%, and 30.34% between 2006 and 2017, respectively. Our results also showed that GECP had different impacts on pastoralists' objective well-being between households with large size rangeland versus small size rangeland (table 2). Before GECP implementation, the t-test analysis showed that values of floor area and total income in households with large rangeland were significantly (P < 0.05) higher than those with small rangeland (table S5). After GECP implementation, except for water source, values of other objective wellbeing indicators of households with large rangeland were all significantly (p < 0.001) and consistently higher than those with small rangeland in both two phases of GECP implementation (table 2). After acquiring an adequate balance in covariates between households with large and small rangeland (table S6 and figures S5-S8), the results of matching also suggested that GECP played a positive and significant (p < 0.01) role in household decoration, floor area, transportation, and total income (table 2). Compared to households with small rangeland size, results of matching analysis showed that households with large rangeland size increased by an extra normalized value of 6.54, 19.89, and 13.89 for household decoration, floor area, and transportation, and 3179.10 USD for total household income in the first phase. In the second phase of the GECP implementation, the significant (P < 0.01) effects on household decoration, floor area, transportation, and total income were 6.78, 21.29, 15.62, and 4981.20 USD, respectively (table 2).
For subjective well-being, most of the indices declined in the first period of GECP implementation and then improved in the second phase ( figure 5(B)). Except for health index, the downtrend of other indices had reversed during the second phase of GECP implementation. The t-test analysis showed that values for security, health, freedom, and overall satisfaction indices of households with large rangeland were significantly (P < 0.05) higher than those with small rangeland before and after GECP implementation (tables 3 and S5). The matching analysis suggested that most of the subjective indices (except for freedom) did not show statistically significant changes between households with large rangeland versus those with small rangeland (table 3). However, all the values of other indices improved in the second phase of GECP implementation, except for the index of health. Based on our survey, the average age of interviewees was 41 in the year of 2017. Thus, for the continuous decline of health status, it is likely due to the deterioration of physical fitness that comes with age. The reasons for the differences between t-test analysis and matching analysis might be because the matching approach has adequately controlled the confounding variates between the two groups (table S6 and figure S9). Except for the sensitivity analysis based on Wilcoxon signed-rank test, we also applied matching analysis with median value to divide the households into a large rangeland group and a small rangeland group for robustness check. The results of the average value (tables 2 and 3) were consistent with those of the median value (tables S7 and S8).
The semi-quantitative results of pastoralists' perceptions on GECP effects showed that pastoralists thought GECP generated positive ecological and socioeconomic effects (table 4). Among a score range of −10 to 10, the surveyed 251 households provided average scores of 3.63 ± 3.09 for ecological effects, 0.16 ± 3.61 for socioeconomic effects, and 2.52 ± 2.59 for overall effects, respectively. Such results suggested that pastoralists felt that GECP had better ecological effects than socioeconomic effects. Furthermore, the large standard deviation of socioeconomic effect scores also suggested that socioeconomic effects largely varied from one household to another, which is consistent with our above comparison results of well-being between households with large rangeland and small rangeland.

Discussion
Our study shows that the GECP plays an important role in grassland restoration as well as poverty alleviation. Combing remotely sensed data with household survey data, our analyses show significant and positive ecological effects of GECP at the pixel, parcel, and county levels but mixed socioeconomic effects at the household level. The puzzle of grassland resource management in Damao County is similar to many other pastoral and agropastoral regions in China and beyond (Cease et al 2015, Huang et al 2013, Liu et al 2018. The practice of GECP in Damao County has important implications for other areas. Below we discussed the experience, challenges, and opportunities of GECP, which not only apply to Damao County but also to other places globally.

Experience from GECP implementation in Damao County
First, strict monitoring and sanction measures are expected to have positive feedback in the short term, whereas the desired prospects of win-win outcomes in the long run call for diverse management approaches. The effective monitoring and graduated sanctions were effective during the early years of GECP implementation in Damao County. However, such simple and coercive enforcement also resulted in a few negative outcomes. The most obvious negative impact was the low satisfaction with GECP, reflected by the negative and insignificant effects of GECP on indices of subjective well-being and relatively low scores for policy effects. Moreover, during our survey, many participants reported that punishments for defying grazing standards were subjective, varied from household to household, and were influenced by nepotism. These injustices in policy sanctions seriously undermined the pastoralists' regard for the policy. This brings out a hidden crisis: how do we maintain the long-  Table 2. Different impacts of GECP on changes in pastoralists' objective well-being using a matching approach and t-test via the comparison of households with large size rangeland versus small size rangeland. Number 1 and 2 in each objective well-being index represent the changes during the first and whole phases of GECP implementation. * p < 0.05, ** p < 0.01 , *** p < 0.001. a Mean values of normalized raw data between households with large rangeland and households with small rangeland in 2013 and 2017, respectively. b Differences of mean values of normalized raw data between households with large rangeland and households with small rangeland in 2013 and 2017 using t-test. c Based on Wilcoxon signed-rank test at p < 0.1. term ecological and socioeconomic effects when monitoring and sanctioning measures attenuate over time or once the GECP ceases? Hence, we came up with several suggestions to address the above obstacles and contribute to long-term winwin prospects based on Damao's experience. The first suggestion is to establish a flexible transformation mechanism between a grazing ban and forage-livestock balance management based on regular monitoring of grassland quantity and quality. The second suggestion is to establish a specific and uniform standard for addressing resistance to grazing rules. For example, the local government should unambiguously clarify and impose punishment (e.g., cash, animal confiscation, or combined penalties) for illegal grazing with different kinds of animals. In the process of law enforcement, government officials must apply a specific and uniform standard in carrying out the punishment. At the same time, a supervisory mechanism also should be applied to rectify the corruption of government officials. The third suggestion is to use self-organization to manage the grassland resources rather than enforcing a strictly centralized natural resource management, especially during the middle and later periods of policy implementation. To establish such a self-organization approach, the government should: (1) encourage pastoralists to actively engage in the policy cycle; (2) rectify the policy's management based on pastoralists' feedback and local experience, and (3) improve the education level of pastoralists to enhance their understanding of policies, adaptability to policy requirements, and thus to adhere to the tenets of the policy.

Objects
Second, it is necessary to link the compensation standard with GECP effects to improve social equity. The pastoralist livelihoods heavily depend on grassland resources and thus there is uneven wealth accumulation across the populace (Liao et al 2020, Liu et al 2018. In addition, the level of compensation is commensurate with the size of the grassland contract for each household. This dependency caused a distinctive discrepancy in compensation payments and widened the income gap (Hou et al 2021, Li et al 2021. For example, pastoralists in Damao County received a compensation range from a minimum level of 583 USD/year to a maximum level of 30 625 USD/year in the first phase of GECP implementation due to variations in grassland size. The widening Figure 5. Pastoralists' objective and subjective well-being before and after the GECP implementation. Table 3. Different impacts of GECP on changes in pastoralists' subjective well-being using a matching approach and t-test via the comparison of households with large size rangeland versus small size rangeland. Number 1 and 2 in each subjective well-being index represent the changes during the first and whole phases of GECP implementation. * p < 0.05, ** p < 0.01 , *** p < 0.001. income gap hindered households from fully embracing the policy, increasing the policy transaction costs, and thus leading to the failure of the policy. Therefore, it is necessary to establish a better policy compensation mechanism that integratively considers the ecological restoration effectiveness, economic efficiency, and social equity (Ding et al 2022).

Challenges and opportunities
Although GECP restored grassland ecological conditions and overall improved pastoralists' well-being, some challenges still exist in grassland resource management and pastoralists' livelihood sustainability. Policymakers should guide pastoralists to transform short-term pains into long-term gains. Unlike the positive ecological effects of GECP, the socioeconomic effects of policy were mixed and varied across different indicators and social groups (Hou et al 2021, Li et al 2014, Li et al 2021. In this study, grazing bans and the accompanying measures forced 50.6% of pastoralists to move from the pastoral areas to the downtown areas, which greatly changed the lifestyle of pastoralists. Most relocated pastoralists could not adapt to the new life both mentally and physically and they had trouble making their living in downtown areas (figure S10). As a result, around 97.6% of relocated pastoralists later returned to pastorals during the middle and late stages of GECP implementation. Many of those who stayed in pastoral areas also had trouble making their living and had to rely on compensation funds to survive. The dramatic change of pastoralists' life led to the low satisfaction in subjective well-being, especially in the first phase of GECP implementation. Moreover, the results of pastoralists' perceptions of GECP effects also implied that pastoralists had a very low rating of the socioeconomic benefits of GECP. Therefore, it remains challenging to achieve long-term positive socioeconomic effects. Despite the challenges in recent PES evaluation and management, some opportunities also arise. Approximately 70% of the 120 million Mongolian minority population live in pastoral areas; their livelihoods largely depend on grassland resources (Chen and Luo 2009, Kemp et al 2020, Zhou et al 2020. Therefore, grassland ecological protection in the arid and semi-arid areas is not only important for environmental sustainability but also crucial for the livelihood sustainability of the pastoralists. Hence, future research should pay more attention to integrated PES portfolios that advocate ecological restoration, social equality, and economic development together in grassland protection to balance the relationship between protection and development. As the Chinese government aggressively promotes renewable energy development (e.g., photovoltaic, wind power) to achieve carbon peaking and carbon neutrality goals, there are also emerging new opportunities in arid and semi-arid areas due to abundant solar energy and low land rent. For instance, recently there have been some successful agrophotovoltaic practices, which not only generate remarkable electricity revenue but also high agropastoral income and vegetation restoration (Barron-Gafford et al 2019, Liu et al 2020). In addition, agrophotovoltaic power plants also provide extra employment opportunities for local households such as regular maintenance, grass mowing, and livestock raising inside the power plants.

Conclusion
The results of our integrated ecological and socioeconomic evaluation of GECP reinforece the belief that PES programs can achieve win-win outcomes. On the one hand, after 15 years of GECP implementation, grassland ecological conditions have significantly improved, with spatial heterogeneity at the pixel and parcel scales. In addition, pastoralists' objective well-being has also improved significantly, and the well-being outcomes differed between households with large and small rangelands. On the other hand, the potential time lag effects in socioeconomic effects of subjective well-being indicate some challenges in transforming the short-term pain into a long-term gain. However, it also gives us faith that the PES program can play a significant positive effect on both objective and subjective well-being in the near future with scientific management. As more and more PES programs at different scales are implemented worldwide, it is crucial to enhance the accountability of both ecological and socioeconomic outcomes with rigorous impact evaluation methods. Providing alternative livelihoods for PES program participants remains the major challenge to long-term policy effectiveness. Notes: # A score range of −10 to 10 was given to represent the pastoralists' perception of GECP effects. A negative number represents a negative effect of the GECP. N = 251.
Measures of education, training, and advertising can increase the policy recognition of participants, which are essential to policy management. The long-term sustainability of PES, particularly socioeconomic sustainability, relies on joint efforts among the government, participants, and third parties. For future research and policy improvement, we call for the establishment of a better policy compensation mechanism that jointly considers ecological effectiveness, economic efficiency, and social equity.